IPMI - 400 Bad Request

IPMI - 400 Bad Request
- IPMI Intelligent Platform Management
Interface Specification
v1.5
Document Revision 1.1
February 20, 2002
6/1/04 MARKUP
Intel Hewlett-Packard NEC Dell
Intelligent Platform Management Interface Specification
Revision History
Date
Ver
Rev
Modifications
9/16/98
8/26/99
1.0
1.0
1.0
1.1
2/21/01
2/20/02
9/12/03
9/26/03
1.5
1.5
1.5
1.5
1.0
1.1
1.1
1.1
1/29/04
6/1/04
1.5
1.5
1.1
1.1
IPMI v1.0 Initial release
Errata Revision. Incorporated errata from revision 1 or the Errata and
Clarifications for the IPMI v1.0 specification.
IPMI v1.5 Initial release
Updated to include addenda and errata
Markup to include 9/12/03 addenda and errata
Markup updated to include missing optional 5th byte on Get Chassis Status
command, per errata E317
Cumulative markup including errata per errata document revision 5
Cumulative markup including errata per IPMI v2.0/v1.5 errata document revision
1. Also added E256, missed in previous markup.
Copyright © 1999, 2000, 2001, 2002, 2003, 2004 Intel Corporation, Hewlett-Packard
Company, NEC Corporation, Dell Computer Corporation, All rights reserved.
INTELLECTUAL PROPERTY DISCLAIMER
THIS SPECIFICATION IS PROVIDED “AS IS” WITH NO WARRANTIES WHATSOEVER INCLUDING ANY
WARRANTY OF MERCHANTABILITY, FITNESS FOR ANY PARTICULAR PURPOSE, OR ANY WARRANTY
OTHERWISE ARISING OUT OF ANY PROPOSAL, SPECIFICATION, OR SAMPLE.
NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY
RIGHTS IS GRANTED OR INTENDED HEREBY.
INTEL, HEWLETT-PACKARD, NEC, AND DELL DISCLAIM ALL LIABILITY, INCLUDING LIABILITY FOR
INFRINGEMENT OF PROPRIETARY RIGHTS, RELATING TO IMPLEMENTATION OF INFORMATION IN THIS
SPECIFICATION. INTEL, HEWLETT-PACKARD, NEC, AND DELL, DO NOT WARRANT OR REPRESENT THAT
SUCH IMPLEMENTATION(S) WILL NOT INFRINGE SUCH RIGHTS.
I2C is a trademark of Philips Semiconductors. All other product names are trademarks, registered trademarks, or servicemarks of
their respective owners.
I2C is a two-wire communications bus/protocol developed by Philips. IPMB is a subset of the I2C bus/protocol and was
developed by Intel. Implementations of the I2C bus/protocol or the IPMB bus/protocol may require licenses from various
entities, including Philips Electronics N.V. and North American Philips Corporation.
Intel, Hewlett-Packard, NEC, and Dell retain the right to make changes to this document at any time, without notice. Intel,
Hewlett-Packard, NEC, and Dell make no warranty for the use of this document and assume no responsibility for any error
which may appear in the document nor does it make a commitment to update the information contained herein.
ii
Intelligent Platform Management Interface Specification
IPMI NON-DISCLOSURE AGREEMENT
DO NOT download these files (collectively, the “Confidential Information”) until you have carefully read the
following terms and conditions. By downloading the Confidential Information you agree to the terms of this
Agreement. If you do not wish to so agree, do not download the Confidential Information.
1. Confidential Information. The confidential, proprietary and trade secret information being disclosed ("Confidential
Information"), is that information marked with a "confidential", "proprietary", or similar legend, and is described as:
Confidential Information: Intelligent Platform Management Interface Specification (v1.5), Intelligent Platform
Management Bus Bridge Specification (v1.0), Intelligent Chassis Management Bus Bridge Specification (v1.0)
CONFIDENTIAL INFORMATION IS PROVIDED SOLELY FOR YOUR INTERNAL EVALUATION AND
REVIEW TO DETERMINE WHETHER TO ADOPT THE SPECIFICATIONS BY SIGNING A SEPARATE
ADOPTER’S AGREEMENT. THE RECEIVING PARTY IS NOT LICENSED TO IMPLEMENT THE
SPECIFICATIONS UNLESS OR UNTIL AN ADOPTER’S AGREEMENT IS EXECUTED.
Disclosing party’s representatives for disclosing Confidential Information is: Fadi Zuhayri ([email protected])
2. Obligations of Receiving Party. The receiving party will maintain the confidentiality of the Confidential Information of
the disclosing party with at least the same degree of care that it uses to protect its own confidential and proprietary
information, but no less than a reasonable degree of care under the circumstances. The receiving party will not
disclose any of the disclosing party’s Confidential Information to employees or to any third parties except to the
receiving party’s employees, parent company and majority-owned subsidiaries who have a need to know and who
agree to abide by nondisclosure terms at least as comprehensive as those set forth herein; provided that the receiving
party will be liable for breach by any such entity. The receiving party will not make any copies of Confidential
Information received from the disclosing party except as necessary for its employees, parent company and majorityowned subsidiaries with a need to know. Any copies which are made will be identified as belonging to the disclosing
party and marked "confidential", "proprietary" or with a similar legend.
3. Period of Non-Assertion. Unless a shorter period is indicated below, the disclosing party will not assert any claims for
breach of this Agreement or misappropriation of trade secrets against the receiving party arising out of the receiving
party’s disclosure of disclosing party’s Confidential Information made more than five (5) years from the date of receipt
of the Confidential Information by the receiving party. However, unless at least one of the exceptions set forth in
Section 4 below has occurred, the receiving party will continue to treat such Confidential Information as the
confidential information of the disclosing party and only disclose any such Confidential Information to third parties
under the terms of a non-disclosure agreement.
4. Termination of Obligation of Confidentiality. The receiving party will not be liable for the disclosure of any Confidential
Information which is: (a) rightfully in the public domain other than by a breach of this Agreement of a duty to the
disclosing party; (b) rightfully received from a third party without any obligation of confidentiality; (c) rightfully known to
the receiving party without any limitation on use or disclosure prior to its receipt from the disclosing party;
(d) independently developed by employees of the receiving party; or (e) generally made available to third parties by
the disclosing party without restriction on disclosure.
5. Title. Title or the right to possess Confidential Information as between the parties will remain in the disclosing party.
6. No Obligation of Disclosure; Termination The disclosing party may terminate this Agreement at any time without
cause upon written notice to the other party; provided that the receiving party’s obligations with respect to information
disclosed during the term of this Agreement will survive any such termination. The disclosing party may, at any time:
(a) cease giving Confidential Information to the other party without any liability, and/or (b) request in writing the return
or destruction of all or part of its Confidential Information previously disclosed, and all copies thereof, and the
receiving party will promptly comply with such request, and certify in writing its compliance.
7. General.
(a) This Agreement is neither intended to nor will it be construed as creating a joint venture, partnership or other form
of business association between the parties, nor an obligation to buy or sell products using or incorporating the
Confidential Information.
(b) No license under any patent, copyright, trade secret or other intellectual property right is granted to or conferred
upon either party in this Agreement or by the transfer of any information by one party to the other party as
contemplated hereunder, either expressly, by implication, inducement, estoppel or otherwise, and that any
license under any such intellectual property rights must be express and in writing.
(c) The failure of either party to enforce any right resulting from breach of any provision of this Agreement will not be
deemed a waiver of any right relating to a subsequent breach of such provision or of any other right hereunder.
(d) This Agreement will be governed by the laws of the State of Delaware without reference to conflict of laws
principles.
(e) This Agreement constitutes the entire agreement between the parties with respect to the disclosure(s) of
Confidential Information described herein, and may not be amended except in a writing signed by a duly authorized
representative of the respective parties. Any other agreements between the parties, including non-disclosure
agreements, will not be affected by this Agreement.
iii
Intelligent Platform Management Interface Specification
Table of Contents
1.
Introduction .......................................................................................................................................... 1
1.1 Audience .........................................................................................................................................................1
1.2 Reference Documents ...................................................................................................................................12
1.3 Conventions and Terminology ........................................................................................................................4
1.4 Background - Architectural Goals...................................................................................................................5
1.5 New for IPMI v1.5 ..........................................................................................................................................6
1.6 IPMI Overview................................................................................................................................................8
1.6.1 Intelligent Platform Management.............................................................................................................8
1.6.2 IPMI Relationship to other Management Standards ................................................................................8
1.6.3 Management Controllers and the IPMB...................................................................................................9
1.6.4 IPMI Messaging.....................................................................................................................................10
1.6.5 Sensor Model .........................................................................................................................................10
1.6.6 System Event Log and Event Messages.................................................................................................11
1.6.7 Sensor Data Records & Capabilities Commands ...................................................................................11
1.6.8 Initialization Agent ................................................................................................................................12
1.6.9 Sensor Data Record Repository .............................................................................................................12
1.6.10 Private Management Busses ..................................................................................................................12
1.6.11 FRU Information....................................................................................................................................12
1.6.12 FRU Devices ..........................................................................................................................................13
1.6.13 Entity Association Records ....................................................................................................................13
1.6.14 Linkage between Events and FRU Information.....................................................................................13
1.6.15 Differentiation and Feature Extensibility...............................................................................................14
1.6.16 System Interfaces ...................................................................................................................................14
1.6.17 Other Messaging Interfaces...................................................................................................................15
1.6.18 LAN Interface ........................................................................................................................................15
1.6.19 Serial/Modem Interface..........................................................................................................................15
1.6.20 IPMI and ASF ........................................................................................................................................15
1.6.21 LAN Alerting .........................................................................................................................................16
1.6.22 Serial/Modem Alerting and Paging........................................................................................................16
1.6.23 Platform Event Filtering (PEF) ..............................................................................................................16
1.6.24 Call Down Lists and Alert Policies ........................................................................................................17
1.6.25 Channel Model, Authentication, Sessions, and Users............................................................................17
1.6.26 Standardized Watchdog Timer...............................................................................................................18
1.6.27 Standardized POH Counter ....................................................................................................................18
1.6.28 IPMI Hardware Components .................................................................................................................18
1.6.29 Configuration Interfaces ........................................................................................................................18
1.7 IPMI and BIOS .............................................................................................................................................18
1.8 System Management Software (SMS) ..........................................................................................................19
1.9 SMI Handler..................................................................................................................................................20
1.10 Overview of Changes from IPMI v1.0..........................................................................................................21
2.
Logical Management Device Types ................................................................................................. 23
3.
Baseboard Management Controller (BMC) ..................................................................................... 27
3.1 Required BMC Functions..............................................................................................................................30
4.
General Mgmt. Controller Required Functions............................................................................... 33
5.
Message Interface Description......................................................................................................... 34
5.1 Network Function Codes...............................................................................................................................34
5.2 Completion Codes .........................................................................................................................................37
5.3 Completion Code Requirements ...................................................................................................................38
5.3.1 Response Field Truncation on non-zero Generic Completion Codes ....................................................38
5.3.2 Summary of Completion Code Use .......................................................................................................38
iv
Intelligent Platform Management Interface Specification
5.4
5.5
5.6
Sensor Owner Identification..........................................................................................................................39
Software IDs (SWIDs) ..................................................................................................................................39
Isolation from Message Content....................................................................................................................40
6.
IPMI Messaging Interfaces................................................................................................................ 42
6.1 Terminology..................................................................................................................................................42
6.2 Channel Model ..............................................................................................................................................42
6.3 Channel Numbers..........................................................................................................................................43
6.4 Channel Protocol Type..............................................................................................................................4344
6.5 Channel Medium Type..............................................................................................................................4445
6.6 Channel Access Modes .............................................................................................................................4546
6.7 Logical Channels.......................................................................................................................................4647
6.8 Channel Privilege Levels ..........................................................................................................................4647
6.9 Users & Password Support........................................................................................................................4748
6.9.1 ‘Anonymous Login’ Convention .......................................................................................................4748
6.9.2 Anonymous Login Status...................................................................................................................4748
6.10 System Interface Messaging......................................................................................................................4849
6.10.1 BMC Channels and Receive Message Queue ....................................................................................4849
6.10.2 Event Message Buffer ........................................................................................................................4849
6.10a System Interface Discovery and Multiple Interfaces .................................................................................4950
6.11 IPMI Sessions............................................................................................................................................4950
6.11.1 Session-less Connections ...................................................................................................................5051
6.11.2 Single-session Connections................................................................................................................5051
6.11.3 Multi-session Connections .................................................................................................................5051
6.11.4 Per-Message and User Level Authentication Disables ......................................................................5051
6.11.5 Link Authentication ...........................................................................................................................5152
6.11.6 Summary of Connection Characteristics............................................................................................5152
6.11.7 Session Activation and IPMI Challenge-Response............................................................................5253
6.11.8 Session Sequence Numbers................................................................................................................5354
6.11.9 Session Sequence Number Generation ..............................................................................................5354
6.11.10 Inbound Session Sequence Number Tracking and Handling.............................................................5354
6.11.11 Out-of-order Packet Handling............................................................................................................5455
6.11.12 Outbound Session Sequence Number Tracking and Handling ..........................................................5455
6.11.13 Session Inactivity Timeouts ...............................................................................................................5455
6.11.13.1 Avoiding ‘Slot Stealing’ ..........................................................................................................5556
6.11.14 Additional Session Specifications and Characteristics ......................................................................5556
6.12 BMC Message Bridging............................................................................................................................5657
6.12.1 BMC LUN 10b Routing.....................................................................................................................5758
6.12.2 Send Message Command From System Interface..............................................................................5758
6.12.3 Send Message Command with Response Tracking ...........................................................................5859
6.12.4 Bridged Request Example..................................................................................................................5960
6.13 Message Size & Private Bus Transaction Size Requirements...................................................................6162
7.
IPMB Interface ................................................................................................................................ 6465
7.1 IPMB Access via Master Write-Read command.......................................................................................6465
7.2 BMC IPMB LUNs ....................................................................................................................................6465
7.3 Sending Messages to IPMB from System Software..................................................................................6465
7.4 Sending IPMB Messages to System Software ..........................................................................................6566
7.5 Testing for Event Message Buffer Support...............................................................................................6667
8.
ICMB Interface................................................................................................................................ 6869
8.1 Virtual ICMB Bridge Device ....................................................................................................................6869
8.2 ICMB Bridge Commands in BMC using Channels ..................................................................................6869
8.2.1 ICMB Bridging from System Interface to Remote IPMB using Channels ........................................6869
8.2.2 ICMB Bridging from Local IPMB to Remote IPMB using Channels ...............................................6970
9.
Keyboard Controller Style (KCS) Interface ................................................................................. 7273
v
Intelligent Platform Management Interface Specification
9.1 KCS Interface/BMC LUNs .......................................................................................................................7273
9.2 KCS Interface-BMC Request Message Format ........................................................................................7273
9.3 BMC-KCS Interface Response Message Format ......................................................................................7374
9.4 Logging Events from System Software via KCS Interface.......................................................................7374
9.5 KCS Interface Registers ............................................................................................................................7374
9.6 KCS Interface Control Codes....................................................................................................................7475
9.7 Status Register...........................................................................................................................................7475
9.7.1 SMS_ATN Flag Usage ......................................................................................................................7576
9.8 Command Register ....................................................................................................................................7677
9.9 Data Registers ...........................................................................................................................................7677
9.10 KCS Control Codes...................................................................................................................................7677
9.11 Performing KCS Interface Message Transfers..........................................................................................7677
9.12 KCS Communication and Non-communication Interrupts .......................................................................7778
9.13 Physical Interrupt Line Sharing.....................................................................................................................78
9.14 Additional Specifications for the KCS interface .......................................................................................7879
9.15 KCS Flow Diagrams .................................................................................................................................7980
9.16 Write Processing Summary .......................................................................................................................8384
9.17 Read Processing Summary........................................................................................................................8384
9.18 Error Processing Summary........................................................................................................................8384
9.19 Interrupting Messages in Progress ............................................................................................................8485
9.20 KCS Driver Design Recommendations.....................................................................................................8485
10. SMIC Interface ................................................................................................................................ 8687
10.1 SMS Transfer Streams ..............................................................................................................................8687
10.2 SMIC Communication Register Overview ...............................................................................................8687
10.3 SMIC/BMC Message Interface Registers .................................................................................................8788
10.3.1 Flags Register.....................................................................................................................................8788
10.3.2 Control/Status Register ......................................................................................................................8889
10.3.2.1 Control and Status Codes..............................................................................................................8889
10.3.3 Data Register......................................................................................................................................8990
10.4 Performing a single SMIC/BMC Transaction...........................................................................................8990
10.5 Performing a SMIC/BMC Message Transfer............................................................................................9091
10.6 Interrupting Streams in Progress ...............................................................................................................9091
10.7 Stream Switching ......................................................................................................................................9192
10.8 DATA_RDY Flag Handling .....................................................................................................................9192
10.9 SMIC Control and Status Code Ranges ....................................................................................................9293
10.10 SMIC SMS Stream Control Codes............................................................................................................9394
10.11 SMIC SMS Stream Status Codes ..............................................................................................................9495
10.12 SMIC Messaging.......................................................................................................................................9596
10.13 SMIC/BMC LUNs ....................................................................................................................................9596
10.14 SMIC-BMC Request Message Format......................................................................................................9596
10.15 BMC-SMIC Response Message Format ...................................................................................................9697
10.16 Logging Events from System Software via SMIC ....................................................................................9697
11. Block Transfer (BT) Interface ....................................................................................................... 9899
11.1 BT Interface-BMC Request Message Format ...........................................................................................9899
11.2 BMC-BT Interface Response Message Format.......................................................................................99100
11.3 Using the Seq Field .................................................................................................................................99100
11.4 Response Expiration Handling..............................................................................................................100101
11.5 Logging Events from System Software via BT Interface .....................................................................100101
11.6 Host to BMC Interface ..........................................................................................................................101102
11.6.1 BT Host Interface Registers ...........................................................................................................101102
11.6.2 BT BMC to Host Buffer (BMC2HOST)........................................................................................101102
11.6.3 BT Host to BMC Buffer (HOST2BMC)........................................................................................101102
11.6.4 BT Control Register (BT_CTRL) ..................................................................................................102103
11.6.5 BT Interrupt Mask Register (INTMASK)......................................................................................104105
vi
Intelligent Platform Management Interface Specification
11.7 Communication Protocol.......................................................................................................................105106
11.8 Host and BMC Busy States ...................................................................................................................106107
11.9 Host Command Power-On/Reset States................................................................................................106107
12. IPMI LAN Interface ..................................................................................................................... 107108
12.1 RMCP....................................................................................................................................................108109
12.1.1 ASF Messages in RMCP................................................................................................................108109
12.1.2 RMCP Port Numbers .....................................................................................................................109110
12.1.3 RMCP Message Format .................................................................................................................110111
12.2 Required ASF/RMCP Messages for IPMI-over-LAN ..........................................................................110111
12.2.1 RMCP ACK Messages...................................................................................................................111112
12.2.2 RMCP ACK Handling .........................................................................................................................112
12.2.3 RMCP/ASF Presence Ping Message..............................................................................................112113
12.2.4 RMCP/ASF Pong Message (Ping Response).................................................................................113114
12.3 IPMI Messages Encapsulation Under RMCP .......................................................................................114115
12.3.1 RMCP/ASF and IPMI Byte Order .................................................................................................114115
12.3.2 Example IPMI over LAN Packet ...................................................................................................115116
12.4 IPMI LAN Message Format..................................................................................................................116117
12.5 LAN Alerting ........................................................................................................................................117118
12.6 IPMI LAN Configuration......................................................................................................................117118
12.6.1 IP and MAC Address Configuration..............................................................................................117118
12.6.2 ‘Teamed’ and Fail-over LAN Channels.........................................................................................117118
12.7 ARP Handling and Gratuitous ARP......................................................................................................118119
12.7.1 OS-Absent problems with ARP .....................................................................................................118119
12.7.2 Resolving ARP issues ....................................................................................................................118119
12.7.3 BMC-generated ARPs....................................................................................................................119120
12.8 Retaining IP Addresses in a DHCP Environment .................................................................................119120
12.8.1 Resolving DHCP issues .................................................................................................................120121
12.9 LAN Session Activation........................................................................................................................120121
13. IPMI Serial/Modem Interface ..................................................................................................... 123124
13.1 Serial/Modem Capabilities ....................................................................................................................123124
13.2 Connection Modes ................................................................................................................................123124
13.2.1 PPP/UDP Proxy Operation ............................................................................................................124125
13.2.2 Asynchronous Communication Parameters ...................................................................................124125
13.2.3 Serial Port Sharing .........................................................................................................................125126
13.2.4 Serial Port Switching .....................................................................................................................126127
13.2.5 Access Modes ................................................................................................................................126127
13.2.6 Console Redirection with Serial Port Sharing................................................................................126127
13.2.6.1 Detecting Who Answered The Phone.......................................................................................127128
13.2.6.2 Connecting to the BMC ............................................................................................................127128
13.2.6.3 Connecting to the Console Redirection ....................................................................................128129
13.2.6.4 Directing the Connection After Power Up / Reset....................................................................128129
13.2.6.5 Interaction with Microsoft ‘Headless’ Operation .....................................................................128129
13.2.6.6 Pre-boot Only Mode .................................................................................................................128129
13.2.6.7 Always Available Mode ...........................................................................................................129130
13.2.6.8 Shared Mode.............................................................................................................................129130
13.2.7 Serial Port Sharing Access Characteristics ....................................................................................129130
13.2.8 Serial Port Sharing Hardware Implementation Notes ....................................................................131132
13.2.9 Connection Mode Auto-detect .......................................................................................................132133
13.2.10 Modem-specific Options................................................................................................................134135
13.2.11 Modem Activation .........................................................................................................................134135
13.3 Serial/Modem Connection Active (Ping) Message ...............................................................................135136
13.3.1 Serial/Modem Connection Active Message Parameters ................................................................136137
13.3.2 Mux Switch Coordination ..............................................................................................................136137
13.3.3 Receive During Ping ......................................................................................................................136137
vii
Intelligent Platform Management Interface Specification
13.3.4 Application Handling of the Serial/Modem Connection Active Message .....................................136137
13.4 Basic Mode............................................................................................................................................137138
13.4.1 Basic Mode Packet Framing ..........................................................................................................137138
13.4.2 Data Byte Escaping ........................................................................................................................137138
13.4.3 Message Fields...............................................................................................................................138139
13.4.4 Message Retries .............................................................................................................................139140
13.4.5 Packet Handshake ..........................................................................................................................139140
13.5 PPP/UDP Mode.....................................................................................................................................140141
13.5.1 PPP/UDP Mode Sessions...............................................................................................................140141
13.5.2 PPP Frame Format .........................................................................................................................140141
13.5.3 PPP Frame Implementation Requirements.....................................................................................140141
13.5.4 Link Control Protocol (LCP) packets ............................................................................................141142
13.5.5 Configuration Requests..................................................................................................................141142
13.5.6 Maximum Receive Unit Handling .................................................................................................143144
13.5.7 Protocol Field Compression Handling ...........................................................................................143144
13.5.8 Address & Control Field Compression Handling ..........................................................................143144
13.5.9 IPMI/RMCP Message Format in PPP Frame.................................................................................144145
13.5.10 Example of IPMI Frame with Field Compression .........................................................................145146
13.5.11 Frame Data Encoding.....................................................................................................................145146
13.5.12 Escaping Algorithm .......................................................................................................................145146
13.5.13 Escaped Character Handling ..........................................................................................................145146
13.5.14 Asynch Control Character Maps (ACCM).....................................................................................145147
13.5.15 IP Network Protocol Negotiation (IPCP).......................................................................................146147
13.5.16 CHAP Operation in PPP Mode ......................................................................................................147148
13.6 Serial/Modem Callback.........................................................................................................................148149
13.6.1 Callback Control Protocol (CBCP) Support ..................................................................................148149
13.6.1.1 CBCP Address Type and Dial String Characters .....................................................................149150
13.7 Terminal Mode......................................................................................................................................149150
13.7.1 Terminal Mode Versus Basic Mode Differences ...........................................................................150151
13.7.2 Terminal Mode Message Format ...................................................................................................150151
13.7.3 IPMI Message Data........................................................................................................................150152
13.7.4 Terminal Mode IPMI Message Bridging .......................................................................................152153
13.7.5 Sending Messages to SMS .............................................................................................................152153
13.7.6 Sending Messages to Other Media.................................................................................................153154
13.7.7 Terminal Mode Packet Handshake ................................................................................................154155
13.7.8 Terminal Mode ASCII Text Commands ........................................................................................154155
13.7.9 Terminal Mode Text Command and IPMI Message Examples .....................................................157158
13.8 Terminal Mode Line Editing.................................................................................................................157158
13.9 Terminal Mode Input Restrictions ........................................................................................................158159
13.10 Page Blackout Interval ..........................................................................................................................158159
13.11 Dial Paging............................................................................................................................................158159
13.11.1 Alert Strings for Dial Paging .........................................................................................................159160
13.11.2 Dialing Digits.................................................................................................................................159160
13.11.3 <Enter> Character (control-M) ......................................................................................................159160
13.11.4 Long Pause Character (control-L)..................................................................................................159160
13.11.5 Empty (delimiter) Character (FFh) ................................................................................................159160
13.11.6 ‘Null’ Terminator Character (00h).................................................................................................159160
13.12 TAP Paging ...........................................................................................................................................160161
13.12.1 TAP Escaping (data transparency).................................................................................................160161
13.12.2 TAP Checksum ..............................................................................................................................160162
13.12.3 TAP Response Codes.....................................................................................................................161162
13.12.4 TAP Page Success Criteria.............................................................................................................161162
13.13 PPP Alerting..........................................................................................................................................161162
14. Event Messages ......................................................................................................................... 162163
viii
Intelligent Platform Management Interface Specification
14.1 Critical Events and System Event Log Restrictions..............................................................................162163
14.2 Event Receiver Handling of Event Messages .......................................................................................163164
14.3 IPMB Seq Field use in Event Messages................................................................................................164165
14.4 Event Status, Event Conditions, and Present State ...............................................................................165166
14.5 System Software use of Sensor Scanning bits & Entity Info ................................................................165166
14.6 Re-arming..............................................................................................................................................166167
14.6.1 ‘Global’ Re-arm .............................................................................................................................166167
15. Platform Event Filtering (PEF).................................................................................................. 168169
15.1 Alert Policies .........................................................................................................................................168169
15.2 Deferred Alerts......................................................................................................................................168169
15.3 PEF Postpone Timer..............................................................................................................................168169
15.4 PEF Startup Delay.................................................................................................................................169170
15.4.1 Last Processed Event Tracking ......................................................................................................169170
15.5 Event Processing When The SEL Is Full ..............................................................................................169170
15.6 PEF Actions ..........................................................................................................................................170171
15.7 Event Filter Table..................................................................................................................................170171
15.8 Event Data 1 Event Offset Mask...........................................................................................................173174
15.9 Using the Mask and Compare Fields ....................................................................................................173174
15.10 Mask and Compare Field Examples......................................................................................................173174
15.11 Alert Policy Table .................................................................................................................................174175
15.12 Alert Testing..........................................................................................................................................175176
15.13 Alert Processing ....................................................................................................................................176177
15.13.1 Alert Processing after Power Loss .................................................................................................176177
15.13.2 Processing non-Alert Actions after Power Loss ............................................................................176177
15.13.3 Alert Processing when IPMI Messaging is in Progress .................................................................176177
15.13.4 Sending Multiple Alerts On One Call ............................................................................................176177
15.13.5 Serial/Modem Alert Processing .....................................................................................................177178
15.14 PEF and Alert Handling Example .........................................................................................................178179
15.15 Event Filter, Policy, Destination, and String Relationships ..................................................................179180
15.16 Populating a PET...................................................................................................................................180181
15.16.1 OEM Custom Fields and Text Alert Strings for IPMI v1.5 PET ...................................................182183
15.17 PEF Performance Target .......................................................................................................................182183
16. Command Specification Information ....................................................................................... 184185
16.1 Specification of Completion Codes.......................................................................................................184185
16.2 Handling ‘Reserved’ Bits and Fields ....................................................................................................184185
16.3 Logical Unit Numbers (LUNs) for Commands.....................................................................................184185
16.4 Command Table Notation .....................................................................................................................184185
17. IPM Device “Global” Commands ............................................................................................. 186187
17.1 Get Device ID Command ......................................................................................................................187188
17.2 Cold Reset Command............................................................................................................................190191
17.3 Warm Reset Command .........................................................................................................................190191
17.4 Get Self Test Results Command ...........................................................................................................191192
17.5 Manufacturing Test On Command........................................................................................................191192
17.6 Set ACPI Power State Command..........................................................................................................192193
17.7 Get ACPI Power State Command .........................................................................................................194195
17.8 Get Device GUID Command ................................................................................................................195196
17.9 Broadcast ‘Get Device ID’....................................................................................................................196197
18. IPMI Messaging Support Commands ...................................................................................... 198199
18.1 Set BMC Global Enables Command.....................................................................................................199200
18.2 Get BMC Global Enables Command ....................................................................................................199200
18.3 Clear Message Flags Command ............................................................................................................200201
18.4 Get Message Flags Command...............................................................................................................200201
18.5 Enable Message Channel Receive Command .......................................................................................201202
ix
Intelligent Platform Management Interface Specification
18.6 Get Message Command.........................................................................................................................201202
18.7 Send Message Command ............................................................................................................................205
18.8 Read Event Message Buffer Command ......................................................................................................207
18.9 Get BT Interface Capabilities Command ....................................................................................................207
18.10 Master Write-Read Command.....................................................................................................................208
18.11 Session Header Fields .................................................................................................................................208
18.12 Get Channel Authentication Capabilities Command...................................................................................209
18.13 Get System GUID Command......................................................................................................................212
18.14 Get Session Challenge Command ...............................................................................................................212
18.15 Activate Session Command.........................................................................................................................213
18.15.1 AuthCode Algorithms ..........................................................................................................................216
18.16 Set Session Privilege Level Command........................................................................................................216
18.17 Close Session Command .............................................................................................................................217
18.18 Get Session Info Command.........................................................................................................................217
18.19 Get AuthCode Command ............................................................................................................................219
18.20 Set Channel Access Command....................................................................................................................221
18.21 Get Channel Access Command...................................................................................................................223
18.22 Get Channel Info Command .......................................................................................................................224
18.23 Set User Access Command .........................................................................................................................225
18.24 Get User Access Command.........................................................................................................................227
18.25 Set User Name Command ...........................................................................................................................228
18.26 Get User Name Command ..........................................................................................................................228
18.27 Set User Password Command .....................................................................................................................229
19. IPMI LAN Commands....................................................................................................................... 230
19.1 Set LAN Configuration Parameters Command ...........................................................................................230
19.2 Get LAN Configuration Parameters Command ..........................................................................................231
19.3 Suspend BMC ARPs Command..................................................................................................................236
19.4 Get IP/UDP/RMCP Statistics Command ....................................................................................................237
20. IPMI Serial/Modem Commands ...................................................................................................... 238
20.1 Set Serial/Modem Configuration Command ...............................................................................................238
20.2 Get Serial/Modem Configuration Command ..............................................................................................239
20.3 Set Serial/Modem Mux Command..............................................................................................................259
20.4 Get TAP Response Codes Command..........................................................................................................260
20.5 Set PPP UDP Proxy Transmit Data Command ...........................................................................................261
20.6 Get PPP UDP Proxy Transmit Data Command ..........................................................................................261
20.7 Send PPP UDP Proxy Packet Command ....................................................................................................262
20.8 Get PPP UDP Proxy Receive Data Command ............................................................................................263
20.9 Serial/Modem Connection Active (Ping) Command...................................................................................264
20.10 Callback Command .....................................................................................................................................264
20.11 Set User Callback Options Command.........................................................................................................265
20.12 Get User Callback Options Command ........................................................................................................266
21. BMC Watchdog Timer Commands................................................................................................. 267
21.1 Watchdog Timer Actions ............................................................................................................................267
21.2 Watchdog Timer Use Field and Expiration Flags .......................................................................................267
21.2.1 Using the Timer Use field and Expiration flags...................................................................................268
21.3 Watchdog Timer Event Logging.................................................................................................................268
21.4 Pre-timeout Interrupt ...................................................................................................................................268
21.4.1 Pre-timeout Interrupt Support Detection..............................................................................................268
21.4.2 BIOS Support for Watchdog Timer .....................................................................................................269
21.5 Reset Watchdog Timer Command ..............................................................................................................269
21.6 Set Watchdog Timer Command ..................................................................................................................269
21.7 Get Watchdog Timer Command .................................................................................................................271
22. Chassis Commands ........................................................................................................................ 273
x
Intelligent Platform Management Interface Specification
22.1
22.2
22.3
22.4
22.5
22.5a
22.6
22.7
22.7a
22.8
22.9
22.10
22.11
22.12
Get Chassis Capabilities Command ............................................................................................................273
Get Chassis Status Command......................................................................................................................275
Chassis Control Command..........................................................................................................................276
Chassis Reset Command .............................................................................................................................276
Chassis Identify Command .........................................................................................................................277
Set Front Panel Enables...............................................................................................................................277
Set Chassis Capabilities Command .............................................................................................................278
Set Power Restore Policy Command...........................................................................................................278
Set Power Cycle Interval.............................................................................................................................279
Remote Access Boot control .......................................................................................................................279
Get System Restart Cause Command..........................................................................................................280
Set System Boot Options Command ...........................................................................................................280
Get System Boot Options Command ..........................................................................................................281
Get POH Counter Command.......................................................................................................................286
23. Event Commands............................................................................................................................. 288
23.1 Set Event Receiver Command.....................................................................................................................288
23.2 Get Event Receiver Command ....................................................................................................................289
23.3 Platform Event Message Command ............................................................................................................289
23.4 Event Request Message Fields ....................................................................................................................290
23.5 IPMB Event Message Formats....................................................................................................................290
23.6 System Interface Event Request Message Format ......................................................................................290
23.7 Event Data Field Formats............................................................................................................................291
24. PEF and Alerting Commands ......................................................................................................... 293
24.1 Get PEF Capabilities Command..................................................................................................................293
24.2 Arm PEF Postpone Timer Command..........................................................................................................293
24.3 Set PEF Configuration Parameters Command ............................................................................................294
24.4 Get PEF Configuration Parameters Command............................................................................................295
24.5 Set Last Processed Event ID Command......................................................................................................300
24.6 Get Last Processed Event ID Command .....................................................................................................300
24.7 Alert Immediate Command .........................................................................................................................300
24.8 PET Acknowledge Command .....................................................................................................................302
25. System Event Log (SEL) ................................................................................................................. 303
25.1 SEL Device Commands ..............................................................................................................................303
25.2 Get SEL Info Command..............................................................................................................................304
25.3 Get SEL Allocation Info Command............................................................................................................305
25.4 Reserve SEL Command ..............................................................................................................................305
25.4.1 Reservation Restricted Commands ......................................................................................................306
25.4.2 Reservation Cancellation .....................................................................................................................306
25.5 Get SEL Entry Command............................................................................................................................307
25.6 Add SEL Entry Command ..........................................................................................................................307
25.6.1 SEL Record Type Ranges ....................................................................................................................308
25.7 Partial Add SEL Entry Command ...............................................................................................................309
25.8 Delete SEL Entry Command.......................................................................................................................309
25.9 Clear SEL Command ..................................................................................................................................310
25.10 Get SEL Time Command ............................................................................................................................310
25.11 Set SEL Time Command.............................................................................................................................310
25.12 Get Auxiliary Log Status Command ...........................................................................................................311
25.13 Set Auxiliary Log Status Command............................................................................................................312
26. SEL Record Formats ....................................................................................................................... 313
26.1 SEL Event Records .....................................................................................................................................313
26.2 OEM SEL Record - Type C0h-DFh............................................................................................................314
26.3 OEM SEL Record - Type E0h-FFh.............................................................................................................314
xi
Intelligent Platform Management Interface Specification
27. SDR Repository................................................................................................................................ 315
27.1 SDR Repository Device ..............................................................................................................................315
27.2 Modal and Non-modal SDR Repositories...................................................................................................316
27.2.1 Command Support while in SDR Repository Update Mode ...............................................................316
27.3 Populating the SDR Repository ..................................................................................................................316
27.3.1 SDR Repository Updating ...................................................................................................................317
27.4 Discovering Management Controllers and Device SDRs ...........................................................................317
27.5 Reading the SDR Repository ......................................................................................................................317
27.6 Sensor Initialization Agent..........................................................................................................................318
27.6.1 System Support Requirements for the Initialization Agent..................................................................318
27.6.2 IPMI and ACPI Interaction ..................................................................................................................318
27.6.3 Recommended Initialization Agent Steps ............................................................................................319
27.7 SDR Repository Device Commands ...........................................................................................................319
27.8 SDR ‘Record IDs’ .......................................................................................................................................320
27.9 Get SDR Repository Info Command...........................................................................................................321
27.10 Get SDR Repository Allocation Info Command...................................................................................321322
27.11 Reserve SDR Repository Command ...........................................................................................................322
27.11.1 Reservation Restricted Commands ......................................................................................................323
27.11.2 Reservation Cancellation .....................................................................................................................323
27.12 Get SDR Command...............................................................................................................................324323
27.13 Add SDR Command....................................................................................................................................325
27.14 Partial Add SDR Command ........................................................................................................................325
27.15 Delete SDR Command ................................................................................................................................326
27.16 Clear SDR Repository Command ...............................................................................................................326
27.17 Get SDR Repository Time Command.........................................................................................................327
27.18 Set SDR Repository Time Command..........................................................................................................327
27.19 Enter SDR Repository Update Mode Command.........................................................................................327
27.20 Exit SDR Repository Update Mode Command...........................................................................................328
27.21 Run Initialization Agent Command ............................................................................................................328
28. FRU Inventory Device Commands ................................................................................................. 329
28.1 Get FRU Inventory Area Info Command....................................................................................................329
28.2 Read FRU Data Command..........................................................................................................................330
28.3 Write FRU Data Command.........................................................................................................................330
29. Sensor Device Commands ............................................................................................................. 332
29.1 Static and Dynamic Sensor Devices............................................................................................................333
29.2 Get Device SDR Info Command.................................................................................................................333
29.3 Get Device SDR Command ........................................................................................................................334
29.4 Reserve Device SDR Repository Command...............................................................................................334
29.5 Get Sensor Reading Factors Command.......................................................................................................335
29.6 Set Sensor Hysteresis Command.................................................................................................................335
29.7 Get Sensor Hysteresis Command ................................................................................................................336
29.8 Set Sensor Thresholds Command................................................................................................................336
29.9 Get Sensor Thresholds Command...............................................................................................................337
29.10 Set Sensor Event Enable Command............................................................................................................338
29.11 Get Sensor Event Enable Command ...........................................................................................................340
29.12 Re-arm Sensor Events Command................................................................................................................341
29.13 Get Sensor Event Status Command.............................................................................................................343
29.13.1 Response According to Sensor Type ...................................................................................................343
29.13.2 Hysteresis and Event Status .................................................................................................................344
29.13.3 High-going versus Low-going Threshold Events ................................................................................344
29.13.4 Get Sensor Event Status Command Format .........................................................................................345
29.14 Get Sensor Reading Command ...................................................................................................................348
29.15 Set Sensor Type Command .........................................................................................................................349
29.16 Get Sensor Type Command ........................................................................................................................349
xii
Intelligent Platform Management Interface Specification
30. Sensor Types and Data Conversion .............................................................................................. 351
30.1 Linear and Linearized Sensors ....................................................................................................................351
30.2 Non-Linear Sensors.....................................................................................................................................351
30.3 Sensor Reading Conversion Formula..........................................................................................................352
30.4 Resolution, Tolerance and Accuracy...........................................................................................................352
30.4.1 Tolerance..............................................................................................................................................352
30.4.2 Resolution ............................................................................................................................................352
30.4.2.1 Resolution for Non-linear & Linearizable Sensors.........................................................................353
30.4.2.2 Offset Constant Relationship to Resolution....................................................................................353
30.5 Management Software, SDRs, and Sensor Display ....................................................................................353
30.5.1 Software Display of Threshold Settings ..............................................................................................353
30.5.2 Notes on Displaying Sensor Readings & Thresholds ..........................................................................354
31. Timestamp Format........................................................................................................................... 356
31.1 Special Timestamp values ...........................................................................................................................356
32. Accessing FRU Devices.................................................................................................................. 357
33. Using Entity IDs ............................................................................................................................... 359
33.1 System- and Device-relative Entity Instance Values ..................................................................................359
33.2 Restrictions on Using Device-relative Entity Instance Values....................................................................360
33.3 Sensor-to-FRU Association.........................................................................................................................360
34. Handling Sensor Associations....................................................................................................... 361
34.1 Entity Presence............................................................................................................................................361
34.2 Software detection of Entities .....................................................................................................................361
34.3 Using Entity Association Records...............................................................................................................362
35. Sensor & Event Message Codes .................................................................................................... 364
35.1 Sensor Type Code .......................................................................................................................................364
35.2 Event/Reading Type Code ..........................................................................................................................364
35.3 SDR Specification of Event Types..............................................................................................................365
35.4 SDR Specification of Reading Types..........................................................................................................365
35.5 Use of Codes in Event Messages ................................................................................................................366
36. Sensor and Event Code Tables ...................................................................................................... 367
36.1 Event/Reading Type Codes.........................................................................................................................367
36.2 Sensor Type Codes and Data ......................................................................................................................371
37. Sensor Data Record Formats ......................................................................................................... 382
37.1 SDR Type 01h, Full Sensor Record ............................................................................................................383
37.2 SDR Type 02h, Compact Sensor Record ....................................................................................................390
37.3 SDR Type 03h, Event-Only Record............................................................................................................396
37.4 SDR Type 08h - Entity Association Record ...............................................................................................398
37.5 SDR Type 09h - Device-relative Entity Association Record ......................................................................400
37.6 SDR Type 0Ah:0Fh - Reserved Records ....................................................................................................401
37.7 SDR Type 10h - Generic Device Locator Record.......................................................................................402
37.8 SDR Type 11h - FRU Device Locator Record............................................................................................403
37.9 SDR Type 12h - Management Controller Device Locator Record .............................................................405
37.10 SDR Type 13h - Management Controller Confirmation Record ................................................................407
37.11 SDR Type 14h - BMC Message Channel Info Record ...............................................................................408
37.12 SDR Type C0h - OEM Record ...................................................................................................................410
37.13 Device Type Codes .....................................................................................................................................411
37.14 Entity IDs ....................................................................................................................................................413
37.15 Type/Length Byte Format ...........................................................................................................................414
37.16 6-bit ASCII Packing Example.....................................................................................................................415
37.17 Sensor Unit Type Codes..............................................................................................................................416
38. Examples .......................................................................................................................................... 417
xiii
Intelligent Platform Management Interface Specification
38.1 Processor Sensor with Sensor-specific States & Event Generation ............................................................417
38.2 Processor Sensor with Generic States & Event Generation ........................................................................419
Appendix A - Previous Sequence Number Tracking........................................................................... 420
Appendix B - Example PEF Mask Compare Algorithm ....................................................................... 421
Appendix C1 - Locating IPMI System Interfaces via SM BIOS Tables............................................... 422
C1-1 IPMI Device Information - BMC Interface.................................................................................................423
C1-1.1 Interface Type ......................................................................................................................................423
C1-1.2 IPMI Specification Revision Field.......................................................................................................423
C1-1.3 I2C Slave Address Field .......................................................................................................................423
C1-1.4 NV Storage Device Address Field .......................................................................................................423
C1-1.5 Base Address Field...............................................................................................................................423
C1-1.6 Base Address Modifier Field ...............................................................................................................424
C1-1.7 System Interface Register Alignment...................................................................................................424
C1-1.7.1 Byte-spaced I/O Address Examples ...............................................................................................424
C1-1.7.2 32-bit Spaced I/O Address Examples .............................................................................................424
C1-1.7.3 Memory-mapped Base Address......................................................................................................424
C1-1.7.4 Interrupt Info Field .........................................................................................................................424
C1-1.8 Interrupt Number Field ........................................................................................................................424
Appendix C2 - Locating IPMI System Interfaces on PCI ..................................................................... 426
Appendix C3 - Locating IPMI System Interfaces with ACPI ............................................................... 427
C3-1 SPMI Description Table and ACPI Control Methods.................................................................................427
C3-2 Locating IPMI System Interfaces in ACPI Name Space.............................................................................429
C3-3 Example IPMI Definition ASL Code..........................................................................................................430
Example 1: SMIC Interface in I/O Space ..........................................................................................................430
Example 2: KCS Interface in 64-bit Address Space..........................................................................................431
Example 3: SMIC Interface in I/O Space ..........................................................................................................432
Appendix D - Determining Message Size Requirements .................................................................... 433
Appendix E - Terminal Mode Grammar................................................................................................. 435
E-1 Notation.......................................................................................................................................................435
E-2 Grammar for Terminal Mode Input.............................................................................................................435
E-3 Grammar for Terminal Mode Output ..........................................................................................................436
Appendix F - TAP Flow Summary ......................................................................................................... 438
Appendix G - Command Assignments ................................................................................................. 442
Index ............................................................................................................................................................. I
xiv
Intelligent Platform Management Interface Specification
List of Figures
Figure 1-1, IPMI and the Management Software Stack ................................................................................................8
Figure 1-2, IPMI Block Diagram ..................................................................................................................................9
Figure 2-1, Intelligent Platform Management Logical Devices ..................................................................................25
Figure 6-1, Session Activation ................................................................................................................................5253
Figure 6-2, LAN to IPMB Bridged Request Example ............................................................................................6061
Figure 7-1, IPMB Request sent using Send Message Command ............................................................................6566
Figure 7-2, Send Message Command Response......................................................................................................6566
Figure 7-3, Response for Set Event Receiver in Receive Message Queue..............................................................6667
Figure 7-4, Get Message Command Response........................................................................................................6667
Figure 9-1, KCS Interface/BMC Request Message Format ....................................................................................7273
Figure 9-2, KCS Interface/BMC Response Message Format..................................................................................7374
Figure 9-3, KCS Interface Event Request Message Format....................................................................................7374
Figure 9-4, KCS Interface Event Response Message Format .................................................................................7374
Figure 9-5, KCS Interface Registers........................................................................................................................7475
Figure 9-6, KCS Interface SMS to BMC Write Transfer Flow Chart.....................................................................8081
Figure 9-7, KCS Interface BMC to SMS Read Transfer Flow Chart......................................................................8182
Figure 9-8, Aborting KCS Transactions in-progress and/or Retrieving KCS Error Status .....................................8283
Figure 10-1, SMIC/BMC Interface Registers..........................................................................................................8788
Figure 10-2, SMIC/BMC Request Message Format ...............................................................................................9596
Figure 10-3, SMIC/BMC Response Message Format .............................................................................................9697
Figure 10-4, SMIC Event Request Message Format ...............................................................................................9697
Figure 10-5, SMIC Event Response Message Format.............................................................................................9697
Figure 11-1, BT Interface/BMC Request Message Format.....................................................................................9899
Figure 11-2, BT Interface/BMC Response Message Format ................................................................................99100
Figure 11-3, BT Interface Event Request Message Format.................................................................................100101
Figure 11-4, BT Interface Event Response Message Format ..............................................................................100101
Figure 11-5, BT_CTRL Register format .............................................................................................................102103
Figure 11-6, BT_INTMASK Register format .....................................................................................................104105
Figure 12-1, Embedded LAN Controller Implementation...................................................................................107108
Figure 12-2, PCI Management Bus Implementation ...........................................................................................108109
Figure 12-3, IPMI LAN Packet Layering............................................................................................................114115
Figure 12-4, IPMI LAN Message Formats..........................................................................................................116117
Figure 12-5, LAN Session Startup ......................................................................................................................122123
Figure 13-1, Serial Port Sharing Logical Diagram..............................................................................................125126
Figure 13-2, Basic Mode Message Fields ...........................................................................................................138139
Figure 13-3, PPP Frame Format ..........................................................................................................................140141
Figure 13-4, Configure-Request, -Ack, -Nak, -Reject Packet Format ................................................................141142
Figure 13-5, IPMI Message in PPP Frame Format..............................................................................................144145
Figure 13-6, IP Frame with Field Compression ..................................................................................................145146
Figure 13-7, Terminal Mode Request to BMC....................................................................................................151152
Figure 13-8, Terminal Mode Response from BMC.............................................................................................151152
Figure 13-9, Terminal Mode Request to SMS.....................................................................................................153154
Figure 13-10, Terminal Mode Response from SMS ...........................................................................................153154
Figure 13-11, Send Message Command for Bridged Request.............................................................................153154
Figure 13-12, Response to Send Message Command for Bridged Request ........................................................153154
Figure 13-13, Bridged Response to Remote Console..........................................................................................153154
Figure 15-1, Alert Processing Example...............................................................................................................179180
Figure 15-2, Event Filter, Alert Policy, and Alert Destination, & String Relationships .....................................180181
xv
Intelligent Platform Management Interface Specification
Figure 17-1, Broadcast Get Device ID Request Message ...................................................................................196197
Figure 18-1, AuthCode Algorithms...........................................................................................................................216
Figure 23-1, IPMB Event Request Message Format .................................................................................................290
Figure 23-2, Example SMIC Event Request Message Format ..................................................................................291
Figure 29-1, High-Going and Low-Going Event Assertion/Deassertion Points .......................................................345
Figure 33-1, Sensor to FRU Lookup .........................................................................................................................360
Figure 37-1, 6-bit Packed ASCII Example................................................................................................................415
Figure B-1, Example Event Data Comparison Algorithm.........................................................................................421
Figure D-1, SMBus Write-Block by Master Write-Read through KCS/SMIC .........................................................433
Figure D-2, Master Write-Read Response via KCS/SMIC ......................................................................................433
Figure D-3, Get Message Response via KCS/SMIC ................................................................................................433
Figure D-4, Master Write-Read Request via LAN/PPP ...........................................................................................434
Figure D-5 Master Write-Read Response via LAN/PPP..........................................................................................434
Figure D-6, Master Write-Read Response via LAN/PPP.........................................................................................434
xvi
Intelligent Platform Management Interface Specification
List of Tables
Table 1-1, Glossary .......................................................................................................................................................4
Table 3-1, Required BMC Functions ..........................................................................................................................30
Table 5-1, Network Function Codes ...........................................................................................................................35
Table 5-2, Completion Codes......................................................................................................................................37
Table 5-3, Sensor Owner ID and Sensor Number Field Definitions ...........................................................................39
Table 5-4, System Software IDs..................................................................................................................................40
Table 6-1, Channel Number Assignments...................................................................................................................43
Table 6-2, Channel Protocol Type Numbers ...........................................................................................................4344
Table 6-3, Channel Medium Type Numbers ...........................................................................................................4445
Table 6-4, Channel Access Modes ..........................................................................................................................4546
Table 6-5, Channel Privilege Levels .......................................................................................................................4647
Table 6-6, Session-less , Single-session and Multi-session Characteristics ............................................................5152
Table 6-7, Default Session Inactivity Timeout Intervals.........................................................................................5556
Table 6-8, Message Bridging Mechanism by Source and Destination....................................................................5859
Table 6-9, IPMI Message and IPMB / Private Bus Transaction Size Requirements...............................................6162
Table 7-1, BMC IPMB LUNs .................................................................................................................................6465
Table 8-1, System Interface Request For Delivering Remote IPMB Request via ICMB........................................6970
Table 8-2, Send Message Response ........................................................................................................................6970
Table 8-3, IPMB Request For Delivering Remote IPMB Request via ICMB.........................................................7071
Table 8-4, Send Message Response ........................................................................................................................7071
Table 8-5, IPMB Response For Remote IPMB Request Delivered via ICMB........................................................7071
Table 9-1, KCS Interface Status Register Bits ........................................................................................................7576
Table 9-2, KCS Interface State Bits ........................................................................................................................7576
Table 9-3, KCS Interface Control Codes.................................................................................................................7677
Table 9-4, KCS Interface Status Codes ...................................................................................................................7677
Table 10-1, SMIC Flags Register Bits.....................................................................................................................8889
Table 10-2, SMS Transfer Stream control codes.....................................................................................................9394
Table 10-3, SMS Transfer Stream Status Codes .....................................................................................................9495
Table 11-1, BT Interface Registers......................................................................................................................101102
Table 11-2, BT_CTRL Register Bit Definitions .................................................................................................102103
Table 11-3, BT_INTMASK Register Bit Definitions .........................................................................................104105
Table 11-4, BT Interface Write Transfer.............................................................................................................105106
Table 11-5, BT Interface Read Transfer..............................................................................................................106107
Table 12-1, RMCP Port Numbers .......................................................................................................................109110
Table 12-2, RMCP Message Format ...................................................................................................................110111
Table 12-3, Message Type Determination Under RMCP....................................................................................110111
Table 12-4, ASF/RMCP Messages for IPMI-over-LAN.....................................................................................111112
Table 12-5, RMCP ACK Message Fields............................................................................................................111112
Table 12-6, RMCP Packet Fields for ASF Presence Ping Message (Ping Request) ...........................................112113
Table 12-7, RMCP Packet Fields for ASF Presence Pong Message (Ping Response)........................................113114
Table 12-8, RMCP Packet for IPMI via Ethernet................................................................................................115116
Table 13-1, Serial Port Switching Triggers .........................................................................................................126127
Table 13-2, Serial Port Sharing Access Characteristics ......................................................................................129130
Table 13-3, Auto-Connection Mode Patterns......................................................................................................133134
Table 13-4, Modem String Summary ..................................................................................................................134135
Table 13-5, Basic Mode Special Characters........................................................................................................137138
Table 13-6, BASIC MODE Data Byte Escape Encoding....................................................................................137138
Table 13-7, LCP Code Fields ..............................................................................................................................141142
xvii
Intelligent Platform Management Interface Specification
Table 13-8, Overview of PPP Configure-Ack, -Nak, & -Reject Packet Use ......................................................141142
Table 13-9, PPP Link Configuration Option Support Requirements ..................................................................142143
Table 13-10, Default Escaped Characters ...........................................................................................................145146
Table 13-11, CBCP Callback Number Options...................................................................................................149150
Table 13-12, Terminal Mode Message Bridge Field...........................................................................................152153
Table 13-13, Terminal Mode Text Commands ...................................................................................................154155
Table 13-14, Terminal Mode Examples ..............................................................................................................157158
Table 13-15, TAP Escaping ................................................................................................................................160161
Table 13-16, TAP Success Codes .......................................................................................................................161162
Table 14-1, Event Message Reception ................................................................................................................163164
Table 15-1, PEF Action Priorities .......................................................................................................................170171
Table 15-2, Event Filter Table Entry...................................................................................................................171172
Table 15-3, Comparison-type Selection according to Compare Field bits ..........................................................173174
Table 15-4, Alert Policy Table Entry ..................................................................................................................175176
Table 15-5, Serial/Modem Alert Destination Priorities.......................................................................................177178
Table 15-6, PET Specific Trap Fields .................................................................................................................180181
Table 15-7 - PET Variable Bindings Field ..........................................................................................................181182
Table 15-8, IPMI PET Multirecord Field Format ...............................................................................................182183
Table 17-1, IPM Device ‘Global’ Commands.....................................................................................................186187
Table 17-2, Get Device ID Command.................................................................................................................187188
Table 17-3, Cold Reset Command.......................................................................................................................190191
Table 17-4, Warm Reset Command ....................................................................................................................190191
Table 17-5, Get Self Test Results Command ......................................................................................................191192
Table 17-6, Manufacturing Test On ....................................................................................................................192193
Table 17-7, Set ACPI Power State Command.....................................................................................................193194
Table 17-8, Get ACPI Power State Command ....................................................................................................194195
Table 17-9, Get Device GUID Command ...........................................................................................................195196
Table 17-10, GUID Format .................................................................................................................................195196
Table 18-1, IPMI Messaging Support Commands ..............................................................................................198199
Table 18-2, Set BMC Global Enables Command................................................................................................199200
Table 18-3, Get BMC Global Enables Command ...............................................................................................200201
Table 18-4, Clear Message Flags Command .......................................................................................................200201
Table 18-5, Get Message Flags Command................................................................................................................201
Table 18-6, Enable Message Channel Receive Command ..................................................................................201202
Table 18-7, Get Message Command .........................................................................................................................203
Table 18-8, Get Message Data Fields........................................................................................................................204
Table 18-9, Send Message Command .......................................................................................................................205
Table 18-10, Message Data for Send Message Command ........................................................................................206
Table 18-11, Read Event Message Buffer Command ...............................................................................................207
Table 18-12, Get BT Interface Capabilities Command .............................................................................................207
Table 18-13, Master Write-Read Command .............................................................................................................208
Table 18-14, Get Channel Authentication Capabilities Command ...........................................................................210
Table 18-15, Get System GUID Command...............................................................................................................212
Table 18-16, Get Session Challenge Command ........................................................................................................213
Table 18-17, Activate Session Command .................................................................................................................214
Table 18-18, Set Session Privilege Level Command ................................................................................................217
Table 18-19, Close Session Command......................................................................................................................217
Table 18-20, Get Session Info Command .................................................................................................................218
Table 18-21, Get AuthCode Command .....................................................................................................................220
Table 18-22, Set Channel Access Command ............................................................................................................221
Table 18-23, Get Channel Access Command............................................................................................................223
xviii
Intelligent Platform Management Interface Specification
Table 18-24, Get Channel Info Command ................................................................................................................224
Table 18-25, Set User Access Command ..................................................................................................................226
Table 18-26, Get User Access Command..................................................................................................................227
Table 18-27, Set User Name Command ....................................................................................................................228
Table 18-28, Get User Name Command ...................................................................................................................228
Table 18-29, Set User Password Command ..............................................................................................................229
Table 19-1, IPMI LAN Commands ...........................................................................................................................230
Table 19-2, Set LAN Configuration Parameters Command......................................................................................230
Table 19-3, Get LAN Configuration Parameters Command .....................................................................................231
Table 19-4, LAN Configuration Parameters .............................................................................................................232
Table 19-5, Suspend BMC ARPs Command ............................................................................................................236
Table 19-6, Get IP/UDP/RMCP Statistics Command ...............................................................................................237
Table 20-1, IPMI Serial/Modem Commands ............................................................................................................238
Table 20-2, Set Serial/Modem Configuration Command ..........................................................................................238
Table 20-3, Get Serial/Modem Configuration Command .........................................................................................239
Table 20-4, Serial/Modem Configuration Parameters...............................................................................................240
Table 20-5, Set Serial/Modem Mux Command.........................................................................................................259
Table 20-6, Get TAP Response Codes Command.....................................................................................................260
Table 20-7, Set PPP UDP Proxy Transmit Data Command ......................................................................................261
Table 20-8, Get PPP UDP Proxy Transmit Data Command .....................................................................................261
Table 20-9, Send PPP UDP Proxy Packet Command ...............................................................................................262
Table 20-10, Get PPP UDP Proxy Receive Data Command.....................................................................................263
Table 20-11, Serial/Modem Connection Active Command ......................................................................................264
Table 20-12, Callback Command ..............................................................................................................................264
Table 20-13, Set User Callback Options Command..................................................................................................265
Table 20-14, Get User Callback Options Command .................................................................................................266
Table 21-1, BMC Watchdog Timer Commands........................................................................................................267
Table 21-2, Reset Watchdog Timer Command .........................................................................................................269
Table 21-3, Set Watchdog Timer Command .............................................................................................................270
Table 21-4, Get Watchdog Timer Command ............................................................................................................271
Table 22-1, Chassis Commands ................................................................................................................................273
Table 22-2, Get Chassis Capabilities Command .......................................................................................................274
Table 22-3, Get Chassis Status Command ................................................................................................................275
Table 22-4, Chassis Control Command.....................................................................................................................276
Table 22-5, Chassis Reset Command ........................................................................................................................277
Table 22-6, Chassis Identify Command ....................................................................................................................277
Table 22-6a, Set Front Panel Button Enables Command ..........................................................................................277
Table 22-7, Set Chassis Capabilities Command........................................................................................................278
Table 22-8, Set Power Restore Policy Command......................................................................................................278
Table 22-8a, Set Power Cycle Interval Command ....................................................................................................279
Table 22-9, Get System Restart Cause Command.....................................................................................................280
Table 22-10, Set System Boot Options Command ....................................................................................................281
Table 22-11, Get System Boot Options Command ...................................................................................................281
Table 22-12, Boot Option Parameters .......................................................................................................................282
Table 22-13, Get POH Counter Command................................................................................................................286
Table 23-1, Event Commands ...................................................................................................................................288
Table 23-2, Set Event Receiver .................................................................................................................................288
Table 23-3, Get Event Receiver Command ...............................................................................................................289
Table 23-4, Platform Event (Event Message) Command ..........................................................................................289
Table 23-5, Event Request Message Fields...............................................................................................................290
Table 23-6, Event Request Message Event Data Field Contents ..............................................................................291
xix
Intelligent Platform Management Interface Specification
Table 24-1, PEF and Alerting Commands.................................................................................................................293
Table 24-2, Get PEF Capabilities Command.............................................................................................................293
Table 24-3, Arm PEF Postpone Timer Command.....................................................................................................294
Table 24-4, Set PEF Configuration Parameters Command .......................................................................................294
Table 24-5, Get PEF Configuration Parameters Command ......................................................................................295
Table 24-6, PEF Configuration Parameters...............................................................................................................295
Table 24-7, Set Last Processed Event ID Command.................................................................................................300
Table 24-8, Get Last Processed Event ID Command ................................................................................................300
Table 24-9, Alert Immediate Command ....................................................................................................................301
Table 24-10, PET Acknowledge Command..............................................................................................................302
Table 25-1, SEL Device Commands .........................................................................................................................303
Table 25-2, Get SEL Info Command.........................................................................................................................304
Table 25-3, Get SEL Allocation Info Command.......................................................................................................305
Table 25-4, Reserve SEL Command .........................................................................................................................306
Table 25-5, Get SEL Entry........................................................................................................................................307
Table 25-6, Add SEL Entry.......................................................................................................................................308
Table 25-7, Partial Add SEL Entry Command ..........................................................................................................309
Table 25-8, Delete SEL Entry ...................................................................................................................................309
Table 25-9, Clear SEL...............................................................................................................................................310
Table 25-10, Get SEL Time Command .....................................................................................................................310
Table 25-11, Set SEL Time Command......................................................................................................................310
Table 25-12, Get Auxiliary Log Status Command ....................................................................................................311
Table 25-13, Set Auxiliary Log Status Command.....................................................................................................312
Table 26-1, SEL Event Records ................................................................................................................................313
Table 26-2, OEM SEL Record (Type C0h-DFh) ......................................................................................................314
Table 26-3, OEM SEL Record (Type E0h-FFh) .......................................................................................................314
Table 27-1, Mandatory SDR Update Mode Commands............................................................................................316
Table 27-2, SDR Repository Device Commands ......................................................................................................320
Table 27-3, Get SDR Repository Info Command .....................................................................................................321
Table 27-4, Get SDR Repository Allocation Info Command....................................................................................322
Table 27-5, Reserve SDR Repository Command ......................................................................................................323
Table 27-6, Get SDR Command................................................................................................................................324
Table 27-7, Add SDR Command ..............................................................................................................................325
Table 27-8, Partial Add SDR Command ...................................................................................................................325
Table 27-9, Delete SDR Command ...........................................................................................................................326
Table 27-10, Clear SDR Repository Command ........................................................................................................326
Table 27-11, Get SDR Repository Time Command..................................................................................................327
Table 27-12, Set SDR Repository Time Command ..................................................................................................327
Table 27-13, Enter SDR Repository Update Mode Command .................................................................................327
Table 27-14, Exit SDR Repository Update Mode Command ...................................................................................328
Table 27-15, Run Initialization Agent.......................................................................................................................328
Table 28-1, FRU Inventory Device Commands ........................................................................................................329
Table 28-2, Get FRU Inventory Area Info Command...............................................................................................329
Table 28-3, Read FRU Data Command.....................................................................................................................330
Table 28-4, Write FRU Data Command....................................................................................................................331
Table 29-1, Sensor Device Commands .....................................................................................................................332
Table 29-2, Get Device SDR Info Command............................................................................................................333
Table 29-3, Get Device SDR Command ...................................................................................................................334
Table 29-4, Reserve Device SDR Repository ...........................................................................................................334
Table 29-5, Get Sensor Reading Factors Command .................................................................................................335
Table 29-6, Set Sensor Hysteresis .............................................................................................................................336
xx
Intelligent Platform Management Interface Specification
Table 29-7, Get Sensor Hysteresis ............................................................................................................................336
Table 29-8, Set Sensor Thresholds ............................................................................................................................336
Table 29-9, Get Sensor Thresholds ...........................................................................................................................337
Table 29-10, Set Sensor Event Enable ......................................................................................................................338
Table 29-11, Get Sensor Event Enable......................................................................................................................340
Table 29-12, Re-arm Sensor Events ..........................................................................................................................342
Table 29-13, Get Sensor Event Status Response Overview ......................................................................................344
Table 29-14, Get Sensor Event Status Command .....................................................................................................345
Table 29-15, Get Sensor Reading Command ............................................................................................................348
Table 29-16, Set Sensor Type Command ..................................................................................................................349
Table 29-17, Get Sensor Type...................................................................................................................................350
Table 32-1, FRU Device Locator Field Usage ..........................................................................................................358
Table 33-1, System and Device-Relative Entity Instance Values .............................................................................359
Table 36-1, Event/Reading Type Code Ranges.........................................................................................................368
Table 36-2, Generic Event/Reading Type Codes ......................................................................................................368
Table 36-3, Sensor Type Codes.................................................................................................................................371
Table 37-1, Full Sensor Record - SDR Type 01h......................................................................................................383
Table 37-2, Compact Sensor Record - SDR Type 02h..............................................................................................390
Table 37-3, Event-Only Sensor Record - SDR Type 03h .........................................................................................396
Table 37-4, Entity Association Record - SDR Type 08h ..........................................................................................399
Table 37-5, Device-relative Entity Association Record - SDR Type 09h.................................................................400
Table 37-6, Generic Device Locator Record - SDR Type 10h..................................................................................402
Table 37-7, FRU Device Locator Record - SDR Type 11h ......................................................................................403
Table 37-8, Management Controller Device Locator - SDR 12h..............................................................................405
Table 37-9, Management Controller Confirmation Record - SDR Type 13h ...........................................................407
Table 37-10, BMC Message Channel Info Record - SDR Type 14h ........................................................................408
Table 37-11, OEM Record - SDR Type C0h ............................................................................................................410
Table 37-12, IPMB/I2C Device Type Codes .............................................................................................................411
Table 37-13, Entity ID Codes....................................................................................................................................413
Table 37-14, 6-bit ASCII definition ..........................................................................................................................415
Table 37-15, Sensor Unit Type Codes.......................................................................................................................416
Table 38-1, Example discrete Processor sensor with Sensor-specific states & event generation .............................418
Table 38-2, Example discrete Processor sensor with Generic states & event generation .........................................419
Table C1-1, SM BIOS IPMI Device Information Record .........................................................................................422
Table C1-2, Interface Type field values ....................................................................................................................423
Table C1-3, Byte-aligned I/O Mapped Register Address examples..........................................................................424
Table C1-4, 32-bit aligned I/O Mapped Register Address examples ........................................................................424
Table C2-1, PCI Class Codes for IPMI .....................................................................................................................426
Table C3-1, Service Processor Management Interface Description Table Format....................................................427
Table C3-2, IPMI Device Object Control Methods...................................................................................................429
Table F-1, TAP Flow Summary ................................................................................................................................438
Table G-1, Command Number Assignments and Privilege Levels ...........................................................................443
xxi
Intelligent Platform Management Interface Specification
xxii
Intelligent Platform Management Interface Specification
1. Introduction
This document presents the base specifications for the Intelligent Platform Management Interface (IPMI)
architecture. The IPMI specifications define standardized, abstracted interfaces to the platform management
subsystem. IPMI includes the definition of interfaces for extending platform management between board within the
main chassis, and between multiple chassis.
The term “platform management” is used to refer to the monitoring and control functions that are built in to the
platform hardware and primarily used for the purpose of monitoring the health of the system hardware. This
typically includes monitoring elements such as system temperatures, voltages, fans, power supplies, bus errors,
system physical security, etc. It includes automatic and manually driven recovery capabilities such as local or
remote system resets and power on/off operations. It includes the logging of abnormal or ‘out-of-range’ conditions
for later examination and alerting where the platform issues the alert without aid of run-time software. Lastly it
includes inventory information that can help identify a failed hardware unit.
This document is the main specification for IPMI. It defines the overall architecture, common commands, event
formats, data records, and capabilities used across IPMI-based systems and peripheral chassis. This includes the
specifications for IPMI management via LAN, serial/modem, PCI Management bus, and the local interface to the
platform management. In addition to this document, there is a set of separate supporting specifications:
•
The Intelligent Platform Management Bus (IPMB) is an I2C*-based bus that provides a standardized
interconnection between different boards within a chassis. The IPMB can also serve as a standardized interface
for auxiliary or ‘emergency’ management add-in cards.
•
IPMB v1.0 Address Allocation documents the different ranges and assignments of addresses on the IPMB.
•
The Intelligent Chassis Management Bus (ICMB) provides a standardized interface for platform management
information and control between chassis. The ICMB is designed so it can be implemented with a device that
connects to the IPMB. This allows the ICMB to be implemented as an add-on to systems that have an existing
IPMB. See [ICMB] for more information.
•
The Platform Event Trap Format specification defines the format of SNMP traps used for alerts.
•
The Platform Management FRU Information Storage Definition defines the format of Field Replaceable Unit
information (information such as serial numbers and part numbers for various replaceable boards and other
components) accessible in an IPMI-based system.
The implementation of certain aspects of IPMI may require access to other specifications and documents that are not
part Refer to the Reference DocumentsReference Documents section below, for these and other supporting
documents.
1.1 Audience
This document is written for engineers and system integrators involved in the design of and interface to platform
management hardware, and System Management Software (SMS) developers. Familiarity with microcontrollers,
software programming, and PC and Intel server architecture is assumed. For basic and/or supplemental
information, refer to the appropriate reference documents.
1
Intelligent Platform Management Interface Specification
1.2 Reference Documents
The following documents are companion and supporting specifications for IPMI and associated interfaces:
[ACPI 1.0] Advanced Configuration and Power Interface Specification, Revision 1.0b, February 8, 1999. ©1999,
Copyright © 1996, 1997, 1998, 1999 Intel Corporation, Microsoft Corporation, Toshiba Corp.
http://www.teleport.com/~acpi/
[ACPI 2.0] Advanced Configuration and Power Interface Specification, Revision 2.0cRevision 2.0, August 25,
2003July 27, 2000. ©1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003 Compaq Computer
Corporation, Intel Corporation, Microsoft Corporation, Phoenix Technologies Ltd., Toshiba
Corporation. http://www.teleport.com/~acpi/http://www.acpi.info
[ADDR]
IPMB v1.0 Address Allocation, ©2001 Intel Corporation, Hewlett-Packard Company, NEC
Corporation, and Dell Computer Corporation. This document specifies the allocation of I2C addresses
on the IPMB. http://developer.intel.com/design/servers/ipmi
[ASF]
Alert Standard Format v1.0 Specification, ©2001, Distributed Management Task Force.
http://www.dmtf.org
[CBCP]
Proposal for Callback Control Protocol (CBCP), draft-ietf-pppext-callback-cp-02.txt, N. Gidwani,
Microsoft, July 19, 1994. As of this writing, the specification is available via the Microsoft
Corporation web site: http://www.microsoft.com
[FRU]
Platform Management FRU Information Storage Definition v1.0, ©1999 Intel Corporation, HewlettPackard Company, NEC Corporation, and Dell Computer Corporation. Provides the field definitions
and format of Field Replaceable Unit (FRU) information.
http://developer.intel.com/design/servers/ipmi
[I2C]
The I2C Bus And How To Use It, ©1995, Philips Semiconductors. This document provides the timing
and electrical specifications for I2C busses.
[ICMB]
Intelligent Chassis Management Bus Bridge Specification v1.0, rev. 1.23, © 2000 2002 Intel
Corporation. Provides the electrical, transport protocol, and specific command specifications for the
ICMB and information on the creation of management controllers that connect to the ICMB.
http://developer.intel.com/design/servers/ipmi
[IPMB]
Intelligent Platform Management Bus Communications Protocol Specification v1.0, ©1998 Intel
Corporation, Hewlett-Packard Company, NEC Corporation, and Dell Computer Corporation. This
document provides the electrical, transport protocol, and specific command specifications for the
IPMB. http://developer.intel.com/design/servers/ipmi
[MSVT]
Windows Platform Design Notes, Building Hardware and Firmware to Complement Microsoft
Windows Headless Operation, ©2001, Microsoft Corporation. http://www.microsoft.com
[PET]
IPMI Platform Event Trap Format Specification v1.0, ©1998, Intel Corporation, Hewlett-Packard
Company, NEC Corporation, and Dell Computer Corporation. This document specifies a common
format for SNMP Traps for platform events.
[RFC826]
An Ethernet Address Resolution Protocol -- or -- Converting Network Protocol Addresses to 48-bit
Ethernet Address for Transmission on Ethernet Hardware, David C. Plummer, November 1982
[RFC1319] RFC 1319, The MD2 Message-Digest Algorithm, B. Kaliski, RSA Laboratories, April 1992.
2
Intelligent Platform Management Interface Specification
[RFC1321] RFC 1321, The MD5 Message-Digest Algorithm, R. Rivest, MIT Laboratory for Computer Science
and RSA Data Security, Inc. April, 1992.
[RFC1332] RFC 1332, The PPP Internet Protocol Control Protocol (IPCP), G. McGreggor, Merit, May 1992.
[RFC1334] RFC 1334, PPP Authentication Protocols, B. Lloyd, L&A, W. Simpson, Daydreamer, October 1992.
Document includes specification for PAP (Password Authentication Protocol).
[RFC1661] RFC 1661, STD 51, The Point-to-Point Protocol (PPP), Simpson, W., Editor, Daydreamer, July 1994.
[RFC1662] RFC 1662, STD 51, PPP in HDLC-like Framing, Simpson, W., Editor, Daydreamer, July 1994.
[RFC1994] RFC 1994, PPP Challenge Handshake Authentication Protocol (CHAP), Simpson, W., Editor,
Daydreamer August 1994.
[RFC2153] RFC 2153, PPP Vendor Extensions, Simpson, W., Daydreamer, May 1997.
[RFC 2433] RFC 2433, Microsoft PPP CHAP Extensions, G. Zorn / S. Cobb, Microsoft Corporation, October 1998
[RFC 2759] RFC 2759, Microsoft PPP CHAP Extensions, Version 2, G. Zorn, Microsoft Corporation, January
2000
[SMBIOS] System Management BIOS Specification, Version 2.3.1, © 1997, 1999 American Megatrends Inc.,
Award Software International, Compaq Computer Corporation, Dell Computer Corporation, HewlettPackard Company, Intel Corporation, International Business Machines Corporation, Phoenix
Technologies Limited, and SystemSoft Corporation.
[SMBUS]
System Management Bus (SMBus) Specification, Version 2.0, ©2000, Duracell Inc., Fujitsu Personal
Systems Inc., Intel Corporation, Linear Technology Corporation, Maxim Integrated Products,
Mitsubishi Electric Corporation, Moltech Power Systems, PowerSmart Inc., Toshiba Battery Co., Ltd.,
Unitrode Corporation, USAR Systems.
[TAP]
Telocator Access Protocol version 1.8, February 04, 1997. ©1997, Personal Communications Industry
Association. http://www.pcia.com (As of this writing, the document is found under ‘Wireless
Resource Center | Protocols’, or: http://www.pcia.com/wireres/protocol.htm.) This document specifies
a protocol for sending an alphanumeric page by connecting to a paging service via a serial modem.
[TIA-602]
TIA/EIA Standard: Data Transmission Systems and Equipment - Serial Asynchronous Automatic
Dialing and Control, TIA/EIA 602, June 1992. © 1992, Telecommunications Industry Association.
Also available from the Electronic Industries Association. This document specifies the dialing protocol
commonly used in asynchronous serial modems.
[WFM]
Wired for Management Baseline Version 2.0 Release, ©1998, Intel Corporation. Attachment A,
UUIDs and GUIDs, provides information specifying the formatting of the IPMI Device GUID and
FRU GUID and the System Management BIOS (SM BIOS) UUID unique IDs.
3
Intelligent Platform Management Interface Specification
1.3 Conventions and Terminology
If not explicitly indicated, bits in figures are numbered with the most significant bit on the left and the least
significant bit on the right.
This document uses the following terms and abbreviations:
Table 1-1, Glossary
Term
Definition
Asserted
Active-high (positive true) signals are asserted when in the high electrical state (near power
potential). Active-low (negative true) signals are asserted when in the low electrical state
(near ground potential).
Baseboard Management Controller
The circuitry that connects one computer bus to another, allowing an agent on one to access
the other.
An 8-bit quantity.
In terms of this specification, this describes the PC-AT compatible region of battery-backed
128 bytes of memory, which normally resides on the baseboard.
A signal is deasserted when in the inactive state. Active-low signal names have “_L”
appended to the end of the signal mnemonic. Active-high signal names have no “_L” suffix.
To reduce confusion when referring to active-high and active-low signals, the terms one/zero,
high/low, and true/false are not used when describing signal states.
A non-maskable interrupt or signal for generating diagnostic traces and ‘core dumps’ from the
operating system. Typically NMI on IA-32 systems, and an INIT on Itanium™-based systems.
Double word is a 32-bit (4 byte) quantity.
Electrically Erasable Programmable Read Only Memory.
Notation for ‘Event Message’. See text for definitions of ‘Event Message’.
Front Panel Controller.
Fault Resilient Booting. A term used to describe system features and algorithms that improve
the likelihood of the detection of, and recovery from, processor failures in a multiprocessor
system.
Field Replaceable Unit. A module or component which will typically be replaced in its entirety
as part of a field service repair operation.
A reset event in the system that initializes all components and invalidates caches.
Inter-Integrated Circuit bus. A multi-master, 2-wire, serial bus used as the basis for the
Intelligent Platform Management Bus.
Intelligent Chassis Management Bus. A serial, differential bus designed for IPMI messaging
between host and peripheral chassis. Refer to [ICMB] for more information.
Internal Error. A signal from the Intel Architecture processors indicating an internal error
condition.
Intelligent Platform Management.
Intelligent Platform Management Bus. Name for the architecture, protocol, and
implementation of a special bus that interconnects the baseboard and chassis electronics and
provides a communications media for system platform management information. The bus is
built on I2C and provides a communications path between ‘management controllers’ such as
the BMC, FPC, HSC, PBC, and PSC.
Industry Standard Architecture. Name for the basic ‘PC-AT’ functionality of an Intel
Architecture computer system.
1024 bytes
Logical Unit Number. In the context of the Intelligent Platform Management Bus protocol, this
is a sub-address that allows messages to be routed to different ‘logical units’ that reside
behind the same I2C slave address.
BMC
Bridge
Byte
CMOS
Deasserted
Diagnostic
Interrupt
Dword
EEPROM
EvM
FPC
FRB
FRU
Hard Reset
I2 C
ICMB
IERR
IPM
IPMB
ISA
KB
LUN
4
Intelligent Platform Management Interface Specification
NMI
Non-maskable Interrupt. The highest priority interrupt in the system, after SMI. This interrupt
has traditionally been used to notify the operating system fatal system hardware error
conditions, such as parity errors and unrecoverable bus errors. It is also used as a Diagnostic
Interrupt for generating diagnostic traces and ‘core dumps’ from the operating system.
MD2
RSA Data Security, Inc. MD2 Message-Digest Algorithm. An algorithm for forming a 128-bit
digital signature for a set of input data.
RSA Data Security, Inc. MD5 Message-Digest Algorithm. An algorithm for forming a 128-bit
digital signature for a set of input data. Improved over earlier algorithms such as MD2.
For this specification, the term ‘payload’ refers to the information bearing fields of a message.
This is separate from those fields and elements that are used to transport the message from
one point to another, such as address fields, framing bits, checksums, etc. In some instances,
a given field may be both a payload field and a transport field.
Platform Event Filtering. The name of the collection of IPMI interfaces in the IPMI v1.5
specification that define how an IPMI Event can be compared against ‘filter table’ entries that,
when matched, trigger a selectable action such as a system reset, power off, alert, etc.
Parity Error. A signal on the PCI bus that indicates a parity error on the bus.
Platform Event Trap. A specific format of SNMP Trap used for system management alerting.
Used for IPMI Alerting as well as alerts using the ASF specification. The trap format is defined
in the PET specification. See [PET] and [ASF] for more information.
Power On Self Test.
Re-arm, in the context of this document, refers to resetting internal device state that tracks
that an event has occurred such that the device will re-check the event condition and regenerate the event if the event condition exists.
Sensor Data Record. A data record that provides platform management sensor type,
locations, event generation, and access information.
System Event Log. A non-volatile storage area and associated interfaces for storing system
platform event information for later retrieval.
System Error. A signal on the PCI bus that indicates a ‘fatal’ error on the bus.
System Management Interrupt.
Server Management Interface Chip. Name for one type of system interface to an IPMI
Baseboard Management Controller.
System Management Mode. A special mode of Intel IA-32 processors, entered via an SMI.
SMI is the highest priority non-maskable interrupt. The handler code for this interrupt is
typically located in a physical memory space that is only accessible while in SMM. This
memory region is typically loaded with SMI Handler code by the BIOS during POST.
System Management Software. Designed to run under the OS.
A reset event in the system which forces CPUs to execute from the boot address, but does
not change the state of any caches or peripheral devices.
A 16-bit quantity.
MD5
Payload
PEF
PERR
PET
POST
Re-arm
SDR
SEL
SERR
SMI
SMIC
SMM
SMS
Soft Reset
Word
1.4 Background - Architectural Goals
A number of goals/principles influence the design and implementation of a platform management subsystem that
works across multiple platforms. The abstracted, modular, extensible interfaces specified in this document seek to
satisfy those goals. The following review is provided to give a framework to assist in the evaluation options in the
implementation of this specification.
Provide Layered Management Value
– Provide management value at each level of integration, and have the net value increase as each level is
added. I.e. progressing from processor, through chip set, BIOS, baseboard, baseboard with
management circuitry, with onboard networking, with intelligent controllers, with managed chassis,
system management software, with Remote Management Cards, etc.
–
Maintain modularity so that one level does not carry undue cost burden for another. Levels should
retain value if separated. Avoid burdening baseboard with cost for chassis-specific management
5
Intelligent Platform Management Interface Specification
functions. Avoid building in functions that unduly impede OEMs in providing their own chassis
management features.
–
Drive intelligence to appropriate level. Don’t put complexity in at a level if the next higher level can
handle it. I.e. don’t do something with a microcontroller if the system’s host processor can do it more
flexibly and economically.
Plan for Evolution and Re-use
– Architect so existing implementations can be cleanly extended with new functionality, without
requiring existing functionality to be re-implemented or redesigned.
–
Architect for product families. Avoid ‘local optimizations’ that benefit one product at the cost of future
projects.
–
Knowledge and understanding of the architecture is also a valuable commodity. “Re-inventing the
wheel” often means retraining the wheel user. Design to preserve the knowledge base of developers,
testers, salespersons, and customers by maintaining consistency in architecture and implementation - in
hardware implementation, firmware, software, protocols, and interfaces.
–
Design for the economic incorporation of changes in the population and implementation of baseboard
and remote sensors. Architect to minimize the impact to hardware and software when sensor
population or sensor hardware interfaces change.
–
Design to maximize ‘self configurability’ in system management software. I.e. ‘Plug ‘N Play’. Provide
platform resident discovery mechanisms, such as standardized tables, discovery mechanisms, etc. to
reduce or eliminate the need to ‘customize’ system software for different platforms.
Provide Scalability
– Architecture should scale from entry through enterprise and data center class server systems.
Architecture should be adaptable from single board and single chassis, through multi-board and multichassis systems.
–
Apply ‘Layered Management Value’ concept. Low-end solutions should be a proper functional subset
of higher end solutions. Entry solutions should not carry undue burden for higher class systems.
Support OEM Extensibility:
– Provide clean points for OEM extension and integration.
–
Provide OEM support in protocol and command specifications. Reserve command numbers, sensor
numbers, etc. for OEM extension.
1.5 New for IPMI v1.5
IPMI v1.5 is an extension of the v1.0 specification. IPMI v1.5 also includes learnings, feedback, and features
gathered from industry review and experiences deploying IPMI v1.0 enabled systems.
The following goals guided the creation of the IPMI v1.5 specification:
6
•
Help enable “Always Available Manageability” by enabling new access media: LAN, Serial/Modem,
and PCI Management Bus.
•
Extend “Autonomous Manageability” by defining new automatic alerting and recovery mechanisms.
•
Synch-up with and support emergent and existing standards such as PPP, the DMTF Pre-OS Working
Group ‘Alert Standard Forum’ specification, SMBus 2.0, Compact PCI / AdvancedTCA, and the PCI
Management Bus.
•
Retain as much backward compatibility with IPMI v1.0 as feasible
Intelligent Platform Management Interface Specification
The following presents a brief summary of some of the more significant additions and enhancements in the IPMI
v1.5 specification:
Serial/Modem Messaging and Alerting
The IPMI v1.5 specification defines how IPMI messaging can be accomplished via a direct serial or
external modem connection to the BMC. It also includes the specifications for generating alerts,
numeric pages, and alphanumeric pages on events.
Serial Port Sharing
This is a capability that works in conjunction with serial/modem messaging and alerting and allows the
baseboard serial connector to be shared between use by the BMC and use by the system. This enables
coordination between system features such as console redirection and allows the serial connection to
be used for run-time applications while still allowing it to be remotely accessed for ‘emergency’
management.
Boot Options
IPMI v1.5 includes a boot options ‘mailbox’ that can be used to direct the operation of BIOS and the
booting process following a system power up or reset operation.
LAN Messaging and Alerting
The specification defines how IPMI messaging can be accomplished via a LAN connection to the
BMC. It also includes the specifications for generating PET format SNMP traps on events.
Extended BMC Messaging ‘Channel Model’
IPMI v1.0 introduced a limited capability to use a ‘Channel Number’ capability that was primarily
used for routing messages to the IPMB. IPMI v1.5 expands on the use of channel numbers as a general
mechanism for routing messages between different media and organizing the access
Additional Sensor and Event Types
Several new sensor types have been added to IPMI v1.5, including a Slot/Connector sensor for
monitoring hot-plug slot status, an ACPI System Power State sensor to support out-of-band
monitoring of the system power state, and an enhanced Watchdog sensor that supports events
generated by the standardized watchdog timer function.
Platform Event Filtering (PEF)
Platform Event Filtering (PEF) provides a mechanism for configuring the BMC to taking selected
actions on event messages that it receives or has internally generated. These actions include operations
such as system power-off, system reset, as well as triggering the generation of an alert.
Alert Policies
Alert policies provide a configurable mechanism for configuring and controlling the delivery of an
alert to multiple destinations. This enables the creation of ‘call down lists’ where one alert destination
is tried first and then another.
7
Intelligent Platform Management Interface Specification
1.6 IPMI Overview
This section presents an overview of IPMI and its main elements and characteristics.
1.6.1
Intelligent Platform Management
The term Intelligent Platform Management refers to autonomous monitoring and recovery features
implemented directly in platform management hardware and firmware. The key characteristic of Intelligent
Platform Management is that inventory, monitoring, logging, and recovery control functions are available
independent of the main processors, BIOS, and operating system. Platform management functions can also be
made available when the system is in a powered down state.
Intelligent Platform Management capabilities are a key component in providing enterprise-class management
for high-availability systems. Platform status information can be obtained and recovery actions initiated under
situations where system management software and normal ‘in-band’ management mechanisms are unavailable.
The independent monitoring, logging, and access functions available through IPMI provide a level of
manageability built-in to the platform hardware. This can support systems where there is no system
management software available for the particular operating system, or the end-user elects not to load or enable
the system management software.
1.6.2
IPMI Relationship to other Management Standards
IPMI is best used in conjunction with system management software running under the operating system. This
provides an enhanced level of manageability by providing in-band access to the IPMI management information
and integrating IPMI with the additional management functions provided by management applications and the
OS. System management software and the OS can provide a more sophisticated control, error handling and
alerting, than can be directly provided by the platform management subsystem.
IPMI is a hardware level interface specification that is ‘management software neutral’ providing monitoring
and control functions that can be exposed through standard management software interfaces such as DMI,
WMI, CIM, SNMP, etc. As a hardware level interface, it sits at the bottom of a typical management software
stack, as illustrated in Figure 1-1Figure 1-1, below.
8
Management S/W
Standards
Management
Applications
‘In-band’
Remote
Access
Service Provider
SP Interface
Instrumentation Code
STANDARD Remote I/F
(e.g. RPC, SNMP)
STANDARD S/W I/F
(e.g. DMI-MI, CIM)
STANDARD S/W I/F
(e.g. DMI-CI, WMI)
IPMI I/F Code
IPMI
HARDWARE
SOFTWARE
Figure 1-1, IPMI and the Management Software Stack
IPMI I/F
IPMI H/W I/F
Baseboard Mgmt.
Controller
Intelligent Platform Management Interface Specification
1.6.3
Management Controllers and the IPMB
Figure 1-2, IPMI Block DiagramFigure 1-2, IPMI Block Diagram, shows the main elements of an IPMI
implementation. At the heart of the IPMI architecture is a microcontroller called the Baseboard Management
Controller, or BMC. The BMC provides the intelligence behind Intelligent Platform Management. The BMC
manages the interface between system management software and the platform management hardware, provides
autonomous monitoring, event logging, and recovery control, and serves as the gateway between system
management software and the IPMB and ICMB.
IPMI supports the extension of platform management by connecting additional management controllers to the
system using the IPMB. The IPMB is an I2C-based serial bus that is routed between major system modules. It is
used for communication to and between management controllers. This provides a standardized way of
integrating chassis features with the baseboard. Because the additional management controllers are typically
distributed on other boards within the system, away from the ‘central’ BMC, they are sometimes referred to as
satellite controllers.
Figure 1-2, IPMI Block Diagram
ICMB (Intelligent Chassis Management Bus)
Remote
Management Card
RS-485
Transceivers
LAN
LAN
Connector
Network
(LAN)
Controller
Serial
Controller
Modem
ICMB
BRIDGE
(optional)
Aux. IPMB
Connector
Serial
Connector
side-band
interface
to NIC,
e.g.
SMBus
PCI Management
Bus
Aux. IPMB
Connector
IPMB
Non-volatile Storage
BASEBOARD
MANAGEMENT
CONTROLLER
(BMC)
y SYSTEM EVENT LOG (SEL)
y SENSOR DATA RECORD (SDR)
REPOSITORY
y BASEBOARD FIELD-REPLACEABLE
UNIT (FRU) INFO
CHASSIS
MANAGEMENT
(SATELLITE
CONTROLLER)
Sensors & Control Circuitry
Serial
Port
Sharing
Motherboard
Serial
Controller
e.g. Voltages, Temperatures,
Fans, Power & Reset control, etc.
Serial
Controller
Private Management Busses
System Bus
Chassis
Sensors
e.g. Fans,
Temperatures, Power
Supplies
System Interface
IPMI
MESSAGES
FRU
SEEPROM
FRU
SEEPROM
FRU
SEEPROM
Temperature
Sensor
FRU
SEEPROM
MOTHERBOARD
MEMORY BOARD
CHASSIS BOARD
PROCESSOR
BOARD
REDUNDANT POWER
BOARD
By standardizing the interconnect, a baseboard can be readily integrated into a variety of chassis that have
different management features. IPMI’s support for multiple management controllers also means that the
architecture is scalable. A complex, multi-board set in an enterprise-class server can use multiple management
controllers for monitoring the different subsystems such as redundant power supplies, hot-swap RAID drive
slots, expansion I/O, etc., while an entry-level system can have all management functions integrated into the
BMC.
IPMI also includes ‘low-level’ I2C access commands that can be used to access ‘non-intelligent’ I2C devices
(devices that don’t handle IPMI commands) on the IPMB or Private Busses accessed via a management
9
Intelligent Platform Management Interface Specification
controller. The IPMB can also support SMBus slave devices, with the restriction that the SMB Alert signal is
not supported on IPMB, and a controller that implements the IPMB protocol cannot serve as the target for an
SMBus Modified Write Word protocol transfer from an SMBus slave. Refer to the IPMB and ICMB
Specifications (see Reference Documents) for additional information on the IPMB and ICMB.
1.6.4
IPMI Messaging
IPMI uses message-based interfaces for the different interfaces to the platform management subsystem such as
IPMB, serial/modem, LAN, ICMB, PCI Management Bus, and the system software-side “System Interface” to
the BMC.
All IPMI messages share the same fields in the message ‘payload’ - regardless of the interface (transport) that
they’re transferred over. This is represented with the double-ended arrows in Figure 1-2, IPMI Block
DiagramFigure 1-2, IPMI Block Diagram, The same core of IPMI messages is available over every IPMIspecified interface, they’re just ‘wrapped’ differently according to the needs of the particular transport. This
enables a piece of management software that works on one interface to be converted to use a different interface
mainly by changing the underlying ‘driver’ for the particular transport. This also enables knowledge re-use: A
developer that understands the operation of IPMI commands over one interface can readily apply that
knowledge to a different IPMI interface.
IPMI messaging uses a request/response protocol. IPMI request messages are commonly referred to as
commands. The use of a request/response protocol facilitates the transfer of IPMI messages over different
transports. It also facilitates multi-master operation on busses like the IPMB and ICMB, allowing messages to
be interleaved and multiple management controllers to directly intercommunicate on the bus.
IPMI commands are grouped into functional command sets, using a field called the Network Function Code.
There are command sets for sensor and event related commands, chassis commands, etc. This functional
grouping makes it easier to organize and manage the assignment and allocation of command values.
All IPMI request messages have a Network Function, command, and optional data fields. All IPMI response
messages carry Network Function, command, optional data, and a completion code field. As inferred earlier,
the differences between the different interfaces has to do with the framing and protocols used to transfer this
payload. For example, the IPMB protocol adds fields for I2C and controller addressing, and data integrity
checking and handling, whereas the LAN interface adds formatting for sending IPMI messages as LAN
packets.
1.6.5
Sensor Model
Access to monitored information, such as temperatures and voltages, fan status, etc., is provided via the IPMI
Sensor Model. Instead of providing direct access to the monitoring hardware IPMI provides access by
abstracted sensor commands, such as the Get Sensor Reading command, implemented via a management
controller. This approach isolates software from changes in the platform management hardware
implementation.
Sensors are classified according to the type of readings they provide and/or the type of events they generate. A
sensor can return either an analog or discrete reading. Sensor events can be discrete or threshold-based.
The different event types, sensor types, and monitored entities are represented using numeric codes defined in
the IPMI specification. IPMI avoids reliance on strings for management information. Using numeric codes
facilitates internationalization, automated handling by higher level software, and reduces management
controller code and data space requirements.
10
Intelligent Platform Management Interface Specification
1.6.6
System Event Log and Event Messages
The same approach is applied to the generation and control of platform events. The BMC provides a
centralized, non-volatile System Event Log, or SEL. Having the SEL and logging functions managed by the
BMC helps ensure that ‘post-mortem’ logging information is available should a failure occur that disables the
systems processor(s).
A set of IPMI commands allows the SEL to be read and cleared, and for events to be added to the SEL. The
common request message (command) used for adding events to the SEL is referred to as an Event Message.
Event Messages can be sent to the BMC via the IPMB. This provides the mechanism for satellite controllers to
detect events and get them logged into the SEL. The controller that generates an event message to another
controller via IPMB is referred to as an IPMB Event Generator. The controller that receives event messages is
called the IPMB Event Receiver.
A generic Event Receiver is a controller that accepts a Platform Event Message command over whatever media
is connected to it, plus internally generated Event Messages. The BMC is typically the only generic Event
Receiver in the system.
Management Controllers that generate Event Messages must know the sensor and event type so it can place that
information in the Event Message. This ensures that Event Messages carry important information that can be
interpreted without requiring a-priori knowledge of the sensor, or access to the Sensor Data Record for the
sensor.
This is often done by ‘hardcoding’ this relationship into the controller’s firmware. However, this approach
binds the Sensor Type and Event Type assignment to the generation of event messages. IPMI also includes
commands that allow the sensor and event type information to be read from the Sensor Data Record and written
into the controller during initialization. This makes it possible to create generic management controllers that do
not have to have hard-coded sensor types. For example, a vendor could create a device that provides a number
of analog, threshold-based sensors that get assigned as voltage, temperature, or other sensor types according to
the type information the system integrator placed in the SDRs for the sensors. An analog input could be
assigned as a “+5V” sensor on one system, and a “-12V” sensor on another just by changing the SDRs.
1.6.7
Sensor Data Records & Capabilities Commands
IPMI’s extensibility and scalability mean that each platform implementation can have a different population of
management controllers and sensors, and different event generation capabilities. The design of IPMI allows
system management software to retrieve information from the platform and automatically configure itself to
the platform’s capabilities. This greatly reduces or eliminates the need for platform-specific configuration of the
platform management instrumentation software - enabling the possibility of “Plug and Play” platformindependent instrumentation software.
Information that describes the platform management capabilities is provided via two mechanisms: capabilities
commands and Sensor Data Records (SDRs). Capabilities commands are commands within the IPMI command
sets that return fields that provide information on other commands and functions the controller can handle.
Sensor Data Records are data records that contain information about the type and number of sensors in the
platform, sensor threshold support, event generation capabilities, and information on what types of readings the
sensor provides.
The primary purpose of Sensor Data Records is to describe the sensor configuration of the platform
management subsystem to system software. Sensor Data Records describe sensors; they do not instantiate
sensors. For example, adding a new Sensor Data Record does not cause management controller firmware to
automatically ‘grow’ or instantiate a new sensor. But they are used to describe sensors that already exist, and
can also be used to tell software to only pay attention to a subset of the available sensors.
11
Intelligent Platform Management Interface Specification
Sensor data records have a limited capability to configure pre-existing sensors. There is information that an
Initialization Agent in the BMC to enable or disable sensors and initialize thresholds. This is described more in
the following section.
Sensor Data Records also include records describing the number and type of devices that are connected to the
system’s IPMB, records that describe the location and type of FRU Devices (devices that contain field
replaceable unit information).
1.6.8
Initialization Agent
SDRs can also hold default threshold values and event generation settings for sensors and management
controllers. During system resets, the BMC performs an initialization agent function and writes these settings to
those sensors that have ‘initialization required’ field set in their SDR. This eliminates the need for satellite
controllers to retain their own non-volatile storage and command interfaces for default settings, and also
provides a mechanism to retrigger any events that may have been transmitted before the BMC was ready to
accept them. The initialization agent can also be used to assign the Sensor Type to a generic sensor. See Section
27.6, Sensor Initialization AgentSensor Initialization Agent, for details on the initialization agent process.
1.6.9
Sensor Data Record Repository
Sensor Data Records are kept in a single, centralized non-volatile storage area that is managed by the BMC.
This storage is called the Sensor Data Record Repository (SDR Repository). Implementing the SDR Repository
via the BMC provides a mechanism that allows SDRs to be retrieved via ‘out-of-band’ interfaces, such as the
ICMB, a Remote Management Card, or other device connected to the IPMB. Like most Intelligent Platform
Management features, this allows SDR information to be obtained independent of the main processors, BIOS,
system management software, and the OS.
1.6.10 Private Management Busses
A Private Management Bus (also referred to as Private Bus) is an I2C bus that is accessed via a management
controller by using IPMI commands for low-level I2C access. Multiple private busses can be implemented
behind a single management controller. IPMI supports using private busses as a mechanism for accessing
24C02-compatible SEEPROMs (Serial Electrically Erasable Programmable ROMs) that hold FRU information.
Private busses may also be used to provide low-level access interface for other I2C or SMBus devices, though
the IPMI specification does not cover the way such devices would be used. Each management controller can
provide up to eight private busses.
1.6.11 FRU Information
The IPMI specifications include support for storing and accessing multiple sets of non-volatile Field
Replaceable Unit (FRU) information for different modules in the system. An enterprise-class system will
typically have FRU information for each major system board (e.g. processor board, memory board, I/O board,
etc.). The FRU data includes information such as serial number, part number, model, and asset tag.
IPMI FRU information can be made accessible via the IPMB and management controllers. The information can
be retrieved at any time, independent of the main processor, BIOS, system software, or OS. This allows FRU
information to be retrieved via ‘out-of-band’ interfaces, such as the ICMB, a Remote Management Card, or
other device connected to the IPMB. FRU information can even be available when the system is powered
down.
These capabilities allow FRU information to be obtained under failure conditions when FRU access
mechanisms that rely on the main processor become unavailable. This facilitates the creation of automated
remote inventory and service applications.
12
Intelligent Platform Management Interface Specification
IPMI does not seek to replace other FRU or inventory data mechanisms, such as those provided by SM BIOS,
and PCI Vital Product Data. Rather, IPMI FRU information is typically used to complement that information or
to provide information access out-of-band or under ‘system down’ conditions.
1.6.12 FRU Devices
IPMI provides FRU information in two ways: via a management controller, or via FRU SEEPROMs. FRU
information that is managed by a management controller is accessed using IPMI commands. This isolates
software from direct access to the non-volatile storage device, allowing the hardware implementer to utilize
whatever type of non-volatile storage they want.
In order to more economically support providing FRU information on multiple platform modules, IPMI also
allows simple 24C02-compatible SEEPROM (Serial Electrically Erasable Programmable ROM) chips to be
used for storing FRU information. (‘24C02’-type devices are non-volatile storage devices that have a built-in
I2C-compatible interface).
FRU SEEPROMs provide a mechanism for implementing FRU information without requiring a management
controller on the field replaceable unit. FRU SEEPROMs can be accessed via a Private Management Bus
connected to a management controller, or if necessary, can be placed directly on the IPMB or PCI Management
Bus. While supported, it is generally recommended that devices with I2C/SMBus interfaces that lack data
integrity checks (e.g. checksums), such 24C02-type SEEPROMs, are not placed on ‘public’ busses such as
IPMB and PCI-SMBus. This is because without data integrity checks it is possible that a misbehaved thirdparty add-in device could cause a bus ‘glitch’ that would result in an undetected error when reading or writing
the SEEPROM. (Note: depending on the type of device, I2C addressing places a limit on the number of devices
that can be placed directly on the IPMB. Refer to the IPMB I2C Address Allocation specification for more
information.)
1.6.13 Entity Association Records
Entity Association Records are a special type of SDR that provides a definition of the relationships among
platform entities. For example, an Entity Association can be set up that groups a set of individual power
supplies into a redundant power unit. A ‘redundancy lost’ event on the power unit can then be correlated with
the individual power supply failure. Without the Entity Association information, the ‘redundancy lost’ and
‘power supply failed’ events would be disjoint events that could only be correlated based on a-priori knowledge
of the system.
1.6.14 Linkage between Events and FRU Information
Included in the SDRs is information that indicates which system entity a sensor is monitoring (e.g. a memory
board) and also provide a link to the FRU information for the entity. SDRs use a set of codes that specify which
controller holds the sensor, the sensor type (e.g. temperature), the particular instance of the sensor (e.g. sensor
#2), the sensor’s event and reading type (e.g. discrete or threshold-based), the set of events it can generate, and
associated bit fields that indicate which specific events a sensor can produce.
The same codes and bit fields directly map to the information that is passed in event messages and logged in the
SEL. Thus, a SEL entry can indicate the controller, sensor, sensor type, and event type associated with the
event. This information provides a useful level of information by itself - but when combined with SDR
information, the event can be correlated to the entity and FRU associated with the event. Correlating an event to
the FRU can help guide a service person to the problem area, or even be used to identify the replacement parts
they should bring to a site.
13
Intelligent Platform Management Interface Specification
1.6.15 Differentiation and Feature Extensibility
Platform management features continue to evolve. While IPMI seeks to provide a standardized interface to
cover the majority of platform management needs, explicit provisions have been made throughout IPMI to
support OEM differentiation and new features. Special ranges of code values and commands have been
reserved to allow OEM sensors, events, and value-added functions to be implemented within the IPMI
framework.
1.6.16 System Interfaces
IPMI defines three standardized system interfaces that system software uses for transferring IPMI messages to
the BMC. In order to support a variety of microcontrollers, IPMI offers a choice of system interfaces. Using
these interfaces is key to enabling cross-platform software. The system interfaces are similar enough so that a
single driver can be created that supports all IPMI system interfaces.
The system interface connects to a system bus that can be driven by the main processor(s). The present IPMI
system interfaces are I/O mapped. Any system bus that allows the main processor(s) to access the specified I/O
locations, and meet the timing specifications, can be used. Thus, an IPMI system interface could be hooked to
the ISA-bus, X-bus, or a proprietary bus off the baseboard chip set.
The three IPMI system interfaces are:
14
Keyboard
Controller
Style (KCS)
The bit definitions, and operation of the registers follows that used in the Intel
8742 Universal Peripheral Interface microcontroller. The term ‘Keyboard
Controller Style’ reflects the fact that the 8742 interface was used as the legacy
keyboard controller interface in PC architecture computer systems. This
interface is available built-in to several commercially available
microcontrollers. Data is transferred across the KCS interface using a per-byte
handshake.
System
Managemen
t Interface
Chip
(SMIC)
The SMIC interface provides an alternative when the implementer wishes to
use a microcontroller for the BMC that does not have the built-in hardware for
a KCS interface. This interface is a three I/O port interface that can be
implemented using a simple ASIC, FPGA, or discrete logic devices. It may
also be built-in to a custom-designed management controller. Like the KCS
interface, a per-byte handshake is also used for transferring data across the
SMIC interface.
Block
Transfer
(BT)
This interface provides a higher performance system interface option. Unlike
the KCS and SMIC interfaces, a per-block handshake is used for transferring
data across the interface. The BT interface also provides an alternative to using
a controller with a built-in KCS interface. The BT interface has three I/Omapped ports. A typical implementation includes hardware buffers for holding
upstream and downstream message blocks. The BT interface can be
implemented using an ASIC or FPGA, or may be built-in to a custom-designed
management controller.
SMBus
System
Interface
(SSIF)
The SMBus System Interface is defined as part of the IPMI v2.0 or later
specifications. However, this interface can be used with IPMI v1.5
implementations. IPMI v1.5 implementations that use SSIF will be considered
to be conformant to the IPMI v1.5 specification if the SSIF implementation
meets the IPMI v2.0 or later specifications for the interface. Use of the SSIF
with an IPMI v1.5 implementation requires using the IPMI v2.0 or later
specification and complying with any licensing requirements for implementing
those specifications.
Intelligent Platform Management Interface Specification
1.6.17 Other Messaging Interfaces
In addition to the System Interface and IPMB, IPMI messaging can be carried over other interfaces, such as
LAN, serial/modem, ICMB, and PCI management bus. IPMI includes a communication infrastructure that
supports transferring messages between these interfaces as well as to the BMC.
1.6.18 LAN Interface
The LAN interface specifications define how IPMI messages can be sent to and from the BMC encapsulated in
RMCP (Remote Management Control Protocol) packets datagrams. This capability is also referred to as “IPMI
over LAN”. IPMI also defines the associated LAN-specific configuration interfaces for setting things such as IP
addresses other options, as well as commands for discovering IPMI-based systems.
The Distributed Management Task Force (DMTF) specifies the RMCP format. This same packet format is used
for non-IPMI messaging via the DMTF’s Alert Standard Forum “ASF” specification. Using the RMCP packet
format enables more commonality between management applications that operate in an environment that
includes both IPMI-based and ASF-based systems. More information on IPMI and ASF is provided below.
1.6.19 Serial/Modem Interface
The Serial/Modem Interface specifications define how IPMI messages can be sent to and form the BMC via a
direct serial or external modem connection. The specification supports three connection modes that define the
protocol for delivering IPMI messages via serial/modem:
•
Basic Mode: The IPMI messages are encapsulated with minimal additional framing and escaping for
transport over a serial/modem connection. Basic Mode provides the highest performance but requires an
‘IPMI-aware’ serial application.
•
PPP Mode: The IPMI messages are encapsulated in the same RMCP format as used for LAN messages,
but are delivered via a PPP connection. PPP mode allows remote applications to take advantage of built-in
PPP support in the OS for things such as dialing and authentication, and provides the highest commonality
with LAN-based software, but at the cost of lower throughput.
•
Terminal Mode: Terminal Mode defines how IPMI messages can be transferred using printable
characters. It also includes a limited number of English ASCII text commands for doing such things as
getting a high level system status and causing a system reset or power state change. Terminal mode is
lower performance than Basic Mode and more limited in capabilities than both Basic Mode and PPP Mode,
but offers a mechanism for those who are transitioning to IPMI and more sophisticated interfaces from a
legacy, character-based environment
1.6.20 IPMI and ASF
IPMI and ASF are complementary specifications that can provide platform management in a ‘pre-boot’ or ‘OS
absent’ environment. IPMI uses a management microcontroller as the main element of the management system,
whereas ASF presently focuses on having an ‘alert sending device’ - typically the network controller - polling
devices on the motherboard and autonomously generating alerts. As of this writing, ASF’s scope primarily
covers the way an alert sending device polls sensor devices and sends alerts, and the specification of ‘LAN’
commands for discovering RMCP-based systems and performing emergency reset and power off actions.
This includes the supporting specification of SMBus interfaces to ‘ASF Sensor Devices’ that can be polled by
the alert sending device, the specification of the RMCP packet format, and the specification of SMBus-based
commands that can be used to send a ‘push’ alert via an alert sending device.
While somewhat of an oversimplification, ASF may be considered to be scoped for ‘desktop/mobile’ class
systems, and IPMI for ‘servers’ where the additional IPMI capabilities such as event logging, multiple users,
15
Intelligent Platform Management Interface Specification
remote authentication, multiple transports, management extension busses, sensor access, etc., are valued.
However there are no restrictions in either specification as to the class of system that the specification can be
used. I.e. you can use IPMI for desktop and mobile systems and ASF for servers if the level of manageability
fits your requirements.
IPMI and ASF share a number of formats, data structures, and enumerations. It is expected that this will
continue to grow.
•
Shared management packet format: IPMI uses ASF ‘RMCP’ packet format for delivering IPMI messages
over LAN and PPP and ASF messages for LAN discovery. The RMCP format includes a message class
explicitly for IPMI use.
•
Common LAN Alert Format: Both generate LAN Alerts using the IPMI PET (Platform Event Trap)
Specification for SNMP Traps
•
Common Flags for boot control: IPMI uses a superset of the boot flags defined in ASF.
•
Common enumerations for sensor types and event types: ASF uses the IPMI enumerations for sensor and
event types. These values are used in Alerts and ASF Sensor Device Status.
•
Common BIOS progress codes: IPMI uses ASF BIOS Error and Progress codes.
•
Hardware: IPMI management controllers and ASF alert sending devices can both use ASF Sensor Devices.
In an IPMI application these can be place on private management busses and polled by the BMC, they can
also be used on the PCI management bus. In an ASF application, the devices would typically always be on
the PCI management bus or main SMBus and polled by the Network Controller(s).
1.6.21 LAN Alerting
IPMI supports LAN Alerting in the form of SNMP Traps that follow the Platform Event Trap (PET) format.
(Refer to [PET] for more information.) SNMP Traps are typically sent as unreliable datagrams. However, IPMI
includes a PET Acknowledge and retry options that allows an IPMI-aware remote application to provide a
positive acknowledge that the trap was received.
1.6.22 Serial/Modem Alerting and Paging
The IPMI specification supports several options for alerting over a serial/modem connection:
•
Dial Page: Sending a numeric page by using an external modem to generate ‘touch tones’.
•
TAP Page: The BMC connects to a TAP 1.8 paging service and delivers an alphanumeric page.
•
PPP Alert: The BMC connects to a remote LAN via PPP and delivers a PET trap to a specified IP address.
1.6.23 Platform Event Filtering (PEF)
Platform Event Filtering (PEF) provides a mechanism for configuring the BMC to take selected actions on
event messages that it receives or has internally generated. These actions include operations such as system
power-off, system reset, as well as triggering the generation of an alert.
The BMC maintains an event filter table that is used to select which events trigger an action and which actions
to perform. Each time the BMC receives an event message (either externally or internally generated) it
compares the event data against the entries in the event filter table. The BMC scans all entries in the table and
collects a set of actions to be performed as determined by the entries that were matched.
16
Intelligent Platform Management Interface Specification
1.6.24 Call Down Lists and Alert Policies
The IPMI specification allows an implementation to support configurable alert policies that determine how an
alert will be processed. These can be used to create a ‘call down list’ of different destinations that an alert gets
sent to. Alert policies can have destinations of different types and on different channels. For example, a policy
could be defined to first try to send an alert to LAN address ‘A’, and if that fails send it to LAN address ‘B’,
and then send a Dial Page via the modem, and if that fails, a TAP page.
IPMI allows the alert destinations to be configured in any order. I.e. you can pick whether an alert goes out via
serial/modem first, or via LAN first. The main limitation comes from the number of policy entries that a given
implementation supports.
1.6.25 Channel Model, Authentication, Sessions, and Users
IPMI v1.5 incorporates a common communication infrastructure referred to as the ‘Channel Model’. This is an
extension of the channels that were used as part of messaging in IPMI v1.0.
Channels provide the mechanism for directing the routing of IPMI messages between different media
connections to the BMC. A channel number identifies a particular connection. For example, 0 is the channel
number for the primary IPMB. Up to nine total channels can be supported (the System Interface and primary
IPMB, plus seven additional channels with a media type assigned by the implementer.) Channels can thus be
used to support multiple IPMB, LAN, Serial, etc., connections to BMC.
Channels can be session-based or session-less. A session is used for two purposes: As a framework for user
authentication, and to support multiple IPMI Messaging streams on a single channel. Session-based channels
thus have at least one user ‘login’ and support user and message authentication. Session-less channels do not
have users or authentication. LAN and serial/modem channels are examples of session-based, while the System
Interface and IPMB are examples of session-less channels.
In order to do IPMI messaging via a session, a session must first be activated. The act of activating a session is
one of authenticating a particular user. This is accomplished using a ‘challenge/response’ mechanism, where a
challenge is requested using a Get Session Challenge command, and the signed challenge returned in an
Activate Session command.
The specification supports different algorithms for the signature - these are referred to as Authentication Types.
Authentication Types include ‘none’, ‘straight password’, the MD2 and MD5 message-digest algorithms, etc.
For consistency, session-based channels always use the Get Session Challenge and Activate Session commands
even if Authentication Type is ‘none’. (In this case, dummy values are used for the signatures.)
A session has a Session ID that is used for tracking the state of a session. The Session ID mechanism allows
multiple sessions to be able to be simultaneously supported on a channel.
The message signature, Session ID, and other session related information is separate from the actual IPMI
message content. Thus, a packet carrying an authenticated IPMI message can be thought of as being comprised
of a ‘Session Packet’ that includes the session-specific fields and carries an IPMI message as its payload.
The concept of user is essentially a way to identify a collection of privilege and authentication information.
User configuration is done on a per channel basis. This means that a given user could have a different password
and set of privileges for accessing the BMC via a LAN channel than via a serial channel.
Privilege Levels determine which IPMI commands a given user can execute over a given channel.
Privilege Limits set the maximum privilege level that a user can operate at. A user is configured with a given
maximum privilege limit for each channel. In addition there is a Channel Privilege Limit that sets the maximum
limit for all users on a given channel. The Channel Privilege Limit takes precedence over the privilege
configured for the user. Thus, a user can operate at a privilege level that is no higher than the lower of the User
Privilege Limit and the Channel Privilege Limit.
17
Intelligent Platform Management Interface Specification
1.6.26 Standardized Watchdog Timer
Watchdog Timer capabilities have been commonly deployed in Enterprise-class servers. IPMI provides a
standardized interface for a system Watchdog Timer. This timer can be used for BIOS, OS, and OEM
applications. The timer can be configured to automatically generate selected actions when it expires, including
power off, power cycle, reset, and interrupt. The timer function can also automatically log the expiration event.
Setting ‘0’ for the timeout interval allows the timeout action to be initiated immediately. This provides a means
for devices on the IPMB, such as Remote Management Cards, to use the Watchdog Timer to initiate emergency
reset and other recovery actions dependent on the capability of the timer.
1.6.27 Standardized POH Counter
This is an optional counter to return a counter value proportional to the system operating (S0) power-on hours.
1.6.28 IPMI Hardware Components
IPMI provides very few restrictions specifications for on the actual hardware components used to implement
the platform management hardware. IPMI seeks to ‘standardize the interface, not the implementation’. IPMI
was designed so that it can be implemented with ‘off-the-shelf’ components. Thus, IPMI does not require
specific microcontrollers to be used for management controllers, nor special ASICs or proprietary logic
devices. As long as the interface, timing and (in the case of IPMB and ICMB) electrical specifications are met,
the choice of components is up to the implementer. It is mandatory to implement a system interface that is
compatible with one of the three specified system interfaces.
1.6.29 Configuration Interfaces
IPMI provides standardized interfaces and commands for configuring the platform management subsystem.
This enables cross-platform software to Sensor Data Records are an example of the interface for configuring
sensor population and behavior on a system. There are also commands for configuring capabilities such as LAN
and serial/modem remote protocols, user passwords and privilege levels, Platform Event Filtering, alert
destinations, and others.
Unless otherwise specified, changes to parameters are required to take effect for the next use. For example,
parameters that affect user access or session operation must take effect for the next time a remote console
attempts to connect to the system. In some implementations, changes to configuration parameters may take
effect immediately. Thus, a remote application should be careful when setting parameters that could cause the
application to become disconnected from the BMC.
For the purpose of conformance checking, up to 5 seconds will be allowed between the time a parameter is
changed to when it must have taken effect.
It is recognized that there are race conditions where a session may already be in the process of being established
before the change can be propagated. It is recommended that a BMC implementation takes steps to ensure that
parameters are used consistently. This specification does not define a specific mechanism, but here are some
possible approaches. An implementation could terminate a session in progress if the user’s parameters change
while the session is being established. Alternatively, an implementation could ‘snap shot’ the user’s
configuration at the time the session is being established and only allow a session to be established if the given
user’s configuration has been unmodified in the last 5 seconds.
1.7 IPMI and BIOS
The level of interaction between BIOS and IPMI is greatly dependent on the implementation and number of
optional capabilities that are to be supported. It is possible to have an IPMI implementation that does not require
18
Intelligent Platform Management Interface Specification
any BIOS support, other than that required to meet any applicable ACPI or Plug ‘N Play requirements for
reporting the I/O and/or interrupt resources used by the IPMI system interface.
In some implementations, BIOS may be responsible for the initialization or startup of certain functions in the
management controllers, such as setting the initial timestamp time in the SEL and/or SDR devices. BIOS may
also perform tests of the platform management hardware and management controllers during POST.
It is recommended that BIOS include provisions for checking and reporting on the basic health of BMC by
executing the Get Self Test Results command and checking the result.
It’s expected that most implementations will provide BIOS features that take advantage of IPMI. For example, it
is expected that many implementations will use IPMI to log POST errors, or to log ‘system boot’ events so that
events can be tracked relative to the last boot time. Another expectation is that many systems utilize the IPMI
Watchdog Timer function with BIOS.
With IPMI v1.5, the BIOS can share in additional capabilities. For example, IPMI v1.5 defines a new LAN-based
interface. The BIOS can help keep the BMC updated with the LAN IP Address assignment. IPMI v1.5 also
includes a serial/modem interface with support for a capability called ‘serial port sharing’ in which the serial
controller can be shared between the BMC and BIOS-based serial console redirection. There is also a set of ‘boot
flags’ that BIOS can read to direct its operation following a system management initiated reset, power cycle, or
power up.
1.8 System Management Software (SMS)
The Management Controllers, Sensors, SEL information, SDR information, etc., are of limited value without
System Management Software to interpret, handle, and present the information. Platform management is only a
subset of systems management. System Management Software takes platform management information and links
it into other aspects of systems management, such as software management and distribution, alerting, remote
console access, etc.
With respect to the platform management architecture and this specification, System Management Software:
19
Intelligent Platform Management Interface Specification
•
Polls the System Event Log for new Event information, acting on it as appropriate. This may include
taking actions such as sending alerts on the network, presenting a local ‘pop-up’ message, shutting
down an application, consolidating the information with other ‘system log’ data, etc. System Critical
Events are primarily communicated to system management software using the System Event Log as a
'mailbox' between the originator of the Event Message and the system management software.
•
Manages the System Event Log. The SEL may contain ‘critical’ event information that should not be
lost. Therefore, the SEL device will not automatically clear the SEL if it gets full. This operation is
based on the assumption that it is the first events that are most indicative of the root cause of a
problem, and that later events may be ‘side-effects’ which, if the SEL were implemented as a ‘FIFO’
could cause the ‘root cause’ events to get lost. Instead, System Management Software has the
responsibility for determining when SEL entries should be cleared. System Management Software can
migrate the SEL contents to disk, to the system’s event log, or even to remote storage as desired.
•
Reads and interprets the SDR Repository information. System Management Software uses this
information to determine the sensor population and capabilities. The Sensor Data information can also
be presented to provide a description of the system’s manageability features.
•
‘Polls’ sensors. System Management Software takes the SDR information and uses it to access the
sensors, either in a polled or ‘on demand’ basis, to monitor and present the system’s current health and
status. Note that whenever possible, System Management Software should rely on event generation for
detecting error conditions, and avoid the overhead associated with polling. ‘Normal’ health status does
not generally need to be polled, but would be delivered ‘on demand’.
•
Potential Event Message Source. System Management Software can also send Event Messages to get
events added to the System Event Log. This allows SMS to record information that may be required
for ‘post-mortem analysis’ should it become necessary for System Management Software to shutdown, power-cycle, reset, or otherwise ‘off line’ the system as a response to a system event. The SEL
should be reserved for ‘critical’ hardware-related errors. The majority OS and software errors should
not be written to the SEL. Candidate errors for the SEL are errors that block normal ‘in-band’
management mechanisms.
1.9 SMI Handler
Not all platform management events come through management controllers or from system software. Some
events come from baseboard interrupts. This may include platform events such as correctable and uncorrectable
ECC errors, critical NMIs (Non-maskable Interrupts) such as PCI PERR (parity error), PCI SERR (system error),
bus timeout interrupts, etc. In some implementations, the platform management hardware maps these ‘critical
interrupts’ to the system SMI (System Management Interrupt) signal. The SMI Handler runs, and, as part of
handling these critical interrupts, generates an Event Message to cause the event to get logged in the SEL. The
SMI Handler can also take autonomous, ‘emergency’ action, such as powering off or resetting the system, or
propagating an NMI to the operating system.
The SMI Handler is typically a routine that is loaded and initialized into a protected area of memory by the BIOS.
SMI is the highest priority non-maskable interrupt in the system. When asserted, it switches the processors into
‘System Management Mode’ (SMM). Upon entry into SMM, the processor state is saved and a memory
configuration is entered where the SMI Handler has full access to system memory and I/O space. This allows the
SMI Handler to implement its management functions in an OS-independent manner. The key aspect to this being
that the SMI Handler code will run even if the OS is ‘hung’. This makes it ideal for implementing certain critical
and emergency management functions.
The explicit interface and functionality between an SMI Handler and the BMC is implementation dependent and
is not covered by this specification. The implementation of system-specific communication interfaces can be
aided using the OEM bits and flags in the BMC-system interface commands.
20
Intelligent Platform Management Interface Specification
1.10 Overview of Changes from IPMI v1.0
This section assumes familiarity with the IPMI v1.0 specification. If you’re new to IPMI, you can skip ahead to
the next section. This is not intended to be a complete “to the bit” list of all the changes, but is provided as a guide
for understanding what’s involved in moving to support IPMI v1.5.
•
Most commands that have version numbers had their version numbers rev'd to 51h for IPMI 1.5
•
The Get Device ID command was extended a couple of optional firmware revision bytes, per NEC request.
Just display them as hex-ASCII.
•
The IPMI sensor commands are the same as in v1.0, though a small amount of typo corrections and
additional clarifications have been made in those sections.
•
The IPMI watchdog commands are backward compatible with IPMI v1.0. A previously reserved bit has been
defined as a new ‘don’t stop’ bit that allows the watchdog timer to be reconfigured without stopping it.
•
The IPMI v1.0 event commands are the same. A couple of new event commands, Set/Get Last Processed
Event have been added to allow someone using the new IPMI v1.5 PEF capability to set or determine
whether or not PEF has pending events to process.
•
Sensor Event/Reading Type codes - The POST Error sensor is now called "System Firmware Progress" and
includes new offsets for POST errors and progress the follow the DMTF ASF specification. There's also a
new 'Management Subsystem Health' sensor type, 28h. Please check the Sensor Event/Reading Type code
table for any other changes.
•
The SEL Event Record format is the same except that four previously reserved bits now hold a channel
number, and the SEL Record version "EvMRev" field goes from 3h to 4h. It’s possible that two events would
be identical except for the channel number field. Software that handles or displays events should interpret the
channel number field in order to differentiate between events coming from different channels.
•
The SDRs are the same with the exception of the version number and a field changes to
accomodateaccommodate a necessary fourth bit for the channel number. This change affected SDRs 01h,
02h, 10h, 11h and 12h. The addition of a channel number in the Type 12h SDR caused the two bytes
following to get pushed down.
•
The Entity Instance value in the Entity Association Record has been split into two ranges: one for 'system
relative' IDs and another for 'device relative’ IDs. Most implementations will be able to use their existing
Entity Instance assignments since the lower range of values are for the 'system relative' Entity Instance
values, which map to the IPMI v1.0 definition of the Entity Instance value.
•
SDR Type 14h is being deprecated. IPMI 1.5 systems and software should not use SDR Type 14h. Software
should use the new Get Channel Info command instead.
•
The SEL Event Record format is the same except that four previously reserved bits now hold a channel
number, and the SEL Record version "EvMRev" field changes from 3h to 4h.
•
The IPMB message format remains the same.
•
The Send Message command is backward compatible with v1.0 with respect to using it to access the IPMB.
I.e. you don't need to make any changes to access the IPMB.
•
The Master Write/Read I2C has had the “I2C” dropped from the name. It is now the Master Write/Read
command. This command is backward compatible with IPMI v1.0. Reserved bits 7:4 in byte one have
become a Channel ID, but 0h is when accessing the IPMB or private management busses as in IPMI v1.0. A
non-zero channel value would only be used for accessing additional IPMBs or a PCI Management Bus.
21
Intelligent Platform Management Interface Specification
•
22
The read/write FRU commands are the same.
Intelligent Platform Management Interface Specification
2. Logical Management Device Types
The Intelligent Platform Management architecture is comprised of a number of ‘logical’ management devices.
These are implemented by and within the ‘physical’ system elements such as the management controllers, I2C bus,
system ASICs, etc.
Each ‘logical device’ type represents the definition of a particular set of mandatory and optional commands. For the
purposes of this specification, the logical management devices are:
IPM Device
Intelligent Platform Management device. This represents the ‘basic’ intelligent
device that responds to the platform sensor and event interface messages. All
Intelligent Platform Management devices on the IPMB are expected to respond to
the mandatory ‘IPM Device’ commands. These are also referred to as the ‘global’
commands. Management Controllers that communicate via compatible messages
to the system are also considered IPM devices.
Sensor Device
The Sensor Device is a device that provides the command interface to one or more
sensors. Sensor Devices provide a set of commands for discovering, configuring
and accessing sensors.
SDR Repository Device
The SDR Device is the logical management device that provides the interface to
the Sensor Data Records (SDR) for the system. The SDR Device provides a set of
commands for storing and retrieving Sensor Data Records.
SEL Device
The SEL Device is the logical management device that provides the interface to
the System Event Log for the system. The SEL Device provides a set of
commands for managing the System Event Log.
FRU Inventory Device
The FRU Inventory Device provides the interface to a particular module’s FRU
Inventory (serial number, part number, asset tag, etc.) information. There will
typically be one set of FRU Inventory information for each major module in the
system. There can be just as many FRU Inventory Devices providing access to
that information. The Primary FRU Inventory Device for a given management
controller is defined as the device that contains the information about the FRU that
holds the management controller itself.
Event Receiver Device
The Event Receiver Device accepts and acknowledges Event Request Messages.
The normal action for the Event Receiver Device is to then pass the Event
Message to the SEL Device for logging. An IPMB Event Receiver refers to an
Event Receiver that accepts Event Messages from the IPMB.
Event Generator Device
The Event Generator Device represents the functionality that is used to deliver
Event Messages to the Event Receiver Device. The Event Generator Device
includes commands to allow configuration of Event Message delivery. The term
IPMB Event Generator refers to the capability to generate an Event Message on
the IPMB. The BMC is typically an IPMB Event Receiver, but not an IPMB Event
Generator.
Application Device
A physical instantiation of an Intelligent Platform Management device will most
likely have some ‘device specific’ functionality that it implements that falls
outside the ‘standard’ sensor and event functions. This functionality is referred to
as the devices ‘Application’ functionality. Commands that address this
functionality are viewed as being handled by an ‘Application’ logical device.
23
Intelligent Platform Management Interface Specification
PEF Device
This logical device represents the functions associated with comparing an event
message against a set of selectable ‘event filters’ and generating a selectable action
on a match.
Alert Processing Device
This logical devices represents the functions associated with queuing up and
processing alerts, and alert policies that determine which destinations an alert will
be sent to.
Chassis Device
This chassis control device represents functions associated with recovery control
actions such as power on/off, power cycle, reset, diagnostic interrupt, chassis
identification indicator, and system boot.
Message Handler
This logical device represents the functions associated with configuration and
operation of message authentication and routing, both internal to the BMC and
among the different interfaces to the BMC.
The Intelligent Platform Management Bus can be considered as defining other ‘logical’ devices as well, such as the
‘Bridge Device’ for the Intelligent Chassis Management Bus (ICMB). Refer to the Intelligent Platform Management
Bus Protocol Specification for more information.
24
Intelligent Platform Management Interface Specification
Figure 2-1, Intelligent Platform Management Logical Devices
INTELLIGENT PLATFORM
MANAGEMENT CONTROLLER 'A'
SENSORS
INTELLIGENT PLATFORM
MANAGEMENT CONTROLLER 'B'
SENSOR DEVICE
APPLICATION DEVICE
SENSOR DEVICE
EVENT GENERATOR
IPM DEVICE
EVENT GENERATOR
APPLICATION DEVICE
IPM DEVICE
MESSAGE
HANDLER
MESSAGE
HANDLER
BRIDGE
DEVICE
IPMB MESSAGE
INTERFACE
ICMB
INTERFACE
SENSORS
IPMB MESSAGE
INTERFACE
INTELLIGENT PLATFORM MANAGEMENT BUS (IPMB)
ICMB
Add-in Card
FRU INVENTORY
EEPROM
NON-INTELLIGENT
I2C SENSOR
IPMB MESSAGE
INTERFACE
SENSORS
IPM DEVICE
NON-INTELLIGENT DEVICES on IPMB
SENSOR DEVICE
PCI Mgmt Bus
PCI MGMT. BUS
INTERFACE
EVENT RECEIVER
LAN
Controller
NIC I/F (e.g.
SMBus)
SEL DEVICE
FRU INVENTORY
DEVICE
PEF DEVICE
SDR REPOSITORY
DEVICE
ALERT PROCESSING
DEVICE
INITIALIZATION
AGENT
APPLICATION DEVICE
Serial
Serial Port Sharing
Logic
SERIAL
INTERFACE
MESSAGE
HANDLER
Privately
Managed
Non-Volatile
Storage
LAN
Baseboard
Serial Controller
BASEBOARD MANAGEMENT
CONTROLLER
SYSTEM MESSAGE
INTERFACE
SYSTEM - BMC INTERFACE
SYSTEM MESSAGE
INTERFACE
SYSTEM SOFTWARE
MESSAGE HANDLER
BIOS
SYSTEM
MANAGEMENT S/W
SMI HANDLER
OS
25
Intelligent Platform Management Interface Specification
26
Intelligent Platform Management Interface Specification
3. Baseboard Management Controller (BMC)
The management architecture can be implemented by centralizing the most common functions into a ‘central’
management controller in the system. This controller is often called the Baseboard Management Controller, or
BMC. In some system implementations, the BMC may be the only management controller. The BMC typically
provides the following platform management functions:
System Interface
The BMC provides the System Interface to the IPMI-based platform management
subsystem. The System Interface is the interface through which system software
sends and receives messages to and from the BMC.
Message Handler
The BMC provides functions for routing messages between the different interfaces,
including the System Interface, IPMB, serial/modem, LAN, etc. The Message
Handler may also be thought of as where shared messaging functions for configuring
channel characteristics and user privileges reside.
SEL Interface
The BMC provides the interface to the System Event Log (SEL). The BMC allows
the SEL to be accessed both from the system side, but also from the Intelligent
Platform Management Bus and other external interfaces to the BMC.
Event Generator
The BMC itself will typically be responsible for monitoring and managing the
system board. For example, monitoring baseboard temperatures and voltages. As
such, the BMC will also be an Event Generator internally, sending the Event
Messages that it generates internally to its Event Receiver functionality. Note the
BMC is not typically and IPMB Event Generator. That is, it does not typically issue
Event Messages onto the IPMB.
SDR Repository Interface The BMC will also provide the interface to the SDR (Sensor Data Record)
Repository. As with the System Event Log, the BMC allows the records in the SDR
Repository to be accessed either via the Intelligent Platform Management Bus or via
the system interface.
IPMB Interface
A BMC will typically support an IPMB connection. The IPMB enables the BMC to
accept IPMI request messages from other management controllers in the system. The
IPMB provides a simple integration point for connecting the ‘chassis’ management
features to the baseboard management. The IPMB can also provide a connection that
enables add-in cards to get access to the platform management subsystem.
A BMC that includes IPMB Interface support also provides the capability for system
software to send and receive messages to and from the IPMB using the BMC as a
kind of communication controller.
IPMB Event Receiver
When an IPMB is implemented, the BMC serves as the primary IPMB Event
Receiver for the system. Event Messages can be sent to the BMC from the system or
from other controllers the IPMB.
Private Bus Controller
FRU SEEPROMs may be provided on Private Management Busses behind the BMC.
The BMC can server as a communication controller that provides access to Private
Management Busses and provide access to FRU SEEPROMs and other nonintelligent devices via the Master Write-Read command.
FRU Information
Interface
The BMC provides access to FRU information for the base system board. The FRU
information for the board holding the BMC is obtained by sending FRU Commands
to the BMC’s LUN 00b.
OEM Commands
A BMC implementation can include special support for OEM-unique features and
27
Intelligent Platform Management Interface Specification
functions. One way of accomplishing this is by implementing OEM commands
through the IPMI messaging interfaces.
Watchdog Timer
IPMI defines common command interfaces for configuring and accessing a watchdog
timer function in the BMC. This timer can be used as an aid in monitoring the health
of BIOS and system software. The watchdog timer can be used by different types of
software such as BIOS, pre-boot, OS, and system management software. Once
started the timer must be periodically reloaded by software in order to keep it from
expiring. If software ceases to run, the timer will expire and generate a timeout
action.
The IPMI definition allows different actions to be selected to occur on a watchdog
timeout. This includes reset, power off, power cycle, etc. and a ‘pre-timeout
interrupt’ option that, if provided, can be used to generate a system interrupt shortly
before the timeout. The definition includes ‘timer use’ fields that keep track of what
type of software (BIOS, OS, System Management Software, etc.) started the timer.
The timeout action and ‘timer use’ information can be automatically logged to the
SEL when the timeout occurs. This provides a record of when the timeout occurred,
what software was using the timer, and what action was taken.
28
Intelligent Platform Management Interface Specification
In addition, a BMC may implement additional functions for messaging and alerting, including:
Serial/Modem Interface
The BMC can provide a serial/modem interface that allows it to receive IPMI
messages over a serial connection to the BMC.
Serial Port Sharing
Serial Port Sharing is a separate capability that works in conjunction with the
serial/modem interface. Serial Port Sharing provides a mechanism where the BMC
can control logic that allows a single serial connector to be shared between a serial
controller on the baseboard and a serial controller for the BMC.
LAN Interface
From the IPMI point-of-view, the interface to the network controller is dedicated to
the BMC. That is, there are no special commands for coordinating the sharing of the
network controller between system software access and BMC access, as there are
with Serial Port Sharing. If the network controller is shared between system software
and the BMC, this is generally accomplished via special hardware in the network
controller that enable BMC traffic and system traffic to be interleaved.
PCI Management Bus
Interface
The BMC can implement a PCI Management Bus Interface that enables the BMC to
accept IPMI request messages from add-in cards that plug into a PCI slot. The PCI
management bus and IPMB can serve complementary roles. The IPMB providing a
mechanism for integrating management functions between baseboard and chassis
board functions, while the PCI Management Bus connection can be used to support
add-in cards. This division allows the inter-board management communications to be
kept separate from add-in card communications.
Platform Event Filtering
(PEF)
Platform Event Filtering is an ability for the BMC to perform a configurable action
based on an event, by matching the event against a set of ‘event filters’. The actions
that a BMC can elect to implement include power off, reset, power cycle, generate
diagnostic interrupt, and send an alert.
Alert Processing
IPMI v1.5 supports the ability for a BMC to deliver alerts such as SNMP Traps in the
Platform Event Trap (PET) format, over media such as LAN and PPP, plus the
ability to perform numeric and/or alphanumeric paging via a serial/modem
connection. Alert processing includes the ability to support sending alerts to multiple
destinations, and to cluster destinations into sets called ‘Alert Policies’. Enabling
alert policies with PEF makes it possible to configure the system so critical events
are delivered to destinations in a ‘high priority’ alert policy, while non-critical events
would go to destinations in a ‘low priority’ alert policy.
29
Intelligent Platform Management Interface Specification
3.1 Required BMC Functions
The following table summarizes the major required and optional functions for an IPMI-conformant BMC.
Table 3-1, Required BMC Functions
Function
M/O
IPM Device
M
System Interface
M
SDR Repository
M
IPMB Interface
O
Watchdog Timer
M
Event Receiver
M
SEL Interface
M
FRU Inventory
M
(v1.5)
Initialization Agent
M
Sensors
O
30
Description
The BMC must implement the mandatory IPM Device commands. If an IPMB
is provided, the mandatory commands must be accessible from the IPMB
unless otherwise noted.
The implementation must provide BMC access via one of the specified IPMI
system interfaces.
The BMC must provide a SDR Repository to hold Sensor, Device Locator, and
Entity Association records for all sensors in the platform management
subsystem. This does not need to include SDRs for sensors that only generate
events. It is recommended that at least 20% additional space is provided for
platform management extensions.
The SDR Repository must be accessible via the system interface. If an IPMB
is provided, the SDR Repository must be readable via that interface as well.
SDR update via the IPMB interface is optional.
SDR Repository access when the system is powered up or in ACPI ‘S1’ sleep
is mandatory, but access when the system is powered-down or in a >S1 sleep
state is optional.
The IPMB is highly recommended, but optional. The BMC must provide the
system interface to the IPMB. If an IPMB is implemented, at least one of the
specified IPMB connectors must be provided. Refer to the IPMB Protocol
specification for connector definition. In addition the BMC must implement a
message channel that allows messages to be sent from the IPMB to the
system interface, and vice-versa, and any other mandatory IPMB support
functions and commands.
The BMC must provide the standardized Watchdog Timer interface, with
support for system reset action. Certain functions within the Watchdog Timer
are optional. Refer to the sections on the Watchdog Timer for information.
The BMC must implement an Event Receiver function and accept Event
Messages via the system interface. If an IPMB is provided, the Event Receiver
function must also accept Event Messages from the IPMB. Event Receiver
operation while the system is powered up or in ACPI ‘S1’ sleep is mandatory,
but operation when the system is powered down or in a >S1 sleep state is
optional.
The BMC must provide a System Event Log interface. The event log must hold
at least 16 entries. SEL access must be provided via the system interface. If
an IPMB is provided, the SDR Repository must be accessible via that interface
as well. SDR Repository access when the system is powered up or in ACPI
‘S1’ sleep is mandatory, but access when the system is powered-down or in a
>S1 sleep state is optional.
The BMC must provide a logical Primary FRU inventory device, accessible via
the Write- and Read FRU Data commands. The FRU Inventory Device Info
command must also be supported. It is highly recommended that all other
management controllers also provide a Primary FRU inventory device. (This
was optional in IPMI v1.0.)
The initialization agent function is one where the BMC initializes event
generation and sensors both internally and on other management controllers
according to initialization settings stored in the SDR for the sensor.
The BMC can provide sensors. A typical server BMC would provide sensors
for baseboard temperature, voltage, and chassis intrusion monitoring.
Intelligent Platform Management Interface Specification
Function
M/O
Internal Event
Generation
M
External Event
Generation
O
PCI Management Bus
Interface
O
LAN Messaging
LAN Alerting
Serial Messaging
O
O
O
Basic Mode
M
PPP Mode
O
Terminal Mode
O
Direct Connect Mode
M
Modem Connect Mode
O
Bridging Support
O/M
Dial Page
O
PPP Alerting
O
Description
The BMC must generate internal events for the Watchdog Timer. It is highly
recommended that sensors generate events to eliminate the need for system
management software to poll sensors, and to provide “post-mortem” failure
information in the SEL. Internal event generation for sensors is optional, but
highly recommended - particularly for ‘environmental’ (e.g. temperature and
voltage) sensors.
The BMC could be designed to accept the Set Event Receiver command to
allow it to be set as an IPMB Event Generator and send its event messages to
another management controller. This would primarily be used for development
and test purposes.
The BMC supports a connection to a PCI Management bus through which the
BMC can send and receive IPMI Messages. System software can also access
the PCI Management Bus by sending commands to the BMC via the System
Interface.
Ability for the BMC to send and receive IPMI Messaging over LAN
Ability to send an Alert over the LAN
Serial messaging is the capability of performing IPMI Messaging over an
asynchronous serial connection to the BMC. If Serial Messaging is supported,
the following sub-functions apply:
Basic Mode is a type of message framing used for IPMI messaging over a
serial connection. Basic Mode support is required if Serial Messaging is
supported.
PPP Mode is support for using PPP protocols and framing for IPMI messaging
over a serial connection.
Terminal Mode is a mechanism for IPMI messaging over serial using printable
ASCII characters. Terminal mode also supports a limited number of text
commands to support legacy ‘text based’ environments.
Direct Connect Mode is support for IPMI messaging over a serial connection
without going through a modem. Direct connect mode is mandatory as part of
Serial Messaging.
Modem Connect Mode is support for IPMI messaging over a serial connection
through a TIA-602-compatible modem, or via modem circuitry that can work
with the IPMI commands defined for modem communication.
The ability to transfer IPMI request and response messages between two
interfaces connected to the BMC.
The following support is required if the corresponding interfaces are supported:
•
serial/modem ÅÆ IPMB
•
serial/modem ÅÆ System Interface
•
LAN ÅÆ IPMB
•
LAN ÅÆ System Interface
Recommended:
•
serial/modem ÅÆ PCI Management Bus
•
LAN ÅÆ PCI Management Bus
Optional:
all other combinations, e.g. serial/modem ÅÆ LAN
Ability to perform a numeric page by dialing. Typically accomplished using an
external modem.
Ability for the BMC to connect to a system
Platform Event Filtering and Serial Messaging with PPP Mode are required if
PPP Alerting is implemented.
31
Intelligent Platform Management Interface Specification
Function
M/O
Callback
O
Basic Mode Callback
M
PPP Mode Callback
O
CBCP Callback
O
Platform Event Filtering
(PEF) and Alert Policies
32
O/M
Description
Callback represents the ability for the BMC to be directed to dial up a selected
or pre-configured destination to establish an IPMI Messaging session. Callback
requires Serial Messaging with Modem Connect Mode.
Required if Callback is supported. BMC uses Basic Mode for IPMI messaging
after connecting to specified destination.
BMC uses PPP Mode for IPMI messaging after connecting to specified
destination.
BMC supports Microsoft CBCP (Callback Control Protocol) for callback. PPP
Mode and PPP Mode Callback support are required if CBCP Callback is
implemented.
Ability for BMC to perform a selectable action on an event. This capability is
mandatory if paging or alerting is supported. Certain actions within PEF are
optional. Refer to the sections on PEF for information. The Alert action and
Alert Policies are mandatory if serial/modem or LAN alerting is supported.
Intelligent Platform Management Interface Specification
4. General Mgmt. Controller Required Functions
All management controllers are required to implement the mandatory IPM Device commands. All other functions
are optional. If a function is implemented, such as Event Generation or Sensors, then the mandatory commands
for that function must be implemented.
It is highly recommended that all controllers that provide sensors also provide event generation for those sensors.
This will eliminate the need for system management software to poll to detect event conditions. It is also highly
recommended that all management controllers provide a Primary FRU Inventory device.
33
Intelligent Platform Management Interface Specification
5. Message Interface Description
The heart of this specification is the definition of the messages and data formats used for implementing sensors,
event messages, Event Generators and Event Receivers, the SDR Repository, and the System Event Log in the
platform management architecture. These messages are designed for delivery via a messaging interface with a
particular set of characteristics. This section presents the general specification of that interface, and of the messages
themselves.
The Message Interface is defined as a ‘request/response’ interface. That is, a request message is used to initiate an
action or set data, and a response message is returned to the Requester . In this document, Request Messages are
often referred to as ‘commands’ or ‘requests’, and Response Messages as ‘responses’.
All messages in this specification share the same common elements as the payload to the ‘command interpreter’ in
the logical device that receives the message. The messaging interfaces differ in the framing, physical addressing,
and data integrity mechanisms that are used to deliver this payload.
The following are the common components of messages specified in this document:
Network Function (NetFn)
A field that identifies the functional class of the message. The Network
Function clusters IPMI commands into different sets. See Section 5.1,
Network Function CodesNetwork Function Codes, for more information.
Request/Response identifier
A field that unambiguously differentiates Request Messages from
Response Messages. In the IPMB Protocol, this identifier is ‘merged’ with
the Network Function code such that ‘Even’ network function codes
identify Request Messages, and ‘Odd’ network function codes identify
Response Messages.
Requester’s ID
Information that identifies the source of the Request. This information
must be sufficient to allow the Response to be returned to the correct
Requester. For example, for the IPMB the Requester’s ID consists of the
Slave Address and LUN of the Requester device. For a multiple stream
system interface the Requester’s ID is the ‘stream id’ for the stream
through which the request was issued.
Responder’s ID
A field that identifies the Responder to the Request. In Request Messages
this field is used to address the Request to the desired Responder, in
Response Messages this field is used to assist the Requester in matching
up a response with a given request.
Command
The messages specified in this document contain a one-byte command
field. Commands are unique within a given Network Function. Command
values can range from 00h through FDh. Code FEh is reserved for future
extension of the specification, and code FFh is reserved for message
interface level error reporting on potential future interfaces.
Data
The Data field carries the additional parameters for a request or a
response, if any.
5.1 Network Function Codes
The network layer in the connection header consists of a six-bit field identifying the function to be accessed. The
remaining two bits are the LUN field. The LUN field provides further sub-addressing within the node.
34
Intelligent Platform Management Interface Specification
The network function is used to cluster commands into functional command sets. In a parsing hierarchy, the LUN
field may be thought of as the selector for a particular Network Function handler in the node, and the Network
Function may be considered the selector for a particular command set handler within the node.
The following table defines the supported network functions. With the exception of the Application and Firmware
Transfer network functions, the commands and responses for a given network function are not node specific. The
format and function for standard command sets is specified later in this specification.
Table 5-1, Network Function Codes
Value(s)
Name
Meaning
Description
00, 01
Chassis
00h identifies the message as a command/request and 01h as a
response, relating to the common chassis control and status functions.
02*, 03*
Bridge
Chassis
Device
Requests and
Responses
Bridge
Requests and
Responses
04, 05
Sensor
/Event
06, 07
App
08, 09
Firmware
0A, 0B
Storage
0C, 0D
Transport
0Eh-2Bh
Reserved
Sensor and
Event
Requests and
Responses
Application
Requests and
Responses
Firmware
Transfer
Requests and
Responses
Non-volatile
storage
Requests and
Responses
Media-specific
configuration
& control
-
02h (request) or 03h (response) identifies the message as containing
data for bridging to the next bus. This data is typically another
message, which may also be a bridging message. This function is
present only on bridge nodes.
This functionality can be present on any node. 04h identifies the
message as a command/request and 05h as a response, relating to
the configuration and transmission of Event Messages and system
Sensors.
06h identifies the message as an application command/request and
07h a response. The exact format of application messages is
implementation-specific for a particular device, with the exception of
App messages that are defined by the IPMI specifications.
Note that it is possible that new versions of this specification will
identify new App commands. To avoid possible conflicts with future
versions of this specification, it is highly recommended that the
OEM/Group network functions be used for providing ‘value added’
functions rather than the App network function code.
The format of firmware transfer requests and responses matches the
format of Application messages. The type and content of firmware
transfer messages is defined by the particular device.
This functionality can be present on any node that provides nonvolatile storage and retrieval services.
Requests (0Ch) and responses (0Dh) for IPMI-specified messages that
are media-specific configuration and operation, such as configuration
of serial and LAN interfaces.
reserved (30 Network Functions [15 pairs])
35
Intelligent Platform Management Interface Specification
Value(s)
Name
Meaning
Description
2Ch-2Dh
Group
Extension
Non-IPMI
group
Requests and
Responses
The first data byte position in requests and responses under this
network function identifies the defining body that specifies command
functionality. Software assumes that the command and completion
code field positions will hold command and completion code values.
The following values are used to identify the defining body:
00h** PICMG - Compact PCI Industrial Computer Manufacturer’s
Group. (www.picmg.com)
01h DMTF Pre-OS Working Group ASF Specification
(www.dmtf.org)
all other Reserved
2Eh-2Fh
30h-3Fh
*
**
OEM/Group
Controllerspecific
OEM/Group
OEM/NonIPMI group
Requests and
Response
-
When this network function is used, the ID for the defining body
occupies the first data byte in a request, and the second data byte
(following the completion code) in a response.
The first three data bytes of requests and responses under this
network function explicitly identify the OEM or non-IPMI group that
specifies the command functionality. While the OEM or non-IPMI group
defines the functional semantics for the cmd and remaining data fields,
the cmd field is required to hold the same value in requests and
responses for a given operation in order to be supported under the
IPMI message handling and transport mechanisms.
When this network function is used, the IANA Enterprise Number for
the defining body occupies the first three data bytes in a request, and
the first three data bytes following the completion code position in a
response.
Vendor specific (16 Network Functions [8 pairs]). The Manufacturer ID
associated with the controller implementing the command identifies the
vendor or group that specifies the command functionality. While the
vendor defines the functional semantics for the cmd and data fields,
the cmd field is required to hold the same value in requests and
responses for a given operation in order for the messages to be
supported under the IPMI message handling and transport
mechanisms.
Network Functions that are only utilized in systems that incorporate Bridge nodes.
This organization value was named ‘Compact PCI’ in revision 1.0 of this document.
5.2 Completion Codes
All Response Messages specified in this document include a completion code as the first byte in the data field of
the response. A management controller that gets a request to an invalid (unimplemented) LUN must return an
error completion code using that LUN as the responder’s LUN (RsLUN) in the response. The completion code
indicates whether the associate Request Message completed successfully and normally, and if not, provides a
value indicating the completion condition.
Completion Codes work at the ‘command’ level. They are responses to the interpretation of the command after it
has been received and validated through the messaging interface. Errors at the ‘network’ (messaging interface)
level are handled with a different error reporting mechanism. For example the SMIC System Interface includes
status codes that are separate from the IPMI message data and used to report changes in communication phase or
errors in the interface.
Completion Code values are split into ‘generic’, ‘device-specific’ (which covers OEM) and ‘command-specific’
ranges. All commands can return Generic Completion Codes. Commands that complete successfully shall return
the 00h, ‘Command Completed Normally’, Completion Code. Commands that produce error conditions, or return
36
Intelligent Platform Management Interface Specification
a response that varies from what was specified by the Request parameters for the command, shall return a nonzero Completion Code, as specified in the following table.
Table 5-2, Completion Codes
Code
Definition
GENERIC COMPLETION CODES 00h, C0h-FFh
00h
C0h
C1h
C2h
C3h
C4h
C5h
C6h
C7h
C8h
C9h
CAh
CBh
CCh
CDh
CEh
CFh
D0h
D1h
D2h
D3h
D4h
D5h
FFh
Command Completed Normally.
Node Busy. Command could not be processed because command processing resources
are temporarily unavailable.
Invalid Command. Used to indicate an unrecognized or unsupported command.
Command invalid for given LUN.
Timeout while processing command. Response unavailable.
Out of space. Command could not be completed because of a lack of storage space
required to execute the given command operation.
Reservation Canceled or Invalid Reservation ID.
Request data truncated.
Request data length invalid.
Request data field length limit exceeded.
Parameter out of range. One or more parameters in the data field of the Request are out
of range. This is different from ‘Invalid data field’ (CCh) code in that it indicates that the
erroneous field(s) has a contiguous range of possible values.
Cannot return number of requested data bytes.
Requested Sensor, data, or record not present.
Invalid data field in Request
Command illegal for specified sensor or record type.
Command response could not be provided.
Cannot execute duplicated request. This completion code is for devices which cannot
return the response that was returned for the original instance of the request. Such
devices should provide separate commands that allow the completion status of the original
request to be determined. An Event Receiver does not use this completion code, but
returns the 00h completion code in the response to (valid) duplicated requests.
Command response could not be provided. SDR Repository in update mode.
Command response could not be provided. Device in firmware update mode.
Command response could not be provided. BMC initialization or initialization agent in
progress.
Destination unavailable. Cannot deliver request to selected destination. E.g. this code can
be returned if a request message is targeted to SMS, but receive message queue
reception is disabled for the particular channel.
Cannot execute command. Insufficient privilege level.
Cannot execute command. Command, or request parameter(s), not supported in present
state.
Unspecified error.
DEVICE-SPECIFIC (OEM) CODES 01h-7Eh
01h-7Eh
Device specific (OEM) completion codes. This range is used for command-specific codes
that are also specific for a particular device and version. A-priori knowledge of the device
command set is required for interpretation of these codes.
COMMAND-SPECIFIC CODES 80h-BEh
80h-BEh
all other
Standard command-specific codes. This range is reserved for command-specific
completion codes for commands specified in this document.
reserved
5.3 Completion Code Requirements
Completion Codes are provided as an aid in system debugging and error handling. All devices meeting the
command specifications of this document shall implement the 00h, ‘Command Completed Normally’ for the
commands specified in this document.
37
Intelligent Platform Management Interface Specification
It is mandatory that devices that produce error conditions, or return a response that varies from what was
specified by the Request parameters for the command, return a non-zero Completion Code from the preceding
table.
In some cases, it is required that a particular completion code be returned for a specified condition. This typically
occurs with command-specific completion codes. These cases are documented in the sections describing the
particular command or function.
Otherwise, if a device implementation produces a completion condition that matches a Generic or Commandspecific completion code for the command, the device shall either return that specific value, or the ‘unspecified
error’ Completion Code, FFh. It is highly desirable that device implementations return an explicit completion
code, rather than ‘unspecified error’, whenever feasible.
In the case that multiple ‘non-zero’ completion conditions occur simultaneously, the implementation should
return whichever completion code the implementer deems to best indicate the condition that the Requester should
correct or handle first.
New for IPMI 1.5 Controllers and software that handle IPMI commands: The value C1h (Invalid Command)
must be returned for unsupported commands, except when the controller or software is in a mode where general
command handling is unavailable. For example, if the controller is in a firmware update mode, it is legal to return
D1h (Command response could not be provided, device in firmware update mode) instead of C1h.
5.3.1
Response Field Truncation on non-zero Generic Completion Codes
The responder may, as an implementation option, truncate fields following a non-zero completion code field.
Typically, a responder will truncate all fields following a non-zero completion code. If additional fields are
returned, however, they should be assumed to have device-specific content.
5.3.2
Summary of Completion Code Use
The following is a summary list of the completion code rules and guidelines.
38
•
A 00h Completion Code must be returned with a normal response to a standard command.
•
It is recommended that a 00h Completion Code also be returned for the normal responses to OEM
commands.
•
A non-zero Completion Code must be returned for an error or atypical response to a standard command.
•
It is recommended that a non-zero Completion Code be returned for an error or atypical response to an
OEM command.
•
The value C1h (Invalid Command) must be returned for unsupported commands, except when in a mode
where general command handling is unavailable.
•
If the specification calls out that a particular completion code must be returned for a given condition, that
code must be returned.
•
Otherwise, it is recommended that an implementation return the closest generic completion code for an
error condition. If an implementation is resource constrained or the error classification is ambiguous, the
FFh (unspecified error) completion code can be returned.
•
Device-specific (OEM) completion codes should only be returned when a suitable generic completion code
is unavailable. Generic software will treat device-specific completion codes as if they were FFh
(unspecified error) completion codes.
Intelligent Platform Management Interface Specification
•
Except for mandatory completion codes, software must not depend on a particular non-zero completion
code to be returned for a given error condition, since it is possible that an FFh or device-specific code
could be returned instead.
•
It is illegal to return a generic or command-specific completion code for a condition that doesn’t exist,
unless it is being used as part of emulating a device or interface. For example, an implementation might
enable the Master Write-Read command to be used to access a Private Management Bus interface that is
not physically an I2C bus. The implementation is allowed to return completion codes related to I2C bus
status as part of the emulation.
5.4 Sensor Owner Identification
The definition for the Request/Response identifier, Requester’s ID, and Responder’s ID are specific to the
particular messaging interface that is used. However, the Sensor Data Record and SEL information must contain
information to identify the ‘owner’ of the sensor. For management controllers, a Slave Address and LUN identify
the owner of a sensor on the IPMB. For system software, a Software ID identifies the owner of a sensor. These
fields are used in Event Messages, where events from management controllers or the IPMB are identified by an
eight-bit field where the upper 7-bits are the Slave Address or System Software ID. The least significant bit is a 0
if the value represents a Slave Address and a 1 if the value represents a System Software ID.
The Sensor Number is not part of the Sensor Owner ID, but is a separate field used to identify a particular sensor
associated with the Sensor Owner. The combination of Sensor Owner ID and Sensor Number uniquely identify a
sensor in the system.
Table 5-3, Sensor Owner ID and Sensor Number Field Definitions
IPMB Sensor Owner ID
System Sensor Owner ID
7:1
0
7:1 System Software ID (7-bits)
0
1b (ID is a Software ID)
See Table 5-4, System Software IDs,
below.
Sensor Number (8 bits, FFh = reserved)
Slave Address (7-bits)
0b (ID is a slave address)
LUN (2-bits)
Sensor Number (8-bits, FFh = reserved)
5.5 Software IDs (SWIDs)
The following table presents a list of the Sensor Owner IDs for ‘system software’ sensor owners or IPMI message
generators. These values are used when system software issues Event Messages via the system interface, and
when remote console software sends messages to the BMC. For example, if BIOS detects a processor failure, it
can generate an Event Message to get the failure event logged. When it formats the Event Message, the BIOS
‘System Owner ID’ is included in the Event Message. Later, System Management Software can access the
System Event Log and tell that the Event Message was generated by BIOS.
For IPMB Messages, the Sensor Owner ID is assumed to be the same as the device that originated the message.
Therefore, the Slave Address and LUN of the Event Generator are used. For system-side sensors, it is assumed
that the class of software that generates the sensor commands is the ‘owner’ of the sensor.
39
Intelligent Platform Management Interface Specification
Table 5-4, System Software IDs
System Software Type
IDs (7-bit)
bit 0
BIOS
SMI Handler
System Management Software
OEM
Remote Console software 1-7
Terminal Mode Remote Console software
reserved
00h-0Fh
10h-1Fh
20h-2Fh
30h-3Fh
40h-46h
47h
remaining
1b
1b
1b
1b
1b
1b
1b
1
Resultant 8-bit value
1
01h, 03h, 05h, …1Fh
21h, 23h, 25h, … 3Fh
41h, 43h, 45h, … 6Fh
61h, 63h, 65h, … 7Fh
81h, 83h, 87h, … 8Dh
8Fh
-
1. The System Software ID is often used in an 8-bit field where the least-significant
bit is a 1b to indicate that the field holds a Software ID rather than a slave
address. One example of this occurs in the first byte of the Generator ID field in
an event message. The last column in the above table illustrates how the 7-bit
Software ID appears in such a 1-byte field.
5.6 Isolation from Message Content
The SEL, SDR, and Event commands are designed so that the ‘devices’ that implement those command sets are
generally isolated from the content of the SEL Entry, Sensor Data Record, and Event Message contents. That is,
the Event Receiver device receives and routes Event Messages, but does not interpret them. Similarly, the SEL
and SDR devices retrieve and store log entries and Sensor Data Records, respectively, without interpreting their
contents.
40
Intelligent Platform Management Interface Specification
41
Intelligent Platform Management Interface Specification
6. IPMI Messaging Interfaces
This section introduces the common characteristics of the messaging interfaces to the BMC and between the BMC
and system software. As mentioned earlier, there are three System Interface implementations specified for the BMC:
SMIC, KCS, and BT. The BMC can also be communicated with through additional interfaces such as the IPMB,
ICMB, LAN, and Serial/Modem interfaces. Information specific to the operation and usage of a particular interface
is given in later sections.
6.1 Terminology
The ICMB, LAN, and Serial/Modem interfaces are typically used to communicate with management software on
another system. The remote software that is used to communicate with the BMC is referred to as the remote console.
Although the word ‘console’ is used, the remote software may or may not provide a user interface or require user
interaction.
Local software running on the managed system and using the System Interface to the BMC will generally be
referred to as system management software or SMS. Unless otherwise indicated, the direction of communications is
given with respect to the BMC. E.g. transmitted, outbound, or outgoing messages are issued from the BMC.
Received, inbound, or incoming messages are accepted by the BMC. Other terminology used for IPMI Messaging
will be introduced as it is used in the following sections.
6.2 Channel Model
IPMI uses a ‘channel model’ for directing communication between different interfaces in the BMC. Channels serve
as the means for identifying the medium for a messaging interface, and for configuring user information and
passwords, message authentication, access modes and privilege limits associated with that interface.
Each channel has its own set of configuration parameters for user information and channel privilege limits. This
allows different sets of user names and passwords and different levels and types of authentication to be used on
different channels. IPMI Messaging and Alerting can also be independently enabled or disabled for an entire
channel on a per channel basis.
Channels share commands related to authentication, access, and configuration. These commands are independent
from the type of communication medium. This reduces the amount of medium-specific information that software
needs to deal with, and simplifies task such as bridging IPMI messages between different media.
42
Intelligent Platform Management Interface Specification
6.3 Channel Numbers
Each interface has a channel number that is used when configuring the channel and for routing messages between
channels. Only the channel number assignments for the primary IPMB and the System Interface are fixed, the
assignment of other channel numbers can vary on a per-platform basis. Software uses a Get Channel Info command
to determine what types of channels are available and what channel number assignments are used on a given
platform. The following table describes the assignment and use of the channel numbers:
Table 6-1, Channel Number Assignments
Channel
Number
0
1-7
Implementationspecific
8h-Dh
Eh
Fh
Type/Protocol
Primary IPMB
Present I/F
System Interface
Description
Channel 0 is assigned for communication with the primary IPMB.
IPMB protocols are used for IPMI messages. Refer to [IPMB] for
more information.
Channels 1-7 can be assigned to different types types of
communication media and protocols for IPMI messages (e.g. IPMB,
LAN, ICMB, etc.), based on the system implementation. For IPMI
1.5, ‘Channel Protocol Type’ and ‘Channel Medium Type’ numbers
identify the channel’s protocol and medium, respectively. Software
can use the Get Channel Info command to retrieve this information.
Reserved
The value Eh is used as a way to identify the current channel that
the command is being received from. For example, if software wants
to know what channel number it’s presently communicating over, it
can find out by issuing a Get Channel Info command for channel E.
Channel ‘F’ is assigned for routing messages to the system
interface.
6.4 Channel Protocol Type
The protocol used for transferring IPMI messages on a given channel is identified using a channel protocol type
number. In earlier versions of the specification, this also implied the particular medium for the channel. The
Channel Medium TypeChannel Medium Type number is now used to explicitly indicate the class of the media.
Both these values are used in the Get Channel InfoGet Channel Info command.
Sensor Data Record 14h - BMC Message Channel Info is superceded by the Get Channel Info command. New
implementations should use the Get Channel Info command instead.
Table 6-2, Channel Protocol Type Numbers
Channel
Protocol
(5-bits)
00h
01h
02h
03h
04h
Name
n/a
IPMB-1.0
ICMB-1.0
reserved
IPMI-SMBus
05h
KCS
06h
07h
SMIC
BT-10
Description
reserved
Used for IPMB, serial/modem Basic Mode, and LAN
ICMB v1.0 - See Section 8, ICMB InterfaceICMB Interface.
reserved
[1]
IPMI on SMBus 1.x - 2.x
[2]
Request = [rsSA, Netfn(even)/rsLUN, 00h, rqSA, rqSeq/rqLUN, CMD , <data>,
PEC]
Response = [rqSA or rqSWID, NetFn(odd)/rqLUN, 00h, rsSA or rsSWID,
rqSeq/rsLUN, CMD, completion code, <data>, PEC]
KCS System Interface Format - See Section 9, Keyboard Controller Style (KCS)
InterfaceKeyboard Controller Style (KCS) Interface
SMIC System Interface Format - See Section 10, SMIC InterfaceSMIC Interface
BT System Interface Format, IPMI v1.0 - See Section 11, Block Transfer (BT)
InterfaceBlock Transfer (BT) Interface
43
Intelligent Platform Management Interface Specification
1.
2.
08h
BT-15
09h
1Ch-1Fh
all other
TMode
n/a
BT System Interface Format, IPMI v1.5 - See Section 11, Block Transfer (BT)
InterfaceBlock Transfer (BT) Interface
Terminal Mode - See Section 13.7, Terminal ModeTerminal Mode
OEM Protocol 1 through 4, respectively
reserved
Note that the IPMI format is intentionally illegal with respect to the SMBus specification protocols in order to provide a way
for a management controller to unambiguously differentiate IPMI messages from SMBus transactions. This enables a
management controller to support both SMBus and IPMI protocols without concern that they would overlap. The PEC
(packet error code) is an 8-bit CRC calculated per the SMBus 2.0 specification. This format makes it simple to use the
same hardware or firmware routines for data integrity checking of both IPMI and SMBus messages.)
Note that certain network functions, such as OEM/Group, require additional standard fields within the <data> portion of a
message.
6.5 Channel Medium Type
The Channel Medium Type number is a seven-bit value that identifies the general class of medium that is being
used for the channel.
Table 6-3, Channel Medium Type Numbers
Channel
Type
0
1
2
3
4
5
6
7
8
9
Ah
Bh
Ch
60h-7Fh
all other
44
Description
reserved
2
IPMB (I C)
ICMB v1.0
ICMB v0.9
802.3 LAN
Asynch. Serial/Modem (RS-232)
Other LAN
PCI SMBus
SMBus v1.0/1.1
SMBus v2.0
reserved for USB 1.x
reserved for USB 2.x
System Interface (KCS, SMIC, or BT)
OEM
reserved
Intelligent Platform Management Interface Specification
6.6 Channel Access Modes
Session-based channels can be configured to provide IPMI Messaging access only when the system is in certain
states. This allows the system user to configure various levels of security and remotely accessible features. The
access modes are summarized in the Table 6-4, Channel Access ModesTable 6-4, Channel Access Modes.
Commands allow the power-up default (non-volatile) Access Mode for a channel to be configured, and allow the
Access Mode setting to be changed dynamically. Channel Access Modes are configured using the Set Channel
AccessSet Channel Access command. The Set Channel AccessSet Channel Access command is also used to enable
or disable Alerting for the entire channel.
Support for any given access mode is implementation specific. It is expected that most implementations will
support Disabled and Always Available, and that serial/modem channels will also support the Shared access
mode.
Table 6-4, Channel Access Modes
Pre-boot
Only
The channel is only available out-of-band while the machine is powered-off and during POST until the
boot process is initiated. This option is primarily used with Serial Port Sharing where it may be desirable
to ensure that the BMC does not take control of the serial port during OS run-time. The BMC will not
claim the port once the port has been switched over to the system using the ‘force mux’ option in the Set
Serial/Modem MuxSet Serial/Modem Mux command, unless the system becomes powered down or is
reset.
As a consequence, run-time software must rely on mechanisms such as the IPMI Watchdog Timer to
power down or reset the system in order to enable communication to the BMC under failure conditions.
There is a Modem Ring Time parameter in the serial/modem channels that configures the amount of time
that the BMC waits for RING before directing the modem to connect. This parameter can be used to
enable the BIOS to ‘answer the phone’ instead of the BMC. See Table 13-2, Serial Port Sharing Access
CharacteristicsTable 13-2, Serial Port Sharing Access Characteristics for more information.
Always
Available
Shared
LAN Channels do not typically allow setting Pre-boot Access Mode. If it is provided, BIOS should disable
the channel at the end of POST (start of boot) by using the Set Channel Access command to set the
channel to ‘disabled’ using the volatile setting.
The Pre-boot Only setting does not affect serial/modem alerting. If alerting is enabled and software does
not handle the event, the BMC will take control of the port/channel for the time it takes to deliver the alert.
Alerting can be enabled or disabled for an entire channel via the Set Channel AccessSet Channel
Access command.
The channel is dedicated to communication with the BMC and is available during all system states
(powered-down, powered-up, pre-boot, sleep, run-time, etc.) For IPMI LAN channels, this means that
RMCP packets are handled by the BMC.
For serial/modem channels on systems that support serial port sharing, the port can still be switched
over to the system, however the BMC will always ‘answer the phone’ and respond to escape sequences
and packets that activate the port. The BIOS will typically disable software access to the serial port when
it sees the BMC configured for Always Available mode. This is done to prevent any possible confusion
between auto-answer applications running on the OS and the BMC’s answering of the phone.
The channel can be shared between system software and the BMC.
Shared Mode is typically only used when there is a need to switch the communication resource between
system software and the BMC because the system and BMC cannot readily interleave their traffic on the
medium, as is the case with Serial Port Sharing.
For IPMI LAN Channels, Shared Mode means that the implementation allows system software to receive
and respond to RMCP packets. However, this does not prevent the BMC from handling IPMI RMCP
packets and RMCP Ping/Pong. If software wanted exclusive access to RMCP Packets, it would need to
temporarily disable IPMI messaging by setting the volatile setting of the access mode to ‘disabled’. Note
that if system software failed, a system reset (e.g. watchdog reset) or power down would be required to
restore LAN communication with the BMC.
For serial/modem channels that support Serial Port Sharing, the system BIOS will typically leave the
baseboard serial port available for software use when it sees this mode set. This allows system software
to use the port and any external modem for ‘outgoing’ traffic, while the BMC can still ‘answer the phone’
for incoming calls. Thus, in shared mode, the mux will be set to ‘system’ whenever the BMC is not in the
process of answering a call or handling or establishing an IPMI messaging session.
45
Intelligent Platform Management Interface Specification
There is a Modem Ring Time parameter in the serial/modem channels that configures the amount of time
that the BMC waits for RING before directing the modem to connect. This parameter can be used to
enable ‘auto answer’ OS applications, while providing a way to connect to the BMC if a failure prevents
the run-time application from answering the phone.
Disabled
If the Modem Ring Time is set to a non-zero wait time, the BMC will leave the mux set to the system until
the Modem Ring Time expires, at which time the BMC can answer the phone. If the Modem Ring Time is
set to a zero wait time, the BMC will take the mux and attempt to answer the phone as soon as it detects
an incoming call. See Table 13-2, Serial Port Sharing Access CharacteristicsTable 13-2, Serial Port
Sharing Access Characteristics for more information.
The channel is disabled from being used to communicate with the BMC. The Disabled setting does not
affect alerting. Alerting is separately enabled or disabled via a separate field in the Set Channel Access
command.
6.7 Logical Channels
From the IPMI Messaging point-of-view, a party that bridges a message from one channel to another only is
mainly concerned that it gets the correct response from the BMC. Often, it doesn’t matter to remote console or
system software whether the target channel and devices are physically implemented or not. For example, a BMC
could implement a logical IPMB where the BMC would respond to messaging commands as if there was a
physical IPMB with other management controllers on it. An implementation might elect to do this for several
reasons. One reason would be that the board vendor wanted to use an alternative bus for interconnecting the
management functions within their board set. Another possibility is that a logical IPMB could provide a way to
organize add-on functions to the BMC, such as embedding a logical ICMB Bridge controller.
6.8 Channel Privilege Levels
Channel privilege limits determine the maximum privilege that a user can have on a given channel. One channel
can be configured to allow users to have up to Administrator level privilege, while another channel may be
restricted to allow no higher than User level. The privilege level limits take precedence over the privilege level
capabilities assigned per user.
Channels can be configured to operate with a particular maximum Privilege Level. Privilege levels tell the BMC
which commands are allowed to be executed via the channel. Table 6-5, Channel Privilege LevelsTable 6-5,
Channel Privilege Levels, lists the currently defined privilege levels. The Set Channel AccessSet Channel Access
command is used to set the maximum privilege level limit for a channel. The Set Session Privilege Level
CommandSet Session Privilege Level Command is used to request the ability to perform operations at a particular
privilege level. The Set Session Privilege Level command can only be used to set privilege levels that are less than
or equal to the privilege level limit for the entire channel, regardless of the privilege level of the user.
Table 6-5, Channel Privilege Levels
Callback
User
Operator
Administrator
46
This may be considered the lowest privilege level. Only commands necessary to support initiating a Callback are
allowed.
Appendix G - Command Assignments, provides a list of the commands that are executable when operating at
Callback Level.
Only ‘benign’ commands are allowed. These are primarily commands that read data structures and retrieve
status. Commands that can be used to alter BMC configuration, write data to the BMC or other management
controllers, or perform system actions such as resets, power on/off, and watchdog activation are disallowed.
Appendix G - Command Assignments, provides a list of the commands that require operating at User level or
higher.
All BMC commands are allowed, except for configuration commands that can change the behavior of the out-ofband interfaces. For example, Operator privilege does not allow the capability to disable individual channels, or
change user access privileges.
Appendix G - Command Assignments, provides a list of the commands that require operating at Operator level or
higher.
All BMC commands are allowed, including configuration commands. An Adminstrator can even execute
configuration commands that would disable the channel that the Administrator is communicating over.
Appendix G - Command Assignments, provides a list of the commands that require operating at Administrator
Intelligent Platform Management Interface Specification
level.
6.9 Users & Password Support
The term user is used in this specification to refer to a collection of data that identifies a password (key) for
establishing an authenticated session, and the privileges associated with that password. For configuration
purposes, the sets of user information are organized and accessed according to a numeric User ID. When
activating a session, user information is looked up using a text username.
User access can be enabled on a per channel basis. Thus, different channels can have different sets of users
enabled.
If desired, a username on one channel can be associated with a different password than the same username on a
different channel. When a session is activated, the BMC will scan the usernames sequentially starting with
User ID 1 and will look for the first user that has a matching username and has access enabled for the given
channel. Thus, having different passwords for a given username requires configuring multiple user entries - one
for each different password that is to be used for a particular set of channels.
The specification allows a number of different implementations for supporting users on a channel. The following
lists the minimum requirements:
•
All authenticated channels are required to support at least one user (User ID 1).
•
Usernames may be fixed or configurable, or a combination of both, at the choice of the implementation.
•
If an implementation supports only one user with a fixed user name, then the fixed user name must be null
(all zeros).
•
Support for configuring user passwords for all User IDs is required.
•
Support for setting per-user privilege limits is optional. If the Set User Access command is not supported, the
privilege limits for the channel are used for all users.
6.9.1
‘Anonymous Login’ Convention
The IPMI convention for enabling an ‘anonymous’ login is to configure the entry for User ID 1 with a null
username (all zero’s) and a null password (all zero’s). Applications may then present this to the user as an
anonymous login and configuration option, knowing what username and password to use if the BMC allows
‘anonymous’ logins. The reason for doing this via User ID 1 is to simplify the task of enabling the BMC to
report whether anonymous login is enabled or not.
6.9.2
Anonymous Login Status
The Get Channel Authentication Capabilities command includes a ‘Anonymous Login Status’ field. This field
indicates to a remote console application whether User ID 1 is presently configured with a null username and
null password. In addition, a bit is provided that indicates whether there are also non-null usernames enabled
for the channel, or whether User ID 1 holds a null username, but a non-null password.
Together, these bits can be used to guide a remote application in presenting connection options to a user. For
example, if a system only has Anonymous Login enabled, the application could immediately connect without
prompting the user, or use that information to enable an ‘anonymous login’ button in the user interface. If a
system has a null username but non-null password, the application could put up just a password dialog box.
Lastly, if the system indicates it has non-null usernames with non-null passwords, the application may put up a
dialog box prompting for both a user name and password.
47
Intelligent Platform Management Interface Specification
6.10 System Interface Messaging
The following sections describe how messaging works across the system interface to the BMC. Later sections go
into detail on message formats and register interfaces for the different physical implementation options for the
system interface.
6.10.1 BMC Channels and Receive Message Queue
Messaging between system software and the other management busses, such as the IPMB, is accomplished
using channels and a Receive Message Queue. A channel is a path through the BMC that allows messages to be
sent between the system interface and a given bus or message transport. The Receive Message Queue is used to
hold message data for system software until system software can collect it. All channels share the Receive
Message Queue for transferring messages to system management software. The Receive Message Queue data
contains channel, session, and IPMI addressing information that allows system software to identify the source
of the message, and to format a message back to the source if necessary.
System management software is responsible for emptying the Receive Message Queue whenever it has data in
it. Messages are rejected if the Receive Message Queue gets full. It is recommended that the Receive Message
Queue have at least two ‘slots’ for each channel. The Receive Message Queue is a logical concept. An
implementation may choose to implement it as an actual queue, or could implement separate internal buffers for
each channel. It is recommended that the implementation attempt to leave a slot open for each channel that does
not presently have a message in the queue. This helps prevent ‘lock out’ by having the queue fill with just
messages from one interface.
The BMC itself can, if necessary, use the Receive Message Queue and Messaging Channels to send
asynchronous messages to system management software. The recommended mechanism for accomplishing this
is to define a unique channel with a protocol type of ‘System’. To send an asynchronous message to system
software, the BMC would place a message from that channel directly into the Receive Message Queue in
‘System’ format. System software would be able to respond back to the BMC using a Send Message command
for that channel.
6.10.2 Event Message Buffer
[Optional] - The Event Message Buffer holds Event Request Messages that have been internally generated by
the BMC, and Event Messages that have been received by the BMC from the IPMB or other channel. The
Event Message Buffer does not hold event messages that have been generated from system software.
The Event Message Buffer holds all 16-bytes of the Event Message as it would be stored in the System Event
Log (see Table 26-1, SEL Event Records). For IPMI v1.5, the Event Message Buffer does not get overwritten if
a new event comes in before system software can empty the buffer. The BMC does clear the buffer when the
BMC is first powered up and whenever the system becomes powered up or is hard reset. A BMC
implementation can support generating a system interrupt when the Event Message Buffer gets filled.
Some implementations will elect to generate an SMI to allow the creation of an SMI Handler that takes
additional actions on Event Messages. If Event Message Buffer interrupts do not generate SMI, or are not
enabled (or not implemented), SMS can use this as a mechanism for examining Event Messages as they are
received. System software must check the status of SMI use before assuming that the Event Message resource
is available. This can be accomplished by using the Get Channel Info command to determine if the interrupt
assignment for the Event Message Buffer is set to SMI.
Note: SMM Messaging and the implementation of SMIs is OPTIONAL. Since SMI operation and functions
are proprietary and not described nor required in this specification, support via the IPMI interfaces is
being deprecated. New implementations should avoid using the IPMI support for SMI.
48
Intelligent Platform Management Interface Specification
6.10a System Interface Discovery and Multiple Interfaces
A BMC device may make available multiple system interfaces, but only one management controller is allowed to
be the ‘active’ BMC that provides BMC functionality for the system (in the case of a ‘partitioned’ system, there
can only one active BMC per partition). Only the system interface(s) for the active BMC are allowed to respond
to the Get Device ID command. If other BMC devices are present, but not being used, they must not respond to
the Get Device ID command.
When system interfaces are available, the driver can select the type interface it wishes to use.
Drivers should not switch system interfaces during system operation or else unexpected results could occur. The
Get Device ID command is required to execute correctly across multiple interfaces to a BMC, but other
commands are not. Once the driver has chosen to use a given interface, all commands beyond Get Device ID
should be delivered to that interface. If it is desired to change the choice of system interfaces, a warm or cold
reset of the platform should be done to ensure that the system can re-initialize BMC operation.
It is recommended that run-time drivers support the IPMI System Interfaces in the following order:
•
A driver should preferentially use the BMC on PCI, via the OS’s native support, if available. (A “Plug and
Play” OS will typically locate and load the appropriate driver for devices it finds on PCI.) Appendix C2 Locating IPMI System Interfaces on PCIAppendix C2 - Locating IPMI System Interfaces on PCI,
summarizes the PCI Class codes for IPMI System Interfaces.
•
If the desired interface is not available on PCI, or the system is in a state where OS support for PCI is
unavailable the next step should be to look for the system interface as a static resource described in ACPI
using the control methods described in Appendix C3, Locating IPMI System Interfaces with ACPI.
•
If the operating environment does not include a mechanism to support executing ACPI control methods,
then look for the system interface at the location described by the SPMI (Service Processor Management
Interface) Table(s) through the ACPI Description Table mechanisms. (The SPMI Table approach supports
BMCs that offer more than one system interface. Therefore, there can be more than one instance of the
SPMI Table.) The SPMI Table is described in Appendix C2 - Locating IPMI System Interfaces on
PCIAppendix C2 - Locating IPMI System Interfaces on PCI.
•
If the SPMI Table is not present, the driver should look for the SMBIOS Type 38 table (See Appendix C1 Locating IPMI System Interfaces via SM BIOS TablesAppendix C1 - Locating IPMI System Interfaces via
SM BIOS Tables) and use the interface described there. Unlike the SPMI Table, there is only one instance
of the Type 38 record allowed, so the driver will not need to look for additional interfaces.
•
Lastly, the driver should look for the IPMI System Interface at the fixed, default I/O addresses specified for
the SMIC, KCS, and BT interfaces. Refer to the individual sections on those interfaces for the addresses.
6.11 IPMI Sessions
Authenticated IPMI communication to the BMC is accomplished by establishing a session. Once established, a
session is identified by a Session ID. The Session ID may be thought of as a handle that identifies a connection
between a given remote user and the BMC, using either the LAN or Serial/Modem connection.
The specification supports having multiple active sessions established with the BMC. It is recommended that a
BMC implementation support at least four simultaneous sessions. This number is shared between the LAN and
Serial/Modem interfaces.
The specification also allows a given endpoint (identified by an IP address) on the LAN to open more than one
session to a BMC. The capability is allowed to allow a single system to serve as a proxy to provide BMC LAN
sessions for other systems. It is not intended for one system to use this provision to open multiple sessions to the
BMC for that system’s sole use.
49
Intelligent Platform Management Interface Specification
An IPMI messaging connection to the BMC fits one of three classifications, session-less, single-session, or multisession.
6.11.1 Session-less Connections
A session-less connection is unauthenticated. There is no ‘user login’ required for performing IPMI messaging.
The System Interface and IPMB are examples of session-less connections.
A special case of a session-less connection can occur over an interface that supports session-based messaging.
Session-based connections have certain commands that are accepted and responded to “outside of a session”.
When that occurs, the channel is effectively operating in a session-less manner for those commands.
Commands that are handled outside of a session have fixed values for session-specific fields in the message.
For example, when the “Get Channel Authentication Capabilities” is sent over a LAN channel outside of a
session, it is sent with the session ID set to NULL and authentication type set to NONE in the IPMI session
header of the message. Note that commands that are accepted “outside of a session” can also be accepted
within the context of a session, in which case they must have valid session IDs, etc., in the session header in
order to be accepted.
6.11.2 Single-session Connections
A single-session connection has a user authentication phase that precedes IPMI messaging. This is
accomplished using the Get Session Challenge and Activate Session commands. A single-session connection is
intended for a physically secure link. Therefore, individual packets are not signed. The serial/modem Basic
Mode is an example of a single-session connection.
6.11.3 Multi-session Connections
A multi-session connection has user authentication and supports multiple interleaved sessions (multiple users).
The multi-session connection is specified to support communication on a shared medium, such as LAN, where
there may be a mix of IPMI and non-IPMI traffic. In order to support multiple sessions, and protect against
attempts to circumvent authentication (such as replay attacks), multi-session packets have a session header in
addition to the IPMI message. The session header carries information to identify the particular session, as well
as other fields such as session sequence numbers and authentication type fields. The LAN and PPP Mode
connections are examples of multi-session IPMI messaging connections.
6.11.4 Per-Message and User Level Authentication Disables
Typically, each packet in a multi-session connection is authenticated (with the exception of the packets for
certain ‘pre-session’ commands such as Get Channel Authentication Capabilities, and Get Session Challenge.)
In some cases however, the connection medium is considered to be trusted even though multiple user sessions
are allowed. Once a session has been activated, the computational overhead of authenticating each packet may
not be necessary.
Thus, there are two options to enable performance improvements in environments where the link is considered
to be secure. The options are to disable ‘Per-Message Authentication’, and to disable ‘User Level
Authentication’. If Per-Message Authentication is disabled, the only packets that are required to be
authenticated are the ones for the Activate Session request and response. Once the session is activated, the
remaining packets will be accepted with the Authentication Type set to NONE. Since the Authentication Code
(signature) is not provided in the packet when the Authentication Type is NONE, this enables a performance
improvement in two ways: fewer bytes are transmitted, and the authentication algorithm doesn’t need to be run.
In many cases, there is little concern about whether User Level commands are authenticated, since the User
privilege allows status to be retrieved, but cannot be used to cause actions such as platform resets, or change
platform configuration. Thus, an option is provided to disable authentication just for User Level commands. If
50
Intelligent Platform Management Interface Specification
User Level Authentication is disabled, then User Level messages will be accepted that have the Authentication
Type set to NONE.
The BMC will always verify any authenticated packets (Authentication Type not NONE) that it receives,
regardless of whether Per-Message Authentication and/or User Level Authentication is disabled. Authenticated
packets will be silently discarded if the signature (AuthCode) is invalid, or the Authentication Type does not
match the authentication type that was negotiated when the session was activated. This is necessary to allow
remote console software to deliver authenticated messages to the Receive Message Queue via the Send
Message command.
Both the Per-Message Authentication and User Level Authentication disable options are configured via the Set
Channel Access command.
6.11.5 Link Authentication
Sometimes connections offer authentication protocols that are applied as part of establishing the
communication link to the BMC. For example, PPP supports authentication protocols such as PAP and CHAP
that are part of link establishment.
Link Authentication is a global characteristic associated with the connection mode for the channel. Link
Authentication is enabled/disabled via the serial/modem configuration parameters. When Link Authentication
is enabled, it is necessary to identify one or more users that will serve as the source of the username (peer ID)
and password information for the link. This is accomplished by setting an ‘Enable User for Link
Authentication’ bit in the Set Channel Access command.
For physically secure connections, these ‘Link Authentication’ protocols may be all that’s considered needed to
authenticate the user. Thus, the BMC supports enabling Link Authentication for PPP using common PPP
authentication algorithms. If Link Authentication is enabled, the Per-Message Authentication Disable, and User
Level Authentication Disable options may be used to improve performance.
6.11.6 Summary of Connection Characteristics
The following table summarizes the key characteristics that differentiate session-less, single-session, and multisession connections:
X
X
X
X
Link
Authentication
User Level
Authentication
Disable
LAN
X
X
X
Serial/Modem:
PPP Mode
X
X
X
Basic Mode
X
X
[1]
Terminal Mode
X
X
IPMB
X
ICMB
X
PCI Management Bus
X
1. Terminal mode only supports ‘straight password’ authentication
Per Message
Authentication
Disable
Authenticated
Access
Session
Header
Session-less
SingleSession
MultiSession
Table 6-6, Session-less , Single-session and Multi-session Characteristics
X
51
Intelligent Platform Management Interface Specification
6.11.7 Session Activation and IPMI Challenge-Response
A session must be activated before general IPMI messaging can occur. The mechanism for accomplishing this
is via a set of IPMI commands that are used to perform an “IPMI Challenge-Response”. The session activation
process involves three IPMI commands: Get Channel Authentication Capabilities, Get Session Challenge, and
Activate Session. Of these three commands, the Get Channel Authentication Capabilities and Get Session
Challenge command must be executable before the session is set up. Therefore, these commands can be thought
of as always being ‘unauthenticated’. The Activate Session command is the first, and in some cases only,
authenticated command for a session.
Figure 6-1, Session Activation
Remote Console
Get Channel
Authentication
Capabilities, Rq
Get Session
Challenge, Rq
Activate Session, Rq
Managed System
1
2
Get Channel
Authentication
Capabilities, Rs
4
Get Session
Challenge, Rs
6
Activate Session, Rs
3
5
Referring to Figure 6-1, Session Activation, the following presents the general steps for activating a session:
52
1.
The Remote Console issues a Get Channel Authentication Capabilities request to the BMC.
2.
The BMC returns a Get Channel Authentication Capabilities response that contains which
authentication types (authentication algorithms) it supports.
3.
The Remote Console sends a Get Session Challenge request to the BMC. The command selects which
of the BMC-supported authentication types the Remote Console would like to use, and a username that
selects which set of user information should be used for the session. This is the only place where the
username is used during the process.
4.
The BMC looks up the user information associated with the username. If the user is found and allowed
access via the given channel, the BMC returns a Get Session Challenge response that includes a
randomly generated Challenge String and a temporary Session ID. The BMC keeps track of the
username associated with the Session ID so that it can use the Session ID to look up the user’s
information in the next step. In some algorithms, the BMC will store challenge string, or a seed that
was used to generate the challenge, for later lookup as well.
5.
The Remote Console then issues an Activate Session request. The request contains the temporary
Session ID plus the authentication information based on the type of authentication that was selected.
For example, a LAN packet would typically include a signature using an authentication algorithm run
Intelligent Platform Management Interface Specification
on elements such as the challenge string, user password/key, IPMI message fields, Session ID, etc.
while a serial/modem connection may only pass a simple clear-text password in the activate session
data. The authentication format for different authentication types is specified in the description of the
Activate Session command. For multi-session connections, the starting Outbound session sequence that
the BMC is to use when sending packets to the remote console is also passed in the request. (Session
sequence numbers are explained in the next section.)
6.
The BMC uses the temporary Session ID to look up the information for the user that was identified in
the Get Session Challenge request. The BMC looks up the user’s password/key data, and potentially
other data such as a stored copy of the earlier challenge string, and uses it to verify that the packet
signature or password is correct. If so, the BMC issues an Activate Session response that provides the
Session ID to use for the remainder of the session. For multi-session connections, the Activate Session
response is itself authenticated (signed). The BMC will also deliver the starting Inbound session
sequence that the Remote Console is to use when sending packets to the BMC.
From this point, whether individual packets for the session are authenticated or not is based on settings such as the
Per User Authentication and User Level Authentication parameters. Refer back to 6.11.4, Per-Message and User
Level Authentication DisablesPer-Message and User Level Authentication Disables.
6.11.8 Session Sequence Numbers
The session sequence number is a 32-bit, unsigned, value. The session sequence number is not used for
matching IPMI requests and responses. The IPMI Sequence (Seq) field is used for that purpose. The session
sequence number is solely for protecting against replay attacks.
There are two Session Sequence Numbers: the Inbound Session Sequence Number and the Outbound Session
Sequence Numbers. The inbound and outbound directions are defined with respect to the BMC. Inbound
messages are from the remote console to the BMC, while outbound messages are from the BMC to the remote
console.
Inbound messages use the inbound session sequence number, while outbound messages use the outbound
session sequence number.
6.11.9 Session Sequence Number Generation
Session sequence numbers are generated on a per-session basis. The inbound and outbound sequence numbers
are updated and tracked independently. The BMC and the remote console independently select the starting
session sequence number for the messages they receive.
The remote console sets the starting values for the outbound session sequence number when it sends the first
Activate Session command for an authenticated session.
The Activate Session response is the first authenticated outbound (BMC to remote console) message. This
response message uses the initial outbound session sequence number value that the remote console delivered in
the prior Activate Session command request. The BMC must increment the outbound session sequence number
by one (1) for each subsequent outbound message from the BMC.
The first authenticated inbound message uses the initial inbound (remote console to BMC) session sequence
number value that was returned by the BMC in the response to the Activate Session command. The remote
console must increment the inbound session sequence number by one (1) for each subsequent message it sends
to the BMC.
6.11.10 Inbound Session Sequence Number Tracking and Handling
Session sequence numbers are tracked on a per-session basis. At a minimum, the BMC is required to track that
the inbound sequence number is increasing, and to silently discard the packet if the sequence number is eight
53
Intelligent Platform Management Interface Specification
counts or more from than the last value received. (An implementation is allowed to contain a proprietary
configuration option that enables a larger sequence number difference, as long as the standard of +eight can be
restored.)
An implementation can elect to terminate the session if it receives a number of sequence numbers that are more
than eight counts from the last value received.
Valid packets (packets with good data integrity checks and signature) to a given session that have the same
inbound sequence number as an earlier packet are considered to be duplicate packets and are silently discarded
(even if the message content is different).
6.11.11 Out-of-order Packet Handling
In order to avoid closing a session because a packet was received out-of-order, the BMC must implement one
of two options:
Option 1: Advancing eight-count (or greater) window. Recommended. Track which packets have been
received that have sequence numbers up to eight counts less than the highest last received sequence
number, tracking which of the prior eight sequence numbers have been received. Also accept packet with
sequence numbers that are up to eight counts greater than the last received sequence number, and set that
number as the new value for the highest sequence number received. This option is illustrated in Appendix A
- Previous Sequence Number TrackingAppendix A - Previous Sequence Number Tracking.
Option 2: Drop any packets with sequence numbers that are lower than the last valid value received. While
simpler than option 1, this option is not recommended except for resource-constrained implementations
due to the fact that any out-of-order packets will require the remote console to timeout and retransmit.
Sequence number wrap-around must be taken into account for both options. When a sequence number
advances from FFFF_FFFFh to 0000_0000h, the value FFFF_FFFFh represents the lesser sequence number.
6.11.12 Outbound Session Sequence Number Tracking and Handling
The remote console is required to handle outbound session sequence number tracking in the same manner as the
BMC handles the inbound session sequence number, except that Option 2 (above) should not be used as a
means of handling out-of-order packets.
6.11.13 Session Inactivity Timeouts
A session is automatically closed if no new, valid, message has been received for the session within the
specified interval since the last message. The session must be re-authenticated to be restored. A remote console
can optionally use the Activate Session command to keep a session open during periods of inactivity.
Note that only an active session will keep the Session Inactivity Timeout from expiring. IPMI message activity
that occurs outside of an active session has no effect. This is to prevent someone from keeping a phone
connection indefinitely while trying to guess different passwords to activate a session.
The BMC only monitors for inactivity while the connection is switched over to the BMC. Note that closing a
session is not always the same as hanging up a modem connection. Serial/modem sessions are also
automatically closed when the connection is switched over to the system, but the phone connection remains
active. The BMC only terminates the phone connection if a session is closed due to an inactivity timeout while
the serial connection is routed to the BMC.
The timeout and tolerance values are specified for the management controller (BMC) that will timeout and
close the session. System software should take this tolerance into account, plus any additional delays due to
media transmission times, etc.
54
Intelligent Platform Management Interface Specification
An implementation can provide an option to allow the timeout to be configurable via a parameter in the
configuration parameters for the given channel type.
Table 6-7, Default Session Inactivity Timeout Intervals
Session Type
LAN
Direct Connect Mode Serial
Modem Mode Serial
6.11.13.1
Default
Expiration
Interval
Tolerance
60 seconds
60 seconds
60 seconds
+/- 3 seconds
+/- 3 seconds
+/- 3 seconds
Notes
The Inactivity Timeout Interval starts
whenever a connection is established with, or
switched to, the BMC. The Phone connection
gets terminated if inactivity timeout occurs
while serial connection is routed to BMC.
Avoiding ‘Slot Stealing’
It is highly recommended that an implementation provide a mechanism for protecting against someone
accidentally or maliciously 'claiming' all the session slots and subsequently locking out access to the BMC. For
example, this could occur by an errant program repeatedly issuing Get Session Challenge commands without
successfully activating a session - causing all available resources for tracking pending sessions to be used up.
One possible solution is to use an LRU algorithm that drops the session ID for the oldest session ID that has a
pending 'Activate Session'. That way, the only way to ‘permanently’ use up slots is by activating and
maintaining sessions for all session slots. A minor refinement may be to provide a few seconds of delay on
returning the response to the Get Session Challenge in order to give opportunity for a well-behaved application
to get a Challenge and return an Activate Session command before the errant software re-issued another Get
Session Challenge. (This is only an improvement for errant applications that wait for the response to the Get
Session Challenge before issuing the request again.)
6.11.14 Additional Session Specifications and Characteristics
•
At least four simultaneous sessions should be supported on a given channel.
•
By default, sessions are automatically closed if no valid activity is detected within the Session Inactivity
Timeout Interval, or if the connection or link is terminated. Valid activity is defined as the receipt of a valid
IPMI message for that session while the connection is routed to the BMC.
•
At least four pending bridged requests should be supported for each bridged interface that requires the BMC
to track pending responses. See 6.12, BMC Message BridgingBMC Message Bridging for more information.
•
The typical BMC is expected to allow only a small number of simultaneous open sessions (on the order of
four to eight). Thus, remote console applications should avoid activating multiple sessions whenever
possible, in order to allow other remote consoles to also get access.
•
The Activate Session command will return a completion code indicating whether the request was rejected
because BMC is presently busy with other open sessions.
•
The specification allows multiple sessions to be activated from a single IP address. The primary reason for
allowing multiple sessions is to allow a system to serve as a proxy agent that provides BMC access for
remote consoles that connect to it instead of directly to the BMC.
•
Multiple sessions are not intended to be used to support access by multiple applications behind an IP
address. If multiple applications require access to the BMC on a given system, a single driver or ‘middleware’ should coordinate that access and use a single session, if possible. The IPMI Software ID and the
IPMI sequence number are two fields that a shared driver can used to identify and route messages to and
from a given application.
55
Intelligent Platform Management Interface Specification
•
There is a 1:1 relationship between a user name and a session. I.e. different usernames cannot share the
same session. However, multiple sessions can be activated using the same username.
•
All sessions start off at User Level privilege. It is necessary to issue a Set Session Privilege to raise the
operating privilege level before commands that required higher privilege can be executed. The maximum
operating privilege for a session is determined by Privilege Limits that are set both for the user and for the
overall channel. The more restrictive setting of the User Privilege Limit and the Channel Privilege Limit
sets the maximum operating privilege available for a session.
•
An Operator can optionally use the Get Channel Info and Get Session Info commands to retrieve the address
of parties with open sessions and their present privilege level. This is to allow a remote console to
coordinate with another remote console that already has an active session. This can be used to allow
software to coordinate access to the system. For example, if management software running at Console “A”
wished to remotely reset a given system, it could first see whether another console had an active session
with the system to be reset. It could then use information from the Get Channel Info and Get Session Info
commands to send a message directly to the other console, notifying it of the pending reset.one system
•
An Administrator can force sessions on any channel to be terminated.
6.12 BMC Message Bridging
BMC Message Bridging provides a mechanism for routing IPMI Messages between different media. Bridging is
only specified for delivering messages between different channels; i.e. it is not specified for delivering messages
between two sessions on the same channel.
In IPMI 1.0, bridging was primarily specified just for providing access between SMS (System Interface) and the
IPMB. With IPMI 1.5, these mechanisms have been extended to support delivering IPMI messages between
active connections / sessions on any IPMI Messaging media connected to the BMC.
There are three mechanisms for bridging messages between different media connected to the BMC, depending on
what the target of the message is:
•
BMC LUN 10b is used for delivering messages to the System Interface. The BMC automatically routes any
messages it receives via LUN 10b to the Receive Message Queue.
•
Send Message command from System Interface is used for delivering messages to other channels, such as
the IPMB. The messages appear on the channel as if they’ve come from BMC LUN 10b. Thus, if the
message is a request message, the response will go to BMC LUN 10b and the BMC will automatically place
the response into the Receive Message Queue for retrieval. System software is responsible for matching the
response up with the original request, thus the ‘No Tracking’ setting in the Send Message command is used.
•
Send Message command with response tracking. This format of Send Message command is used with
response tracking for bridging request messages to all other channels except when the System Interface is the
source or destination of the message.
The following sections provide additional information on the operation and use of these bridging mechanisms.
56
Intelligent Platform Management Interface Specification
6.12.1 BMC LUN 10b Routing
Because messages to SMS are always routed to the Receive Message Queue, the Send Message command is not
typically used to send messages to SMS. Instead, messages to SMS are delivered via the BMC SMS LUN, 10b.
The BMC automatically reformats and places any messages that are addressed to LUN 10b into the Receive
Message Queue for SMS to retrieve using the Get Message command.
Thus, sending a request to SMS just requires formatting the command so that it is addressed to BMC LUN 10b.
SMS can retrieve the request from the Receive Message Queue, extract the originator’s address and channel
info, and then use the Send Message command to deliver a response.
The BMC does not track requests and responses for messages to system software because the Receive Message
Queue provides the channel and session information necessary to format the Send Message command to deliver
the response. Similarly, system software is capable of tracking the channel and session information it used
when generating a request. Thus, the ‘No Tracking’ option is used for Send Message commands from system
software.
The responder then delivers its response message to BMC LUN 10b and the response gets routed to the Receive
Message Queue. Conversely, if a channel wants to deliver a message to SMS, it sends the request message to
BMC LUN 10b, and later SMS uses a Send Message command to return the response from BMC LUN 10b.
6.12.2 Send Message Command From System Interface
The operation of the Send Message command when issued via the System Interface is different than when the
Send Message command is issued from other interfaces. This is because the IPMI System Interfaces were
specified as always returning an immediate command response. In order to avoid tying up the System Interface
waiting for a bridged response, a response to the Send Message command is returned as soon as the request is
bridged to the target channel. This response only indicates that the Send Message command was executed. It is
not the response to the bridged request.
Later, the response to the bridged request is received by the BMC and routed into the Receive Message Queue
and it is retrieved using a Get Message command. For example, here are the typical steps involved in delivering
a request from the System Interface to a device on the IPMB, and receiving a corresponding response:
1.
System software formats a Send Message request that encapsulates information for the request to be
placed on IPMB. The requester’s LUN in the data is set to 10b so when the response comes back, the
BMC will place it in the Receive Message Queue. The encapsulated request is also given a sequence
number by the system software. System software will use this number later, along with other fields, to
match up the Receive Message Queue data with the original request.
2.
System software delivers the Send Message request to the BMC via the System Interface.
3.
The BMC returns an ‘OK’ response to the Send Message command, indicating that it has received the
request and delivered it to the IPMB.
4.
Sometime later, the target IPMB device delivers a response to the request. The response is sent back to
the same requester’s LUN that was used in the request, 10b. The BMC routes message data received
on 10b to the receive message queue, and also tags it with information such as the channel number that
the message was received from.
5.
System software detects that there is data in the Receive Message Queue. This is either done by
polling for messages by periodically checking the SMS_ATN bit, or for interrupt driven
implementations, getting an interrupt when SMS_ATN becomes set. Software then uses the Get
Message Flags command to discern whether the SMS_ATN condition was from getting data into the
Receive Message Queue or some other event.
6.
System software then issues a Get Message request. The response returns a message from the queue. If
the data is for a response, software then checks the message fields, such as sequence number, channel
57
Intelligent Platform Management Interface Specification
number, CMD, etc., to see if the response matches up with an earlier request. In this example, software
would be looking for a response to the request it had bridged onto IPMB. If the Receive Message
Queue holds a request for system software, it processes it accordingly.
7.
If software has not received a response by the timeout intervals specified for IPMB, it can retry the
request. Also note that IPMB sequence numbers generally expire after 5 Seconds. This number comes
from the sequence number expiration interval on IPMB. Software can generally discard requests that
are more than 5 seconds old and re-use their sequence numbers.
If the target channel uses sessions, the Send Message command data will require a Session Handle value to
select which session on the channel the message will be sent to. Software can use the Get Channel Info and Get
Channel Sessions commands to determine what channels are present and to obtain the Session Handle for a
given session.
6.12.3 Send Message Command with Response Tracking
The Send Message command is used primarily to direct the BMC to act as a proxy that translates a message
from one IPMI messaging protocol to another. The BMC formats the data for the target channel type and
protocol and delivers it to the selected medium.
Media such as the IPMB do not include channel number and session information as part of their addressing
information. As a result, request messages from another channel must be delivered as if they originated from
the BMC itself.
If the bridged message is a request, it is necessary for the BMC to hold onto certain data, such as originating
channel and session information, so that when the response message comes back it can reformat the response
and forward it back to the originator of the request. The primary way the BMC accomplishes this is by
assigning a unique sequence number to each request that it generates, and saving a set of information in a
‘Pending Bridged Response’ table that is later used to reformat and route a response back to the originator of
the request.
The sequence number returned in the response is then used to look up who generated the original response, plus
the saved formatting and addressing information. The BMC then reformats and delivers the response to the
original requester and deletes the request from its list of ‘pending responses’. The Send Message command
includes a parameter that directs the BMC to save translation information for and track outstanding request
messages for the purpose of routing the response back to the originator of the Send Message command.
Note that, with the exception of messages to SMS, when the Send Message command is used to deliver a
message to a given medium the message appears to have been originated by the BMC. This means that a
controller on the IPMB can’t generically distinguish a bridged request from SMS from a bridged request from
LAN.
Table 6-8, Message Bridging Mechanism by Source and Destination
Message Type and direction
Request or Response from System Interface to any other channel
Request or Response to System Interface from any other channel
Request from any channel except System Interface to IPMB
Response from IPMB to any channel except System Interface
Request from any channel (except System Interface) to PCI
Management Bus
Request from PCI Management bus to any channel except System
Interface
Request from Serial to LAN
Response LAN to Serial
58
Delivery
Mechanism
Send Message
BMC LUN 10b
Send Message
BMC LUN 00b
Send Message
BMC tracks
pending
responses
no
no
yes
yes
yes
BMC LUN 00b
yes
Send Message
BMC LUN 00b
yes
yes
Intelligent Platform Management Interface Specification
Request from LAN to Serial
Response from Serial to LAN
Send Message
BMC LUN 00b
yes
yes
6.12.4 Bridged Request Example
The example illustrates a Send Message command from LAN being used to deliver a request to IPMB.
Bridged requests to the IPMB can come from several different channels: LAN, serial/modem, and the ICMB.
The BMC uses the sequence number that it places on the bridged request to identify which channel and to
which address on that channel the response is to go back to. It is therefore important for the BMC to ensure that
unique sequence numbers are used for pending requests from the different channels. It is also important that
sequence numbers are unique for successive requests to a given responder. One way to manage sequence
numbers to the IPMB is to track sequence numbers on a per responder basis. This can be kept in a table of
‘Pending Bridged Response’ info.
In order to get the response back to the LAN, the IPMB response must return the same sequence number that
was passed in the request. (This is a basic rule of IPMI Messaging, so there’s nothing special about that
requirement.) The management controller uses the sequence number to look up the channel type specific
addressing, sequence number, and security information that it stored when the request was forwarded. For
example, if the channel type is ‘LAN’ then the response message must be formatted up in an RMCP/UDP
packet with the IP address of the requester, the sequence number passed in the original request, the appropriate
security ‘key’ information, etc.
When a request message is bridged to another channel by encapsulating it in a Send Message command (from a
source channel other than the system interface), the BMC immediately returns a response to the Send Message
command itself. Meanwhile, the request is extracted from the Send Message command and forwarded to the
specified target channel.
The Send Message command must be configured to direct the BMC to keep track of data in the request so when
the response comes back from the target device it can be forwarded by the BMC back to the channel that
delivered the original Send Message command to the BMC. When the response comes back from the target, the
BMC uses the tracking information to format the response for the given channel. To the party that initiated the
Send Message command, the response will appear as if the encapsulated request was directly executed by the
BMC. I.e. it will look like an asynchronously generated response message.
For example, suppose a Get Device ID command has been encapsulated in a Send Message command directed
to the IPMB from a LAN channel. The BMC will immediately send a response to the Send Message command
back on LAN. The BMC will extract the encapsulated Get Device ID message content and format it as a Get
Device ID request for IPMB. The target device on IPMB responds with a Get Device ID response message in
IPMB format. The BMC takes the tracking information that was stored when the Send Message command was
issued, and uses it to create a Get Device ID response in LAN format. The Responder’s address information in
that response will be that of the BMC, not of the device on IPMB that the request was targeted to.
The following figure and steps present an example high-level design for handling a bridged request. Note that
the example shows information that is generated and stored, but it does not show any particular code module
that would perform that operation. That is, the choice of which functions are centralized, which are in a ‘LAN’
module, and which are in an ‘IPMB’ module (or whether you even have such modularity) is left to the
implementer.
59
Intelligent Platform Management Interface Specification
Figure 6-2, LAN to IPMB Bridged Request Example
LAN REQUEST
Source
IP / MAC
Address
Session
ID
(0047h)
Encapsulated data for IPMB Request
Responder's
Address
(RsA) [BMC]
NetFn/
RsLUN
Requester's
Address
(RqA)
Chk1
A
CMD =
Send
Message
RqSeq/
RqLUN
Channel
=0
(IPMB)
...
2
3
Responder's
Address (RsA)
e.g. 24h
NetFn/RsLUN
e.g. S/E, 00b
Chk1
B
Requester's
Address
(RqA)
RqSeq/
RqLUN
CMD =
Get Sensor
Reading
Sensor
Number
Chk2
B
Chk2
A
1
Src IP
& MAC
Addr
0047h
4
2
3
Ch 0
(IPMB)
Seq #s
Ch1
Seq #s
Destination
Destination Source Session Requester's Destination Channel Destination Source
Address / Channel Responder's Channel Channel
Channel Channel ID
SWID
Number
RsLUN
RqA
Address
RqSeq# Number Handle
Seq #
Allocator
0017h
Ch 1
3
AAh
Ch 0
24h
00b
(RqA)
Source
Channel
RqSeq/
RqLUN
RqSeq/
RqLUN
Source
Channel
CMD
Get
Sensor
Reading
Seq #
Expiration
500
4
Internally Stored Info for tracking
and formatting response back to
requester
5
IPMB REQUEST
Responder's
Slave Addr.
(RsSA)
=24h
* BMC S
60
NetFn/
RsLUN
=S/E,
00b
th
i
Chk1*
Requester's
Slave Addr
(RqSA)
=20h*
(BMC)
RqSeq/
RqLUN
LUN=
17h*/ 00b*
CMD =
Get Sensor
Reading
Sensor
Number
Chk2*
d Fi ld
1.
When the BMC receives the Send Message command with the ‘Bridged Request’ parameter bit set, it checks
for an available entry in a Pending Bridged Response table and copies parameters from the request to be
bridged. When the response comes back, these parameters will be used to validate that the response matches
the earlier request and to reformat the response for the originating channel. The bold outlined boxes
represent parameters and data in the Send Message command that will ultimately be copied to the resulting
request on the target channel (the IPMB in this example).
2.
Any channel session information necessary to get the response back to the original requester will also need
to be recorded. In this example, the BMC maintains a separate table of session information for the LAN
channel. An offset into that table is used as a ‘handle’ for identifying the session information associated with
the request. This handle is used in the Pending Bridged Response table in lieu of copying all the session
information. Note that with such an implementation, it is important to remember details such as invalidating
and freeing any bridge table entry associated with that session if the session should get closed while
responses are pending.
3.
In this example, the BMC has a separate ‘Sequence Number Allocator’ routine that ensures that sequence
numbers used in bridged requests are kept unique for a given channel. This is done so when the response
comes back, the sequence number can be used to look up the corresponding request info entries from the
Pending Bridged Response table.
4.
Responses have a five second ‘sequence number expiration’ interval. If a response is not received by the
expiration interval, the corresponding entry in the Pending Bridged Response entry is deleted and the
sequence number associated with the request can be reused. The Seq # Expiration column in this example
represents a possible implementation where the Seq # Expiration value is decremented nominally once every
10 ms. The entry is considered to be free when the number hits 0. Thus, in this example the Seq # Expiration
field could be used both for tracking sequence number expiration as well as a mechanism for marking
whether a table entry is available or not.
Intelligent Platform Management Interface Specification
5.
The BMC takes the indicated values and uses them to construct the bridged request. The request is a
combination of field values copied from the original Send Message command and values generated by the
BMC. The BMC generated values are shown with a bold underlined typeface with an asterisk.
6.13 Message Size & Private Bus Transaction Size Requirements
The following table summarizes the message size and transaction size requirements for the various interfaces to
the BMC and IPMI Management controllers. The IPMI message sizes include any IPMI-level addressing and data
integrity information required for the interface. For example, the IPMB Message lengths include the requester and
responder addressing information, sequence number, and checksums. The message sizes do not include counts for
additional encapsulation, data escaping, or framing used to transport the IPMI message on the given media.
The IPMB standard overall message length for ‘non-bridging’ messages is specified as 32 bytes, maximum,
including slave address. This sets the upper limit for typical IPMI messages. With the exception of messages used
for bridging messages to other busses or interfaces (e.g. Master Write-Read and Send Message) IPMI messages
should be designed to fit within this 32-byte maximum. In order to support bridging, the Master Write-Read and
Send Message commands are allowed to exceed the 32-byte maximum transaction on IPMB.
Refer to Appendix D - Determining Message Size RequirementsAppendix D - Determining Message Size
Requirements for information on how these values were derived.
Table 6-9, IPMI Message and IPMB / Private Bus Transaction Size Requirements
Interface
KCS/SMIC Input
KCS/SMIC Output
BT Input
BT Output
IPMB Input
IPMB Output
SMBus 2.0 Input
SMBus 2.0 Output
Private Bus Input
Requirement Description
Required:
40 bytes IPMI Message, minimum
Required:
38 bytes IPMI Message, minimum
Required:
42 bytes IPMI Message, minimum, (including BT Interface ‘length’
byte). The BT interface has length and Seq fields in addition to the
fields used by the KCS and SMIC interfaces. This adds two bytes to
the message size support requirements.
Required:
40 bytes IPMI Message, minimum, (including BT Interface ‘length’
byte)
Required:
32 bytes IPMI Message, minimum (including slave address)
Recommended: 36 bytes bus transaction, minimum (including slave address), to
support an OEM option to allow the BMC to be the target of an SMBus
2.0 Block-Write with PEC.
Required:
36 bytes bus transaction, minimum (including slave address) to allow
the BMC to be able to issue access slave devices that use the SMBus
2.0 Block-Write with PEC protocol. Note that the IPMB standard
message length is shorter than the SMBus 2.0 message.
The IPMB standard message length is specified as 32 bytes,
maximum, including slave address.
Required:
36 bytes bus transaction, minimum (including slave address) to allow
the BMC to be target of an SMBus 2.0 Block-Write with PEC protocol
transaction.
Required:
36 bytes (including slave address) to allow the BMC to generate an
SMBus 2.0 Block-Write with PEC transaction.
Recommended: 34 bytes bus transaction, minimum, if the Private Bus is implemented
2
as a physical I C or SMBus. The 34 bytes supports accessing slave
devices that use the SMBus 2.0 Block-Read protocol. (The count
excludes any slave address, since for this type of transaction the slave
address is output by the management controller, rather than being an
input to the management controller.)
61
Intelligent Platform Management Interface Specification
Private Bus Output
Required:
23 bytes This is only required when a controller indicates that it has a
Private Bus that includes FRU SEEPROMs that are accessible via the
Master Write-Read command must support a Master Write-Read
command equivalent to the largest Master Write-Read command that
could be delivered as a 32-byte IPMB message.
Otherwise, the Private Bus is only required to support the transaction
size required for the private or OEM devices that are used in the
particular implementation.
LAN/PPP Input
LAN/PPP Output
62
An implementation will typically implement a private bus using an
2
actual I C or SMBus connection. However, the private bus
implementation could be ‘virtual’ - where the management controller
responds to the Master Write-Read command as if a physical private
bus were present. For a physical private bus implementation, a 32-byte
Master Write-Read command in IPMB format results in one byte of
slave address and 22 bytes of write data going to the private bus.
Recommended: 36 bytes bus transaction, minimum (including slave address). A
2
private bus that truly implements a physical I C or SMBus interface
should support system management software access to slave devices
that use the SMBus 2.0 Block Write protocol. This means supporting a
Master Write-Read command over the system interface that can be
used to perform a full, 36-byte SMBus 2.0 Block-Write protocol
transaction.
Required:
45 bytes IPMI Message content, minimum. IPMI LAN and PPP
interfaces must accept an RMCP Packet containing an IPMI Message
that would allow the remote console to submit a Master Write-Read
message to perform an SMBus 2.0 Block-Write protocol transaction.
Since the LAN interface uses a message format that follows the IPMB
message format, there are additional bytes for source and destination
addressing, sequence number, and checksums.
Required:
42 bytes IPMI Message content, minimum. IPMI LAN and PPP
interface must support delivering an RMCP Packet containing an IPMI
Message that would allow the BMC to return the response to a Master
Write-Read message that returns data from an SMBus 2.0 Block-Read
protocol transaction.
Intelligent Platform Management Interface Specification
63
Intelligent Platform Management Interface Specification
7. IPMB Interface
7.1 IPMB Access via Master Write-Read command
When an IPMB is implemented in the system, the BMC serves as a controller to give system software access to
the IPMB. The IPMB allows non-intelligent devices as well as management controllers on the bus. To support
this operation, the BMC provides the Master Write-Read command via its interface with system software. The
Master Write-Read command provides low-level access to non-intelligent devices on the IPMB, such as FRU
SEEPROMs.
The Master Write-Read command provides a subset of the possible I2C and SMBus operations that covers most
I2C/SMBus-compatible devices.
In addition to supporting non-intelligent devices on the IPMB, the Master Write-Read command also provides
access to non-intelligent devices on Private Busses behind management controllers. The main purpose of this is to
support FRU SEEPROMs on Private Busses.
7.2 BMC IPMB LUNs
A BMC supports several LUNs (Logical Units) to which messages can be sent via the IPMB interface. These
LUNs are used to identify different sub-addresses within the BMC that messages can be sent to.
Table 7-1, BMC IPMB LUNs
LUN
Short Description
Long Description
00b
BMC commands and Event
Request Messages
01b
10b
OEM LUN 1
SMS Message LUN
(Intended for messages to
System Management
Software)
OEM LUN 2
Event Request Messages received on this LUN are routed to the Event
Receiver function in the BMC, and automatically logged if SEL logging
is enabled .
OEM - reserved for BMC implementer / system integrator definition.
Messages received on this LUN are routed to the Receive Message
Queue and retrieved using a Read Message command. The
SMS_Avail flag is set whenever the Receive Message Queue has valid
contents.
OEM - reserved for BMC implementer / system integrator definition.
11b
7.3 Sending Messages to IPMB from System Software
System Management Software (SMS) can use the BMC for sending and receiving IPMB messages. Both IPMB
request and response messages can be sent and received using this mechanism. Therefore, not only can system
software send requests to the IPMB and receive responses from the IPMB, it is also possible for system software
to receive requests from the IPMB to send back IPMB responses.
System software sends messages to the IPMB through the system interface using the BMC as an IPMB controller.
This is accomplished by using the Send Message command to write the message to the IPMB (channel 0). The
BMC does not place any restrictions on the type or content of the IPMB message being sent. System management
software can send any IPMB request or response message it desires provided that the message meets the
maximum length requirements of the Send Message command.
System Management Software is responsible for providing all fields for the IPMB message, including Requester
and Responder Slave addresses and checksums. The following figures show an example using the Send Message
command to send a Set Event Receiver command to an IPMB device at slave address 52h, LUN 00b, via the
64
Intelligent Platform Management Interface Specification
system interface (see Table 23-2, Set Event ReceiverTable 23-2, Set Event Receiver). The example command sets
the Event Receiver address to 20h = BMC.
The heavy bordered fields show the bytes for the IPMB message carried in the Send Message command. The
requester’s LUN field (rqLUN) is set to 10b (BMC SMS LUN). This directs the responder to send the response to
the Set Event Receiver command to the BMC’s Receive Message Queue.
Figure 7-1, IPMB Request sent using Send Message Command
NetFn
LUN
Command
Channel
(06h = App request)
(00b)
(Send Message)
(00h)
Slave address for write
NetFn/rsLUN
check 1
(52h = rsSA)
( 04h / 00b = Sensor/Event, LUN 00b)
(9Eh)
rqSA
rqSeq/rqLUN
Cmd
(20h = BMC)
(000001b / 10b, 10b = SMS LUN)
(00h = Set Event Receiver)
event receiver slave address
event receiver LUN
check 2
(20h = BMC)
(00h)
(BAh)
Figure 7-2, Send Message Command Response
NetFn
(07h = App response)
LUN
(00b)
Command
(Send Message)
Completion Code
(00h)
Note that the response is for the Send Message command, not for the Set Event Receiver command. The response
to the Set Event Receiver command will be returned later in the Receive Message Queue. System software uses
the Get Message command to read messages from the Receive Message Queue. System software keeps track of
any outstanding responses and matches responses up with corresponding requests as they come in. System
software must also keep track of the protocol assigned to the particular channel in order to interpret the response
to the Get Message command.
7.4 Sending IPMB Messages to System Software
It is possible for devices on the IPMB to autonomously send messages to system management software via the
BMC. IPMB messages that are addressed to the SMS LUN (10b) in the BMC are placed into the Receive
Message Queue. The contents can then be retrieved using the Get Message command. System management can
then interpret the message and use the Send Message command to return a response.
The BMC does not place any restrictions on the type of content of the IPMB message being received, as long as it
is properly formatted, is addressed to the SMS LUN, and meets the maximum length requirements of the Get
Message command.
The BMC sets the corresponding ‘ATN’ flag in the system interface when a message is received into the Receive
Message Queue. System software must poll for the ‘ATN’ flag, or receive an interrupt to determine that when a
message is available.
Event Messages can also be made directly available to system software via the optional Event Message Buffer
and retrieved using the Read Event Message Buffer command.
In the example shown in the preceding section, a Set Event Receiver command was sent out on the IPMB using
the Send Message command. In the IPMB command, the requester’s slave address (rqSA) was set to 20h (BMC),
and the requester’s LUN (rqLUN) set to 10b (SMS LUN). This means that the response will be sent to the SMS
Message Buffer in the BMC.
65
Intelligent Platform Management Interface Specification
The IPMB response to a Set Event Receiver command consists of just a Completion Code byte in the data portion
of the IPMB message. Assuming a Completion Code of 00h = OK, the Receive Message Queue will eventually
wind up a response message with the following contents:
Figure 7-3, Response for Set Event Receiver in Receive Message Queue
NetFn
rqLUN
Check 1
(05h = Sensor/Event Response)
(10b = SMS LUN)
(CAh)
rsSA
rqSeq/rsLUN
Cmd
Completion Code
(52h)
(000001b / 00b)
(00h = Set Event Receiver)
(00h = OK)
Check 2
(AAh)
Note that this is the entire IPMB response message, with the leading slave address stripped off (the leading slave
address does not need to be stored, since its known to be the BMC slave address = 20h).
The response to the Get Message command would then look like the following. The heavy border fields show the
data portion of the response that came from the Receive Message Queue.
Figure 7-4, Get Message Command Response
NetFn
LUN
Command
Completion
Channel
(07h = App response)
(00b)
(Get Message)
Code (00h)
Number (00h)
NetFn
rqLUN
Check 1
(05h = Sensor/Event Response)
(10b = SMS LUN)
(CAh)
rsSA
rqSeq/rsLUN
Cmd
Completion Code
Check 2
(52h)
(000001b / 00b)
(00h = Set Event Receiver)
(00h = OK)
(AAh)
7.5 Testing for Event Message Buffer Support
System software must test for Event Message Buffer support. If software issues a Get BMC Global Enables
command, and finds the buffer enabled, it can assume the controller supports the buffer. Otherwise, it must test by
attempting to enable the Event Message buffer.
If the BMC does not support the desired buffer, it shall return an Invalid Data Field (CCh) error completion code
when an attempt is made to enable the respective buffer using the Set BMC Global Enables command. An error
completion code shall also be returned if an attempt is made to enable Event Message Buffer interrupts when that
option is not supported.
66
Intelligent Platform Management Interface Specification
67
Intelligent Platform Management Interface Specification
8. ICMB Interface
The ICMB Specification (see [ICMB]) describes the interfaces for implementing access via an ICMB Bridge
Controller. ICMB was specified so that an ICMB Bridge controller could be added to an existing IPMI
Implementation that contained an IPMB.
In some implementations, the BMC can serve as the ICMB Bridge Controller. There are two ways to implement the
interface for such a controller:
a.
b.
Implement a virtual ICMB Bridge Controller within the BMC.
(IPMI v1.5 only) Implement the ICMB Bridge commands directly as BMC commands, but use IPMI v1.5
channels and the Send Message command to replace the ICMB Bridge Request command.
8.1 Virtual ICMB Bridge Device
In this implementation, the ICMB Bridge Device functionality appears to system software as if there was a
separate ICMB Bridge Controller on a physical IPMB. Instead, the BMC implements the ICMB Bridge Device
functions internally on a ‘virtual IPMB’. This option can provide backward-compatibility with software that was
designed to work with a non-integrated ICMB Bridge Device.
The BMC reports that the Chassis Bridge Device is not part of the BMC by returning an address in the Get
Chassis Capabilities that is different than the BMC address (20h). This indicates to software that it need to access
the Bridge Device function by using Send Message commands to deliver messages to the Bridge Device via the
primary IPMB. The BMC then monitors the Send Message command for messages directed to the Bridge Device
address. When the BMC sees Send Message commands to the Bridge Device address, it handles them internally
instead of routing them out to a physical IPMB. Responses from the virtual Bridge Device are placed into the
Receive Message Queue as if they were received from the IPMB.
It is optional for the BMC to provide ICMB access from the IPMB for this implementation. If such support is
desired, there are two implementation options. The first option is for the BMC to respond to two slave addresses
(the BMC address and the Bridge Device address). The second option is for the BMC to report the BMC address
as the Bridge Device address whenever the Get Chassis Capabilities command is received from IPMB, and
implement the bridge commands directly when accessed via the IPMB.
8.2 ICMB Bridge Commands in BMC using Channels
In this implementation, the BMC directly responds to the commands for the Chassis Bridge Device. The BMC
reports its own address in the Get Chassis Capabilities command. This tells software that it does not need to use
the Send Message command to encapsulate messages in order to access the Chassis Bridge function itself. This
also tells software that is must use the Send Message command instead of the Bridge Request command.
8.2.1
ICMB Bridging from System Interface to Remote IPMB using Channels
The behavior with the Send Message command is somewhat different than the operation of the Bridge Request
command implemented by a separated bridge controller on the IPMB. When the Bridge Request command was
used to access the ICMB, the response to that command would hold the response from the ICMB.
From the System Interface, bridging with the Send Message command is a multi-step operation: First you issue
the Send Message command with the data to be sent to the ICMB. You then get a response to the Send Message
command indicating that the data was successfully bridged onto the ICMB. This response does not contain the
response data from the ICMB. Later (assuming that a request was bridged) the device on ICMB will respond
and the response data will appear in the Receive Message Queue (if the System Interface was the source of the
68
Intelligent Platform Management Interface Specification
original Send Message command). The software can then use a Get Message command to retrieve the response
message data.
The following tables show the KCS formats of the Send Message command request and response for bridging a
request to a device on a remote IPMB and the later corresponding Receive Message Queue contents for the
response from the remote device.
Table 8-1, System Interface Request For Delivering Remote IPMB Request via ICMB
NetFn/RsLUN
CMD
Data 1
Data 2
Data 3:4
Data 5
Data 6
Data 7
Data 8
Data 9:N
Checksum
App (even=Rq) / BMC LUN = 00b
Send Message
Channel Number = ICMB, track request = 1b
rqSeq = sequence number selected by system
software / 00b
rmtBrXA
Bridge Request CMD
rsSA for remote IPMB device
netFn / rsLUN for remote IPMB device
CMD for remote IPMB device
Data for remote IPMB device
Checksum for Send Message Command
Table 8-2, Send Message Response
NetFn/RsLUN
CMD
Data 1
Checksum
App (odd=Rs) / BMC LUN = 00b
Send Message
Completion Code for Send Message command
Checksum for Send Message Command
Table 8-2a, Get Message Response Data for Remote IPMB Request Delivered via ICMB
NetFn/RsLUN
CMD
Data 1
Data 2
Data 3
Data 4
Data 5
Data 6:N
Checksum
8.2.2
App (odd=Rs) / BMC LUN = 00b
Get Message
Completion Code for Get Message command
Channel Number = ICMB
rqSeq = Sequence number from original request
/ 00b
rmtBr Completion Code
remote IPMB completion code
remote IPMB data
Checksum for Send Message Command
ICMB Bridging from Local IPMB to Remote IPMB using Channels
From the IPMB, bridging with the Send Message command operates in a manner similar to the way it operates
over the system interface. The main difference being that the device that originated the request later receives an
asynchronous response message that appears as if the BMC is responding directly to the remote IPMB
command.
Note that the same rqSeq is used both in the response to the Send Message command and in the asynchronous
response from the BMC.
Bridging with this approach introduces a five-byte overhead on the request, and a 0-byte overhead on the
response.
69
Intelligent Platform Management Interface Specification
Table 8-3, IPMB Request For Delivering Remote IPMB Request via ICMB
RsSA
NetFn/RsLUN
RqSA
RqSeq/RqLUN
CMD
Data 1
Data 2:3
Data 4
Data 5
Data 6
Data 7
Data 8:N
Checksum
20h (BMC)
App (even=Rq) / 00b (BMC LUN)
Address of local IPMB device issuing the request
Sequence number selected by IPMB device
issuing the request
/ LUN of IPMB device issuing the request
Send Message
Channel Number = ICMB, track request = 1b
rmtBrXA
Bridge Request CMD (Tells BMC to deliver
this to ICMB a message to be bridged to a
remote IPMB)
rsSA for remote IPMB device
netFn / rsLUN for remote IPMB device
Remote CMD (CMD for remote IPMB device)
Remote Data (data for remote IPMB device)
Checksum for Send Message Command
Table 8-4, Send Message Response
RqSA
NetFn/RqLUN
RsSA
RqSeq /
RsLUN
NetFn/RsLUN
CMD
Data 1
Checksum
Address of local IPMB device that issued the
original request
App (odd-Rs) / LUN of device that issued the
original request
20h (BMC)
Sequence number from original request
/ 00b (BMC LUN)
App (odd=Rs) / BMC LUN = 00b
Send Message
Completion Code for Send Message command
Checksum for Send Message Command
Table 8-5, IPMB Response For Remote IPMB Request Delivered via ICMB
RqSA
NetFn/RqLUN
RsSA
RqSeq /
RsLUN
NetFn/RsLUN
CMD
Data 1
Data 4:N
Checksum
70
Address of local IPMB device that issued the
original request
App (odd-Rs) / LUN of device that issued the
original request
20h (BMC)
Sequence number from original Send Message
request
/ 00b (BMC LUN)
App (odd=Rs) / BMC LUN = 00b
Remote CMD
Remote CMD Completion code
Remote CMD data
Checksum for Send Message Command
Intelligent Platform Management Interface Specification
71
Intelligent Platform Management Interface Specification
9. Keyboard Controller Style (KCS) Interface
This section describes the Keyboard Controller Style (KCS) Interface. The KCS interface is one of the supported
BMC to SMS interfaces. The KCS interface is specified solely for SMS messages. SMM messages between the
BMC and an SMI Handler will typically require a separate interface, though the KCS interface is designed so that
system software can detect if a transaction was interrupted. Any BMC-to-SMI Handler communication via the KCS
interface is implementation specific and is not covered by this specification.
The KCS Interface is designed to support polled operation. Implementations can optionally provide an interrupt
driven from the OBF flag, but this must not prevent driver software from the using the interface in a polled manner.
This allows software to default to polled operation. It also allows software to use the KCS interface in a polled
mode until it determines the type of interrupt support. Methods for assigning and enabling such an interrupt are
outside the scope of this specification.
9.1 KCS Interface/BMC LUNs
LUN 00b is typically used for all messages to the BMC through the KCS Interface. LUN 10b is reserved for
Receive Message Queue use and should not be used for sending commands to the BMC. Note that messages
encapsulated in a Send Message command can use any LUN in the encapsulated portion.
9.2 KCS Interface-BMC Request Message Format
Request Messages are sent to the BMC from system software using a write transfer through the KCS Interface.
The message bytes are organized according to the following format specification:
Figure 9-1, KCS Interface/BMC Request Message Format
Byte 1
NetFn/LUN
Byte 2
Cmd
Byte 3:N
Data
Where:
72
LUN
Logical Unit Number. This is a sub-address that allows messages to be routed to different
‘logical units’ that reside behind the same physical interface. The LUN field occupies the least
significant two bits of the first message byte.
NetFn
Network Function code. This provides the first level of functional routing for messages received
by the BMC via the KCS Interface. The NetFn field occupies the most significant six bits of the
first message byte. Even NetFn values are used for requests to the BMC, and odd NetFn values
are returned in responses from the BMC.
Cmd
Command code. This message byte specifies the operation that is to be executed under the
specified Network Function.
Data
Zero or more bytes of data, as required by the given command. The general convention is to
pass data LS-byte first, but check the individual command specifications to be sure.
Intelligent Platform Management Interface Specification
9.3 BMC-KCS Interface Response Message Format
Response Messages are read transfers from the BMC to system software via the KCS Interface. Note that the
BMC only returns responses via the KCS Interface when Data needs to be returned. The message bytes are
organized according to the following format specification:
Figure 9-2, KCS Interface/BMC Response Message Format
Byte 1
NetFn/LUN
Byte 2
Cmd
Byte 3
Completion Code
Byte 4:N
Data
Where:
LUN
Logical Unit Number. This is a return of the LUN that was passed in the Request
Message.
NetFn
Network Function. This is a return of the NetFn code that was passed in the Request
Message. Except that an odd NetFn value is returned.
Cmd
Command. This is a return of the Cmd code that was passed in the Request Message.
Completion Code
The Completion Code indicates whether the request completed successfully or not.
Data
Zero or more bytes of data. The BMC always returns a response to acknowledge the
request, regardless of whether data is returned or not.
9.4 Logging Events from System Software via KCS Interface
The KCS Interface can be used for sending Event Messages from system software to the BMC Event Receiver.
The following figures show the format for KCS Interface Event Request and corresponding Event Response
messages. Note that only Event Request Messages to the BMC via the KCS Interface have a Software ID field.
This is so the Software ID can be saved in the logged event.
Figure 9-3, KCS Interface Event Request Message Format
NetFn
LUN
(04h = Sensor/Event Request)
(00b)
EvMRev
Sensor Type
Sensor #
Software ID (Gen ID)
1
Command
(02h = Platform Event)
7-bits
Event Dir
Event Type
Event Data 1
Event Data 2
Event Data 3
Shading designates fields that are not stored in the event record.
Figure 9-4, KCS Interface Event Response Message Format
NetFn
(05h = Sensor/Event Response)
00
Command
(02h = Platform Event)
Completion Code
9.5 KCS Interface Registers
The KCS Interface defines a set of I/O mapped communication registers. The bit definitions, and operation of
these registers follows that used in the Intel 8742 Universal Peripheral Interface microcontroller. The term
‘Keyboard Controller Style’ reflects the fact that the 8742 interface is used as the system keyboard controller
interface in PC architecture computer systems.
The specification of the KCS Interface registers is given solely with respect to the ‘system software side’ view of
the interface in system I/O space. The functional behavior of the management controller to support the KCS
Interface registers is specified, but the physical implementation of the interface and the organization of the
73
Intelligent Platform Management Interface Specification
interface from the management controller side is implementation dependent and is beyond the scope of this
specification.
On the system side, the registers are mapped to system I/O space and consists of four byte-wide registers.
•
Status Register - provides flags and status bits for use in various defined operations.
•
Command Register - provides port into which ‘Write Control Codes’ may be written.
•
Data_In - provides a port into which data bytes and ‘Read Control Codes’ may be written.
•
Data_Out - provides a port from which data bytes may be read.
The default system base address for an I/O mapped KCS SMS Interface is CA2h. Refer to Appendix C1 Locating IPMI System Interfaces via SM BIOS TablesAppendix C1 - Locating IPMI System Interfaces via SM
BIOS Tables for information on using SM BIOS tables for describing optional interrupt usage, memory mapped
registers, 32-bit and 16-byte aligned registers, and alternative KCS interface addresses. Software can assume the
KCS interface registers are I/O mapped and byte aligned at the default address unless other information is
provided.
Figure 9-5, KCS Interface Registers
Status (ro)
7
6
5
4
3
2
1
0
S1
S0
OEM
2
OEM
1
C/D#
SMS_ATN
IBF
OBF
Command (wo)
Data_Out (ro)
Data_In (wo)
I/O address
base+1
base+1
base+0
base+0
Reserved bits must be written as ‘0’ and ignored during reads. Software should not assume that
reserved bits return a constant value.
9.6 KCS Interface Control Codes
Control codes are used for ‘framing’ message data transferred across the KCS Interface. Control Codes are used
to:
•
Identify the first and last bytes of a packet.
•
Identify when an error/abort has occurred.
•
Request additional data bytes.
9.7 Status Register
System software always initiates a transfer. If the BMC has a message for SMS, it can request attention by setting
the SMS_ATN bit in the status register. System software then detects the flag and initiates the transfer.
Other bits in the status register are used to arbitrate access to the command and data registers between the BMC
and system software and to indicate the “state” (write, read, error, or idle) of the current transaction. The
following tables summarize the functions of the Status Register bits.
74
Intelligent Platform Management Interface Specification
Table 9-1, KCS Interface Status Register Bits
Bit
Name
7
S1
6
5
4
3
S0
OEM2
OEM1
C/D#
2
SMS_ATN
1
IBF
0
OBF
1.
2.
[1]
Description
R/W
State bit 1. Bits 7 & 6 are used to indicate the current state of the KCS Interface.
Host Software should examine these bits to verify that it’s in sync with the BMC.
See below for more detail.
State bit 0. See bit 7.
OEM - reserved for BMC implementer / system integrator definition.
OEM - reserved for BMC implementer / system integrator definition.
Specifies whether the last write was to the Command register or the Data_In
register (1=command, 0=data). Set by hardware to indicate whether last write
from the system software side was to Command or Data_In register.
Set to 1 when the BMC has one or more messages in the Receive Message
Queue, or when a watchdog timer pre-timeout, or event message buffer full
[2]
condition exists . OEMs may also elect to set this flag is one of the OEM 1, 2,
or 3 flags from the Get Message Flags command becomes set.
R/O
This bit is related to indicating when the BMC is the source of a system
interrupt. Refer to sections 9.12, KCS Communication and Non-communication
InterruptsKCS Communication and Non-communication Interrupts, 9.13,
Physical Interrupt Line SharingPhysical Interrupt Line Sharing, and 9.14,
Additional Specifications for the KCS interfaceAdditional Specifications for the
KCS interface for additional information on the use and requirements for the
SMS_ATN bit.
Automatically set to 1 when either the associated Command or Data_In register
has been written by system-side software.
Set to 1 when the associated Data_Out register has been written by the BMC.
R/O
R/O
R/O
R/O
R/O
R/O
R/O
R/W direction is with respect to the system side of the interface. Reads move data from the BMC to system software,
writes move data from system software to the BMC.
The event message buffer full condition contributes to SMS_ATN only if the event buffer full condition is intended to
be handled by system management software. Otherwise, the event message buffer full condition should not
contribute to SMS_ATN. For interrupt driven interfaces, the condition is required to contribute to SMS_ATN if the
event message buffer full condition generates the same interrupt as the KCS Communications interrupt.
Bits 7:6 are state bits that provide information that is used to ensure that the BMC and system software remain in
sync with one another. Following are the possible states and their meaning:
Table 9-2, KCS Interface State Bits
S1
(bit 7)
S0
(bit 6)
0
0
0
1
Definition
IDLE_STATE. Interface is idle. System software should not be expecting nor sending any data.
READ_STATE. BMC is transferring a packet to system software. System software should be in the
“Read Message” state.
1
0
WRITE_STATE. BMC is receiving a packet from system software. System software should be
writing a command to the BMC.
1
1
ERROR_STATE. BMC has detected a protocol violation at the interface level, or the transfer has
been aborted. System software can either use the “Get_Status’ control code to request the nature of
the error, or it can just retry the command.
Note: Whenever the BMC is reset (from power-on or a hard reset), the State Bits are initialized to “11 - Error State”. Doing so
allows SMS to detect that the BMC has been reset and that any message in process has been terminated by the BMC.
9.7.1
SMS_ATN Flag Usage
The SMS_ATN flag is used to indicate that the BMC requires attention from system software. This could either
be because a message was received into the Receive Message Queue and ready for delivery to system software,
the Event Message Buffer is full (if the Event Message Buffer is configured to generate an interrupt to system
management software), a watchdog pre-timeout occurred, or because of an OEM event. Flags in the BMC
75
Intelligent Platform Management Interface Specification
identify which conditions are causing the SMS_ATN flag to be set. All conditions must be cleared (i.e. all
messages must be flushed) in order for the SMS_ATN bit to be cleared.
The SMS_ATN bit is also used when the KCS interface is interrupt driven, or when OEM events or watchdog
pre-timeouts generate a system interrupt. Refer to sections 9.12, KCS Communication and Non-communication
InterruptsKCS Communication and Non-communication Interrupts, 9.13, Physical Interrupt Line
SharingPhysical Interrupt Line Sharing, and 9.14, Additional Specifications for the KCS interfaceAdditional
Specifications for the KCS interface for additional information on the use and requirements for the SMS_ATN
bit.
9.8 Command Register
The Command register must only be written from the system side when the IBF flag is clear. Only
WRITE_START, WRITE_END, or GET_STATUS/ABORT Control Codes are written to the command register.
9.9 Data Registers
Packets to and from the BMC are passed through the data registers. These bytes contain all the fields of a
message, such as the Network Function code, Command Byte, and any additional data required for the Request or
Response message.
The Data_In register must only be written from the system side when the IBF flag is clear. The OBF flag must be
set (1) before the Data_Out register can be read (see status register).
9.10 KCS Control Codes
The following table details the usage of ‘Control Codes’ by the KCS interface.
Table 9-3, KCS Interface Control Codes
Code
Name
60h
61h
GET_STATUS /
ABORT
WRITE_START
62h
WRITE_END
63h-67h
68h
69h6Fh
reserved
READ
reserved
Description
Target
register
Output Data
Register
Request Interface Status / Abort Current
operation
Write the First byte
of an Write Transfer
Write the Last byte
of an Write Transfer
reserved
Request the next data byte
reserved
Command
Status Code
Command
N/A.
Command
N/A
Data_In
Next byte
Table 9-4, KCS Interface Status Codes
Code
00h
01h
02h
06h
C0h-FEh
FFH
all other
76
Description
No Error
Aborted By Command (Transfer in progress was aborted by SMS issuing the Abort/Status
control code)
Illegal Control Code
Length Error (e.g.overrun)
OEM Error (Error must not fit into one of above categories.)
Unspecified Error
Reserved
Intelligent Platform Management Interface Specification
9.11 Performing KCS Interface Message Transfers
System Management Software that uses the KCS Interface will typically be running under a multi-tasking
operating system. This means transfers with the BMC may be interrupted by higher priority tasks or delayed by
other System Management Software processing. The SMS channel handshake is optimized to allow the BMC to
continue to perform tasks between data byte transfers with System Management Software. The BMC does not
time out data byte transfers on the SMS interface.
Request and Response Messages are paired together as a Write Transfer to the BMC to send the request followed
by a Read Transfer from the BMC to get the response.
The process, as seen from the system perspective is depicted in Figure 9-6, KCS Interface SMS to BMC Write
Transfer Flow Chart, and Figure 9-7, KCS Interface BMC to SMS Read Transfer Flow Chart, below.
During the write transfer each write of a Control Code to the command register and each write of a data byte to
Data_In will cause IBF to become set, triggering the BMC to read in the corresponding Control Code or data
byte.
If the KCS interface uses an interrupt, the BMC will write a dummy value of 00h to the output data register after
it has updated the status register and read the input buffer. This generates an ‘OBF’ interrupt. The points at which
this would occur are shown as “OBF” in Figure 9-6, KCS Interface SMS to BMC Write Transfer Flow Chart,
below.
During the read phase, each write of a READ Control Code to Data_In will cause IBF to become set, causing the
BMC to read in the Control Code and write a data byte to Data_Out in response. If the KCS interface uses an
interrupt, the write of the data byte to Data_Out will also generate an interrupt. The point at which this would
occur during the READ_STATE is shown as “OBF” in Figure 9-7, KCS Interface BMC to SMS Read Transfer
Flow Chart, below.
Note that software does not need to use the Get Status/Abort transaction to return the interface to the
IDLE_STATE or handle an error condition. The interface should return to IDLE_STATE on successful
completion of any full command/response transaction with the BMC. Thus, since the interface will allow a
command transfer to be started or restarted at any time when the input buffer is empty, software could elect to
simply retry the command upon detecting an error condition, or issue a ‘known good’ command in order to clear
ERROR_STATE.
9.12 KCS Communication and Non-communication Interrupts
The following lists some general requirements and clarifications to support both KCS communication and KCS
non-communication interrupts on the same interrupt line using the OBF signal. A KCS communications interrupt
is defined as an OBF-generated interrupt that occurs during the process of sending a request message to the BMC
and receiving the corresponding response. This occurs from the start of the write (request) phase of the message
(issuing WRITE_START to the command register) through to the normal conclusion of the corresponding read
(response) phase of the message. (The conclusion of the communications interval is normally identified by the
interface going to IDLE_STATE). KCS communications interrupts are also encountered during the course of
processing a GET_STATUS/ABORT control code.
A KCS non-communication interrupt is defined as an OBF-generated interrupt that occurs when the BMC is not
in the process of transferring message data or getting error status. This will typically be an interrupt that occurs
while the interface is in the IDLE_STATE.
There are several options in the BMC that can be enabled to cause KCS non-communication interrupts as
described in the Set BMC Global Enables command, and Get Message Flags commands. These are the watchdog
timer pre-timeout interrupt, event message buffer interrupt, receive message queue interrupt, and the OEM
interrupts. Software can detect which of the standard interrupts are supported by attempting to enable them using
the Set BMC Global Enables command and checking for an error completion code.
77
Intelligent Platform Management Interface Specification
9.13 Physical Interrupt Line Sharing
A typical interrupt-driven implementation will assert a physical interrupt line when OBF is asserted. In order to
allow a single interrupt line to serve for both communication and non-communication interrupts, the physical
interrupt line must be automatically deasserted by the BMC whenever a communication phase begins, even if
there is a pending non-communications interrupt to be serviced. This is necessary so the interrupt line can be used
for signaling communication interrupts . Once the communication operations have completed (return to idle
phase) the controller must re-assert the interrupt line if the non-communications interrupt is still pending.
9.14 Additional Specifications for the KCS interface
This section lists additional specifications for the KCS interface.
•
The BMC must generate an OBF whenever it changes the status to ERROR_STATE. This will ensure that any
transition to ERROR_STATE will cause the interrupt handler to run and catch the state.
•
The BMC generates an OBF upon changing the status to IDLE_STATE. An IPMI 1.5 implementation is
allowed to share this interrupt with a pending KCS non-communication interrupt, or it elect to always generate
a separate OBF interrupt for non-communications interrupts.
•
A BMC implementation that elects to always generate a separate non-communications interrupt must wait for
the OBF interrupt that signals entering the IDLE_STATE to be cleared before it asserts an OBF interrupt for
the non-communications interrupt.
•
IPMI v1.5 systems are allowed to generate a single OBF that covers both the last communications interrupt
(when the BMC status goes to IDLE_STATE) and a pending non-communications interrupt. I.e. it is not
required to generate a separate OBF interrupt for the non-communications interrupt if a non-communications
interrupt was pending at the time the BMC status goes to IDLE_STATE. In order to support this, an IPMI v1.5
KCS interface implementation must set SMS_ATN for all standard (IPMI defined) non-communication interrupt
sources.
•
For IPMI v1.5, the BMC must set the SMS_ATN flag if any of the standard message flags become set. This
includes Receive Message Available, Event Message Buffer Full (if the Event Message Buffer Full condition is
intended to be handled by System Management Software), and Watchdog Timer pre-timeout flags, as listed in
the Get Message Flags command. This is independent of whether the corresponding interrupt is enabled or not.
•
The BMC must change the status to ERROR_STATE on any condition where it aborts a command transfer in
progress. For example, if the BMC had an OEM command that allowed the KCS interface to be asynchronously
reset via IPMB, the KCS interface status should be put into the ERROR_STATE and OBF set, not
IDLE_STATE, in order for software to be notified of the change. However, the BMC does not change the
status to the ERROR_STATE, but to the IDLE_STATE, when the BMC executes the Get Status/Abort control
code from SMS I/F, even if the Get Status/Abort control code is used to abort a transfer.
•
A cross-platform driver must be able to function without handling any of the OEM bits. Therefore, enabling
SMS_ATN on OEM interrupts/states must not be enabled by default, but must be explicitly enabled either by
the Set BMC Global Enables command or by an OEM-defined command.
•
The SMS_ATN bit will remain set until all standard interrupt sources in the BMC have been cleared by the
Clear Message Flags command, or by a corresponding command. For example, the Read Message command
can automatically clear the Receive Message Queue interrupt if the command empties the queue.
•
A KCS interface implementation that allows its interrupt to be shared with other hardware must set SMS_ATN
whenever it generates a KCS interrupt. A system will typically report whether it allows an interrupt to be shared
or not via resource usage configuration reporting structures such as those in ACPI.
•
OEM non-communications interrupts should be disabled by default. They must be returned to the disabled state
whenever the controller or the system is powered up or reset. This is necessary to allow a generic driver to be
used with the controller. A driver or system software must be explicitly required to enable vendor-specific non-
78
Intelligent Platform Management Interface Specification
communications interrupt sources in order for them to be used. OEM non-communications interrupt sources
must not contribute to SMS_ATN when they are disabled.
•
The OEM 0, 1, and 2 flags that are returned by the Get Message Flags command may also cause the
SMS_ATN flag to be set. A platform or system software must not enable these interrupts/flags unless there is a
corresponding driver that can handle them. Otherwise, a generic cross-platform driver could get into a situation
where it would never be able to clear SMS_ATN.
•
It is recommended that any OEM generated non-communications interrupts cause at least one of the OEM flags
in the Get Message Flags to become set. This will enable improving system efficiency by allowing a crossplatform driver to pass the value of the Get Message Flags to an OEM extension, saving the OEM extension
software from having to issue an additional command to determine whether it has an anything to process.
•
It is recommended that an OEM that uses the OEM flags sets the SMS_ATN flag if one or more of the OEM
flags (OEM 0, OEM 1, or OEM 2) becomes set, especially if those flags can be the source of a KCS noncommunications interrupt. The driver can use SMS_ATN as the clue to execute the Get Message Flags
command and pass the data along to an OEM extension routine.
•
OEM non-communications interrupts may elect to either share the IDLE_STATE OBF interrupt with the noncommunications interrupt OBF, or generate a separate non-communications OBF interrupt. If the OEM noncommunications interrupt implementation shares the IDLE_STATE OBF interrupt, the OEM noncommunications interrupt must also set SMS_ATN.
9.15 KCS Flow Diagrams
The following flow diagrams have been updated from corresponding diagrams in the original IPMI v1.0, rev. 1.1
specification. This information applies to the following flow diagrams:
•
All system software wait loops should include error timeouts. For simplicity, such timeouts are not shown
explicitly in the flow diagrams. A five-second timeout or greater is recommended.
•
The phase values represent state information that could be kept across different activations of an interrupt
handler, and corresponding entry points. Based on the 'phase' the interrupt handler would branch to the
corresponding point when an OBF interrupt occurred. The information may also be useful for error reporting
and handling for both polled- and interrupt-driven drivers. Note that other state may need to be kept as well.
For example, during the 'wr_data’ phase, the handler may also need to preserve a byte counter in order to track
when the last byte of the write was to be sent.
•
The symbol of a circle with an arrow and the text ‘OBF’ inside the circle represents the points where the BMC
would write a dummy data byte to the output buffer in order to create an OBF interrupt. The label above the
circle indicates where an interrupt handler would branch to when the OBF interrupt occurs under in the
corresponding phase. An interrupt handler would exit upon completing the step that occurs before where the
OBF interrupt symbol points.
79
Intelligent Platform Management Interface Specification
Figure 9-6, KCS Interface SMS to BMC Write Transfer Flow Chart
WRITE
BMC sets status to
WRITE_STATE
immediately after
receiving any control code
in the command register
unless it needs to force an
ERROR_STATE. The
status is set before
reading the control code
from the input buffer.
wait for IBF=0
clear OBF
wr_start
WR_START to CMD
phase=wr_start
OBF
WRITE_STATE?
OBF
wait for IBF=0
In the unlikely event that
an asynchronous interrupt
occurs after clearing OBF
the interrupt handler may
spin waiting for IBF=0.
wait for IBF=0
No
Yes
WRITE_STATE?
Error_Exit
data byte to DATA
phase=read
READ
OBF
wait for IBF=0
Yes
No
WRITE_STATE?
Yes
Clear OBF
Last write byte?
80
No
Error Exit
Error Exit
Clear OBF
wr_data
data byte to DATA
No
Yes
BMC updates state after
receiving data byte in
DATA_IN, but before
reading the byte out of
the input buffer. I.e. it
changes state while
IBF=1
Clear OBF
phase=wr_data
wr_end_cmd
WR_END to CMD
phase=wr_end_cmd
The BMC sets state
to READ_STATE
before reading data
byte from data
register. This ensures
state change to
READ_STATE
occurs while IBF=1.
Intelligent Platform Management Interface Specification
Figure 9-7, KCS Interface BMC to SMS Read Transfer Flow Chart
This OBF is normally caused by the BMC returning a data byte for the read
operation. After the last data byte, the BMC sets the state to IDLE_STATE
while IBF=1 and then reads the input buffer to check the control code =
READ. The status will be set to ERROR_STATE if the control code is not
READ. The BMC then writes a dummy data byte to the output buffer to
generate an interrupt so the driver can see the status change.
READ
read
Note that software must track that it has received an interrupt from
'IDLE_STATE' while it is still in the 'read' phase in order to differentiate it
from a non-communication interrupt. If the BMC needs to set the status to
ERROR_STATE it will do so before writing a dummy 00h byte to the output
buffer. (The BMC always places a dummy byte in the output buffer
whenever it sets the status to ERROR_STATE.)
OBF
wait for IBF=0
READ_STATE?
No
IDLE_STATE?
Yes
Yes
wait for OBF=1
wait for OBF=1
Read data byte from DATA_OUT
Read dummy data byte from
DATA_OUT
write READ byte to DATA_IN
phase = idle
Exit
No
Error Exit
The BMC must wait for software to read
the output buffer before writing OBF to
generate a non-communications interrupt.
That is, if there are any pending interrupts
while in IDLE_STATE, but OBF is already
set, it must hold off the interrupt until it
sees OBF go clear. Software must be
careful, since missing any read of the
output buffer will effectively disable
interrupt generation. It may be a prudent
safeguard for a driver to poll for OBF
occassionallywhen waiting for an interrupt
from the IDLE state.
Note that for IPMI v1.5, the last OBF
interrupt is allowed to be shared with a
pending non-communications interrupt.
See text.
81
Intelligent Platform Management Interface Specification
The following figure shows a flow diagram for aborting KCS transactions in progress and/or retrieving KCS error
status.
Figure 9-8, Aborting KCS Transactions in-progress and/or Retrieving KCS Error Status
Error Exit
wait for IBF=0
error1
BMC writes dummy byte.
Required for interrupt
driven systems. Optional
if BMC supports polled
access only.
GET_STATUS/ABORT to CMD
phase = error1
OBF
wait for IBF=0
clear OBF
error2
BMC sets status to
READ_STATE and writes
the error status byte to
the DATA_OUT register.
00h to DATA_IN
phase = error2
OBF
BMC sets status to
'WRITE_STATE' (BMC
always sets status to
WRITE_STATE upon
getting a control code in
the command register).
The BMC then generates
OBF interrupt to signal
that it has read the byte
from the command
register.
This dummy byte
interrupts the BMC and
tells it that the software
has handled the OBF
interrupt and is ready for
the next state.
wait for IBF=0
READ_STATE?
Yes
wait for OBF=1
No
Read error status code byte from
DATA_OUT
This write interrupts the
BMC and tells it that the
software has retrieved
the error status byte
Write READ dummy byte to DATA_IN
phase = error3
error3
wait for IBF=0
OBF
Note that this last interrupt
occurs when the BMC is in
IDLE_STATE. The driver
must track that this interrupt
is expected, otherwise it
might interpret it as a noncommunications interrupt.
IDLE_STATE?
No
Yes
wait for OBF=1
increment retry count
clear OBF
RETRY LIMIT?
phase = idle
Yes
Exit
82
Comm Failure
No
Intelligent Platform Management Interface Specification
9.16 Write Processing Summary
The following summarizes the main steps write transfer from system software to the BMC:
•
Issue a ‘WRITE_START’ control code to the Command register to start the transaction.
•
Write data bytes (NetFn, Command, Data) to Data_In.
•
Issue an ‘WRITE_END’ control code then the last data byte to conclude the write transaction.
9.17 Read Processing Summary
The following summarizes the main steps for a read transfer from the BMC to system software:
•
Read Data_Out when OBF set
•
Issue READ command to request additional bytes
•
If READ_STATE (after IBF = 0), repeat previous two steps.
9.18 Error Processing Summary
The following summarizes the main steps by which system software processes KCS Interface errors:
•
Issue a ‘GET_STATUS/ABORT’ control code to the Command register. Wait for IBF=0. State should be
WRITE_STATE.
•
If OBF=1, Clear OBF by reading Data_Out register.
•
Write 00h to data register, wait for IBF=0. State should now be READ_STATE.
•
Wait for OBF=1. Read status from Data_Out
•
Conclude by writing READ to data register, wait for IBF=0. State should be IDLE.
83
Intelligent Platform Management Interface Specification
9.19 Interrupting Messages in Progress
If, during a message transfer, the system software wants to abort a message it can do so by the following methods:
1.
Place another “WRITE_START” command into the Command Register (a WRITE_START Control Code
is always legal). The BMC then sets the state flags to “WRITE_STATE” and sets its internal flags to
indicate that the stream has been aborted.
2.
Send a “GET_STATUS/ABORT” request. This is actually the same as #1 above but is explicitly stated to
indicate that this command will cause the current packet to be aborted. This command allows a stream to
be terminated and the state to be returned to IDLE without requiring a complete BMC request and response
transfer.
9.20 KCS Driver Design Recommendations
•
A generic, cross-platform driver that supports the interrupt-driven KCS interface is not required to handle
interrupts other than the interrupt signal used for IPMI message communication with the BMC. The message
interrupt may be shared with other BMC interrupt sources, such as the watchdog timer pre-timeout interrupt,
the event message buffer full interrupt, and OEM interrupts.
•
A cross-platform driver should use the Get BMC Global Enables and Set BMC Global Enables commands in a
‘read-modify-write’ manner to avoid modifying the settings of any OEM interrupts or flags.
•
It is recommended that cross-platform driver software provide a ‘hook’ that allows OEM extension software to
do additional processing of KCS non-communication interrupts. It is highly recommended that the driver
execute the Get Message Flags command whenever SMS_ATN remains set after normal processing and
provide the results to the OEM extension software.
•
The driver cannot know the whether the pre-existing state of any OEM interrupts or flags is correct. Therefore,
a driver that supports OEM extensions should allow for an OEM initialization routine that can configure the
OEM flags/interrupts before KCS OBF-generated interrupts are enabled.
•
It is recommended that cross-platform drivers or software make provision for BMC implementations that may
miss generating interrupts on a command error condition by having a timeout that will activate the driver or
software in case an expected interrupt is not received.
•
A driver should be designed to allow for the possibility that an earlier BMC implementation does not set the
SMS_ATN flag except when there is data in the Receive Message Queue. If the driver cannot determine
whether SMS_ATN is supported for all enabled standard flags or not, it should issue a Get Message Flags
command whenever it gets a KCS non-communication interrupt.
•
A driver or system software can test for whether the Watchdog Timer pre-timeout and/or Event Message Buffer
Full flags will cause SMS_ATN to become set. This is accomplished by disabling the associated interrupts (if
enabled) and then causing a corresponding action that sets the flag. This is straightforward by using the
watchdog timer commands in conjunction with the Set BMC Global Enables and Get Message Flags
commands.
For example, to test for the Event Message Buffer Full flag setting SMS_ATN, first check to see if the Event
Message Buffer feature is implemented by attempting to enable the event message buffer using the Set and Get
BMC Global Enables command. If the feature is not implemented, an error completion code will be returned.
Next, disable event logging and use the watchdog timer to generate an SMS/OS ‘no action’ timeout event, then
see if the SMS_ATN becomes set. If so, use the Get Message Flags command to verify that the Event Message
Buffer Full flag is the only one set (in case an asynchronous message came in to the Receive Message Queue
during the test.) The pre-timeout interrupt can be testing in a similar manner.
84
Intelligent Platform Management Interface Specification
•
It is possible (though not recommended) for a BMC implementation to include proprietary noncommunication interrupt sources that do not set SMS_ATN. These sources must not be enabled by default. It
is recommended that a generic cross-platform driver have provisions for OEM extensions that get called
whenever a non-communication interrupt occurs. It is recommended that the extension interface provides the
last reading of the KCS flags so that an OEM extension can see the state of SMS_ATN.
•
Software should be aware that IPMI v1.0 implementations were not required to set SMS_ATN for all noncommunication interrupts. If a BMC implementation does not set SMS_ATN for all non-communication
interrupts, it must generate a separate OBF interrupt for non-communication interrupts. A controller that does
not set SMS_ATN for all non-communication interrupts is not allowed to use the same OBF interrupt to
signal the both completion of communications and a non-communications interrupt.
•
Regardless of whether the IDLE_STATE OBF interrupt is shared with a pending non-communications
interrupt, software drivers must examine SMS_ATN after clearing OBF. If SMS_ATN is asserted the driver
must process the non-communications interrupt sources.
85
Intelligent Platform Management Interface Specification
10. SMIC Interface
This section provides the specifications of the SMIC (Server Management Interface Chip) interface. The SMIC
interface is one of the physical interfaces specified for transferring IPMI messages between the system management
software and the system’s primary management controller (BMC).
The interface can be readily implemented using an external ASIC or standard programmable logic to provide a byte
I/O-mapped messaging interface to standard microcontrollers.
The SMIC Interface is designed to support polled operation. Implementations can optionally provide an interrupt
driven from the BUSY bit, but this must not prevent driver software from using the interface in a polled manner.
This allows software to default to polled operation. It also allows software to use the KCS interface in a polled
mode until it determines the type of interrupt support. Methods for assigning and enabling such an interrupt are
outside the scope of this specification.
The specification of the SMIC interface registers is given solely with respect to the ‘system software side’ view of
the interface in system I/O space.
The functional behavior of the management controller to support the SMIC registers is specified, but the physical
implementation of the interface and the organization of the interface from the management controller side is
implementation dependent and is beyond the scope of this specification.
10.1 SMS Transfer Streams
The SMIC interface is designed to be interruptible to allow the one physical interface to be shared by two types of
system software: SMM (System Management Mode) software that runs from within an SMI Handler, and SMS
(System Management Software) that runs under the OS.
If an SMS transaction is interrupted, system management software will need to restart the Request/Response
transaction it had in progress.
To support this sharing, the interface provides mechanisms that allow system management software to detect that
its use of the interface has been interrupted. The protocol for messaging between SMM and the BMC over the
SMIC interface is implementation specific and not covered by this specification.
10.2 SMIC Communication Register Overview
The SMIC registers are mapped into system I/O space. This shared register space consists of three byte-wide
registers:
•
Flags Register - provides flags for use in various defined operations
•
Control/Status Register - accepts control codes and returns status codes
•
Data Register - provides a port for transactions that exchange message data
Message contents are passed through the data register. This includes the fields of a message, such as the Network
Function code, Command Byte, and any additional data required for the Request or Response message.
The control register is loaded with control code values that are used for framing the message data (indicating
message start, middle, and end) and for indicating message data transfer direction.
Status codes are returned through the control/status register. A control code is required to initiate for each data
byte transferred through the data register.
86
Intelligent Platform Management Interface Specification
The Flags register contains bits that indicate whether the controller has a message for system software, generated
an SMI, or is ready for a transfer operation. The Flags register also contains a special BUSY bit, that is used by
system software to initiate and handshake data byte transfers through the interface.
The SMIC interface is used as a polled interface. System software is always the “Master” for transfers between
system software and the BMC. The BMC can signal that data is available via bits in the flags register, but data
bytes will not be moved to or from the data register until the transaction is initiated by system software.
10.3 SMIC/BMC Message Interface Registers
The following figure illustrates the SMIC/BMC Interface Registers and register bits. These registers are located at
three consecutive 8-bit port addresses in I/O space.
The data, control/status, and flags registers appear at an I/O addresses 0CA9h, 0CAAh, and 0CABh,
respectively.
Reserved bits should be written as ‘0’ and ignored during reads. Software should not assume that a reserved bit
will return a constant value.
Figure 10-1, SMIC/BMC Interface Registers
7
flags
6
RX
DATA
READY
(ro)
TX
DATA
READY
(ro)
5
rsvd
4
SMI
(ro)
3
EVT
ATN
2
SMS
ATN
(ro)
(ro)
1
rsvd
0
BUSY
I/O address
base+2
(r/w)
control/status
(r/w)
base+1
data
(r/w)
base+0
10.3.1 Flags Register
System software always initiates the SMIC transfers, regardless of direction. The management controller uses
the SMS_ATN bit in the Flags register to indicate to system software that is has a message to be read. This bit
will be set whenever data is present in the Receive Message Queue.
Other bits in the Flags register are used to arbitrate access to the control/status and data registers between the
BMC and system software. Bits 7::2 are read-only from the system bus and write-only from the BMC; these
bits are used by the BMC as communication and status flags. Bit 0, the BUSY bit, is used as a semaphore for
coordinating access to the control/status and data registers between the system bus and the BMC. The following
table summarizes the functions of Flags Register bits.
87
Intelligent Platform Management Interface Specification
Table 10-1, SMIC Flags Register Bits
Bit
Name
Description
7
RX_DATA_RDY
6
TX_DATA_RDY
Indicates that the BMC has data that can be delivered in response to a
‘READ’ control code. RX_DATA_RDY must be ‘1’ before a control code that
causes a data byte to be transferred (read) from the BMC into the data
register can be issued. RX_DATA_RDY shall also be set to ‘1’ whenever a
‘READY’ status code is returned. RX_DATA_RDY should be ignored when
issuing control codes that do not cause a data byte to be read from the BMC.
Indicates that the BMC is ready to accept a ‘WRITE’ control code and data.
TX_DATA_RDY must be ‘1’ before a control code that causes data to be
transferred (written) to the BMC from the data register can be issued.
TX_DATA_RDY can be ignored when issuing control codes that do not cause
a data byte to be written to the BMC.
5
4
Reserved
SMI
3
EVT_ATN
2
SMS_ATN
1
0
Reserved
BUSY
Indicates that the BMC has asserted the SMI signal and has internal ‘SMI
event’ flags that are set.
Indicates that an Event Message has been received by the BMC and is ready
to be read from the Event Message Buffer in the BMC. If SMIs are used, this
flag may also set when an OEM message for the SMI Handler is available.
Indicates that the BMC has messages that are ready to be read from the
Receive Message Queue in the BMC. An implementation can use the 0-to-1
transition of this bit to provide an interrupt. Clearing the Receive Message
Queue clears this bit and clears ‘Receive Message Queue not empty’ as an
interrupt source.
Provides the arbitration mechanism for SMIC mailbox register access. This
bit is only set (1) from the system side and only cleared (0) by the BMC. The
system side sets the BUSY bit whenever it wishes to send a control code
(and data, if appropriate) to the BMC. The BMC acknowledges that it has
accepted and acted on the control code (and performed the data byte
transfer) by clearing the BUSY bit. An implementation can use the 1-to-0
transition of BUSY to provide an interrupt.
10.3.2 Control/Status Register
The Control/Status register is the destination for control codes written from the system bus, and Status codes
returned by the BMC.
SMS transfers have a specific numeric range for control codes and status codes. This provides a ‘stream ID’
that allows the BMC to tell whether SMS or some other message stream issued a control code. This also allows
system software to detect interruption by examining which the range of values for the most recent status code.
The control and status codes used for SMS transactions are defined in the control code and status code tables in
the sections following Section 10.9, SMIC Control and Status Code RangesSMIC Control and Status Code
Ranges.
10.3.2.1
Control and Status Codes
Message transfer control and framing codes (control codes) are transferred via the Control/Status register
while message content, such as command and data bytes, is transferred the Data register. control codes are
unique to each transfer stream and defined transaction and for each phase (beginning, intermediate, and end)
of a message.
Status Codes confirm the message phasing, identify the active stream, and provide error status.
When a message is transferred between system software and the BMC, each byte of the message that is
passed through the data register is accompanied by a control code written to the Control/Status register. The
BMC acknowledges reception of the control code and data byte by writing a corresponding Status Code to
88
Intelligent Platform Management Interface Specification
the Control/Status register before clearing the BUSY bit. Note that Status Codes are returned for each
transaction, regardless of whether a data byte is transferred or not.
10.3.3 Data Register
The message bytes for all requests (commands) and responses between system software and the BMC pass
through the Data register. The data register must only be written or read from the system side when the BUSY
bit is clear.
Messages to the BMC contain the same types of message body fields as messages on the IPMB. This includes
Network Function, Command byte, and Data fields.
10.4 Performing a single SMIC/BMC Transaction
The following steps describe how system software issues a control code to the BMC and transfers a data byte
through the SMIC interface.
1.
System software polls the BUSY bit of the Flags register until it reads back as cleared (0) by the BMC.
When the BUSY bit is 0, the BMC is ready to accept a new control code. System software is not
allowed to access the Control/status or Data Registers when the BUSY bit is high.
2.
System software writes the control code to the Control/Status register. See Section 10.9, SMIC Control
and Status Code RangesSMIC Control and Status Code Ranges and following, for control and status
code specifications.
If the transfer is a ‘Write’ transfer, and the control code is for ‘WR_NEXT’ or ‘WR_END’ operation,
system software waits for the TX_DATA_RDY bit to be set become set. This indicates that the BMC
is ready for the next write data byte. If the transfer is a ‘Read’ transfer, wait for the RX_DATA_RDY
bit to be set. (The exception to this is the ‘GET_STATUS’ control code, which though it causes a data
byte to be returned (the error code) does not require RX_DATA_RDY to be high first.
3.
If the transfer is a ‘Write’ transfer, system software loads the data to be written to the BMC into the
Data register.
4.
System software then initiates the operation by setting the BUSY bit. Setting the BUSY bit causes the
BMC to read the SMIC control code register and act on the control code. If the transfer is a Write
transfer, the BMC reads the data from the data register at this time. If the transfer is a Read transfer,
the BMC writes the data to the data register. The controller then returns a status code in the
control/status register. data or an error code in the data register (as appropriate), and clears the BUSY
bit.
5.
System software waits for the BUSY bit to clear, indicating the completion of the control code
operation.
6.
System software reads the Control/Status register for the completion status of the transaction. If the
operation was successful, the status code will reflect the next step in the transaction, or the successful
completion of the transaction. If the operation was not successful, the status code will be set to
‘READY’ and the data register will hold an error code.
89
Intelligent Platform Management Interface Specification
10.5 Performing a SMIC/BMC Message Transfer
Multiple transactions are required to transfer a message between system software and the BMC. In this case, a
message transfer refers to the sequence of steps required to transfer a series of data bytes to or from the BMC.
One control code transaction is required for each message data byte transferred via the SMIC interface.
A message transfer can be restarted at any time. Issuing an SMS_WR_START control code immediately aborts
any message transfer in progress and begins a new write transfer. Issuing WRITE_START control codes does
not require the RX_DATA_RDY or TX_DATA_RDY flags to be set.
The control code/status code sequences for the SMS-to-BMC transactions follows a “Transfer Start, Transfer
Middle, Transfer End” pattern:
•
Issue a ‘Start’ control code to start the transaction. Signifying ‘Transfer Start’
•
Issue ‘Next’ control codes to transfer the body of the data bytes. Signifying ‘Transfer Middle’. These
are either ‘Write_Next’ or ‘Read_Next’ control codes, dependent on the transfer direction.
•
Issue an ‘End’ control code to conclude the transaction and return the stream to the ‘Ready’ status.
This signifies ‘Transfer End’
The following summarizes these steps:
1.
If the transfer is a write transfer (system software to BMC), load the data register with the appropriate
data and issue the ‘Write Start’ control code for the stream. There is no need to check
TX_DATA_RDY.
If the transfer is a read transfer (a data byte transfer from the BMC to system software) wait for the
RX_DATA_RDY flag to become set, then issue the ‘Read Start’ control code for the stream.
2.
After each transaction, check the status code to see if the operation was successful. For write transfers,
the status code will generally be a ‘Write Next’, indicating that the interface is ready to accept more
data. For read transfers, the status code will typically be either a ‘Read Next’, indicating that there is
more data to be read, or a ‘Read End’ indicating that the last byte of data was transferred. If the
operation was aborted or an error occurred the transaction will need to be restarted from the beginning.
3.
Continue the transfer based on the status code. For read transfers, wait for the RX_DATA_RDY flag
and perform read operations until a ‘Read End’ status code is encountered (or an abort). For write
transfers, wait for the TX_DATA_RDY flag and perform write operations until you conclude the
transfer with a ‘Write End’ control code.
4.
Issue any additional control codes to return the transfer stream to the ‘Ready’ condition (indicated by
the ‘Ready’ status code). For read transfers from the SMM/SMS streams, this requires issuing a ‘Read
End’.
10.6 Interrupting Streams in Progress
Any software that interrupts a transfer in progress and switches to another stream is responsible saving and
restoring the status and data register values for the transaction that was in effect at the time of the interrupt. The
interrupting routine must first wait for the BUSY bit to clear and then save the control/status and data register
contents. Before exiting, the interrupting routine must wait for the BUSY bit to clear following its last transaction,
then restore the control/status and data register values. The interrupting routine can then perform its transfer(s).
After the interrupting routine concludes its last transaction, it must wait for BUSY to clear and restore the original
control/status and data register contents before returning from the interrupt.
90
Intelligent Platform Management Interface Specification
The following summarizes the steps for an interrupting routine, e.g. an SMI Handler:
1.
Poll the BUSY bit until cleared by the BMC.
2.
Save the contents of the Control/Status and Data registers.
3.
Perform the desired message transfers.
4.
Wait for the BUSY bit to clear, then restore the Control/Status and Data register values that were
saved in step 2, and return to the interrupted routine. If the interrupted routine has additional bytes to
transfer, the succeeding control code will be ‘out-of-phase’ with the state expected by the BMC. The
BMC will then return a ‘READY’ status code with an ‘Aborted’ return value. This indicates to the
interrupted routine that it needs to restart the transaction. If the interrupt happened to occur between
transfers, or on the last transfer of a transaction, the Control/Status and Data registers will have the
correct values and the interrupted routine will be able to start a new transaction.
10.7 Stream Switching
A stream switch occurs when the BMC receives a WR_START control code with a ‘stream ID’ that is different
than the stream ID for the previous control code. This is the mechanism that SMS uses to restart an interrupted
transaction. If a control code, other than WR_START is issued, and the stream does not match the stream ID for
the previous control code, the BMC shall return a ‘READY’ status code with an ‘Aborted’ return value.
10.8 DATA_RDY Flag Handling
The BMC shall set the TX_DATA_RDY whenever it is ready to accept a ‘WR_START’, ‘WR_NEXT’, or
‘WR_END’ control code that transfers a data byte from the SMIC data register. Note that system software does
not need to check for TX_DATA_RDY in order to issue a WR_START.
The BMC shall set the RX_DATA_RDY flag whenever it has a data byte that is ready to be requested with a
‘Read Start’, ‘Read Next’, or ‘Read End’ control code, or when it is prepared to return a ‘Ready’ status code.
The BMC shall set the RX_DATA_RDY flag whenever it returns a ‘Ready’ status code to the
stream. This includes when a stream is interrupted or when other errors occur during a transfer.
This is to ensure that a routine that may be spinning on the RX_DATA_RDY bit will proceed and
attempt its next transaction.
The BMC shall only deassert (0) the RX_DATA_RDY or TX_DATA_RDY flags while BUSY is asserted (1).
The BMC can assert the RX_DATA_RDY or TX_DATA_RDY flags any time that the associated conditions
become true.
91
Intelligent Platform Management Interface Specification
10.9 SMIC Control and Status Code Ranges
Specific Control Code ranges are used to identify transactions using the SMS transfer stream. This allows the
BMC to tell when the SMS stream is in use. Status Codes are returned by the BMC in the SMIC Control/Status
register to reflect the completion status of a previously issued control code. Like the control codes, the Status
Codes occupy a specific range for SMS transactions. Another set of control and status code ranges is reserved for
OEM / SMI Handler.
The presently defined ranges are:
•
40h-5Fh SMS (System Management Software) Transfer Stream Control Codes
•
C0h-DFh SMS (System Management Software) Transfer Stream Status Codes
•
60h-7Fh Available SMM (System Management Mode) / OEM Transfer Stream Control Codes
•
E0h-FFh Available SMM (System Management Mode) / OEM Transfer Stream Status Codes
All unspecified codes are reserved.
92
Intelligent Platform Management Interface Specification
10.10 SMIC SMS Stream Control Codes
Table 10-2, SMS Transfer Stream control codes
Code
Name
Description
SMS STREAM CONTROL CODES
40h
CC_SMS_GET_STATUS
41h
CC_SMS_WR_START
42h
CC_SMS_WR_NEXT
43h
CC_SMS_WR_END
44h
CC_SMS_RD_START
45h
CC_SMS_RD_NEXT
46h
CC_SMS_RD_END
47h-5Fh
reserved
Get status related to the SMS (system mgt. software) transfer stream. An
‘SC_SMS_RDY’ status code will be returned as the completion status for this
control code, along with the last error code for the stream in the data register.
Write the first message byte of an SMS write transfer. This is also used to
switch to the SMS stream. The SMIC data register must be loaded with the
data byte to be written to the BMC. The non-error completion status for this
control code will be an ‘SC_SMS_WR_START’ status code. The data register
contents will remain unaltered if no error occurred.
Write a ‘middle’ message byte in an SMS write transfer. The SMIC data
register must be loaded with the data byte to be written to the BMC. The user
must wait for the TX_DATA_RDY=1 before issuing the control code. The nonerror completion status for this control code will be an ‘SC_SMS_WR_NEXT’
status code. The data register contents will also remain unaltered if no error
occurred.
Indicates the last message byte for an SMS write transfer. The SMIC data
register must be loaded with the last data byte to be written to the BMC for the
current message. The user must wait for TX_DATA_RDY=1 before issuing
this control code. The non-error completion status for this control code will be
an ‘SC_SMS_WR_END’ status code with an error code of ‘00’ in the data
register, indicating ‘OK’.
Get the first byte of a read transfer from the BMC. The user must wait for
RX_DATA_RDY = 1 before issuing the control code. The non-error completion
status for this control code will be an ‘SC_SMS_RD_START’ status code in
the status register and the data byte in the data register.
Get a ‘middle’ message byte for an SMS read transfer. The user must wait for
RX_DATA_RDY = 1 before issuing the control code. The non-error completion
status for this control code will be an ‘SC_SMS_RD_NEXT’ status code in the
status register if there is more data to read or an ‘SC_SMS_RD_END’ status
code if the last byte was transferred, and the data byte in the data register.
Used to tell the BMC that the last byte of an SMS read transfer has been read
from the data register. It is not necessary to check the RX_DATA_RDY flag
before performing this operation. The non-error completion status for this
control code will be an ‘SC_SMS_RDY’ status code with an error code of ‘00’
in the data register, indicating ‘OK’.
reserved
93
Intelligent Platform Management Interface Specification
10.11 SMIC SMS Stream Status Codes
Table 10-3, SMS Transfer Stream Status Codes
Code
Name
Description
SMS STREAM STATUS CODES
C0h
SC_SMS_RDY
C1h
SC_SMS_WR_START
C2h
SC_SMS_WR_NEXT
C3h
SC_SMS_WR_END
C4h
SC_SMS_RD_START
C5h
SC_SMS_RD_NEXT
C6h
SC_SMS_RD_END
C7hDFh
reserved
94
BMC is ready for next SMS transfer. An error code is returned in the data register:
00 = NO ERROR
01 = UNSPECIFIED ERROR / ABORTED
02 = ILLEGAL or unexpected control code
03 = NO RESPONSE - response timeout. This will occur if the BMC cannot supply
a command response.
04 = ILLEGAL command. The request message is not recognized as being a legal
BMC request.
05 = BUFFER FULL. Attempt to write too many bytes to the BMC.
This status code indicates that the BMC has accepted first byte of a write transfer
and is ready for the next transaction.
The BMC has accepted next data byte of the write transfer, and is ready for the
next transaction.
The BMC has accepted the byte as being the last byte of the write transfer and is
ready for next SMS transfer. An error code is returned in the data register:
00 = NO ERROR
01 = ABORTED
02 = ILLEGAL or unexpected control code
03 = NO RESPONSE - response timeout. This will occur if the BMC cannot supply
a command response.
04 = ILLEGAL command. The request message is not recognized as being a legal
BMC request.
05 = BUFFER FULL. Last byte could not be accepted.
BMC has accepted the start of an SMS stream read transfer. The first data byte of
the read transfer is returned in the data register.
The BMC acknowledges a CC_SMS_RD_NEXT control code and is indicating that
there is more data to be read. The requested data byte is in the data register.
The BMC acknowledges a CC_SMS_RD_NEXT control code and is indicating that
there is no more data to be read. The last data byte is in the data register.
reserved
Intelligent Platform Management Interface Specification
10.12 SMIC Messaging
The SMIC message interface is essentially a ‘single master’ interface, where the ‘Master’ is the system software
on the system side of the interface. System software can only write Request Messages to the BMC, and can only
receive Response Messages from the BMC.
This does not mean that the system software cannot receive downstream ‘requests’ from the IPMB, or even the
BMC. Downstream requests can be ‘wrapped’ in a BMC Response Message. For example, a downstream request
could be placed in the Receive Message Queue where the data is retrieved using the Get Message command. The
Response Message would contain the downstream request data - which could then be extracted from the
Response Message by system software.
Since the SMIC interface is a ‘point-to-point’ connection, a ‘Requester’s ID’ is not required in a Request
Message to identify which physical interface to return a message response to. Only SMIC Event Request
messages include a Requester ID in the form of the Software ID field.
10.13 SMIC/BMC LUNs
LUN 00b is typically used for all messages to the BMC through the SMIC interface. LUNs 01b is reserved for
Receive Message Queue use and should not be used for sending other commands to the BMC. Note that messages
encapsulated in a Send Message command can use any LUN in the encapsulated portion.
10.14 SMIC-BMC Request Message Format
Request Messages are sent to the BMC from system software using a write transfer through the SMIC. The
message bytes are organized according to the following format specification:
Figure 10-2, SMIC/BMC Request Message Format
Byte 1
NetFn/LUN
Byte 2
Cmd
Byte 3:N
Data
Where:
LUN
Logical Unit Number. This is a sub-address that allows messages to be routed to different
‘logical units’ that reside behind the same physical interface. The LUN field occupies the least
significant two bits of the first message byte.
NetFn
Network Function code. This provides the first level of functional routing for messages received
by the BMC via the SMIC interface. The NetFn field occupies the most significant six bits of
the first message byte.
Cmd
Command code. This message byte specifies the operation that is to be executed under the
specified Network Function.
Data
Zero or more bytes of data, as required by the given command. The general convention is to
pass data LS-byte first, but check the individual command specifications to be sure.
95
Intelligent Platform Management Interface Specification
10.15 BMC-SMIC Response Message Format
Response Messages are read transfers from the BMC to system software via the SMIC. Note that the BMC only
returns responses via the SMIC interface when Data needs to be returned. The message bytes are organized
according to the following format specification:
Figure 10-3, SMIC/BMC Response Message Format
Byte 1
NetFn/LUN
Byte 2
Cmd
Byte 3
Completion Code
Byte 4:N
Data
Where:
LUN
Logical Unit Number. This is a return of the LUN that was passed in the Request
Message.
NetFn
Network Function. This is a return of the NetFn code that was passed in the Request
Message.
Cmd
Command. This is a return of the Cmd code that was passed in the Request Message.
Completion Code
The Completion Code indicates whether the request completed successfully or not.
Data
Zero or more bytes of data. The BMC always returns a response to acknowledge the
request, regardless of whether data is returned or not.
10.16 Logging Events from System Software via SMIC
The SMIC interface can be used for sending Event Messages from system software to the BMC Event Receiver.
The following figures show the format for SMIC Event Request and corresponding Event Response messages.
Note that only Event Request Messages to the BMC via the SMIC interface have a Software ID field. This is so
the Software ID can be saved in the logged event.
Figure 10-4, SMIC Event Request Message Format
NetFn
LUN
(04h = Sensor/Event Request)
(00b)
EvMRev
Sensor Type
Sensor #
Command
(02h = Platform Event)
Event Dir
Event Type
Software ID (Gen ID), 7-bits
1
(20h-2Fh = system sw)
Event Data 1
Event Data 2
Shading designates fields that are not stored in the event record.
Figure 10-5, SMIC Event Response Message Format
NetFn
(05h = Sensor/Event Response)
96
00
Command
(02h = Platform Event)
Completion Code
Event Data 3
Intelligent Platform Management Interface Specification
97
Intelligent Platform Management Interface Specification
11. Block Transfer (BT) Interface
This section describes the Block Transfer (BT) Interface. The BT interface is one of the supported BMC to SMS
system interfaces. The BT interface is specified for SMS or OEM Defined messages. Messaging between the BMC
and an SMI Handler is not specified for this interface.
The BT Interface is so named because an entire block of message data is buffered before the management controller
is notified of available data. This is different from the SMIC and KCS interfaces, which are byte-transfer oriented.
A BT Interface Capabilities command provides supplementary information about extended buffer sizes and other
elements of the interface.
The host side of the BT Interface is designed for interrupt or polled operation. Implementations can elect to provide
a system interrupt from the assertion of the B2H_ATN or SMS_ATN (BMC-to-Host attention or System
Management Software attention) states. Note that implementing an interrupt must not preclude driver software from
the using the interface in a polled manner.
The BT Interface is designed for efficient interrupt operation via assertion of H2B_ATN by the host.
Provision for operation in a polled mode is optional.
Methods for assigning, enabling, and determining the system interrupt are outside the scope of this specification.
The BT interface provides support for implementations that allow the submission and asynchronous completion of
commands.
11.1 BT Interface-BMC Request Message Format
Request Messages are sent to the BMC from system software using a write transfer through the BT Interface. The
message bytes are organized according to the following format specification:
Figure 11-1, BT Interface/BMC Request Message Format
Byte 1
Length
Byte 2
NetFn/LUN
Byte 3
Seq
Byte 4
Cmd
Byte 5:N
Data
Where:
98
Length
This is not actually part of the message, but part of the framing for the BT Interface. This value
is the 1-based count of message bytes following the length byte. The minimum length byte
value for a command to the BMC would be 3 to cover the NetFn/LUN, Seq, and Cmd bytes.
LUN
Logical Unit Number. This is a sub-address that allows messages to be routed to different
‘logical units’ that reside behind the same physical interface. The LUN field occupies the least
significant two bits of the first message byte.
NetFn
Network Function code. This provides the first level of functional routing for messages received
by the BMC via the BT Interface. The NetFn field occupies the most significant six bits of the
first message byte.
Seq
Used for matching responses up with requests. The BT interface can support interleaved ‘multithreaded’ communications. There can be multiple simultaneous outstanding requests from SMS
with responses returned asynchronously (and in any order). The Requester (SMS) sets the value
for this field. The Responder returns the value in the corresponding response. The Seq field is
used in combination with the NetFn and Command fields to form a unique value. I.e. the same
Seq value could be used in multiple outstanding requests, as long as the combinations of Seq
value, NetFn, and Command were unique among the requests.
Intelligent Platform Management Interface Specification
Cmd
Command code. This message byte specifies the operation that is to be executed under the
specified Network Function.
Data
Zero or more bytes of data, as required by the given command. The general convention is to
pass data LS-byte first, but check the individual command specifications to be sure.
11.2 BMC-BT Interface Response Message Format
Response Messages are read transfers from the BMC to system software via the BT Interface. Note that with a
few exceptions (e.g., Cold Reset command) the BMC always returns response to a request delivered via the BT
interface in order to deliver the completion code, regardless of whether the response has data in the Data field.
The message bytes are organized according to the following format specification:
Figure 11-2, BT Interface/BMC Response Message Format
Byte 1
Length
Byte 2
NetFn/LUN
Byte 3
Seq
Byte 4
Cmd
Byte 5:N
Completion Code
Byte 6:N
Data
Where:
Length
This is not actually part of the message, but part of the framing for the BT Interface.
This value is the 1-based count of message bytes following the length byte. The
minimum length byte value for a response from the BMC would be 4 to cover the
NetFn/LUN, Seq, Cmd, and Completion Code bytes.
LUN
Logical Unit Number. This is a return of the LUN that was passed in the Request
Message.
NetFn
Network Function. This is a return of the NetFn code that was passed in the Request
Message.
Seq
Used for matching responses up with requests. The BT interface can support
interleaved ‘multi-threaded’ communications. There can be multiple simultaneous
outstanding requests from SMS with responses returned asynchronously (and in any
order). The Requester (SMS) sets the value for this field. The Responder returns the
value in the corresponding response. The Seq field is used in combination with the
NetFn and Command fields to form a unique value. I.e. the same Seq value could be
used in multiple outstanding requests, as long as the combinations of Seq value,
NetFn, and Command were unique among the requests.
Cmd
Command. This is a return of the Cmd code that was passed in the Request Message.
Completion Code
The Completion Code indicates whether the request completed successfully or not.
Data
Zero or more bytes of data. The BMC always returns a response to acknowledge the
request, regardless of whether data is returned or not.
11.3 Using the Seq Field
System Management Software is expected to use the Seq field in the following manner. SMS maintains a list of
the outstanding requests it has sent. This list holds the Seq, NetFn, and Command values that were used to send
the request. There should be one entry in the list for each possible simultaneous outstanding request. When SMS
generates a Seq value for a new request, it must ensure that the combination of Seq, Command, and NetFn values
do not match any entries already in the outstanding request list.
When a response is received from the BMC, SMS looks for a match between the Seq value, Command, and
NetFn values in the response and an entry in the outstanding request list. If there is a match, the response is
99
Intelligent Platform Management Interface Specification
processed normally and the outstanding request list entry freed for a new request. If the response does not match,
the response can be ignored or passed on to error tracking procedures.
11.4 Response Expiration Handling
It is possible that conditions could occur where a response will not return for a given request. The Seq number
associated with the request must be freed so it can be reused. To support this, SMS should implement a response
expiration interval.
The BMC must return a response within the specified response time seconds (per the Get BT Interface
Capabilities command). If the response is not received in this time the corresponding entry in the SMS
outstanding response list can be cleared. If retries are not recommended at the interface, a missing response
constitutes an immediate error condition. If the interface recommends retries (per the Get BT Interface
Capabilities command) SMS should retry the request up to the specified count. If the response is still not
provided, an error has occurred.
The typical BT Interface is expected to be fundamentally reliable without retries. The retry
specification is to support possible commands within the controller that may occasionally exceed
the Request-to-Response specification. An application can elect to implement retry counts that
exceed the recommendation.
The BMC must not return a given response once the corresponding Request-to-Response interval has passed. The
BMC can ensure this by maintaining its own internal list of outstanding requests through the interface. The BMC
could age and expire the entries in the list by expiring the entries at an interval that is somewhat shorter than the
specified Request-to-Response interval. The BMC can define its own internal Seq value or tracking number for
this purpose, or it could use the Seq, NetFn, and Command values in the same manner as SMS.
11.5 Logging Events from System Software via BT Interface
The BT Interface can be used for sending Event Messages from system software to the BMC Event Receiver. The
following figures show the format for BT Interface Event Request and corresponding Event Response messages.
Note that only Event Request Messages to the BMC via the BT Interface have a Software ID field. This is so the
Software ID can be saved in the logged event.
Figure 11-3, BT Interface Event Request Message Format
Length
EvMRev
NetFn
(04h = Sensor/Event Request)
Sensor Type
Sensor #
LUN
(00b)
Event Dir
Seq
Command
(02h = Platform Event)
Event Type
Event Data 1
Software ID (Gen ID)
1
7-bits
Event Data 2
Event Data 3
Shading designates fields that are not stored in the event record.
Figure 11-4, BT Interface Event Response Message Format
Length
100
NetFn
(05h = Sensor/Event Response)
00
Seq
Command
(02h = Platform Event)
Completion Code
Intelligent Platform Management Interface Specification
11.6 Host to BMC Interface
The Host interface to the baseboard management controller (BMC) requires a block of 3 contiguous I/O locations
on the system board. (A reference implementation fixes this at locations E4h:E6h. The interface circuitry will
decode the lower 2 address lines, SA[1..0] ). A general-purpose chip select will be used to generate the select line
for the interface, which is to reside in system I/O space. The I/O address offsets are defined as follows:
Table 11-1, BT Interface Registers
Offset
0
1
2
Read
Write
BT_CTRL - control register
BMC2HOST buffer
HOST2BMC buffer
BT_INTMASK - interrupt mask register
The two buffers must meet the specified maximum message size requirements for all protocols supported on the
messaging channels implemented on the BMC. Implementations can choose to provide more depth optionally. The
GET_BT_INTERFACE_CAPABILITIES command is used to query for the actual implementation buffer depth.
The messaging protocol involves the host writing the command stream to the BT buffer, followed by setting a
“attention” bit in the BT control register. This automatically generates an interrupt to the baseboard management
controller (BMC). The BMC then reads command packet from the BT buffer, and clears the attention bit. After
processing the command, the BMC then writes the response data to the host-bound buffer. Finally, the BMC sets an
outbound attention bit and generates an interrupt to the host (the host may optionally poll the attention bits, and may
enable/disable the interrupts via a MASK register). Refer to Section 11.7 for a walk-through of the sequence of
operations used for transfers on the BT interface.
There is no explicit requirement or recommendation for the hardware used to implement the interface. A discrete,
custom, programmable array, or other implementation may be used at the discretion of the designer. As an example,
some implementations have used a Xilinx* XC4003E Field Programmable Gate Array (FPGA) to implement the
interface circuit because it provides on-chip user RAM that can be effectively used to implement the interface’s
buffers. This implementation was able to provide 64-byte buffers.
11.6.1 BT Host Interface Registers
The Host BT interface provides an independent set of registers and interrupts to allow the Host driver to
communicate with the baseboard management controller without conflicting with the O/S ACPI driver.
11.6.2 BT BMC to Host Buffer (BMC2HOST)
From the host side, this is a read-only buffer, which contains a command response stream from the embedded
controller. The buffer must be a minimum of 64-bytes deep. This shares offset 1 of the I/O space with the
HOST2BMC buffer. Hence I/O read cycles from the host CPU remove data from this buffer, whereas write cycles
from the BMC load data into this buffer.
11.6.3 BT Host to BMC Buffer (HOST2BMC)
From the host side, this is a write-only buffer to which the host writes a command stream to the baseboard
management controller. The buffer must be a minimum of 64-bytes deep. This shares offset 1 of the I/O space with
the BMC2HOST buffer. Hence an I/O write cycles from the host CPU load data into this buffer, whereas read
cycles from the BMC remove data from this buffer.
101
Intelligent Platform Management Interface Specification
11.6.4 BT Control Register (BT_CTRL)
The host and the BMC use this register for various control functions defined below.
Figure 11-5, BT_CTRL Register format
7
B_BUSY
6
H_BUSY
5
OEM0
4
EVT_ATN
3
B2H_ATN
2
H2B_ATN
1
CLR_RD_PTR
0
CLR_WR_PTR
Table 11-2, BT_CTRL Register Bit Definitions
BIT
R/W*
By
Host
R/W*
By
BMC
0
W
W
CLR_WR_PTR
1
W
W
CLR_RD_PTR
2
R/S
Write 1 to
set bit;
0 no effect
R/C
Write 1 to
clear bit;
0 no effect
H2B_ATN
Reset State=0
3
R/C
Write 1 to
clear bit;
0 no effect
R/S
Write 1 to
set bit;
0 no effect
B2H_ATN
Reset State=0
4
R/C
Write 1 to
clear bit:
0 no effect
R/S
Write 1 to
set bit;
0 no effect
SMS_ATN
Reset State=0
102
NAME
FUNCTION
Clear Write Pointer. The host writes a 1 to clear the write pointer to
the BT HOST2BMC buffer; this bit is always read back as 0.
Writing a 0 has no effect. Similarly, the BMC writes a 1 to clear the
write pointer to the BT BMC2HOST buffer; this bit is always read
back as 0. Writing a 0 has no effect. Clearing the pointer is defined
as moving it to point to the start of the next valid buffer (typically
the top of a single FIFO buffer).
Clear Read Pointer. The host writes a 1 to clear the read pointer to
the BT BMC2HOST buffer; this bit is always read back as 0.
Writing a 0 has no effect. Similarly, the BMC writes a 1 to clear the
read pointer to the BT HOST2BMC buffer; this bit is always read
back as 0. Writing a 0 has no effect. Clearing the pointer is defined
as moving it to point to the start of the next valid buffer (typically
the top of a single FIFO buffer).
Host to BMC Attention. When the host writes a 1 to this bit, an
interrupt is generated to the baseboard management controller.
The host should set this bit when it has completed writing a
message stream to the HOST2BMC buffer. The baseboard
management controller clears this bit after it has set the B_BUSY
bit. The host may poll the H2B_ATN bit to determine that the
baseboard management controller has acknowledged the
command. The capability to operate in a polled mode by the BMC
is optional.
BMC to Host Attention. The BMC sets this bit when it has
completed writing a message response stream to the BMC2HOST
buffer. The host may poll the B2H_ATN bit to determine that the
baseboard management controller has finished writing a message
response stream to the BMC2HOST buffer. After setting H_BUSY,
the host should clear this bit to acknowledge receipt of the
message response. This bit can be enabled to generate an
interrupt to the host by setting the B2HI_EN bit in the INTMASK
register.
SMS Attention. The BMC sets this bit when it has detected and
queued an SMS message that must be reported to the host. This
allows the host to distinguish between command responses and
SMS messages from the baseboard management controller. This
bit can be enabled to generate an interrupt to the host by a host set
of the B2HI_EN bit in the INTMASK register. The host clears this
bit by writing a 1 to it.
Intelligent Platform Management Interface Specification
BIT
R/W*
By
Host
R/W*
By
BMC
5
R/S
Write 1 to
set bit;
0 no effect
R/C
Write 1 to
clear bit:
0 no effect
OEM0
Reset State=0
R/S/C
R
H_BUSY
Reset State=0
R/S/C
B_BUSY
Reset State=1
6
Write 1 to
toggle
7
R
Write 1
to toggle
NAME
FUNCTION
Reserved for definition by platform. Generic IPMI software must
write this bit as 0, and ignore the value on read. The OEM0 bit
should be able to generate an interrupt to the BMC when written by
the host but is not required (polled mode is acceptable). Typical
usage is a “heartbeat” mechanism from/to the host; the host sets
OEM0 to interrupt the BMC and then polls this bit to be cleared
(BMC is alive and responded to the interrupt). The BMC FW
completes the acknowledge cycle by clearing OEM0 upon receipt
of the interrupt (host is alive).
Host Busy. This bit is set/cleared by the Host to indicate that it is
busy processing response/event data from the BMC or cannot
accept response/event data at this time. It is set to 1 if the host
writes a 1 when H_BUSY=0, cleared if the host writes a 1 when
H_BUSY=1; there is no effect if the host writes a 0 to this bit
(toggle implementation). The BMC will need to verify that this bit is
cleared before sending a response or event message.
Baseboard Management Controller Busy. This bit is set/cleared
by the BMC to indicate that it is busy processing command/request
data from the Host or cannot accept command/request data at this
time. It is set to 1 if the BMC writes 1 when B_BUSY=0, cleared if
the BMC writes 1 when B_BUSY=1; there is no effect if the BMC
writes 0 to this bit (toggle implementation). . The initial state of
this bit should be set to 1 so that the BMC side driver can initialize
and prepare to accept Host traffic before the Host attempts to use
it the first time.
* R=read; W=write; S=set; C=clear
103
Intelligent Platform Management Interface Specification
11.6.5 BT Interrupt Mask Register (INTMASK)
This register is used by the host to control which interrupts can be generated by the baseboard management
controller.
Figure 11-6, BT_INTMASK Register format
7
BMC_HWRST
6
rsvd
5
rsvd
4
OEM3
3
OEM2
2
OEM1
1
B2H_IRQ
0
B2H_IRQ_EN
Table 11-3, BT_INTMASK Register Bit Definitions
BIT
R/W
NAME
FUNCTION
0
R/W
B2H_IRQ_EN
1
R/W
B2H_IRQ
2
R/W
OEM1
3
R/W
OEM2
4
R/W
OEM3
5
6
7
R/W
R/W
R/W
Reserved
Reserved
BMC_HWRST
BMC to HOST Interrupt Enable. The interrupt is generated by the BMC-BT
interface if B2H_IRQ_EN is set (1) and either the B2H_ATN or EVT_ATN bits are
set by the BMC.
BMC to HOST Interrupt Active. This bit reflects the state of the interrupt line to the
host, and therefore can only become set (1) if by B2H_IRQ_EN is set and the
interrupt condition has occurred.
On a read: 0 = interrupt to host not active; 1 = interrupt to host active
On a write: 0 = no effect; 1 = clear interrupt (this is the source of the INT, and is
immediately cleared by the O/S driver). This only clears the interrupt for the
system interface. Other interrupts may require clearing flags internal to the BMC.
If bit is 0, then a rising edge on B2H_ATN or EVT_ATN sets this to 1. If already 1,
then no affect.
Reserved for definition by platform manufacturer for BIOS/SMI Handler use.
Generic IPMI software must write this bit as 0, and ignore the value on read.
Reserved for definition by platform manufacturer for BIOS/SMI Handler use.
Generic IPMI software must write this bit as 0, and ignore the value on read.
Reserved for definition by platform manufacturer for BIOS/SMI Handler use.
Generic IPMI software must write this bit as 0, and ignore the value on read.
Reserved for future definition by IPMI. Write as 0, ignore value on read.
Reserved for future definition by IPMI. Write as 0, ignore value on read.
Host to Baseboard Management Controller Reset. (OPTIONAL)
Always read back as zero. Writing a 1 to this bit will cause a hardware reset of the
BMC. This is non-sticky; writing zero has no effect. This bit, if provided, is intended
for to be used for error recovery by the host if loss of communication with the BMC
occurs.
104
Intelligent Platform Management Interface Specification
11.7 Communication Protocol
In the context of the BT Interface, the term Write Transfer refers to the Host writing data to the BMC, while Read
Transfer refers to the Host reading data from the BMC.
If the interface implementation supports multithreaded operation, the interface driver should always be looking
for the B2H_ATN or EVT_ATN condition. In an interrupt driven implementation, this means the interrupt
handler should always check for responses or asynchronous requests. In a polled implementation, the driver
should periodically poll the state of these bits.
Table 11-4, BT Interface Write Transfer
Operation
Host
Start
“Command”
(Write
Transfer)
Wait for B_BUSY clear (BMC
ready to accept a request) &
H2B_ATN clear (signifying
acknowledge of previous
command)
Write 1 to CLR_WR_ PTR bit in
BT_CNTRL (reset pointer to start
of buffer)
Write bytes 1 to n of command
(request) to HOST2BMC buffer
Set H2B_ATN attention (tell BMC
that write data is available)
BMC
H2B_
ATN
B2H_
ATN
B_BUSY
H_BUSY
Enable host interface (Clear
B_BUSY)
Wait for H2B_ATN
(indicating data has been
loaded into HOST2BMC
buffer)
0
0
1
0
0
0
0
0
“
0
0
0
0
“
0
0
0
0
“
1
0
0
0
Set B_BUSY (indicating
BMC is preparing to transfer
data from the HOST2BMC
buffer)
Clear H2B_ATN (the ACK)
Read HOST2BMC buffer
Clear B_BUSY (indicating
BMC is done transferring
data)
Process command
1
0
1
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
105
Intelligent Platform Management Interface Specification
Table 11-5, BT Interface Read Transfer
Operation
Host
BMC
“Response”
(Read
Transfer)
Wait for B2H_ATN attention to be set
(or wait for interrupt from BMC),
signaling BMC has data available for
Host.
“
“
Set H_BUSY (indicating Host is in
process of reading data from the
interface)
Clear B2H_ATN
Write 1 to CLR_RD_PTR bit in
BT_CTRL
Read bytes 1 to n of response phase
from BMC2HOST buffer
Clear H_BUSY (indicating Host has
completed reading data from the
buffer)
H2B_
ATN
B2H_
ATN
B_ BUSY
H_BUSY
Waits for H_BUSY to
be cleared
0
0
0
0
Write bytes 1 to n of
response to
BMC2HOST buffer
Set B2H_ATN
(indicating BMC has
put data in BMC2HOST
buffer)
Wait for B2H_ATN
clear (ACK of BMC
response message)
“
“
0
0
0
0
0
1
0
0
0
1
0
1
0
0
0
0
0
0
1
1
“
0
0
0
1
“
0
0
0
0
“
0
0
0
0
Idle
11.8 Host and BMC Busy States
The host and BMC can set H_BUSY and B_BUSY, respectively, as necessary to indicate they are not able to accept
response/event or command data from the BMC or host, respectively for any reason. This allows for asynchronous
housekeeping functions that might take an extended period of time (seconds or minutes) to be accomplished in a
controlled manner and minimize the chance of getting out of synchronization - which might occur if the host or
BMC "timed out" and assumed the other side was hung or not responding.
11.9 Host Command Power-On/Reset States
The BMC sets B_BUSY to 1 whenever it is initializing from a cold reset and following BMC power up. The
interface will initialize with H_BUSY, H2B_ATN, and B2H_ATN set to 0 (reset state = 0).
106
Intelligent Platform Management Interface Specification
12. IPMI LAN Interface
This section describes the mechanisms specific to transferring IPMI messages between the BMC and a remote
management system (remote console) over an Ethernet LAN connection using UDP under IPv4. The UDP
datagrams are formatted to contain IPMI request and response messages, plus additional messages for discovery and
authentication.
While an IPMI LAN interface can be accomplished using a LAN Controller that is dedicated to the BMC, it will
usually be accomplished using LAN Controller that can be shared for both BMC and system use.
There are two implementations that are likely to be used to deploy an IPMI LAN Interface using a shared LAN
controller. The first implementation is using an embedded LAN controller, as shown in Figure 12-1Figure 12-1,
and the second is using a LAN controller on an add-in card, as shown in Figure 12-2Figure 12-2.
Both examples show a LAN Controller that has the capability to detect UDP datagrams sent to a ‘management port’.
Any datagrams received on that port are forwarded to a ‘side-band’ interface that allows them to be delivered to, or
retrieved by, the BMC. As Figure 12-1Figure 12-1 shows, these incoming ‘platform management’ datagrams may
also be delivered to system software in parallel with being delivered to the BMC.
The BMC can use this same interface to inject datagrams onto the LAN. These datagrams are interleaved with the
network packets that are generated by system software.
The LAN Controller can be designed in such a way that the interface for the ‘management port’ is powered by
standby power and remains operative even when the system is powered down. This provides a mechanism that
allows IPMI LAN messaging to occur independent from system software and the system’s power state. A LAN
controller dedicated to the BMC can also be used.
Figure 12-1, Embedded LAN Controller Implementation
Managed System
UDP datagrams to
'Mgmt. Port'
Remote
Management
System
Datagrams
gen'd by
BMC
LAN
LAN
Controller
PCI
Outgoing packets
from system
software
'side band'
connection.
E.g. SMBus
or I2C
SEL,
SDR,
FRU
BMC
Satellite
Controller
IPMB
System Bus
All incoming
packets
Figure 12-2Figure 12-2 shows an implementation where the LAN Controller is implemented as a PCI add-in card
connected to the BMC via a PCI Management Bus connection. This approach avoids the need to have the LAN
Controller built-into the system, allowing the LAN Controller portion of the IPMI LAN Interface to be added or
updated at a later time.
107
Intelligent Platform Management Interface Specification
Figure 12-2, PCI Management Bus Implementation
LAN
Controller A
LAN
Controller B
Add-in Card
PCI
SMBus
IPMB
BMC
PCI
System Bus
12.1 RMCP
The Distributed Management Task Force (DMTF) has defined a ‘Remote Management Control Protocol’
(RMCP) for supporting pre-OS and OS-absent management. RMCP is a simple request-response protocol that
can be delivered using UDP datagrams. IPMI-over-LAN uses version 1 of the RMCP protocol and packet format.
RMCP includes a field that indicates the class of messages that can be embedded in an RMCP message packet,
including a class for IPMI messages. Other message classes are ‘ASF’ and ‘OEM’.
IPMI LAN messages are encapsulated in RMCP packets using the IPMI message class. An IPMI LAN
implementation can also use ASF-class ‘Ping’ and ‘Pong’ messages to support the discovery of IPMI managed
systems on the network.
12.1.1 ASF Messages in RMCP
The term ‘ASF’ is commonly used in RMCP. ASF originally stood for ‘Alerting Standard Forum’. This is the
original name of a group that has moved into the DMTF as the Pre-OS Working Group. The group is
standardizing a set of messages under RMCP that are oriented towards non-intelligent management hardware
supporting basic LAN Alerting and recovery control (e.g. system reset) capabilities via the LAN.
RMCP uses ‘ASF’ to denote the fields and values that support these messages.
108
Intelligent Platform Management Interface Specification
12.1.2 RMCP Port Numbers
RMCP uses two well-known ports under UDP. The following table describes these ports and summarizes their
use.
Table 12-1, RMCP Port Numbers
Port #
Name
Description
623
(26Fh)
Aux Bus Shunt
(Primary RMCP
Port)
Hereon referred to as the Primary RMCP Port - This port and the
required RMCP messages must be provided to be conformant with the
RMCP specifications.
Secure Aux Bus
(Secondary
RMCP Port)
There is a mandatory set of messages that are required to be supported
on this port. These messages are always sent ‘in the clear’ so that
system software can discover systems that have RMCP support.
Hereon referred to as the Secondary RMCP Port or Secure Port. This
port is only used when it is necessary to encrypt packets using an
algorithm or specification that prevents also sending unencrypted
packets from being transferred via the same port. Since discovery
requires sending ‘in the clear’ RMCP Ping/Pong packets, the secondary
port is used to transfer encrypted transfers while the primary port
continues to support unencrypted packets.
664
(298h)
An implementation that utilizes this port must still support the Primary
RMCP Port and the required messages on that port in order to be
conformant with the RMCP specifications.
Note that the common IPMI messaging protocols and authentication
mechanisms in this specification do not use encrypted packets,
therefore IPMI messaging does not need to use the secondary port.
109
Intelligent Platform Management Interface Specification
12.1.3 RMCP Message Format
There are two types of RMCP messages: Data or ‘Normal’ RMCP messages, and RMCP Acknowledge
Messages. Data messages and ACK messages are differentiated by the ACK/normal bit of the Class of Message
field.
Table 12-2, RMCP Message Format
size in
bytes
Field
RMCP Header
Version
Reserved
Sequence Number
Class of Message
1
1
1
1
RMCP Data
Data
Description
06h = RMCP Version 1.0
00h
varies, see text
This field identifies the format of the messages that follow this header.
All messages of class ASF (6) conform to the formats defined in this
specification and can be extended via an OEM IANA.
Bit 7 RMCP ACK
0 - Normal RMCP message
1 - RMCP ACK message
Bit 6:5 Reserved
Bit 4:0 Message Class
0-5 = Reserved
6 = ASF
7 = IPMI
8 = OEM defined
all other = Reserved
Variable data based class of message
The following table presents how the ACK/Normal Bit and the Message Class combine to identify the type of
message under RMCP and which specification defines the format of the associated message data.
Table 12-3, Message Type Determination Under RMCP
ACK/Normal
bit
Message
Class
Message Type
Message Data
ACK
ASF
RMCP ACK
ACK
normal
normal
all other
ASF
OEM
undefined
ASF Messages
OEM Message
under RMCP
No Data. Message just contains RMCP Header with the
Sequence Number set to the sequence number from the
last message that was received.
not allowed
Per ASF Specification
bytes 0:3 = OEM IANA
normal
IPMI
IPMI Messages
bytes 4:N = OEM Message Data (defined by manufacturer
or organization identified by the OEM IANA field value)
Per this specification
12.2 Required ASF/RMCP Messages for IPMI-over-LAN
The following class=ASF messages under RMCP must be supported in a system implementing the IPMI LAN
interfaces over TCP/IP-UDP. This is just a specification of the minimum ASF message support required for
IPMI LAN implementations. IPMI LAN messaging can coexist with additional ASF messaging on a system.
110
Intelligent Platform Management Interface Specification
Therefore, a system can support additional ASF messages and functions without being non-conformant with the
IPMI LAN specifications.
There is no IPMI requirement for the BMC to respond to RMCP Messages of class=ASF other than the RMCP
Ping message. However, additional message support may be required if the system is also to be conformant
with the ASF specification. Refer to [ASF].
Table 12-4, ASF/RMCP Messages for IPMI-over-LAN
Message
Description
RMCP ACK
RECOMMENDED.
Per the ASF specifications the RMCP ACK
message should be returned whenever a ‘normal’
RMCP message with an RMCP sequence number
of 0-254 is received. This is recommended, but not
required for a system to be conformant with the
IPMI LAN specification. (Note, however, that a
system that does not return the ACK is not fully
conformant with the RMCP specification). See
sections 12.2.1, RMCP ACK MessagesRMCP ACK
Messages, and 12.2.2, RMCP ACK
HandlingRMCP ACK Handling for more
information.
REQUIRED.
This message returns information about the
interfaces supported via RMCP. It is used both to
discover managed systems that support RMCP
and to determine whether the system supports
IPMI LAN messaging and/or additional ASF
commands.
This message must be supported on the Primary
RMCP port.
REQUIRED.
This message must be returned from the managed
system in response to the Presence Ping message
on the Primary RMCP port.
ASF Presence Ping message
ASF Presence Pong Message (Ping response)
12.2.1 RMCP ACK Messages
Table 12-5, RMCP ACK Message Fields, shows the RMCP header and data values for the RMCP ACK
message. This message is used to acknowledge receipt of a ‘normal’ RMCP messages that were transmitted
with a 0-254 RMCP sequence number. RMCP ACK messages are not generated if the RMCP sequence number
is 255 (FFh). The RMCP ACK message does not indicate that an action has been completed, only that a
specific RMCP packet has been received.
The RMCP ACK operation is defined as being symmetric. That is, any party that receives a normal RMCP
message with a 0-254 RMCP sequence number is supposed to respond with an RMCP ACK message. Thus,
RMCP ACK messages can be generated by remote consoles and managed systems.
Table 12-5, RMCP ACK Message Fields
Field
Value
Version
Reserved
Sequence Number
Class of Message
Copied from received message.
Copied from received message.
Copied from received message.
7 Set to 1 to indicate ‘ACK’ packet
6:0 Copied from received message.
none
RMCP Data
111
Intelligent Platform Management Interface Specification
12.2.2 RMCP ACK Handling
RMCP ACK messages are not required for IPMI messaging, since IPMI already has its own messaging retry
policies. In addition, some Network Controllers usable for IPMI messaging do not automatically generate
RMCP ACK messages. In these implementations, the BMC would have to generate the RMCP ACK, resulting
in additional, unnecessary traffic from the BMC. Therefore, RMCP ACK messages should not be used for IPMI
messaging. This leads to the following requirements and recommendations:
•
RMCP messages with class=IPMI must have their RMCP sequence number set to 255 (FFh) to indicate
that RMCP ACK messages are not to be generated by the message receiver.
•
Console software should also set the RMCP sequence number to 255 (FFh) for non-IPMI messages,
whenever possible. Some systems may not respond with an RMCP ACK for non-IPMI messages even if
one was requested using a 0-254 RMCP sequence number. Console software should be prepared for this
occurrence. The software can discover which systems support RMCP ACK by checking to see whether
RMCP ACKs are generated as the result of sending RMCP Presence Ping messages. If RMCP ACKs are
not received, the software should proceed without requiring RMCP ACK messages.
•
Regardless of whether RMCP ACK messages are received from a system, console software should still
send RMCP ACKs whenever it receives an RMCP message with a 0-254 RMCP sequence number.
12.2.3 RMCP/ASF Presence Ping Message
This message returns information about the interfaces supported via RMCP. It is used both to discover managed
systems that support RMCP and to determine whether the system supports IPMI LAN messaging and/or
additional ASF commands. The following table illustrates the specific fields to be used for Presence Ping
Message to a system implementing IPMI LAN messaging.
Table 12-6, RMCP Packet Fields for ASF Presence Ping Message (Ping Request)
size in
bytes
Field
UDP Header
RMCP Header
ASF Message
Source Port
Destination Port
UDP Length
UDP Checksum
Version
Reserved
RMCP Sequence Number
2
2
2
2
1
1
1
Class of Message
IANA Enterprise Number
Message Type
Message Tag
1
4
1
1
Reserved
Data Length
1.
112
1
1
Value
per UDP
26Fh
per UDP
per UDP
06h = RMCP Version 1.0
00h
0-254 if RMCP ACK desired. 255 for no
RMCP ACK. See sections 12.2.1, RMCP
ACK MessagesRMCP ACK Messages, and
12.2.2, RMCP ACK HandlingRMCP ACK
Handling for more information.[1]
06h for ASF
4542 (ASF IANA)
80h = Presence Ping
0-FEh, generated by remote console.
This is an RMCP version of a sequence
number. Values 0-254 (0-FEh) are used
for RMCP request/response messages.
255 indicates the message is unidirectional
and not part of a request/response pair.
00h
00h
Some systems may not generate RMCP ACKs even if requested. Software should be designed to
handle this occurrence.
Intelligent Platform Management Interface Specification
12.2.4 RMCP/ASF Pong Message (Ping Response)
This message must be returned from the managed system in response to the Presence Ping message.
Table 12-7, RMCP Packet Fields for ASF Presence Pong Message (Ping Response)
size in
bytes
Field
UDP Header
RMCP Header
ASF Message
Source Port
Destination Port
UDP Length
UDP Checksum
Version
Reserved
RMCP Sequence Number
Class of Message
IANA Enterprise Number
Message Type
Message Tag
Reserved
Data Length
IANA Enterprise Number
OEM-defined
4
Supported Entities
Supported Interactions
Reserved
2
2
2
2
1
1
1
1
4
1
1
1
1
4
1
[1]
1
6
Value
26Fh
from Ping request
per UDP
per UDP
6 = RMCP Version 1.0
00h
[2]
FFh for IPMI
06h = ASF
4542 = ASF IANA
40h = Presence Pong
from Ping request
00h
16 (10h)
If no OEM-specific capabilities exist, this
field contains the ASF IANA (4542) and
the OEM-defined field is set to all zeroes
(00000000h). Otherwise, this field
contains the OEM’s IANA Enterprise
Number and the OEM-defined field
contains the OEM-specific capabilities.
Not used for IPMI.
This field can contain OEM-defined values;
the definition of these values is left to the
manufacturer identified by the preceding
IANA Enterprise number.
81h for IPMI
[7]
1b = IPMI Supported
[6:4] Reserved
[3:0] 0001b = ASF Version 1.0
Reserved for future definition by ASF
specification, set to 00000000b
Reserved for future definition by ASF
specification, set to 00 00 00 00 00 00h
113
Intelligent Platform Management Interface Specification
12.3 IPMI Messages Encapsulation Under RMCP
For LAN transfers, IPMI messages are a special class of data encapsulated in an IPMI Session packet. The IPMI
Session packets are encapsulated in RMCP packets, which are encapsulated in UDP datagrams. This is illustrated
in the following figure. The same type of encapsulation is used for IPMI serial/modem messages via PPP, except
the Ethernet Framing is replaced with a packet that uses PPP Framing and IP protocol type.
Figure 12-3, IPMI LAN Packet Layering
Ethernet Framing
MAC Address
IP/UDP
IP Address, RMCP Port #
RMCP message
Class=IPMI
RMCP Sequence# = FFh
IPMI Session
Header
Session ID
Session Sequence #
Authentication Type
AuthCode
IPMI Message
NetFn
LUN
Seq#
CMD
Data
12.3.1 RMCP/ASF and IPMI Byte Order
Please take note of the following:
Multi-byte fields in RMCP/ASF fields are specified as being transmitted in ‘Network Byte Order’ - meaning
most-significant byte first.
RMCP and ASF-specified fields are therefore transferred most-significant byte first.
The IPMI convention is to transfer multi-byte numeric fields least-significant Byte first. Therefore, unless
otherwise specified:
Data in the IPMI Session Header and IPMI Message fields are transmitted least-significant byte first.
114
Intelligent Platform Management Interface Specification
12.3.2 Example IPMI over LAN Packet
The following table shows the format and fields for IPMI messages encapsulated in an RMCP packet that is
itself encapsulated within an IPv4 UDP packet delivered over Ethernet:
Table 12-8, RMCP Packet for IPMI via Ethernet
MAC Header
IP Header
Field
size in
bytes
Byte
Offset
Value
Destination Address
Source Address
Frame Type
Version
Header Length
6
6
2
4-bits
4-bits
00h
06h
0Ch
0Eh
0Eh
0800h
4h for IPv4
5h
Precedence
Service Type (Type of Service)
3-bits
4-bits
0Fh
0Fh
000b
[4]
1000b
reserved
Total Length
[5]
Identification
Flags
1-bit
2
2
3-bits
0Fh
10h
12h
14h
13-bits
1
1
2
4
4
2
2
2
2
1
1
1
1
1
4
4
16
14h
16h
17h
18h
1Ah
1Eh
22h
24h
26h
28h
2Ah
2Bh
2Ch
2Dh
2Eh
2Fh
33h
37h
1
varies
47h
48h:xx
(length of IP header in units of 4-bytes)
[4]
(minimize delay)
0b
[5]
note
[6]
010b
(don’t fragment)
Fragment Offset
Time-to-Live
Protocol
Header Checksum
Source IP Address
Destination IP Address
Source Port
Destination Port
UDP Length
UDP Checksum
Version
Reserved
RMCP Sequence Number
Class of Message
Authentication Type
Session Sequence #
Session ID
Message Authentication Code
(AuthCode)
This field is not present when
Authentication Type set to ‘none’.
IPMI Message Length
Per Section 12.4, IPMI LAN
Message FormatIPMI LAN
Message Format
[1]
PAD
CRC
UDP Header
RMCP Header
IPMI Session
IPMI Message
MAC level
1.
[7]
0_0000_0000_0000b
[3]
40h
11h
26Fh
[2]
FFh for IPMI
07h for IPMI
[8]
note
[8]
note
1
4
Some LAN adapter chips may have a problem where packets of overall lengths 56, 84, 112, 128, or
156 are not handled correctly. The PAD byte is added as necessary to avoid these overall lengths.
Remote console software must use the PAD byte when formatting packets to any 10/100 Ethernet
device that accepts RMCP packets.
115
Intelligent Platform Management Interface Specification
2.
RMCP Messages with class=IPMI should be sent with an RMCP Sequence Number of FFh to
indicate that an RMCP ACK message should not be generated by the message receiver.
3.
Default value for packets transmitted from the BMC. Can be overridden via a configuration
parameter setting.
4.
Value used for packets transmitted from the BMC. The BMC ignores the value of this parameter
(except for checksum calculations) on received packets.
5.
BMC should increment this field each time it sends a new packet.
6.
Default value for packets transmitted from the BMC. Bit offset 1 (fragment bit) can be overridden via
a configuration parameter setting.
7.
Default value for packets transmitted from the BMC. The BMC is not required to support receiving
fragmented packets. Packets with a non-zero fragment offset and/or a flags field bit 2 = 1b (“more
fragments” may be silently discarded.)
8.
The Session ID and Session Sequence Number must be non-zero for commands executed during
an active session. All 0’s for the Session ID and/or Session Sequence Number (null Session ID, null
Session Sequence Number) are special values only used for commands that can be executed prior
to without establishing a session, e.g. Get System GUID, Get Channel Authentication Capabilities,
and Get Session Challenge. The Activate Session uses a null Session Sequence Number before a
session is activated, but does not use a null Session ID. Instead, it must use the Temporary Session
ID given by the BMC in the response to the Get Session Challenge command.
12.4 IPMI LAN Message Format
The encapsulated IPMI Messages are based on the same format as specified for the IPMB. This is done for
consistency and simplification of bridging operations. There is one significant difference. For IPMB messages,
the requester and responder addresses are always 7-bit I2C slave addresses. For IPMI LAN messages, the
addresses can be either slave addresses or software IDs. The least significant bit of the responder’s address and
requester’s address field indicates which type of address is being used, as described below.
There is no linkage between inbound and outbound messages and whether the message is a request or a response
message. Inbound messages can be either request or response messages and outbound messages can be request or
response messages.
The following table presents the formats for request and response messages:
Figure 12-4, IPMI LAN Message Formats
Request
rsAddr.
(SA or sw ID)
net Fn
(even) / rsLUN
rqAddr.
(SA or sw ID)
rqSeq / rqLUN
checksum
cmd
request data bytes
(0 or more)
checksum
Response
rqAddr.
(SA or sw ID)
rsAddr.
(SA or sw ID)
net Fn
(odd) / rqLUN
rqSeq / rsLUN
checksum
cmd
completion
code
response data
bytes (0 or more)
checksum
Where:
checksum
116
2's complement checksum of preceding bytes in the connection header or between the
previous checksum. 8-bit checksum algorithm: Initialize checksum to 0. For each byte,
checksum = (checksum + byte) modulo 256. Then checksum = - checksum. When the
checksum and the bytes are added together, modulo 256, the result should be 0.
Intelligent Platform Management Interface Specification
cmd
completion code
data
LUN
netFn
rq
rqLUN
rqAddr
rqSeq
rs
rsLUN
rsAddr
Seq
Command Byte
Completion code returned in the response to indicated success/failure status of the request.
As required by the particular request or response for the command
The lower 2-bits of the netFn byte identify the logical unit number, which provides further
sub-addressing within the target node.
Network Function code
Abbreviation for ‘Requester’.
Requester’s LUN.
Requester's Address. 1 byte. LS bit is 0 for Slave Addresses and 1 for Software IDs.
Upper 7-bits hold Slave Address or Software ID, respectively. This byte is always 20h
when the BMC is the requester.
Sequence number, generated by the requester.
Abbreviation for ‘Responder’.
Responder’s LUN
Responder's Slave Address. 1 byte. LS bit is 0 for Slave Addresses and 1 for Software
IDs. Upper 7-bits hold Slave Address or Software ID, respectively. This byte is always 20h
when the BMC is the responder.
Sequence number. This field is used to verify that a response is for a particular instance of
a request. Refer to [IPMB] for additional information on use and operation of the Seq field.
12.5 LAN Alerting
LAN Alerts are accomplished by generating a UDP Datagram that contains an SNMP Trap formatted per the
IPMI Platform Event Trap (PET) Format specification. This same format is used for PPP alerts generated over the
serial/modem interface when operating in PPP/UDP mode. Information for the PET trap comes from the Event
Message that generated the alert and from the LAN configuration parameters for PET.
12.6 IPMI LAN Configuration
12.6.1 IP and MAC Address Configuration
The BMC in the managed system needs the system’s IP Address and MAC Address in order to be able to
respond to UDP/IP packets or generate LAN alerts.
A BMC implementation is not required to be able to run DHCP or other protocols to keep it’s IP address
assignment up-to-date. In such implementations, it is the responsibility of system software to keep this address
information current in case it might change (as could be the case if the lease expired on an IP address, perhaps
because the system was unplugged for a long time).
It is recommended that system software periodically check the BMC’s address assignment to see if it is current,
and to update it if it’s not. It is also recommended that the BIOS run DHCP and initialize the BMC IP address
on startup if the BMC implementation does not include built-in DHCP support.
12.6.2 ‘Teamed’ and Fail-over LAN Channels
It is possible that an implementation may have multiple network controllers connected to the BMC. In such a
configuration, it may be desirable to support a configuration where multiple network controllers share the same
IP address. This ‘teamed’ configuration provides a bandwidth improvement by allowing messages to that IP
address to be sent and received by multiple NICs. Similar arrangements can be used to offer ‘fail-over’
capability where one NIC will be activated if another fails.
117
Intelligent Platform Management Interface Specification
Teaming and fail-over require special system software and driver support that is outside the scope of this
specification. However, it should be noted that IPMI Sessions could be implemented in a manner that can
facilitate such applications. One useful approach is to design the implementation such that Session IDs are
unique across all channels. That way, if two LAN channels were configured with the same IP address, the BMC
could accept session traffic that was split across the two channels. Since user information is channel-specific, it
would also be necessary for the user data and other configuration options to be identically configured. An
alternative implementation may provide a proprietary option where the two LAN connections are combined
into a single logical channel when teaming is in effect.
Note: The maximum operating privilege level and authentication will be determined by the user privilege
and channel privilege limit settings. Since these can vary on a per channel basis, it is possible that unless
the channels are configured identically a different maximum operating privilege level will be seen based
on which channel a message is received on.
12.7 ARP Handling and Gratuitous ARP
For Ethernet, the Address Resolution Protocol (ARP) [RFC 826] allows a host to find the physical address (MAC
address) of a target host on the same network, given only the target's Internet address. Systems and routers cache
IP Address-to-MAC Address information so they do not need to perform ARP requests every time they
communicate with another system. This cache is commonly referred to as an ARP Cache.
ARP Requests are broadcast. The request contains the IP Address for which an ARP Response containing the
MAC address is desired. The sender's IP Address-to-MAC Address mapping is also in every ARP request
broadcast. This allows receivers to update their own caches with the sender’s information before responding to
the ARP request.
A Gratuitous ARP is an ARP Response where the responder sends out the internet-to-physical mapping of its own
IP Address. Since the sender’s internet-to-physical address mapping is part the request, receivers use that
information to update their own caches with the sender’s address mapping.
It is common for systems to do a Gratuitous ARP on startup to inform other machines of its address (possibly a
new address). This gives the other systems a chance to update their ARP cache entries immediately. A Gratuitous
ARP at startup can also be used as a way to check whether another system is already using the system’s IP
address.
For this version of the specification, Gratuitous ARP capability is only described for Ethernet LAN channels.
12.7.1 OS-Absent problems with ARP
Some BMC LAN implementations may only have the ability to only receive UDP packets that are addressed to
the RMCP ports. Since Ethernet ARP packets are not UDP packets, Ethernet ARP request packets would not
get routed to the BMC. Thus, when the system is in a powered down state, the system may not accept ARP
Requests, or the request may not be able to be seen by the BMC, and the ARP Request will not get responded
to. This means that a remote application that relies on ARP to get the MAC address will not be able to connect
to the managed system once the system has powered down or is in a sleep state and the remote application’s or
intermediate router’s ARP cache entries expire.
12.7.2 Resolving ARP issues
The following are possible approaches to eliminating or reducing issues that can occur if the BMC LAN
implementation cannot receive or respond to ARP Requests while the system is powered down or sleeping. It is
also possible for this to happen if the run-time software does not use the network. This could happen while in a
failed state or if the system has booted to ‘DOS’ or a local diagnostic partition.
•
118
Increase ARP Cache expiration intervals in routers and applications.
Intelligent Platform Management Interface Specification
•
Implement Proxy ARP on the subnet - implement Proxy ARP software (software that responds to ARP
Requests on behalf of the managed systems) on one or more systems on the subnet. At least one of the
Proxy ARP systems must remain powered up.
•
Have the application maintain the ARP table - This only works if the remote console application is on the
same subnet as the managed system. Some network stacks include an arp utility program that allows ARP
entries to be manually entered into the ARP table (cache) for that system. An application could use this
mechanism to maintain the ARP table with a ‘fixed’ IP-to-MAC address association for the system.
•
Use a router with Proxy ARP capability - Some routers can be configured to provide a Proxy ARP
capability
•
Wake-On-LAN - If the managed system supports Wake-On-LAN it may be used to wake the system in
order to allow system software to respond to a later ARP Request.
•
Use a Network Controller with built-in ARP Response capability. As out-of-band management using
RMCP becomes more popular, network controller vendors may offer controllers with the ability to directly
respond to ARP Requests when the system is powered down or sleeping.
•
Provide Gratuitous ARPs from the BMC. If the BMC LAN connection allows the BMC to send ARP
Requests, then the BMC could periodically issue Gratuitous ARPs. Many routers and network stacks will
accept this Gratuitous ARP in place of an actual ARP response packet.
The best solution is to have an implementation where the BMC or Network Controller directly responds to ARP
Requests during times that the OS does not. If this is not possible, having the BMC issue Gratuitous ARPs can
often work well as a substitute. Because BMC-generated Gratuitous ARPs and ARP Responses may be
common, this specification includes commands that can be used for configuring and controlling those
capabilities if they exist in the implementation.
12.7.3 BMC-generated ARPs
A BMC LAN implementation may support BMC-generated Gratuitous ARPs or BMC-generated ARP
responses. If either of these options are supported, the BMC shall also support the BMC-generated ARP
Control LAN configuration parameter.
The term “BMC-generated” in this case means that the Gratuitous ARP or ARP Response generation is under
direct control of the BMC. The actual logic for sending the ARP packet may be in another device. For example,
a Network Controller chip may have the ability to be enabled by the BMC through a private interface.
It is possible that run-time software will want to take over the responsibilities for ARP handling during runtime. A BMC implementation that supports BMC-generated ARPs should also support the Suspend BMC ARPs
command. This command allows system management software to suspend BMC-generated ARPs while the
Watchdog Timer is running. Refer 19.3, Suspend BMC ARPs CommandSuspend BMC ARPs Command for
more information.
12.8 Retaining IP Addresses in a DHCP Environment
DHCP (Dynamic Host Configuration Protocol) is an UDP-based protocol that is primarily used to allow
systems to obtain an IP Address from a DHCP Server on the network. This address assignment is ‘leased’ and
will expire if the assignment is not refreshed by the time the lease expires. The BMC LAN implementation may
not be able to run DHCP. This could be because the BMC LAN implementation may only have the ability to
send and receive via the RMCP port addresses, preventing it from running standard DCHP.
If the BMC itself cannot run DHCP, the BMC must rely on the IP Address assignment that is configured into
the LAN Configuration Parameters. Typically, system software be able to keep the address assignment while
the system is running. This can either occur as a consequence of having sufficient IP traffic activity occur to
119
Intelligent Platform Management Interface Specification
keep the lease, or if the system may be idle for long periods of time, a software agent could be written that
periodically refreshes the assignment.
A more serious issue can occur while the system is powered down or sleeping. If the system is powered down
or is sleeping for a sufficiently long time, the IP Address could be lost due to expiration of the DHCP lease.
When the system starts up again, the BMC will need to get a new IP address assignment into its configuration
parameters.
12.8.1 Resolving DHCP issues
The following are possible approaches to eliminating or reducing issues that can occur if the BMC LAN
implementation cannot perform DHCP while the system is powered down or is sleeping:
•
If possible, configure Static IP addresses for your managed server systems. DHCP Servers can typically be
configured to deliver fixed IP addresses for a given MAC address.
•
If you have to use leased IP addresses, configure long lease intervals for the addresses.
•
Have a system management software agent that checks the IP Address assignment and updates the BMC if
the assignment changes.
•
Have the BIOS perform DHCP and update the BMC when the system powers up or resets. This helps
safeguard against changes to the IP address that may have occurred when the system was powered down or
sleeping. It also helps ensure that the BMC as an IP Address assignment if booting to an alternate OS or
service partition, and provides a mechanism for getting an IP address for BMC LAN even before the system
has an OS loaded.
•
Enable Wake-On-LAN capabilities. This capability can be used to allow a remote console to occasionally
wake the system to ensure that the IP Address assignment is retained or updated.
In general, a system in a DHCP environment will typically be used frequently enough to never lose its address
assignment. If run-time software and BIOS can keep the BMC up-to-date with IP address assignment changes,
the need to refresh assignments while the system is powered-down or sleeping may not be an issue in many
environments.
12.9 LAN Session Activation
The LAN Channel is an authenticated multi-session connection. Messages delivered to the BMC via LAN are
optionally authenticated using the session authentication mechanisms and challenge/response protocol described
in section 6.11.7, Session Activation and IPMI Challenge-ResponseSession Activation and IPMI ChallengeResponse. Also refer to sections 6.9, Users & Password SupportUsers & Password Support, and 6.11.3, Multisession ConnectionsMulti-session Connections.
In addition, a LAN implementation supports discovery via the RMCP Ping/Pong mechanism as a step that
typically precedes the session activation phase.
The following presents an overview of the steps that are used by a remote console to establish a IPMI Session via
IPMI LAN. These are also illustrated in Figure 12-5, LAN Session Start, below.
1.
The remote console discovers the system by issuing an RMCP Presence Ping message. The response called
the Presence Pong message, returns a bit indicating whether the platform supports IPMI, and whether the
platform uses just the Primary RMCP Port (26Fh) or both the Primary RMCP Port and the Secondary/Secure
RMCP Port (298h).
2.
If the system supports IPMI, the remote console starts the process of establishing a session by sending a Get
Channel Authentication Capabilities command packet with Authentication Type = none (“in clear”). The
120
Intelligent Platform Management Interface Specification
response packet will contain information regarding which type of challenge/response authentication is
available to be utilized.
3.
The console then requests a session challenge by issuing a Get Session Challenge request, also with
Authentication Type = none. The request contains information indicating what type of authentication type the
console wants to use. This must be one of the supported types returned by the Get Channel Authentication
Capabilities command. The response packet will contain a challenge string and a Session ID.
4.
The console then activates the session by issuing an Activate Session request. The Activate Session packet is
typically authenticated. For message-digest algorithms, the packet includes a signature (AuthCode) that is a
hash of the challenge, the Session ID, the password, and the message data, using one of the supported
algorithms from the Get Channel Authentication Capabilities command. The console also sets the initial
value for the Outbound sequence number that it wants the BMC to use for packets it sends to the console.
5.
The BMC returns a response confirming that the Session has been successfully activated. It also returns the
Session ID to be used for the remainder of the session, and the initial Inbound session sequence number that
it wants the remote console to use for subsequent messages it sends to the BMC for that session. The Activate
Session response is also authenticated (signed) in the same manner as the request was. This allows the remote
console to validate that it has a correct Session ID. Note that IPMI does not support switching authentication
algorithms ‘mid stream’. The algorithm used with the Activate Session command is the algorithm that will be
used for subsequent authenticated messages for the session. The exception to this is that the ‘none’
authentication type is allowed if options such as ‘Per-Message Authentication’ and/or ‘User Authentication’
are disabled.
121
Intelligent Platform Management Interface Specification
Figure 12-5, LAN Session Startup
Discovery
Remote Console
Managed System
RMCP Ping
RMCP "Pong"
(Ping response)
Get Session Challenge, Rq
SessionID=0, Sess. seq# = 0,
Auth. type =None, AuthCode =
not present [username, Auth Type
= MD2]
(Console requests a challenge using MD2
authentication type)
Activation
SessionID=0, Sess. seq# = 0,
Auth. type =None, AuthCode =
not present, [Challenge, temp
session ID]
Get Session Challenge, Rs
(BMC returns challenge string and
temporary session ID)
Activate Session, Rq
(Console delivers signed request using
temporary session ID, console also selects
the outbound sequence number it want the
BMC to use)
SessionID=temp, Sess. seq# = 0,
Auth. type =MD2, AuthCode =
MD2(), [challenge, Outbound
Seq=mm]
SessionID=temp, Sess. seq# = 0,
Auth. type =None, AuthCode =
MD2(), [SessionID, Inbound Sess.
seq#=nn]
Activate Session, Rs
(BMC returns a signed packet with the
session ID to be used for the active session.
BMC also returns the inbound sequence
number it wants the console to use for
messages to the BMC)
Active
Set Privilege Level, Rq
122
(from this point, all packets in the session
use the assigned Session ID and session
sequence numbers starting from the
inbound and outbound sequence numbers
that were exchanged using the Activate
Session command)
SessionID, Sess. seq# = nn,
Auth. type =MD2, AuthCode =
MD2(), [desired priv. level]
SessionID, Sess. seq# = mm,
Auth. type =MD2, AuthCode =
MD2(), [completion code]
Set Privilege Level, Rs
Intelligent Platform Management Interface Specification
13. IPMI Serial/Modem Interface
This section describes the mechanisms specific to transferring IPMI messages between the BMC and a remote
management system (remote console) over modem or direct serial connection. It also describes the mechanism that
support the Serial Port Sharing capability.
13.1 Serial/Modem Capabilities
The following is a review of the capabilities that can be provided via an IPMI serial/modem connection:
IPMI messaging Transmission of IPMI messages between a remote console and the BMC in one of three
configurable modes: Basic Mode, PPP Mode, and Terminal Mode.
Dial Paging
Ability to generate a numeric page by sending a dial string to a modem.
TAP Paging
Ability to automatically generate a configurable alphanumeric page by automatically
connecting to a TAP v1.8 -based paging service.
Dial-out PET
Alerting
Ability to automatically dial up a remote PPP-to-LAN gateway, connect, and place a Platform
Event Trap onto the remote LAN. Also known as PPP Alerting.
Callback
Ability for a remote console to trigger the BMC to call the remote console back and establish
a system management session. There are two types of Call-back: “IPMI” callback, which is
initiated via an IPMI command to the BMC, and callback using Microsoft’s CBCP (callback
control protocol). CBCP is an option that is only available in PPP Mode.
PPP UDP Proxy Option to allow the BMC to function as a low-performance communication bridge to allow
software to sending and receiving UDP data via a pre-established BMC PPP connection. If the
call-back option is supported, local management software or BIOS can trigger the BMC to
dial up the remote console.
Serial Port
Sharing
Ability to share a serial connector between the BMC’s serial controller and a system serial
controller by using circuitry to allow it to be switched between the two.
13.2 Connection Modes
The specification for the serial/modem interface supports IPMI Messaging in three possible connection modes.
Support for Basic Mode is mandatory if serial/modem support is provided. A given implementation can
implement any number or combination of the other connection mode options.
Basic Mode
This mode uses a simple clear text password to activate a session. IPMI messages are
encoded and delimited using a simple framing scheme based on ‘escaped’ characters.
Basic Mode is the most efficient standard operating mode for enabling a remote console
application to communicate with the BMC using IPMI messages.
PPP/UDP Mode
This mode uses the same session and authentication operation as IPMI over LAN. It uses
PPP as the protocol for establishing a point-to-point communications link over which
IPMI messages are sent encapsulated in UDP datagrams. This mode incurs significant
overhead in message size and handshake complexity beyond that required for Basic
Mode IPMI messaging, but has the advantage of using a widely supported standard.
Terminal Mode
This mode is intended primarily for direct serial connection operation. The mode is
designed so that a simple terminal or terminal emulator can be used to generate requests
and get responses from the BMC. The IPMI messages are entered using printable ASCII
123
Intelligent Platform Management Interface Specification
characters. While a user can enable a ‘line edit mode’ and directly enter the codes for an
IPMI message, the main purpose of this mode is to facilitate the development of scripts
that work with available terminal emulation programs.
Terminal Mode also supports a small number of ASCII Text Commands that can be used
for operations such as getting a high-level hardware health status for the system, and
doing system reset and power on/off operations.
13.2.1 PPP/UDP Proxy Operation
The BMC can support a mode that allows local system software (e.g. BIOS) to send and receive UDP
datagrams via the BMC connection to the remote console. This operation is supported using two special
message buffers associated with the channel: the PPP UDP Proxy Transmit Buffer and the PPP UDP Proxy
Receive Buffer.
When PPP/UDP Proxy Operation is supported (and enabled) the BMC will check the destination port address
used in incoming UDP datagrams. After removing any data escaping and checking the FCS, the BMC will
check the destination port address in the UDP packet. If the packet is not addressed to either the primary or
secondary RMCP Port addresses, the BMC will place the contents the packet into the PPP UDP Proxy Receive
Buffer (assuming the packet fits, and the buffer is already empty). Otherwise, the packet will be silently
discarded.
When sending messages to the remote console, local software loads the PPP UDP Proxy Transmit Buffer with
the contents for the UDP message and then directs the BMC to deliver that message as a UDP datagram from
the given serial/modem channel. The BMC fills in remaining data for the UDP and IP Header according to data
passed in the Send PPP UDP Proxy Packet command and from the LAN Configuration parameters and then
transmits the packet.
PPP/UDP Proxy Operation is only specified for execution via the BMC system interface. This capability is
OPTIONAL for serial/modem channels that support PPP mode.
13.2.2 Asynchronous Communication Parameters
The asynchronous communication parameters consist of elements such as bit rate, type of handshake, parity,
and other settings related to the configuration of the BMC’s serial controller. These setting are configured via
the serial/modem configuration parameters.
The number of different sets of parameters for a given channel depends on which messaging and alerting
features are implemented:
124
•
There is one set for use by IPMI Messaging (Basic Mode, PPP Mode, or Terminal Mode) for the entire
channel.
•
There is one set of asynchronous communication parameters for each Alert or Callback Destination
supported by a channel, used according to the Alert Type (Dial Page, TAP Page, PPP Alert, Callback).
Intelligent Platform Management Interface Specification
13.2.3 Serial Port Sharing
Serial Port Sharing is an option where the BMCs serial controller and a baseboard serial controller can share the
same serial connector. Figure 13-1, Serial Port Sharing Logical DiagramFigure 13-1, Serial Port Sharing
Logical Diagram, presents a logical example of how this could be accomplished using a multiplexer to switch
serial connector signals between the BMC and the baseboard serial controller.
Figure 13-1Figure 13-1 is referred to as a logical diagram because this specification does not require a
particular physical implementation as long as the commands function as described in this specification.
Figure 13-1, Serial Port Sharing Logical Diagram
TxD
RxD
DCD
RI
MODEM
Platform Status and Control
Serial Port Connector
(Typically 'COM2')
Power Control
Power Status
Serial Port Transceivers
Reset Control
Serial
Connector
Mux
Chassis Intrusion
Non-volatile Storage
TxD
System Event Log
(SEL)
DCD
UART
RxD
BMC
Sensor Data Record
(SDR) Repository
RI
FRU Inventory info
System Interface
Baseboard Serial
Controller
System Bus (e.g. X-bus, LPC)
System Bus
(e.g. PCI)
125
Intelligent Platform Management Interface Specification
13.2.4 Serial Port Switching
The following can cause a switch of the serial port:
Table 13-1, Serial Port Switching Triggers
[1]
From - To
Cause
BMC to Baseboard
IPMI Set Serial/Modem Mux command.
Microsoft VT100 ‘Exit Exit UPS, ASIC or
Service Processor’ Escape sequence. (see
ref: [MSVT]). Per [MSVT] the BMC should
immediately acknowledge the switch to
baseboard by returning <ESC>*
IPMI Set Serial/Modem Mux command.
Microsoft VT100 ‘Invoke ASIC/Service
Processor’ Escape sequence (see ref:
[MSVT])
yes
yes
Detection of basic mode Get Channel
Authentication Capabilities request
yes
Detection of leading bytes in a PPP IPv4UDP Packet addressed to BMC’s IP
address and RMCP primary or secondary
port, ending at the RMCP message class
field.
PPP Protocol = Internet Protocol (e.g.
0021h)
Packet =
IPv4 UDP Datagram
IP Address = BMC IP Address
Port =
Primary or Secondary
RMCP port, as set in the
serial/modem configuration
parameters
Initial packet data = RMCP v1.0 header
with message class field =
IPMI
RI Signal
yes
Baseboard to BMC
disable
no
yes
Notes
<ESC>Q
<ESC>(
Pattern is also used for
Connection Mode Auto-detect
capability. See 13.2.9,
Connection Mode AutodetectConnection Mode Autodetect.
Requires Basic Mode to be
enabled. Pattern is also used for
Connection Mode Auto-detect
capability. See 13.2.9,
Connection Mode AutodetectConnection Mode Autodetect.
Requires PPP mode to be
enabled. Pattern is also used for
Connection Mode Auto-detect
capability. See 13.2.9,
Connection Mode AutodetectConnection Mode Autodetect.
yes
In Modem Connect mode only.
See 13.2.11, Modem
ActivationModem Activation for
more information.
DCD Signal
yes
Used to cause a mux switch to
BMC when in Direct Connect
mode.
1. This indicates whether a configuration option to disable switching on this action exists. Note that switching may also be
disabled as part of the operation of some access modes, such as ‘Pre-boot Only’.
13.2.5 Access Modes
BMC channels used for serial/modem access can be configured for several Access Modes using the Set
Channel AccessSet Channel Access command. The command determines which states of system operation (e.g.
pre-boot) the channel can be used for BMC communication. Refer to section 6.6, Channel Access
ModesChannel Access Modes for more information.
126
Intelligent Platform Management Interface Specification
13.2.6 Console Redirection with Serial Port Sharing
A common use for Serial Port Sharing is as a mechanism to allow the serial connection to be shared between
the BMC and with BIOS Console Redirection. Serial Port sharing includes commands to help facilitate this
application. This section presents an overview of those commands and how they might be used. This is just a
starting point. The actual specification of the BMC with BIOS console redirection is outside the scope of this
specification.
13.2.6.1
Detecting Who Answered The Phone
Since console redirection is normally used with a remote console, we’ll start from when the remote console
first connects to the system. Depending on the configuration, either the system or the BMC may be the party
that ‘answers the call’. Whether the BMC connects or not is determined by the access mode settings set via
the Set Channel Access command, plus activation settings in the serial/modem configuration parameters.
There is also a configuration parameter that allows the BMC to wait for a ‘ring interval’ before answering a
call in order to give system software an opportunity to connect first.
Thus, when a remote console connects to a given system, it should first try to determine whether it connected
with the BMC or with console redirection. This avoids the possibility that the remote console will send IPMI
message data down only to have it be mis-interpreted as console-redirection keystrokes.
For Basic Mode and Terminal Mode, the BMC can be configured to periodically issue a Serial/Modem
Connection Active message whenever it has the port and right after the port is switched to the BMC. The
remote console can wait for this message as a confirmation that the BMC has the port before attempting to
send any messages to the BMC. If that message is not received, it can assume that the system answered the
phone instead of the BMC.
If PPP is used, PPP communication software on the remote console will typically initiate the PPP negotiation
without waiting for the managed system. Because of this, there is less need to send a Serial/Modem
Connection Active message first, since by the time the message is generated the remote console has already
sent in characters. Similarly, because there are separate port addresses that would be used for RMCP traffic
to the BMC, there’s no strong need for the Serial/Modem Connection Active message to be periodically sent.
Thus, the BMC does not send Serial/Modem Connection Active messages in PPP Mode except when the
serial connection is being switched to or from the BMC.
When the serial connection is switched over to the BMC, the Serial/Modem Connection Active message will
be delivered to the Primary RMCP port address and IP address of the remote peer that was negotiated during
IPCP. If the BMC did not negotiate IPCP, the Serial/Modem Connection Active message will not be sent.
When the serial connection is being switched over to the system, the Serial/Modem Connection Active
message will be delivered to the Primary RMCP port address and IP address of the remote peer that was
negotiated with IPCP, and to each active session on that PPP channel. If the BMC did not negotiate IPCP,
then the Serial/Modem Connection Active message will only be sent to the active sessions.
If remote console software wishes to detect the presence of a BMC, it can do so by sending a Get Channel
Authentication Capabilities message after IPCP has been negotiated. [Note that if console redirection uses
‘ASCII’ then the remote console may have to assume that console redirection is occurring if it cannot
establish a PPP Link. (Generally, ASCII text console redirection and PPP communication with the BMC
don’t share well together)]
If the system includes PPP Link authentication, the remote console could distinguish between whether the
system or the BMC established the link based on the peer name that is used in the link negotiation.
127
Intelligent Platform Management Interface Specification
13.2.6.2
Connecting to the BMC
The remote console can cause the connection to be switched back to the BMC by the mechanisms listed in
Table 13-1, Serial Port Switching TriggersTable 13-1, Serial Port Switching Triggers. Whether switching is
allowed is based on what the Access Mode setting is for the channel. For example, if the channel is set to
‘pre-boot only’ - then the remote console will not be able to remotely switch the mux over to the BMC if the
system is presently in run-time operation.
13.2.6.3
Connecting to the Console Redirection
The Set Serial/Modem Mux command provides the mechanism for a remote console to direct the BMC to
switch the serial connection over to the system serial controller. The <ESC>Q sequence can also be enabled
for this purpose.
13.2.6.4
Directing the Connection After Power Up / Reset
The remote console can send commands to the BMC to initiate a system power up or reset operation. After
that operation, the remote console may want to see console redirection, or it may want to stay connected to
the BMC. The Set Serial/Modem Mux command can be used to direct whether the remote console stays
connected to the BMC or not. The Set Serial/Modem Mux command includes an option to allow a mux
switch to be requested or to be forced. A mux switch that is requested may be denied (blocked). A BIOS
using console redirection would typically request that the mux be switch over to the system during POST, so
that the remote console could block that request if necessary. Thus, if the remote console wants to keep the
connection, it simply issues a Set Serial/Modem Mux command to block requests to switch the mux to the
system before sending the power up or reset command.
13.2.6.5
Interaction with Microsoft ‘Headless’ Operation
Microsoft has specified an interface [MSVT] for text-based console redirection to support pre-boot
operations with the operating system. This specification includes escape sequences for activating and
deactivating the connection to a ‘service processor’, as well as an escape sequence for hard resetting the
system. See [MSVT] for more information.
IPMI v1.5 includes optional serial/modem configuration parameters for supporting [MSVT] in a system that
implements serial port sharing along with [MSVT]. These parameters provide a common way for the
[MSVT] activate/deactivate and reset sequences to be enabled or disabled in the system. Supporting these
options in IPMI does not imply that a given implementation is conformant with the [MSVT] specification.
Refer to [MSVT] for the full system requirements.
Note that the present [MSVT] specification calls out for a timeout on the escape sequence filtering. If an
<ESC> is received, subsequent characters in the sequence must be received within 20 seconds.
13.2.6.6
Pre-boot Only Mode
The definition of Pre-boot Only access mode is that the BMC serial connection becomes disabled when the
system starts to boot in order to guarantee that system software has full use of the serial connection without
concern that incoming calls would be able to connect to the BMC. In order to provide emergency
management coverage, someone using pre-boot only mode would typically also configure the watchdog
timer and PEF so that a system power down or reset would occur on critical system failures, thus allowing a
remote console to connect to the BMC.
The remote console has the ability to use the Set Serial/Modem Mux command to block mux switch requests,
but allow mux switch ‘forces’. This is typically used with the Pre-boot Only access mode. At the start of
POST BIOS requests the mux. If a remote console is connected, it can block that request in order to continue
128
Intelligent Platform Management Interface Specification
to communicate with the BMC during POST, if desired, or the remote console can let BIOS take the mux in
order to see BIOS console redirection.
At the conclusion of POST and start of boot, BIOS will typically force the mux away from the BMC and to
the system. Once the mux has been forced away from the BMC in Pre-boot Only mode, the BMC is not
allowed to take the port back until the next time the system is powered down or is reset.
Note that Alerting is not affected by Pre-boot Only mode. Alerting, if enabled, will ‘take over the port’ and
cause an alert to be sent even if the system was using the port at the time. BIOS can use the Get Channel
Access command to determine that the BMC is configured to operate in Pre-boot Only mode for the serial
connection.
13.2.6.7
Always Available Mode
In Always Available Mode the serial connection is considered to be dedicated to the BMC. In order to avoid
confusion with run-time software, BIOS will typically hide or disable the serial port when the OS load
process starts. BIOS can know when to do this by reading the access mode setting from the BMC using the
Get Channel Access command.
13.2.6.8
Shared Mode
In Shared Mode the BMC is allowed to ‘answer the phone’ but run-time software is also able to use the serial
connection when it’s not being used by the BMC. BIOS can use the Get Channel Access command to see
when the BMC is configured for Shared mode. In this case, it can leave the serial port enabled for run-time
software access. The serial/modem configuration parameters include a ‘ring interval’ parameter that can be
used to enable the BMC to only answer the phone if system software doesn’t. This is accomplished by
simply setting a ring interval for the BMC that is longer than the time it takes system software to answer.
13.2.7 Serial Port Sharing Access Characteristics
The following table lists the mux control and modem-answering characteristics according to the type of Access
Mode and state of Serial Port Sharing. Corresponding mux switching steps that would typically be used in
BIOS for supporting console redirection also listed.
In general, if a remote console application wishes to keep communication with the BMC after a power-up or
reset, it should issue a Set Serial Modem/Mux command to block mux requests before issuing a Chassis Control
command to cause the system to power-up or reset. However, it should not block mux ‘force’ operations
because this could interfere with run-time access of the serial connection.
Table 13-2, Serial Port Sharing Access Characteristics
Serial Port
Sharing
disabled
Access
Mode
disabled
Characteristics
Same behavior for both Modem and Direct Connect Mode
• If system power is On, Mux always set to system. When power is off Mux setting is
unspecified.
• Set Serial/Modem Mux command is rejected. (See response data for the Set
Serial/Modem Mux command).
• Escape sequence / pattern triggered switching is not available.
• Alerting Unavailable.
• BMC Power-on Default = mux set to system when system power is On. When power is
off, mux setting is unspecified.
BIOS Action at POST start: none required
BIOS Action at POST end: none required
129
Intelligent Platform Management Interface Specification
disabled
any
except
‘disabled’
enabled
disabled
enabled
pre-boot
130
Same behavior for both Modem and Direct Connect Mode
• Mux always set to BMC.
• Set Serial/Modem Mux command is rejected. (See response data for the Set
Serial/Modem Mux command).
• Escape sequence / pattern triggered switching is not available.
• Alerting available.
• BMC Power-on Default = mux set to BMC.
BIOS Action at POST start: none required.
BIOS Action at POST end: Recommend hiding/disabling baseboard serial controller.
Same behavior for both Modem and Direct Connect Mode
• Mux always set to system (except during alerting).
• Escape sequence / pattern triggered switching is disabled.
• Set Serial/Modem Mux command available.
• Alerting available.
• BMC Power-on Default = mux set to system.
BIOS Action at POST start: none required.
BIOS Action at POST end: none required.
BMC pays attention to Modem Ring Time parameter until mux is forced to system using
Set Serial/Modem Mux command. Afterwards, BMC will not automatically take over mux for
IPMI messaging (will not answer the phone) until next power down or system reset (unless
commanded).
• Escape sequence / pattern triggered switching is available.
• Set Serial/Modem Mux command available.
• Alerting available. BMC will terminate call and automatically take the mux in order to
send an alert, unless an IPMI Messaging Session is already in progress on the channel in which case alert will be “deferred” until channel becomes available for dial-out.
•
BMC Power-on Default = Mux set to system iIf system power is off, or if Modem
Ring Time >00h and <3Fh the power-on default mux setting is unspecified until RI
or DCD is detected (see below), otherwise set to BMC.
For Modem Mode, the BMC automatically takes over the connection upon power down,
after system resets, and on detecting Ring based on Modem Ring Time parameter,
except if a session is active - in which case the BMC will keep the connection (until the
mux is forced to system using the Set Serial/Modem Mux command).
If Modem Ring Time parameter is >00h and <3Fh, If system power is on and the
Ring Time countdown is running, the mux will be set to system to allow the system to
answer the modem call. Mux will be set to system. BMC will take over mux if Ring
Time expires while Ring is being detected via the RI signal. If system power is on,
tThe mux will be returned to system when loss of connection (loss of DCD) is
detected, or if the BMC takes the mux but is unable to establish a connection.
If Ring Time = 00h, BMC will take mux during power down and after system resets
as necessary to be able to answer the call via the modem. BMC will also take the
mux and answer phone connect with the modem immediately when a Ring is
detected via the RI signal. Mux will be claimed by the BMC whenever loss of DCD
connection is detected. To the BMC, this is essentially the same ‘phone answer’ and
power down/reset behavior as in ‘Always Available’ mode.
For Direct Connect Mode the BMC automatically takes the connection upon power down,
after system resets, and whenever loss of DCD is detected (if DCD-based switching is
enabled) except if a session is active - in which case the BMC will keep the connection
(until the mux is forced to system using the Set Serial/Modem Mux command).
BIOS Action at POST start: Request mux to system if BIOS console redirection enabled.
BIOS Action at POST end: Force to system. Keep baseboard serial controller enabled.
Intelligent Platform Management Interface Specification
enabled
always
available
enabled
shared
• Escape sequence / pattern triggered switching is available.
• Set Serial/Modem Mux command available.
• Alerting available. BMC will terminate call and automatically take the mux in order to
send an alert, unless an IPMI Messaging Session is already in progress on the channel
in which case alert will be “deferred” until channel becomes available for dial-out.
• BMC Power-on Default = Mux set to BMC.
For Modem Mode, the BMC automatically takes over mux on power down, system resets,
when loss of DCD is detected, and upon detecting initial activity of RI. The BMC also
initializes the modem whenever DCD loss is detected. The BMC ignores the Modem
Ring Time parameter.
For Direct Connect Mode, the BMC automatically takes the mux upon power down, after
system resets, and whenever DCD is absent (if DCD-based switching is enabled).
BIOS Action at POST start: Request mux to system if BIOS console redirection enabled.
BIOS Action at POST end: Force mux to BMC. Recommend BIOS hides/disables
baseboard serial controller.
•
Escape sequence / pattern triggered switching is available.
•
Set Serial/Modem Mux command available.
•
Alerting available. BMC will terminate call and take mux in order to send an alert,
unless an IPMI Messaging Session is already in progress on the channel - in
which case alert will be “deferred” until channel becomes available for dial-out.
•
BMC Power-on Default = Mux set to system iIf system power is off, or if Modem
Ring Time >00h and <3Fh the power-on default mux setting is unspecified until RI
or DCD is detected (see below), otherwise set to BMC.
For Modem Mode, the BMC controls mux upon power down, after system resets, and on
detecting Ring based on Modem Ring Time parameter, except if a session is active - in
which case the BMC will keep the connection:
If Modem Ring Time parameter is >00h, <3Fh. If system power is on and the Ring
Time countdown is running, the mux will be set to system to allow the system to
answer the modem call. Mux will be set to system. BMC will take over mux if Ring
Time expires while Ring is being detected via the RI signal. If system power is on,
tThe mux will be returned to system when loss of connection (loss of DCD) is
detected, or if the BMC takes the mux but is unable to establish a connection.
If Ring Time = 00h, BMC will take mux during power down and after system resets
as necessary to be able to answer the call via the modem. BMC will also take the
mux and answer phone connect with the modem immediately when a Ring is
detected via the RI signal. Mux will be claimed by the BMC whenever loss of DCD
connection is detected. To the BMC, this is essentially the same ‘phone answer’ and
power down/reset behavior as in ‘Always Available’ mode.
For Direct Connect Mode, the BMC takes the mux upon power down and after system
resets, except if a session is active - in which case the BMC will keep the connection.
Once power is up, the BMC will leave the mux in the state last commanded by software
or an escape sequence and will not automatically take the mux unless DCD loss is
detected (if DCD-based switching is enabled), or an alert needs to be sent.
BIOS Action at POST start: Request mux to system if BIOS console redirection enabled.
BIOS Action at POST end: None. BIOS leaves mux setting alone. Note that the Boot
Options contain flags that remote software can use to request BIOS to place the mux
into a given setting at POST end.
13.2.8 Serial Port Sharing Hardware Implementation Notes
There are a number of characteristics that should be considered when designing hardware that aids in the
implementation of serial/modem remote access and Serial Port Sharing
•
The BMC needs the ability to monitor the DCD and RI signals from the serial connector in order to
detect incoming modem calls.
131
Intelligent Platform Management Interface Specification
•
The physical implementation is required to ensure that the baseboard serial controller does not receive
characters when the serial connector is switched to the BMC. The lines in to the baseboard serial
controller should also be placed in an appropriate ‘idle’ level.
•
In order to prevent signal transitions from causing interrupts to the baseboard communication routines, it
is recommended that the remaining serial input signals to the baseboard serial controller also be
switched and placed into an appropriate ‘idle’ level when the BMC is using the connector.
•
The physical implementation is required to handle additional serial signal lines, such as RTS and DTR,
in order to ensure that those signals remain in the active state to keep the modem connection active when
switching between the baseboard and the BMC. If the BMC does not have control over those signal
levels, it may be necessary to accomplish this using additional baseboard circuitry.
•
The implementation is required to ensure that the serial connection does not see glitches or signal drops
on the RTS, CTS, DSR, DTR, DCD, and RI lines due to switching between the baseboard and BMC
serial controllers.
•
It is up to the implementation to determine how it handles any ‘Wake On Ring’ options for the serial
connector.
•
The BMC may have other RS-232 lines under its control (DTR, RTS, CTS, and DSR). Hardware
handshake via RTS and CTS is an implementation option. A BMC implementation may also optionally
use DTR as an additional hang-up mechanism.
•
The serial/modem feature is more valuable if the BMC can be communicated with when the system is in
a powered-down or sleep state. This may require the port transceivers to be powered via Standby Power.
13.2.9 Connection Mode Auto-detect
Connection Mode Auto-detect refers the capability for the BMC to automatically detect and enter a particular
Connection Mode (Basic mode, PPP mode, or Terminal mode) based on detecting an appropriate data pattern in
the serial traffic from the remote console. Implementing Connection Mode Auto-detect is optional.
The configuration of this capability is handled via the Connection Mode parameter in the serial/modem
configuration parameters. The parameters allow this capability to be disabled, and the BMC set to use just one
connection mode for direct IPMI messaging to the BMC.
The following is the description and specification of the operation of Connection Mode Auto-detect.
132
•
The BMC will auto-detect whenever an IPMI messaging session is not active. When a session is not active,
the BMC constantly snoops for the different data patterns that will tell the BMC what connection mode to
use, but also will cause the serial connection to be switched over to the BMC. The pattern matching routine
must check for all supported mode patterns in parallel. For example, suppose the BMC supports auto-detect
for all three connection modes. Even if it detects what looks like the start of the PPP Mode pattern, the
pattern detection routine must continue to look for Basic Mode and Terminal Mode patterns in parallel
until the PPP mode pattern is confirmed.
•
For modem mode, the BMC detects that a connection has been established by detecting DCD or by
receiving a ‘CONNECT’ string from the modem, depending on the implementation. For direct connect
mode, the BMC uses DCD if DCD is enabled, otherwise it assumes that a connection exists any time the
mux is switched to the BMC.
•
If Basic Mode is enabled, and the mux is set to the BMC, the BMC will assume that Basic Mode is the
desired connection type. The BMC will send out a Serial/Modem Connection Active “Ping” messages (if
enabled) after the connection is established. A remote console that wishes to use Basic Mode should wait
for the Ping before sending any packets to the BMC. This will avoid the possibility of the packet from
Intelligent Platform Management Interface Specification
being interpreted as keystroke input if the remote console happens to connect to text-based console
redirection instead of the BMC.
•
If the mux is already switched to the BMC, the BMC can detect a PPP Link Negotiation request and use
that to set the connection mode to PPP Mode.
•
The process is more complicated if system software performed the negotiation and the BMC is ‘snooping’
to see if it should be activated and enter PPP mode. The BMC needs to snoop for both compressed and
uncompressed versions of the address and protocol fields. Note that PPP already specifies that a receiver
must accept uncompressed headers even if compressed headers were negotiated, so this support should
already be part of the BMC’s PPP routines. The BMC also needs to snoop according to the escaping that
the system software negotiated. This is more problematic. The BMC needs to know what the system
negotiated for its transmit ACCM (Asynchronous Control Character Mask) in order to know what control
characters to ignore in the data stream. The following are options for handling this situation:
a.
Pre-configure the BMC to match the escape negotiation that software will use. The serial/modem
configuration parameters contain a ‘Snoop ACCM’ parameter that the BMC can be directed to use.
The Snoop ACCM indicates which control characters the BMC should ignore when snooping in PPP
mode.
b.
Have the BMC snoop the Link Negotiation process. The BMC monitors the transmit ACCM that the
system is using.
The following table lists the patterns that the BMC will look for when auto-detecting the connection mode. The
patterns will also trigger a switch of the serial connector to the BMC if Serial Port Sharing is enabled.
Table 13-3, Auto-Connection Mode Patterns
Connection Mode
Basic Mode
PPP Mode
Pattern
BMC looks for a complete Get Channel Authentication Capabilities
command in basic mode format. The BMC should also respond to
this command. Once a basic session is established the BMC will
stay in basic mode until the session is terminated.
If the mux is already connected to the BMC, the BMC will enter
PPP mode if it detects the start of a PPP LCP packet after the
connection is established. The BMC shall check the PPP packet
bytes up to and including the LCP Packet Code field. An
implementation can elect to ignore the Identifier and Data field
values, but the Length and FCS (Frame check sequence) must be
correct.
If the mux is connected to the system, the BMC will switch and
attempt to use PPP mode if it detects a PPP packet with the
following characteristics:
Protocol =
IPCP
Packet =
IPv4 UDP Datagram
IP Address =
BMC IP Address
Port =
Primary or Secondary RMCP port, as set in
the serial/modem configuration parameters
Initial packet data = RMCP v1.0 header with message class
field = IPMI
An implementation can elect to either switch immediately on
detecting this pattern without additional data integrity checks, or
wait until it has verified the checksums and FCS on the packet.
Terminal Mode
The BMC is not required to respond to the IPMI message that was
encapsulated in the packet.
Terminal Mode can be enabled to be entered on receiving an
“<ESC>(“ sequence (if enabled). The BMC will respond with
[TMODE OK] and will operate in terminal mode until the connection
133
Intelligent Platform Management Interface Specification
is terminated or the data pattern for Basic Mode or PPP Mode
IPMI-RMCP packet is detected.
13.2.10 Modem-specific Options
The serial/modem configuration parameters (set using the Set Serial/Modem ConfigurationSet Serial/Modem
Configuration command) support various modem configuration strings. These strings are set into non-volatile
storage managed by the BMC. The BMC uses the strings to configure the modem for out-of-band access use.
Due to the limited length of these strings, it may not be possible to configure all necessary modem parameters.
Rather than relying solely on these strings, it is recommended that the user pre-configure the modem for out-ofband operation and save those settings in the modem as the default. The Modem Strings can then be used just to
hold strings that trigger the modem to restore its defaults.
The BMC automatically sends an <Enter> character (carriage return = 0Dh) after sending the Modem
Initialization and Hang Up Line strings. <Enter> is not sent after the Escape Sequence string.
Table 13-4, Modem String Summary
String Name
Default String
Minimum.
String
Length
Modem Init String
Implementation
dependent. The string
ATE1Q0V1X4&D0S0=0
is a good starting point.
64 bytes,
including
termination
character
Modem Hang Up
Sequence
(unused if DTR
hang-up is available
and selected)
ATH
8 characters,
plus
termination
character
Modem Escape
Sequence
(Unused if DTR
hang-up is available)
+++
4 characters,
plus
termination
character
String Usage
Transmitted every time the
serial/modem connection becomes
activated. The BMC automatically
sends an <Enter> character after this
string.
Sent to modem whenever the BMC
wants to terminate the session (i.e.
password retry count is exceeded, etc).
The BMC automatically sends an
<Enter> character after this string.
NOTE: If the DTR hang-up option is
selected, this field will not be used.
Informs modem that next stream of
bytes should be interpreted as
command bytes and not sent to the
remote software. The escape sequence
must be sent prior to sending any other
command if the modem is currently
connected with another modem. Note:
This may cause the modem to hang up
unless it has been configured
otherwise.
Escape sequence is preceded by a 2
second pause in transmission from the
BMC, and is follow by a 2 second pause
in transmission.
This sequence precedes the Modem
Initialization string, except after a RI or
connection after DCD loss.
134
Intelligent Platform Management Interface Specification
13.2.11 Modem Activation
The BMC will monitor RI and claim the serial connection (switch the mux to the BMC) according to the Modem
Ring Time parameter in the serial/modem configuration parameters1 and Section 13.2.7, Serial Port Sharing Access
CharacteristicsSerial Port Sharing Access Characteristics. After getting the connection, if DCD is already asserted,
the BMC will monitor the incoming data stream for the start of an IPMI session. If DCD is not already asserted, the
implementation can use one of the following mechanisms:
1.
The BMC will initialize the modem with the initialization string from the serial/modem configuration
parameters. The initialization string must be configured to set the modem to answer the phone. The BMC
then waits for DCD to become active.
2.
The BMC initializes the modem by sending the initialization string. The BMC then listens for a “RING”
result code from the modem and sends out an “ATA” to answer the phone.
Of these two methods, method #2 is the preferred implementation since it does require leaving the modem in an
auto-answer state.
13.3 Serial/Modem Connection Active (Ping) Message
If terminal-based console redirection is used, it is important for a remote console application to know whether the
system or the BMC is connected to the serial connector before it sends any messages. Otherwise, if the serial port
was already connected to the system, an incoming IPMI message could be interpreted by the system as redirected
key strokes.
Therefore, there is a configuration option for Basic Mode and Terminal Mode that can direct the BMC to send out
a Serial/Modem Connection Active request message once every two seconds whenever the serial connection is
switched to the BMC, with the first message starting immediately after the connection has been switched.
This message is also referred to as the “Serial/Modem Ping”. The message is used both to get the attention of the
remote software and to allow the remote software to determine whether it is connected to the BMC or not. The
Serial/Modem Connection Active message is primarily required when system console redirection is using a
terminal-based format for input from the remote console, where incoming IPMI message characters could be
misinterpreted as redirected input. For example, if console redirection was operating using a ‘VT100’ terminal
emulation, the characters in an incoming IPMI message might be interpreted as a command or terminal control
escape sequence.
If PPP is used, PPP communication software on the remote console will typically initiate the PPP negotiation
without waiting for the managed system. Because of this, there is less need to send a Serial/Modem Connection
Active message first, since by the time the message is generated the remote console has already sent in characters
in an attempt to do link negotiation for PPP. Similarly, because there are separate port addresses that would be
used for RMCP traffic to the BMC in PPP mode, there’s no strong need for the Serial/Modem Connection Active
message to be periodically sent. Thus, the BMC does not send Serial/Modem Connection Active messages in PPP
Mode except when the serial connection is being switched to or from the BMC.
When the serial connection is switched over to the BMC, the Serial/Modem Connection Active message will be
delivered to the Primary RMCP port address and IP address of the remote peer that was negotiated during IPCP.
If the BMC did not establish the PPP Link, the Serial/Modem Connection Active message will not be sent.
When the serial connection is being switched over to the system, the Serial/Modem Connection Active message
will be delivered to the Primary RMCP port address and IP address of the remote peer that was negotiated with
IPCP, and to each active session on that PPP channel. If the BMC did not establish the PPP Link, then the
Serial/Modem Connection Active message will only be sent to the active sessions.
1
A ring duration is used instead of a ring count in order to simplify handling the variations in RI that occur between different national
telephone systems and modems.
135
Intelligent Platform Management Interface Specification
Note: If the BMC configured for any mode other than Direct Connect Mode, the Serial/Modem
Connection Active message will not be sent out unless DCD is asserted. Sending Serial/Modem
Connection Active messages while the modem is on-hook has been shown to prevent some
modems from answering. This also implies that the modem should not be configured to hold
DCD asserted.
13.3.1 Serial/Modem Connection Active Message Parameters
The Serial/Modem Connection Active message includes a parameter that indicates that a mux switch from BMC
to the system serial controller is about to occur. This is provided to give a remote application some notification
that the system is switching the port back over to the baseboard serial port. The Serial/Modem Connection Active
Message with the ‘switching to system’ parameter will be sent out before the mux is switched and before the
response is returned for the Set Serial/Modem Mux command.
13.3.2 Mux Switch Coordination
It is possible that the remote application committed an IPMI message for delivery to the BMC at the time that the
switchover to the system occurred. If the mux switch occurred immediately, this means that the message might be
delivered to the system instead of the BMC. To protect against this occurrence, the BMC can be configured to
look for the remote console to acknowledge the Serial/Modem Connection Active message before the mux switch
occurs.
There is a configuration option that directs the BMC to retry sending the Serial/Modem Connection Active
message up to three times with 20 ms between retries if it does not get an acknowledge from the remote
console. The BMC will then switch the mux if it has not received an acknowledge-message from the remote
console within three seconds of sending the last retry.
The remote console acknowledges the switch by sending a Serial/Modem Connection Active request message of
its own back to the BMC. The reason for this approach is so the BMC will return a response to the message,
allowing the remote console to receive positive confirmation of the acknowledge message or to retry the message
if the response is not received.
Note that the remote console should not send messages if it has not received a Serial/Modem Connection Active
message or other message from the BMC in the last two seconds. The three second delay provides margin to help
ensure that the console will not transmit with a Serial/Modem Connection Active message right when BMC times
out and the mux switch occurs.
13.3.3 Receive During Ping
The serial/modem interface operates in a ‘full duplex’ mode. Thus, the BMC must continue to receive message
characters while it is transmitting a Serial/Modem Connection Active 'Ping' message, or any other IPMI
message.
For Basic Mode operation, the BMC must continue to handshake each message that it receives. This means that
the BMC may insert an 'ACK' character in the middle of a Ping message transmission.
13.3.4 Application Handling of the Serial/Modem Connection Active Message
A robust Remote Console Application should be prepared to handle serial/modem remote access connection
becoming deactivated or activated at any time.
A cessation of Serial/Modem Connection Active messages indicates that the serial/modem remote access
connection is no longer active, while the occurrence of Serial/Modem Connection Active messages indicates
that the connection is active. Thus, if a Remote Console Application should always monitor the
136
Intelligent Platform Management Interface Specification
presence/absence of Serial/Modem Connection Active messages, whether the serial/modem connection is active
or not.
If the application is connected to the BMC, and does not receive an Serial/Modem Connection Active message
within 2 seconds of its last transaction with the BMC should assume that the serial/modem connection has
become deactivated.
Conversely, if the application is communicating with the system (e.g. console redirection) and an Serial/Modem
Connection Active message is received, the application should recognize that the serial/modem connection has
become reactivated.
13.4 Basic Mode
Basic Mode eliminates much of the overhead associated with PPP/UDP mode. Instead of encapsulating IPMI
messages within an RMCP message in a UDP datagram in a PPP frame, the IPMI messages are simply encoded
and framed for serial transmission. The price of this efficiency is that the remote console application cannot take
advantage of built-in support for PPP and UDP in the operating system, but will need to implement IPMI
communications routines on top of the OS’s generic support for asynchronous serial communications.
Since Session IDs are not part of the basic IPMI message, only a single IPMI session is supported in Basic Mode.
The BMC can use whatever Session ID value it wants for the Get Session Challenge and Activate Session
commands.
13.4.1 Basic Mode Packet Framing
Framing is done with special characters to delimit the start and end of a Basic Mode packet, and to indicate the
sequence for an escaped data byte (see following section). Framing and data byte escaping are applied after the
message fields have been formatted. These special characters are specified in the following table.
Table 13-5, Basic Mode Special Characters
Description
Start Character
Stop Character
Packet Handshake Character
Data Escape Character
Value
A0h
A5h
A6h
AAh
Basic Mode messages can be thought of as IPMB messages with the I2C start and stop condition framing
replaced with start and stop characters, and with the addition of data byte escaping to ensure that the framing
characters are not encountered within the body of the packet. The Packet Handshake character is a special
value that is used for implementing a level of software flow control with the remote application accessing the
BMC. See Section 13.4.5, Packet HandshakePacket Handshake.
13.4.2 Data Byte Escaping
The Start, Stop, and Escape characters are disallowed within the body of the message. This is done to ensure
that the start and end of a message is unambiguously delimited. If a byte matching one of the special characters
is encountered in the data to be transmitted, it is encoded into a corresponding two-character sequence for
transmission. This encoding is summarized in the following table.
137
Intelligent Platform Management Interface Specification
Table 13-6, BASIC MODE Data Byte Escape Encoding
Data Byte
Encoded Sequence
A0h
A5h
AAh
A6h
1Bh <ESC>
AAh (ESC), B0h
AAh (ESC), B5h
AAh (ESC), BAh
AAh (ESC), B6h
AAh (ESC), 3Bh
The first character of the sequence is always the Escape character. Only the special Basic Mode characters plus
the ASCII Escape <ESC> character, 1Bh, are escaped. (The ASCII Escape <ESC> character, 1Bh, is escaped to
enable the BMC to snoop for certain escape sequences in the data stream, such as the <ESC>( and <ESC>Q
patterns.) All other byte values in the message are transmitted without escaping.
When the packet is received, the process is reversed. If the two-byte ‘escaped’ sequence is detected in the
packet, it is converted to the corresponding data byte value. The BMC shall reject any messages that have
illegal character combinations or exceed message buffer length limits. The BMC may not send an error
response for these conditions.
13.4.3 Message Fields
The message fields follow those used for the IPMB, as specified in the Intelligent Platform Management Bus
Communications Protocol v1.0 Specification, with the exception of the Requester’s (rq) Slave Address and
Responder’s (rs) Slave Address fields, which have slightly different definitions. Note, framing and data byte
escaping are applied after the message fields have been formatted. The general message format is illustrated in
the following figure:
Figure 13-2, Basic Mode Message Fields
Request
rsAddr
(SA or SWID)
rqAddr
(SA or SWID)
net Fn
(even) / rsLUN
rqSeq / rqLUN
checksum
cmd
data bytes
(0 or more)
checksum
Response
rqAddr
(SA or SWID)
rsAddr
(SA or SWID)
net Fn
(odd) / rqLUN
rqSeq / rsLUN
checksum
cmd
completion
code
response data
bytes (0 or more)
checksum
Where:
checksum
cmd
completion code
data
LUN
138
2's complement checksum of preceding bytes in the connection header or between the
previous checksum. 8-bit checksum algorithm: Initialize checksum to 0. For each byte,
checksum = (checksum + byte) modulo 256. Then checksum = - checksum. When the
checksum and the bytes are added together, modulo 256, the result should be 0.
Command Byte
Completion code returned in the response to indicated success/failure status of the request.
As required by the particular request or response for the command
The lower 2-bits of the netFn byte identify the logical unit number, which provides further
sub-addressing within the target node.
Intelligent Platform Management Interface Specification
netFn
rq
rqLUN
rqAddr
rqSeq
rs
rsLUN
rsAddr
Seq
Network Function code
Abbreviation for ‘Requester’.
Requester’s LUN.
Requester's Address. 1 byte. LS bit is 0 for Slave Addresses and 1 for Software IDs.
Upper 7-bits hold Slave Address or Software ID, respectively. This byte is 20h when the
BMC is the requester.
Sequence number, generated by the requester.
Abbreviation for ‘Responder’.
Responder’s LUN
Responder's Slave Address. 1 byte. LS bit is 0 for Slave Addresses and 1 for Software
IDs. Upper 7-bits hold Slave Address or Software ID, respectively. This byte is 20h when
the BMC is the responder.
Sequence number. This field is used to verify that a response is for a particular instance of
a request. Refer to [IPMB] for additional information on use of the Seq field.
13.4.4 Message Retries
Basic Mode Messaging utilizes the same retry mechanisms used for the IPMB, as specification in the Intelligent
Platform Management Bus Communications Protocol v1.0 Specification. The remote application timeout
should be based on the IPMB timeout specifications, with additional time added for delay due to the phone
system. A remote application can determine this additional delay for a given connection based on the time it
takes to receive the Handshake character.
13.4.5 Packet Handshake
The handshake character is used to signal that the BMC has freed space in its input buffers for a new, incoming
IPMI Message. The BMC typically returns a Handshake character within one millisecond of being able to
accept a new message, unless the controller has already initiated a message transmission, or an operation such
as firmware update has been initiated.
An implementation can either send the handshake character in the middle of the transmission or elect to wait to
transmit the handshake character until the transmission in-progress has completed. If the implementation waits
for the transmission to complete, the handshake character will typically be sent within one millisecond after the
message transmission completed.
If the implementation elects to send the Handshake character in the middle of an outgoing message
transmission, it must not insert the Handshake character immediately following a Data Escape character. The
reason for this is to allow the remote console application some flexibility in whether it processes the Handshake
character before or after removing data escaping.
139
Intelligent Platform Management Interface Specification
13.5 PPP/UDP Mode
This mode of operation uses PPP [RFC1661] (point-to-point protocol) messaging for transmitting IP packets on
an asynchronous link per [RFC1662] The following sections provide overview material on PPP operation in
and explicit requirements for an IPMI implementation. The overview material is to provide context and a
starting point for understanding the implementation.
The material does not supercede the PPP specifications. Designers are required to refer to the PPP RFC
reference documents for information on implementing PPP, especially regarding the states involved in opening
and terminating a PPP link.
All values are delivered most-significant byte first unless otherwise specified.
PPP/UDP mode transfers IPMI Messages encapsulated in RMCP Packets. This enables RMCP ASF Messages
as well as IPMI Messages to be delivered to the BMC. The RMCP Packets are carried within UDP datagrams
using the same format as the IPMI LAN messages. The resultant UDP datagrams are transferred within PPP
frames. Specifications on this message formatting are provided in the follow sections.
13.5.1 PPP/UDP Mode Sessions
The BMC is only required to support one session on the PPP/UDP interface. A BMC implementation may elect
to support multiple sessions.
13.5.2 PPP Frame Format
PPP/UDP mode framing follows [RFC1662]. [RFC1662] specifies an ‘HDLC-like’ format for PPP frames
using on an asynchronous serial communication media. The following figure presents an overview of this
format.
Figure 13-3, PPP Frame Format
Flag
(7Eh)
Address
(FFh)
Control
(03h)
Protocol
1 or 2
bytes
Information
Padding
FCS
2 or 4 bytes
Flag
(7Eh)
Inter-frame Fill or next Address
13.5.3 PPP Frame Implementation Requirements
Since the flag (7Eh) indicates both the start and end of a packet, it’s possible that another flag could
immediately follow a flag. However, the protocol also allows the ‘end’ flag to serve as both the end of one
packet and the beginning of the next. The BMC must be able to handle both occurrences.
In order to reduce differences between implementations, it is recommended that the BMC must explicitly
transmit a flag on both ends of the packet2.
Support for the 16-bit (2-byte) FCS (frame check sequence) is mandatory.
2
Per [RFC1662], only one Flag is required between frames, but if a two flag sequence is received, it is viewed as if an empty frame
were received between two frames where the empty frame is silently discarded. Since the flag character delimits both the start and
end of the frame, this requirement eliminates the need for the BMC to track that it had already sent a flag on the end of the
previous frame and thus can skip sending a flag to start the current frame.
140
Intelligent Platform Management Interface Specification
13.5.4 Link Control Protocol (LCP) packets
The following table presents a summary of the LCP Fields used in PPP. This is provided for reference only.
Table 13-7, LCP Code Fields
Code
1
2
3
4
5
6
7
8
9
10
11
LCP Packet Type
Configure-Request
Configure-Ack
Configure-Nak
Configure-Reject
Terminate-Request
Terminate-Ack
Code-Reject
Protocol-Reject
Echo-Request
Echo-Reply
Discard-Request
13.5.5 Configuration Requests
The first step in opening a PPP link is to establish the connection through the exchange of Configure packets.
The PPP Configure-Request message is used to request changes to the link defaults. A Configure-Request is
responded to with a Configure-Ack, Configure-Nak, or Configure-Reject packet. A link is considered
established once the negotiation of link options has completed. If a Configure packet is received later, the link
will be returned to the link establishment phase.
Table 13-8, Overview of PPP Configure-Ack, -Nak, & -Reject Packet Use
LCP Packet
Configure-Ack
Configure-Nak
Configure-Reject
…
All options are recognized and accepted. The Options field is a copy of the Options field from the
corresponding Configure-Request. Receipt of a Configure-Ack from both ends of the link signals that the link is
opened and other (non-LCP) protocols may be accepted.
All options are recognized, but one or more are not acceptable. The Options field returns a list of the
unacceptable options in the same order that the options were given in the Configure-Request. An
implementation may append other options to prompt the peer to include those options in its next ConfigureRequest packet.
Some of the configuration options are not recognizable, or are not acceptable for negotiation.
…
The format of the Configure-Request, -Ack, -Nak, and -Reject packets follow the same format, as illustrated in
Figure 13-4, Configure-Request, -Ack, -Nak, -Reject Packet Format.
Figure 13-4, Configure-Request, -Ack, -Nak, -Reject Packet Format
Options
Code
Type1
Type2
Identifier
(Seq)
Len1
Len2
Length
Data1
Data2
TypeN
•••
LenN
DataN
The Code field identifies whether the Link Control Packet is a Configure-Request, -Ack, -Nak, or -Reject
packet, per Table 13-7, LCP Code FieldsTable 13-7, LCP Code Fields.
The Identifier field is similar to the IPMI ‘Seq’ field. The value must be changed for new requests, and the
value in the request must be returned in the corresponding Configure-Ack, Configure-Nak, or Configure-Reject
response.
141
Intelligent Platform Management Interface Specification
The Options field holds a list of 0 or more link configuration parameters to be changed, and the corresponding
values for those parameters. The Options field should only be filled with requests for non-default values.
Options are not required to be in any given order in a Configure-Request. But the Configure-Ack, -Nak, and Reject packets do need to return the options in the same order they were given in the corresponding Configure
Request.
All Configuration Options that a sender wishes to negotiate are negotiated simultaneously.
Table 13-9, PPP Link Configuration Option Support Requirements
ID
1
2
3
Type
Maximum Receive Unit
Asynch Control
Character Map
Authentication Protocol
Len
4
≥4
Quality Protocol
≥4
5
Magic Number
6
8
1.
142
BMC Support Requirement
Request: Recommended.
It is recommended that the BMC request that
packets smaller than 1500 bytes be used. The
requested value must be ≥ XX bytes. IPMI only
requires XX bytes, but OEM value-added
features may require a larger maximum packet
size.
Response: Optional.
The BMC can respond to a request that larger
or smaller packets be used. If implemented, the
BMC must reject/nak any request to use a value
smaller than XX bytes. The implementation can
elect to accept >1500 bytes, at the discretion of
the implementer.
Note that per PPP an implementation must
accept at least 1500 bytes. See Section 13.5.6,
Maximum Receive Unit HandlingMaximum
Receive Unit Handling for more information.
Request and Response: Optional
4
7
Data
bytes 1:2 - 2-byte value indicating
request
bytes 1:2 - protocol ID
C023 = PAP
C223 = CHAP [RFC 1994]
(Algorithm:
#5 = CHAP w/MD5 [RFC 1994]
#128 = MS-CHAP v1
#129 = MS-CHAP v2
bytes 3:N - data according to
protocol ID
bytes 1:2 - protocol ID
C025 = Link Quality Report
bytes 3:N - data according to
protocol ID
bytes 1:4 - magic number
Request and Response: based on firmware
support for PAP, CHAP, and MS-CHAP
Request and Response: Optional
Request:Optional.
Response: Recommended.
It is recommended that the BMC indicate
support for Magic Number.
Protocol Field
2
none
Request and Response: Recommended.
Compression (PFC)
It is highly recommended that the BMC support
being configured to accept compressed protocol
fields, and request that it can use compressed
Protocol Fields when it transmits.
Address & Control Field
2
none
Request and Response: Recommended.
Compression
It is highly recommended that the BMC support
being configured to receive compressed
Address and Control Fields, and for the BMC to
request that it can use compressed Address
and Control Fields when it transmits.
PPP requires that implementations must be able to receive a full 1500-byte information field in case link synchronization is lost.
If an MRU value is not specified, it is assumed to be 1500 bytes.
Intelligent Platform Management Interface Specification
13.5.6 Maximum Receive Unit Handling
For full PPP compatibility, the BMC must accept up to 1500 bytes in the information field. The storage
requirements could be even greater if the implementation elects to buffer the frame first and process escaped
characters after getting the entire frame, instead of handling escaped characters as they’re received.
Ideally, the BMC would have internal storage to hold this full amount of data, but this may not be economical
in some implementations. If the BMC cannot directly buffer a 1500-byte MRU frame, the BMC must still
continue to accept bytes until the end of the frame, and must not lose track of framing, terminate the link, or
issue any error response just because of the overall frame length.
It is possible, though unlikely, that a UDP packet could contain an RMCP message that meets the BMC buffer
size requirements, but is padded with additional bytes that cause the PPP Frame to exceed the BMC’s buffer.
An implementation can handle this possibility by continuing to calculate the FCS on characters received after
the buffer has become full, before discarding those characters. Once the frame was completed, the BMC could
check the leading contents of the buffer to see if a complete, valid message was contained in the initial bytes.
Note that there is a UDP Checksum that would also need to be tracked. However, the BMC could opt to ignore
the UDP Checksum field. Barring formatting errors, the data-integrity-checking role of the UDP Checksum
should be covered by the PPP FCS. The UDP Length field should also be able to be ignored for IPMI-RMCP
messages, since the number of bytes preceding the IPMI Message Length field is constant, and the IPMI
Message Length will indicate the number of valid bytes remaining in the message.
It is recommended that, for PPP frames containing RMCP/UDP packets, the implementation accept PPP frames
greater than its buffer size, track and verify the Frame Check Sequence, and attempt to validate and interpret the
leading, buffered data.
13.5.7 Protocol Field Compression Handling
The least significant bit of a protocol field indicates that the last byte of the protocol has been sent. Therefore, if
the ls-bit of the first protocol byte position is a ‘1’, the implementation can simply assume that the protocol
field has been compressed to one byte.
Accepting a configuration request for Protocol Field Compression indicates that the implementation supports
receiving compressed protocol field values. This does not obligate the transmitter to send them. Thus, the
receiver must be able to receive frames that use both compressed and non-compressed formats.
It is recommended that the BMC request that it can use Protocol Field Compression for the frames it sends. If
this configuration option is accepted, the BMC itself should use it. Note there may be some cases where the
BMC may need to transmit without using compressed fields, even though it has negotiated for compressed
fields to be accepted. This is allowable in PPP and is also allowable in a BMC implementation.
13.5.8 Address & Control Field Compression Handling
Per the PPP specification, when Address & Control Field compression is used the Address and Control fields
are simply omitted. On reception, the Address and Control fields are decompressed by examining the first two
bytes. If they contain the values 0xff and 0x03, they are assumed to be the Address and Control fields. If not, it
is assumed that the fields were compressed and were not transmitted.
This works because the first byte of a two byte Protocol field will never be 0xff (since it is not even), and the
Protocol field value 0x00ff is not allowed (reserved) to avoid ambiguity when Protocol-Field-Compression is
enabled and the first Information field byte is 0x03.
LCP Packets are not allowed to be sent with compressed Address and Control fields.
Accepting a configuration request for Address and Control Field Compression indicates that the
implementation accepts frames using Address and Control Field Compression. This does not obligate the
143
Intelligent Platform Management Interface Specification
transmitter to send them. Thus, the receiver must be able to receive frames the use both compressed and noncompressed formats.
It is recommended that the BMC request that it can use Address and Control Field Compression for the frames
it sends. If this configuration option is accepted, the BMC itself should use it. Note there may be some cases
where the BMC may need to transmit without using compressed fields, even though it has negotiated for
compressed fields to be accepted. This is allowable in PPP and is also allowable in a BMC implementation.
13.5.9 IPMI/RMCP Message Format in PPP Frame
IPMI Messages are carried in RMCP Packets in UDP using the same format as the IPMI LAN messages. This
enables RMCP ASF Messages as well as IPMI Messages to be delivered to the BMC. RMCP support adds only
four bytes overhead to an IPMI Session message in UDP.
Figure 13-5, IPMI Message in PPP Frame Format
Field
PPP Frame
IP Header
UDP Header
RMCP Header
IPMI Session
IPMI Message
Size
Flag
Address
Control
Protocol
Version and Header Length
Service Type
Total Length
Identification
Flags & Fragment Offset
Time to Live
Protocol
Header Checksum
Source IP Address
Destination IP Address
Source Port
Destination Port
UDP Length
UDP Checksum
Version
Reserved
RMCP Sequence Number
Class of Message
Authentication Type
Session Sequence #
Session ID
Message Authentication Code
(AuthCode) Not present when
Authentication Type = none.
IPMI Message Length
Per Section 12.4, IPMI LAN
1
[1]
1 or 0
[1]
1 or 0
[2]
1 or 2
1
1
2
2
2
1
1
2
4
4
2
2
2
2
1
1
1
1
1
4
4
16
Value
7Eh
[1]
FFh
[1]
03h
0021 = IPv4
11h
26Fh
[3]
FFh for IPMI
07h for IPMI
[5]
note
[5]
note
1
varies
Message FormatIPMI LAN
Message Format
PPP Frame
144
FCS
2
[4]
Flag
1
1. Dependent on whether Address & Control Field Compression is used
2. Dependent on whether Protocol Field Compression is used
3. RMCP Messages with class=IPMI should be sent with an RMCP Sequence Number of FFh to indicate that an
RMCP ACK message should not be generated by the message receiver.
4. Per [RFC 1662] “Each frame begins and ends with a Flag sequence… Only one Flag Sequence is required
between two frames. Two consecutive Flag sequences constitute an empty frame, which is silently discarded
an not counted as an FCS error.” The implementation should take care to track that a single flag character may
indicate both the end of the present packet, and the start of the next.
Intelligent Platform Management Interface Specification
5.
The Session ID and Session Sequence Number must be non-zero for commands executed during an active
session. All 0’s for the Session ID and/or Session Sequence Number (null Session ID, null Session Sequence
Number) are special values only used for commands that can be executed prior to establishing a session, e.g.
Get System GUID, Get Channel Authentication Capabilities, and Get Session Challenge. The Activate Session
command uses a null Session Sequence Number before a session is activated, but does not use a null
Session ID. Instead, it must use the Temporary Session ID given by the BMC in the response to the Get
Session Challenge command.
13.5.10 Example of IPMI Frame with Field Compression
A PPP frame for IPMI in UDP/RMCP that uses both Protocol and Address-and-Control Field compression will
have the following format. Note that per PPP, uncompressed frames must also be accepted at any time. Buffer
sizes must take this into account.
Figure 13-6, IP Frame with Field Compression
Flag
(7Eh)
Protocol
(21h = IPv4)
UDP/RMCP/IPMI Packet data
FCS
2 bytes
Flag
(7Eh)
13.5.11 Frame Data Encoding
In order for the Flag and Control-Escape bytes to be utilized, they must not appear directly in the data stream.
The encoding of utilizes an escaping mechanism where bytes in packet are replaced with a two-character
sequence in order to prevent them from being mis-interpreted as being flag or Control-Escape bytes. The
escaping mechanism can also be used to prevent bytes from being interpreted as ASCII control characters.
Only bytes between flag bytes are escaped. The flag byte themselves are never escaped.
13.5.12 Escaping Algorithm
To ‘escape’ a character, N, the BMC simply emits a 7Dh character, followed by N exclusive-OR’d with 20h.
To convert the escaped-pair back to the original data byte, the 7Dh is thrown away and the second character
exclusive-OR’d with 20h.
13.5.13 Escaped Character Handling
By default, the following characters are escaped:
Table 13-10, Default Escaped Characters
Character
Control Escape
Flag
ASCII Control Characters
value
7Dh
7Eh
00h-20h
Escaped as:
7Dh, 7Dh
7Dh, 5Eh
7Dh, (value XOR 20h)
The BMC must ignore non-escaped versions of the above characters as part of the frame data, unless there has
been a negotiation that allows some of the characters to be sent without escaping. (See Section 13.5.14, Asynch
Control Character Maps (ACCM)Asynch Control Character Maps (ACCM), below) The control-escape
character and flag characters still need to be interpreted, of course.
The reason that non-escaped characters are dropped is that an intervening communication device may have
inserted the characters.
Only non-escaped characters are eligible to be dropped on receipt as spurious characters in the frame data. The
BMC must accept all escaped characters received within flag delimiters as part of the frame data.
145
Intelligent Platform Management Interface Specification
13.5.14 Asynch Control Character Maps (ACCM)
PPP includes an option that allows the negotiation of which characters require escaping and which can be
optionally escaped. This is accomplished by negotiating ACCMs (Asynch Control Character Maps) between
both ends of the link. ACCM negotiation can potentially improve data throughput by reducing the number of
characters that require escaping.
The BMC is not required to support ACCM negotiation. If ACCM negotiation is not supported, the BMC must
handle character escaping and escaped characters as described in Section 13.5.13, Escaped Character
HandlingEscaped Character Handling.
ACCMs are negotiated using a 32-bit parameter where each bit corresponds to a character from 00h to 1Fh,
with the least-significant bit corresponding to 00h. When sent in a configure-request, the ACCM indicates
which values that the originator of the request must receive in escaped format, and which can optionally be sent
without escaping. This ACCM is referred to as the ‘receiving ACCM’.
The ‘sending ACCM’ is the set of characters that will be sent with escaping, barring any additional
configuration due to a configure-request. By default, the sending ACCM is FFFFFFFFh - indicating the set of
default escaped characters listed in Table 13-10, Default Escaped CharactersTable 13-10, Default Escaped
Characters. The sending ACCM can only change as the result of receiving a configure-request indicating that
fewer characters need to be escaped. Note that a transmitter can send escaped codes for values >1Fh (in
addition to the required escaping of the flag and control-escape values). These cannot be configured via
negotiation, however.
The receiver of the configuration-request can respond with a configure-ack for the option, or it can responds
with a configure-nak and return the union of the requested ACCM with the ACCM that it will be using for
transmission. For example, suppose the remote console was hard-coded to always escape character 0Dh for
some reason. If the BMC submitted an ACCM indicating that it required only characters 03h and 04h to be
escaped, the remote console could respond with a configure-nak that it would always escape 03h, 04h, and
0Dh. This would tell the BMC that it should also ignore 0Dh characters in the data.
The receiving ACCM is assumed to be FFFFFFFFh by default. That is, a transmitter must escape values 00h1Fh (plus flag and control-escape values encountered in the frame data) unless it receives a configure-request
indicating that certain values do not need to be escaped. This also means that the receiver can expect to receive
00h-1Fh in escaped format until it has successfully configured an alternative.
13.5.15 IP Network Protocol Negotiation (IPCP)
Once the PPP link has been established, it is necessary to send NCP (network control packets) to choose and
configure one or more network layer protocols.
[RFC1332] describes IP Control Protocol (IPCP). IPCP is a network control protocol used to choose transfer
of IPv4 packets via PPP. The BMC must both accept IPCP configuration requests and generate IPCP
configuration requests.
146
•
BMC shall Configure-Ack an IPCP Configure-Request that contains 0 (zero) configuration options.
•
BMC shall issue a Configure-Request for IPCP option 3 (IP Address) to request that the IP Address
specified by the PPP IP Address parameter (see Table 20-4, Serial/Modem Configuration ParametersTable
20-4, Serial/Modem Configuration Parameters) be used as the BMC’s IP Address.
−
If IP Address Assignment is enabled in the serial/modem configuration parameters, the BMC shall
accept an address assignment that is returned via a corresponding Configure-Nak from the remote
console.
−
If IP Address Assignment is not enabled (Fixed IP Address), the BMC shall not accept a different
address assignment returned via a corresponding Configure-Nak from the remote console. If the BMC
PPP IP Address is not accepted, the BMC shall issue a new Configure-Request with the same PPP IP
Intelligent Platform Management Interface Specification
Address value (but a new identifier value). This will be repeated until the PPP IP Address parameter is
accepted, or at least three Configure-Requests for setting the PPP IP Address parameter have been
issued.
•
The BMC shall silently discard later IP Protocol (0021h) UDP packets that are addressed to the Primary or
Secondary RMCP ports, but do not match the negotiated PPP IP Address.
•
BMC shall accept IPCP option 3 (IP Address) in an IPCP Configure-Request, unless that message matches
the BMC’s PPP IP Address.
•
BMC shall Configure-Nak IPCP option 3 if IP Address Assignment is disabled, and return the PPP IP
Address value from the serial/modem configuration parameters in the Configure-Nak.
•
BMC shall Configure-Reject IPCP option 1, IP Addresses. This option has been deprecated in IPCP.
•
BMC shall Terminate-ACK a Terminate-Request for IPCP.
•
The BMC shall not transmit IPv4 protocol (0021h) packets after an IPCP Terminate-Request has been
received, until IPCP is renegotiated.
•
The BMC shall silently discard any IPv4 protocol packets received after an IPCP Terminate-Request has
been received, until IPCP is renegotiated (the BMC receives its first IPCP Configure-Request).
•
BMC may accept other IPCP options to support OEM features. For example, Option 2 is Van Jacobsen
compression for TCP/IP. Remote stacks must be prepared for the potential that a BMC implementation
might request network protocols and/or configuration options beyond those specified in this document.
•
BMC shall not enable OEM framing extensions alongside PPP mode. It is possible that an OEM may want
to include proprietary serial framing formats or special handshake or escape sequences that are not
specified in this document, but that work as a proprietary extension to PPP mode. The BMC will not be
considered to be conformant for PPP mode if these extensions are active. It is acceptable for the BMC to
have an OEM-specific option to enable/disable OEM extensions. In this case, conformance will only be
assessed when such OEM extensions are disabled. Remote stacks and remote console applications
designed for IPMI may break when OEM extensions are enabled.
•
The BMC shall Request the remote console IP Address by issuing a configure-request for Option 3 with an
IP Address value of 00.00.00.00 [This provides a mechanism for the BMC to obtain the remote console
connection’s IP Address in order to enable the BMC to asynchronously send UDP datagrams to the remote
console.]
•
The BMC shall accept IPv4 Protocol Packets (0021h) once it has received and responded to an IPCP
Configure-Request from the remote console.
13.5.16 CHAP Operation in PPP Mode
An implementation can support CHAP as a mechanism for authenticating the serial/modem connection at the
link level. This option is separate from the whether or not RMCP/IPMI Message packets are authenticated once
the link has been established. Serial/modem configuration options to select and support either standard CHAP
[RFC 1994], MS-CHAP v1 [RFC 2433], or MS-CHAP v2 [RFC 2759] are provided.
There are two classes of configuration options for CHAP:
•
IPMI Messaging: There is a single set of options that configures what type of CHAP, if any, is used for
serial/modem IPMI messaging with the BMC in PPP Mode. These are referred to as PPP Link options in
the serial/modem configuration parameters. See Table 20-4, Serial/Modem Configuration ParametersTable
20-4, Serial/Modem Configuration Parameters.
147
Intelligent Platform Management Interface Specification
•
Callback and Dial-out Alerting. Another class of PPP options relates to user names and accounts for
connecting with a remote system via a PPP-to-LAN connection. The specification supports multiple sets of
these options. The option sets are grouped under a PPP Account Selector number in the configuration
parameters. The PPP Account Selector provides the link that associates a set of PPP account parameters
with a particular serial/modem Alert or Callback destination.
The Set/Get User Name, Set/Get User Access, and Set/Get Channel Access commands are used to configure the
password and username associated with CHAP link-level authentication for IPMI messaging in PPP mode.
Note that the same User Names and passwords that are used for link authentication can either be the same as
those used for IPMI Messaging, or they can be different. There is a bit setting associated with the Set User
Access command that determines whether the information associated with a given User ID is to be used for PPP
Link Authentication (e.g. CHAP), or IPMI Messaging Authentication, or both.
PPP Account 1 is used to hold information for both IPMI Messaging via PPP and for callback, such as the IP
Address that the BMC will attempt to negotiate for itself.
13.6 Serial/Modem Callback
Callback provides a serial/modem channel mechanism that enables a remote console to direct the BMC to call a
pre-configured destination and attempt to establish an IPMI Messaging connection. Callback provides both a
security enhancement and a way to ‘reverse’ phone charges associated with managing a system.
Callback is primarily for use under the Modem connection mode. It can, however, be used with Direct Connect
mode for testing and development purposes. For example, PPP destination parameters could be tested locally
without requiring going through a modem. It’s potentially possible to locally verify parameters or do testing by
looping back from one system serial port to another using a ‘null modem’ cable.
Once the callback connection has been established, the BMC waits for the remote application to activate a session
with the BMC by issuing a Get Session Challenge command, etc.
If the Serial/Modem Connection Active (Ping) message is enabled, the BMC will announce its presence by
periodically sending the Ping once the connection has been established. The call will be automatically terminated
if the remote system does not activate a session with the BMC within the Session Inactivity Timeout interval for
the channel (See Section 6.11.13, Session Inactivity TimeoutsSession Inactivity Timeouts).
IPMI Messaging and Callback use the same mode setting (basic mode, PPP mode, or terminal mode). I.e. you
can’t request callback using one messaging mode, and have the BMC connect using a different messaging mode.
The Callback function is implemented at the IPMI Message level. PPP Callback (I.e. PPP LCP option 0D) is not
used. For callback, the PPP Account Set settings parameters are only used if IPMI Messaging for the channel is
set to PPP Mode.
In order to initiate a callback, the remote console first connects to the BMC using a pre-configured User ID and
then issues the Callback command. The User ID can be restricted to ‘Callback level’ privilege so that the only
operation that can be performed is to initiate a callback using the Callback command. A User ID can also be
restricted to only be accessible while a callback connection is active. Together, this provides the option to allow
one User ID and password to initiate the callback, while making it necessary to have a callback connection active
in order to perform any higher-privilege level connections to the BMC.
The Set User Access command is used to configure, on a per channel basis, whether a given user is enabled, what
the user’s limits are, and whether user access is restricted to only being available during a callback connection.
The Callback, Operator, and Administrator privilege levels can be used to initiate a Callback, but the User Level
cannot. This is consistent with the definition of User privilege.
148
Intelligent Platform Management Interface Specification
13.6.1 Callback Control Protocol (CBCP) Support
An implementation that supports PPP can elect to support the Microsoft Callback Control Protocol (CBCP).
CBCP is a Microsoft Corporation specification for supporting callback from a Microsoft RAS (Remote Access
Services) PPP connection. Other devices such as serial-to-LAN gateways may also support CBCP. See [CBCP]
for specification information.
With respect to the BMC, CBCP provides a protocol by which a remote console can request the BMC to initiate
a callback to the remote console.
CBCP is negotiated during the initial LCP phase. Once CBCP has been negotiated, per [CBCP] the BMC
initiates the callback process by issuing a request that tells the remote console what callback number options are
available from the implementation.
A BMC implementation can support one or any combination of the callback number options listed in the
following table. An implementation may elect to implement CBCP callback numbers such that different users
can have different callback numbers, or where callback numbers are shared across users.
Table 13-11, CBCP Callback Number Options
Option
No Callback
Callback to caller-specified
number
Callback to a pre-specified
number
Callback to one from a list
of numbers
Description
The remote console requests not to be called back at all.
The BMC indicates that it allows the remote console to specify which number
is to be called back.
The BMC calls back a pre-configured phone number.
If a PPP Link authentication protocol such as CHAP is used, the BMC uses
the user id string from the authentication negotiation to look up which phone
number to use for the given user. Otherwise, a global number associated with
the serial/modem configuration parameters for the channel will be used.
The BMC offers up a list of possible phone numbers that the callback can be
directed to. The remote console picks one and returns it to the BMC. If the
number matches one from the list, the BMC calls that number.
If a PPP Link authentication protocol such as CHAP is used, the BMC uses
the user id string from the authentication negotiation to look up which set of
phone numbers to offer to the given user. Otherwise, a global set of numbers
associated with the serial/modem configuration parameters for the channel will
be used.
13.6.1.1
CBCP Address Type and Dial String Characters
CBCP includes an Address Type field that indicates the format used for callback addresses. Address Type = 1
indicates PSTN/ISDN. No other Address Type values are specified, therefore this field is, by default, a fixed
field for IPMI implementations.
Per [CBCP] callback strings are null terminated ASCII strings formed from the following set of characters:
0-9, *, #, T, P, W, @, comma, space, dash, and parentheses.
This specification applies to using NT RAS as the dialer. For IPMI 1.5, however, the BMC is the dialer. Thus,
additional characters specified in section 13.11.1, Alert Strings for Dial PagingAlert Strings for Dial Paging
can also be used in the Dial String for CBCP callback.
IPMI 1.5 implementations do not check for illegal characters in dial strings. It is the responsibility of
configuration software to ensure that correct characters are entered.
149
Intelligent Platform Management Interface Specification
13.7 Terminal Mode
Terminal Mode is an operating mode of a serial/modem channel used for the following purposes:
•
It provides a printable text-based mechanism for delivering IPMI between a terminal or remote console and
the BMC. The text-based approach makes it simpler to develop script-based tools for generating and
handling IPMI messages.
•
It provides a small number of ASCII-text based commands to enable a small number of basic recovery and
status functions to be executed when only a dumb terminal is available in lieu of real system management
software.
13.7.1 Terminal Mode Versus Basic Mode Differences
Terminal Mode is primarily intended for local use rather than remote use via a modem. The following are the
main differences between terminal mode and basic mode operation:
•
If password protection is desired, only ‘plain text’ passwords can be used. Password characters are
restricted to be from the printable set of ASCII characters as defined in Appendix E - Terminal Mode
GrammarAppendix E - Terminal Mode Grammar.
•
Passwords can be entered two ways in Terminal Mode: either via the Get Session Challenge / Activate
Session command, or via an ASCII Text command.
•
Terminal Mode does not utilize checksums on IPMI messages or ASCII Text commands. If a modem
connection is used, the modem should be configured for error correction, or Basic Mode should be used
instead.
•
Terminal Mode remote console is limited to a single, fixed, single Software ID (SWID). See Table 5-4,
System Software IDsTable 5-4, System Software IDs. The fixed SWID is used where a requester’s SWID
would have been extracted from the IPMI Message. For example, if Terminal Mode IPMI Messaging is
used to generate a Platform Event Request message (Event Message) the SEL Record would contain the
fixed SWID identifying the Terminal Mode remote console.
•
The Terminal Mode remote console is limited to a single LUN (00b). This LUN is implicit in the message
format. When Terminal Mode request or response messages are bridged to other media, the value 00b is
used as requester’s or responders LUN, respectively.
•
Terminal Mode messages delivered to SMS via BMC LUN 10b always go to SMS Software ID 20h (41h)
LUN 00b, unless the Send Message command is used to put the message in the receive message queue.
•
Callback is not supported for Terminal Mode. You can trigger a callback from Terminal Mode, but the
party that is called must support either Basic Mode or PPP Mode.
13.7.2 Terminal Mode Message Format
Terminal mode messages are of the general format:
[<message data>]<newline>
The left-bracket and right-bracket+<newline> characters serve as START and STOP delimiters for the
message. Note that the right-bracket and <newline> characters together form the sequence that indicates the end
of the message. <newline> characters may appear within the message as a result of input line editing and multiline output message data.
150
Intelligent Platform Management Interface Specification
13.7.3 IPMI Message Data
IPMI Messages are sent and received in Terminal Mode <message data> as a series of case-insensitive hexASCII pairs, where each is optionally separated from the preceding pair by a single <space> character. The
following is an example of an IPMI Request message in Terminal Mode:
[18 00 22]<newline>
The Terminal Mode Request Message field definitions follow those used for the Basic Mode except that there
is no Slave address / Software ID field or LUN information for the requester. The software ID and LUN for the
remote console are fixed and implied by the command. The SWID for messages to the remote console is always
40h, and the LUN is 00b.
Instead, there is a ‘bridge’ field that is used to identify whether the message should be routed to the BMC’s
bridged message tracking functionality or not.
Figure 13-7, Terminal Mode Request to BMC
Byte 1
NetFn (even) / rsLUN=00b (BMC)
Byte 2
rqSeq / Bridge=00b (BMC)
Byte 3
cmd
Byte 4:N
data
The following figure shows the corresponding format of a response message from the BMC.
Figure 13-8, Terminal Mode Response from BMC
Byte 1
NetFn (odd) / rsLUN=00b (BMC)
Byte 2
rqSeq/Bridge=00b (BMC)
Byte 3
Cmd
Byte 4
Completion Code
Byte 5:N
Data
151
Intelligent Platform Management Interface Specification
13.7.4 Terminal Mode IPMI Message Bridging
The terminal mode message includes a ‘bridge’ field that is used to determine whether the message is going to
or coming from the BMC’s command functionality, or to/from the BMC’s ‘bridge’ tracking functionality.
The message is interpreted based on the value of the bridge field, whether the message is a request or response,
and the message direction per the following table.
Note that messages to and from the system interface are transferred using the BMC SMS LUN,
10b, with the bridge field set to 00b.
Support for Terminal Mode IPMI Message Bridging is optional.
Table 13-12, Terminal Mode Message Bridge Field
Bridge
Field
Request/
Response
Message
Direction
(to BMC)
00b
Request
In
00b
Response
Out
00b
Request
00b
LUN
Message Interpretation
Remote Console request to BMC functionality
Message is a request from the remote console to the BMC
In
00b,
01b,
11b
00b,
01b,
11b
10b
Response
Out
10b
00b
Request
Out
00b
Response
In
00b
00b
01b
Request
Response
Request
Out
In
In
00b,
01b,
11b
00b,
01b,
11b
10b
10b
any
01b
Response
Out
any
01b
Request
Out
any
01b
Response
In
any
Response to Remote Console from BMC functionality
Message is a response to an earlier request from the remote console to the BMC
Remote Console request to SMS
Message is a request from the remote console to SMS via the Receive Message
Queue
SMS Response to Remote Console
Message is a response to an earlier request from SMS
Asynchronous Request to Remote Console from BMC
Remote Console Response to earlier Asynchronous Request from BMC
Asynchronous Request from SMS to Remote Console
Remote Console Response to earlier Asynchronous Request from SMS
ILLEGAL COMBINATION
The remote console bridges requests to other media by encapsulating the
message content in a Send Message command to the BMC functionality.
Response to earlier Bridged Request from Remote Console
Message is the asynchronous response from an earlier bridged request that was
encapsulated in a Send Message command issued to the BMC by the remote
console.
Asynchronous, bridged request to remote console from other media
Message is a bridged request to the remote console from another media, e.g. the
system interface. BMC assigns the sequence number as part of bridging.
Remote Console response to earlier asynchronous request from another
media Message is a response from the remote console to an earlier bridged
request from another media. BMC uses the sequence number in the response to
determine how to route the response to the original requester.
13.7.5 Sending Messages to SMS
Terminal Mode uses BMC LUN 10b to send messages to SMS (system interface) via the Receive Message
Queue in the BMC. The following shows the format of a request message delivered to SMS, and the
corresponding response.
152
Intelligent Platform Management Interface Specification
Figure 13-9, Terminal Mode Request to SMS
Byte 1
NetFn (even) /
rsLUN=10b (BMC to SMS)
Byte 2
rqSeq =XX /
Bridge=00b (BMC)
Byte 3
cmd
Byte 4:N
data
Figure 13-10, Terminal Mode Response from SMS
Byte 1
NetFn (odd) /
rsLUN=10b (BMC to SMS)
Byte 2
rqSeq =XX /
Bridge=00b (BMC)
Byte 3
cmd
Byte 4
completion
code
Byte 5:N
data
13.7.6 Sending Messages to Other Media
The Send Message command is used to deliver a message to a different media, e.g. IPMB. The following figure
illustrates the data contents that would be used in a Send Message command to deliver a Terminal Mode request
to another (non-system interface) channel:
This data would be carried in a Send Message command of the following format:
Figure 13-11, Send Message Command for Bridged Request
NetFn (even) / BMC_LUN
rqSeq / bridge=00b
cmd = Send Message
channel #
session handle
rsSWID
chk1
rqSWID=81h
netFn (even)/rsLUN
rqSeq/rqLUN=00b
(terminal mode console)
<data>
chk2
cmd
The BMC will return a Send Message response matching the Send Message request. This will normally be
returned immediately after the request.3
Figure 13-12, Response to Send Message Command for Bridged Request
NetFn (odd) / BMC_LUN
rqSeq / bridge=00b
cmd = Send Message
completion code
= 00h (OK)
Later, the bridged response will be returned. The following figure shows the contents of a corresponding
bridged response to the Remote Console:
Figure 13-13, Bridged Response to Remote Console
Byte 1
netFn (odd) / rsLUN
Byte 2
rqSeq / bridge=01b
Byte 3
Cmd
Byte 4
Completion Code
Byte 5:N
Data
Note that much of the targets addressing information (rqSWID, rqLUN) is absent from the response. The
remote console must use the original request’s sequence number (rqSeq), netFn/rsLUN, and command values to
match bridged response up with the earlier bridged request. These fields are highlighted with bold in the
preceding Send Message and Bridged Response figures.
3
Note that because IPMI messaging allows for other messages to appear between requests and responses, it is possible that one
or more asynchronous messages could appear between the Send Message request and response. Console software should be
prepared to handle such occurrences.
153
Intelligent Platform Management Interface Specification
13.7.7 Terminal Mode Packet Handshake
There is a configuration option that allows the BMC to output a character sequence that indicates when its input
buffer is ready to accept another IPMI Message via Terminal Mode. This option is typically used with
automated applications that send and receive IPMI Messages using Terminal Mode. The BMC outputs the
following character sequence whenever there is space for a new input message from Terminal Mode, and the
‘handshake’ option is enabled:
[SYS]<newline>
If a message transmission from the BMC is already in progress, the handshake sequence will be held-off until
the present message transmission has completed. The BMC will typically output the handshake sequence within
1ms of the buffer space becoming available and the present message transmission (if any) completing.
13.7.8 Terminal Mode ASCII Text Commands
A small number of ASCII-text commands can be delivered while in terminal mode. The following table lists
these commands. Commands are CASE SENSITIVE. Appendix E - Terminal Mode GrammarAppendix E Terminal Mode Grammar, lists the rules for the format of terminal mode input and output for both IPMI
messages and text commands. Refer to Table 13-13, Terminal Mode Examples for some examples of Terminal
Mode text command and IPMI messages.
Table 13-13, Terminal Mode Text Commands
Command Text
Description
SYS PWD -U USERNAME
<password>
Used to activate a terminal mode session. USERNAME corresponds to the ASCII text
for the username. <password> represents a printable password (up to 16 characters). If
<password> is not provided, then a Null password (all binary 0’s) is submitted.
Passwords are case sensitive.
SYS PWD -N <password>
SYS PWD -X
SYS TMODE
SYS SET BOOT XX YY
ZZ AA BB
SYS SET BOOTOPT NN
XX…NN
SYS GET BOOTOPT XX
YY ZZ
154
Either the SYS PWD command (or Activate Session IPMI message) must be
successfully executed before any command or IPMI messages will be accepted. Note
that a modem connection may be automatically dropped if multiple bad passwords are
entered.
-N represents a Null username. <password> represents a printable password (up to 16
characters). If <password> is not provided, then a Null password (all binary 0’s) is
submitted. Passwords are case sensitive.
Either the SYS PWD command (or Activate Session IPMI message) must be
successfully executed before any command or IPMI messages will be accepted. Note
that a modem connection may be automatically dropped if multiple bad passwords are
entered.
-X immediately ‘logs out’ any presently active session. Entering an invalid password with
-U or -N will also have the same effect.
Used as a ‘no-op’ confirm that Terminal Mode is active. BMC returns an OK response
followed by “TMODE”.
Sets the boot flags to direct a boot to the specified device following the next IPMI
command or action initiated reset or power-on. XX…BB are five hex-ASCII bytes for the
boot flags parameter in the Boot Options Parameters. See Table 22-12, Boot Option
ParametersTable 22-12, Boot Option Parameters.
Upon receiving this command, the BMC will also set the ‘valid bit’ in the boot options,
and will set all the Boot Initiator Acknowledge data bits to 1b.
This is essentially a text version of the IPMI “Set System Boot Options” command,
allows any of the boot option parameters to be set, not just the boot flags. XX…NN
represents the hex-ascii for the data bytes that are passed in the Set System Boot
Options request.
This is essentially a text version of the IPMI “Get System Boot Options” command,
allows any of the boot option parameters to be set. XX YY ZZ represents the hex-ascii
for the data bytes that are passed in the Get System Boot Options request. The BMC
Intelligent Platform Management Interface Specification
SYS SET TCFG
SYS SET TCFG -V XX YY
SYS SET TCFG -N XX YY
SYS RESET
SYS POWER OFF
SYS POWER ON
SYS HEALTH QUERY
returns the data from the command in hex-ascii format, with a maximum of four hexascii pairs per line.
Returns the Terminal Mode Configuration bytes where XX and YY represent hex-ascii
encodings for the volatile version of data bytes 1 and 2 as specified in the Terminal
Mode Configuration parameter (#29) listed in Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters, and AA BB represent
hex-ascii encoding of the non-volatile version.
V:XX YY<output termination sequence>
N:AA BB<output termination sequence>
This command sets the volatile Terminal Mode Configuration. XX and YY represent
hex-ascii encodings for data bytes 1 and 2 as specified in the Terminal Mode
Configuration parameter (#29) listed in Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters. The BMC returns the
same output as for SYS SET TCFG, above.
This command sets the non-volatile Terminal Mode Configuration. XX and YY represent
hex-ascii encodings for data bytes 1 and 2 as specified in the Terminal Mode
Configuration parameter (#29) listed in Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters. The BMC returns the
same output as for SYS SET TCFG, above.
Directs the BMC to perform an immediate system hard reset.
Directs the BMC to perform an immediate system power off.
Causes the BMC to initiate an immediate system power on
Causes the BMC to return a high level version of the system health status in ‘terse’
format. The BMC returns a string with the following format if command is accepted.
PWR:zzz H:xx T:xx V:xx PS:xx C:xx D:xx S:xx O:xx
Where:
PWR
H
T
V
PS
F
D
S
O
is system POWER state
is overall Health
is Temperature
is Voltage
is Power Supply subsystem
is cooling subsystem (Fans)
is Hard Drive / RAID Subsystem
is physical Security
is Other (OEM)
zzz is: “ON”, “OFF” (soft-off or mechanical off), “SLP” (sleep - used when can’t
distinguish sleep level), “S4”, “S3”, “S2”, “S1”, “??” (unknown)
SYS HEALTH QUERY -V
and xx is: ok, nc, cr, nr, uf, or ?? where:
“ok” = OK
(monitored parameters within normal operating ranges)
“nc” = non-critical
(‘warning’: hardware outside normal operating range)
“cr” = critical
(‘fatal’ :hardware exceeding specified ratings)
“nr” = non-recoverable (‘potential damage’: system hardware in jeopardy or
damaged)
“uf” = unspecified fault (fault detected, but severity unspecified)
“??” = status not available/unknown (typically because system power is OFF)
Causes the BMC to return a high level version of the system health status in multi-line
‘verbose’ format. The BMC returns a string of the following format:
SYS Health:xx<output termination sequence>
Power: “ON”, “OFF” (soft-off or mechanical off), “SLEEP” (sleep - used when can’t
distinguish sleep level), “S4”, “S3”, “S2”, “S1”, “Unknown”
Temperature:xx<output termination sequence>
Voltage:xx<output termination sequence>
155
Intelligent Platform Management Interface Specification
PowerSystem:xx<output termination sequence>
Cooling:xx<output termination sequence>
Drives:xx<output termination sequence>
Security:xx<output termination sequence>
Other:xx<output termination sequence>
Where xx is:
“OK”
“Non-critical”
“Critical”
“Non-recoverable”
“Unspecified fault”
“Unknown”
SYS IDENTIFY
SYS IDENTIFY –ON
<XX>
SYS IDENTIFY –OFF
SYS XXXXXX yy..zz
156
(monitored parameters within normal operating ranges)
(‘warning’: hardware outside normal operating range)
(‘fatal’ :hardware exceeding specified ratings)
(‘potential damage’: system hardware in jeopardy or damaged)
(fault detected, but severity unspecified)
(status not available/unknown (typically because system power
is OFF)
Causes the BMC to signal the system’s location (e.g. with a blinking led or beep). This
is intended to locate the system amongst a rack of systems. The BMC will signal the
system’s location for 15 seconds and then stop signaling. This is a text version of the
optional Chassis Identify command.
Causes the BMC to signal the system’s location (e.g. with a blinking led or beep) for a
specific amount of time. XX is an optional hex-ASCII byte representing the number of
seconds the BMC is to cause the system to identify itself. If XX is not supplied, the
BMC will signal the system’s location for 15 seconds and then stop signaling. This is a
text version of the optional Chassis Identify command.
Causes the BMC to stop signaling the system’s location. This has no effect if the
system is not currently identifying itself. This is a text version of the optional Chassis
Identify command.
OEM Text Commands (optional, vendor-specific). All OEM text commands are prefixed
with SYS followed by XXXXXX where XXXXXX is the OEM ID expressed as a six-digit
hex-ASCII number. For example, the IANA OEM IDs for Intel, HP, Dell, and NEC are
000157h (343), 00000B (11), 0002A2h (674), and 000077 (119), respectively. yy..zz
represents OEM-specific text.
It is recommended that OEM Text Command implementations use the same OK and
ERROR completion returns be used for OEM Commands as for the IPMI-specified text
commands.
Intelligent Platform Management Interface Specification
13.7.9 Terminal Mode Text Command and IPMI Message Examples
The following table presents some examples of terminal mode commands and IPMI messages.
Table 13-14, Terminal Mode Examples
Console input:
[SYS TMODE]<crlf>
BMC responds:
Console input:
BMC responds:
Console input:
[OK TMODE]<crlf>
[SYS PWD -U Fred letME1n]<crlf>
[OK]<crlf>
[SYS PWD -N]<crlf>
BMC responds:
Console input:
BMC responds:
Console input:
BMC responds:
Console input:
BMC responds:
Console input:
[ERR CC]
[SYS RESET]<crlf>
[OK]<crlf> and resets system.
[sys blah]<crlf>
[ERR C1]<crlf>
[sys health query -V]<crlf>
[OK<crlf>
Health:Critical<crlf>
Temperature:OK<crlf >
Voltages:OK<crlf>
Drive Subsystems:OK<crlf>
Power System:OK<crlf>
Cooling:Critical<crlf>
Security:OK<crlf>
Other:OK]<crlf>
[18 xx 22]<crlf>
BMC responds:
[1C xx 22 00]<crlf>
Console input:
BMC responds:
[SYS 000157 My Command]<crlf>
[OK 000157 My Response]<crlf>
TMODE is a ‘no-op’ command used to confirm the
BMC is operating in terminal mode.
User submits password for username Fred.
User attempts to activate session with anonymous
login (null username, null password)
BMC returns error, e.g. ‘invalid data field’.
User resets system.
User enters an invalid command.
Verbose system health query.
IPMI Reset Watchdog Timer request message to
BMC. xx represents the console selected sequence
number and LUN field for the request.
Reset Watchdog Timer response message from
BMC. The same sequence number and LUN passed
in the request is returned in the response.
Submit an OEM text command
Get an OEM text response
13.8 Terminal Mode Line Editing
Since direct human input is likely to be used with Terminal Mode, it is useful to support a limited amount of
editing to reduce the effort required to recover from the inevitable typo’s that occur during text entry. Line editing
is an operating mode of Terminal Mode. Line editing should be enabled when direct human entry is used, and
disabled when automated entry is used.
•
Line editing is enabled or disabled via an option in the serial/modem configuration parameters.
•
Enabling line editing disables input time-outs.
•
When line editing is enabled, echo should also be enabled.
•
When line editing is enabled, the Serial/Modem Connection Active (Ping) message should be disabled.
Otherwise, unrequested Ping messages will appear in the data stream.
157
Intelligent Platform Management Interface Specification
•
The <backspace> or <delete> key can be used to delete the last character entered.
•
The <ESC> character can be used to delete the entire message. An <ESC> (1Bh) character received by the
BMC immediately flushes any pending input message data. If line editing is enabled, and the <ESC> is
followed by an input newline, the BMC responds by putting out an output newline sequence (typically <crlf>). Otherwise, the BMC just silently flushes the data and goes back to looking for a START character.
•
Any illegal characters received after the START character will silently flush the message in progress. The
difference between this and <ESC> is that
•
Long IPMI message lines can be split across multiple lines by using a line continuation <backslash>
character following immediately by the input newline sequence.
•
Line continuation character support is optional for the text commands, because they’re considered to be short
enough to fit one line.
•
Line continuation character support for OEM messages is an implementation option.
13.9 Terminal Mode Input Restrictions
The following restrictions and characteristics apply to terminal mode:
•
Up to 80 printable characters are required to be supported for one line. The BMC can stop accepting new
characters and stop echoing input when the 80 character limit is reached (with the exception of the <ESC>,
<backspace>/<delete>, illegal character, and input <newline> characters, which will still be handled).
•
The interface must support the maximum IPMI input message length that is supported on the given BMC. Per
Section 6.13, Message Size & Private Bus Transaction Size RequirementsMessage Size & Private Bus
Transaction Size Requirements, this will typically be 40 bytes. Since each message byte can require three
input characters (two hex-ASCII digits, plus a <space> character) a 40-byte IPMI message could require 120
characters, plus the starting and ending brackets, or 122 characters, total, for the message.
13.10 Page Blackout Interval
The Page Blackout interval determines the minimum number of minutes between successive pages. The purpose
for this parameter is to provide a mechanism to prevent someone from getting back-to-back pages if a flurry of
events occurs. The interval applies to Dial Pages and TAP Pages. It does not apply to Dial-out PET Alerting.
The Page Blackout Interval does not turn off Platform Event Filtering or associated actions. Platform Event
filtering continues while a page or blackout interval is in progress. The BMC will accumulate the set of pending
actions that occur during the page and blackout interval. If an event triggers a power-down action, the page will
be aborted, the power-down will occur, and the page restarted after the power down. If a reset or power-cycle
action is triggered, that action will be held off until the paging process and blackout interval have concluded.
The Page Blackout Interval setting is kept in the serial/modem configuration parameters managed by the BMC.
13.11 Dial Paging
Dial Paging is accomplished by the BMC using the dialing capabilities of an external modem to connect to a
paging service and enter a page using telephone numeric keypad numbers. Once a connection is established, the
BMC delivers the specified Alert String from the PEF Configuration Parameters to the modem. The paging
string directs the modem to deliver a fixed number to the paging service. The paging string is often used to
deliver the phone number of the system that is generating the alert. An administrator receiving the string can then
call the system with a management application and use IPMI messaging to retrieve system status, SEL entries,
and other information about the alert.
158
Intelligent Platform Management Interface Specification
13.11.1 Alert Strings for Dial Paging
Modern modems deploying the modem ‘AT’ command set [TIA-602] contain character options associated with
the dial command that utilize built-in call detection features of the modem. Thus, the call progress detection
requirements on the BMC are greatly simplified. Though the majority of modems will support all of the
following options, some are not mandatory in [TIA-602]. The modem’s documentation should be consulted to
verify support for character options prior to configuring non-volatile dial strings. The Alert Immediate
command can be used to test dial strings before committing them to non-volatile storage.
The following character options can be used in the Alert String for a Dial Page. These options follow the
modem dial command (default =‘ATD’) issued by the BMC:
P
R
S=n
T
W
@
,
;
!
Dial using Pulse. Dialing digits after the ‘P’ will be sent using pulse dialing
Reverse Frequencies. Forces modem to dial out at the answer frequency
Dial pre-stored phone number, n
Dial using Tone. Dialing digits after the ‘T’ will be sent using touch tones
Wait for dial tone
Wait for quiet (answer)
Pause (2 seconds)
Return to command mode after dialing
Flash the switch hook
13.11.2 Dialing Digits
Per [TIA-602] the dialing digits consist of the ASCII characters 0..9, *, #, A, B, C, and D.
13.11.3 <Enter> Character (control-M)
The BMC recognizes the “control-M” character (0Dh) as an <ENTER> character. When the BMC encounters
this character it transfers it to the modem and then delays 1 second before sending any remaining characters in
the page string. Note that the BMC automatically issues an <ENTER> character after sending out the Dial
String when performing a Dial Page.
13.11.4 Long Pause Character (control-L)
The BMC also recognizes the “control-L” character (0Ch) as the trigger for generating a 10-second ‘long
pause’ sequence. When the BMC encounters this character, it doesn’t send it to the modem but instead delays
10 seconds before sending any remaining characters in the page string.
13.11.5 Empty (delimiter) Character (FFh)
The BMC recognizes FFh in the page string as a ‘delimiter’ character used by BIOS. The character is provided
as a potential aid to for interfaces, such as BIOS setup, that may split the page string into multiple fields for
presentation to the user. The BMC ignores the character and does not transfer it to the modem.
13.11.6 ‘Null’ Terminator Character (00h)
The BMC recognizes this character as a terminator for the Dial String. This terminator is used whenever the
Dial String data consists of fewer characters than the maximum length for the Dial String. Note that the BMC
automatically issues an <ENTER> character after sending out the Dial String when performing a Dial Page.
159
Intelligent Platform Management Interface Specification
13.12 TAP Paging
TAP (Telocator Access Protocol) is a popular protocol for sending an alphanumeric page by connecting to a
paging service using a serial modem. IPMI supports TAP as an option for delivering a short alert page to a remote
paging service. This capability is referred to as TAP Paging. TAP Paging is triggered from the IPMI Platform
Event Filtering capability. It can also be triggered ‘manually’ via the Serial/Modem Connection Active
(Ping)Serial/Modem Connection Active (Ping) command.
The protocol is described in the [TAP]. Per [TAP], there are two types of remote page entry: automated and
manual. The TAP implementation for IPMI operates using the management controller as an automated entry
device.
A TAP Paging transaction consists of one or more blocks. A block can contain up to 250 characters of
information. Each block contains one or more short message strings that form a message sequence. Message data
can span blocks. The short message strings are also referred to as ‘Fields’ in the TAP specification. Since only a
small number of characters are delivered in an IPMI TAP Page, only a single block will be transmitted.
There are typically two fields within the first block. The first field, Field #1, is called the Pager ID field. Some
paging services refer to this as the PIN (Pager ID Number). This field is used to identify the target pager. The
second field, Field #2, is the alphanumeric paging message. TAP only directly supports delivery of 7-bit ASCII
characters. There is an associated protocol for transmitting 8-bit characters via TAP, but that protocol is not
supported in this specification.
TAP includes provision for an optional alphanumeric six-character password for the paging service. The
password is also set via the serial/modem configuration parameters.
Appendix F - TAP Flow SummaryAppendix F - TAP Flow Summary, presents an additional overview and
implementation notes for TAP paging via a BMC.
13.12.1 TAP Escaping (data transparency)
TAP allows ASCII control characters (00h to 20h) to be sent in the alphanumeric paging transaction as twocharacter sequences. TAP requires that the characters be escaped per the following table. The BMC
automatically performs escaping when transmitting TAP messages.
Table 13-15, TAP Escaping
Character
<EOT>
<STX>
<ETX>
<LF>
<CR>
<ETB>
<SUB>
<ESC>
Other Control Characters
value
04h
02h
03h
0Ah
0Dh
17h
1Ah
1Bh
-
Escaped as:
1Ah, 44h
1Ah, 42h
1Ah, 43h
1Ah, 4Ah
1Ah, 4Dh
1Ah, 57h
1Ah, 5Ah
1Ah, 5Bh
1Ah, (value + 40h)
e.g. 20h Æ 1Ah, 60h
Escaping mandatory?
yes
yes
yes
yes
yes
yes
yes
yes
optional
[TAP] states that “any [optional] control character may be made transparent at the implementor’s discretion”.
The IPMI serial/modem configuration options for TAP paging include a control-character map similar to that
used in PPP so that the user can configure which control characters get escaped for delivery to a particular TAP
service.
13.12.2 TAP Checksum
TAP messages include their own checksum format. The checksum is transmitted using three printable ASCII
characters. Refer to [TAP] for the checksum algorithm.
160
Intelligent Platform Management Interface Specification
13.12.3 TAP Response Codes
Per [TAP], each handshake message that is returned from the paging service is specified to start with a 3character response code. The values for these codes are specified in [TAP]. The implementation can optionally
return a set of the last TAP Response Codes using the Get TAP Response Codes CommandGet TAP Response
Codes Command as an aid to TAP connection setup and debugging.
13.12.4 TAP Page Success Criteria
The management controller can use the TAP response codes to determine whether a page was successfully sent
or not. One of the following response codes, received after the management controller sends an ‘End-ofTransaction’ <ETX> sequence, are used a positive indication of a successful page. Optionally, the page can be
considered successful if an <ACK> is received following the end-of-transaction. The serial/modem
configuration parameters include a setting that selects which of these confirmation mechanisms is used.
Table 13-16, TAP Success Codes
code
211
213
TAP Definition
Page(s) Sent Successfully
Message accepted - held for deferred delivery.
13.13 PPP Alerting
PPP Alerts are accomplished by the BMC connecting to a remote LAN via a PPP account and then delivering a
UDP Datagram that contains an SNMP Trap formatted per the IPMI Platform Event Trap (PET) Format
specification. Information for the PET trap comes from the Event Message that generated the alert and from the
serial/modem configuration parameters for PET.
161
Intelligent Platform Management Interface Specification
14. Event Messages
Event Messages are special messages that are sent by management controllers when they detect significant or
critical system management events. This includes messages for events such as ‘temperature threshold exceeded’,
‘voltage threshold exceeded’, ‘power fault’, etc. The Event Message generator (the device generating an Event
Message) notifies the system of the event by sending an “Event Request Message” to the Event Receiver Device.
When the Event Receiver gets a valid Event Message, it sends a response message to the generator of the Event
Message. It then typically transfers the message to the System Event Log. The Event Receiver does not interpret the
Event Messages it receives. Thus, new Event Message types can be added into the system without impacting the
Event Receiver implementation.
In some systems, the Event Receiver will need to interrupt the system to notify it that there is an Event Message to
be logged. It is desirable for the implementation to have verified and buffered Event Messages in their entirety
before issuing such an interrupt. This way, the interrupt handler will not need to wait for the Event Message
transmission to complete first.
14.1 Critical Events and System Event Log Restrictions
The platform’s System Event Log is typically of limited size (~3 to ~8 KB, depending on implementation).
Therefore, it is important to refrain from filling the System Event Log with non-critical ‘clutter’.
The System Event Log is primarily intended for capturing Critical Events. These include events that require
immediate logging to guarantee that they’re available for ‘post-mortem’ analysis, and events that may require
quick system responses, such as system power off, or shutdown.
Critical events include out-of-range temperature and voltage events, hardware failures such as power supply or
fan failures, interrupts and signals that affect system operation such as NMIs and PCI PERR (parity error) and
SERR (system error). Critical Events also include events that impact system data integrity, such as the
uncorrectable ECC errors, or system security, such as ‘chassis intrusion’.
In addition to events that indicate ‘failure’ conditions, events that indicate impending failures are also considered
to be critical events. This includes events for reaching ‘warning levels’ for things such as system temperature or
error counts. The assertion of ‘Predictive Fault’ information is also considered critical, particularly if the
monitored device does not have a direct ‘failure’ indication.
Non-critical events, such as the return to an ‘OK’ state from a ‘Warning’ state should not be sent as critical
events. Non-critical system information is normally obtained by System Management Software polling sensors
and management controllers for their status.
162
Intelligent Platform Management Interface Specification
IPMB Events
PCI Mgmt. Bus
events
System Interface
SMS Events
BIOS Events
Table 14-1, Event Message Reception
IPMB Interface
PCI Mgmt. Bus
Event Msg. Buffer
Event Receiver
SEL Mgr.
PEF
BMC Internal
Events
NV Storage I/F
External Event
Messages
NV Storage
SEL
Data
The preceding figure presents a conceptual illustration of the manner in which Event Messages can be handled by a
Baseboard Management Controller device that uses an external non-volatile storage device to hold the System
Event Log.
The figure shows a BMC with a shared system messaging interface where Event Messages can be delivered from
either BIOS, SMS (system management software / OS), or an SMI Handler, and an IPMB interface and through
which it can receive Event Messages from the Intelligent Platform Management bus. The BMC can also generate
‘internal’ Event Messages.
When the BMC receives a message via the system or IPMB interfaces, a ‘Message Handler’ function recognizes the
message as being for the ‘Event’ functionality in the BMC and passes the message information on to the ‘Event
Receiver’ function. The Event Receiver function then takes the message content and issues a request to a ‘SEL
Mgr.’ function that formats the message as an SEL Entry and calls the FLASH Interface to have the data stored.
The Event Receiver function is also responsible for driving the response message back through the messaging
system. This way, message acknowledgment or error reporting can be provided.
14.2 Event Receiver Handling of Event Messages
This section presents some implementation advice for the Event Receiver device. Please refer to the Intelligent
Platform Management Bus Communications Protocol Specification for additional information on Event Message
handling.
Since retries of Event Messages are part of the IPMB protocol, there is the potential for the Critical Event Handler
to receive more than one Event Messages for the same event. The Seq field allows repeated Event Messages to be
discriminated from new Event Messages. Event Messages from a Event Generator that match an earlier Event
Message can be ignored.
The option to disable SEL Logging only affects events that are received from the IPMB and PCI Management
Bus interfaces. Devices on the IPMB and PCI Management Bus are more likely to generate events ‘automatically’
while the other interfaces are primarily driven by either local or remote software which is assumed to have more
control as to whether it generates events or not.
163
Intelligent Platform Management Interface Specification
It is recommended that Event Receiver keep a table or queue of the Event Messages it has received. Any new
event message from the same source and of the same type, but with a different sequence number, would replace
the previous entry.
There are many ways to implement such a table or queue. Any implementation should provide enough tracking
support to handle previously received Event Messages for all the ‘known’ Event Generators in the basic system.
For example, a system that has four management controllers on the IPMB that can generate Event Messages
should track the previously received Event Messages from those devices.
It is desired that the Event Receiver can track at least six additional Event Generators to cover additional Event
Generators that are added into the system. (One common add-on would be an emergency management. Other
possible ‘add-on’ event generators would be other systems and peripheral boxes in a “managed cluster”
arrangement).
The Event Receiver implementation should account for the possibility that there can be more different Event
Generators than there are slots in the table. This can be managed by implementing the table with an ‘LRU’
deletion algorithm, where the oldest tracked Event Messages are deleted if a new Event Message comes in and the
table or queue is full. It can be assumed that there will rarely be more than two event messages that would be in
the state where they are to be re-transmitted because of a lost acknowledge.
With this type of design, the most anomalous behavior would be the multiple recording of the same event. This
would only be seen under artificially generated ‘stress’ testing and would only be able to occur if there were more
event message sources than table slots.
It is also recommended that the Event Receiver implement the ‘Seq Timeout’ as specified in the IPMB
Communications Protocol specification.
14.3 IPMB Seq Field use in Event Messages
This section presents a review of the IPMB Seq field and the manner in which it is used when Event Messages are
delivered via the IPMB.
The Event Receiver uses the Seq field to reject retried (duplicate) Event Request Messages that it may receive.
The Event Generator will re-send an Event Request Message if it does not receive the Event Response Message.
It is possible that the response could get corrupted, causing the Event Generator to re-send the original request
even though the Event Receiver had already successfully received it. This is one way that an Event Receiver
could get more than one Event Request Message for the same event. When the Event Generator re-sends the
Event Request Message, it does so with the same Seq value that it used for the original try. The Event Generator
will increment the Seq value the next time it has a new Event Request Message to send.
When Event Messages are delivered via the IPMB, the IPMB message’s Seq field is used to allow Event Receiver
to discriminate whether the Event Message is for a new occurrence of a given event, or is a re-transmission of a
previous Event Message for that event. The IPMB Seq field should not be confused with being a sequence
number for tracking multi-message transfers, as might be its use in other serial protocols.
If the Event Receiver receives an Event Message where the Cmd, NetFn, LUN, and Seq fields match the previous
event message from the same Requester, it can assume that the latter message is a re-transmission and return a
‘normal completion’ (00h) as a response to valid, duplicated requests. The Event Receiver does not log duplicate
events.
If the Event Receiver does not return a response, the Event Generator retries up to its retry limit count and then
concludes that the Event Request failed. Event Generator devices on the IPMB do not send new Event Messages
until they’ve finished sending the previous Event Message (including retries). This eliminates the need for the
Event Receiver to maintain status for multiple Seq numbers from a single Event Generator.
164
Intelligent Platform Management Interface Specification
The data fields for the Event Request Message are not included in the comparison. This is because the Event
Request Message may return a data field that reflects a ‘present state’ or data value that could vary with each
retry.
Refer to the Intelligent Platform Management Bus v1.0 Communications Protocol Specification for more
information on the Seq field.
14.4 Event Status, Event Conditions, and Present State
A sensor tracks present state and Event Conditions. An Event Condition is that set of comparisons applied to the
present state and previous state that produces a given Event Status.
A management controller typically polls for Event Conditions. When it sees a condition become active, it updates
the Event Status for the sensor. The process of updating the present state Event Status is referred to as Scanning
or Sensor Scanning.
The Event Status is those bits that are reported in the Get Sensor Event Status command. As long as scanning is
enabled, the Event Status bits will be updated according to changes in Event Status. This is independent of
whether Event Messages are generated on a given event. That is, turning off Event Message Generation for a
particular state does not turn off scanning or updates of the Event Status.
The Get Sensor Reading command returns State Bits reflecting the present state of the sensor. If the sensor is an
‘auto- re-arm’ sensor, these bits can also represent the Event Status if hysteresis is factored in. Thus, the Get
Sensor Events command is optional for auto- re-arm sensors. An application uses the masks in the SDR to
determine which bits reflect both current state and event status, and which bits reflect current state only.
The condition that causes an Event Message to be sent is referred to as the 'Event Trigger'. The classification of a
sensor indicates whether the corresponding event was discrete, or threshold-based. The sensor classification is
part of the Event/Reading Type Code (see section 36.1, Event/Reading Type CodesEvent/Reading Type Codes).
A reading/state unavailable (formerly “initial update in progress”) bit is provided with the Get Sensor Reading
and Get Sensor Event Status commands to indicate to software that it must ignore the reading and/or state
information because the BMC cannot obtain a valid reading and/or state information. This can occur in situations
where the sensor is monitoring an entity that may or may not be present, such as with hot-swap devices. For
example, if a sensor monitors the temperature of a hot-swap power supply, the reading/state unavailable bit can
be used to indicate that no valid temperature reading is available because the power supply is not installed. The
bit can also indicate when a reading or state is unavailable because a sensor is re-arming (see Section 14.6, RearmingRe-arming, below.
14.5 System Software use of Sensor Scanning bits & Entity Info
System software must ignore any sensor that has the sensor scanning bit disabled - if system software didn’t
disable the sensor. This provides an alternate mechanism to allow the management controller to automatically
adjust the sensor population without requiring a corresponding change of the sensor data records. For example,
suppose the management controller has a way of automatically knowing that a particular temperature sensor will
be absent in a given system configuration if a given processor is also absent. The management controller could
elect to automatically disable scanning for that temperature sensor. System management software would ignore
that sensor even if it was reported in the SDRs.
Note that this is an alternate mechanism that may be useful in some circumstances. The primary mechanism is to
use the Entity ID information in the SDRs, and combine that information with presence detection for the entity.
If there is a presence detection sensor for a given entity, then system management software should ignore all other
sensors associated with that entity. Some sensors have intrinsic support for this. For example, a sensor-specific
Processor sensor has a ‘Processor Presence’ bit. If that bit is implemented, and the processor is absent, any other
sensors and non-presence related bits associated with that processor can be ignored. If the sensor type doesn’t
165
Intelligent Platform Management Interface Specification
have an intrinsic presence capability, you can implement an ‘Entity Presence’ sensor. This sensor solely reports
whether a given Entity is present or not.
14.6 Re-arming
Re-arm refers to resetting internal device state that tracks that an event has occurred on the sensor. After a sensor
is re-armed the device will re-check the event condition and re-generate the event if the event condition exists.
If the event condition already exists at the time that the re-arm is initiated, then it is possible that the event will be
regenerated immediately following the conclusion of the re-arm. The delay from the re-arming of a sensor to the
regeneration of the event is device implementation dependent. An initial update in progressreading/state
unavailable (formerly “initial update in progress”) bit is provided with the Get Sensor Reading and Get Sensor
Event Status commands to help software avoid getting incorrect event status due to a re-arm.
14.6.1 ‘Global’ Re-arm
A device that receives a Set Event Receiver command shall ‘re-arm’ event generation for all its internal sensors.
166
Intelligent Platform Management Interface Specification
167
Intelligent Platform Management Interface Specification
15. Platform Event Filtering (PEF)
Platform Event Filtering (PEF) provides a regular mechanism for configuring the BMC to take selected actions on
event messages that it receives or has internally generated. These actions include operations such as system poweroff, system reset, as well as triggering the generation of an Alert.
The BMC maintains an event filter table that is used to select which events trigger a page (or other action) and
which actions to perform. Each time the BMC receives an event message (either externally or internally generated)
it compares the event data against the entries in the event filter table. The BMC scans all entries in the table and
collects a set of actions to be performed as determined by the entries that were matched.
Event filtering is independent of Event Logging. That is, Event Logging and Event Filtering (and associated
actions) are enabled/disabled independent of one another.
15.1 Alert Policies
When an Alert is triggered via PEF the alerting process is directed by an Alert Policy. An alert policy is a
collection of one or more alert destinations. An alert policy can support a mix of different alert destination types
and channels. For example, one policy could include LAN, dial page, and TAP alerts to different locations. The
destinations in a policy are processed in sequence. Whether a given destination will be used or not can be
configured to be dependent on whether the alert to the previous destination was successful or not.
Alert Policy data is stored in an Alert Policy Table that is part of the PEF configuration parameters. An
implementation can support multiple policies. A policy number identifies different policies in the table. The alert
policy number is used in the Event Filter Entry to select what alert policy is used when a match occurs. This
mechanism allows different alert policies to be associated with different classes of events. For example, one
policy to be used for ‘high priority’ events and a different policy for ‘low priority’ events.
Some alerts, such as alphanumeric pages, can be associated with Alert Strings. The combination of Event Filter
Entry and alert destination are used to select a given Alert String from a set of strings kept in the PEF
configuration parameters. This enables different strings to be sent based on what event filter was matched and
where the alert is being sent.
15.2 Deferred Alerts
When an alert policy is initiated, it’s possible that the communication path to the destination could already be
busy processing an earlier alert. To handle this situation, the implementation internally queues up information that
tracks alert policies and destinations to be processed. Alerts that have been postponed are referred to as Deferred
Alerts.
A BMC that supports alerting is required to support deferred alert policies for at least eight events.
15.3 PEF Postpone Timer
Event logging takes precedence over PEF actions. That is, BMC logging of the event is completed prior to
initiating any PEF actions. PEF can occur immediately after the event is logged, or it may be postponed by the
PEF Postpone Timer. The PEF Postpone Timer is a separate timer that allows system software time to process
events instead of PEF. PEF will occur if system software does not handle the event before the PEF Postpone
Timer expires.
168
Intelligent Platform Management Interface Specification
15.4 PEF Startup Delay
Platform Event Filtering is active whenever the BMC is in a state to receive events, either internally or externally
generated. This includes events received over the system interface. Platform Event Filtering is not available when
the BMC is in manufacturing test, modal SDR update, or firmware update modes.
PEF triggered actions may be postponed during certain intervals of BMC operation. The PEF Startup Delay
causes Platform Event Filtering triggered power-down, reset, and power-cycle actions to be postponed when the
system is either powered up or is reset locally via a pushbutton or other local ‘front-panel’ user interface. OEM
actions may or may not be postponed, at the choice of the OEM implementation. Alerts are not postponed by the
PEF Startup Delay. There is a separate, optional, PEF configuration parameter that can control whether Alerts are
delayed on system startup. An implementation may allow the time delay for the PEF Startup Delay to be
configured via the Set PEF Configuration ParametersSet PEF Configuration Parameters command.
It is recommended that the act of entering BIOS setup automatically disables Platform Event Filtering, and that
exiting BIOS setup automatically restores the prior PEF enabled/disabled state (provided that the user does not
explicitly change the PEF configuration while in setup).
The combination of the delay and BIOS setup gives the user the opportunity to enter setup and disable PEF.
These provisions are to allow recovery in case an incorrectly configured filter/action prevents the system from
running by powering it off, power cycling, or resetting it whenever the BMC initializes. Disabling PEF must
immediately cancel any pending PEF actions and deferred alerts.
15.4.1 Last Processed Event Tracking
A non-volatile ‘Last Software Processed Event’ storage location holds the Record ID for the last SEL Record
that system software has processed. System software writes to that location to identify which records it has
processed. A corresponding ‘Last BMC Processed Event’ value holds the Record ID for the last event in the
SEL that the BMC processed. These values can be set and retrieved by software using the Set Last Processed
Event ID and Get Last Processed Event ID commands, respectively.
If PEF is disabled, the Last BMC-processed Event holds the Record ID for the last event that was received.
Clearing the SEL automatically clears the Last Software Processed Event and Last BMC Processed Event
values.
If PEF is enabled and the BMC loses power or is reset before the PEF Postpone Timer expires, the BMC will
automatically perform PEF against any existing, unprocessed events in the SEL once the BMC has restarted
and reinitialized.
Once enabled, the PEF Postpone timer starts running as soon as an unprocessed event is detected in the SEL. If
the SEL already contains unprocessed events, the timer will start immediately.
The timer does not automatically reset on events received while the timer is running, but is reset by system
software after it sets the Last Software Processed Event value.
15.5 Event Processing When The SEL Is Full
If the SEL is full, new events will still be put into the Event Message Buffer (if the optional Event Message
Buffer implemented). The Event Message Buffer for IPMI v1.5 is not overwritten if new events come in.
Therefore, if the Event Message Buffer is full, further events will not go into the event message buffer until its
cleared.
If PEF is implemented, events will also be accepted into a ‘hidden’ internal queue or buffer so they can be
processed by PEF. That buffer is only required to hold a single event. Thus, if that internal buffer gets full,
event messages will be rejected until a new space opens up.
169
Intelligent Platform Management Interface Specification
If neither an Event Message Buffer nor PEF are implemented, events will be rejected by the BMC once the SEL
gets full.
An implementation is allowed to provide a proprietary ‘SEL Aging’ option that automatically clears out SEL
entries if they’re more than a certain age old. If this is done, the algorithm must set the SEL ‘most recent erase
timestamp’ to reflect the time entries were deleted. It must also be possible to configure the system to operate
with the aging algorithm turned off.
15.6 PEF Actions
BMC will scan entire list, collecting a set of actions. Actions will then be executed in priority order. An alert
action can occur in combination with any other action (in priority order). The power down, power cycle, and
reset actions are mutually exclusive. If a combination of power down, power cycle, and/or reset actions results,
only the highest priority action will be taken.
Table 15-1, PEF Action Priorities
Action
power down
power cycle
Priority
1
2
reset
3
Diagnostic
Interrupt
Send Alert
4
OEM
5
OEM
Additional Information
(optional)
(optional) Will not be executed if a power down
action was also selected.
(mandatory) Will not be executed if a power down or
power cycle action was also selected.
(optional) The diagnostic interrupt will not occur if a
higher priority action is also selected to occur.
(mandatory if alerting is supported) Send alerts in
order based on the selected Alert Policy.
Alert actions will be deferred until after the power
down has completed.
There is an additional prioritization within alerts
being sent: based on the Alert Policy Table entries
for the alert. This is described further in Section
15.11, Alert Policy TableAlert Policy Table.
(optional) Priority determined by OEM.
15.7 Event Filter Table
The Event Filter Table consists of a set of rows or ‘entries’ that define each filter. The following table specifies
the fields that comprise a row in the Event Filter Table. These entries include a series of masks that the BMC
applies to the event data. The fields are designed such that a combination of absolute and ‘wildcarded’
comparisons can be used for matching fields in the event record. Thus, either a single event or multiple events can
match up with a single Event Filter Table entry.
A PEF implementation is recommended to provide at least 16 entries in the event filter table. A subset of these
entries should be pre-configured for common system failure events, such as over-temperature, power system
failure, fan failure events, etc. Remaining entries can be made available for ‘OEM’ or System Management
Software configured events. Note that individual entries can be tagged as being reserved for system use - so this
ratio of pre-configured entries to run-time configurable entries can be reallocated if necessary.
A match occurs when there are event filter table matches (exact or wild-carded) for all compared fields in the
event message.
There are two things that can kick off PEF: the arrival of a new event or BMC startup with pending events.
170
Intelligent Platform Management Interface Specification
Table 15-2, Event Filter Table Entry
Byte
1
Field
Filter Configuration
2
Event Filter Action
3
Alert Policy Number
4
Event Severity
5
Generator ID Byte 1
Description
[7] 1b =
0b =
[6:5] - 11b =
10b =
enable filter
disable filter
reserved
manufacturer pre-configured filter. The filter entry has been
configured by the system integrator and should not be
altered by software. Software is allowed to enable or
disable the filter, however.
01b = reserved
00b = software configurable filter. The filter entry is available for
configuration by system management software.
[4:0] - reserved
All actions are optional for an implementation, with the exception of Alert
which is mandatory if alerting is supported for one or more channels.
The BMC will return 0b for unsupported actions. Software can test for
which actions are supported by writing 1’s to the specified fields and
reading back the result. (Note that reserved bits must be written with
0’s)
[7:6] - reserved
[6] 1b = group control operation (see [ICMB])
0b = no group control operation (see [ICMB])
[5] 1b = Diagnostic Interrupt (NMI)
0b = no Diagnostic Interrupt
[4] 1b = OEM action
0b = no OEM
[3] 1b = power cycle
0b = no power cycle
[2] 1b = reset
0b = no reset
[1] 1b = power off
0b = no power off
[0] 1b = Alert
0b = no Alert
Used to select an alerting policy set from the Alert Policy Table. The Alert
Policy Table holds different policies that configure the order in which
different alert destinations and alerting media are tried.
[7:4] - reserved
[6:4] - group control selector (1-based). Selects entry from group control
table. (see [ICMB)
[3:0] - policy number. Value is ‘don’t care’ if Alert is not selected in the
Event Filter Action.
This field can be used to fill in the Event Severity field in a PET alert. The
severity values are based on the ‘DMI’ severity values used for the
generic sensor event/reading type code. In the case that more than one
event filter match occurs for a given Alert Policy Number, the numerically
highest severity value will be used.
00h = unspecified
01h = Monitor
00 0001
02h = Information
00 0010
04h = OK (return to OK condition) 00 0100
08h = Non-critical condition
00 1000 a.k.a. ‘warning’
10h = Critical condition
01 0000
20h = Non-recoverable condition
10 0000
Slave Address or Software ID from Event Message.
FFh = match any
171
Intelligent Platform Management Interface Specification
6
Generator ID Byte 2
7
8
9
Sensor Type
Sensor #
Event Trigger (Event/Reading
Type)
Event Data 1 Event Offset Mask
10,
11
12
Event Data 1 AND Mask
13
Event Data 1 Compare 1
14
Event Data 1 Compare 2
15
16
17
18
19
20
Event Data 2 AND Mask
Event Data 2 Compare 1
Event Data 2 Compare 2
Event Data 3 AND Mask
Event Data 3 Compare 1
Event Data 3 Compare 2
172
Channel Number / LUN to match. FFh = match any see section 26, SEL
Record FormatsSEL Record Formats.
Type of sensor. FFh = match any
FFh = match any
FFh = match any
This bit field is used to match different values of the least significant
nibble of the Event Data 1 field. This enables a filter to provide a match
on multiple event offset values.
Bit positions 15 through 0 correspond to the offset values Fh - 0h,
respectively. A 1 in a given bit position will cause a match if the value in
bits 3:0 of the Event Data 1 hold the corresponding value for the bit
position. Multiple mask bits can be set to 1, enabling a match to multiple
values. A match must be made with this field in order to have a match for
the filter.
data 1
7:0 - mask bit positions 7 to 0, respectively.
data 2
15:8 - mask bit positions 15 to 8, respectively.
This value is applied to the entire Event Data 1 byte. The field is Used to
indicate ‘wildcarded’ or ‘compared’ bits. This field must be used in
conjunction with Compare 2. To match any Event Data field value, just set
the corresponding AND Mask, Compare 1, and Compare 2 fields to 00h.
(See Section 15.8, Event Data 1 Event Offset MaskEvent Data 1 Event
Offset Mask for more information). Note that the Event Data 1 AND mask,
Compare 1 mask, and Compare 2 masks will typically be set to wild-card
the least significant of Event Data 1 in order to allow the Event Data 1
Event Mask field to determine matches to the event offset.
Bits 7:0 all have the following definition:
0 = Wildcard bit. (drops this bit position in the Event Data byte out of
the comparison process) Corresponding bit position must be a 1 in
Compare 1, and a 0 in Compare 2.
(Note - setting a 0 in this bit, a 1 and Compare 1 and a 1 in
Compare 2 guarantees that you’ll never have a match.)
1 = use bit for further ‘exact’ or ‘non-exact’ comparisons based on
Compare 1 and Compare 2 values.
Used to indicate whether each bit position’s comparison is an exact
comparison or not. (See Section 15.8, Event Data 1 Event Offset
MaskEvent Data 1 Event Offset Mask for more information). Here, ‘test
value’ refers to the Event Data value after the AND Mask has been
applied.
Bits 7:0 all have the following definition:
1 = match bit in test value exactly to correspond bit position in
Compare 2
0 = contributes to match if corresponding bit in test value matches
corresponding bit in Compare 2.
(See Section 15.8, Event Data 1 Event Offset MaskEvent Data 1 Event
Offset Mask for more information). Here, ‘test value’ refers to the Event
Data value after the AND Mask has been applied.
Bits 7:0 all have the following definition:
1 = match a ‘1’ in corresponding bit position in test value.
0 = match a ‘0’ in corresponding bit position in test value.
Intelligent Platform Management Interface Specification
15.8 Event Data 1 Event Offset Mask
The Event Data 1 Event Offset Mask field in the Event Filter is used to match multiple bits in the Event Offset
field of the Event Data 1 byte of an event. The least significant nibble of event data 1 typically holds an event
offset value. This offset selects among different possible events for a sensor. For example, a ‘button’ sensor
supports a set of sensor-specific event offsets: 0 for Power Button pressed, 1 for Sleep Button pressed, and 2 for
Reset Button pressed. When an event is generated, it could have a 0, 1, or 2 in the event offset field depending on
what button press occurred.
The Event Offset Mask makes it simple to have a filter match a subset of the possible event offset values. Each bit
in the mask corresponds to a different offset values starting with bit 0 in the mask corresponding to offset 0. For
example, if it is desired to have a filter match offsets 0 and 2, but not 1, the mask would be configured to
000_0000_0000_0101b.
15.9 Using the Mask and Compare Fields
The AND Mask and the Compare 1 and Compare 2 fields are used in combination to allow wildcarding, ‘one or
more bit(s)’, and exact comparisons to be made between bits in the corresponding event data byte. One way to
understanding the bits is to look at the way they’re used in combination. First the AND Mask is applied to the test
value. The result, referred to below as the test value, is then bit-wise matched based on the values in the Compare
1 and Compare 2 fields, as summarized in the following table.
Table 15-3, Comparison-type Selection according to Compare Field bits
Compare
1
Compare
2
comparison
1
1
exact compare
to 1
This bit in the test value must be = 1 for a match.
If it’s 0, there is no match. All ‘exact’ comparison bits must match
the corresponding bits in the test value for a match.
1
0
exact compare
to 0
0
1
‘non exact’
compare to 1
0
0
‘non exact’
compare to 0
This bit in the test value must be = 0 for a match.
If it’s 1, there is no match. All ‘exact’ comparison bits must match
the corresponding bits in the test value for a match.
If this bit in the test value is 1, it contributes to a match. There will
be a match if any one of the bit positions that has a ‘0’ in
Compare 1 field has a bit in the test value that matches the
polarity given in the corresponding bit position in the Compare 2
field. - unless there are exact comparisons that don’t match.
If this bit in the test value is 0, it contributes to a match. There will
be a match if any one of the bit positions that has a ‘0’ in
Compare 1 field has a bit in the test value that matches the
polarity given in the corresponding bit position in the Compare 2
field. - unless there are exact comparisons that don’t match.
description
15.10 Mask and Compare Field Examples
The following examples show how the fields are used. See Appendix B - Example PEF Mask Compare
AlgorithmAppendix B - Example PEF Mask Compare Algorithm for example matching algorithm.
Example 1:
AND Mask:
0000 0110
Compare 1:
Compare 2:
1111 1111
0000 0110
Match (bit 2 = 1) AND (bit 1 = 1), and ignore all other bits.
Force all bits except bits 2 and 1 to 0. (Forcing to 0 and comparing exactly to 0 makes
the other bits ‘don’t care’)
Compare all bits exactly.
Compare for bits 2 and 1 both = 1, and remaining bits = 0.
173
Intelligent Platform Management Interface Specification
Example 2:
AND Mask:
Compare 1:
0000 0110
1111 1001
Compare 2:
0000 0110
Match (bit 2 = 1) OR (bit 1=1), and ignore all other bits.
Force all bits except bits 2 and 1 to 0.
Compare for at least one of bit 2 or bit 1being polarity specified in the corresponding bit
position in Compare 2. Compare remaining bits exactly.
Compare for bit 2 or bit 1 = 1, and remaining bits = 0 exactly.
Example 3:
AND Mask:
Compare 1:
Compare 2:
0000 0110
1111 1111
0000 0100
Match (bit 2 = 1) AND (bit 1 = 0)
Force all bits except bits 2 and 1 to 0
Compare all bits (after AND) exactly
Compare for bit 2 = 1 and bit 1 = 0, and remaining bits = 0 exactly.
Example 4:
AND Mask:
Compare 1:
Compare 2:
0000 0110
1111 1001
0000 0100
Match bit 2 = 1 OR bit 1 = 0
Force all bits except bits 2 and 1 to 0.
Compare all bits except bits 2 and 1 exactly.
Compare for bit 2 = 1 OR bit 1 = 0, and remaining bits = 0 exactly.
Example 5:
AND Mask:
Compare 1:
Compare 2:
1111 1111
1111 0000
1010 1111
Match most significant nibble = 1010 exactly, and any bit in LSN = 1.
Look at all bits
Compare all bits in MSN exactly.
Compare MSN = 1010 exactly, and for a 1 in one or more positions in LSN
Example 6:
AND Mask:
Compare 1:
Compare 2:
1111 1111
1111 0000
1010 0000
match MSN = 1010 exactly, and any bit in LSN = 0.
Look at all bits
Compare all bits in MSN exactly.
Compare MSN = 1010 exactly, and for a 0 in one or more positions in LSN
15.11 Alert Policy Table
Platform Event Filtering supports alerting as one of the selectable actions that can occur when an event matches
an event filter table entry. The alerting media and the different alert destinations that are tried are determined by
the settings in the Alert Policy Table.
The Alert Policy Table definition enables implementations to offer multiple policy sets. For example, one policy
could be configured to generate LAN Alerts and Pages and be associated with critical and non-recoverable
events, while another may generate only LAN Alerts and be associated with non-critical events. It would even be
possible to configure a ‘Chassis Security’ event to notify one party, while an ‘Over-temperature’ event could be
delivered to a different destination.
The table entry also contains a parameter that can be used to select whether the alert is delivered to all enabled
destinations, or that different destinations are tried until the alert is successfully delivered. Altering the policy
table entries can change the whether certain destinations are processed or not.
A policy number is used to group multiple table entries into a policy set. The collection of entries determines
which different media and alert types can be tried when an alert action occurs. When a Platform Event Filter table
entry is configured to perform an alert action on an event, an alert policy number is also configured for the action.
The BMC takes this policy number and uses it to look up the entries for the corresponding policy set in the Alert
Policy Table.
The policy number also determines the priority of alert policies. Only one starting Alert policy will be used for a
given event. If an event matches more than one alert policy, the policy with the lowest number will be used.
Implementations that support alerting are required to provide at least one table entry. It is recommended that an
implementation supports at least one policy table entry for each different channel over which an alert can be
delivered.
174
Intelligent Platform Management Interface Specification
Table 15-4, Alert Policy Table Entry
Byte
1
Field
Description
DATA BYTES
Policy Number /
Policy
This value identifies the entries belonging to a particular policy set. When an Alert Action is
taken, the BMC will scan the Alert Policy Table and will attempt to generate alerts based on the
entries that form the policy set.
2
Channel / Destination
3
Alert String Key
[7:4] - policy number. 1 based. 0000b = reserved.
[3] - 0b = this entry is disabled. Skip to next entry in policy, if any.
1b = this entry is enabled.
[2:0] - policy
0h = always send alert to this destination.
1h = if alert to previous destination was successful, do not send alert to this destination.
Proceed to next entry in this policy set.
2h = if alert to previous destination was successful, do not send alert to this destination.
Do not process any more entries in this policy set.
3h = if alert to previous destination was successful, do not send alert to this destination.
Proceed to next entry in this policy set that is to a different channel.
4h = if alert to previous destination was successful, do not send alert to this destination.
Proceed to next entry in this policy set that is to a different destination type.
Channel that the alert is to be sent over. Channel determines which set of destination
addresses or phone numbers is used. Destination addresses and/or phone numbers are set via
the LAN and/or serial/modem configuration parameter commands. The Alert Type (e.g. PET,
TAP, Dial Page, etc.) is specified in the configuration parameters associated with the specified
destination.
[7:4] = Channel Number.
[3:0] = Destination selector.
This field holds information that is used to look up the Alert String to send for this Alert Policy
entry.
00h = no alert string.
[7] - Event-specific Alert String
1b = Alert String look-up is event specific. The following Alert String Set / Selector subfield is interpreted as an Alert String Set Number that is used in conjunction with
the Event Filter Number to lookup the Alert String from the PEF Configuration
Parameters.
0b = Alert String is not event specific. The following Alert String Set / Selector sub-field
is interpreted as an Alert String Selector that provides a direct pointer to the
desired Alert String from the PEF Configuration Parameters.
[6:0] - Alert String Set / Selector. This value identifies one or more Alert Strings in the Alert
String table. When used as an Alert String Set Number, it is used in conjunction with the
Event Filter Number to uniquely identify an Alert String. When used as an Alert String
Selector it directly selects an Alert String from the PEF Configuration Parameters.
The Alert String Key and lookup mechanism allows the Alert String to be ‘Event Specific’ meaning the string selection is determined by both the Event Policy Entry and Event Filter, or,
the string can be selected by the Event Policy alone. An Alert String can be pointed to by
multiple policy entries.
The Alert Policy Entry identifies a particular channel and destination for an alert. This in turn,
identifies the alert type. Thus, the binding of an Alert Policy Entry and an Alert String effectively
provides a mechanism for allowing different Alert Strings to be selected based on the alert
destination, or the type of alert destination. For example, a single Alert String could be shared
among all Alert Policy Entries for ‘Dial Page’ destinations, while event-specific Alert Strings
could be used for alerts to LAN destinations.
15.12 Alert Testing
BMC includes an ability to test alert configurations. This is accomplished through the Alert Immediate command.
This command allows a particular alert destination to be selected and an alert sent to it. Typically, software will
use the volatile settings in the configuration parameters for the channel to hold alert destination information until
the setting is verified by the user, at which time it can be moved into one of the non-volatile positions.
175
Intelligent Platform Management Interface Specification
15.13 Alert Processing
The BMC starts from the beginning of the Alert Policy Table, scanning for entries that match the event. The BMC
will scan all entries in the Event Filter table and initiate the highest priority Alert Policy for which there was a
match. If multiple filters match and have the same alert policy priority, the first matched event filter will be used.
(Since an Alert String can be associated with an event filter, it may be important to order the event filter table
such that the more ‘granular’ filters occupy the earlier entries, and the more generic filters occupy the later
entries.)
BMC implementations are allowed to have multiple alerts simultaneously in progress on the same channel or
across multiple channels.
If an alert was successfully delivered, the BMC will either stop processing the policy, or will continue to process
alert policy table entries - based on whether the destination was configured to stop alert processing on success or
not. Each entry in the alert policy is processed in order until all entries have been processed (meaning the alert
was either sent, failed, or the alert was skipped).
15.13.1 Alert Processing after Power Loss
It is possible that more events will come in while an alert is being processed. Each time a SEL Record is
completely processed for alerts, the BMC saves a copy of the Last BMC Processed Record ID to NV storage.
(“Completely processed” means that there are no incompletely processed alert policies for the event). That way,
if AC power is lost the BMC can tell whether there are events that may not have been processed for alerting by
comparing the Last BMC Processed Record ID with the Record ID of the last SEL Entry.
It’s possible that alerts were sent to some destinations, but not all, when power was lost. When power comes
back up, the BMC will start processing the entire record and as a result some alerts may get re-sent.
15.13.2 Processing non-Alert Actions after Power Loss
After the BMC powers up, and before restoring the system power state, it uses PEF to check pending events for
matches that would have yielded a ‘power off’ action. If there is a match, it sets the previous power Restore
State to Off and leaves the system powered off. The BMC skips processing reset or power cycle actions that
were pending when AC was lost.
In order to know how many events to process, the BMC should copy the Record ID or timestamp of the last
SEL Entry and only process records up to that point as events that were pending when AC was lost. That way,
if a new event comes in, it will be handled as a new event rather than as an event that was pending when AC
was lost.
15.13.3 Alert Processing when IPMI Messaging is in Progress
All alerts are deferred if IPMI Messaging is in progress on the channel. The remote console software can direct
the BMC to skip processing deferred events by setting the Last BMC Processed Record ID value for all filters
to the Record ID of the last SEL Record.
15.13.4 Sending Multiple Alerts On One Call
To avoid unnecessary phone calls, it is desirable to have multiple alerts be delivered to a given PPP Account
before hanging up, rather than hanging-up after each event. The serial/modem configuration parameters and the
Alert Policy entries can be configured to support this. To have this accomplished, the configuration must fit the
following rules.
176
Intelligent Platform Management Interface Specification
•
The Connection Hold Time parameter for the PPP Accounts should be set to a value that covers the time
that you’d like the call to be maintained waiting for the next event to occur. Since a new alert will restart
the connection time time-out, the value should be set to the time required to maintain the connection
between alerts.
•
All event policies should be configured to have alerts to each channel delivered in priority order starting
with the highest priority destination first.
15.13.5 Serial/Modem Alert Processing
Alerts on the same given serial/modem channel processed according to priority. Alerts to PPP Accounts are
processed with higher priority than Dial Page or TAP Page alerts. PPP Accounts are prioritized according to the
account selection, with the lowest account selector corresponding to the highest priority with the lowest account
selector as summarized in the following table:
Table 15-5, Serial/Modem Alert Destination Priorities
Destination Type
PPP Account #1
PPP Account #2
PPP Account #N
Dial Page and TAP Page
Priority
(0 = highest)
1
2
N
N+1
Comments
All destinations behind a given PPP Account are at equal
priority. Destinations behind an account are handled in
the order that they occur in the Alert Policy Table entry
associated with the account.
All Dial Page and TAP Pages are at equal priority. They
are handled in the order that they occur in the Alert Policy
Table entry.
The following specifies how serial/modem alerts from the same channel are handled:
•
The BMC will not prematurely terminate a PPP Alert, Dial Page, or TAP Page in progress to a given
destination, with the exception that an implementation is allowed to in order to handle a power off, power
cycle, or system reset transition. In this case, the alert should be resumed once the transition has completed.
•
The BMC checks the event filters for matches to the event. If there is more than one alert policy selected
by the match, the BMC will only execute the highest priority (lowest numbered) alert policy.
•
The BMC must keep track of how far it has processed the alert policy associated with the event. It does this
in case it needs to prematurely terminate a call because of a power off, power cycle, or system reset
operation.
•
The BMC processes each alert policy through to completion. Completion of processing means that all
alerts were either sent or deferred.
•
If there is no presently active connection, a connection will be made for the first alert destination in the
alert policy sent and the alert sent.
•
The PPP Account Connection Hold Time parameter determines how long a PPP call will be held open
waiting for another alert. The PPP connection will only close when the time expires, or if an alert to a
higher priority PPP Account occurs.
•
The call to a PPP Account will be dropped without waiting for the Connection Hold Time to expire if alerts
fail to all destinations for a given policy.
•
The Connection Hold Time time-out must be restarted whenever a new alert is sent to the account.
•
If a PPP account connection is already active and the alert is to a destination behind that PPP account, the
BMC will wait for any alerts in progress to that account to complete and then send the present alert.
177
Intelligent Platform Management Interface Specification
•
If the alert is to a destination that is behind a higher priority PPP account, the present connection will be
terminated as soon as the present alert to that account has completed, regardless of the connection hold
time. The BMC will then dial the higher priority account and sent the alert.
•
If a PPP account connection is already open it will be used unless the alert is to a lower-priority destination
type, in which case the alert will be deferred until connection hold time for the present connection expires.
•
If a Dial Page or TAP Page is already active, the BMC will wait for the alert in progress to complete and
then send the present alert, regardless of whether the present alert is of higher priority. (Completion of an
alert in progress for an unacknowledged alert means that the alert has been sent. For an acknowledged
alert, completion means the alert has been sent and the acknowledge received or all retries and timeouts
have concluded and the alert is considered to have failed.)
•
The Page Blackout Interval is essentially a ‘throttle’ that prevents pages from being sent ‘back-to-back’.
See 13.10, Page Blackout IntervalPage Blackout Interval.
15.14 PEF and Alert Handling Example
The following figure presents a snapshot of event and alert processing using PEF. It also helps illustrate the
relationship between entries in the serial/modem configuration parameters.
1.
The present event being processed is identified as event with Record ID = “N+2”
2.
The BMC scans all filter table entries for matches to the present event. When done, it finds three entries with
Alert actions that have been matched: event filters X, P, and T.
3.
The BMC handles matched filters in priority order based on the action associated with the filter. The Power
Off action is higher priority than Alert actions, so filter X is acted on immediately and the power OFF action
performed.
4.
There are two matched filters left, both with Alert actions. Filter P is acted on because it has the lower policy
number. The BMC ‘queues’ information for the alert policy and the event filter so it can be processed later.
5.
The event triggers Alert Policy #3. The BMC sends the alert to LAN destination #1, and then sends the alert
to serial/modem destination #1. The figure shows that serial/modem destination #1 is a PPP Alert
destination, therefore the BMC looks up the corresponding PPP Account information from the serial/modem
configuration parameters. The PPP account is account #1. This is the highest priority account. If the account
is not already active, the BMC will terminate any lower priority call in progress and then call the account #1
and send the alert.
6.
The completion of the alert policy for event N+2 may cause the Last BMC Processed Record ID to be set to
N+2 - but it also may not. Whether the Last BMC Processed Record ID is advanced is based on whether all
deferred alerts have been processed. Due to prioritization, it’s possible that in some cases the alert policy
triggered for event N+2 could complete while an alert policy associated with an earlier event ‘N’ could still
have destinations to be processed.
178
Intelligent Platform Management Interface Specification
Figure 15-1, Alert Processing Example
Serial/Modem Configuration Parameters
Event Filter X,
Action = Power OFF
Policy = 0
Filter P,
Action = Alert
Policy = 3
Filter T
Action = Alert
Policy = 4
Alert Policy
info for filter
match 'queued'
serial/modem
destinations table
to LAN Configuration
Parameters
Alert String Key
Event Filter Table
Policy Number
Present
Event = N+2
Alert Policy Table
Destination Type = PPP Alert
1
q
Destination1 = LAN 1
3
n
Destination1 = serial 1
3
n
Destination2 = serial 2
3
m Destination3 = serial 4
3
m Destination4 = serial 3
Destination Type = PPP Alert
2
Server Account Auth. Type
dial
strings
Server Account Password
Server User Name
2 Server User Domain
1 dial string X
Server IP Address
2 dial string Y
Dial String Selector=3
3 dial string Z
Connection Hold Time
Alert ACK timeout
4 dial string Q
PPP Account Selector = 2
5 dial string Y
IP Address Selector=2
6 dial string Y
Destination Type = PPP Alert
3
n
TAP Service / PPP Account Selector = 2
IP Address Selector=1
3
4
Alert ACK timeout
PPP accounts table
Destination1 = serial 2
IP Addresses
Alert ACK timeout
PPP Account Selector = 2
IP Address Selector=3
1 IP Address A
2 IP Address B
Destination Type = Dial Page
Alert String Set
Event Filter #
4 Alert ACK timeout
3 IP Address C
4 IP Address D
Dial String Selector=4
Destination Type = TAP Page
Alert Strings
5
P
n
P
m string Y
string X
T
q
TAP Services
Alert ACK timeout
Dial String Selector=5
TAP Service Selector = 1
1
string Z
15.15 Event Filter, Policy, Destination, and String Relationships
The following figure illustrates the relationship between the different structures and configuration parameters
related to platform event filtering and alerting. Note that the number of table entries and support for different alert
types is implementation dependent.
The figure shows the lookup process that occurs when an event matches an Event Filter Table entry. In this
example, the entry triggers an Alert that activates alert policy number 3. The policy table is scanned for the first
entry with a matching policy number. The first matching entry is for a LAN channel. First, a LAN PET trap to
xxx.212.123.67, then, if that fails a LAN PET trap to xxx.100.200.1 will be tried. If that fails, the next matching
policy entry will be tried.
The next policy is for a serial/modem channel. The first alert on this channel will be a Dial Out PET Alert (a.k.a.
PPP Alert). The second attempt will be a TAP Page. Note that a short ASCII Alert String “System FRED
Intrusion” is selected by the Alert Policy Entry for the TAP page. The third attempt will be a dial page. Since a
dial-page is restricted to submitting ‘key pad’ digits via the modem command set, we see that the final attempt is
used to deliver a numeric page with the number of the system in trouble, and a user-defined ‘911’ to indicate that
there’s a serious condition.
In order to simplify the figure, certain parameters associated with the Alert destinations have been left out. For
example, there are destination phone numbers and modem init strings associated with the paging destinations, and
MAC address and Gateway Address values associated with the LAN Alert destinations.
179
Intelligent Platform Management Interface Specification
Figure 15-2, Event Filter, Alert Policy, and Alert Destination, & String Relationships
Actions = Alert
Policy Set = 3
...
N
Actions = xx
Assume Filter 2 matches
"Chassis Intrusion" events
Policy Set 3
xxx.100.200.1
Alert Type = PET Alert
xxx.100.210.12
3
Alert Type = PET Alert
xxx.212.123.67
X
Alert Type = YY
...
destination = xx
alert string key = yy
1
Ch# = N
(serial/modem)
destination = pp
3
Ch# = M
(LAN)
destination = 3
3
Ch# = M
(LAN)
destination = 1
2
Ch# = N
(serial/modem)
destination = aa
alert string key = bb
3
Ch # = N
(serial/modem)
destination = 2
3
Ch # = N
(serial/modem)
destination = 1
3
Ch # = N
(serial/modem)
destination = 3
2
Ch# = N
(serial/modem)
destination = aa
2
Ch # = N
(serial/modem)
destination = cc
X
Ch# = M (LAN)
xxx.xxx.xxx.xxx
alert string key = qq
alert string key = 0
alert string key = 0
alert string key = 0
Serial/Modem Configuration Parameters
Destination Alert
Type
Serial/Modem
Destination
Addresses
1
Alert Type = TAP
9,1,800,555,1212
2
Alert Type = PPP Alert
6968080
3
Alert Type = Dial Page
18002255288
...
alert string key = 85h
X
Alert Type = XX
xxxxxxxxxx
alert string key = 2
alert string key = bb
alert string key = dd
...
destination 1 = rr
alert string key 1 = ss
Alert String 2 is an example of a nonevent-specific Alert String
Alert String 3 is an example of an event-specific Alert String.
Note that bit 7 is set in the Alert String Key field from the Alert Policy Table.
The value '2' in the Event Filter field matches up with Event Filter Number 2, which
in this example is a filter that matches up to Chassis Intrusion events.
Alert String Set
Actions = reset
Alert Type = PET Alert
2
LAN Alert
Destination
Addresses
Event Filter #
2
Ch# = N
(serial/modem)
1
1
Destination
Selector
1
Channel
Destination Alert
Type
String Selector
Event Filter
Number
Event
Match
Entry Data
Policy
Number
Alert Policy Table
Event Filter Table
Destination
Selector
LAN Configuration Parameters
1
N
M
2
0
0
",(503) 555-1212 911"
3
2
5
"System FRED Intrusion"
Alert Strings (from PEF
configuration parameters)
"System FRED Overtemperature"
...
X
X
X
"xxxxxxxx"
15.16 Populating a PET
The following table outlines the way PET fields are populated for an IPMI alert. See [PET] for more information.
The Community String portion of the PET can be obtained from the configuration parameters associated with
the channel from which the trap is issued. If the Community String parameter is not supported, the string ‘public’
should be sent.
The following table lists the population of the PET Specific Trap fields for an IPMI alert:
Table 15-6, PET Specific Trap Fields
PET Field
Event Sensor Type
Event Type
Event Offset
180
IPMI Source
Sensor Type code from event message
Event/Reading Type code from event message
[7] = Event Dir bit from event message
[3:0] = Event Offset for Event Data 1 byte of event message
Intelligent Platform Management Interface Specification
The following table lists the IPMI source and formatting for fields that go into the ‘variable bindings’ fields of a
PET.
Table 15-7 - PET Variable Bindings Field
PET Field
GUID
Sequence # /
Cookie
Local Timestamp
size/
type
16
bytes
word
dword
UTC Offset
word
Trap Source
Type
byte
Event Source
Type
byte
Event Severity
byte
IPMI Source
Recommended that BMC populate this with the System GUID. If a system GUID
is not available, a device GUID for the BMC may be substituted.
BMC should increment the value in this field for each new PET issued, but leave
the value unchanged for PET retries.
BMC should populate this field with the time value that would be used to log the
event in the SEL. Per PET, this needs to be converted to represent number of
seconds from 0:00 1/1/98.
0000 0000 = unspecified.
Optional. UTC Offset in minutes (two’s complement, signed. -720 to +720,
0xFFFF=unspecified)
Class of the device or software that originated the trap on the network. Use 20h
for PETs that are issued from the BMC.
Use 20h for events that are automatically generated by the BMC (e.g. by PEF)
It is recommended that 21h be used for IPMI-format PETs that are generated by
system software instead of automatically by the BMC.
Severity (based on DMI Event Severity).
If PEF specifies an event severity for the event filter that triggered the Alert, that
severity should be used instead.
0x00 = unspecified
0x01 = Monitor
0x02 = Information
0x04 = OK (return to OK condition)
0x08 = Non-critical condition
0x10 = Critical condition
0x20 = Non-recoverable condition
Sensor Device
byte
Sensor Number
Entity
byte
byte
Entity Instance
byte
Event Data
octet
string
(8)
Language Code
byte
00_0001b
00_0010b
00_0100b
00_1000b a.k.a. ‘warning’
01_0000b
10_0000b
2
In IPMI this holds an ID (I C address or SWID) of the controller or software
entity that generated the event. This comes from the event message. I.e. if the
BMC received the event from controller C2h, this value would be set to C2h, not
to the BMC’s address.
Sensor number from the event message.
Entity ID from IPMI v1.5specification. (Optional). An implementation can elect to
look up the Entity ID associated with the sensor and send that information in the
PET.
00h = unspecified
This field can hold is the Entity Instance value associated with the preceding
Entity field.
00h = unspecified
Additional parametric data byte - formatted as specified by Event Type in
combination with Event Source. Interpreted as individual octet fields.
Event Data 1 - Populate this field with the Event Data 1 byte from the Event
Message
Event Data 2 - Populate this field with the Event Data 2 byte from the Event
Message
Event Data 3 - Populate this field with the Event Data 3 byte from the Event
Message
Event Data 4:8 - These bytes are not used with IPMI messages. They should be
set to 00h. Software should ignore their content.
Per IPMI v1.0 FRU Information Format. FFh = ‘unspecified’. This field can be
used in conjunction with the OEM fields, below, to indicate the language that any
strings are in. Note that language is different than character set. Character sets
are specified as ASCII or UNICODE, per type/length bytes.
181
Intelligent Platform Management Interface Specification
PET Field
Manufacturer ID
size/
type
dword
System ID
word
OEM Custom
Fields
octet
string
(max.
64)
IPMI Source
Manufacturer ID using Private Enterprise IDs per IANA. This should reflect the
ID of the manufacturer of the System from which the alert it being issued.
Specified by manufacturer given by Manufacturer ID field, this number can be
used to identify the particular system/product model or type.
One or more fields given in IPMI v1.0 FRU Information field format:
Type/length code byte followed by N data bytes for each field.
Fields end when type/length byte indicates ‘no more records’ (C1h). A C1h in
octet 47 indicates no OEM Custom Fields.
15.16.1 OEM Custom Fields and Text Alert Strings for IPMI v1.5 PET
An IPMI format PET (PET Event Source = 20h or 21h) provides additional specification on the use of the
Type/Length Byte in the OEM Custom Fields by defining additional special values as follows:
PET OEM Field Type/Length Byte special values
00h Æ reserved
40h Æ reserved
80h Æ typed field (‘PET multirecord’ field)
C0h Æ empty field
C1h Æ end of fields
With this encoding, a Type/Length value of 80h indicates the start of a ‘PET multirecord’ field where the field
format is as follows. This format enables the PET trap to carry an Alert String from PEF, while allowing OEM
data to coexist in the custom fields as well.
Table 15-8, IPMI PET Multirecord Field Format
byte
1
2
Field
Type/Length
Encoding/Length
3
Record Type
4:N
Record Data
Definition
80h (PET multirecord field)
7:6 Record Encoding
00b = binary / unspecified
01b = ASCII
10b = UNICODE
11b = reserved
5:0 Record Data Length in bytes (number of bytes in Record Data field)
7:4 reserved
3:0 Record Type
0h = reserved
1h = Text Alert String
2h = OEM Data per OEM Identified by IANA in first three bytes of
Record Data
3h = OEM Data per OEM Identified by Manufacturer ID field
Data per Record Type
15.17 PEF Performance Target
Excluding PEF Action Delays, a PEF action should nominally occur within two seconds of the corresponding
Event Message being received by the BMC. For events generated internal to the BMC, this should occur within
two seconds of the event being written to the SEL or delivered to the Event Message Buffer. Note that this does
not include delays for event detection and formatting the event record, nor the time to poll and accumulate the
182
Intelligent Platform Management Interface Specification
data that triggers the event. For example, it may take the BMC several seconds to detect a fan failure, that time is
not included in this performance target.
For alerts, this target also represents the time to initiate the processing of an alert policy. The actual time it takes
to complete an alert policy is widely variable, dependent on factors such as whether the alert is deferred, plus
elements such as telephone system and network response delays, thus a performance target is not specified for
alert completion at this time.
183
Intelligent Platform Management Interface Specification
16. Command Specification Information
This section provides specifications for elements that apply to all requests and responses presented later in this
document.
16.1 Specification of Completion Codes
Completion codes are specified in Section 5.2, Completion CodesCompletion Codes. Additional commandspecific completion codes, if any are listed in the ‘completion code’ field description for the command. In some
cases, use of certain command-specific completion codes is mandatory. This will be listed alongside the
description of the completion code in the command table. If no command-specific completion codes are listed, the
description will solely indicate that the field is the ‘completion code’ field. Note that the generic completion code
values can be used with any command, regardless of whether additional command-specific completion codes are
defined. Therefore generic completion codes are not explicitly listed in the command tables. Refer to Section 5.2
for additional requirements and guidelines.
16.2 Handling ‘Reserved’ Bits and Fields
Unless otherwise noted, Reserved bits and fields in commands (request messages) and responses shall be written
as ‘0’. Applications must ignore the state of reserved bits when they are read.
16.3 Logical Unit Numbers (LUNs) for Commands
Unless otherwise specified, commands that are listed as mandatory must be accessed via LUN 00b. An
implementation may elect to make any command available on any LUN or channel as long as it does not conflict
with other requirements in this specification.
16.4 Command Table Notation
The following section includes command tables that list the data that is included in a request or a response for
each command. The completion code for a response is included as the first byte of the response data field for each
command. The Network Function (NetFn) and command byte values for each command are specified in separate
tables.
The following notation is used in the command tables.
Request Data
Identifies portion of the table that lists the fields that are included in the data portion of a
request message for the given command.
Response Data
Identifies portion of the table that lists the fields that are included in the data portion of a
response message for the given command. Note that the completion code is always listed as the
first byte in the response data field.
-
Empty Field. A dash (-) in the byte column indicates that there is no request data for the
command.
4
Single Byte Field. A single value in the byte column of a command table is used to identify a
single byte field. The value represents the offset to the field within the data portion of the
message. In some cases a single byte field with follow a variable length field (see following),
in which case the single byte offset will be represented with an alphabetic variable and number
184
Intelligent Platform Management Interface Specification
representing the single byte field’s location relative to the end of the variable length field. E.g.
N+1.
5:7
Multi-byte Field. Indicates a multi-byte field. The byte column indicates the byte offset(s) for
a given field. For a multi-byte field, the first value indicates the starting offset, the second
value (following the colon) indicates the offset for the last byte in the field. For example, 5:7
indicates a three-byte field spanning byte offsets 5, 6, and 7.
In some cases, multi-byte fields may be variable length, in which case an alphabetic variable
will be used to represent the ending offset, e.g. 5:N. Similarly, a field may following a variable
length field. In this case the starting value will be shown as an offset relative to the notation
used for the previous field, e.g. if the previous field were 5:N, the next field would be shown
starting at N+1.
Lastly, a variable length field may follow a variable length field, in which case a relative
starting offset will be shown with an alphabetic value indicating a relative ending offset, e.g.
N+1:M.
(3)
Optional Fields. When used in the byte column of the command tables, parentheses are used
to indicate optional data byte fields. These can be absent or present at the choice of the party
generating the request or response message. Devices receiving the message are required to
accept any legal combination of optional data byte fields.
Unless otherwise indicated, if an optional byte field is present all prior specified byte fields
must also be present. Similarly, if an optional byte field is absent all following byte fields must
also be absent. For example, suppose a request accepts 4 data bytes. If data byte 3 was shown
in parentheses as ‘(3), it would indicate that byte 3 and following were optional. A legal
request could consist of just bytes [1 and 2], bytes [1, 2, and 3,] or bytes [1, 2, 3 and 4]. It is to
eliminate just byte 3, but include byte 4. I.e. a request with data bytes [1, 2, and 4], would be
illegal.
Multi-byte fields that are shown as optional cannot be split. Either all bytes for the field are
present or absent. I.e. if a four byte multi-byte field is listed as optional, it is illegal to include
the first two bytes, but not the second two bytes.
185
Intelligent Platform Management Interface Specification
17. IPM Device “Global” Commands
This section presents the commands that are common to all Intelligent Platform Management (IPM) devices that
follow this specification’s message/command interface. This includes management controllers that connect to the
system via a compatible message interface, as well as ‘IPMB Devices’.
IPMI Management Controllers shall recognize and respond to these commands via LUN 0. Refer to Appendix G Command Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands. O/M = Optional/Mandatory.
Table 17-1, IPM Device ‘Global’ Commands
Command
Get Device ID
Cold Reset
Warm Reset
Get Self Test Results
Manufacturing Test On
Set ACPI Power State
Get ACPI Power State
Get Device GUID
Section
O/M
17.1
17.2
17.3
17.4
17.5
17.6
17.7
17.8
M
[1]
O
O
M
O
O
[3]
O
O
17.9
M
Broadcast Commands
Broadcast ‘Get Device ID’
[1]
[2]
[3]
186
[2]
This command is not required to return a response in all implementations.
Broadcast is over IPMB channels only. Request is formatted as an entire IPMB
application request message, from the RsSA field through the second checksum,
with the message prefixed with the broadcast slave address, 00h. Response
format is same as the regular ‘Get Device ID’ response.
Mandatory if Set ACPI Power State command is implemented on given
management controller.
Intelligent Platform Management Interface Specification
17.1 Get Device ID Command
This command is used to retrieve the Intelligent Device’s Hardware Revision, Firmware/Software Revision, and
Sensor and Event Interface Command specification revision information. The command also returns information
regarding the additional ‘logical device’ functionality (beyond ‘Application’ and ‘IPM’ device functionality) that
is provided within the intelligent device, if any.
While broad dependence on OEM-specific functionality is discouraged, two fields in the response allow software
to identify controllers for the purpose of recognizing controller specific functionality. These are the Device ID
and the Product ID fields. A controller that just implements standard IPMI commands can set these fields to
‘unspecified’.
Table 17-2, Get Device ID Command
byte
Request Data
Response Data
data field
-
-
1
Completion Code
2
Device ID. 00h = unspecified.
3
Device Revision
[7] 1 = device provides Device SDRs
0 = device does not provide Device SDRs
[6:4] reserved. Return as 0.
[3:0] Device Revision, binary encoded.
Firmware Revision 1
[7] Device available: 0=normal operation, 1= device firmware, SDR
Repository update or self-initialization in progress. [Firmware / SDR
Repository updates can be differentiated by issuing a Get SDR
command and checking the completion code.]
[6:0] Major Firmware Revision, binary encoded.
Firmware Revision 2: Minor Firmware Revision. BCD encoded.
4
5
6
7
8:10
11:12
IPMI Version. Holds IPMI Command Specification Version. BCD encoded.
00h = reserved. Bits 7:4 hold the Least Significant digit of the revision, while
bits 3:0 hold the Most Significant bits. E.g. a value of 51h indicates revision
1.5.
Additional Device Support (formerly called IPM Device Support). Lists the
IPMI ‘logical device’ commands and functions that the controller supports that
are in addition to the mandatory IPM and Application commands.
[7] Chassis Device (device functions as chassis device per ICMB spec.)
[6] Bridge (device responds to Bridge NetFn commands)
[5] IPMB Event Generator (device generates event messages [platform
event request messages] onto the IPMB)
[4] IPMB Event Receiver (device accepts event messages [platform event
request messages] from the IPMB)
[3] FRU Inventory Device
[2] SEL Device
[1] SDR Repository Device
[0] Sensor Device
Manufacturer ID, LS Byte first. The manufacturer ID is a 20-bit value that is
derived from the IANA ‘Private Enterprise’ ID (see below).
Most significant four bits = reserved (0000b).
000000h = unspecified. 0FFFFFh = reserved. This value is binary encoded.
E.g. the ID for the IPMI forum is 7154 decimal, which is 1BF2h, which would
be stored in this record as F2h, 1Bh, 00h for bytes 8 through 10, respectively.
Product ID, LS Byte first. This field can be used to provide a number that
identifies a particular system, module, add-in card, or board set. The number
is specified according to the manufacturer given by Manufacturer ID (see
below).
0000h = unspecified. FFFFh = reserved.
187
Intelligent Platform Management Interface Specification
(13:16)
Auxiliary Firmware Revision Information. This field is optional. If present, it
holds additional information about the firmware revision, such as boot block or
internal data structure version numbers. The meanings of the numbers are
specific to the vendor identified by Manufacturer ID (see below). When the
vendor-specific definition is not known, generic utilities should display each
byte as 2-digit hexadecimal numbers, with byte 13 displayed first as the mostsignificant byte.
The following presents additional specifications and descriptions for the Device ID response fields:
Device ID/Device Instance This number is specified by the manufacturer identified by the Manufacturer ID
field. The Device ID field allows controller-specific software to identify the
unique application command, OEM fields, and functionality that are provided
by the controller.
Controllers that have different application commands, or different definitions of
OEM fields, are expected to have different Device ID values. Controllers that
implement identical sets of applications commands can have the same Device
ID in a given system. Thus, a ‘standardized’ controller could be produced where
multiple instances of the controller are used in a system, and all have the same
Device ID value. [The controllers would still be differentiable by their address,
location, and associated information for the controllers in the Sensor Data
Records.]
The Device ID is typically used in combination with the Product ID field such
that the Device IDs for different controllers are unique under a given Product
ID. A controller can optionally use the Device ID as an ‘instance’ identifier if
more than one controller of that kind is used in the system. Though
implementing a Device GUID is the preferred method for uniquely identifying
controllers. (See section 17.8, Get Device GUIDGet Device GUID) This field is
binary encoded.
Device Revision
The least significant nibble of the Device Revision field is used to identify when
significant hardware changes have been made to the implementation of the
management controller that cannot be covered with a single firmware release.
That is, this field would be used to identify two builds off the same code
firmware base, but for different board fab levels. For example, device revision
"1" might be required for 'fab X and earlier' boards, while device revision "2"
would be for 'fab Y and later' boards. This field is binary encoded and unsigned.
Firmware Revision 1
Major Revision. 7-bits. This field holds the major revision of the firmware. This
field shall be incremented on major changes or extensions of the functionality of
the firmware - such as additions, deletions, or modifications of the command
set. This field is binary encoded and unsigned.
The Device Available bit is used to indicate whether normal command set
operation is available from the device, or it is operating in a state where only a
subset of the normal commands are available. This will typically be because the
device is in a firmware update state. It may also indicate that full command
functionality is not available because the device is in its initialization phase or
an SDR update is in progress.
Note that the revision information obtained when the Device Available bit is ‘1’
shall be indicative of the code version that is in effect. Thus, the version
information may vary with the Device Available bit state.
Firmware Revision 2
188
Minor Revision. This field holds the minor revision of the firmware. This field
will increment for minor changes such as bug fixes. This field is BCD encoded.
Intelligent Platform Management Interface Specification
IPMI Version
This field holds the version of the IPMI specification that the controller is
compatible with. This indicates conformance with this document, including
event message formats and mandatory command support. This field is BCD
encoded with bits 7:4 holding the Least Significant digit of the revision and bits
3:0 holding the Most Significant bits.
The value shall be 51h indicating conformance with this specification, version
1.5.
Additional Device
Support
This field indicates the logical device support that the device provides in addition
to the IPM and Application logical devices.
Manufacturer ID
This field uses the Internet Assigned Numbers Authority (http://www.iana.org/)
SMI Network Management Private Enterprise Codes a.k.a. “Enterprise
Numbers” for identifying the manufacturer responsible for the specification of
functionality of the vendor (OEM) -specific commands, codes, and interfaces
used in the controller.
For example, an event in the SEL could have OEM values in the event record.
An application that parses the SEL could extract the controller address from the
event record contents and use it to send the ‘Get Device ID’ command and
retrieve the Manufacturer ID. A manufacturer-specific application could then do
further interpretation based on a-priori knowledge of the OEM field, while a
generic cross-platform application would typically just use the ID to present the
manufacturer’s name alongside uninterpreted OEM event values.
The manufacturer that defines the functionality is not necessarily the
manufacturer that created the physical microcontroller. For example, Vendor A
may create the controller, but it gets loaded with Vendor B’s firmware. The
Manufacturer ID would be for Vendor B, since they’re the party that defined the
controller’s functionality.
The Manufacturer ID value from the Get Device ID command does not override
Manufacturer or OEM ID fields that are explicitly defined as part of a command
or record format.
If no vendor-specific functionality is defined, it is recommended that the field
can either be loaded with the Manufacturer ID of the party that is responsible for
the firmware for the controller, or the value FFFFh to indicate ‘unspecified’.
This field is binary encoded, and unsigned.
Product ID
This value can be used in combination with the Manufacturer ID and Device ID
values to identify the product-specific element of the controller-specific
functionality. This number is specified by the manufacturer identified by the
Manufacturer ID field.
Typically, a controller-specific application would use the Product ID to identify
the type of board, module, or system that the controller is used in, instead of
using the data from the FRU information associated with the controller.
Auxiliary Firmware
Revision Information
This field is optional. If present, it holds additional information about the firmware
revision, such as boot block or internal data structure version numbers. The
meanings of the numbers are specific to the vendor identified by Manufacturer ID
(above). When the vendor-specific definition is not known, generic utilities should
display each byte as 2-digit hexadecimal numbers, with byte 13 displayed first as
the most-significant byte.
189
Intelligent Platform Management Interface Specification
17.2 Cold Reset Command
This command directs the Responder to perform a ‘Cold Reset’ of itself. This causes default setting of interrupt
enables, event message generation, sensor scanning, threshold values, and other ‘power up default’ state to be
restored. That is, the device reinitializes its event, communication, and sensor functions. If the device incorporates
a Self Test, the Self Test will also run at this time.
Table 17-3, Cold Reset Command
Request Data
Response Data
byte
1
data field
Completion Code
Note: The Cold Reset command is provided for platform development, test, and platformspecific initialization and recovery actions. The system actions of the Cold Reset
command are platform specific. Issuing a Cold Reset command could have adverse
effects on system operation, particularly if issued during run-time. Therefore, the Cold
Reset command should not be used unless all the side-effects for the given platform are
known.
It is recognized that there are conditions where a given controller may not be able to
return a response to a Cold Reset Request message. Therefore, though recommended,
the implementation is not required to return a response to the Cold Reset command.
Applications should not rely on receiving a response as verification of the completion of a
Cold Reset command.
17.3 Warm Reset Command
This command directs the Responder to perform a ‘Warm Reset’ of itself. Communications interfaces shall be
reset, but current configurations of interrupt enables, thresholds, etc. will be left alone. A warm reset does not
initiate the Self Test. The intent of the Warm Reset command is to provide a mechanism for cleaning up the
internal state of the device and its communication interfaces. A Warm Reset will reset communication state
information such as sequence number and retry tracking, but shall not reset interface configuration information
such as addresses, enables, etc. An application may try a Warm Reset if it determines a non-responsive
communication interface - but it must also be capable of handling the side effects.
Table 17-4, Warm Reset Command
Request Data
Response Data
190
byte
1
data field
Completion Code
Intelligent Platform Management Interface Specification
17.4 Get Self Test Results Command
This command directs the device to return its Self Test results, if any. A device implementing a Self Test will
normally run that test on device power up as well as after Cold Reset commands. A device is allowed to update
this field during operation if it has tests that run while the device is operating. Devices that do not implement a
self test shall always return a 56h for this command.
While the Self Test only runs at particular times, the Get Self Test Results command can be issued any time the
device is in a ‘ready for commands’ state.
Table 17-5, Get Self Test Results Command
Request Data
Response Data
byte
1
2
3
data field
Completion Code
55h =
No error. All Self Tests Passed.
56h =
Self Test function not implemented in this controller.
57h =
Corrupted or inaccessible data or devices
58h =
Fatal hardware error (system should consider BMC
inoperative). This will indicate that the controller
hardware (including associated devices such as sensor
hardware or RAM) may need to be repaired or
replaced.
FFh =
reserved.
all other:
Device-specific ‘internal’ failure. Refer to the particular
device’s specification for definition.
For byte 2 = 55h, 56h, FFh:
00h
For byte 2 = 58h, all other:
Device-specific
For byte 2 = 57h: self-test error bitfield. Note: returning 57h does not
imply that all tests were run, just that a given test has failed. I.e. 1b
means ‘failed’, 0b means ‘unknown’.
[7]
1b = Cannot access SEL device
[6]
1b = Cannot access SDR Repository
[5]
1b = Cannot access BMC FRU device
[4]
1b = IPMB signal lines do not respond
[3]
1b = SDR Repository empty
[2]
1b = Internal Use Area of BMC FRU corrupted
[1]
1b = controller update ‘boot block’ firmware corrupted
[0]
1b = controller operational firmware corrupted
17.5 Manufacturing Test On Command
If the device supports a ‘manufacturing test mode’, this command is reserved to turn that mode on. The
specification of the functionality of this command is device dependent. A Cold Reset command shall, if accepted,
take the device out of ‘manufacturing test mode’ - as shall a physical reset of the device. Device-specific
commands to exit manufacturing test mode are also allowed.
Note that it may be possible to ‘lock out’ the command interface while in manufacturing test mode, in which case
the Cold Reset command or other mechanism for exiting manufacturing test mode may fail and a physical reset of
the device will be necessary to restore the device to normal operation.
The request parameters for this command are device specific. Typically, the parameters will be used for
transmitting a password or key that prevents manufacturing test mode from being entered unless the correct
values are provided.
191
Intelligent Platform Management Interface Specification
Table 17-6, Manufacturing Test On
Request Data
Response Data
byte
1:N
1
data field
device specific parameters. See text.
Completion Code
17.6 Set ACPI Power State Command
This command is provided to allow system software to tell a controller the present ACPI power state of the
system. The Set ACPI Power State command can also be used as a mechanism for setting elements of the platform
management subsystem to a particular power state. This is an independent setting that may not necessarily match
the actual power state of the system. This command is used to enable the reporting of the power state, it does not
control or change the power state.
There is corresponding information in sensor data record for the controller that tells system software which
controllers require this notification. The BMC does not automatically inform controllers of changes in the system
power state.
Since system management software does not run when the system is in a sleep state, the impact of sleep state on
the platform management subsystem is mainly one of changes in the automatic handling of sensor scanning and
events. For example, the system may shut down fans when in a particular power state. If the fans were monitored,
shutting down the fans without notifying the platform management subsystem could cause a false failure event to
be generated. Here are two possible ways to handle this:
a.
Have the management controller perform the fan shut down operation after receiving the Set ACPI
Power State command. In this case, the controller needs an SDR entry indicating that the controller
needs to receive notification via the Set ACPI Power State command.
b.
Have the controller monitor the system power state by proprietary means, such as a signal line directly
from the power control hardware to the management controller. The management controller uses the
signal to directly control the fans without receiving an Set ACPI Power State command. Note that that
controller should still report the power state using the Set ACPI Power State command. This is to aid outof-band applications that may directly access the controller to get sensor information.
Out-of-band applications should be prepared to find sensors or controllers that may have become disabled
because of a sleep state. Ideally, all management controllers should remain enabled while the system is in a sleep
state so that the sleep state information can be retrieved. Information in the SDR can be used to determine
whether a controller gets disabled in a particular sleep state. A system will normally power up to a Legacy On
state prior to the initialization of ACPI, at which time the system power state is known to be ACPI S0/G0.
192
Intelligent Platform Management Interface Specification
Table 17-7, Set ACPI Power State Command
Request Data
byte
1
2
Response Data
1
data field
ACPI System Power State to set
Power states are mutually exclusive. Only one state can be set at a
time.
[7] - 1b = set system power state
0b = don’t change system power state
[6:0] - System Power State enumeration
00h set S0 / G0
working
01h set S1
hardware context maintained, typically equates
to processor/chip set clocks stopped
02h set S2
typically equates to stopped clocks with
processor/cache context lost
03h set S3
typically equates to “suspend-to-RAM”
04h set S4
typically equates to “suspend-to-disk”
05h set S5 / G2
soft off
06h set S4/S5
sent when message source cannot differentiate
between S4 and S5
07h set G3
mechanical off
08h sleeping
sleeping - cannot differentiate between S1-S3.
09h G1 sleeping
sleeping - cannot differentiate between S1-S4
0Ah set override
S5 entered by override
20h set Legacy On Legacy On (indicates On for system that don’t
support ACPI or have ACPI capabilities
disabled)
21h set Legacy Off Legacy Soft-Off
2Ah set unknown system power state unknown
7Fh no change
Use this value when communicating a change
the device power state without indicating a
change to the system power state.
ACPI Device Power State to set
Power states are mutually exclusive. Only one state can be set at a
time.
[7] - 1 = set device power state
0 = don't change device power state
[6:0] - Device Power State enumeration
00h set D0
01h set D1
02h set D2
03h set D3
2Ah set unknown
7Fh no change Use this value when communicating a change the
system power state without indicating a change to the device
power state.
Completion Code
The BMC is allowed to return an error completion code if an attempt is
made to set states it knows the system doesn’t support.
193
Intelligent Platform Management Interface Specification
17.7 Get ACPI Power State Command
The command can also be used to retrieve the present power state information that has been set into the
controller. This is an independent setting from the system power state that may not necessarily match the actual
power state of the system. Unspecified bits and codes are reserved and shall be returned as 0.
Table 17-8, Get ACPI Power State Command
Request Data
Response Data
byte
1
2
3
194
data field
Completion Code
ACPI System Power State
[7] - reserved
[6:0] - System Power State enumeration
00h S0 / G0
working
01h S1
hardware context maintained, typically equates to
processor/chip set clocks stopped
02h S2
typically equates to stopped clocks with
processor/cache context lost
03h S3
typically equates to “suspend-to-RAM”
04h S4
typically equates to “suspend-to-disk”
05h S5 / G2
soft off
06h S4/S5
soft off, cannot differentiate between S4 and S5
07h G3
mechanical off
08h sleeping
sleeping - cannot differentiate between S1-S3.
09h G1 sleeping sleeping - cannot differentiate between S1-S4
0Ah override
S5 entered by override
20h Legacy On Legacy On (indicates On for system that don’t
support ACPI or have ACPI capabilities disabled)
21h Legacy Off Legacy Soft-Off
2Ah unknown
power state has not been initialized, or device
lost track of power state.
ACPI Device Power State
[7] reserved
[6:0] Device Power State enumeration
00h D0
01h D1
02h D2
03h D3
2Ah unknown - power state has not been initialized, or device lost
track of power state.
Intelligent Platform Management Interface Specification
17.8 Get Device GUID Command
This command returns a Globally Unique ID (GUID) for the management controller. The format of the ID
follows that specified in the Attachment A of the Wired for Management Baseline, Version 2.0 specification. The
ID can be used to uniquely identify an instance of a management controller. This information can be used in
conjunction with SDR information to verify that a particular controller is still present in the system. IPMI uses
version 1 of the GUID format, with a three bit variant field of 10x (where x indicates ‘don’t care’).
Table 17-9, Get Device GUID Command
Request Data
Response Data
1
2:17
Completion Code
GUID bytes 1 through 16.
Note that the individual fields within the GUID are stored least-significant byte first, and in the order illustrated
in the following table. This is the reverse of convention described in the Wired for Management Specification
where GUID bytes are transmitted in ‘network order’ (most-significant byte first) starting with the time low field.
Table 17-10, GUID Format
GUID byte
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Field
node
node
node
node
node
node
clock seq and reserved
clock seq and reserved
time high and version
time high and version
time mid
time mid
time low
time low
time low
time low
MSbyte
MSbyte
MSbyte
MSbyte
MSbyte
MSbyte
195
Intelligent Platform Management Interface Specification
17.9 Broadcast ‘Get Device ID’
This is a broadcast version of the ‘Get Device ID’ command that is provided for the ‘discovery’ of Intelligent
Devices on the IPMB. It is only specified for use on the IPMB. Discovery of management controllers on a PCI
Management Bus is handled via the SMBus 2.0 ‘ARP’ protocol. See [SMBUS] for more information.
The Broadcast ‘Get Device ID’ command is not bridged but can be delivered to the IPMB using Master WriteRead commands.
To perform a ‘discovery’ the command is repeatedly broadcast with a different rsSA ‘slave address parameter’
field specified in the command. The device that has the matching physical slave address information shall respond
with the same data it would return from a ‘regular’ (non-broadcast) ‘Get Device ID’ command. Since an IPMB
response message carries the Responder’s Slave Address, the response to the broadcast provides a positive
confirmation that an Intelligent Device exists at the slave address given by the rsSA field in the request.
An application driving discovery then cycles through the possible range of IPMB Device slave addresses to find
the population of intelligent devices on the IPMB. Refer to [ADDR] for information on which slave address
ranges are allocated for different uses on IPMB.
Refer to the description of the Get Device ID command, above, for information on the fields returned by the
Broadcast Get Device ID command response. The IPMB message format for the Broadcast Get Device ID
request exactly matches that for the Get Device ID command, with the exception that the IPMB message is
prefixed with the 00h broadcast address. The following illustrates the format of the IPMB Broadcast Get Device
ID request message:
Figure 17-1, Broadcast Get Device ID Request Message
Broadcast
(00h)
rsSA
netFn/rsLUN
check1
rqSA
rqSeq/rqLUN
Cmd
(01h)
check2
Addresses 00h-0Fh and F0h-FFh are reserved for I2C functions and will not be used for IPM devices on the
IPMB. These addresses can therefore be skipped if using the Broadcast Get Device ID command to scan for IPM
devices. The remaining fields follow the regular IPMB definitions.
In order to speed the discovery process on the IPMB, a controller should drop off the bus as soon as it sees that
the rsSA in the command doesn’t match its rsSA.
196
Intelligent Platform Management Interface Specification
197
Intelligent Platform Management Interface Specification
18. IPMI Messaging Support Commands
This section defines the commands used to support the system messaging interfaces. This includes control bits for
using the BMC as an Event Receiver and SEL Device. SMM Messaging and Event Message Buffer support is
optional. Use of IPMI support for SMI’s and SMM Messaging is deprecated. Configuration interface support for
enabling/disabling SMM Messaging and corresponding SMI has been removed from the specification. If SMM
Messaging were implemented using the IPMI infrastructure, it would now be done as an OEM-proprietary
capability.
System software that is not explicitly aware of the particular platform’s use of SMI Messaging must assume that
the any SMI options have been pre-configured by the controller, system BIOS, or other software. Therefore, runtime system software should not reconfigure SMI options, nor should it access the Event Message Buffer if it
finds that Event Message Buffer interrupt is mapped to SMI. The effects of SMS accessing the Event Message
Buffer when it is configured for SMI are unspecified. Refer to Appendix G - Command Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 18-1, IPMI Messaging Support Commands
Command
Set BMC Global Enables
Get BMC Global Enables
Clear Message Flags
Get Message Flags
Enable Message Channel Receive
Get Message
Send Message
Read Event Message Buffer
Get BT Interface Capabilities
Master Write-Read
Get System GUID
Get Channel Authentication Capabilities
Get Session Challenge
Activate Session
Set Session Privilege Level
Close Session
Get Session Info
Get AuthCode
Set Channel Access
Get Channel Access
Get Channel Info Command
Set User Access Command
Get User Access Command
Set User Name
Get User Name Command
Set User Password Command
1.
2.
3.
4.
198
Section
Defined
O/M
18.1
18.2
18.3
18.4
18.5
18.6
18.7
18.8
18.9
18.10
18.13
18.12
18.14
18.15
18.16
18.17
18.18
18.19
18.20
18.21
18.22
18.23
18.24
18.25
18.26
18.27
M
M
M
M
O
[1]
M
[1]
M
O
[2]
M
[6]
M
[5]
O
[3]
O
[4]
O
[4]
O
[4]
O
[4]
O
[4]
O
O
[4]
O
[4]
O
[4]
O
[4]
O
[4]
O
[5]
O
[4]
O
[4]
O
Optional if the System Interface is the only channel that’s implemented.
Mandatory only if BT (block transfer) System Interface is used.
Mandatory if Private Bus FRU SEEPROMs or IPMB implemented.
Mandatory if session-based channels are supported
Intelligent Platform Management Interface Specification
5.
6.
Highly recommended for session-based channels. It is also recommended that
the implementation support multiple of users with configurable usernames.
Mandatory for a BMC that includes IPMB or PCI SMBus channels, or for any
BMC or satellite controller that implements a private management bus for FRU
SEEPROM access.
18.1 Set BMC Global Enables Command
This command is used to enable message reception into Message Buffers, and any interrupt associated with that
buffer getting full. The OEM0, OEM 1, and OEM 2 flags are available for definition by the OEM/System
Integrator. Generic system management software must not alter these bits.
Table 18-2, Set BMC Global Enables Command
Request Data
byte
1
Response Data
1
data field
This field is set to xxxx_100xb on power-up and system resets. If the
implementation allows the receive message queue interrupt to be
enabled/disabled, the default for bit 0 should be 0b, otherwise it should always
be 1b.
[7] OEM 2 Enable. Generic system mgmt. software must do a ‘read-modifywrite’ using the Get BMC Global Enables and Set BMC Global
Enables to avoid altering this bit.
[6] OEM 1 Enable. Generic system mgmt. software must do a ‘read-modifywrite’ using the Get BMC Global Enables and Set BMC Global
Enables to avoid altering this bit.
[5] OEM 0 Enable. Generic system mgmt. software must do a ‘read-modifywrite’ using the Get BMC Global Enables and Set BMC Global
Enables to avoid altering this bit.
[4] reserved
[3] 1b = Enable System Event Logging (enables/disables logging of events
to the SEL - with the exception of events received over the system
interface. Event reception and logging via the system interface is
always enabled.) SEL Logging is enabled by default whenever the
BMC is first powered up. It’s recommended that this default state
also be restored on system resets and power on.
[2] 1b = Enable Event Message Buffer. Error completion code returned if
written as ‘1’ and Event Message Buffer not supported.
[1] 1b = Enable Event Message Buffer Full Interrupt
[0] 1b = Enable Receive Message Queue Interrupt (this bit also controls
whether KCS communication interrupts are enabled or disabled. An
implementation is allowed to have this interrupt always enabled.)
Completion Code.
18.2 Get BMC Global Enables Command
This command is used to retrieve the present setting of the Global Enables. The OEM0, OEM 1, and OEM 2 flags
are available for definition by the OEM/System Integrator. Generic system management software must ignore these
bits.
199
Intelligent Platform Management Interface Specification
Table 18-3, Get BMC Global Enables Command
Request Data
Response Data
byte
1
2
data field
Completion Code
[7] [6] [5] [4] [3] [2] [1] [0] -
1b = OEM 2 Enabled.
1b = OEM 1 Enabled.
1b = OEM 0 Enabled.
reserved
1b = System Event Logging Enabled
1b = Event Message Buffer Enabled
1b = Event Message Buffer Full Interrupt Enabled
1b = Receive Message Queue Interrupt Enabled (this bit also indicates
whether KCS communication interrupt are enabled or disabled.)
18.3 Clear Message Flags Command
This command is used to flush unread data from the Receive Message Queue or Event Message Buffer. This will
also clear the associated buffer full / message available flags. See Get Message Flags command.
Table 18-4, Clear Message Flags Command
Request Data
byte
1
Response Data
1
data field
[7] - 1b = Clear OEM 2
[6] - 1b = Clear OEM 1
[5] - 1b = Clear OEM 0
[4] - reserved
[3] - 1b = Clear watchdog pre-timeout interrupt flag
[2] - reserved
[1] - 1b = Clear Event Message Buffer.
[0] - 1b = Clear Receive Message Queue.
Completion Code.
Implementations are not required to return an error completion code if an
attempt is made to clear the Event Message Buffer flag but the Event
Message Buffer is not supported.
18.4 Get Message Flags Command
This command is used to retrieve the present ‘message available’ states. The OEM0, OEM 1, and OEM 2 flags are
available for definition by the OEM/System Integrator. Generic system management software must ignore these
bits.
200
Intelligent Platform Management Interface Specification
Table 18-5, Get Message Flags Command
Request Data
Response Data
byte
1
data field
-
1
Completion Code.
2
Flags
[7] - 1b = OEM 2 data available.
[6] - 1b = OEM 1 data available.
[5] - 1b = OEM 0 data available.
[4] - reserved
[3] - 1b = Watchdog pre-timeout interrupt occurred
[2] - reserved
[1] - 1b = Event Message Buffer Full. (Return as 0 if Event Message Buffer is
not supported, or when the Event Message buffer is disabled.)
[0] - 1b = Receive Message Available. One or more messages ready for
reading from Receive Message Queue
18.5 Enable Message Channel Receive Command
This command is used to enable/disable message reception into the Receive Message Queue from a given
message channel. The command provides a mechanism to allow SMS to only receive messages from channels
that it intends to process, and provides a disable mechanism in case the receive message queue is being
erroneously or maliciously flooded with requests on a particular channel. It does not affect the ability for SMS to
transmit on that channel. Only the SMS Message channel is enabled by default. All other channels must be
explicitly enabled by BIOS or system software, as appropriate. It is recommended that a ‘Destination
Unavailable’ completion code be returned if a request message to SMS is rejected because reception has been
disabled.
Table 18-6, Enable Message Channel Receive Command
Request Data
byte
1
2
Response Data
1
2
3
data field
Channel Number
[7:4] - reserved
[3:0] - channel number
Channel State
[7:2] - reserved
[1:0] - 00b = disable channel
01b = enable channel
10b = get channel enable/disable state
11b = reserved
Completion Code
Channel Number
[7:4] - reserved
[3:0] - channel number
Channel State
[7:1] - reserved
[0] 1b = channel enabled
0b = channel disabled
18.6 Get Message Command
This command is used to get data from the Receive Message Queue. Refer to Table 6-8, IPMI Message and
IPMB / Private Bus Transaction Size Requirements, for information that can be used to determine the sizes of
messages that need to be supported for a given Receive Message Queue implementation.
201
Intelligent Platform Management Interface Specification
Software is responsible for reading all messages from the message queue even if the message is not the expected
response to an earlier Send Message. System management software is responsible for matching responses up with
requests.
The Get Message command includes an “inferred privilege level” that is returned with the message. This can help
avoid the need for software to implement a separate privilege-level and authentication mechanism. This works as
follows: Suppose a user activates a session with a maximum privilege level of Administrator on a multi-session
channel, and that an MD5 authentication type was negotiated. Also suppose that User-level authentication is
disabled. A user that has User or higher privilege can place messages into the receive message queue by sending
them to LUN 10b, or by using the Send Message command. If the packet has Authentication Type = MD5, the
packet will be assigned an inferred privilege level based the on the present operating privilege level for the user
(set using the Set Session Privilege Level command). If, before sending the packet, the user had set their privilege
level to Operator, the packet would be assigned an inferred privilege level of Operator. (Note that this means an
authenticated (signed) packet can be assigned different inferred privilege levels based on the present operating
privilege set by the Set Session Privilege Level command.) If the message is received in a packet that has
Authentication Type = None, the packet will be assigned an inferred privilege level of ‘User’, since that is the
lowest privilege level for which that type of authentication is accepted.
202
Intelligent Platform Management Interface Specification
Now suppose that the remote user had used the receive message queue as a way to send a message to system
management software that requests a soft shutdown of the operating system. The message would either have an
inferred privilege level of ‘Operator’ or ‘User’ depending on whether it was sent as an authenticated message or
not. System Management Software can then use that inferred privilege level as part of deciding whether to accept
and process the message or not. For single-session channels, the inferred privilege level is always set to the
present operating privilege level. For session-less channels, the inferred privilege level is set to ‘None’, indicating
that there was no IPMI-specified authentication operating on the channel from which the message was received.
Table 18-7, Get Message Command
Request Data
Response Data
byte
1
2
data field
Completion Code
generic, plus following command specific completion code:
80h = data not available (queue / buffer empty)
Implementation of this completion code is Mandatory. The code eliminates
the need for system software to always check the Message Buffer Flags to
see if there data left in the Receive Message Queue. If a non-OK, non-80h
completion is encountered - software will need to check the Message Flags
to get the empty/non-empty status of the Receive Message Queue.
Channel Number
[7:4] Inferred privilege level for message.
When the BMC receives a message for the receive message queue, it
assigns an ‘inferred privilege level’ to the message as follows:
If the message is received from a session-based channel, it will initially be
assigned a privilege level that matches the ‘maximum requested privilege
level’ that was negotiated via the Activate Session command.
If per-message authentication is enabled, but User-level authentication is
disabled, the BMC will assign a level of ‘User’ to any messages that are
received with an Authentication Type = none. (Note that per-message and
user-level authentication options only apply to multi-session channels)
The BMC will then lower the assigned privilege limit, if necessary, based on
the present session privilege limit that was set via the Set Session Privilege
Level command.
If the channel is session-less (e.g. IPMB), the BMC will return ‘None’ for the
privilege level.
0h = None (unspecified)
1h = Callback level
2h = User level
3h = Operator level
4h = Administrator level
5h = OEM Proprietary level
3:N
[3:0] channel number
Message Data. Max. Length & format dependent on protocol associated with
channel.
203
Intelligent Platform Management Interface Specification
The following table indicates the contents of the Message Data field from the Get Message response according to
the Channel Type and Channel Protocol that was used to place the message in the Receive Message Queue.
Table 18-8, Get Message Data Fields
1
Originating Channel Type
2
IPMB (I C)
Channel
Protocol
IPMB
2
ICMB v1.0
ICMB-1.0
3
ICMB v0.9
ICMB-0.9
4
802.3 LAN
IPMB
5
Asynch. Serial/Modem
(RS-232)
6
7
8
9
Other LAN
PCI SMBus
SMBus v1.0/1.1
SMBus v2.0
10
11
12
reserved for USB 1.x
reserved for USB 2.x
System Interface
204
IPMB (Basic
Mode, Terminal
Mode, and PPP
Mode)
IPMB
IPMI-SMBus
n/a
n/a
BT, KCS, SMIC
Message Data for received requests(RQ)
and responses (RS)
RQ: netFn/rsLUN, chk1, rqSA, rqSeq/rqLUN, cmd,
<data>, chk2
RS: netFn/rqLUN, chk1, rsSA, rqSeq/rsLUN, cmd,
completion code, <data>, chk2
See Section 8.2, ICMB Bridge Commands in BMC
using ChannelsICMB Bridge Commands in BMC using
Channels
See Section 8.2, ICMB Bridge Commands in BMC
using ChannelsICMB Bridge Commands in BMC using
Channels
RQ: Session Handle, rsSWID, netFn/rsLUN, chk1,
rqSWID or rqSA, rqSeq/rqLUN, cmd, <data>,
chk2
RS: Session Handle, rqSWID, netFn/rsLUN, chk1,
rsSWID or rsSA, rqSeq/rsLUN, cmd, completion
code, <data>, chk2
RQ/RS: See row for Originating Channel Type = 802.3
LAN
Note: When LUN 10b is used to deliver a message to
SMS from a Terminal Mode remote console, the
BMC inserts fixed values for the SWIDs and LUNs
in the message. Messages from the remote
console are always returned as coming from
SWID 40h (81h) LUN 00b, and going to SMS
SWID 20h (41h) LUN 00b.
See row for Originating Channel Type = 802.3 LAN
RQ: rsSA, Netfn(even)/rsLUN, 00h, rqSA,
rqSeq/rqLUN, CMD, <data>, PEC
RS: rqSA or rqSWID, NetFn(odd)/rqLUN, 00h, rsSA or
rsSWID, rqSeq/rsLUN, CMD, completion code,
<data>, PEC
n/a
n/a
RQ/RS: See row for Originating Channel Type = 802.3
LAN.
Intelligent Platform Management Interface Specification
18.7 Send Message Command
The Send Message command is used for bridging IPMI messages between channels, and between the system
management software (SMS) and a given channel. Refer to 6.12, BMC Message BridgingBMC Message
Bridging, for information on how the Send Message command is used.
Table 18-9, Send Message Command
byte
Request Data
1
data field
Channel Number
[7:6] 00b = No tracking. The BMC reformats the message for the selected channel but does not track
the originating channel, sequence number, or address information. This option is typically
used when software sends a message from the system interface to another media.
Software will typically use ‘no tracking’ when it delivers sends a message from the system
interface to another channel, such as IPMB. In this case, software will format the
encapsulated message so that when it appears on the other channel, it will appear to have
been directly originated by BMC LUN 10b. See 6.12.1, BMC LUN 10b RoutingBMC LUN
10b Routing.
01b = Track Request. The BMC records the originating channel, sequence number, and
addressing information for the requester, and then reformats the message for the protocol
of the destination channel. When a response is returned, the BMC looks up the requester’s
information and format the response message with the framing and destination address
information and reformats the response for delivery back to the requester. This option is
used for delivering IPMI Request messages from non-SMS (non-system interface)
channels. See 6.12.3, Send Message Command with Response TrackingSend Message
Command with Response Tracking.
10b = Send Raw. (optional) This option is primarily provided for test purposes. It may also be
used for proprietary messaging purposes. The BMC simply delivers the encapsulated data
to the selected channel in place of the IPMI Message data. Note that if the channel uses
sessions, the first byte of the Message Data field must be a Session Handle. The BMC
should return a non-zero completion code if an attempt is made to use this option for a
given channel and the option is not supported. It is recommended that completion code
CCh be returned for this condition.
11b = reserved
2:N
Response Data
1
(2:N)(
2)
[5:4] reserved
[3:0] channel number to send message to.
Message Data. Format dependent on target channel type. See Table 18-10, Message Data for Send
Message CommandTable 18-10, Message Data for Send Message Command
Completion Code
generic, plus additional command-specific completion codes:
80h = Invalid Session Handle. The session handle does not match up with any currently active
sessions for this channel.
If channel medium = IPMB, SMBus, or PCI Management Bus:
2
(This status is important for applications that need to access low-level I C or SMBus devices and
should be implemented.)
81h = Lost Arbitration
82h = Bus Error
83h = NAK on Write
Response Data
This data will only be present when using the Send Message command to originate requests from
IPMB or PCI Management Bus to other channels such as LAN or serial/modem. It is not present in the
response to a Send Message command delivered via the System Interface.
NOTE: The BMC does not parse messages that are encapsulated in a Send Message command. Therefore,
it does not know what privilege level should associated with an encapsulated message. Thus,
messages that are sent to a session using the Send Message command are always output using the
Authentication Type that was negotiated when the session was activated.
205
Intelligent Platform Management Interface Specification
The following table summarizes the contents of the Message Data field when the Send Message command is used
to deliver an IPMI Message to different channel types. Note that in most cases the format of message information
the Message Data field follows that used for the IPMB, with two typical exceptions: When the message is
delivered to channels without physical slave devices, a software ID (SWID) field takes the place of the slave
address field. When the message is delivered to a channel that supports sessions, the first byte of the message data
holds a Session Handle.
Table 18-10, Message Data for Send Message Command
1.
206
1
Target Channel Type
2
IPMB (I C)
Target
Channel
Protocol
IPMB
2
ICMB v1.0
ICMB-1.0
3
ICMB v0.9
ICMB-0.9
4
802.3 LAN
IPMB+session
header
5
Asynch. Serial/Modem
(RS-232)
IPMB (Basic
Mode, Terminal
Mode, and PPP
Mode)
6
7
8
9
10
11
12
Other LAN
PCI SMBus
SMBus v1.0/1.1
SMBus v2.0
reserved for USB 1.x
reserved for USB 2.x
System Interface
IPMB
IPMI-SMBus
n/a
n/a
Message Data field contents for Send Message
command for sending requests(RQ) and responses
(RS)
RQ: rsSA, netFn/rsLUN, chk1, rqSA, rqSeq/rqLUN,
cmd, <data>, chk2
RS: rqSA, netFn/rqLUN, chk1, rsSA, rqSeq/rsLUN,
cmd, completion code, <data>, chk2
See Section 8.2, ICMB Bridge Commands in BMC
using ChannelsICMB Bridge Commands in BMC
using Channels
See Section 8.2, ICMB Bridge Commands in BMC
using ChannelsICMB Bridge Commands in BMC
using Channels
[1]
RQ: Session Handle , rsSWID, netFn/rsLUN, chk1,
rqSWID or rqSA, rqSeq/rqLUN, cmd, <data>,
chk2
[1]
RS: Session Handle , rqSWID, netFn/rsLUN, chk1,
rsSWID or rsSA, rqSeq/rsLUN, cmd,
completion code, <data>, chk2
RQ/RS: See Target Channel Type = 802.3 LAN
Note: Terminal mode has a single, fixed SWID for
the remote console, software using Send
Message to deliver a message to a terminal
mode remote console should use their SWID or
slave address as the source of the request or
response, and the Terminal Mode SWID (40h)
as the destination.
See Target Channel Type = 802.3 LAN
See Target Channel Type = IPMB
n/a
n/a
RQ/RS: See Target Channel Type = IPMB
Note: BMC adds Session Handle info when it puts
the message into the Receive Message Queue.
Session Handle. Identifies the particular active session for this channel. The session handle identifies a particular active
session on the given channel. The BMC assigns a different value to each time a new session is activated. A typical
implementation will keep track of the last value that was assigned and increment it before assigning it to a new active session
when the Activate Session command has been accepted. Software must include this field for channels where the Get Channel
Info command indicates that the channel supports sessions.
Intelligent Platform Management Interface Specification
18.8 Read Event Message Buffer Command
This command is used to retrieve the contents of the Event Message Buffer. Reading the buffer automatically
clears the Event Message Buffer Full flag from the Get Message Flags command.
Table 18-11, Read Event Message Buffer Command
Request Data
Response Data
byte
1
2:17
data field
Completion Code.
generic, plus additional command-specific completion codes:
80h = data not available (queue / buffer empty)
Message Data. 16 bytes of data in SEL Record format, per Table 26-1, SEL
Event RecordsTable 26-1, SEL Event Records. A dummy Record ID of FFFFh
should be returned for events placed in the Event Message Buffer while event
logging is disabled or if the SEL is full. System management software should
ignore the record ID when event logging is disabled.
18.9 Get BT Interface Capabilities Command
The BT interface includes a Get BT Interface Capabilities command that returns various characteristics of the
interface, including buffer sizes, and multithreaded communications capabilities.
Table 18-12, Get BT Interface Capabilities Command
Request Data
Response Data
byte
-
-
1
Completion Code
2
Number of outstanding requests supported (1 based. 0 illegal)
[1]
Input (request) buffer message size in bytes. (1 based.)
[1]
Output (response) buffer message size in bytes. (1 based.)
BMC Request-to-Response time, in seconds, 1 based. 30 seconds,
maximum.
Recommended retries (1 based). (see text for BT Interface)
3
4
5
6
1.
data field
For Bytes 3 and 4 (Input and Output Buffer size), the buffer message size is the
largest value allowed in first byte (length field) of any BT request or response
message. For a send, this means if Get BT Interface Capabilities returns 255 in
byte 3 (input buffer size) the driver can actually write 256 bytes to the input buffer
(adding one for the length byte (byte 1) that is sent in with the request.)
207
Intelligent Platform Management Interface Specification
18.10 Master Write-Read Command
This command can be used for low-level I2C/SMBus write, read, or write-read accesses to the IPMB or private
busses behind a management controller. The command can also be used for providing low-level access to devices
that provide an SMBus slave interface.
Table 18-13, Master Write-Read Command
Request Data
byte
1
2
3
4:N
Response Data
1
(2:M)
data field
bus ID:
[7:4] channel number
[3:1] bus ID, 0-based (always 000b for public bus)
[0] bus type:
0 = public (e.g. IPMB or PCI Management Bus)
1 = private bus
[7:1] - Slave Address
[0] - reserved. Write as 0.
Read count. Number of bytes to read, 1 based. 0 = no bytes to read. The
maximum read count should be at least 34 bytes. This allows the command to
be used for an SMBus Block Read. This is required if the command provides
access to an SMBus or IPMB. Otherwise, if FRU SEEPROM devices are
accessible, at least 31 bytes must be supported. Note that an implementation
can support fewer bytes can be supported if none of the devices to be
accessed can handle the recommended minimum.
Data to write. This command should support at least 35 bytes of write data.
This allows the command to be used for an SMBus Block Write with PEC.
Otherwise, if FRU SEEPROM devices are accessible, at least 31 bytes must
be supported. Note that an implementation is allowed to support fewer bytes if
none of the devices to can handle the recommended minimum.
Completion Code
A management controller shall return an error Completion Code if an attempt
is made to access an unsupported bus.
generic, plus following command specific codes:
81h = Lost Arbitration
82h = Bus Error
83h = NAK on Write
84h = Truncated Read
Bytes read from specified slave address. This field will be absent if the read
2
count is 0. The controller terminates the I C transaction with a STOP condition
after reading the requested number of bytes.
18.11 Session Header Fields
Whether the session header fields are present in a packet is based on whether the channel is specified as
supporting multiple sessions or not. In addition, which session fields are present is based on the authentication
type. Single-session connections and session-less channels do not include session header fields.
Session header fields are present on all packets where the channel and connection mode is specified as
supporting multiple sessions, even if the particular implementation only supports one session. The following
tables for the Get System GUID, Get Channel Authentication Capabilities, Get Session Challenge, and Activate
Session commands explicitly show ‘session header’ fields in gray. This is done because those commands can be
executed prior to a session being activated, and therefore certain session header field values are handled
differently than they are after a session is established.
The grayed session header fields illustrate which session header fields are present and what their values are
required to be, but they do not serve to specify the order or organization of those fields in the packet for a
particular medium. Refer to the example packet format figures for (e.g. Table 12-8, RMCP Packet for IPMI via
208
Intelligent Platform Management Interface Specification
EthernetTable 12-8, RMCP Packet for IPMI via Ethernet) for the specification of the organization and position
of the session header bytes for a particular medium.
The following applies to packets on connections that are specified with support for multiple sessions:
•
Session header fields are present on all packets where the channel and connection mode is specified as
supporting multiple sessions, even if the particular implementation only supports one session.
•
Note that the command tables do not show the session header fields except for the Get Channel
Authentication Capabilities, Get Session Challenge, and Activate Session commands. However, they are
still required for all commands on a multi-session connection.
•
The Authentication Code field in the session header may or may not be present based on the Authentication
Type. The authentication code field is absent whenever the Authentication Type is NONE. Whether the
authentication code field is present or not when the Authentication Type = OEM is dependent on the OEM
identified in the Get Channel Authentication Capabilities command.
•
If per-message authentication is turned off for the channel, only the Activate Session command would use a
non-NONE authentication type and include an AuthCode signature. All other commands under the session
are sent with Authentication Type = NONE.
•
If per-message authentication is turned off and a packet is received that has a non-NONE authentication
type, it will be accepted (if the authentication type is supported), however the implementation is not
required to authenticate the packet. Note that an implementation may authenticate the packet, therefore the
Authentication Code must be correct.
•
If User authentication is turned off for the channel, the behavior is the same as if per-message
authentication is turned off, except that only packets for commands that are enabled at User privilege level
are sent with Authentication Type = NONE. All other packets must be authenticated per the Authentication
Type used to activate the session.
18.12 Get Channel Authentication Capabilities Command
This command is sent in unauthenticated (clear) format. This command is used to retrieve capability information
about the channel that the message is delivered over, or for a particular channel. The command returns the
authentication algorithm support for the given privilege level. When activating a session, the privilege level
passed in this command will normally be the same Requested Maximum Privilege level that will be used for a
subsequent Activate Session command.
The Get Channel Authentication Capabilities command can also be used as a no-op ‘Ping’ to keep a session from
timing out.
209
Intelligent Platform Management Interface Specification
Table 18-14, Get Channel Authentication Capabilities Command
Session Request Data
IPMI Request Data
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
1
2
210
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
Requested Maximum Privilege Level
[7:4] - reserved
[3:0] - requested privilege level
0h = reserved
1h = Callback level
2h = User level
3h = Operator level
4h = Administrator level
5h = OEM Proprietary level
Intelligent Platform Management Interface Specification
Session Response Data
IPMI Response Data
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
1
2
3
4
Completion Code
Channel Number
Channel number that the Authentication Capabilities is being returned
for. If the channel number in the request was set to Eh, this will return
the channel number for the channel that the request was received on.
Authentication Type Support
Returns the setting of the Authentication Type Enable field from the
configuration parameters for the given channel that corresponds to the
Requested Maximum Privilege Level.
[7:6] - reserved
[5:0] - Authentication type(s) enabled for given Requested Maximum Privilege
Level
All bits: 1b = supported
0b = authentication type not available for use.
[5] OEM proprietary (per OEM identified by the IANA OEM ID in
the RMCP Ping Response)
[4] straight password / key
[3] reserved
[2] MD5
[1] MD2
[0] none
[7:5] - reserved
[4] - Per-message Authentication status
0b = Per-message Authentication is enabled. Packets to the BMC must
be authenticated per Authentication Type used to activate the
session, and User Level Authentication setting, following.
1b = Per-message Authentication is disabled. Authentication Type ‘none’
accepted for packets to the BMC after the session has been
activated.
[3] - User Level Authentication status
0b = User Level Authentication is enabled. User Level commands must
be authenticated per Authentication Type used to activate the
session.
1b = User Level Authentication is disabled. Authentication Type ‘none’
accepted for User Level commands to the BMC.
5
6:8
9
[2:0] - Anonymous Login status
(see sections 6.9.1, ‘Anonymous Login’ Convention‘Anonymous Login’
Convention and 6.9.2, Anonymous Login Anonymous Login )
[2] - 1b = Non-null user names enabled
[1] - 1b = Null user name enabled (A user that has a null username, but
non-null password, is presently enabled)
[0] - 1b = Anonymous Login enabled (A user that has a null username
and null password is presently enabled)
reserved
OEM ID.
IANA Enterprise Number for OEM/Organization that specified the
particular OEM Authentication Type. Least significant byte first.
Return 00h, 00h, 00h if no OEM authentication type available.
OEM auxiliary data.
Additional OEM-specific information for the OEM Authentication Type.
Return 00h if no OEM authentication type available.
211
Intelligent Platform Management Interface Specification
18.13 Get System GUID Command
This optional, though highly recommended, command can be used to return a globally unique ID (GUID) for the
system to support the remote discovery process and other operations. Since the GUID is typically ‘permanently’
assigned to a system, an interface that would allow the GUID to be configured or changed is not specified. The
session header (Session Request data and Session Response Data) values shown illustrate the values that would
be used to execute a Get System GUID Command outside of an active session. The Get System GUID will always
be accepted outside of an active session.
The Get System GUID command can also be executed in the context of an active session (providing the user is
operating at higher than ‘Callback’ privilege). When the Get System GUID command is executed in the context of
an active session, the session header fields must have correct values according to the authentication, session ID,
and session sequence number information that was negotiated for the session.
Table 18-15, Get System GUID Command
Session Request Data
Request Data
Session Response Data
Response Data
-
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
-
1
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
Completion Code
2:17
GUID bytes 1 through 16. See Table 17-10, GUID FormatTable 17-10,
GUID Format.
18.14 Get Session Challenge Command
This command is sent in unauthenticated format. While a session ID is returned from the response to the Get
Session Challenge command, the session must be activated using the Activate Session command before it can be
used for sending other authenticated commands. The Activate Session command provides the starting sequence
number for subsequent messages under the session.
When the management controller looks up user names the controller scans the names sequentially by user ID
starting from User ID 1. Disabled user names are skipped. The scan stops when the first matching enabled user
name is made.
212
Intelligent Platform Management Interface Specification
Table 18-16, Get Session Challenge Command
byte
Session Request Data
IPMI Request Data
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
1
2:17
Session Response Data
IPMI Response Data
data field
Authentication Type for Challenge
[7:4] - reserved
[3:0] - requested Authentication Type
0h = none. No hashing or authentication done on session packets.
Authentication Code field is not present.
1h = MD2
2h = MD5
3h = reserved
4h = straight password / key
5h = OEM proprietary
all other = reserved
User Name. Sixteen-bytes. All 0’s for null user name (User 1)
authentication type = NONE
session seq# = null (0’s)
session ID = null (0’s)
AuthCode = NOT PRESENT
1
2:5
6:21
Completion Code
81h = invalid user name
82h = null user name (User 1) not enabled
Temporary Session ID. LS byte first.
This is a provision for a temporary Session ID that can be given out to
parties that have requested challenges, but have not yet activated a
session. It can be used as a mechanism to help protect against denial of
service attacks by grabbing all free session IDs.
Challenge string data
18.15 Activate Session Command
While a Session ID is returned from the response to the Get Session Challenge command, the session must be
activated using the Activate Session command before it can be used for sending other authenticated commands.
The initial Activate Session command is used by the remote console to set the starting sequence number for
subsequent messages under the session. When the Activate Session command is issued (for a given session ID)
the outbound session sequence number is set by the remote console and can be any random value.
For a given temporary session ID, the BMC must accept Activate Session commands with a null session sequence
number and silently discard all other commands targeted to that session ID. This provision is to enable a remote
console to retry the Activate Session command in case the response was lost. The BMC will continue to accept the
Activate Session command with a null session sequence number until the first valid and appropriately
authenticated command with a non-null session sequence number is received. (The non-null sequence number
must also be within the range specified by the initial inbound sequence number). After which, all subsequent
commands for the session must have appropriately incremented, non-null sequence number values, including any
Activate Session commands that may be received during session operation.
The remote console can use an Activate Session command to change the outbound session sequence number
during session operation. The BMC may also elect to change its inbound session sequence number at that time, or
may continue with the inbound session sequence number sequence already in progress.
213
Intelligent Platform Management Interface Specification
Table 18-17, Activate Session Command
byte
Session Request Data
IPMI Request Data
1
2
3:18
19:22
data field
authentication type = from corresponding Get Session Challenge command.
session seq# = null (0’s) when in ‘pre-session’ phase, non-null afterward. See
text.
session ID = Temporary Session ID value from corresponding Get Session
Challenge command, or present session ID if session already active
AuthCode = present unless authentication type = None. See 18.15.1,
AuthCode AlgorithmsAuthCode Algorithms for information on calculating
this field for authentication types that are not “None”.
Authentication Type for session. The selected type will be used for session
activation and for all subsequent authenticated packets under the
session, unless ‘Per-message Authentication’ or ‘User Level
Authentication’ are disabled. (See 6.11.4, Per-Message and User Level
Authentication DisablesPer-Message and User Level Authentication
Disables, for more information.)
[7:4] - reserved
[3:0] - Authentication Type. This value must match with the Authentication
Type used in the Get Session Challenge request for the session. In
addition, for multi-session channels this value must also match the
authentication type used in the Session Header.
0h = none. No hashing or authentication done on session packets.
Authentication Code field is not present.
1h = MD2
2h = MD5
3h = reserved
4h = straight password / key
5h = OEM proprietary
all other = reserved
Maximum privilege level requested. Indicates the highest privilege level that
may be requested for this session. This privilege level must be less than
or equal to the privilege limit for the channel and the privilege limit for the
user in order for the Activate Session command to be successful
(completion code = 00h). Once the Activate Session command has been
successful, the requested privilege level becomes a ‘session limit’ that
cannot be raised beyond the requested level, even if the user and/or
channel privilege level limits would allow it. I.e. it takes precedence over
the channel and user privilege level limits.
[7:4] - reserved
[3:0] - Requested Maximum Privilege Level
0h = reserved
1h = Callback level
2h = User level
3h = Operator level
4h = Administrator level
5h = OEM Proprietary level
all other = reserved
For multi-session channels: (e.g. LAN channel):
Challenge String data from corresponding Get Session Challenge
response.
For single-session channels that lack session header (e.g. serial/modem in
Basic Mode):
Clear text password or AuthCode. See 18.15.1, AuthCode
AlgorithmsAuthCode Algorithms.
Initial Outbound Sequence Number = Starting sequence number that remote
console wants used for messages from the BMC. (LS byte first). Must be
non-null in order to establish a session. 0000_0000h = reserved.
If the Activate Session command is executed after a session has been
established, the Outbound Sequence Number will be reset to the given
value. This will take effect for the corresponding Activate Session
response and subsequent commands under the session.
214
Intelligent Platform Management Interface Specification
215
Intelligent Platform Management Interface Specification
Session Response Data
IPMI Response Data
1
2
3:6
7:10
11
216
session ID = value from request
authentication type = value passed in from request data
session seq# = Initial outbound sequence number from corresponding
Activate Session request.
AuthCode = present unless authentication type = None. See 18.15.1,
AuthCode AlgorithmsAuthCode Algorithms for information on calculating
this field for authentication types that are not “None”.
Completion Code
00h = success
81h = No session slot available (BMC cannot accept any more sessions)
82h = No slot available for given user. (Limit of user sessions allowed under
that name has been reached)
83h = No slot available to support user due to maximum privilege capability.
(An implementation may only be able to support a certain number of
sessions based on what authentication resources are required. For
example, if User Level Authentication is disabled, an implementation
may be able to allow a larger number of users that are limited to User
Level privilege, than users that require higher privilege.)
84h = session sequence number out-of-range
85h = invalid session ID in request
86h = requested maximum privilege level exceeds user and/or channel
privilege limit
Authentication Type for remainder of session
The primary use of this parameter is to report whether per-message
authentication will be used for IPMI message packets that follow the
Activate Session packet. Per-message authentication is a channel
configuration option that is set using the Get User Name command. If
per-message authentication is disabled, the Authentication Type will be
returned as ‘none’, and all subsequent packets for the session are
required to can either use ‘none’ as the authentication type or use the
Authentication Type that was used in the request. Otherwise this value
will be set to the Authentication Type that was used in the request. Note
that Activate Session requests and responses are always required to be
authenticated per what is returned by the Get Session Challenge
command for the user.
[7:4] - reserved
[3:0] - Authentication Type
0h = none. No hashing or authentication done on session packets.
Authentication Code field is not present.
1h = MD2
2h = MD5
3h = reserved
4h = straight password / key
5h = OEM proprietary
all other = reserved
Session ID - use this for remainder of session. While atypical, the BMC is
allowed to change the session ID from the one that passed in the
request.
Initial inbound seq# = Sequence number that BMC wants remote console to
use for subsequent messages in the session. The BMC returns a nonnull value for multi-session connections and returns null (all 0’s) for
single-session connections.
Maximum privilege level allowed for this session
[7:4] - reserved
[3:0] - Maximum Privilege Level allowed
0h = reserved
1h = Callback level
2h = User level
3h = Operator level
4h = Administrator level
5h = OEM Proprietary level
all other = reserved
Intelligent Platform Management Interface Specification
18.15.1 AuthCode Algorithms
The following table lists the AuthCode calculation mechanism and field usage for the Activate Session
command, authenticated packets, and the Get AuthCode command. The Get AuthCode command uses the same
algorithm as
•
Refer to the [RFC1319] and [RFC1321] for information on the MD2 and MD5 algorithms, respectively.
•
For the following table, ‘+’ indicates concatenation of data, and H( ) represents the application of the
message digest algorithm to that data.
•
The data bytes are passed to the message-digest algorithm in the same order that they’re transmitted in the
message / packet.
•
The password/key is 0 padded to 16-bytes for all specified authentication types.
Figure 18-1, AuthCode Algorithms
Authentication
Type
straight
password
MD2
MD5
straight
password
MD2
MD5
straight
password
MD2
MD5
Algorithm
Single Session AuthCode carried in IPMI message data for Activate Session Command
AuthCode = password
AuthCode = H(password + temporary session ID + challenge string+ password)
AuthCode = H(password + temporary session ID + challenge string+ password)
Multi-Session AuthCode carried in session header for all ‘authenticated’ packets
AuthCode=password
[1]
AuthCode = H(password + session ID + IPMI Message data + session_seq# + password)
[1]
AuthCode = H(password + session ID + IPMI Message data + session_seq# + password)
Get AuthCode AuthCode carried in IPMI message data, per command description
See description of Get AuthCode command.
AuthCode = H(password + Get AuthCode data + password)
AuthCode = H(password + Get AuthCode data + password)
1. This will be the Temporary Session ID when calculating the AuthCode for the initial Activate Session command.
18.16 Set Session Privilege Level Command
This command is sent in authenticated format. When a session is activated, the session is set to an initial privilege
level. A session that is activated at a maximum privilege level of Callback is set to an initial privilege level of
Callback and cannot be changed. All other sessions are initially set to USER level, regardless of the maximum
privilege level requested in the Activate Session command. The remote console must ‘raise’ the privilege level of
the session using this command in order to execute commands that require a greater-than-User level of privilege.
This command cannot be used to set a privilege level higher than the lowest of the privilege level set for the user
(via the Set User Access command) and the privilege limit for the channel that was set via the Set Channel Access
command. Note that the specification allows a session to be used across multiple channels. The maximum
privilege limit and authentication are based on the user privilege and channel limits. Since these can vary on a per
channel basis, an implementation cannot simply assign a single privilege limit to a given session but must
authenticate incoming messages according to the specific settings for the channel and the user on a per-channel
basis.
217
Intelligent Platform Management Interface Specification
Table 18-18, Set Session Privilege Level Command
IPMI Request Data
1
IPMI Response Data
1
Requested Privilege Level
[7:4] - reserved
[3:0] - Privilege Level
0h - no change, just return present privilege level
1h - reserved
2h - change to USER level
3h - change to OPERATOR level
4h - change to ADMINISTRATOR level
5h - change to OEM Proprietary level
all other = reserved
Completion Code. Generic, plus following command specific:
80h = Requested level not available for this user
81h = Requested level exceeds Channel and/or User Privilege Limit
82h = Cannot disable User Level authentication
New Privilege Level (or present level if ‘return present privilege level’ was
selected.)
2
18.17 Close Session Command
This command is used to immediately terminate a session in progress. It is typically used to close the session that
the user is communicating over, though it can be used to other terminate sessions in progress (provided that the
user is operating at the appropriate privilege level, or the command is executed over a local channel - e.g. the
system interface).
Table 18-19, Close Session Command
Request Data
Response Data
byte
1:4
1
data field
Session ID
Completion Code
87h = invalid session ID in request
18.18 Get Session Info Command
This command is used to get information regarding which users presently have active sessions, and, when
available, addressing information for the party that has established the session. Note that a portion of the response
is dependent on the type of channel.
218
Intelligent Platform Management Interface Specification
Table 18-20, Get Session Info Command
IPMI Request Data
1
2
2:5
IPMI Response Data
1
2
3
4
5
6
7
8:11
12:17
18:19
8
9
10:13
14:15
Session Index:
This value is used to select entries in a logical ‘sessions’ table
maintained by the management controller. Info for all active sessions
can be retrieved by incrementing the session index from 1 to N, where
N is the number of entries in the Active Sessions table.
00h = Return info for active session associated with session this command
was received over.
N = get info for Nth active session
FEh = Look up session info according to Session Handle passed in this
request.
FFh = Look up session info according to Session ID passed in this request.
Present if Session Index = FEh:
Session Handle. 00h = reserved.
Present if Session Index = FFh:
Session ID. ID of session to look up session information for.
Completion Code
Session Handle presently assigned to active session. FFh = reserved. Return
00h if no active session associated with given session index.
Number of possible active sessions. This value reflects the number of
possible entries (slots) in the sessions table.
[7:6] - reserved
[5:0] - session slot count. 1-based.
Number of currently active sessions on all channels on this controller.
[7:6] - reserved
[5:0] - active session count. 1-based. 0 = no currently active sessions.
The following parameters are returned only if there is an active session
corresponding to the given session index:
User ID for selected active session
[7:6] - reserved.
[5:0] - User ID. 000000b = reserved.
Operating Privilege Level
[7:4] - reserved
[3:0] - present privilege level that user is operating at.
Channel that session was activated over.
[7:4] - reserved
[3:0] - channel number
The following bytes 8:18 are optionally returned if Channel Type = 802.3 LAN:
IP Address of remote console (MS-byte first). Address that was received in the
Activate Session command that activated the session.
MAC Address (MS-byte first). Address that was received in the Activate
Session command that activated the session.
Port Number of remote console (LS-byte first). Port Number that was received
in UDPUPD packet that held the Activate Session command that
activated the session.
The following bytes 8:13 are returned if Channel Type = asynch.
serial/modem:
Session / Channel Activity Type:
0 = IPMI Messaging session active
1 = Callback Messaging session active
2 = Dial-out Alert active
3 = TAP Page active
Destination Selector for active call-out session. 0 otherwise.
[7:4] - reserved
[3:0] - Destination selector. Destination 0 is always present as a volatile
destination that is used with the Alert Immediate command.
If PPP connection:
IP address of remote console. (MS-byte first) 00h, 00h, 00h, 00h otherwise.
The following additional bytes 14:15 are returned if Channel Type = asynch.
serial/modem and connection is PPP:
Port Address of remote console (LS-byte first). Address that was received in
the Activate Session command that activated the session.
219
Intelligent Platform Management Interface Specification
18.19 Get AuthCode Command
This command is used to send a block of data to the BMC, whereupon the BMC will return a hash of the data
together concatenated with the internally stored password for the given channel and user. This command allows a
remote console to send an AuthCode and data block to system software on a remote platform, whereby the system
software can validate the AuthCode by comparing it with the AuthCode returned by the BMC. This enables the
BMC to serve as a validation agent for remote requests that come through local system software instead of
through a remote session directly with the BMC.
The application of this command is beyond this specification. However, the following is an outline of potential
use of this capability. Remote console software could request that system software perform a particular operation.
In response, local system software could deliver a challenge string to the remote console, which would be
required to hash it with the desired password and return the AuthCode to the local system software. The local
system software would then perform the requested operation only if it found that the AuthCode matched the one
returned by the BMC. The local software would typically implement mechanisms to bind the challenge string to
the requested operation to ensure that the challenge string and AuthCode combination only applied to a given
instance of the requested operation, and even from a particular remote console.
•
Managed system delivers a random number token, S, to the Console. In this example, the Console uses S to
identify a particular request. The managed system tracks outstanding S values, and expires them either
because a valid message was received from a Console that used that token, or because the token was not
used within a specified interval.
•
Console determines: X = data to be authenticated
K1 = 16-byte ‘signature’ of X and a sequence number = hash(X, S, SW_Authentication_Type). Where
SW_Authentication_Type is any signature algorithm management software wishes to use for
providing a signature given X and S.
K2 = 16-byte hash of K1 and the password = hash(K1, PWD, Authentication_Type). Where
Authentication_Type in this case is one of the supported Authentication Types for the given
BMC. Figure 18-1, AuthCode AlgorithmsFigure 18-1, AuthCode Algorithms, specifies how the
“Get AuthCode Data” (K1) and password data (PWD) are concatenated for processing according
to Authentication_Type. Note that the hash algorithm for K1 does not need to be a BMC
supported algorithm or match the algorithm used for K2.
220
•
Console sends X, S, and K2 to software agent on managed system.
•
Software agent on the managed system calculates K1 from X and S that it received by locally calculating
K1=hash1(X, S, SW_Authentication_Type). The software also verifies that S is a valid outstanding token.
•
Managed system passes K1 to BMC. BMC internally looks up password based on the user ID passed in the
Get Authcode Command and produces: K2BMC = hash(K1, PWD, Authentication_Type)
•
Managed system accepts data if software agents finds that K2 = K2BMC.
Intelligent Platform Management Interface Specification
Table 18-21, Get AuthCode Command
IPMI Request Data
byte
1
2
3
4:19
IPMI Response Data
1
2:17
data field
Authentication Type
[7:4] - reserved
[3:0] - hash type
0h = reserved
1h = MD2
2h = MD5
3h = reserved
4h = straight password / key 5h = OEM proprietary
all other = reserved
Channel Number
[7:4] - reserved
[3:0] - Channel number
User ID. (software will typically have to use the Get User Info command to
look up the User ID from a username)
[7:6] - reserved
[5:0] - User ID
hash data (must be 16 bytes), or zero padded clear text password if the
Authentication Type = straight password
Completion Code
If authentication type = straight password, BMC returns ‘OK’ if password was
correct for specified user, or error completion code if it is not.
AuthCode = See 18.15.1, AuthCode AlgorithmsAuthCode Algorithms.
221
Intelligent Platform Management Interface Specification
18.20 Set Channel Access Command
This command is used to configure whether channels are enabled or disabled, whether alerting is enabled or
disabled for a channel, and to set which system modes channels are available under. This configuration is saved
in non-volatile storage associated with the BMC. The choice of factory default setting for the non-volatile
parameters is left to the implementer or system integrator.
The active (volatile) settings can be overwritten to allow run-time software to make temporary changes to the
access. The volatile settings are overwritten from the non-volatile settings whenever the system is reset or
transitions to a powered off state.
An implementation can elect to provide a subset of the possible Access Mode options. If a given Access Mode
is not supported, the command-specific completion code 83h, access mode not supported, must be returned.
Table 18-22, Set Channel Access Command
byte
Request Data
1
2
data field
[7:4] - reserved
[3:0] - Channel number
[7:6] - 00b = don’t set or change Channel Access
01b = set non-volatile Channel Access according to bits [5:0]
10b = set volatile (active) setting of Channel Access according to bits
[5:0]
11b = reserved
[5] - PEF Alerting Enable/Disable
0b = enable PEF Alerting
1b = disable PEF Alerting on this channel (the Alert Immediate
command can still be used to generate alerts)
[4] - 0b = enable Per-message Authentication
1b = disable Per-message Authentication. [Authentication required to
activate any session on this channel, but authentication not used
on subsequent packets for the session.]
[3] - User Level Authentication Enable/Disable.
Optional. Return a CCh ‘invalid data field’ error completion code if an
attempt is made to set this bit, but the option is not supported.
0b = enable User Level Authentication. All User Level commands are
to be authenticated per the Authentication Type that was
negotiated when the session was activated.
1b = disable User Level Authentication. Allow User Level commands to
be executed without being authenticated.
If the option to disable User Level Command authentication is
accepted, the BMC will accept packets with Authentication Type
set to None if they contain user level commands.
For outgoing packets, the BMC returns responses with the same
Authentication Type that was used for the request.
[2:0] - Access Mode for IPMI messaging (PEF Alerting is enabled/disabled
separately from IPMI messaging, see bit 5)
000b = disabled
channel disabled for IPMI messaging
001b = pre-boot only
channel only available when system is in a powered down state or
in BIOS prior to start of boot.
010b = always available
channel always available for communication regardless of system
mode. BIOS typically dedicates the serial connection to the BMC.
011b = shared
same as always available, but BIOS typically leaves the serial port
available for software use.
222
Intelligent Platform Management Interface Specification
3
Channel Privilege Level Limit. This value sets the maximum privilege level
that can be accepted on the specified channel.
[7:6] - 00b = don’t set or change channel Privilege Level Limit
01b = set non-volatile Privilege Level Limit according to bits [3:0]
10b = set volatile setting of Privilege Level Limit according to bits [3:0]
11b = reserved
[5:4] - reserved
Response Data
1
[3:0] - Channel Privilege Level Limit
0h = reserved
1h = CALLBACK level
2h = USER level
3h = OPERATOR level
4h = ADMINISTRATOR level
5h = OEM Proprietary level
Completion Code
generic, plus following command-specific completion codes:
82h = set not supported on selected channel (e.g. channel is sessionless.)
83h = access mode not supported
223
Intelligent Platform Management Interface Specification
18.21 Get Channel Access Command
This command is used to return whether a given channel is enabled or disabled, whether alerting is enabled or
disabled for the entire channel, and under what system modes the channel can be accessed.
Table 18-23, Get Channel Access Command
byte
Request Data
1
2
Response Data
1
2
224
data field
[7:4] - reserved
[3:0] - Channel number.
[7:6] - 00b = reserved
01b = get non-volatile Channel Access
10b = get present volatile (active) setting of Channel Access
11b = reserved
[5:0] - reserved
Completion Code
generic, plus following command-specific completion codes:
82h = Command not supported for selected channel (e.g. channel is
session-less.)
[7:6] - reserved
[5] - 0b = Alerting enabled
1b = Alerting disabled
[4] - 0b = per message authentication enabled
1b = per message authentication disabled
[3] - User Level Authentication Enable
0b = User Level Authentication enabled.
1b = User Level Authentication disabled.
[2:0] - Access Mode
0h = disabled
channel disabled for communication
1h = pre-boot only
channel only available when system is in a powered down state or
in BIOS prior to start of boot.
2h = always available
channel always available for communication regardless of system
mode. BIOS typically dedicates the serial connection to the BMC.
3h = shared
same as always available, but BIOS typically leaves the serial port
available for software use.
Intelligent Platform Management Interface Specification
3
Channel Privilege Level Limit. This value returns the maximum privilege level
that can be accepted on the specified channel.
[7:4] - reserved
[3:0] - Channel Privilege Level Limit
0h = reserved
1h = CALLBACK level
2h = USER level
3h = OPERATOR level
4h = ADMINISTRATOR level
5h = OEM Proprietary level
18.22 Get Channel Info Command
This command returns media and protocol information about the given channel. The channel protocol may vary
with changes to the configuration parameters associated with the channel.
Table 18-24, Get Channel Info Command
IPMI Request Data
1
IPMI Response Data
1
2
3
4
5
6:8
[7:4] - reserved
[3:0] - channel number. Use Eh to get information about the channel this
command is being executed from.
Completion Code
[7:4] - reserved
[3:0] - actual channel number. This value will typically match the channel
number passed in the request, unless the request is for channel E, in
which case the response returns the actual channel number.
[7] reserved
[6:0] - 7-bit Channel Medium type: per Table 6-3, Channel Medium Type
NumbersTable 6-3, Channel Medium Type Numbers
Channel Protocol Type:
[7:5] - reserved
[4:0] - 5-bit Channel IPMI Messaging Protocol Type per Table 6-2, Channel
Protocol Type NumbersTable 6-2, Channel Protocol Type Numbers
Session support
[7:6] - 00b = channel is session-less
01b = channel is single-session
10b = channel is multi-session
11b = channel is session-based (return this value if a channel could
alternate between single- and multi-session operation, as can
occur with a serial/modem channel that supports connection
mode auto-detect)
Number of sessions that have been activated on given channel.
[5:0] - active session count. 1-based.
00_0000b = no sessions have been activated on this channel.
Vendor ID (IANA Enterprise Number) for OEM/Organization that specified the
Channel Protocol.
Least significant byte first.
Returns the IPMI IANA for IPMI-specification defined, non-OEM protocol type
numbers other than OEM.
The IPMI Enterprise Number is: 7154 (decimal).
This gives the values F2h, 1Bh, 00h for bytes 6 through 8, respectively. This
value is returned for all channel protocols specified in this document, including
PPP.
225
Intelligent Platform Management Interface Specification
9:10
Auxiliary Channel Info
For Channel = Fh (System Interface) :
byte 1: SMS Interrupt Type
00h-0Fh = IRQ 0 through 15, respectively
10h-13h = PCI A-D, respectively
14h = SMI
15h = SCI
20h-5Fh = system interrupt 0 through 63, respectively
60h = assigned by ACPI / Plug ‘n Play BIOS
FFh = no interrupt / unspecified
all other = reserved
byte 2: Event Message Buffer Interrupt Type
see values for byte 1
For OEM channel types:
byte 1:2 = OEM specified per OEM identified by Vendor ID field.
All other channel types:
byte 1:2 = reserved.
18.23 Set User Access Command
This command is used to configure the privilege level and channel accessibility associated with a given user ID.
If this command is not supported, then a single ‘null user’ (User 1) per channel is assumed and the privilege
level and channel access are determined solely by the settings returned by the Get Channel Access Limits
command. If implemented, this command must support at least the null user (User 1). The number of additional
users supported is left to the implementer.
Note: The limits set using the Set Channel Access command take precedence over the Set User Access
command settings. That is, if a given channel is limited to User level then all users will be limited to User level
operation regardless of what their User Access levels were set to using the Set User Access command.
Note that changes made to the user access and privilege levels may not take affect until the next time the user
establishes a session.
226
Intelligent Platform Management Interface Specification
Table 18-25, Set User Access Command
byte
Request Data
1
2
3
(4)
Response Data
1
data field
[7] -
0b = do not change any of the following bits in this byte
1b = enable changing the following bits in this byte
[6] -
User Restricted to Callback
0b = User Privilege Limit is determined by the User Privilege Limit
parameter, below, for both callback and non-callback connections.
1b = User Privilege Limit is determined by the User Privilege Limit
parameter for callback connections, but is restricted to Callback
level for non-callback connections. Thus, a user can only initiate a
Callback when they ‘call in’ to the BMC, but once the callback
connection has been made, the user could potentially establish a
session as an Operator.
[5] -
User Link authentication enable/disable (used to enable whether this
user’s name and password information will be used for link
authentication, e.g. PPP CHAP) for the given channel. Link
authentication itself is a global setting for the channel and is
enabled/disabled via the serial/modem configuration parameters.
0b = disable user for link authentication
1b = enable user for link authentication
[4] -
User IPMI Messaging enable/disable (used to enable/disable whether
this user’s name and password information will be used for IPMI
Messaging)
0b = disable user for IPMI Messaging
1b = enable user for IPMI Messaging
[3:0] - Channel Number
User ID
[7:6] - reserved.
[5:0] - User ID. 000000b = reserved.
User Limits
[7:4] - reserved
[3:0] - User Privilege Limit. (Determines the maximum privilege level that the
user is allowed to switch to on the specified channel.)
0h = reserved
1h = Callback
2h = User
3h = Operator
4h = Administrator
5h = OEM Proprietary
Fh = NO ACCESS
User Session Limit. (Optional) Sets how many simultaneous sessions can be
activated with the username associated with this user. If not supported, the
username can be used to activate as many simultaneous sessions as the
implementation supports.
Return a CCh ‘invalid data field’ error completion code if an attempt is made to
set a non-zero value in this field, but the option is not supported.
[7:4] - reserved
[3:0] - User simultaneous session limit. 1-based. 0h = only limited by the
implementations overall support for simultaneous sessions.
Completion Code.
Note: an implementation will not return an error completion code if the user
access level is set higher than the privilege limit for a given channel. If it is
desired to bring attention to this condition, it is up to software to check the
channel privilege limits set using the Set Channel Access command and
provide notification of any mismatch.
227
Intelligent Platform Management Interface Specification
18.24 Get User Access Command
This command is used to retrieve channel access information and enabled/disabled state for the given User ID.
The command also returns information about the number of supported users.
Table 18-26, Get User Access Command
byte
Request Data
1
2
Response Data
1
2
3
4
5
data field
[7:4] - reserved
[3:0] - Channel Number
[7:6] - reserved
[5:0] - User ID. 000000b = reserved.
Completion Code.
Note: an implementation will not return an error completion code if the user
access level is set higher than the privilege limit for a given channel. If it is
desired to bring attention to this condition, it is up to software to check the
channel privilege limits and provide notification of the mis-match.
Maximum number of User IDs. 1-based. Count includes User 1. A value of 1
indicates only User 1 is supported.
[7:6] - reserved
[5:0] - maximum number of user IDs on this channel
Count of currently enabled User IDs (1-based). A value of 0 indicates that all
users, including User 1, are disabled. This is equivalent to disabling access to
the channel.
[7:6] - reserved
[5:0] - count of currently enabled user IDs on this channel (Indicates how
many User ID slots are presently in use.)
Count of User IDs with fixed names, including User 1(1-based). Fixed names
in addition to User 1 are required to be associated with sequential user IDs
starting from User ID 2.
[7:6] - reserved.
[5:0] - count of user IDs with fixed names on this channel
Channel Access
[7] - reserved.
[6] - 0b = user access available during call-in or callback direct connection
1b = user access available only during callback connection
bits 5:4, following, are used for determining the ‘count of currently enabled
user IDs’ in byte 3. Either bit being set to 1b represents an ‘enabled user ID’.
[5] - 0b = user disabled for link authentication
1b = user enabled for link authentication
[4] - 0b = user disabled for IPMI Messaging
1b = user enabled for IPMI Messaging
[3:0] - User Privilege Limit for given Channel
0h = reserved
1h = Callback
2h = User
3h = Operator
4h = Administrator
5h = OEM Proprietary
Fh = NO ACCESS (Note: this value does not add to, or subtract from,
the number of enabled user IDs)
228
Intelligent Platform Management Interface Specification
18.25 Set User Name Command
This command allows user names to be assigned to a given User ID. The names are stored as a logical array
within non-volatile storage associated with the management controller. Names are stored and retrieved using
the User ID as the index into the logical array. There is no configurable name for User ID 1. User ID 1 is
reserved for the null user name, User 1.
The management controller does not prevent duplicate usernames from being enabled for the same channel. It
is the responsibility of configuration software to ensure that duplicate user names are not enabled
simultaneously for the same channel.
Having duplicate usernames will not cause functional problems with the BMC because the BMC will just use
the first username match that it finds. However, it could be confusing to the user if they have duplicate
usernames enabled for a given channel, since only the settings for the first encountered username would be
used by the BMC. See 6.9, Users & Password SupportUsers & Password Support for more information.
Table 18-27, Set User Name Command
byte
Request Data
1
2:17
Response Data
1
data field
User ID
[7:6] - reserved.
[5:0] - User ID. 000000b = reserved. (User ID 1 is permanently associated
with User 1, the null user name).
User Name String in ASCII, 16 bytes, max. Strings with fewer than 16
characters are terminated with a null (00h) character and 00h padded to 16
bytes. When the string is read back using the Get User Name command,
those bytes shall be returned as 0’s.
Completion Code.
18.26 Get User Name Command
This command is used to retrieve user name information that was set using the Set User Name command.
Configuration software can use this command to retrieve user names.
Table 18-28, Get User Name Command
byte
Request Data
Response Data
1
1
2:17
data field
User ID
[7:6] - reserved.
[5:0] - User ID to return name for. 000000b = reserved.
Completion Code.
User Name String in ASCII, 16 bytes, max. Strings of fewer than 16
characters are returned with null (00h) characters filling in the remaining
bytes. BMC does not check to see whether string data is ‘printable’ or not.
Only character that BMC interprets is null (00h).
229
Intelligent Platform Management Interface Specification
18.27 Set User Password Command
This command is used to set and change user passwords and to enable and disable User IDs. If no password
protection is desired for a given user, the password must be stored as an ASCII null-string. The management
controller firmware will force the remaining fifteen bytes to 00h and store the password as sixteen bytes of 00h.
If this command is not supported, then the implication is that only User 1 with a fixed, null password is
supported.
The password is stored as a 16-byte ‘octet string’. All values (0-255) are allowed for every byte. The
management controller does not check the format or interpret values that are passed in with the Set User
Password command.
Software is allowed to place additional restrictions on what passwords can be entered, in which case it is the
responsibility of configuration software and console software to stay in synch with that definition. For example,
remote console software could restrict passwords to the printable ASCII character set in order to simplify direct
keyboard entry. If this were done, a companion configuration utility should ensure that the user does not
configure the managed system with non-printable passwords. Otherwise, it would be possible for the
management controller to be configured with passwords that could not be entered via the remote console utility.
Table 18-29, Set User Password Command
byte
Request Data
1
2
3:18
Response Data
230
1
data field
User ID.
[7:6] - reserved.
[5:0] - User ID. 000000b = reserved. (User ID 1 is permanently associated
with User 1, the null user name).
[7:2] - reserved
[1:0] - operation
00b = disable user
01b = enable user
10b = set password
11b = test password
Password data. This is a required, fixed length field when used for the setand test password operations. If the password is entered as an ASCII string, it
must be null (00h) terminated and 00h padded if the string is shorter than 16
bytes. This field need not be present if the operation is ‘disable user’ or
‘enable user’. If this field is present for those operations, the BMC will ignore
the data.
Completion Code. Generic, plus following command-specific completion
codes:
80h = password test failed (mandatory)
Intelligent Platform Management Interface Specification
19. IPMI LAN Commands
This section defines the configuration and control commands that are specific to LAN channels. None of the
commands in the following table are required unless a LAN channel is implemented. Refer to Appendix G Command Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 19-1, IPMI LAN Commands
Command
Set LAN Configuration Parameters
Get LAN Configuration Parameters
Suspend BMC ARPs
Get IP/UDP/RMCP Statistics
1.
2.
Section
Defined
O/M
19.1
19.2
19.3
19.4
M
[1]
M
[2]
O
O
[1]
Mandatory if LAN channel is supported.
Mandatory if BMC autonomously generates Gratuitous ARPs
19.1 Set LAN Configuration Parameters Command
This command is used for setting parameters such as the network addressing information required for IPMI LAN
operation.
Table 19-2, Set LAN Configuration Parameters Command
byte
Request Data
1
2
3:N
Response Data
1
data field
[7:4] - reserved
[3:0] - Channel number.
Parameter selector
Configuration parameter data, per Table 19-4, LAN Configuration
ParametersTable 19-4, LAN Configuration Parameters
Completion Code
80h = parameter not supported.
81h = attempt to set the ‘set in progress’ value (in parameter #0) when not in
the ‘set complete’ state. (This completion code provides a way to
recognize that another party has already ‘claimed’ the parameters)
82h = attempt to write read-only parameter
231
Intelligent Platform Management Interface Specification
19.2 Get LAN Configuration Parameters Command
This command is used for retrieving the configuration parameters from the Set LAN Configuration command.
Table 19-3, Get LAN Configuration Parameters Command
byte
Request Data
1
2
3
4
Response Data
1
2
3:N
232
data field
[7] -
0b = get parameter
1b = get parameter revision only.
[6:4] - reserved
[3:0] - Channel number.
Parameter selector
Set Selector. Selects a given set of parameters under a given Parameter
selector value. 00h if parameter doesn’t use a Set Selector.
Block Selector (00h if parameter does not require a block number)
Completion Code.
Generic codes, plus following command-specific completion code(s):
80h = parameter not supported.
[7:0] - Parameter revision.
Format: MSN = present revision. LSN = oldest revision parameter is backward
compatible with. 11h for parameters in this specification.
The following data bytes are not returned when the ‘get parameter revision
only’ bit is 1b.
Configuration parameter data, per Table 19-4, LAN Configuration
ParametersTable 19-4, LAN Configuration Parameters
If the rollback feature is implemented, the BMC makes a copy of the existing
parameters when the ‘set in progress’ state becomes asserted (See the Set In
Progress parameter #0). While the ‘set in progress’ state is active, the BMC
will return data from this copy of the parameters, plus any uncommitted
changes that were made to the data. Otherwise, the BMC returns parameter
data from non-volatile storage.
Intelligent Platform Management Interface Specification
Table 19-4, LAN Configuration Parameters
[1]
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
Set In Progress
(volatile)
0
Authentication Type
Support (Read Only)
1
data 1 - This parameter is used to indicate when any of the following parameters are being
updated, and when the updates are completed. The bit is primarily provided to alert software
than some other software or utility is in the process of making changes to the data.
An implementation can also elect to provide a ‘rollback’ feature that uses this information to
decide whether to ‘roll back’ to the previous configuration information, or to accept the
configuration change.
If used, the roll back shall restore all parameters to their previous state. Otherwise, the
change shall take effect when the write occurs.
[7:2] - reserved
[1:0] - 00b = set complete. If a system reset or transition to powered down state occurs
while ‘set in progress’ is active, the BMC will go to the ‘set complete’ state. If
rollback is implemented, going directly to ‘set complete’ without first doing a
‘commit write’ will cause any pending write data to be discarded.
01b = set in progress. This flag indicates that some utility or other software is
presently doing writes to parameter data. It is a notification flag only, it is not a
resource lock. The BMC does not provide any interlock mechanism that would
prevent other software from writing parameter data while.
10b = commit write (optional). This is only used if a rollback is implemented. The
BMC will save the data that has been written since the last time the ‘set in
progress’ and then go to the ‘set in progress’ state. An error completion code
will be returned if this option is not supported.
11b = reserved
This ‘read only’ field returns which possible Authentication Types (algorithms) can be
enabled for the given channel. The following Authentication Type Enables parameter selects
which Authentication Types are available when activating a session for a particular maximum
privilege level.
[7:6] - reserved
[5:0] - Authentication type(s) enabled for this channel (bitfield):
All bits:
1b = supported
0b = authentication type not available for use.
[5] - OEM proprietary (per OEM identified by the IANA OEM ID in the RMCP Ping
Response)
[4] - straight password / key
[3] - reserved
[2] - MD5
[1] - MD2
[0] - none
233
Intelligent Platform Management Interface Specification
[1]
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
Authentication Type
Enables
2
This field is used to configure which Authentication Types are available for use when a
remote console activates an IPMI messaging connection to the BMC for a given requested
maximum privilege level. Once the session has been activated, the accepted authentication
type will be the only one used for authenticated packets, regardless of the present operating
privilege level, or the privilege level associated with the command.
Depending on configuration of per-message and user-level authentication disables,
unauthenticated packets (authentication type = none) may also be accepted. The BMC
makes no attempt to check or ensure that stricter authentication types are associated with
higher requested maximum privilege levels. E.g. it is possible to configure the BMC so
activating a session with a maximum privilege level of ‘User’ requires MD5 while ‘Admin’
requires ‘none’.
Note: An implementation that has fixed privilege and authentication type assignments, in
which case this parameter can be implemented as Read Only. It is recommended that an
implementation that implements a subset of the possible authentication types returns a CCh
error completion code if an attempt is made to select an unsupported authentication type.
byte 1: Authentication Types returned for maximum requested privilege = Callback level.
[7:6] - reserved
[5:0] - Authentication type(s) enabled for this channel (bitfield):
All bits:
1b = authentication type enabled for use at given privilege level
0b = authentication type not available for use at given privilege level.
[5] - OEM proprietary (per OEM identified by the IANA OEM ID in the RMCP Ping
Response)
[4] - straight password / key
[3] - reserved
[2] - MD5
[1] - MD2
[0] - none
byte 2: Authentication Type(s) for maximum privilege = User level
(format follows byte 1)
byte 3: Authentication Type (s) for maximum privilege = Operator level
(format follows byte 1)
byte 4: Authentication Type (s) for maximum privilege = Administrator level
(format follows byte 1)
IP Address
3
IP Address Source
4
MAC Address
5
Subnet Mask
6
234
byte 5: Authentication Type (s) for maximum privilege = OEM level
(format follows byte 1)
data 1:4 - IP Address
MS-byte first.
data 1
[7:4] - reserved
[3:0] - address source
0h = unspecified
1h = static address (manually configured)
2h = address obtained by BMC running DHCP
3h = address loaded by BIOS or system software
4h = address obtained by BMC running other address assignment protocol
data 1:6 - MAC Address for messages transmitted from BMC.
MS-byte first.
data 1:4 - Subnet Mask. MS-byte first.
Intelligent Platform Management Interface Specification
[1]
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
IPv4 Header Parameters
7
data 1 - Time-to-live. 1-based. (Default = 40h)
Value for time-to-live parameter in IP Header for RMCP packets and PET Traps
transmitted from this channel.
data 2
[7:5] - Flags. Sets value of bit 1 in the Flags field in the IP Header for packets transmitted
by this channel. (Default = 010b “don’t fragment”)
[4:0] - reserved
Primary RMCP Port
Number (optional)
Secondary RMCP Port
Number (optional)
BMC-generated ARP
control
[2]
(optional )
8
9
10
Gratuitous ARP
interval (optional)
11
Default Gateway
Address
12
Default Gateway MAC
Address
Backup Gateway
Address
13
Backup Gateway MAC
Address
Community String
15
Number of Destinations
(Read Only)
14
16
17
data 3
[7:5] - Precedence (Default = 000b)
[4:1] - Type of Service (Default = 1000b, “minimize delay”)
[0] reserved
data 1:2 - Primary RMCP Port Number, LSByte first.
Default = 26Fh (RMCP ‘Aux Bus Shunt’ port)
data 1:2 - Secondary Port Number, LSByte first.
Default = 298h (RMCP ‘Secure Aux Bus’ port)
data 1 - BMC-generated ARP control. Note: the individual capabilities for BMC-generated
ARP responses and BMC-generated Gratuitous ARPs are individually optional. The BMC
should return an error completion code if an attempt is made to enable an unsupported
capability.
[7:2] - reserved
[1] 1b = enable BMC-generated ARP responses
0b = disable BMC-generated ARP responses
[0] 1b = enable BMC-generated Gratuitous ARPs
0b = disable BMC-generated Gratuitous ARPs
data 1 - Gratuitous ARP interval
Gratuitous ARP interval in 500 millisecond increments. 0-based.
Interval accuracy is +/- 10%.
If this configuration parameter is not implemented, gratuitous ARPs shall be issued at a
rate of once every 2 seconds.
data 1:4 - IP Address
MS-byte first. This is the address of the gateway (router) used when the BMC sends a
message or alert to a party on a different subnet than the one the BMC is on.
data 1:6 - MAC Address. MS-byte first.
data 1:4 - IP Address
MS-byte first. This is the address of an alternate gateway (router) that can be selected
when a sending a LAN Alert.
data 1:6 - MAC Address. MS-byte first.
data 1:18 - Community String
Default = ‘public’. Used to fill in the ‘Community String’ field in a PET Trap. This string may
optionally be used to hold a vendor-specific string that is used to provide the network name
identity of the system that generated the event. Printable ASCII string-. If a full 18 non-null
characters are provided, the last character does not need to be a null. 18 characters must be
written when setting this parameter, and 18 will be returned when this parameter is read. The
null character, and any following characters, will be ignored when the Community String
parameter is placed into the PET. The BMC will return whatever characters were written. I.e.
it will not set bytes following the null to any particular value.
data 1 - Number of LAN Alert Destinations supported on this channel. (Read Only). At least
one set of non-volatile destination information is required if LAN alerting is supported.
Additional non-volatile destination parameters can optionally be provided for supporting an
alert ‘call down’ list policy. A maximum of fifteen (1h to Fh) non-volatile destinations are
supported in this specification. Destination 0 is always present as a volatile destination that is
used with the Alert ImmediateAlert Immediate command.
[7:4] - reserved.
[3:0] - Number LAN Destinations. A count of 0h indicates LAN Alerting is not supported.
235
Intelligent Platform Management Interface Specification
Parameter
Destination Type
(volatile / non-volatile see description)
#
18
Parameter Data (non-volatile unless otherwise noted)
[1]
Sets the type of LAN Alert associated with the given destination. This parameter is not
present if the Number of Destinations parameter is 0.
data 1 - Set Selector = Destination selector, 0 based.
[7:4] - reserved
[3:0] - Destination selector. Destination 0 is always present as a volatile destination that is
used with the Alert ImmediateAlert Immediate command.
data 2 - Destination Type
[7] Alert Acknowledge.
0b = Unacknowledged. Alert is assumed successful if transmission occurs
without error. This value is also used with Callback numbers.
1b = Acknowledged. Alert is assumed successful only if acknowledged is
returned. Note, some alert types, such as Dial Page, do not support an
acknowledge.
[6:3] - reserved
[2:0] - Destination Type
000b
= PET Trap destination
001b - 101b = reserved
110b
= OEM 1
111b
= OEM 2
data 3 - Alert Acknowledge Timeout / Retry Interval, in seconds, 0-based (i.e. minimum
timeout = 1 second).
This value sets the timeout waiting for an acknowledge, or the time between
automatic retries depending on whether the alert is acknowledge or not.
Recommended factory default = 3 seconds. Value is ignored if alert type does not
support acknowledge, or if the Alert Acknowledge bit (above) is 0b.
Destination Addresses
19
data 4 - Retries
[7:4] - reserved
[3] reserved
[2:0] - Number of times to retry alert to given destination. 0 = no retries (alert is only sent
once). If the alert is acknowledged (Alert Acknowlege bit = 1b) the alert will only be
retried if a timeout occurs waiting for the acknowledge. Otherwise, this value selects
the number of times an unacknowledged alert will be sent out. The timeout interval
or time between retries is set by the Alert Acknowledge Timeout / Retry Interval
value (byte 3 of this parameter).
Sets/Gets the list of IP addresses that a LAN alert can be sent to. This parameter is not
present if the Number of Destinations parameter is 0.
data 1 - Set Selector = Destination Selector.
[7:4] - reserved
[3:0] - Destination selector. Destination 0 is always present as a volatile destination that is
used with the Alert ImmediateAlert Immediate command.
data 2 - Address Format
[7:4] Address Format.
0h = IPv4 IP Address followed by DIX Ethernet/802.3 MAC Address
[3:0] reserved
For Address Format = 0h:
data 3 - Gateway selector
[7:1] reserved
[0] 0b = use default gateway
1b = use backup gateway
OEM Parameters
192
:
255
data 4:7 - Alerting IP Address (MS-byte first)
data 8:13 - Alerting MAC Address (MS-byte first)
This range is available for special OEM configuration parameters. The OEM is identified
according to the Manufacturer ID field returned by the Get Device ID command.
1. Choice of system manufacturing defaults is left to the system manufacturer unless otherwise specified.
2. This configuration parameter must be supported if the controller autonomously issues gratuitous ARPs or ARP responses.
236
Intelligent Platform Management Interface Specification
19.3 Suspend BMC ARPs Command
This command can be used to suspend BMC-generated Gratuitous ARPs or ARP responses (if implemented and
enabled) while run-time software is handling them. ARPs will automatically resume and the ARP suspend option
will be automatically disabled whenever the watchdog timer stops or when the Set Watchdog Timer command is
executed. Software must explicitly enable the suspension after starting or re-starting the Watchdog Timer. This is
to ensure that BMC-generated ARPs will resume if the watchdog expires (indicating that software may no longer
be handling the ARPs).
Software can examine the LAN Configuration Parameters for the desired channel to see if Gratuitous ARPs or
BMC-generated ARP Responses are enabled.
Table 19-5, Suspend BMC ARPs Command
byte
Request Data
1
2
data field
[7:4] reserved
[3:0] Channel number
[7:2] - reserved
ARP Response suspend
[1] - 0b = do not suspend BMC-generated ARP responses while the
Watchdog Timer is running.
1b = suspend BMC-generated ARP responses while Watchdog Timer
is running. This value must be set after the Watchdog has been
configured using the Set Watchdog Timer command and, if it is
desired that the suspension of ARPs be continued, must be rewritten after any subsequent Set Watchdog Timer commands.
Response Data
1
2
Gratuitous ARP suspend
[0] - 0b = do not suspend BMC-generated Gratuitous ARPs while the
Watchdog Timer is running.
1b = suspend BMC-generated Gratuitous ARPs while the Watchdog
Timer is running. This value must be set after the Watchdog has
been configured using the Set Watchdog Timer command and, if it
is desired that the suspension of ARPs be continued, must be rewritten after any subsequent Set Watchdog Timer commands.
Completion Code
Present state of ARP suspension:
[7:2] - reserved
ARP Response status
[1] - 1b = BMC-generated ARP Responses are occurring
0b = BMC-generated ARP Responses are presently being suspended
or are disabled.
Gratuitous ARP Response status
[0] - 1b = BMC-generated Gratuitous ARPs are occurring
0b = BMC-generated Gratuitous ARPs are presently being suspended
or are disabled.
237
Intelligent Platform Management Interface Specification
19.4 Get IP/UDP/RMCP Statistics Command
This command is used retrieving information about the IP connections on the given channel. The info is
cumulative but volatile. I.e. it is not required to keep these statistics across management controller power cycles.
It is recommended that this information be kept across system resets and power cycles. The statistics values are
initialized to 0 unless otherwise noted.
Table 19-6, Get IP/UDP/RMCP Statistics Command
byte
Request Data
1
2
Response Data
1
2:3
4:5
6:7
8:9
10:11
12:13
14:15
16:17
18:19
238
data field
[7:4] reserved
[3:0] Channel number.
Clear Statistics
[7:1]- reserved
[0] - 0b = don’t clear statistics
1b = clear all statistics values to 0
Completion Code
IP Packets Received. All statistics returned by this command are 1-based
unless otherwise noted. All statistics stop accumulating at FFFFh unless
otherwise noted.
Received IP Header Errors
Received IP Address Errors
Fragmented IP Packets Received
IP Packets Transmitted
UDP Packets Received
Valid RMCP Packets Received
UDP Proxy Packets Received
UDP Proxy Packets dropped
Intelligent Platform Management Interface Specification
20. IPMI Serial/Modem Commands
This section defines the configuration and control commands that are specific to serial/modem channels. None of
the commands in the following table are required unless a serial/modem channel is implemented. Refer to
Appendix G - Command Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 20-1, IPMI Serial/Modem Commands
Command
Set Serial/Modem Configuration
Get Serial/Modem Configuration
Set Serial/Modem Mux
Get TAP Response Codes
Set PPP UDP Proxy Transmit Data
Get PPP UDP Proxy Transmit Data
Send PPP UDP Proxy Packet
Get PPP UDP Proxy Receive Data
Serial/Modem Connection Active
Callback
Set User Callback Options
Get User Callback Options
1.
2.
3.
4.
5.
Section
Defined
O/M
20.1
20.2
20.3
20.4
20.5
20.6
20.7
20.8
20.9
20.10
20.11
20.12
M
[1]
M
[2]
O
[3]
O
[4]
O
[4]
O
[4]
O
[4]
O
[1]
M
O
[5]
O
[5]
O
[1]
Mandatory if serial/modem channel(s) supported.
Mandatory if Serial Port Sharing is supported.
Mandatory if TAP Paging is supported. If TAP Paging is supported it is
recommended, but not mandatory, that it be supported on all serial/modem
channels that could support a modem connection. (Some serial/modem channels
may never be connected to a modem)
Mandatory if PPP UDP Proxy capability is supported.
Mandatory if IPMI Callback is supported. Note that CBCP callback support is
optional. Whether CBCP is supported or not is determined from the serial/modem
configuration parameters.
20.1 Set Serial/Modem Configuration Command
This command is used for setting parameters such as the string used for initializing the modem, communication
bit rates, and selecting configuration options such as Direct Connect versus Modem Connect.
239
Intelligent Platform Management Interface Specification
Table 20-2, Set Serial/Modem Configuration Command
byte
Request Data
1
2
3:N
Response Data
1
data field
[7:4] - reserved
[3:0] - Channel number.
Parameter selector
Configuration parameter data, per Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
81h = attempt to set the ‘set in progress’ value (in parameter #0) when not in
the ‘set complete’ state. (This completion code provides a way to
recognize that another party has already ‘claimed’ the parameters)
82h = attempt to write read-only parameter
83h = attempt to read write-only parameter
20.2 Get Serial/Modem Configuration Command
This command is used for retrieving the configuration parameters from the Set Serial/Modem Configuration
command.
Table 20-3, Get Serial/Modem Configuration Command
byte
Request Data
1
2
3
4
Response Data
1
2
3:N
240
data field
[7] -
0b = get parameter
1b = get parameter revision only
[6:4] - reserved
[3:0] - Channel number.
Parameter selector
Set Selector. Selects a particular set or block data under the given parameter
selector. 00h if parameter does not use a set selector.
Block Selector (00h if parameter does not require a block number)
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
[7:0] - Parameter revision.
Format: MSN = present revision. LSN = oldest revision parameter is
backward compatible with. 11h for parameters in this specification.
The following data bytes are not returned when the ‘get parameter revision
only’ bit is 1b.
Configuration parameter data, per Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters
If the rollback feature is implemented, the BMC makes a copy of the existing
parameters when the ‘set in progress’ state becomes asserted (See the Set In
Progress parameter #0). While the ‘set in progress’ state is active, the BMC
will return data from this copy of the parameters, plus any uncommitted
changes that were made to the data. Otherwise, the BMC returns parameter
data from non-volatile storage.
Intelligent Platform Management Interface Specification
Table 20-4, Serial/Modem Configuration Parameters
[1]
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
Set In Progress
(volatile)
0
Authentication Type
Support (Read Only)
1
data 1 - This parameter is used to indicate when any of the following parameters
are being updated, and when the updates are completed. The bit is primarily
provided to alert software than some other software or utility is in the process of
making changes to the data.
An implementation can also elect to provide a ‘rollback’ feature that uses this
information to decide whether to ‘roll back’ to the previous configuration
information, or to accept the configuration change.
If used, the roll back shall restore all parameters to their previous state. Otherwise,
the change shall take effect when the write occurs.
[7:2] - reserved
[1:0] - 00b = set complete. If a system reset or transition to powered down state
occurs while ‘set in progress’ is active, the BMC will go to the ‘set
complete’ state. If rollback is implemented, going directly to ‘set
complete’ without first doing a ‘commit write’ will cause any pending
write data to be discarded.
01b = set in progress. This flag indicates that some utility or other software
is presently doing writes to parameter data. It is a notification flag
only, it is not a resource lock. The BMC does not provide any
interlock mechanism that would prevent other software from writing
parameter data while.
10b = commit write (optional). This is only used if a rollback is
implemented. The BMC will save the data that has been written
since the last time the ‘set in progress’ and then go to the ‘set in
progress’ state. An error completion code will be returned if this
option is not supported.
11b = reserved
This ‘read only’ field returns which possible Authentication Types (algorithms) can
be enabled for the given channel. The following Authentication Type Enables
parameter selects which Authentication Types are available when activating a
session for a particular maximum privilege level.
[7:6] - reserved
[5:0] - Authentication type(s) enabled for this channel (bitfield):
All bits:
1b = supported
0b = authentication type not available for use.
[5] - OEM proprietary (per OEM identified by the IANA OEM ID in the
RMCP Ping Response)
[4] - straight password / key
[3] - reserved
[2] - MD5
[1] - MD2
[0] - none
241
Intelligent Platform Management Interface Specification
[1]
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
Authentication Type
Enables
2
This field is used to configure which Authentication Types are available for use
when a remote console activates an IPMI messaging connection to the BMC for a
given requested maximum privilege level. Once the session has been activated,
the accepted authentication type will be the only one used for authenticated
packets, regardless of the present operating privilege level, or the privilege level
associated with the command.
Depending on configuration of per-message and user-level authentication
disables, unauthenticated packets (authentication type = none) may also be
accepted. The BMC makes no attempt to check or ensure that stricter
authentication types are associated with higher requested maximum privilege
levels. E.g. it is possible to configure the BMC so activating a session with a
maximum privilege level of ‘User’ requires MD5 while ‘Admin’ requires ‘none’.
Note: An implementation that has fixed privilege and authentication type
assignments, in which case this parameter can be implemented as Read Only. It is
recommended that an implementation that implements a subset of the possible
authentication types returns a CCh error completion code if an attempt is made to
select an unsupported authentication type.
byte 1: Authentication Types returned for maximum requested privilege = Callback
level.
[7:6] - reserved
[5:0] - Authentication type(s) enabled for this channel (bitfield):
All bits:
1b = authentication type enabled for use at given privilege level
0b = authentication type not available for use at given privilege level.
[5] - OEM proprietary (For PPP, per OEM identified by the IANA OEM ID in
the RMCP Ping Response. For other serial/modem modes, a-priori
knowledge of the algorithm is required.)
[4] - straight password / key
[3] - reserved
[2] - MD5
[1] - MD2
[0] - none
byte 2: Authentication Type(s) for maximum privilege = User level
(format follows byte 1)
byte 3: Authentication Type (s) for maximum privilege = Operator level
(format follows byte 1)
byte 4: Authentication Type (s) for maximum privilege = Administrator level
(format follows byte 1)
byte 5: Authentication Type (s) for maximum privilege = OEM level
(format follows byte 1)
242
Intelligent Platform Management Interface Specification
Parameter
#
Connection Mode
3
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1 - connection mode - This parameter determines the protocols used
when performing IPMI messaging to the BMC.
[7] 0b = Modem Connect mode
1b = Direct Connect mode
[6] reserved
Connection mode enables.
Sets which mode or modes can be used for establishing an IPMI
Messaging connection with the BMC. If more than mode is enabled, the
BMC will attempt to auto-detect the appropriate connection mode based
on snooping traffic from the remote console.
Supporting connection mode auto-detect is optional. If an implementation
does not support the capability, it shall return an “Illegal Data Field”
completion code (CCh) if an attempt is made to enable more than one
connection mode at a time. An ‘Illegal Data Field’ code shall also be
returned if an attempt is made to enable a connection mode that the
implementation does not support.
[5:3] [2] -
Session Inactivity
Timeout (optional)
4
[1] [0] [7:4] [3:0] -
reserved
1b = enable Terminal mode
(Note: Terminal mode auto-detect also requires that the
“Enable Baseboard-to-BMC switch on <ESC>(“ option be
enabled in the Mux Switch Configuration parameters,
below.)
1b = enable PPP mode
1b = enable Basic mode
reserved
Inactivity timeout in 30 second increments. 1-based. 0h =
session does not timeout and close due to inactivity.
243
Intelligent Platform Management Interface Specification
Parameter
#
Channel Callback
Control
5
[1]
Parameter Data (non-volatile unless otherwise noted)
This is parameter determines which callback options are enabled or
disabled for the channel. These parameters take precedence over any
user-specific callback settings configured using the Set User Callback
Options command. An option must be enabled in this global parameter in
order to be able to be enabled in the user-specific callback settings. (see
20.11, Set User Callback Options CommandSet User Callback Options
Command).
data 1 - callback enable
[7:2] - reserved
[1] - 1b = enable CBCP callback protocol (see 13.6.1, Callback Control
Protocol (CBCP) SupportCallback Control Protocol (CBCP)
Support)
[0] - 1b = enable IPMI callback
data 2 - CBCP Negotiation Options.
[7:4] - reserved.
[3] - 1b = enable callback to one from list of possible numbers
[2] - 1b = enable user-specifiable callback number. Allow caller to
specify number to be used for callback.
[1] - 1b = enable Pre-specified number. Allow caller to request that
callback occur to a single, pre-specified number for the
user.
[0] - 1b = enable No Callback. Allow caller to request that callback not
be used.
data 3 Callback destination 1. This field holds a Destination Selector that
picks which Destination Dial String from the serial/modem
configuration parameters to use for callback. This selector is used
when the ‘pre-specified number’ option is used. Otherwise, this is
the first number in the list when the “caller selects one number
from a list of numbers” option is used. Refer to 13.6.1, Callback
Control Protocol (CBCP) SupportCallback Control Protocol
(CBCP) Support for characters supported in dial strings for CBCP.
FFh = unspecified.
Note, if this field is set to FFh, the BMC should reject negotiation
for the ‘pre-specified number’ option, even if it is enabled in the
CBCP Negotiation Options field, above.
data 4 Callback destination 2. This is the second number in the list when
the “caller selects one number from a list of numbers” option is
used.
FFh = unspecified.
Note, at least one destination must be specified in order for the
‘callback to one from a list of numbers’ option to be negotiated,
even it that option is enabled in the CBCP Negotiation Options
field, above.
data 5 Callback destination 3. This is the third number in the list when the
“caller selects one number from a list of numbers” option is used.
FFh = unspecified.
Note, at least one destination must be specified in order for the
‘callback to one from a list of numbers’ option to be negotiated,
even it that option is enabled in the CBCP Negotiation Options
field, above.
244
Intelligent Platform Management Interface Specification
Parameter
#
Session Termination
6
IPMI Messaging Comm
Settings
7
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1 - connection termination. This parameter determines whether
serial/modem connections are terminated by inactivity or by a loss of
DCD. For modem mode, the line is hung-up when the specified
termination condition occurs. For both modem and direct connect mode,
the session will be terminated and will need to be reactivated and
authenticated (if authentication is enabled) in order for IPMI messaging
communications to be re-established.
[7:2] - reserved
[1] - 1b = enable session inactivity timeout
0b = disable session inactivity timeout
[0] 1b = close session on loss of DCD (this should be used as the
default setting for both Modem Connect and Direct Connect
mode) [Also see bit to enable mux switch on DCD assertion,
in Mux Switch Control parameter, below]
0b = ignore DCD (DCD is never ignored in Modem Mode)
This parameter is used for IPMI messaging in PPP Mode, Basic Mode,
and Terminal Mode. These settings can be overridden on a perdestination basis for Dial-out LAN Alerting, Dial-Paging, TAP Paging, and
Callback Security, according to the Destination Comm Settings
parameter, below.
IPMI Messaging always occurs with 8 bits/character, no parity, and 1 stop
bit.
data 1 - flow control, DTR hang-up, asynch format
[7:6] - Flow control
00b = No flow control
01b = RTS/CTS flow control (a.k.a. hardware handshake)
10b = XON/XOFF flow control (optional) [if implemented, may not
be supported for all connection modes]
11b = Reserved.
[5] DTR hang-up
0b = disable DTR hang-up
1b = enable DTR hang-up
[4:0] - reserved.
data 2 - bit rate
[7:4] - reserved
[3:0] - 0-5h = reserved. Support for bit rates other than 19.2 kbps is
optional. The BMC must return an error completion if a requested
bit rate is not supported. It is recommended that the ‘parameter
out-of-range’ (C9h) code be used for this situation.
6h = 9600 bps
7h = 19.2 kbps (required)
8h = 38.4 kbps
9h = 57.6 kbps
Ah = 115.2 kbps
245
Intelligent Platform Management Interface Specification
246
Parameter
#
Mux Switch Control
8
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1
See 13.2.4, Serial Port SwitchingSerial Port Switching for additional
information on these bits. Bit [3] is only applicable if PPP Mode is
supported.
[7] reserved
[6] 0b = Disable system power-up/wakeup via [MSVT] <ESC>^
escape sequence
1b = Enable system power-up/wakeup via [MSVT] escape
[3] [5]
sequence
[5] 0b = Disable hard reset on [MSVT] <ESC>R<ESC>r<ESC>R
escape sequence
[3]
1b = Enable hard reset on [MSVT] escape sequence
[4] 0b = Disable Baseboard-to-BMC switch on detecting basic mode
Get Channel Authentication Capabilities message pattern in serial
stream.
1b = Enable Baseboard-to-BMC switch on detecting basic mode
Get Channel Authentication Capabilities message pattern in serial
stream.
[3] 0b = Disable switch to BMC on PPP IPMI-RMCP pattern
1b = Enable switch on PPP IPMI-RMCP pattern
[2] 0b = Disable BMC-to-Baseboard switch on [MSVT] <ESC>Q
[3]
1b = Enable BMC-to-Baseboard switch on [MSVT] <ESC>Q
[1] 0b = Disable Baseboard-to-BMC switch on [MSVT] <ESC>(
[3] [5]
1b = Enable Baseboard-to-BMC switch on [MSVT]<ESC>(
[0] Following only used in Direct Connect Mode (ignored in Modem
Mode)
0b = Disable mux switch to BMC on DCD loss
1b = Enable mux switch on DCD loss
data 2
[7:4] - reserved
[3] - 0b =
Disable Serial Port Sharing. (cannot force mux setting via
Set Serial/Modem Mux command)
The serial connection is assigned to the BMC whenever
the channel is enabled, and cannot be switched to the
baseboard UART. Note: if this setting is 0b and the
serial/modem channel is disabled, the mux will be
connected to the baseboard UART and will not be able to
be switched to the BMC by IPMI command. If Serial Port
Sharing is not implemented, this bit will always be set to
‘disabled’ and will not be changeable.
1b = Enable Serial Port Sharing (can force mux setting using
Set Serial/Modem Mux command)
[2] - 0b = Disable Serial/Modem Connection Active message during
Callback connection.
1b = Enable Serial/Modem Connection Active message during
Callback connection.
[1] - 0b = Disable Serial/Modem Connection Active message during
direct-call
1b = Enable Serial/Modem Connection Active message during
direct-call
[0] - 0b = Send Serial/Modem Connection Active message only once
before switching mux to system
1b = Mux switch acknowledge. Retry Serial/Modem Connection
Active message with retry counts and interval as specified
in Section 13.3.2, Mux Switch CoordinationMux Switch
Coordination.
Intelligent Platform Management Interface Specification
Parameter
#
Modem Ring Time
9
[1]
Parameter Data (non-volatile unless otherwise noted)
Configures the amount of time that the BMC needs to see transitions or
an active state on RI before the BMC claims the mux in Modem Mode.
This setting only applies when the Access Mode is set to “Shared” or
“Pre-boot Only”, Serial Port Sharing is enabled, the channel is enabled
for IPMI Messaging. This includes when the system is powered down, in
order to allow the possibility for using “Wake On Ring” to trigger a wake of
the system without causing the BMC answering the phone. See 13.2.7,
Serial Port Sharing Access CharacteristicsSerial Port Sharing Access
Characteristics for additional information.
data 1 - Ring Duration
[7:6] - reserved
[5:0] - Ring duration in 500 ms increments. 1 based.
00_0000b = BMC switches mux immediately on first detected
transition of RI.
11_1111b (3Fh) = reserved
Modem Init String
10
Modem Escape
Sequence
(optional)
11
Modem Hang-up
Sequence (optional)
12
Modem Dial Command
(optional)
13
Page Blackout Interval
14
data 2 - Ring Dead Time
[7:4] - reserved
[3:0] - Amount of time, in 500 ms increments, that the RI signal must be
deasserted before the BMC determines that ringing has stopped.
0h = 500 ms.
Sets the modem initialization string data. The BMC automatically follows
this string with an <enter> character when sending it to the modem.
set selector = 16-byte block number to set, 1 based. Two
data 1 blocks required, at least three recommended.
data 2:N - Modem Init string data. String is stored as null terminated
ASCII string.
data1:5- Null terminated ASCII string for the Escape string to be sent to
the modem. If this parameter is empty, or this configuration option is not
implemented, the default ‘+++’ sequence will be used. [If a full five
characters are provided, the last character does not need to be null]
data1:8 - Null terminated ASCII string for the hang-up string to be sent to
the modem. The BMC automatically follows this string with an <enter>
character when sending it to the modem. If this parameter is empty, or
this configuration option is not implemented, the default ‘ATH’ sequence
will be used. [If a full eight characters are provided, the last character
does not need to be null]
data1:8 - Null terminated ASCII string for the modem string used to
initiate a dial sequence with the modem. If this parameter is empty, or this
configuration option is not implemented, the default ‘ATD’ sequence will
be used. [If a full eight characters are provided, the last character does
not need to be null]
data 1 - Dial Page, Directed Alert, or TAP Blackout Interval in minutes. 1
based. 00h = no blackout. See Section 13.10, Page Blackout
IntervalPage Blackout Interval for more information.
247
Intelligent Platform Management Interface Specification
Parameter
#
Community String
15
Number of Alert
Destinations
(READ ONLY)
16
Destination Info
(volatile) &
(non-volatile) - see
description.
17
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1:18 - Community String
Default = ‘public’. Used to fill in the ‘Community String’ field in a PET
format trap. This string may optionally be used to hold a vendor-specific
string that is used to provide the network name identity of the system that
generated the event. Printable ASCII string. If 18 non-null characters are
provided, the last character does not need to be a null. 18 characters
must be written when setting this parameter, and 18 will be returned when
this parameter is read. The null character, and any following characters,
will be ignored when the Community String parameter is placed into the
PET. The BMC will return whatever characters were written. I.e. it will not
set bytes following the null to any particular value.
(Community strings are supported on a ‘per channel’ basis in order to
allow the possibility that a different Community String would be used
based on the type of connection.)
data 1 - Number of non-volatile Alert Destinations for this channel.
Destination 0 is always present as a volatile destination that is used with
the Alert ImmediateAlert Immediate command.
[7:5] - reserved.
[3:0] - Number of non-volatile alert destinations. One minimum, fifteen
non-volatile destinations maximum. It is recommended that an
implementation provide at least two destination numbers for each
page/alert type supported, plus two for callback if callback is
supported.
0h = Page Alerting not supported.
Sets the type of page associated with the given destination. For Dial
Page, TAP Page, and Callback, this also selects the dial string
associated with the destination. Destination 0 is used to set a temporary,
RAM-based, value. This value is used with the Alert ImmediateAlert
Immediate command. The value is not guaranteed to be retained across
BMC or system hard resets or power on/off transitions.
data 1 - Destination Selector
A minimum of one and a maximum of fifteen non-volatile destinations
are supported in the specification. If callback is supported, the callback
number is also a type of destination. Destination 0 is always present as
a volatile destination that is used with the Alert ImmediateAlert
Immediate command.
[7:4] - reserved
[3:0] - destination selector.
0h = volatile destination.
1-Fh = non-volatile destination.
data 2 - Destination Type
[7] Alert Acknowledge. Note, some alert types, such as Dial Page, do
not support acknowledge, in which case this bit is ignored
and should be written as 0b.
0b = Unacknowledged. Alert is assumed successful if
transmission occurs without error. This value is also used
with Callback numbers.
1b = Acknowledged. Alert is assumed successful only if
acknowledged is returned.
[6:4] - reserved
[3:0] - Destination Type:
0000b = Dial Page
0001b = TAP Page
0010b = PPP Alert (PET Alert delivered via a PPP-to-LAN connection)
0011b = Basic Mode Callback
0100b = PPP Mode Callback
248
Intelligent Platform Management Interface Specification
Parameter
#
Parameter Data (non-volatile unless otherwise noted)
0101b:1101b = reserved
1110b = OEM 1
1111b = OEM 2
[1]
data 3 - Alert Acknowledge Timeout, in seconds, 0-based (i.e. minimum
timeout = 1 second). Recommended factory default = 5 seconds. Value is
ignored if alert type does not support acknowledge, or if the Alert
Acknowledge bit (above) is 0b.
data 4: Retries
[7] - reserved
[6:4] - Number of times to retry alert once call connection has been
made. (Does not apply to TAP Page or Dial Page alerts)
1-based. 000b = no retries (alert is only sent once).
[3] reserved
[2:0] - Number of times to retry call to given destination. (See below for
Call Retry Interval parameter) 1-based. 000b = no retries (call is
only tried once).
data 5: Destination Type Specific:
For Destination Type = Dial Page:
[7:4] - Dial String Selector
[3:0] - reserved
For Destination Type = TAP Page:
Indicates which set of TAP Service Settings should be used for
communication with this destination.
[7:4] - reserved
[3:0] - TAP Account Selector
For Destination Type = PPP Alert:
Indicates which set of PPP Account settings should be used for
communication with the selected destination.
[7:4] - Destination IP Address Selector
[3:0] - PPP Account Set Selector
Call Retry Interval
(non-volatile)
18
For Destination Type = PPP Mode Callback or Basic Mode Callback:
[7:4] - = Destination IP Address Selector for PPP Mode Callback (The
IP Address is used to enable the BMC to send a
Serial/Modem Connection Active message once the
connection has been established.)
= Dial String Selector for Basic Mode Callback
[3:0] - PPP Account Set Selector (PPP Mode Callback only, reserved
otherwise)
[7:0] - Number of seconds between call (‘busy signal’) retries.
249
Intelligent Platform Management Interface Specification
Parameter
Destination Comm
Settings
(volatile) &
(non-volatile) - see
description.
#
19
Parameter Data (non-volatile unless otherwise noted)
data 1 - Destination Selector
Note that each destination has its own comm settings.
[7:4] - reserved
[3:0] - Destination Selector.
0 = volatile destination.
1-Fh = non-volatile destination.
[1]
Destination comm settings. These settings override the IPMI Messaging
Comm Setting configuration parameter.
data 2 - flow control, DTR hang-up, asynch format
[7:6] - flow control
00b = No flow control
01b = RTS/CTS flow control
10b = XON/XOFF flow control
11b = reserved
[5] - reserved
[4] - stop bits
0b = 1 stop bit (default)
1b = 2 stop bits
[3] - character size
0b = 8 bits (must be 8-bit for PPP)
1b = 7-bits (most TAP services use 7-bit)
[2:0] - parity
000b = no parity.
001b = odd parity.
010b = even parity
Number of Dial Strings
(READ ONLY)
250
20
data 3 - bit rate
[7:4] - reserved
[3:0] - bit rate
0-5h = reserved
6h = 9600 bps
7h = 19.2 kbps (required)
8h = 38.4 kbps
9h = 57.6 kbps
Ah = 115.2 kbps
data 1 - Number of non-volatile Dial Strings for this channel. Dial String 0
is always present and is typically used as a volatile destination that is
used with the Alert ImmediateAlert Immediate command.
[7:5] - reserved.
[3:0] - Number of non-volatile dial strings. One minimum, fifteen nonvolatile dial strings maximum. An implementation should support
one dial string for each destination.
0h = Serial/Modem Alerting and Callback not supported.
Intelligent Platform Management Interface Specification
Parameter
Destination Dial Strings
(volatile) &
(non-volatile) - see
description.
#
21
[1]
Parameter Data (non-volatile unless otherwise noted)
Sets the phone number that the page, alert is to be sent to. The BMC
automatically precedes this string with the Modem Init String sequence,
when not using direct connect mode. The string can contain embedded
modem control sequence characters.
data 1 - destination selector
[7:4] - reserved
[3:0] - Dial String Selector.
0 = volatile dial string
1-Fh = non-volatile dial string.
data 2 - block number to set, 1 based.
Blocks are 16-bytes. At least two blocks are required per number,
supporting a dial string of 31 characters plus terminator.
Number of Alert
Destination IP
Addresses
(READ ONLY)
22
Destination IP
Addresses
23
Number of TAP
Accounts
(READ ONLY)
24
TAP Account
25
TAP Passwords
(WRITE ONLY)
26
TAP Pager ID Strings
27
data 3:N - Dial string data. Null terminated ASCII string.
data 1 - Number of non-volatile Alert Destination IP Addresses for this
channel. Address 0 is always present and is typically used as a volatile
destination that is used with the Alert ImmediateAlert Immediate
command. It is recommended that there be at least one destination IP
Address per PPP Account.
[7:5] - reserved.
[3:0] - Number of Destination IP Addresses. 0h = PPP Alerting and
Callback are not supported.
data 1 - destination selector
[7:4] - reserved
[3:0] - Destination IP Address Selector.
0 = volatile IP Address location
1-Fh = non-volatile IP Address
data 2:5 - destination IP Address. MS-byte first.
data 1 - Number of non-volatile TAP Accounts for this channel. Account 0
is always present and is typically used as a volatile destination that is
used with the Alert ImmediateAlert Immediate command. It is not included
in the count.
[7:5] - reserved.
[3:0] - Number of TAP Accounts. 0h = TAP not supported.
data 1 - set selector = TAP Account Selector, 1-based. At least one set of
TAP Account parameters must be provided for each TAP destination
supported. Account 0 is always present and is typically used as a volatile
destination that is used with the Alert ImmediateAlert Immediate
command.
data 2 - TAP Dial String and Service Setting selectors
[7:4] - Dial String Selector
[3:0] - TAP Service Settings Selector. 1-based. 0h if Destination Type is
not ‘TAP Page’
data 1 - set selector = TAP Account selector, 1 based.
data 2:8 - Password. This string is up to six ASCII characters. Null
terminated if fewer than six characters are used.
This parameter sets and returns the TAP Pager ID (also referred to as
‘Field 1’) for the specified destination. This typically holds the phone
number of the party to be paged. Note that some paging services will
reject transactions that have an empty Field 1.
data 1 - set selector = TAP Account selector, 1 based.
data 2:17 - Pager ID String. This string is up to 16 ASCII characters. Null
terminated if fewer than 16 characters are used. The string will be
transmitted with escaping as specified by the control-character
escaping mask for the given destination.
251
Intelligent Platform Management Interface Specification
Parameter
TAP Service Settings
#
28
[1]
Parameter Data (non-volatile unless otherwise noted)
This parameter is used to configure one or more sets of values related to
strings, escaping, and timeouts and retries associated with a TAP paging
service. The timing parameters are per [TAP], with the exception of T6
and N4, which are extended parameters for this specification. There must
be at least one set of TAP Service Setting parameters supported if TAP
paging is supported on this channel.
data 1 - set selector = TAP Service Setting Selector
There is a 1:1 association between the TAP Parameter selector in this
row, and the selector in the previous row. Parameter fields that share the
same parameter selector form a parameter set.
[7:4] - reserved
[3:0] - TAP Parameter selector. 1-based. (0 = volatile paramters)
data 2 - TAP Confirmation
[7:2] - reserved.
[1:0] - confirmation. This parameter determines what criteria is used by
PEF and the Alert Immediate command to determine that a TAP
Page was successfully delivered to the paging service.
00b = ACK received after end-of-transaction only
01b = code 211 and ACK received after ETX
10b = code 211or 213, and ACK, received after ETX
11b = reserved
data 3:5 - TAP ‘SST’ Service Type field characters, in ASCII. Default =
“PG1”.
Three characters must be provided.
data 6:9 - TAP Control-character escaping mask. (Default =
FFFF_FFFFh)
[31:0] - each bit position represents escaping for corresponding control
characters 31h through 00h. A bit value of 1b = escape the character. 0b
= don’t escape the character. This bit value is ignored for characters that
a required to be escaped by TAP. By default, all control characters are
escaped.
data 10 - timeout parameters 1
[7:4]
TAP T2 - timeout in 500 ms. 0-based (0h = 500 ms). Default = 1h
(1 second)
[3:0]
TAP T1 - timeout in seconds. 0-based (0h = 1 second). Default =
1h (2 seconds)
data 11 - timeout parameters 2
[7:4]
TAP T4 - timeout in seconds. 0-based (0h = 1 second). Default =
3h (4 seconds)
[3:0]
TAP T3 - timeout in 2 second increments. 0-based (0h = 2
seconds). Default = 4h (10 seconds)
data 12 - timeout parameters 3
[7:4] IPMI T6 - IPMI timeout waiting for end-of-transaction
acknowledge, in seconds. 0-based (0 = 1 second). Default = 1h
(2 seconds).
[3:0] TAP T5 - timeout in 2 second increments. 0-based (0h = 2
seconds). Default = 3h (4 seconds)
data 13 - retry parameters 1
[7:4]
TAP N2 - retries. 1-based. (0 = no retry). Default = 3.
[3:0]
TAP N1 - retries. 1-based. (0 = no retry). Default = 3.
data 14 - retry parameters 2
[7:4]
IPMI N4 - number of retries for end-of-transaction. Default = 3.
252
Intelligent Platform Management Interface Specification
Parameter
Terminal Mode
Configuration
#
29
[1]
Parameter Data (non-volatile unless otherwise noted)
[3:0]
TAP N3 - retries. 1-based. (0 = no retry). Default = 3.
This parameter and its fields only apply when Terminal Mode is enabled.
The non-volatile parameters are the initial values used whenever a
terminal mode session is first established. The settings are returned to
the non-volatile settings when a loss of DCD is detected and whenever
the Terminal Mode session is deactivated.
data 1
Parameter Operation
[7:6] - 00b = Set volatile version of data 1 bits 5:0 and data 2
01b = Set non-volatile version of data 1 bits 5:0 and data 2
10b = Copy non-volatile setting to volatile setting (restore default).
11b = reserved
Terminal mode options
[5] - 0b = disable line editing
1b = enable line editing
[4] - reserved
[3:2] - delete control (only applies when line editing is enabled)
00b = BMC outputs a <del> character when <bksp> or <del> is
received
01b = BMC outputs a <bksp><sp><bksp> sequence when <bksp>
or <del> is received
[1] - 0b = no echo
1b = echo (BMC echoes characters it receives)
[0] - 0b = disable handshake (See 13.7.7, Terminal Mode Packet
HandshakeTerminal Mode Packet Handshake)
1b = enable handshake
data 2 - newline sequences
[7:4] - output newline sequence (BMC to console). Selects what
characters the BMC uses as a <newline> sequence when the
BMC writes a line to the console in Terminal Mode.
0h = no termination sequence
1h = <cr-lf> (default)
2h = <NULL>
3h = <CR>
4h = <LF-CR>
5h = <LF>
all other = reserved.
PPP Protocol Options
30
[3:0] - input newline sequence (Console to BMC). Selects what
characters the console uses as the <newline> sequence when
writing to the BMC in Terminal Mode.
0h = reserved
1h = <cr> (default)
2h = <NULL>
all other = reserved.
data 1 - Snoop Control
[7:3] - reserved
[2] - System Negotiation Snooping
1b = BMC snoops system’s PPP negotiation (optional)
0b = BMC doesn’t snoop system’s PPP negotiation
[1:0] - Snoop ACCM Control
00b = BMC uses Transmit ACCM when snooping (mandatory if
connection mode Auto-detect is supported)
253
Intelligent Platform Management Interface Specification
Parameter
#
[1]
Parameter Data (non-volatile unless otherwise noted)
01b = BMC uses Snoop ACCM when snooping (mandatory if
connection mode Auto-detect is supported)
10b = reserved
11b = reserved
data 2 - Negotiation Control
[7:6] - reserved
[5:4] - Negotiation Control
00b = BMC Negotiates link parameters (runs LCP) on initial
connection and whenever mux becomes switched to BMC
and a connection is present.
01b = BMC Negotiates link parameters on initial connection only.
Upon a mux switch to the BMC, the BMC continues using the
parameters it had originally negotiated. If BMC did not do the
negotiation, BMC uses pre-configured settings, following unless system negotiation snooping is enabled, in which case
BMC uses system parameters.
10b = BMC never negotiates link parameters. BMC always uses
pre-configured settings unless system negotiation snooping is
enabled, in which case BMC uses system parameters.
11b = (optional)
[3] - reserved
Pre-configured link settings
[2] - 1b = BMC uses Transmit ACCM to filter received characters
0b = BMC assumes all control characters 00h-1Fh are escaped
[1] - 1b = BMC transmits with Address and Control Field Compression
0b = BMC transmits without Address and Control Field
Compression
[0] - 1b = BMC transmits with Protocol Field Compression
0b = BMC transmits without Protocol Field Compression
data 3 - Negotiation Configuration. This parameter selects what the BMC
negotiates for when it runs LCP.
[7:5] - reserved
[4:3] - BMC PPP IP Address Negotiation.
00b = Request PPP IP Address Assignment. BMC issues an IPCP
Configure-Request for IPCP Option 3 “IP Address”. The BMC
uses the PPP Account #1’s IP Address parameter (below) as
the initial value in the request. If the remote console responds
with a different address in a Configure-Nak for option 3, the
BMC shall accept that IP Address value and use it as its PPP
IP Address.
Per [RFC1332], an address of 00.00.00.00 indicates a request
to the peer (remote console) to provide the IP Address. If
option 3 is rejected, the BMC shall use the PPP Account #1’s
IP Address parameter setting for any IP Protocol (0021h)
packets it sends to the remote console. The BMC may silently
discard any IP Protocol packets addressed to an IP Address
other than the negotiated PPP IP Address.
01b = Request Fixed PPP IP Address. This is the same as
negotiation option 00b “Request PPP IP Address Assignment”
except that the BMC will reject any alternative address offered
by the remote console, and will continue to request PPP
Account #1’s IP Address as the IP Address it will use.
254
Intelligent Platform Management Interface Specification
Parameter
#
[1]
Parameter Data (non-volatile unless otherwise noted)
10b = No PPP IP Address Negotiation. The BMC does not issue a
Configure-Request to request a PPP IP Address. If this option
is selected, the BMC shall accept any IP Protocol (0021h)
message delivered to the Primary or Secondary RMCP Port
addresses. The BMC shall use the PPP IP Address parameter
setting for any IP Packets it generates.
11b = reserved.
PPP Primary RMCP
Port Number (optional)
PPP Secondary RMCP
Port Number (optional)
31
32
[2] - 1b = Enable ACCM negotiation
0b = Disable ACCM negotiation (also use 0b if this option not
supported)
[1] - 1b = Enable Address and Control Field Compression
0b = Disable Address and Control Field Compression (also use 0b
if this option not supported)
[0] - 1b = Enable Protocol Field Compression
0b = Disable Protocol Field Compression(also use 0b if this option
not supported)
data 1:2 - Primary RMCP Port Number, LS-byte first.
Default = 26Fh (RMCP ‘Aux Bus Shunt’ port)
data 1:2 - Secondary Port Number, LS-byte first.
Default = 298h (RMCP ‘Secure Aux Bus’ port)
255
Intelligent Platform Management Interface Specification
Parameter
PPP Link Authentication
#
33
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1 - Link Authentication Type. This configuration option selects
whether the PPP Link itself is authenticated or not. Used with
IPMI Messaging in PPP Mode, this parameter selects which
type of Link Authentication will be used when a remote console
initiates the connection and the BMC acts as the ‘authenticator’.
For PAP, CHAP, and MS-CHAP: The usernames (peer names /
peer IDs) and passwords (peer password) used for Link
Authentication for IPMI Messaging are obtained from users for
which the “Enable User for Link Authentication” bit has been set
using the Set User Access command.
For PAP: The ’peer ID’ field in the Authenticate Request from the
remote console is expected to hold the username, and the
password field the password. The BMC uses the peer ID field
contents. Assuming the user is appropriately enabled for the
channel, the BMC then compares the stored password with the
password that was submitted in the Authenticate Request.
For CHAP and MS-CHAP v1 & v2: The remote console responds to
a challenge generated by the BMC. The BMC takes the name
field from that response and uses it as the username to look up
the user and password information from the user configuration
information. Assuming the user is appropriately enabled for the
channel, the BMC will then use that password to verify the
response. If the name field is empty, the BMC attempts to look
up the password using the Null username. Note that the BMC
also inserts the CHAP Name (parameter 34) in the name field of
the challenge it generates.
CHAP Name
(required if CHAP
supported)
256
34
[7:4] - reserved
[3:0] -PPP Link Authentication protocol
0h = none
1h = CHAP
2h = PAP [RFC 1334]
3h = MS-CHAP v1 [RFC 2433] BMC requires challenge response to
be in Windows NT format.
4h = MS-CHAP v1 [RFC 2433] BMC generates challenge response
in LAN Manager format. (LAN Manager format is deprecated in
RFC 2433, this option is only provided for implementations that
may wish to support connecting to older systems that do not
support Windows NT format.)
5h = MS-CHAP v2 [RFC 2759]
data 1:16 - Null terminated ASCII string for the “system name” used to
represent the BMC when it emits a challenge during CHAP. This is only
used when dialing in to the BMC. If this parameter is provided, it will also
be used by MS-CHAP v1 & v2.
Intelligent Platform Management Interface Specification
Parameter
PPP ACCM
(optional)
#
35
[1]
Parameter Data (non-volatile unless otherwise noted)
data 1:4 - Receive ACCM, MS-byte first. (ls-bit of ls-byte corresponds to
character 00h, ms-bit of ms-byte corresponds to character 1Fh). The
BMC uses this field as part of link negotiation. A 1b in a bit position
identifies a character the must be escaped in order to be accepted by the
BMC.
The BMC will ignore any corresponding characters that are not escaped.
Note that per [RFC 1662] the BMC is required to accept all escaped
characters regardless of whether they’re part of the set that the BMC
required to be escaped.
(If XON/XOFF is used, be sure to include the XON/XOFF characters in
the ACCM.)
If ACCM Negotiation is not enabled (or this parameter is not supported),
the BMC will require that all control characters (00h-1Fh) be escaped.
data 5:8 - Transmit ACCM, MS-byte first (ls-bit of ls-byte corresponds to
character 00h, ms-bit of ms-byte corresponds to character 1Fh). If ACCM
Negotiation is enabled, and this field is supported, this field will determine
which characters the BMC will always transmit with escaping. Characters
that match the value for the PPP flag character (7Eh) and escape
character (7Dh) are always escaped when encountered in the data, so
the values in the corresponding bit positions are ‘don’t care’. I.e. if you set
this field to all 0’s, the 7Eh and 7Dh will still be escaped before being
transmitted.
If ACCM Negotiation is enabled, but this field is not supported, the BMC
will negotiate to transmit all control characters (00h-1Fh) with escaping.
PPP Snoop ACCM
(optional. Required if
Connection Mode Autodetect is supported for
PPP mode)
36
Number of PPP
Accounts
(READ ONLY)
37
PPP Account Dial String
Selector
38
If ACCM Negotiation is not enabled, the BMC will transmit all control
characters (00h-1Fh) with escaping.
data 1:4 - Snoop Receive ACCM, MS-byte first. (ls-bit of ls-byte
corresponds to character 00h, ms-bit of ms-byte corresponds to character
1Fh). A 1b in a bit position identifies a character the must be escaped in
order to be accepted by the BMC. The BMC can be directed to use this
receive ACCM when snooping for a PPP Packet for Connection Mode
Auto-detect. This ACCM is used while snooping when the mux is
switched over to the system.
data 1 - Number of non-volatile destination IP Addresses for this channel.
Account 0 is always present and is typically used as a volatile destination
that is used with the Alert ImmediateAlert Immediate command. Account
1 is used for IPMI Messaging via PPP.
[7:4] - reserved.
[3:0] - 0h = PPP Alerting and Callback are not supported.
9h to Fh = reserved.
data 1 - set selector = account set selector.
data 2 - Dial String Selector. Selects which dial string from the
Destination Dial Strings to use for calling the given PPP account.
257
Intelligent Platform Management Interface Specification
Parameter
PPP Account IP
Addresses, BMC IP
Address
#
39
[1]
Parameter Data (non-volatile unless otherwise noted)
This is the IPv4 Address used to connect to a PPP Server for dial-out
alerting or callback. This value will be assumed to be the IP Address of
the PPP Server unless the PPP Server requests a different address by
negotiating IPCP Option 3 (IP Address). The BMC will offer this address
to the PPP Server if the PPP Server passes 00.00.00.00 as the
requested IP Address when negotiating IPCP option 3. Otherwise, the
BMC will accept the IP Address requested by the PPP Server.
Account 0 is always present and is typically used as a volatile destination
that is used with the Alert ImmediateAlert Immediate command.
Account 1 holds the IP Address used for IPMI Messaging via PPP (the
BMC’s IP Address) instead of a PPP Server’s IP Address. It is also used
as the BMC’s IP Address when connecting to a remote system for
callback or PPP Alerts. The Account 1 IP Address is handled according
to the PPP Protocol Options parameter, above.
PPP Account User
Names
40
data 1 - set selector = account set selector.
data 2:5 - IP Address. MS-byte first. 0000_0000h = unspecified.
This parameter holds the username data for dial-out alerting or callback.
For MS-CHAP: The BMC will prefix the PPP Account Domain
(parameter 41) to this parameter an use the result in the name
field of the response to the challenge. The challenge response
is based on the specified algorithm and the PPP Account User
Password (parameter #42) for the account.
For PAP and CHAP: The BMC uses this parameter to populate the
peer ID when generating a PAP authentication request, or in the
name field of the response to a CHAP challenge. The challenge
is signed using the specified algorithm and the PPP Account
User Password (parameter #42) for the account.
PPP Account User
Domains
41
PPP Account User
Passwords
(Write Only)
42
data 1 - set selector = account set selector.
data 2:N - User Name data. ASCII string. 16 characters, max. Null
terminated if fewer than 16 characters are used.
Required for dial-out alerting using MS-CHAP v1 or v2. If string is nonempty it will be transmitted as a prefix to the user name. Per [RFC 2433]
& [RFC 2759] the domain and user name are separated by a backslash ‘\’
character. This character is not automatically added by the BMC and
should be entered as the last character of the domain.
data 1 - set selector = account set selector.
data 2:N - User Domain data. ASCII string. 16 characters, max. Null
terminated if fewer than 16 characters are used.
The PPP Account parameters (selected by the account set selector
value) are used for connecting to remote systems for dial-out alerting or
callback using PPP/UDP mode.
Note, the usernames (peer names) and passwords used for Link
Authentication for ‘call in’ IPMI Messaging are obtained from users for
which the “Enable User for Link Authentication” bit has been set using the
Set User Access command.
data 1 - set selector = account set selector.
data 2:N - password data. ASCII string. 16 characters max. Null
terminated if fewer than 16 characters are used.
258
Intelligent Platform Management Interface Specification
Parameter
PPP Account
Authentication Settings
#
43
PPP Account
Connection Hold Times
44
PPP UDP Proxy IP
Header data
PPP UDP Proxy
Transmit Buffer Size
(READ ONLY)
45
PPP UDP Proxy
Receive Buffer Size
(READ ONLY)
47
PPP Remote Console IP
Address
[4]
(optional)
System Phone Number
(optional)
48
46
49
Parameter Data (non-volatile unless otherwise noted)
These parameters are used for ‘dial-out’ connections.
[1]
data 1 - set selector = account set selector
data 2 - Link Authentication Type
[7:4] - reserved
[3:0] -PPP Link Authentication protocol.
0h = none (Link Authentication not used)
1h = CHAP
2h = PAP
3h = MS-CHAP v1 [RFC 2433] BMC generates challenge response
in Windows NT format
4h = MS-CHAP v1 [RFC 2433] BMC generates challenge response
in LAN Manager format. (LAN Manager format is deprecated in
RFC 2433, this option is only provided for implementations that
may connect and send alerts to older systems)
5h = MS-CHAP v2 [RFC 2759]
Minimum number of seconds that the call to the given account will be
held prior to automatically hanging up the call. Note the connection will
only stay open for this time if no other alert or action needs to call a
different location or use the channel. Note that an implementation is
allowed to terminate the connection on system resets, power on/off
transitions, and power cycles.
data 1 - set selector = account set selector
data 2 - connection hold time in seconds. 1-based.
data 1:4 - Source IP Address. MS-byte first.
data 5:8 - Destination IP Address. MS-byte first.
data 1:2 - Transmit buffer size in bytes. 1-based.
This parameter is used to return the size of the PPP UDP Proxy Data
transmit buffer. 0000h if PPP UDP Proxy not supported on given
channel.
data 1:2 - Receive buffer size in bytes. 1-based.
This parameter is used to return the size of the PPP UDP Proxy Data
transmit buffer. 0000h if PPP UDP Proxy not supported on given
channel.
data 1:4 - IP Address to offer remote peer if it requests the BMC to
provide it an address as part of IPCP Negotiation. MS-byte first.
This parameter can be used to store an ASCII string representing the
‘dial in’ number for this channel. This can enable an application such as a
LAN remote console to retrieve the phone number for access when a
LAN connection to the managed system becomes unavailable.
data 1 - block number to set, 1 based.
Blocks are 16-bytes. At least two blocks are required per number
string, supporting a dial string of 31 characters plus terminator.
OEM Parameters
1.
2.
3.
4.
5.
192
:
255
data 2:N - Dial string data. Null terminated ASCII string.
This range is available for special OEM configuration parameters. The
OEM is identified according to the Manufacturer ID field returned by the
Get Device ID command.
Choice of system manufacturing defaults is left to the system manufacturer unless otherwise specified.
These settings are copied from the corresponding non-volatile values whenever the system is powered up or hard reset.
Optional but recommended if [MSVT] is implemented in conjunction with IPMI serial port sharing on the same serial interface.
Optional but recommended if PPP supported.
Per [MSVT] The BMC should put out an <ESC>* to the remote console after being switched by the <ESC>( sequence and
after powering up/waking the system using the <ESC>^ sequence. Refer to [MSVT] for timing requirements.
259
Intelligent Platform Management Interface Specification
20.3 Set Serial/Modem Mux Command
This command is used to force or request the selected serial mux to connect the serial connector to the baseboard
serial port or the BMC serial port. The command also returns the present setting of the mux.
Table 20-5, Set Serial/Modem Mux Command
byte
Request Data
1
2
data field
Channel number. This must correspond to the channel number that the
desired serial/modem mux is on.
[7:4] - reserved
[3:0] - Channel number.
Mux setting <VOLATILE> The BMC can override these settings on power
down, power on, and system resets, and change it during system operation
when a serial/modem connection is activated or deactivated.
Otherwise, the ‘mux block’ settings are set back to ‘allowed’ on system power
up, power down, power cycles, and resets except when those actions are
initiated by the Chassis Control command. This enables a remote console to
use the ‘block’ settings to keep connected to the BMC after causing a reset or
power state change using the Chassis Control command.
The BMC power-on default (i.e. when the BMC first gets powered/initialized) is
based on the Access Mode setting for the channel (See Table 13-2, Serial
Port Sharing Access CharacteristicsTable 13-2, Serial Port Sharing Access
Characteristics).
The blocking of ‘switch requests’ and ‘switch forces’ only affects the operation
of the Set Serial/Modem Mux command. Switching caused by other
mechanisms such as snooping and changes to system or connection states
are not blocked.
Response Data
260
1
[7:4] - reserved
[3:0] - 0h = get present mux setting/status only
1h = request switch of mux to system
2h = request switch of mux to BMC
3h = force switch of mux to system
4h = force switch of mux to BMC
5h = block requests to switch mux to system
6h = allow requests to switch mux to system
7h = block requests to switch mux to BMC
8h = allow requests to switch mux to BMC
Completion Code
Intelligent Platform Management Interface Specification
2
Mux setting. This returns the present state of the mux and the mux change
bits from the last Set Serial/Modem Mux command.
switch request enable settings
[7] 0b = requests to switch mux to system are allowed
1b = requests to switch mux to system are blocked
[6] 0b = requests to switch mux to BMC are allowed
1b = requests to switch mux to BMC are blocked
switch status
[5:4] - reserved
[3] 0b = no alert presently in progress
1b = alert in progress on channel
[2] 0b = no IPMI or OEM messaging presently active on channel
1b = IPMI or OEM messaging session active on channel
[1] 0b = request was rejected
1b = request was accepted (see note, below) or switch was forced
present mux setting
[0] 0b = mux is set to system (system can transmit and receive)
1b = mux is set to BMC (BMC can transmit. System can neither
transmit nor receive)
Note: Bit 1 will immediately indicate whether the request was accepted.
However, if ‘mux switch acknowledge’ is enabled, it may take seconds
before the actual switch occurs. Software that needs to confirm a
change of the present mux setting must poll the ‘present mux setting’
bit until it changes to the new state.
20.4 Get TAP Response Codes Command
This command returns the values for up to the last five TAP response codes as an aid to verifying TAP settings.
The values are volatile and are not guaranteed to be retained across system or management controllers resets or
power on/off changes. The values are automatically cleared to ‘0’, ‘0’, ‘0’ at the start of a TAP page. The
command is provided to aid in verifying and debugging the TAP configuration settings.
Table 20-6, Get TAP Response Codes Command
byte
Request Data
Response Data
1
1
2:4
5:7
8:10
11:13
14:16
data field
Channel number.
[7:4] - reserved
[3:0] - Channel number.
Completion Code
Most recent (last received) 3-character ASCII response code. MS-char. first.
Second to last code.
Third.
Fourth.
Fifth.
261
Intelligent Platform Management Interface Specification
20.5 Set PPP UDP Proxy Transmit Data Command
This command is used to load data into the PPP UDP Proxy transmit data buffer. This data is expected to consist
of UDP Packet Data starting with the first UDP data byte (byte following UDP checksum) through the last UDP
data byte. The BMC fills in the remaining PPP and IP/UDP header information, and takes care of framing and
escaping for delivering the data over PPP per [RFC 1662]. The BMC does not verify the correctness of the data.
Table 20-7, Set PPP UDP Proxy Transmit Data Command
byte
Request Data
1
2
3:18
Response Data
1
data field
Channel number.
[7:4] - reserved
[3:0] - Channel number.
Block number. 1-based.
Block Data.
16-byte block of packet data to set. Note the management controller does not
check to see that the block is filled. All writes start at a 16-byte boundary in the
buffer specified by the block number. If fewer than 16-bytes are sent, the BMC
will not overwrite any prior data remaining in the block.
Completion Code
20.6 Get PPP UDP Proxy Transmit Data Command
This command is used to retrieve data that has been written into the PPP UDP Proxy transmit data buffer. The
command is primarily to aid in the test and debug of software that uses the PPP UDP Proxy capability.
Table 20-8, Get PPP UDP Proxy Transmit Data Command
byte
Request Data
Response Data
1
2
Channel number.
[7:4] - reserved
[3:0] - Channel number
Block number to get. 1-based.
1
Completion Code
2:17
262
data field
Block Data. Note, the BMC always returns 16-bytes of data, even if fewer data
bytes were written to the specified block.
Intelligent Platform Management Interface Specification
20.7 Send PPP UDP Proxy Packet Command
This command is used to initiate the transmission of the PPP UDP Proxy Packet using the data stored in the PPP
UDP Proxy transmit data buffer.
Table 20-9, Send PPP UDP Proxy Packet Command
byte
Request Data
1
2:3
4:5
6:9
10:13
14:15
Response Data
1
data field
Channel number.
[7:4] - reserved
[3:0] - Channel number.
UDP Source Port Number. LS-byte first.
UDP Destination Port Number. LS-byte first.
Source IP Address. MS-byte first. 00 00 00 00h = Use PPP IP Address
associated with this channel. (See Table 20-4, Serial/Modem Configuration
ParametersTable 20-4, Serial/Modem Configuration Parameters)
Destination IP Address. MS-byte first. MS-byte first. The Get Session Info
command can be used to look this up for a given session. Software using the
Send PPP UDP Proxy command will usually get a Session ID or Session
Handle from the Boot Options or from a message retrieved via a Get Message
command.
Number of bytes to send. 1-based.
Completion Code. Generic, plus the following command specific.
80h = PPP Link is not up
81h = IP Protocol is not up
263
Intelligent Platform Management Interface Specification
20.8 Get PPP UDP Proxy Receive Data Command
This command is used to retrieve data from the PPP UDP Proxy receive data buffer. The data buffer holds the
complete received PPP IP Packet, from the byte following protocol field up to, but excluding, the FCS field. The
BMC handles PPP Framing and extracting the encapsulated IP data, including checking the PPP Header
information and the FCS, and translating any escaped data in the packet. The BMC does not check the correctness
of the encapsulated IP data. The packet is silently discarded if a bad FCS or partial packet is received.
Table 20-10, Get PPP UDP Proxy Receive Data Command
byte
Request Data
1
2
Response Data
1
2:3
2:17
264
data field
Channel number.
[7:4] - reserved
[3:0] - Channel number.
[7] - Clear Buffer
1b = clear buffer after returning response to this command.
0b = don’t clear buffer after completing this command.
[6:0] - Block Number. 1-based.
000_0000b = Get received data length.
Completion Code.
80h = No packet data available. (Returned when a non-zero block number is
used but there’s no packet data available.)
If block number = 000_0000b:
Number of received data bytes.
0000h after buffer is emptied until a full packet received. This value can be
polled to see when a new packet is available. Software must explicitly clear
the buffer after completing the read of each packet.
Received packets are volatile. The controller may discard packets on
controller resets, system resets, system or controller power on/off changes,
the enable/disable of the associated channel, or the enable/disable of PPP
mode on the associated channel, on changes to the link up/down state, or
changes to the IP protocol up/down state.
If block number non-zero:
Block Data.
Note, the BMC implementation is allowed to always return a return a full 16byte block of data, even if fewer bytes were received in the last block.
Intelligent Platform Management Interface Specification
20.9 Serial/Modem Connection Active (Ping) Command
This command is also referred to as the “Serial/Modem Ping”. It is sent by the BMC to tell a remote console
application whether the system or the BMC is connected to the serial connector before the remote console sends
any messages. If Serial Port Sharing is implemented, this command is also sent out before a mux switch from
BMC to the system occurs, and immediately after a switch from the system to the BMC occurs. Refer to 13.3,
Serial/Modem Connection Active (Ping) MessageSerial/Modem Connection Active (Ping) Message for details
about the operation of this command.
When enabled, the Serial/Modem Connection Active message is sent out at a nominal rate of once every two
seconds, +/- 10% for Basic Mode and Terminal Mode. The BMC is required to send it’s first Serial/Modem
Connection Active message out within 100 milliseconds of the serial connection to the BMC being established.
For PPP Mode, the Serial/Modem Connection Active message will only be sent out before a mux switch from
BMC to the system occurs, and immediately after a switch from the system to the BMC occurs,
When the BMC issues the Serial/Modem Connection Active command, it will typically be addressed to remote
console software. Thus, for IPMI serial/modem and LAN connections the responder’s address byte should be set
to 81h, which is the software ID (SWID) for remote console software. See Section 5.5,Software IDs
(SWIDs)Software IDs (SWIDs), for more information.
Table 20-11, Serial/Modem Connection Active Command
byte
Request Data
Response Data
1
data field
2
Session state
[7:4] - reserved
[3:0] - session state
0h - No session active (password required)
1h - Session active (sent after mux switch to BMC or <ESC>( [if enabled]
detected - and then periodically afterward)
2h - Switching mux to system
IPMI Version in hexadecimal, LSN first. 51h corresponds to IPMI 1.5.
1
Completion Code
20.10 Callback Command
This command is used to initiate a callback to the selected destination. An error completion code will be returned
if the specified destination has not been configured to be a callback destination for the selected channel. This
callback is accomplished using IPMI commands. Note that there is also a PPP option to perform callback using
CBCP. CBCP callback does not use this command. See 13.6.1, Callback Control Protocol (CBCP)
SupportCallback Control Protocol (CBCP) Support.
If the callback command is initiated over the same connection that the callback is to occur over, the BMC will
deliver the response to the callback command, and if the Completion Code is 00h (OK) the BMC will terminate
the session, hang-up the phone, and initiate the callback.
Table 20-12, Callback Command
byte
Request Data
1
2
data field
Channel number. (This value is required to select which configuration
parameters are to be used for callback.)
[7:4] - reserved
[3:0] - Channel number
Destination Selector
Selects which alert destination the callback should go to.
[7:4] - reserved
[3:0] - destination selector. 0 = use volatile destination info. 1-Fh = nonvolatile destination.
265
Intelligent Platform Management Interface Specification
Response Data
1
Completion Code. Generic codes, plus following command-specific
completion codes:
81h = Callback rejected due to alert in progress on this channel.
82h = Callback rejected due to IPMI messaging session active on the callback
channel.
20.11 Set User Callback Options Command
This command is used to configure the callback options associated with a specific user. Note that the options are
also channel-specific. An implementation can allow three different callback numbers to be offered as part of the
callback negotiation.
Table 20-13, Set User Callback Options Command
byte
Request Data
1
2
3
4
5
6
7
Response Data
266
1
data field
User ID. (00h = reserved. 01h=Set password and enable/disabled User 1)
7:6 - reserved.
5:0 - User ID. 000000b, 000001b = reserved. (User ID 1 is permanently
associated with User 1, the null user name).
Channel Number
User callback capabilities
[7:2] - reserved
[1] 1b = user enabled for CBCP callback
[0] 1b = user enabled for IPMI callback
CBCP Negotiation Options. Used when user enabled for CBCP callback, and
CBCP is globally enabled in the serial/modem configuration parameters.
[7:4] - reserved.
[3] 1b = enable callback to one from list of possible numbers. Allow caller
to pick one of a set of phone numbers offered by the BMC.
[2] 1b = enable user-specifiable callback number. Allow caller to specify
number to be used for callback.
[1] 1b = enable pre-specified number. Allow caller to request that
callback occur to a single, pre-specified number for the user.
[0] 1b = enable No Callback. Allow caller to request that callback not be
used.
Callback destination 1. This field holds a Destination Selector that picks which
Destination Dial String from the serial/modem configuration parameters to use
for callback. This selector is used when the ‘pre-specified number’ option is
used. Otherwise, this is the first number in the list when the “caller selects one
number from a list of numbers” option is used.
FFh = unspecified.
Note, if this field is set to FFh, the BMC should reject CBCP negotiation for
the ‘pre-specified number’ option, even if it is enabled in the CBCP
Negotiation Options field, above.
Callback destination 2. This is the second number in the list when the “caller
selects one number from a list of numbers” option is used.
FFh = unspecified.
Note, at least one destination must be specified in order for the ‘callback to
one from a list of numbers’ option to be negotiated, even it that option is
enabled in the CBCP Negotiation Options field, above.
Callback destination 3. This is the third number in the list when the “caller
selects one number from a list of numbers” option is used.
FFh = unspecified.
Note, at least one destination must be specified in order for the ‘callback to
one from a list of numbers’ option to be negotiated, even it that option is
enabled in the CBCP Negotiation Options field, above.
Completion Code.
Intelligent Platform Management Interface Specification
20.12 Get User Callback Options Command
This command is used to return the present settings for the User Callback Options.
Table 20-14, Get User Callback Options Command
byte
Request Data
1
2
Response Data
1
2
3
4
5
6
data field
User ID. (00h = reserved. 01h=Set password and enable/disabled User 1)
[7:6] - reserved.
[5:0] - User ID. 000000b, 000001b = reserved. (User ID 1 is permanently
associated with User 1, the null user name).
Channel Number
Completion Code.
User callback capabilities
[7:2] - reserved
[1] 1b = user enabled for CBCP callback
[0] 1b = user enabled for IPMI callback
CBCP Negotiation Options. Used when user enabled for CBCP callback, and
CBCP is globally enabled in the serial/modem configuration parameters.
[7:4] - reserved.
[3] 1b = callback to one from list of possible numbers enabled
[2] 1b = user-specifiable callback number enabled.
[1] 1b = callback to pre-specified number enabled.
[0] 1b = No Callback enabled. Allow caller to negotiate that callback not
be used.
Callback destination 1. This field holds a Destination Selector that picks which
Destination Dial String from the serial/modem configuration parameters to use
for callback. This selector is used when the ‘pre-specified number’ option is
used. Otherwise, this is the first number in the list when the “caller selects one
number from a list of numbers” option is used.
FFh = unspecified.
Callback destination 2. This is the second number in the list when the “caller
selects one number from a list of numbers” option is used.
FFh = unspecified.
Callback destination 3. This is the third number in the list when the “caller
selects one number from a list of numbers” option is used.
FFh = unspecified.
267
Intelligent Platform Management Interface Specification
21. BMC Watchdog Timer Commands
The BMC implements a standardized ‘Watchdog Timer’ that can be used for a number of system timeout
functions by system management software or by the BIOS. Setting a timeout value of ‘0’ allows the selected
timeout action to occur immediately. This provides a standardized means for devices on the IPMB, such as
Remote Management Cards, to perform emergency recovery actions. Refer to Appendix G - Command
Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 21-1, BMC Watchdog Timer Commands
Command
Reset Watchdog Timer
Set Watchdog Timer
Get Watchdog Timer
Section
Defined
O/M
21.5
21.6
21.7
M
M
M
21.1 Watchdog Timer Actions
The following actions are available on expiration of the Watchdog Timer:
• System Reset
• System Power Off
• System Power Cycle
• Pre-timeout Interrupt (OPTIONAL)
The System Reset on timeout, System Power Off on timeout, and System Power Cycle on timeout action
selections are mutually exclusive. The watchdog timer is stopped whenever the system is powered-down. A
command must be sent to start the timer after the system powers up.
21.2 Watchdog Timer Use Field and Expiration Flags
The watchdog timer provides a ‘timer use’ field that indicates the current use assigned to the watchdog timer. The
watchdog timer provides a corresponding set of ‘timer use expiration’ flags that are used to track the type of
timeout(s) that had occurred.
The timeout use expiration flags retain their state across system resets and power cycles, as long as the BMC
remains powered. The flags are normally cleared solely by the ‘Set Watchdog Timer’ command; with the
exception of the “don’t log” flag, which is cleared after every system hard reset or timer timeout.
The Timer Use fields indicate:
BIOS FRB2 timeout
BIOS POST timeout
4
An FRB-2 (fault-resilient booting, level 2) timeout has occurred. This
indicates that the last system reset or power cycle was due to the system
timeout during POST, presumed to be caused by a failure or hang related to
the bootstrap processor4.
In this mode, the timeout occurred while the watchdog timer was being used
by the BIOS for some purpose other than FRB-2 or OS Load Watchdog.
In a multiprocessor system, the bootstrap processor is defined as the processor that, on system power-up or hard reset, is allowed
to run and execute system initialization (BIOS POST) while the remaining processors are held in a idle state awaiting startup by
the multiprocessing OS.
268
Intelligent Platform Management Interface Specification
OS Load timeout
SMS ‘OS Watchdog’ timeout
OEM
The last reset or power cycle was caused by the timer being used to
‘watchdog’ the interval from ‘boot’ to OS up and running. This mode requires
system management software, or OS support. BIOS should clear this flag if it
starts this timer during POST.
This indicates that the timer was being used by System Management
Software. During run-time, System Management Software (SMS) starts the
timer, then periodically resets it to keep it from expiring. This periodic action
serves as a ‘heartbeat’ that indicates that the OS (or at least the SMS task) is
still functioning. If SMS hangs, the timer expires and the BMC generates a
system reset. When SMS enables the timer, it should make sure the ‘SMS’ bit
is set to indicate that the timer is being used in its ‘OS Watchdog’ role.
Indicates that the timer was being used for an OEM-specific function.
21.2.1 Using the Timer Use field and Expiration flags
The software that sets the Timer Use field is responsible for managing the associated Timer Use Expiration
flag. For example, if system management software sets the timer use to ‘SMS/OS Watchdog’, then that same
system management software is responsible for acting on and clearing the associated Timer Use Expiration
flag.
In addition, software should only interpret or manage the expiration flags for watchdog timer uses that it set.
For example, BIOS should not report watchdog timer expirations or clear the expiration flags for non-BIOS
uses of the timer. This is to allow the software that did set the Timer Use to see that a matching expiration
occurred.
21.3 Watchdog Timer Event Logging
By default, the BMC will automatically log the corresponding sensor-specific watchdog sensor event when a
timer expiration occurs. A “don’t log” bit is provided to temporarily disable the automatic logging. The “don’t
log” bit is automatically cleared (logging re-enabled) whenever a timer expiration occurs.
21.4 Pre-timeout Interrupt
The Watchdog Timer offers a ‘Pre-timeout Interrupt’ option. This option is enabled whenever the ‘Interrupt on
timeout’ option is selected coincident with any of the other Watchdog Timer actions.
If this option is enabled, the BMC generates the selected interrupt a fixed interval before the timer expires. This
feature can be used to allow an interrupt handler to intercept the timeout event before it actually occurs.
The default pre-timeout interrupt interval is one (1) second.
The watchdog timeout action and the pre-timeout interrupt functions are individually enabled. Thus, the
Watchdog Timer can be configured so that when it times out it provides just an interrupt, just the selected action,
both an interrupt and selected action, or none.
If the pre-timeout interval is set to zero, the pre-timeout action occurs concurrently with the timeout action. Note
that if a power or reset action is selected with a pre-timeout interval of zero there is no guarantee that a pretimeout interrupt handler would have time to execute, or to run to completion.
21.4.1 Pre-timeout Interrupt Support Detection
An application that wishes to use a particular pre-timeout interrupt can check for its support by issuing a Set
Watchdog Timer command with the desired pre-timeout interrupt selection. If the controller does not return an
269
Intelligent Platform Management Interface Specification
error completion code, then a Get Watchdog Timer command should be issued to verify that the interrupt
selection was accepted.
While it can be assumed that a controller that accepts a given interrupt selection supports the associated
interrupt, it is recommended that, if possible, an application also generate a test interrupt and verify that the
interrupt occurs and the handler executes correctly.
21.4.2 BIOS Support for Watchdog Timer
If a system ‘Warm Reset’ occurs, the watchdog timer may still be running while BIOS executes POST.
Therefore, BIOS should take steps to stop or restart the watchdog timer early in POST. Otherwise, the timer
may expire later during POST or after the OS has booted.
21.5 Reset Watchdog Timer Command
The Reset Watchdog Timer command is used for starting and restarting the Watchdog Timer from the initial
countdown value that was specified in the Set Watchdog Timer command.
If a pre-timeout interrupt has been configured, the Reset Watchdog Timer command will not restart the timer once
the pre-timeout interrupt interval has been reached. The only way to stop the timer once it has reached this point
is via the Set Watchdog Timer command.
Table 21-2, Reset Watchdog Timer Command
Request Data
Response Data
byte
1
data field
Completion Code. Generic plus the following command-specific completion
codes:
80h = Attempt to start un-initialized watchdog. It is recommended that a
BMC implementation return this error completion code to indicate to
software that a Set Watchdog Timer command has not been issued
to initialize the timer since the last system power on, reset, or BMC
reset. Note that since many systems may initialize the watchdog timer
during BIOS operation, this condition may only be seen by software if
a BMC gets re-initialized during system operation (as might be the
case if a firmware update occurred, for example).
21.6 Set Watchdog Timer Command
The Set Watchdog Timer command is used for initializing and configuring the watchdog timer. The command is
also used for stopping the timer.
If the timer is already running, the Set Watchdog Timer command stops the timer (unless the “don’t stop” bit is
set) and clears the Watchdog pre-timeout interrupt flag (see Get Message Flags command). BMC hard resets,
system hard resets, and the Cold Reset command also stop the timer and clear the flag.
Byte 1 is used for selecting the timer use and configuring whether an event will be logged on expiration.
Byte 2 is used for selecting the timeout action and pre-timeout interrupt type.
Byte 3 sets the pre-timeout interval. If the interval is set to zero, the pre-timeout action occurs concurrently with
the timeout action.
Byte 4 is used for clearing the Timer Use Expiration flags. A bit set in byte 4 of this command clears the
corresponding bit in byte 5 of the Get Watchdog Timer command.
270
Intelligent Platform Management Interface Specification
Bytes 5 and 6 hold the least significant and most significant bytes, respectively, of the countdown value. The
Watchdog Timer decrement is one count/100 ms. The counter expires when the count reaches zero. If the counter
is loaded with zero and the Reset Watchdog command is issued to start the timer, the associated timer events
occur immediately.
Table 21-3, Set Watchdog Timer Command
Request Data
byte
1
2
3
4
5
6
1
Response Data
1.
data field
Timer Use
[7] - 1b = don’t log
[6] - 1b = don’t stop timer on Set Watchdog Timer command (new for IPMI
v1.5) new parameters take effect immediately. If timer is already
running, countdown value will get set to given value and
countdown will continue from that point. If timer is already
stopped, it will remain stopped. If the pre-timeout interrupt bit is
[1]
set, it will get cleared.
0b = timer stops automatically when Set Watchdog Timer command is
received
[5:3] - reserved
[2:0] - timer use (logged on expiration when “don’t log” bit = 0b)
000b =
reserved
001b =
BIOS FRB2
010b =
BIOS/POST
011b =
OS Load
100b =
SMS/OS
101b =
OEM
110b -111b = reserved
Timer Actions
[7] - reserved
[6:4] - pre-timeout interrupt (logged on expiration when “don’t log” bit = 0b)
000b =
none
001b =
SMI
010b =
NMI / Diagnostic Interrupt
011b =
Messaging Interrupt (this is the same interrupt as
allocated to the messaging interface)
100b,111b = reserved
[3] - reserved
[2:0] - timeout action
000b =
no action
001b =
Hard Reset
010b =
Power Down
011b =
Power Cycle
100b,111b = reserved
Pre-timeout interval in seconds. ‘1’ based.
Timer Use Expiration flags clear
(0b = leave alone, 1b = clear timer use expiration bit)
[7] - reserved
[6] - reserved
[5] - OEM
[4] - SMS/OS
[3] - OS Load
[2] - BIOS/POST
[1] - BIOS FRB2
[0] - reserved
Initial countdown value, lsbyte (100 ms/count)
Initial countdown value, msbyte
Completion Code
Potential race conditions exist with implementations of this option. If the Set Watchdog Timer
command is sent just before a pre-timeout interrupt or timeout is set to occur, the timeout
could occur before the command is executed. To avoid this condition, it is recommended that
software set this value no closer than 3 counts before the pre-timeout or timeout value is
reached.
271
Intelligent Platform Management Interface Specification
21.7 Get Watchdog Timer Command
This command retrieves the current settings and present countdown of the watchdog timer. The Timer Use
Expiration flags in byte 5 retain their states across system resets and system power cycles. With the exception of
bit 6 in the Timer Use byte, the Timer Use Expiration flags are cleared using the Set Watchdog Timer command.
They may also become cleared because of a loss of BMC power, firmware update, or other cause of BMC hard
reset. Bit 6 of the Timer Use byte is automatically cleared to 0b whenever the timer times out, is stopped when the
system is powered down, enters a sleep state, or is reset.
Table 21-4, Get Watchdog Timer Command
Request Data
Response Data
byte
1
2
3
4
5
6
7
8
9
272
data field
Completion Code
Timer Use
[7] - 1b =don’t log
[6] - 1b = timer is started (running)
0b = timer is stopped
[5:3] - reserved
[2:0] - timer use (logged on expiration if don’t log bit = 0b)
000b
= reserved
001b
= BIOS FRB2
010b
= BIOS/POST
011b
= OS Load
100b
= SMS/OS
101b
= OEM
110b,111b = reserved
Timer Actions
[7] - reserved
[6:4] - pre-timeout interrupt
000b =
none
001b =
SMI
010b =
NMI / Diagnostic Interrupt
011b =
Messaging Interrupt (this would be the same interrupt as
allocated to the messaging interface)
100b,111b = reserved
[3] - reserved
[2:0] - timeout action
000b =
no action
001b =
Hard Reset
010b =
Power Down
011b =
Power Cycle
100b,111b = reserved
Pre-timeout interval in seconds. ‘1’ based.
Timer Use Expiration flags (1b = timer expired while associated ‘use’ was
selected.)
[7] - reserved
[6] - reserved
[5] - OEM
[4] - SMS/OS
[3] - OS Load
[2] - BIOS/POST
[1] - BIOS FRB2
[0] - reserved
Initial countdown value, lsbyte (100 ms/count)
Initial countdown, msbyte
Present countdown value, lsbyte. The initial countdown value and present
countdown values should match immediately after the countdown is
initialized via a Set Watchdog Timer command and after a Reset Watchdog
Timer has been executed.
Note that internal delays in the BMC may require software to delay up to 100
ms before seeing the countdown value change and be reflected in the Get
Watchdog Timer command.
Present countdown value, msbyte
Intelligent Platform Management Interface Specification
273
Intelligent Platform Management Interface Specification
274
Intelligent Platform Management Interface Specification
22. Chassis Commands
The following chassis commands are specified for IPMI v1.5. These commands are primarily to provide
standardized chassis status and control functions for Remote Management Cards and Remote Consoles that access
the BMC. They can also be used for ‘emergency’ management control functions by system management software.
Refer to Appendix G - Command Assignments
Appendix G - Command Assignments
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 22-1, Chassis Commands
Command
Get Chassis Capabilities
Get Chassis Status
Chassis Control
Chassis Reset
Chassis Identify
Set Front Panel Enables
Set Chassis Capabilities
Set Power Restore Policy
Set Power Cycle Interval
Get System Restart Cause
Set System Boot Options
Get System Boot Options
Get POH Counter
1.
2.
Section
Defined
O/M
22.1
22.2
22.3
22.4
22.5
22.5a
22.6
22.7
22.7a
22.9
22.10
22.11
22.12
M
[1]
M
[1]
M
O
O
O
O
O
O
[2]
O
[2]
O
[2]
O
O
These commands are mandatory for standalone server motherboards that
include ACPI-based power control capabilities.
Highly recommended. These commands should be supported on host systems
that support remote reset and power on/off capabilities, since these commands
enable remote coordination of the booting process with BIOS.
22.1 Get Chassis Capabilities Command
The Get Chassis Capabilities command returns information about which main chassis management functions are
present on the IPMB (or virtual IPMB) and what addresses are used to access those functions. This command is
used to find the devices that provide functions such as SEL, SDR, and ICMB Bridging so that they can be
accessed via commands delivered via a physical or logical IPMB. Note that the command does not include a
channel number for the individual functions, therefore all reported functions must be located on the primary
IPMB.
Refer to [ICMB] for additional information.
The Chassis Capabilities information is non-volatile. There is no requirement that the information be
configurable. The Chassis Device function in a peripheral chassis may be hardcoded with this information. For
example, a system that implements the ICMB as an add-on bridge to a BMC will typically be able to have the
well known address for the BMC (20h) hardcoded as the address for the Chassis SDR, SEL, and SM Devices,
while the Chassis FRU Info Device address could be set with the chassis devices own address.
275
Intelligent Platform Management Interface Specification
An add-in device that serves as a bridge device that could be used in different vendors systems may want to
provide a way for this information to be configured. The Set Chassis Capabilities command is one option for
providing this.
Table 22-2, Get Chassis Capabilities Command
byte
Request Data
Response Data
-
1
2
Completion Code
Capabilities Flags
[7:4] - reserved
[3] - 1b = provides power interlock (IPMI 1.5)
[2] - 1b = provides Diagnostic Interrupt (FP NMI) (IPMI 1.5)
[1] - 1b = Provides “Front Panel Lockout” (this indicates that the chassis
has capabilities to lock out external power control and reset
button or front panel interfaces and/or detect tampering with
those interfaces)
[0] - 1b = Chassis provides intrusion (physical security) sensor
Chassis FRU Info Device Address.
2
Note: all IPMB addresses used in this command are have the 7-bit I C slave
address as the most-significant 7-bits and the least significant bit set to 0b.
00h = unspecified.
Chassis SDR Device Address
Chassis SEL Device Address
Chassis System Management Device Address
Chassis Bridge Device Address. Reports location of the ICMB bridge
function. If this field is not provided, the address is assumed to be the BMC
address (20h). Implementing this field is required when the Get Chassis
Capabilities command is implemented by a BMC, and whenever the Chassis
Bridge function is implemented at an address other than 20h.
3
4
5
6
(7)
276
data field
-
Intelligent Platform Management Interface Specification
22.2 Get Chassis Status Command
The following command returns information regarding the high-level status of the system chassis and main power
subsystem.
Table 22-3, Get Chassis Status Command
byte
Request Data
Response Data
data field
-
-
1
Completion Code
2
Current Power State
[7] - reserved
[1]
[6:5] - power restore policy
00b = chassis stays powered off after AC/mains returns
01b = after AC returns, power is restored to the state that was in effect
when AC/mains was lost
10b = chassis always powers up after AC/mains returns
11b = unknown
[4] - power control fault
1b = Controller attempted to turn system power on or off, but system
did not enter desired state.
[3] - power fault
1b = fault detected in main power subsystem.
[2] - 1b = Interlock (chassis is presently shut down because a chassis
panel interlock switch is active). (IPMI 1.5)
[1] - Power overload
1b = system shutdown because of power overload condition.
[0] - Power is on
1b = system power is on
0b = system power is off (soft-off S4/S5 or mechanical off)
3
Last Power Event
[7:5] - reserved
[4] - 1b = last ‘Power is on’ state was entered via IPMI command
[3] - 1b = last power down caused by power fault
[2] - 1b = last power down caused by a power interlock being activated
[1] - 1b = last power down caused by a Power overload
[0] - 1b = AC failed
4
Misc. Chassis State
[7:4] - reserved
[3] - 1b = Cooling/fan fault detected
[2] - 1b = Drive Fault
[1] - 1b = Front Panel Lockout active (power off and / reset via chassis
push-buttons disabled.)
[0] - 1b = Chassis intrusion active
(5)
Front Panel Button Capabilities and disable/enable status (Optional)
(“Button” actually refers to the ability for the local user to be able to perform
the specified functions via a pushbutton, switch, or other ‘front panel’ control
built into the system chassis.)
[7] - 1b = Standby (sleep) button disable allowed
[6] - 1b = Diagnostic Interrupt button disable allowed
[5] - 1b = Reset button disable allowed
[4] - 1b = Power off button disable allowed (in the case there is a single
combined power/standby (sleep) button, disabling power off also
disables sleep requests via that button.)
[3] - 1b = Standby (sleep) button disabled
[2] - 1b = Diagnostic Interrupt button disabled
[1] - 1b = Reset button disabled
[0] - 1b = Power off button disabled (in the case there is a single combined
power/standby (sleep) button, then this indicates that sleep
requests via that button are also disabled.)
1. In some installations, the chassis’ main power feed may be DC based. For
example, -48V. In this case, the power restore policy for AC/mains refers to the
loss and restoration of the DC main power feed.
277
Intelligent Platform Management Interface Specification
22.3 Chassis Control Command
The following command provides a mechanism for providing power up, power down, and reset control.
Table 22-4, Chassis Control Command
byte
1
Request Data
1
Response Data
1.
data field
[7:4] - reserved
[3:0] - chassis control
0h = power down. Force system into soft off (S4/S45) state. This is for
‘emergency’ management power down actions. The command
does not initiate a clean shut-down of the operating system prior to
powering down the system.
1h = power up.
2h = power cycle (optional). This command provides a power off interval
of at least 1 second following the deassertion of the system’s
POWERGOOD status from the main power subsystem. It is
recommended that no action occur if system power is off (S4/S5)
when this action is selected, and that a D5h “Request parameter(s)
not supported in present state.” error completion code be returned.
Note that some implementations may cause a system power up if a
power cycle operation is selected when system power is down. For
consistency of operation, it is recommended that system
management software first check the system power state before
issuing a power cycle, and only issue the command if system
power is ON or in a lower sleep state than S4/S5.
3h = hard reset. In some implementations, the BMC may not know
whether a reset will cause any particular effect and will pulse the
system reset signal regardless of power state. If the implementation
can tell that no action will occur if a reset is delivered in a given
power state, then it is recommended (but still optional) that a D5h
“Request parameter(s) not supported in present state.” error
completion code be returned.
4h = pulse Diagnostic Interrupt. (optional) Pulse a version of a diagnostic
interrupt that goes directly to the processor(s). This is typically used
to cause the operating system to do a diagnostic dump (OS
dependent). The interrupt is commonly an NMI on IA-32 systems
and an INIT on Intel® Itanium™ processor based systems.
5h = Initiate a soft-shutdown of OS via ACPI by emulating a fatal
overtemperature. (optional)
all other = reserved
[1]
Completion Code
The implementation is allowed to return the completion code prior to performing
the selected control action if necessary.
22.4 Chassis Reset Command
This command was used with early versions of the ICMB. It has been superceded by the Chassis Control command
and is not recommended for new implementations. Refer to [ICMB] for more information. The Chassis Reset
command allows chassis logic (excluding the chassis device itself) to be reset. For host systems, this corresponds to
a system hard reset.
278
Intelligent Platform Management Interface Specification
Table 22-5, Chassis Reset Command
byte
Request Data
Response Data
data field
-
-
1
Completion Code
22.5 Chassis Identify Command
This command causes the chassis to physically identify itself by a mechanism chosen by the system
implementation; such as turning on blinking user-visible lights or emitting beeps via a speaker, LCD panel, etc.
Unless the optional “Force Identify On” capability is supported and used, the Chassis Identify command
automatically times out and deasserts the indication after a configurable time-out. Software must periodically
resend the command to keep the identify condition asserted. This will restart the timeout.
Table 22-6, Chassis Identify Command
byte
[1]
(1)
Request Data
[2]
(2)
Response Data
1.
2.
1
data field
[7:0] - Identify Interval in seconds. 1-based. Timing accuracy = -0/+20%. This
field is optional. If this byte is not provided the default timeout shall be
15 seconds -0/+20%. Note that this byte can be overridden by optional
byte 2.
00h = Turn off Identify
Force Identify On. This optional field enables software to command the
Identify to be On indefinitely. The BMC implementation should return
an error completion code if this byte is not supported.
[7:1] - reserved
[0] - 1b = Turn on Identify indefinitely. This overrides the values in byte 1.
0b = Identify state driven according to byte 1.
Completion Code
This parameter byte is optionally present. If not provided, the Chassis Identify can be used to turn
on the Identify indication for the default timeout interval, but cannot be used to turn the indication
off.
This parameter byte is optionally present. If provided, it is highly recommended that the chassis
provides a local manual mechanism that enables a user or service personnel to turn off Identify. If a
local manual mechanism is not provided, AC removal (BMC reset) should remove the indication.
22.5a Set Front Panel Enables
The following command is used to enable or disable the buttons on the front panel of the chassis. (Button actually
refers to the ability for the local user to be able to perform the specified functions via a pushbutton, switch, or
other ‘front panel’ control built into the system chassis.) These values will be returned in the Front Panel Button
capabilities and disable/enable status (byte 5) of the Get Chassis Status command.
Table 22-6a, Set Front Panel Button Enables Command
byte
1
Response Data
1
data field
Front Panel Button Enables
[7:4] - reserved
[3] - 1b = disable Standby (sleep) button for entering standby (sleep)
(control can still be used to wake the system)
[2] - 1b = disable Diagnostic Interrupt button
[1] - 1b = disable Reset button
[0] - 1b = disable Power off button for power off only (in the case there is a
single combined power/standby (sleep) button, then this also
disables sleep requests via that button)
Completion Code
279
Intelligent Platform Management Interface Specification
22.6 Set Chassis Capabilities Command
This command is used to set the values that will be returned for the Get Chassis Capabilities command into nonvolatile storage associated with the Chassis Device.
This command is recommended for all add-on bridge applications.
Table 22-7, Set Chassis Capabilities Command
byte
Request Data
1
2
3
4
5
(6)
Response Data
1
data field
Capabilities Flags
[7:2] - reserved
[1] 1b = Provides Front Panel Lockout (see 22.1, Get Chassis
CapabilitiesGet Chassis Capabilities)
[0] 1b = Provides intrusion
Chassis FRU Info Device Address (see 22.1, Get Chassis CapabilitiesGet
Chassis Capabilities for a description of these addresses, their use,
and the field formatting)
Chassis SDR Device Address
Chassis SEL Device Address
Chassis SM Device Address
Chassis Bridge Device Address
Completion Code. Note, this command does not return an error completion
code if an attempt is made to change a ‘read-only’ parameter.
Software must check which fields in the response match the value
from the request by using the Get Chassis Capabilities command.
22.7 Set Power Restore Policy Command
This command can be used to configure the power restore policy. This configuration parameter is kept in nonvolatile storage. The power restore policy determines how the system or chassis behaves when AC power returns
after an AC power loss. The Get Chassis Status command returns the power restore policy setting.
Table 22-8, Set Power Restore Policy Command
byte
Request Data
1
Response Data
1
2
1.
280
data field
[7:3] - reserved
[2:0] - power restore policy
011b = no change (just get present policy support)
010b = chassis always powers up after AC/mains is applied or
returns
001b = after AC/mains is applied or returns, power is restored to the
state that was in effect when AC/mains was removed or lost
000b = chassis always stays powered off after AC/mains is applied,
power pushbutton or command required to power on system
all other = reserved
Completion Code. A non-zero completion code should be returned if an
attempt is made to set a policy option that is not supported.
power restore policy support (bitfield)
[7:3] - reserved
[2] - 1b = chassis supports always powering up after AC/mains returns
[1] - 1b = chassis supports restoring power to state that was in effect when
AC/mains was lost
[0] - 1b = chassis supports staying powered off after AC/mains returns
In some installations, the chassis’ main power feed may be DC based. For example, -48V.
In this case, the power restore policy for AC/mains refers to the loss and restoration of the
DC main power feed.
Intelligent Platform Management Interface Specification
22.7a Set Power Cycle Interval
This command provides a mechanism for configuring the power cycle interval for the system. This interval
determines the time that system power will be powered down during a power cycle operation initiated by the
Chassis Control command or the watchdog time. The setting is non-volatile.
Table 22-8a, Set Power Cycle Interval Command
byte
Request Data
1
Response Data
1
data field
[7:0] - Power Cycle Interval in seconds. 1-based.
00h = no delay.
Completion code.
22.8 Remote Access Boot control
The BMC allows a remote console application to optionally direct the boot process following a command to reset,
power-up, or power-cycle the system. The remote console sets Boot Option flags prior to issuing a command to
reset, power up, or power-cycle the system. The system BIOS can then read these flags after the system restarts
and perform the requested boot operation. This will typically be used to direct the system to boot to an alternative
partition or source in order to perform emergency remote recovery operations.
The Boot Option parameter definitions follow the set of Boot Option parameters defined by the DMTF Pre-OS
Working Group.
Implementing Remote Access Boot control is optional.
281
Intelligent Platform Management Interface Specification
22.9 Get System Restart Cause Command
This command returns information about what action last caused the system to restart. BIOS can use this
command in conjunction with the System Boot Options as additional information in determining whether to
perform the requested boot operation.
Table 22-9, Get System Restart Cause Command
byte
Request Data
Response Data
data field
-
-
1
2
Completion Code
Restart Cause
[7:4] - reserved
[3:0] - 0h = unknown (system start/restart detected, but cause unknown)
[required if this condition exists]
1h = Chassis Control command [required]
2h = reset via pushbutton [optional]
3h = power-up via power pushbutton [optional]
4h = Watchdog expiration (see watchdog flags) [required]
5h = OEM [optional]
6h = automatic power-up on AC being applied due to ‘always restore’
power restore policy (see 22.7, Set Power Restore Policy
CommandSet Power Restore Policy Command) [optional]
7h = automatic power-up on AC being applied due to ‘restore previous
power state’ power restore policy (see 22.7, Set Power Restore
Policy CommandSet Power Restore Policy Command) [optional]
8h = reset via PEF [required if PEF reset supported]
9h = power-cycle via PEF [required if PEF power-cycle supported]
Ah = soft reset (e.g. CTRL-ALT-DEL) [optional]
Bh = power-up via RTC (system real time clock) wakeup [optional]
all other = reserved
Channel number. (Channel that command was received over)
3
22.10 Set System Boot Options Command
This command is used to set parameters that direct the system boot following a system power up or reset. The
boot flags only apply for one system restart. It is the responsibility of the system BIOS to read these settings
from the BMC and then clear the boot flags.
It is possible that a remote console application could set the boot option flags and then be terminated either
accidentally or intentionally. In this circumstance, it’s possible that a user initiated system restart could occur
hours or even days later. If the boot options were used without examining the reset cause, this could cause an
unexpected boot sequence. Thus, the BMC will automatically clear a ‘boot flags valid bit’ if a system restart is
not initiated by a Chassis Control command within 60 seconds +/- 10% of the valid flag being set. The BMC
will also clear the bit on any system resets or power-cycles that are not triggered by a System Control
command. This default behavior can be temporarily overridden using the ‘BMC boot flag valid bit clearing’
parameter.
282
Intelligent Platform Management Interface Specification
Table 22-10, Set System Boot Options Command
byte
Request Data
1
(2:N)
Response Data
1
data field
Parameter valid
[7] - 1b = mark parameter invalid / locked
0b = mark parameter valid / unlocked
[6:0] - boot option parameter selector
Boot option parameter data, per Table 22-12, Boot Option Parameters.
Passing 0-bytes of parameter data allows the parameter valid bit to be
changed without affecting the present parameter setting.
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
81h = attempt to set the ‘set in progress’ value (in parameter #0) when not in
the ‘set complete’ state. (This completion code provides a way to
recognize that another party has already ‘claimed’ the parameters)
82h = attempt to write read-only parameter
22.11 Get System Boot Options Command
This command is used to retrieve the boot options set by the Set System Boot Options command.
Table 22-11, Get System Boot Options Command
byte
Request Data
1
2
3
Response Data
1
2
3
4:N
data field
Parameter selector
[7] - reserved
[6:0] - boot option parameter selector
[7:0] - Set Selector
Selects a particular block or set of parameters under the given parameter
selector.
Write as 00h if parameter doesn’t use a Set Selector
[7:0] - Block Selector
Selects a particular block within a set of parameters.
Write as 00h if parameter doesn’t use a Block Selector.
Note: As of this writing, there are no IPMI-specified Boot Options parameters
that use the block selector. However, this field is provided for consistency with
other configuration commands and as a placeholder for future extension of the
IPMI specification.
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
[7:4] - reserved
[3:0] - parameter version. 1h for this specification unless otherwise specified.
Parameter valid
[7] - 1b = parameter marked invalid / locked
0b = parameter marked valid / unlocked
[6:0] - boot option parameter selector
Configuration parameter data, per Table 22-12, Boot Option ParametersTable
22-12, Boot Option Parameters.
If the rollback feature is implemented, the BMC makes a copy of the existing
parameters when the ‘set in progress’ state becomes asserted (See the Set In
Progress parameter #0). While the ‘set in progress’ state is active, the BMC
will return data from this copy of the parameters, plus any uncommitted
changes that were made to the data. Otherwise, the BMC returns parameter
data from non-volatile storage.
283
Intelligent Platform Management Interface Specification
Table 22-12, Boot Option Parameters
Parameter
Set In Progress
(volatile)
#
0
service partition
selector
[1]
(semi-volatile)
1
service partition scan
[1]
(nonsemi-volatile)
2
BMC boot flag valid bit
clearing
[1]
(semi-volatile)
3
284
Parameter Data (non-volatile unless otherwise noted)
data 1 - This parameter is used to indicate when any of the following parameters are
being updated, and when the updates are completed. The bit is primarily provided to alert
software that some other software or utility is in the process of making changes to the
data.
An implementation can also elect to provide a ‘rollback’ feature that uses this information
to decide whether to ‘roll back’ to the previous configuration information, or to accept the
configuration change.
If used, the roll back shall restore all parameters to their previous state. Otherwise, the
change shall take effect when the write occurs.
[7:2] - reserved
[1:0] - 00b = set complete. If a system reset or transition to powered down state occurs
while ‘set in progress’ is active, the BMC will go to the ‘set complete’ state. If
rollback is implemented, going directly to ‘set complete’ without first doing a
‘commit write’ will cause any pending write data to be discarded.
01b = set in progress. This flag indicates that some utility or other software is
presently doing writes to parameter data. It is a notification flag only, it is not
a resource lock. The BMC does not provide any interlock mechanism that
would prevent other software from writing parameter data while.
10b = commit write (optional). This is only used if a rollback is implemented. The
BMC will save the data that has been written since the last time the ‘set in
progress’ and then go to the ‘set in progress’ state. An error completion
code will be returned if this option is not supported.
11b = reserved
data 1
[7:0] - service partition selector. This value is used to select which service partition BIOS
should boot using. This document doesn’t specify which value corresponds to a
particular service partition.
00h = unspecified.
data 1
[7:2] - reserved
[1] - 1b = Request BIOS to scan for specified service partition. BIOS clears this bit after
the requested scan has been performed.
[0] - 1b = Service Partition discovered. BIOS sets this bit to indicate it has discovered
the specified service partition. The bit retains the value from the last scan.
Therefore, to get up-to-date status of the discovery state, a scan may need to
be requested.BIOS must clear this bit on all system resets and power ups,
except when a scan is requested.
data 1- BMC boot flag valid bit clearing. Default = 0000b.
[7:5] - reserved
[4] - 1b = don’t clear valid bit on reset/power cycle caused by PEF
[3] - 1b = don’t automatically clear boot flag valid bit if Chassis Control command not
received within 60-second timeout (countdown restarts when a Chassis
Control command is received)
[2] - 1b = don’t clear valid bit on reset/power cycle caused by watchdog timeout
[1] - 1b = don’t clear valid bit on pushbutton reset / soft-reset (e.g. “Ctrl-Alt-Del”)
[0] - 1b = don’t clear valid bit on power up via power pushbutton or wake event
Intelligent Platform Management Interface Specification
boot info acknowledge
[1]
(semi-volatile)
4
These flags are used to allow individual parties to track whether they’ve already seen and
handled the boot information. Applications that deal with boot information should check
the boot info and clear their corresponding bit after consuming the boot options data.
data 1: Write Mask (‘write-only’. This field is returned as 00h when read. This is to
eliminate the need for the BMC to provide storage for the Write Mask field.)
[7] - 1b = enable write to bit 7 of Data field
[6] - 1b = enable write to bit 6 of Data field
[5] - 1b = enable write to bit 5 of Data field
[4] - 1b = enable write to bit 4 of Data field
[3] - 1b = enable write to bit 3 of Data field
[2] - 1b = enable write to bit 2 of Data field
[1] - 1b = enable write to bit 1 of Data field
[0] - 1b = enable write to bit 0 of Data field
boot flags
[1]
(semi-volatile)
5
data 2:Boot Initiator Acknowledge Data
The boot initiator should typically write FFh to this parameter prior to initiating the boot.
The boot initiator may write 0’s if it wants to intentionally direct a given party to ignore the
boot info. This field is automatically initialized to 00h when the management controller is
first powered up or reset.
[7] - reserved. Write as 1b. Ignore on read.
[6] - reserved. Write as 1b. Ignore on read.
[5] - reserved. Write as 1b. Ignore on read.
[4] - 0b = OEM has handled boot info.
[3] - 0b = SMS has handled boot info.
[2] - 0b = OS / service partition has handled boot info.
[1] - 0b = OS Loader has handled boot info.
[0] - 0b = BIOS/POST has handled boot info.
data 1
[7] - 1b = boot flags valid. The bit should be set to indicate that valid flag data is present.
This bit may be automatically cleared based on the boot flag valid bit clearing
parameter, above.
[6:0] - reserved
BIOS support for the following flags is optional. If a given flag is supported, it must cause
the specified function to occur in order for the implementation to be considered to be
conformant with this specification.
The following parameters represent temporary overrides of the BIOS default settings.
BIOS should only use these parameters for the single boot where these flags were set. If
the bit is 0b, BIOS should use its default configuration for the given option.
data 2
[7] - 1b = CMOS clear
[6] - 1b = Lock Keyboard
[5:2] - Boot device selector
0000b = No override
0001b = Force PXE
[2]
0010b = Force boot from default Hard-drive
[2]
0011b = Force boot from default Hard-drive, request Safe Mode
[2]
0100b = Force boot from default Diagnostic Partition
[2]
0101b = Force boot from default CD/DVD
0110b-1110b = Reserved
1111b = Force boot from Floppy/primary removable media
[1] - 1b = Screen Blank
[0] - 1b = Lock out Reset buttons
data 3
[7] - 1b = Lock out (power off/ sleep request) via Power Button
[6:5] - Firmware (BIOS) Verbosity (Directs what appears on POST display)
00b = system default
01b = request quiet display
10b = request verbose display
11b = reserved
[4] - 1b = Force progress event traps. When set to 1b, the BMC transmits PET traps for
285
Intelligent Platform Management Interface Specification
BIOS progress events to the LAN or serial/modem destination for the session
that set the flag. Since this capability uses PET traps, this bit will be ignored if
for connection modes that do not support PET such as Basic Mode and
Terminal Mode.
[3] - 1b = User password bypass. When set to 1b, the managed client’s BIOS boots the
system and bypasses any user or boot password that might be set in the
system. This option allows a system administrator to, for example, force a
system boot via PXE in an unattended manner. It is recommended that
ADMINISTRATOR Privilege Level be required to set this bit if this feature is
supported in a given BIOS.
[2] - 1b = Lock Sleep Button. When set to 1b, directs BIOS to disable the sleep button
operation for the system, normally until the next boot cycle. Client
instrumentation might provide the capability to re-enable the button
functionality without rebooting.
[1:0] - 00b = console redirection occurs per BIOS configuration setting
01b = suppress (skip) console redirection if enabled
10b = request console redirection be enabled
11b = reserved
data 4
[7:4] - reserved
[3]
[3] - BIOS Shared Mode Override
Can be used to request BIOS to temporarily place the channel into Shared access mode.
Per the recommendations in Table 13-2, Serial Port Sharing Access CharacteristicsTable
13-2, Serial Port Sharing Access Characteristics, ‘Shared’ access would cause the
baseboard serial controller to both remain enabled after POST/start of OS boot, while also
allowing the BMC to be accessible. This can be useful when booting to an alternative
device such as a Diagnostic Partition since it means the partition can use the serial port
but that communication with the BMC can remain available if the partition software fails.
Note: BIOS should only pay attention this field if when the ‘valid’ flag is set and the
‘BIOS/POST has handled boot info’ flag is set.
1b = Request BIOS to temporarily set the access mode for the channel specified in
parameter #6 to ‘Shared’. This is typically accomplished by sending a ‘Set
Channel Access’ command to set the volatile access mode setting in the
[4]
BMC .
0b = No request to BIOS to change present access mode setting.
[2:0] - BIOS Mux Control Override
Can be used to request BIOS to force a particular setting of the serial/modem mux at the
conclusion of POST / start of OS boot. This override takes precedence over the mux
settings for the access mode even if the BIOS Shared Mode Override is set.
Note: BIOS should only pay attention this field if when the ‘valid’ flag is set and
BIOS/POST has handled boot info flag is set.
000b = BIOS uses recommended setting of the mux at the end of POST (See
Table 13-2, Serial Port Sharing Access CharacteristicsTable 13-2, Serial
Port Sharing Access Characteristics for more info.)
001b = Requests BIOS to force mux to BMC at conclusion of POST/start of OSboot. If honored, this will override the recommended setting of the mux at
the end of POST (See Table 13-2, Serial Port Sharing Access
CharacteristicsTable 13-2, Serial Port Sharing Access Characteristics for
more info.)
010b = Requests BIOS to force mux to system at conclusion of POST/start of OSboot. If honored, this will override the recommended setting of the mux at
the end of POST (See Table 13-2, Serial Port Sharing Access
CharacteristicsTable 13-2, Serial Port Sharing Access Characteristics for
more info.)
data 5 - reserved
286
Intelligent Platform Management Interface Specification
boot initiator info
[1]
(semi-volatile)
6
Address & Identity information for the party that initiated the boot. The party that initiates
the boot writes this parameter and the boot info acknowledge parameter prior to issuing
the command that causes the system power up, power cycle, or reset. This data is
normally written by the remote console application, not the BMC.
boot source
data 1- Channel Number. Channel that will deliver the boot command (e.g. chassis
control). BIOS and boot software (e.g. service partition or OS loader) can use the
Get Channel Sessions to find out information about the party that initiated the boot.
[7:4] - reserved
[3:0] - Channel Number
data 2:5 - Session ID. Session ID for session that the boot command will be issued over.
This value can be used with the Get Channel Sessions command to find out
information about the party that initiated the boot.
boot initiator mailbox
[1][2]
(semi-volatile)
7
data 6:9 - Boot Info Timestamp. This timestamp is used to help software determine
whether the boot information is ‘stale’ or not. A service partition or OS loader may
elect to ignore the boot information if it is older than expected.
The boot initiator should load this field with the timestamp value from the Get SEL
Time command prior to issuing the command that initiates the boot.
This parameter is used as a ‘mailbox’ for holding information that directs the operation of
the OS loader or service partition software. The data content is specified by the software
vendor.
Note: Since this information will be retained by the BMC and may be readable by other
software entities, care should be taken to avoid using it to carry ‘secret’ data.
data1: Set Selector = block selector
Selects which 16-byte info block to access. 0-based.
data 2:(17) block data
The first three bytes of block #0 are required to be an IANA Enterprise ID Number (least
significant byte first) for the company or organization that has specified the loader.
Up to 16-bytes per block of information regarding boot initiator, based on protocol and
medium.
An implementation is required to support at least 80-bytes (five blocks) of storage for this
command. Previous values are overwritten. The BMC does not automatically clear any
remaining data bytes if fewer than 16 bytes are written to a given block.
OEM Parameters
(optional. Non-volatile
or volatile as specified
by OEM)
1.
2.
3.
4.
96:127
This range is available for special OEM configuration parameters. The OEM is
identified according to the Manufacturer ID field returned by the Get Device ID
command.
The designation ‘semi-volatile’ means that the parameter will be kept across system power cycles, resets, system power
on/off, and sleep state changes, but will not be preserved if the management controller loses standby power or is cold reset.
Parameters designated as ‘semi-volatile’ are initialized to 0’s upon controller power up or hard reset, unless otherwise
specified.
IPMI allows software to use the boot initiator mailbox as a way for a remote application to pass OEM parameters for
additional selection of the boot process and direction of the startup of post-boot software. If additional parameters are not
included, the system boots the primary/first-scanned device of the type specified.
When BIOS temporarily changes the access mode to ‘Shared’, the BMC should operate according to the description for that
mode provided in Table 13-2, Serial Port Sharing Access CharacteristicsTable 13-2, Serial Port Sharing Access
Characteristics. Because this is a volatile setting, the BMC will return to operating according to the non-volatile setting on the
next system power down or hard reset. A remote application that uses this bit should be aware of possible differences in
operation between the non-volatile setting and Shared mode. For example, the differences in answering behavior between
“Shared” mode and “Always Available” mode.
BIOS should set this access mode and, if serial port sharing is enabled, configure the system UART according to Table 13-2,
Serial Port Sharing Access CharacteristicsTable 13-2, Serial Port Sharing Access Characteristics prior to launching the load
(boot) of the operating system. It is recommended that this operation be performed as early in POST as feasible. In any case,
a remote application should be aware that the BIOS may be operating according to the non-volatile setting during a
significant portion of POST until it reaches the point where it acts on the BOOT options.
287
Intelligent Platform Management Interface Specification
22.12 Get POH Counter Command
This version of IPMI provides a specification for an optional, POH (Power-On Hours) counter. The management
controller automatically increments non-volatile storage at the specified rate whenever the system is powered up.
It is recommended that this command be implemented in the BMC to provide a standardized location for this
function.
Note that in a power-managed system, the definition of ‘powered up’ can be somewhat ambiguous. The definition
used here is that the power-on hours shall accumulate whenever the system is in the operational (S0) state. An
implementation may elect to increment power-on hours in the S1 and S2 states as well.
‘Clear’ or ‘Set’ commands are not specified for this counter. This is because the counter is most typically used for
warranty tracking or replacement purposes where changing or clearing the counter would defeat the purpose.
The following command is used for accessing the POH Counter. This command returns the present reading of the
counter, plus the number of counts per hour.
Table 22-13, Get POH Counter Command
Request Data
Response Data
byte
-
data field
-
1
Completion Code
2
Minutes per count.
3:6
Counter reading. LS Byte first.
When the system is powered down between counts, the counter either picks up incrementing at the offset at which
the power down occurred, or starts counting at 0 minutes from the last counter reading, depending on the choice
of the implementer. In any case, the time does not get ‘rounded up’ to the next count as a result of powering down
between counts.
288
Intelligent Platform Management Interface Specification
289
Intelligent Platform Management Interface Specification
23. Event Commands
The ‘Sensor/Event’ Network Function is used for device functionality related to the transmission, reception, and
handling of ‘Event Messages’ and platform sensors.
What is commonly referred to as an ‘Event Message’ is actually a Sensor/Event Message with a command byte of
‘02h’. The request is also referred to as an ‘Event Request Message’, while the corresponding response is referred to
as an ‘Event Response Message’.
The following presents the list of the Event commands under the ‘Sensor/Event’ Network Function. Refer to
Appendix G - Command Assignments for the specification of the Network Function and Command (CMD) values
and privilege levels for these commands.
Table 23-1, Event Commands
Section
Defined
Command
Set Event Receiver
Get Event Receiver
Platform Event (a.k.a. “Event Message”)
Mandatory/Optional
Event
Event
Generator
Receiver
23.1
23.2
23.3
M
M
M
O
O
M
23.1 Set Event Receiver Command
This global command tells a controller where to send Event Messages. The slave address and LUN of the Event
Receiver must be provided. A value FFh for the Event Receiver Slave Address disables Event Message
generation entirely. This command is only applicable to management controllers that act as IPMB Event
Generators.
A device that receives a ‘Set Event Receiver’ command shall ‘re-arm’ event generation for all its internal sensors.
This means internally re-scanning for the event condition, and updating the event status based on the result. This
will cause devices that have any pre-existing event conditions to transmit new event messages for those events.
⇒ A initial update in progressreading/state unavailable (formerly “initial update in progress”)
bit is provided with the Get Sensor Reading and Get Sensor Event Status commands to help
software avoid getting incorrect event status due to a re-arm. For example, suppose a
controller only scans for an event condition once every four seconds. Software that accessed
the event status using the Get Sensor Reading command could see the wrong status for up to
four seconds before the event status would be correctly updated. A controller that has slow
updates must implement the initial update in progress bitreading/state unavailable bit, and
should not generate event messages until the update has completed. Software should ignore
the Event Status bits while the initial update in progressreading/state unavailable bit is set.
Table 23-2, Set Event Receiver
byte
Request Data
1
2
Response Data
290
1
data field
Event Receiver Slave Address. 0FFh disables Event Message Generation,
Otherwise:
2
[7:1] - IPMB (I C) Slave Address
2
[0] - always 0b when [7:1] hold I C slave address
[7:2] - reserved
[1:0] - Event Receiver LUN
Completion Code
Intelligent Platform Management Interface Specification
23.2 Get Event Receiver Command
This global command is used to retrieve the present setting for the Event Receiver Slave Address and LUN. This
command is only applicable to management controllers that act as IPMB Event Generators.
Table 23-3, Get Event Receiver Command
byte
-
Request Data
Response Data
data field
-
1
Completion Code.
2
Event Receiver Slave Address. 0FFh indicates Event Message Generation
has been disabled. Otherwise:
2
[7:1] IPMB (I C) Slave Address
2
[0] always 0b when [7:1] hold I C slave address
[7:2] - reserved
[1:0] - Event Receiver LUN
3
23.3 Platform Event Message Command
This command may be thought of as a request for the BMC to process the event data that the command contains.
Typically, the data will be logged to the System Event Log (SEL). Depending on the implementation, the data
may also go to the Event Message Buffer and processed by Platform Event Filtering (PEF).
Table 23-4, Platform Event (Event Message) Command
IPMB MESSAGING
SYSTEM INTERFACE
(IPMB, LAN, Serial/Modem, PCI Mgmt. Bus)
byte
-
Request Data
Response Data
data field
Generator ID (RqSA, RqLUN)
byte
1
data field
Generator ID
1
EvMRev
2
EvMRev
2
Sensor Type
3
Sensor Type
3
Sensor #
4
Sensor #
4
Event Dir | Event Type
5
Event Dir | Event Type
5
Event Data 1
6
Event Data 1
6
Event Data 2
7
Event Data 2
7
Event Data 3
8
Event Data 3
1
Completion Code.
1
Completion Code.
The Generator ID field is a required element of an Event Request Message. For IPMB messages, this field is
equated to the Requester’s Slave Address and LUN fields. Thus, the Generator ID information is not carried in
the data field of an IPMB request message.
For ‘system side’ interfaces, it is not as useful or appropriate to ‘overlay’ the Generator ID field with the message
source address information, and so it is specified as being carried in the data field of the request.
291
Intelligent Platform Management Interface Specification
23.4 Event Request Message Fields
An Event Request Message contains the following fields for the Event Receiver, regardless of whether the
message is received from the IPMB or from a ‘system side’ messaging interface, such as the SMIC. Most of the
information is passed in the data field of the message, however, in some cases field information is extracted from
the ‘message header’.
Table 23-5, Event Request Message Fields
Field
Generator ID
EvMRev
Sensor Type
Sensor #
Event Dir
Event Type
Event Data
Description
This field identifies the device that has generated the Event Message. This is the 7-bit Requester’s
Slave Address (RqSA) and 2-bit Requester’s LUN (RqLUN) if the message was received from the
IPMB, or the 7-bit System Software ID if the message was received from system software.
One byte. Event Message Revision. This field is used to identify different revisions of the Event
Message format. The revision number shall be 04h for Event Messages that comply with the format
given in this specification. IPMI v1.0 messages use 03h. It is recommended that software be able to
interpret both versions.
One byte. Indicates the event class or type of sensor that generated the Event Message. The
Sensor Type Codes are specified in Table 36-3, Sensor Type CodesTable 36-3, Sensor Type
Codes.
One byte. A unique number (within a given sensor device) representing the ‘sensor’ within the
management controller that generated the Event Message. Sensor numbers are used for both
identification and access of sensor information, such as getting and setting sensor thresholds.
1-bit. Indicates the event transition direction. (0 = Assertion Event, 1 = Deassertion Event)
7-bits. This field indicates the type of threshold crossing or state transition (trigger) that produced the
event. This is encoded using the Event/Reading Type Code. See Section 36, Sensor and Event
Code TablesSensor and Event Code Tables.
One to three Bytes. The remainder of the Event Message data according to the class of the Event
Type for the sensor (threshold, discrete, or OEM). The contents and format of this field is specified in
Table 23-6, Event Request Message Event Data Field ContentsTable 23-6, Event Request Message
Event Data Field Contents, below.
The following illustrates which fields from the Event Request Message get transferred to the System Event
Record.
23.5 IPMB Event Message Formats
The following figure illustrates the formatting of an Event Request Message as an ‘IPMB’ message on an I2C bus,
per the Intelligent Platform Management Bus Communications Protocol v1.0.
Figure 23-1, IPMB Event Request Message Format
RsSA NetFn /RsLUN Chk1
RqSA** RqSeq /RqLUN**
Cmd=02
EvMRev Sensor Type
**
Sensor #
Event Dir
Event Type
Event Data
Chk2
These fields constitute the ‘Generator ID’ field for the Event Request Message.
Shading designates fields that are not stored in the event record.
The Event Receiver device responds to IPMB Event Request Messages by simply issuing the Event Response
Message with a single ‘Completion Code’ byte in the data field and a command code of 02h in IPMB Response
Message format.
292
Intelligent Platform Management Interface Specification
23.6 System Interface Event Request Message Format
Event Request Messages are formatted differently over the System Interface than they are over the IPMB or
interfaces that use the IPMB message format. The following figure illustrates the formatting of an Event Request
Message as it would be transmitted over the SMIC interface. This is provided for illustration purposes only. Refer
to the individual sections for the System Interfaces for more information: Section 10.16, Logging Events from
System Software via SMICLogging Events from System Software via SMIC, Section 9.4, Logging Events from
System Software via KCS InterfaceLogging Events from System Software via KCS Interface, and Section 11.5,
Logging Events from System Software via BT InterfaceLogging Events from System Software via BT Interface.
Figure 23-2, Example SMIC Event Request Message Format
NetFn /00
Cmd=02
7-bit Software ID** 1
EvMRev Sensor Type Sensor #
**
Event Dir
Event Type
Event Data
This field constitutes the ‘Generator ID’ field for the Event Request Message.
Shading designates fields that are not stored in the event record.
293
Intelligent Platform Management Interface Specification
23.7 Event Data Field Formats
The contents of the Event Data field in an Event Request Message (Event Message) is dependent on the sensor
class of the sensor. The sensor class obtained from the Event/Reading Type Code specifies whether the sensor
event is threshold based, discrete, or OEM defined. Each Event Type is associated with a sensor class. An
application can extract the sensor class, and determine the corresponding Event Data format, from the
Event/Reading Type Code that was received in the Event Type field in the Event Message. See section 36.1,
Event/Reading Type CodesEvent/Reading Type Codes, for more information.
Table 23-6, Event Request Message Event Data Field Contents
Sensor
Class
threshold
Event Data
Event Data 1
[7:6] -
[5:4] -
[3:0] -
discrete
00b = unspecified byte 2
01b = trigger reading in byte 2
10b = OEM code in byte 2
11b = sensor-specific event extension code in byte 2
00b = unspecified byte 3
01b = trigger threshold value in byte 3
10b = OEM code in byte 3
11b = sensor-specific event extension code in byte 3
Offset from Event/Reading Code for threshold event.
Event Data 2 reading that triggered event, FFh or not present if unspecified.
Event Data 3 threshold value that triggered event, FFh or not present if unspecified. If present, byte 2 must be
present.
Event Data 1
[7:6] -
[5:4] -
[3:0] -
00b = unspecified byte 2
01b = previous state and/or severity in byte 2
10b = OEM code in byte 2
11b = sensor-specific event extension code in byte 2
00b = unspecified byte 3
01b = reserved
10b = OEM code in byte 3
11b = sensor-specific event extension code in byte 3
Offset from Event/Reading Code for discrete event state
Event Data 2
[7:4] [3:0] -
Optional offset from ‘Severity’ Event/Reading Code. (0Fh if unspecified).
Optional offset from Event/Reading Type Code for previous discrete event state. (0Fh if unspecified.)
Event Data 3 Optional OEM code. FFh or not present if unspecified.
OEM
Event Data 1
[7:6] -
[5:4] -
[3:0] -
00b = unspecified in byte 2
01b = previous state and/or severity in byte 2
10b = OEM code in byte 2
11b = reserved
00b = unspecified byte 3
01b = reserved
10b = OEM code in byte 3
11b = reserved
Offset from Event/Reading Type Code
Event Data 2
[7:4] [3:0] -
Optional OEM code bits or offset from ‘Severity’ Event/Reading Type Code. (0Fh if unspecified).
Optional OEM code or offset from Event/Reading Type Code for previous event state. (0Fh if unspecified).
Event Data 3 Optional OEM code. FFh or not present or unspecified.
O/M = Optional/Mandatory. Mandatory indicates that the byte must be present in all messages. Optional bytes may be left
out of messages, as specified. If an optional byte is not present, the Event Receiver shall substitute the value FFh in the
corresponding Event Data byte position when transferring the information to the System Event Log function.
294
Intelligent Platform Management Interface Specification
24. PEF and Alerting Commands
This section describes the formats of the commands related to configuring and controlling the Platform Event
Filtering (PEF) and Alerting capabilities. None of the commands in the following table are required unless PEF or
Alerting is supported. Refer to Appendix G - Command Assignments for the specification of the Network Function
and Command (CMD) values and privilege levels for these commands.
Table 24-1, PEF and Alerting Commands
Command
Get PEF Capabilities
Arm PEF Postpone Timer
Set PEF Configuration Parameters
Get PEF Configuration Parameters
Set Last Processed Event ID
Get Last Processed Event ID
Alert Immediate
PET Acknowledge
1.
2.
3.
Section
Defined
O/M
24.1
24.2
24.3
24.4
24.5
24.6
24.7
24.8
M
[1]
M
[1]
M
[1]
M
[1]
M
[1]
M
[2]
O
[3]
O
[1]
Mandatory if PEF or Alerting is supported
Mandatory if Alerting is supported
Mandatory if LAN or PPP Alerting is supported
24.1 Get PEF Capabilities Command
This command returns the information about the implementation of PEF on the BMC.
Table 24-2, Get PEF Capabilities Command
byte
Request Data
Response Data
1
2
3
4
data field
Completion Code
PEF Version (BCD encoded, LSN first, 51h for this specification. 51h Æ
version 1.5)
Action Support
[7:6] - reserved
[5] - 1b = diagnostic interrupt
[4] - 1b = OEM action
[3] - 1b = power cycle
[2] - 1b = reset
[1] - 1b = power down
[0] - 1b = Alert
Number of event filter table entries (1 based)
24.2 Arm PEF Postpone Timer Command
This command is used by software to enable and arm the PEF Postpone Timer. The command can also be used by
software to disable PEF indefinitely during run-time. Once enabled, the timer automatically starts counting down
whenever the last software-processed event Record ID is for a record that is not equal to the most recent (last)
SEL record. The countdown will begin immediately if the Record IDs are already different when the timer is
armed.
In order to keep the PEF Postpone Timer from expiring, software must use the Set Last Processed Event ID
command to update the last software-processed Record ID to match the value for the last SEL record. This will
cause the BMC to stop the timer and rearm it to start counting down from the value that was passed in the Arm
PEF Postpone Timer command.
295
Intelligent Platform Management Interface Specification
The Get Last Processed Event ID command can be used to retrieve the present value for the last SEL record’s
Record ID, the last BMC-processed Record ID, and the last software-processed Record ID.
Table 24-3, Arm PEF Postpone Timer Command
byte
Request Data
1
Response Data
1
2
data field
[7:0] - PEF Postpone Timeout, in seconds. 01h Æ 1 second.
00h = disable Postpone Timer (PEF will immediately handle events, if
enabled). The BMC automatically disables the timer whenever the
system enters a sleep state, is powered down, or reset.
01h - FDh = arm timer. Timer will automatically start counting down from given
value when the last-processed event Record ID is not equal to the last
received event’s Record ID.
FEh = Temporary PEF disable. The PEF Postpone timer does not countdown
from the value. The BMC automatically re-enables PEF (if enabled in
the PEF configuration parameters) and sets the PEF Postpone timeout
to 00h whenever the system enters a sleep state, is powered down, or
reset. Software can cancel this disable by setting this parameter to 00h
or 01h-FDh.
FFh = get present countdown value
Completion Code
Present timer countdown value
24.3 Set PEF Configuration Parameters Command
This command is used for setting parameters such as PEF enable/disable and for entering the configuration of the
Event Filter table and the Alert Strings.
Table 24-4, Set PEF Configuration Parameters Command
byte
Request Data
1
2:N
Response Data
296
1
data field
Parameter selector
[7] - reserved
[6:0] - Parameter selector
Configuration parameter data, per Table 24-6, PEF Configuration Parameters.
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
81h = attempt to set the ‘set in progress’ value (in parameter #0) when not in
the ‘set complete’ state. (This completion code provides a way to
recognize that another party has already ‘claimed’ the parameters)
82h = attempt to write read-only parameter
Intelligent Platform Management Interface Specification
24.4 Get PEF Configuration Parameters Command
This command is used for retrieving the configuration parameters from the Set PEF Configuration command.
Table 24-5, Get PEF Configuration Parameters Command
byte
Request Data
1
2
3
Response Data
1
2
3:N
data field
[7] -
1b = get parameter revision only.
0b = get parameter
[6:0] - Parameter selector
Set Selector (00h if parameter does not require a Set Selector)
Block Selector (00h if parameter does not require a block number)
Completion Code. Generic plus the following command-specific completion
codes:
80h = parameter not supported.
[7:0] - Parameter revision.
Format: MSN = present revision. LSN = oldest revision parameter is
backward compatible with. 11h for parameters in this specification.
The following data bytes are not returned when the ‘get parameter revision
only’ bit is 1b.
Configuration parameter data, per Table 24-6, PEF Configuration Parameters.
If the rollback feature is implemented, the BMC makes a copy of the existing
parameters when the ‘set in progress’ state becomes asserted (See the Set In
Progress parameter #0). While the ‘set in progress’ state is active, the BMC
will return data from this copy of the parameters, plus any uncommitted
changes that were made to the data. Otherwise, the BMC returns parameter
data from non-volatile storage.
Table 24-6, PEF Configuration Parameters
Parameter
#
Set In Progress
(volatile)
0
Parameter Data
data 1 - This parameter is used to indicate when any of the following
parameters are being updated, and when the updates are completed.
The bit is primarily provided to alert software than some other software or
utility is in the process of making changes to the data.
An implementation can also elect to provide a ‘rollback’ feature that uses
this information to decide whether to ‘roll back’ to the previous
configuration information, or to accept the configuration change.
If used, the roll back shall restore all parameters to their previous state.
Otherwise, the change shall take effect when the write occurs.
[7:2] - reserved
[1:0] - 00b = set complete. If a system reset or transition to powered
down state occurs while ‘set in progress’ is active, the
BMC will go to the ‘set complete’ state. If rollback is
implemented, going directly to ‘set complete’ without first
doing a ‘commit write’ will cause any pending write data to
be discarded.
01b = set in progress. This flag indicates that some utility or other
software is presently doing writes to parameter data. It is a
notification flag only, it is not a resource lock. The BMC
does not provide any interlock mechanism that would
prevent other software from writing parameter data while.
10b = commit write (optional). This is only used if a rollback is
implemented. The BMC will save the data that has been
written since the last time the ‘set in progress’ and then go
to the ‘set in progress’ state. An error completion code will
be returned if this option is not supported.
11b = reserved
297
Intelligent Platform Management Interface Specification
Parameter
#
PEF control
(non-volatile)
1
PEF Action global control
(non-volatile)
2
PEF Startup Delay
(optional, non-volatile)
3
Parameter Data
data 1
[7:4] - reserved
[3] PEF Alert Startup Delay disable. (optional)
1b = enable PEF Alert Startup delay
0b = disable PEF startup delay.
[2] PEF Startup Delay disable. (optional)
An implementation that supports this bit should also provide
a mechanism that allows the user to Disable PEF in case
the filter entries are programmed to cause an ‘infinite loop’
of PEF actions (such as system resets or power cycles)
when the PEF startup delay is disabled. If this bit is not
implemented the PEF startup delay must always be
enabled.
1b = enable PEF startup delay on manual (pushbutton) system
power-ups (from S4/S5) and system resets (including
system resets initiated by PEF).
0b = disable PEF startup delay.
[1] 1b = enable event messages for PEF actions. If this bit is set,
each action triggered by a filter will generate an event
message for the action. These allow the occurrence of PEFtriggered actions to be logged (if event logging is enabled).
The events are logged as System Event Sensor 12h, offset
04h. See Table 36-3, Sensor Type CodesTable 36-3,
Sensor Type Codes.) These event messages are also
subject to PEF.
0b = disable event messages for PEF actions.
[0] 1b = enable PEF.
0b = disable PEF.
data 1
[7:6] - reserved
[5] 1b = enable diagnostic interrupt
[4] 1b = enable OEM action
[3] 1b = enable power cycle action (No effect if power is already off)
[2] 1b = enable reset action
[1] 1b = enable power down action
[0] 1b = enable Alert action
data 1 - time to delay PEF after a system power-ups (from S4/S5) and
resets. Default = 60 seconds. If this parameter is not provided, the
default PEF Startup Delay must be implemented. Enable/disable of the
delay is configured using the PEF Control parameter, above. If this
parameter is supported, a 00h value can also be used to disable the
delay if necessary. See Section 15.4, PEF Startup DelayPEF Startup
Delay, for more information.
Note: An implementation that supports this parameter should also
provide a mechanism that allows the user to Disable PEF in case the
filter entries are programmed to cause an ‘infinite loop’ of PEF actions
under the situation where this parameter is set to too short an interval to
allow a user to locally disable PEF. An implementation is allowed to force
this parameter to a minimum, non-zero value.
PEF Alert Startup Delay
(optional, non-volatile)
4
PEF Startup Delay
[7:0] - PEF Startup Delay in seconds, +/- 10%. 1-based. 00h = no delay.
data 1 - time to delay Alerts after system power-ups (from S4/S5) and
resets. Default = platform-specific. 60-seconds typical, though may be
longer on systems that require more startup time before user can take
action to disable PEF. If this parameter is not provided, a default PEF
Startup Delay, appropriate for the platform, must be implemented.
Enable/disable of the delay can also be optionally configured using the
PEF Control parameter, above. An implementation can separately
implement this parameter and/or the enable/disable bit.
PEF Alert Delay
[7:0] - PEF Alert Startup Delay in seconds, +/- 10%. 1-based.
00h = no delay.
298
Intelligent Platform Management Interface Specification
Parameter
#
Number of Event Filters
(READ ONLY)
5
Event Filter Table, (nonvolatile)
6
Event Filter Table Data 1
(non-volatile)
7
Number of Alert Policy
Entries
(READ ONLY)
8
Alert Policy Table
(non-volatile)
9
System GUID
(non-volatile)
10
Number of Alert Strings
(READ ONLY)
11
Alert String Keys
(volatile) & (non-volatile) see description
12
Parameter Data
Number of event filters supported. 1-based. This parameter does not
need to be supported if Alerting is not supported.
[7] - reserved
[6:0] - number of event filter entries. 0 = alerting not supported.
data 1 - Set Selector = filter number.
[7] reserved.
[6:0] - Filter number. 1-based. 00h = reserved.
data 2:21 - filter data
This parameter provides an aliased access to the first byte of the event
filter data. This is provided to simplify the act of enabling and disabling
individual filters by avoiding the need to do a read-modify-write of the
entire filter data.
data 1 - Set Selector = filter number
[7] - reserved
[6:0] - Filter number. 1-based. 00h = reserved.
data 2 - data byte 1 of event filter data
Number of alert policy entries supported. 1-based. This parameter does
not need to be supported if Alerting is not supported.
[7] reserved
[6:0] - number of alert policy entries. 0 = alerting not supported.
data 1 - Set Selector = entry number
[7] reserved
[6:0] - alert policy entry number. 1-based.
data 2:4 - entry data
data 1 Used to fill in the GUID field in a PET Trap. Stored per Table
17-10, GUID FormatTable 17-10, GUID Format.
[7:1] - reserved
[0] - 1b = BMC uses following value in PET Trap.
0b = BMC ignores following value and uses value returned from
Get System GUID command instead.
2:17 - System GUID
Number of alert strings supported in addition to Alert String 0. 1-based.
This parameter does not need to be supported if Alerting is not
supported.
[7] - reserved
[6:0] - number of alert strings.
Sets the keys used to look up Alert String data in PEF. This parameter
does not need to be supported if Alerting is not supported.
data 1 - Set Selector = Alert string selector.
[7] reserved.
[6:0] - string selector.
0 = selects volatile string parameters
01h-7Fh = non-volatile string selectors
PEF uses the following Event Filter Number and the Alert String Key
fields to look up the string associated with a particular event. String 0 is a
special, volatile string reserved for use by the Alert ImmediateAlert
Immediate command.
The following two fields are used by PEF to look up a particular Alert
String based on information obtained from the alert policy entry. The
fields should typically be set to 0’s (unspecified) for string selector 0. PEF
will scan the values for string 0 when doing a look up, so the string 0
values can be set to non-zero values for PEF testing/debug purposes in
order to avoid writes to non-volatile storage.
data 2 - Event Filter Number
[7] reserved.
[6:0] - Filter number. 1-based. 00h = unspecified.
data 3 - Alert String Set
[7] reserved
[6:0] - Set number for string. 1-based. 00h = unspecified.
299
Intelligent Platform Management Interface Specification
Parameter
Alert Strings
(volatile) &
(non-volatile) - see
description.
#
13
Parameter Data
Sets the Alert String data. The string data that should be used is
dependent on the Channel and Alert Type. This parameter does not need
to be supported if Alerting is not supported.
For Dial paging, the BMC automatically follows the string with a <CR>
(carriage return) character when sending it to the modem.
For TAP paging the string corresponds to ‘Field 2’, the Pager Message.
Note that while the string accepts 8-bit ASCII data, the TAP
implementation only supports 7-bit ASCII.
th
The BMC shall automatically zero the 8 bit when transmitting the string
during TAP paging.
String 0 is a special, volatile string reserved for use by the Alert
ImmediateAlert Immediate command.
Set Selector = string selector.
data 1 [7] reserved.
[6:0] string selector.
0 = selects volatile string
01h-7Fh = non-volatile string selectors
Block Selector = string block number to set, 1 based. Blocks
data 2 are 16 bytes.
data 3:N -
Number of Group Control
Table entries
(optional. Present if BMC
supports automatic ICMB
Group Power Control.
See ICMB specification
for details.)
Group Control Table
(optional, non-volatile.
Present if BMC supports
automatic ICMB Group
Power Control. See ICMB
specification for details.)
14
data 1 -
15
data 1 [7] [6:0] -
String data. Null terminated 8-bit ASCII string. 16-bytes max.
per block.
Number of group control table entries. 1-based (4 min, 8
max)
Set Selector = group control table entry selector.
reserved.
group control table entry selector.
data 2 [7:6] reserved
[5] Request/Force
0b = request control operation. A requested operation will only
complete once the same operation has been requested for
all control groups and all enabled control members for the
given chassis.
1b = force control operation. A forced operation will occur
regardless of whether the same operation has been
requested for all control groups and all enabled control
membership for the given chassis.
[4] Immediate/Delayed. Selects whether the BMC requests an
immediate or delayed control operation. Note: whether this
operation is initiated at the time the command is received is
dependent on the request/force bit, see above.
0b = immediate control. BMC sends command that requests an
immediate control operation.
1b = delayed control. BMC sends control command to request a
delayed control operation. This is conditioned by the
request/force bit.
[3:0] Channel Number (channel number for ICMB that group
control operation is to be delivered over)
data 3: Group ID 0 (1-based)
00h = unspecified
FFh = all groups
data 4: Member ID 0 (0-based)
300
Intelligent Platform Management Interface Specification
Parameter
#
Parameter Data
[7:5] reserved
[4] 0b = enable member ID check.
[1]
1b = disable member ID check .
[3:0] member ID. ID of this chassis within specified group. (value
is ignored if Group ID 0 = FFh)
data 5: Group ID 1 (1-based)
00h = unspecified
FFh = all groups
data 6: Member ID 1 (0-based)
[7:5] reserved
[4] 0b = enable member ID check.
[1]
1b = disable member ID check .
[3:0] member ID. ID of this chassis within specified group. (value
is ignored if Group ID 1 = FFh)
data 7: Group ID 2 (1-based)
00h = unspecified
FFh = all groups
data 8: Member ID 2 (0-based)
[7:5] reserved
[4] 0b = enable member ID check.
[1]
1b = disable member ID check .
[3:0] member ID. ID of this chassis within specified group. (value
is ignored if Group ID 2 = FFh)
data 9: Group ID 3 (1-based)
00h = unspecified
FFh = all groups
data 10: Member ID 3 (0-based)
[7:5] reserved
[4] 0b = enable member ID check.
[1]
1b = disable member ID check .
[3:0] member ID. ID of this chassis within specified group. (value
is ignored if Group ID 3 = FFh)
OEM Parameters
(optional. Non-volatile or
volatile as specified by
OEM)
96:127
data 11: - Retries and Operation
[7] reserved
[6:4] number of times to retry sending the command to perform
the group operation [For ICMB, the BMC broadcasts a
Group Chassis Control command] (1-based)
[3:0] - operation
0h = power down. Force system into soft off (S4/S45) state.
This is for ‘emergency’ management power down actions.
The command does not initiate a clean shut-down of the
operating system prior to powering down the system.
1h = power up.
2h = power cycle (optional). This command provides a power off
interval of at least 1 second.
3h = hard reset. Some systems may accept this option even if
the system is in a state (e.g. powered down) where resets
are unavailable.
4h = pulse Diagnostic Interrupt. (optional) Pulse a version of a
diagnostic interrupt that goes directly to the processor(s).
This is typically used to cause the operating system to do
a diagnostic dump (OS dependent). The interrupt is
commonly an NMI on IA-32 systems and an INIT on Intel®
Itanium™ processor based systems.
5h = Initiate a soft-shutdown of OS via ACPI by emulating a
fatal overtemperature. (optional)
This range is available for special OEM configuration parameters. The
OEM is identified according to the Manufacturer ID field returned by the
Get Device ID command.
301
Intelligent Platform Management Interface Specification
24.5 Set Last Processed Event ID Command
This command is used to set the Record ID for the last event that was processed by system software. For test and
debug purposes, it can also be used to set the Record ID for the last event processed by the BMC. See sections
15.3, PEF Postpone TimerPEF Postpone Timer and 15.4.1, Last Processed Event TrackingLast Processed Event
Tracking for more information. The Last Processed Event ID value is automatically set to FFFFh whenever the
SEL is cleared using the Clear SEL command. If the Delete SEL Entry command is used to either clear the SEL
or delete the last event, software must set the Last Processed event manually by using the Set Last Processed
Event ID command.
Of the two Record IDs (software-processed or BMC-processed) PEF uses the Record ID for the most recent event
that was added to the SEL as the indicator of events that have yet to be processed. Both the last BMC-processed
and last software-processed IDs are kept in NV storage.
Table 24-7, Set Last Processed Event ID Command
byte
Request Data
1
2:3
Response Data
1
data field
[7:1] - reserved.
[0] - 0b = set Record ID for last record processed by software.
1b = set Record ID for last record processed by BMC.
Record ID. LS-byte first.
Completion Code
81h = cannot execute command, SEL erase in progress
24.6 Get Last Processed Event ID Command
This command is used to retrieve the Record ID for the last event that was processed by system software and the
BMC. See sections 15.3, PEF Postpone TimerPEF Postpone Timer and 15.4.1, Last Processed Event
TrackingLast Processed Event Tracking for more information.
Table 24-8, Get Last Processed Event ID Command
byte
Request Data
Response Data
data field
-
-
1
Completion Code
81h = cannot execute command, SEL erase in progress
Most recent addition timestamp. LS byte first.
Record ID for last record in SEL. Returns FFFFh if SEL is empty.
Last SW Processed Event Record ID.
Last BMC Processed Event Record ID. Returns 0000h when event has been
processed but could not be logged because the SEL is full or logging has
been disabled.
2:5
6:7
8:9
10:11
24.7 Alert Immediate Command
This command is used to send an alert to the destination specified by the destination selector. The kind of alert
that will be sent is determined by Destination Type associated with the destination. Alerts that are initiated via
this command are never logged as events. This command is to support utilities or BIOS setup options that allow
the user to test their alerting configuration for a given destination. The command can also be used by system
software as a run-time mechanism to trigger the delivery of an alert.
These alerts are not subject to the Page Blackout intervals, although an alert must complete before the next Alert
ImmediateAlert Immediate command will be accepted. Alert ImmediateAlert Immediate commands are also
rejected with an error completion code if an IPMI messaging session or automatic page is already in progress.
302
Intelligent Platform Management Interface Specification
Table 24-9, Alert Immediate Command
Byte
Request Data
1
data field
Channel number. (This value is required to select which configuration
parameters are to be used to send the Alert.)
[7:4] - reserved
[3:0] - Channel number.
Note: BMC stores the ‘Alert immediate status’ for each channel that
can send alert.
2
Destination Selector/ Operation
[7:6] - Operation
00b = Initiate alert
01b = Get Alert Immediate status
10b = Clear Alert Immediate status (sets status to 00h)
11b = reserved
[5:4] - Reserved
[3:0] - destination selector. Selects which alert destination should go to.
0h = use volatile destination info. 1h-Fh = non-volatile destination.
Note: If Operation is ‘Get Alert Immediate status’ or ‘Clear Alert
Immediate Status’ bits [3:0] are reserved.
3
Alert String Selector
Selects which Alert String, if any, to use with the alert.
[7] -
0b = don’t send an Alert String
1b = send Alert String identified by following string selector.
[6:0] - string selector.
000_0000b = use volatile Alert String.
01h-7Fh = non-volatile string selector.
Response Data
1
Completion Code. Generic codes, plus following command-specific
completion codes:
81h = Alert Immediate rejected due to alert already in progress.
82h = Alert Immediate rejected due to IPMI messaging session active on this
channel.
Following byte is only returned when Operation in request is set to “Get Alert
Immediate status”
2
Alert Immediate Status
SMS can poll this status to determine present state of the immediate
alert.
00h = No status.
Note: A BMC implementation is allowed (but not required) to abort
the Alert Immediate command due to a channel parameter
configuration, power, or reset state changes that occur while the
Alert Immediate command is being processed. In which case the
BMC will return the ‘no status’ state.
01h = Alert was Normal End. This will also be returned if one or more
attempts failed, but the last attempt was successful.
02h = “Call Retry” (Dial connection) retries failed.
03h = Alert failed due to timeouts waiting for acknowledge on all retries.
FFh = Alert by this command is in progress. Status pending.
303
Intelligent Platform Management Interface Specification
24.8 PET Acknowledge Command
This message is used to acknowledge a Platform Event Trap (PET) alert. PET alerts are SNMP Traps that are
delivered by LAN or PPP alerting, see [PET] for more info. The PET Acknowledge message is an IPMI Request
Message that is sent by the remote console that has received the trap.
Note: The PET Acknowledge command does not require that an IPMI Messaging session be established with the
BMC. It is in the same class as the Get Channel Authentication Capabilities command. In addition, if Alerting is
enabled and the configuration parameters for the Alert Destination require the PET Alert to be acknowledged, the
BMC will wait for and accept the PET Acknowledge command until the selected retry interval has expired, even if
IPMI Messaging is not available according to the present Access Mode for the channel. For systems using Serial
Port Sharing, the BMC will stay switched to the serial connector while waiting for the PET Acknowledge.
Table 24-10, PET Acknowledge Command
byte
Request Data
1:2
3:6
7
8
9
10:12
Response Data
1
1.
304
data field
Sequence Number. Value from the Sequence Number field of the PET. LS[1]
byte first .
Local Timestamp. Value from the Local Timestamp field of the PET. LS-byte
[1]
first .
Event Source type. From corresponding field in the PET.
Sensor Device. From corresponding field in the PET.
Sensor Number. From corresponding field in the PET.
Event Data 1:3. From corresponding field in the PET.
Completion Code.
Note: The sequence number and local timestamp fields in the actual PET on the network are
in network byte order, therefore filling in these values may require software to re-order the
bytes as they get them from the trap.
Intelligent Platform Management Interface Specification
25. System Event Log (SEL)
The System Event Log is a non-volatile repository for system events and certain system configuration information.
The device that fields the commands to access the SEL is referred to as the System Event Log Device or SEL Device.
Event Message information is normally written into the SEL after being received by the Event Receiver
functionality in the Event Receiver Device.
The SEL Device commands are structured in such a way that the SEL Device could actually be separated from the
Event Receiver Device. In which case it would be the responsibility of the Event Receiver Device to send the
appropriate ‘Add SEL Entry’ message directly to the SEL Device, or to pass the equivalent request through an
intermediary.
SEL Entries have a unique ‘Record ID’ field. This field is used for retrieving log entries from the SEL. SEL reading
can be done in a ‘random access’ manner. That is, SEL Entries can be read in any order assuming that the Record
ID is known.
SEL Record IDs 0000h and FFFFh are reserved for functional use and are not legal ID values. Record IDs are
handles. They are not required to be sequential or consecutive. Applications should not assume that SEL Record IDs
will follow any particular numeric ordering.
SEL Records are kept as an ordered list. That is, appending and deleting individual entries does not change the
access order of entries that precede or follow the point of addition or deletion.
25.1 SEL Device Commands
The following table summarizes the commands that are required for implementing a System Event Log device.
Note that this specification allows the System Event Log device to be implemented as a separate device from the
Event Receiver and Event Generator devices. If this is done, it is up to the implementer to create the method by
which Event Messages are passed from the Event Receiver Device to the System Event Log Device. Refer to
Appendix G - Command Assignments for the specification of the Network Function and Command (CMD) values
and privilege levels for these commands.
Table 25-1, SEL Device Commands
Command
Get SEL Info
Get SEL Allocation Info
Reserve SEL
Get SEL Entry
Add SEL Entry
Partial Add SEL Entry
Delete SEL Entry
Clear SEL
Get SEL Time
Set SEL Time
Get Auxiliary Log Status
Set Auxiliary Log Status
1.
2.
Section
O/M
25.2
25.3
25.4
25.5
25.6
25.7
25.8
25.9
25.10
25.11
25.12
25.13
M
O
[1]
O
M
[2]
M
[2]
M
O
M
M
M
O
[3]
O
Mandatory if multiple entities have overlapping access to the SEL. If system
mechanisms or conventions are defined that preclude this operation, then this
command is optional.
Either Add SEL Entry or Partial Add SEL Entry must be provided. Providing both is
optional.
305
Intelligent Platform Management Interface Specification
3.
Set Auxiliary Log Status cannot be implemented without also supporting Get Auxiliary
Log Status. However, Get Auxiliary Log Status is allowed to be implemented without
Set Auxiliary Log Status.
25.2 Get SEL Info Command
This command returns the number of entries in the SEL, SEL command version, and the timestamp for the most
recent entry and delete/clear. The timestamp format is provided in section 31,Timestamp FormatTimestamp
Format. The Most Recent Addition timestamp field returns the timestamp for the last add or log operation, while
the Most Recent Erase field returns the timestamp for the last delete or clear operation.
These timestamps are independent of timestamps that may be returned by other commands, such as those returned
by the Get SDR Repository Info command. The timestamp reflects when the most recent SEL add or erase
occurred, not when the last add or erase occurred on the physical storage device.
For example, the SEL Info Most Recent Addition timestamp would reflect the last time a new event was added to
the SEL. This would be independent of the Most Recent Addition time for an SDR - even if the implementation
elected to implement the SEL and SDR Repository in the same storage device.
Table 25-2, Get SEL Info Command
Request Data
Response Data
byte
1
3
Completion Code
81h = cannot execute command, SEL erase in progress
SEL Version - version number of the SEL command set for this SEL Device.
51h for this specification.
(BCD encoded).BCD encoded with bits 7:4 holding the Least Significant
digit of the revision and bits 3:0 holding the Most Significant bits.
Entries LS Byte - number of log entries in SEL, LS Byte
4
Entries MS Byte - number of log entries in SEL, MS Byte
2
5:6
7:10
11:14
15
306
data field
-
Free Space in bytes, LS Byte first. FFFFh indicates 65535 or more bytes of
free space are available.
Most recent addition timestamp. LS byte first.
Returns FFFF_FFFFh if no SEL entries have ever been made or if a
component update or error caused the retained value to be lost.
Most recent erase timestamp. Last time that one or more entries were
deleted from the log. LS byte first.
Operation Support
[7] - Overflow Flag. 1=Events have been dropped due to lack of space in
the SEL.
[6:4] - reserved. Write as 000
[3] - 1b = Delete SEL command supported
[2] - 1b = Partial Add SEL Entry command supported
[1] - 1b = Reserve SEL command supported
[0] - 1b = Get SEL Allocation Information command supported
Intelligent Platform Management Interface Specification
25.3 Get SEL Allocation Info Command
Returns the number of possible allocation units, the amount of usable free space (in allocation units), the
allocation unit size (in bytes), and the size of the largest contiguous free region (in allocation units). The
‘allocation unit size’ is the number of bytes in which storage is allocated. For example, if a 16-byte record is to be
added, and the SEL has a 32-byte allocation unit size, then the record would take up 32-bytes of storage.
The SEL implementation shall, at a minimum, support an allocation unit size of ≥16 bytes.
Table 25-3, Get SEL Allocation Info Command
byte
Request Data
Response Data
-
data field
-
1
Completion Code
2
3
Number of possible allocation units, LS Byte
Number of possible allocation units, MS Bytes
This number indicates whether the total number of possible allocation units is
equal to, or some number less than the log size divided by the allocation unit
size.
0000h indicates ‘unspecified’.
Allocation unit size in bytes, LS Byte. 0000h indicates ‘unspecified’.
Allocation unit size in bytes, MS byte.
Number of free allocation units, LS Byte
Number of free allocation units, MS Byte
Largest free block in allocation units, LS Byte
Largest free block in allocation units, MS Byte
Maximum record size in allocation units.
4
5
6
7
8
9
10
25.4 Reserve SEL Command
This command is used to set the present ‘owner’ of the SEL, as identified by the Software ID or by the
Requester’s Slave Address from the command. The reservation process provides a limited amount of protection
on repository access from the IPMB when records are being deleted or incrementally read.
The Reserve SEL command is provided to help prevent deleting the wrong record when doing deletes, to provide
a mechanism to avoid clearing the SEL just after a new event has been received, and to prevent receiving
incorrect data when doing incremental reads.
The Reserve SEL command does NOT guarantee access to the SEL. That is, the case exists that a pair of
requesters could vie for access to the SEL in such a manner that they alternately cancel the reservation that is held
by the other - effectively ‘deadlocking’ each other.
A ‘Reservation ID’ value is returned in response to this command. This value is required in other requests, such
as the ‘Clear SEL’ command. These commands will not execute unless the correct Reservation ID value is
provided.
The Reservation ID is used in the following manner. Suppose an application wishes to clear the SEL. The
application would first ‘reserve’ the repository by issuing a Reserve SEL command. The application would then
check that all SEL entries have been handled prior to issuing the Clear SEL command.
307
Intelligent Platform Management Interface Specification
If an new event had been placed in the SEL after the records were checked, but before the Clear SEL command, it
is possible for the event to be lost. However, the addition of a new event to the SEL causes the present
Reservation ID to be ‘canceled’. This would prevent the Clear SEL command from executing. If this occurred,
the application would repeat the reserve-check-clear process until successful.
Table 25-4, Reserve SEL Command
Request Data
Response Data
byte
1
data field
Completion Code
81h = cannot execute command, SEL erase in progress
2
Reservation ID, LS Byte 0000h reserved.
3
Reservation ID, MS Byte
25.4.1 Reservation Restricted Commands
A Requester must issue a ‘Reserve SEL’ command prior to issuing any of the following SEL commands. Note
that the ‘Reserve SEL’ command only needs to be reissued if the reservation is canceled. These commands
shall be rejected if the Requester’s reservation has been canceled.
•
Delete SEL Entry command
•
Clear SEL command
•
Get SEL Entry command (if ‘get’ is from an offset other than 00h)
•
Partial Add SEL Entry command
If the given reservation has been canceled, a ‘reservation canceled’ completion code shall be returned in the
response to the above commands.
Note that the Record ID associated with a given record could change between successive offset 0 ‘Gets’ to that
Record ID. For example, the first SEL Entry could change if the SEL were cleared and a new event came in. It
is thus the responsibility of the device accessing the SEL to verify that the retrieved record information matches
up with the ID information (timestamp, slave address, LUN, sensor ID, etc.) of the event record.
25.4.2 Reservation Cancellation
The SEL Device shall automatically cancel the present SEL reservation after any of the following events occur:
308
•
A SEL entry is added.
•
A SEL entry is deleted such that other Record IDs change. As a simplification, an implementation is
allowed to cancel the reservation on any SEL entry deletion.
•
The SEL is cleared.
•
The SEL Device is reset (via hardware or Cold Reset command)
•
A new ‘Reserve SEL’ command is received.
Intelligent Platform Management Interface Specification
25.5 Get SEL Entry Command
This command is used to retrieve entries from the SEL. The record data field in the response returns the 16 bytes
of data from the SEL Event Record.
Table 25-5, Get SEL Entry
Request Data
byte
1:2
3:4
Response Data
5
6
1
2:3
4:N
data field
Reservation ID, LS Byte first. Only required for partial Get. Use
[1]
0000h otherwise.
SEL Record ID, LS Byte first.
0000h = GET FIRST ENTRY
FFFFh = GET LAST ENTRY
Offset into record
Bytes to read. FFh means read entire record.
Completion Code
Return an error completion code if the SEL is empty.
81h = cannot execute command, SEL erase in progress.
Next SEL Record ID, LS Byte first (return FFFFh if the record just
returned is the last record.)
Note: FFFFh is not allowed as the record ID for an actual record. I.e.
the Record ID in the Record Data for the last record should not be
FFFFh.
Record Data, 16 bytes for entire record
1. The reservation ID should be set to 0000h for implementations that don’t implement the Reserve
SEL command.
25.6 Add SEL Entry Command
This command is provided to enable BIOS to add records to the System Event Log. Normally, the SEL Device
and the Event Receiver Device will be incorporated into the same management controller. In this case, BIOS or
the system SMI Handler adds its own events to the SEL by formatting an Event Message and transmitting it to the
SEL Device, rather than by using this command.
Records are added on after the last record in the SEL. The SEL Device adds the timestamp according to the SEL
Record Type (see 25.6.1, SEL Record Type RangesSEL Record Type Ranges, following) when it creates the
record. Thus, in some cases the timestamp bytes in the record data are ignored. However, ‘dummy’ timestamp
bytes must still be present in the data.
The record data field that is passed in the request consists of all bytes of the SEL event record. The Record ID
field that is passed in the request is just a placeholder. The Record ID field that was passed in the request will be
overwritten with a Record ID value that the SEL Device generates before the record is stored. Depending on the
Record Type, the entry may also be automatically timestamped (see following section). If the entry is
automatically timestamped, the SEL Device will also over-write the four bytes of the record’s timestamp field.
Note: The normal mechanism for adding entries to the SEL is via an Event Request message to the Event
Receiver device.
309
Intelligent Platform Management Interface Specification
Table 25-6, Add SEL Entry
Request Data
Response Data
byte
1:16
1
2:3
data field
Record Data, 16 bytes. Refer to section 26, SEL Record
FormatsSEL Record Formats
Completion Code. Generic, plus following command specific:
80h = operation not supported for this Record Type
81h = cannot execute command, SEL erase in progress
Record ID for added record, LS Byte first.
25.6.1 SEL Record Type Ranges
The following lists the ranges used for SEL Record types:
310
00h - BFh
Range reserved for standard SEL Record Types. As of this writing, only type 02h is defined.
Records are automatically timestamped unless otherwise indicated.
C0h - DFh
Range reserved for timestamped OEM SEL records. These records are automatically
timestamped by the SEL Device.
E0h - FFh
Range reserved for non-timestamped OEM SEL records. The SEL Device does not automatically
timestamp these records. The four bytes passed in the byte locations for the timestamp will be
directly entered into the SEL.
Intelligent Platform Management Interface Specification
25.7 Partial Add SEL Entry Command
This command is a version of the Add SEL Entry command that allows the record to be incrementally added to
the SEL. The Partial Add SEL Entry command must be preceded by a Reserve SEL command. The first partial
add must be to offset 0000h, and subsequent partial adds must be done sequentially, with no gaps or overlap
between the adds.
The add must be completed before any of its contents can be retrieved from the SEL. If the reservation is
canceled before the add is completed, the information is discarded and the add must be redone starting at offset
0000h.
When the Record Type directs the BMC to automatically timestamp the record, the BMC will set the timestamp
when the last record is transferred.
Note: The normal mechanism for adding entries to the SEL is via an Event Request message to the Event
Receiver device.
Table 25-7, Partial Add SEL Entry Command
Request Data
byte
1:2
3:4
5
6
Response Data
7:N
1
2:3
data field
Reservation ID, LS Byte first. Only required for partial add. Use
[1]
0000h for Reservation ID otherwise.
Record ID, LS Byte first. Used when continuing a partial add (nonzero offset into record). Use 0000h for Record ID otherwise.
Offset into record.
In progress.
[7:4] - reserved
[3:0] - in progress
0h = partial add in progress.
1h = last record data being transferred with this request
SEL Record Data
Completion Code
80h = Record rejected due to mismatch between record length in
header data and number of bytes written. (Verifying the
length is an optional operation for the management
controller)
81h = cannot execute command, SEL erase in progress
Record ID for added record, LS Byte first.
1. The reservation ID should be set to 0000h for implementations that don’t implement the Reserve
SEL command.
25.8 Delete SEL Entry Command
Table 25-8, Delete SEL Entry
Request Data
Response Data
byte
1:2
3:4
1
2:3
data field
[1]
Reservation ID, LS Byte first.
SEL Record ID to delete, LS Byte first.
0000h = FIRST ENTRY
FFFFh = LAST ENTRY
Completion Code - Generic plus following command specific:
80h = operation not supported for this Record Type
81h = cannot execute command, SEL erase in progress
Record ID for deleted record, LS Byte first.
1. The reservation ID should be set to 0000h for implementations that don’t implement the Reserve
SEL command.
311
Intelligent Platform Management Interface Specification
25.9 Clear SEL Command
The command ‘erases’ all contents of the System Event Log. Since this process may take several seconds, based
on the type of storage device, the command also provides a means for obtaining the status of the erasure.
Table 25-9, Clear SEL
Request Data
Response Data
byte
1:2
3
4
5
6
1
2
data field
[1]
Reservation ID, LS Byte first.
‘C’ (43h)
‘L’ (4Ch)
‘R’ (52h)
AAh = initiate erase.
00h = get erasure status.
Completion Code
Erasure progress.
[7:4] - reserved
[3:0] - erasure progress
0h = erasure in progress.
1h = erase completed.
1. The reservation ID should be set to 0000h for implementations that don’t implement the Reserve
SEL command.
25.10 Get SEL Time Command
This command returns the time from the SEL Device. This time is used by the SEL Device for Event
Timestamping.
Table 25-10, Get SEL Time Command
Request Data
Response Data
byte
1
2:5
data field
Completion Code
Present Timestamp clock reading. LS byte first. See Section 31,
Timestamp FormatTimestamp Format.
25.11 Set SEL Time Command
This command initializes the time in the SEL Device. This time is used by the SEL Device for Event
Timestamping.
Table 25-11, Set SEL Time Command
Request Data
Response Data
312
byte
1:4
1
data field
Time in four-byte format. LS byte first. See Section 31, Timestamp
FormatTimestamp Format.
Completion Code
Intelligent Platform Management Interface Specification
25.12 Get Auxiliary Log Status Command
This command originated primarily to provide a mechanism that would allow remote software to know whether
new information has been added to Machine Check Architecture (MCA) Log. that can be provided. The MCA
Log is a storage area that can be implemented in Intel® Itanium™-based computer systems and holds information
from an MCA Handler running from system firmware.
For systems that lack MCA, the command can be used to return information about similar OEM-specified logs
that may hold extended event information for the platform. Since such logs are usually central resources, this
command will typically be implemented by a BMC in a host system, or the chassis controller in a managed
peripheral chassis.
Table 25-12, Get Auxiliary Log Status Command
Request Data
Response Data
byte
1
1
2:5
6:9
2:5
6:8
9:16
data field
Log Type
[7:4] - reserved
[3:0] - Log Type
00h = MCA Log
01h = OEM 1
02h = OEM 2
all other = reserved
Completion Code. An error completion code will be returned if the
given log type is not supported.
For Log Type = MCA Log :
IPMI Timestamp for when last entry was added to MCA Log, per
section 31,Timestamp Format.
32-bit count of number of entries in MCA Log, LSByte first.
FFFF_FFFFh = unspecified.
For Log Type = OEM 1 or OEM 2:
IPMI Timestamp for when last entry was added to log, per section
31,Timestamp Format.
OEM ID = three byte OEM IANA. IANA Enterprise Number for
OEM/Organization that specifies the following log status bytes. Least
significant byte first.
Log status bytes per OEM identified by OEM ID
313
Intelligent Platform Management Interface Specification
25.13 Set Auxiliary Log Status Command
This command can be used by system software or firmware to set the status returned by the Get Auxiliary Log
Status command. Some implementations may elect to implement solely private mechanism for setting this status,
in which case this command may not be provided even if the Get Auxiliary Log Status is.
Table 25-13, Set Auxiliary Log Status Command
Request Data
byte
1
2:5
6:9
2:5
6:8
Response Data
314
9:16
1
data field
Log Type
[7:4] - reserved
[3:0] - Log Type
00h = MCA Log
01h = OEM 1
02h = OEM 2
all other = reserved
For Log Type = MCA Log :
IPMI Timestamp for when last entry was added to MCA Log, per
section 31,Timestamp Format.
32-bit count of number of entries in MCA Log, LSByte first.
FFFF_FFFFh = unspecified.
For Log Type = OEM 1 or OEM 2:
IPMI Timestamp for when last entry was added to log, per section
31,Timestamp Format.
OEM ID = three byte OEM IANA. IANA Enterprise Number for
OEM/Organization that specifies the following log status bytes. Least
significant byte first.
Log status bytes per OEM identified by OEM ID
Completion Code. An error completion code will be returned if the
given Log Type is not supported.
Intelligent Platform Management Interface Specification
26. SEL Record Formats
The following sections present the record formats for SEL entries. Note that these are the ‘external’ specifications
for the records. The actual storage format within the SEL Device implementation may be different.
26.1 SEL Event Records
The following table presents the format of SEL Event Records This is the stored information from Event
Messages, as described in 23.4, Event Request Message FieldsEvent Request Message Fields.
Table 26-1, SEL Event Records
Byte
Field
Description
1
2
Record ID
3
Record Type
4
5
6
7
8
9
Timestamp
ID used for SEL Record access. The Record ID values 0000h and FFFFh have
special meaning in the Event Access commands and must not be used as Record ID
values for stored SEL Event Records.
[7:0] - Record Type
02h = system event record
C0h-DFh = OEM timestamped, bytes 8-16 OEM defined
E0h-FFh = OEM non-timestamped, bytes 4-16 OEM defined
Time when event was logged. LS byte first.
10
EvM Rev
Generator ID
11
12
13
Sensor Type
Sensor #
Event Dir |
Event Type
14
Event Data 1
15
Event Data 2
16
Event Data 3
RqSA & LUN if event was generated from IPMB. Software ID if event was generated
from system software.
Byte 1
2
[7:1] - 7-bit I C . Slave Address, or 7-bit system software ID
[0]
0b = ID is IPMB Slave Address
1b = system software ID
Byte 2
[7:4] - Channel number. Channel that event message was received over. 0h if the
event message was received via the system interface, primary IPMB, or
internally generated by the BMC. (New for IPMI v1.5. These bits were reserved
in IPMI v1.0)
[3:2] - reserved. Write as 00b.
[1:0] - IPMB device LUN if byte 1 holds Slave Address. 00b otherwise.
Event Message format version (=04h for events in this specification, 03h for IPMI
v1.0 Event Messages.)
Note: the BMC must accept Platform Event request messages that are in IPMI v1.0
format (EvMRev=03h) and log them as IPMI v1.5 Records by setting the EvMRev
field to 04h and setting the Channel Number in the Generator ID field appropriately
for the channel that the event was received from.
Sensor Type Code for sensor that generated the event
Number of sensor that generated the event
Event Dir
[7 ] -0b = Assertion event.
1b = Deassertion event.
Event Type
Type of trigger for the event, e.g. critical threshold going high, state asserted, etc.
Also indicates class of the event. E.g. discrete, threshold, or OEM. The Event Type
field is encoded using the Event/Reading Type Code. See section 36.1,
Event/Reading Type CodesEvent/Reading Type Codes.
[6:0] - Event Type Code
Per Table 23-6, Event Request Message Event Data Field ContentsTable 23-6, Event
Request Message Event Data Field Contents
Per Table 23-6, Event Request Message Event Data Field ContentsTable 23-6, Event
Request Message Event Data Field Contents
Per Table 23-6, Event Request Message Event Data Field ContentsTable 23-6, Event
Request Message Event Data Field Contents
315
Intelligent Platform Management Interface Specification
26.2 OEM SEL Record - Type C0h-DFh
C0h - DFh
Range reserved for timestamped OEM SEL records. These records are automatically
timestamped by the SEL Device. These records are entered via the Add SEL or Partial Add SEL
commands.
Table 26-2, OEM SEL Record (Type C0h-DFh)
Byte
Field
Description
1
2
Record ID
3
Record Type
ID used for SEL Record access. The Record ID values 0000h and FFFFh have
special meaning in the event access commands, and are not to be used as Record
ID values for stored SEL Event Records.
[7:0] Record Type
C0h-DFh = OEM system event record
Time when event was logged (automatically added by SEL device). LS byte first.
4
5
6
7
8:10
11:16
Timestamp
Manufacturer ID
OEM Defined
Manufacturer ID (see Get Device ID command for definition)
OEM Defined. This is defined according to the manufacturer identified by the
Manufacturer ID field.
26.3 OEM SEL Record - Type E0h-FFh
E0h - FFh
Range reserved for non-timestamped OEM SEL records. The SEL Device does not automatically
timestamp these records. The four bytes passed in the byte locations normally used for the
timestamp will be directly entered into the SEL. SEL viewer applications should not interpret
byte positions 4:7 in this record as a timestamp. These records are entered via the Add SEL or
Partial Add SEL commands.
Table 26-3, OEM SEL Record (Type E0h-FFh)
Byte
Field
Description
1
2
Record ID
3
Record Type
ID used for SEL Record access. The Record ID values 0000h and FFFFh have
special meaning in the event access commands, and are not to be used as Record
ID values for stored SEL Event Records.
[7:0] - Record Type
E0h-FFh = OEM system event record
OEM Defined. This is defined by the system integrator.
4:16
316
OEM
Intelligent Platform Management Interface Specification
27. SDR Repository
This section describes the logical SDR Repository Device, and the commands that are used to access the SDR
Repository. This section also describes a companion set of functionality, the Internal Sensor Initialization Agent,
that is part of a system that implements this platform and sensor instrumentation specification.
The SDR Repository is intended to hold information indicating the set of management controllers, sensors, and FRU
Devices that is expected to be in the system. Platform management often requires knowledge of what devices are
supposed to be there, as opposed to what devices are detected. This is because an undetected device may be
unintentionally absent, which in platform management usually constitutes a failure condition.
For example, suppose the baseboard had connectors for five fans, but only the first four were supposed to be
populated. The SDRs for the system would report four fan sensors, one for each of the first four connectors. This
tells system management software that any fewer than four fans on the designated connectors would be an error
condition. Thus, if the system user unintentionally disconnected a fan, system management software would see an
error when it tried to get the fan status. Keeping this information also enables features that allow the platform
management hardware itself to take automatic actions based on the ‘missing’ fan.
The SDR records can be used to represent a custom configuration. Using the same example, suppose a system
integrator wanted to attach only three fans to the baseboard, and use them on the last three connectors. The SDRs
could be changed to report only three fans, and indicate they’re on the last three connectors. System management
software only pays attention to sensors for which SDRs are present, thus by just adding, deleting, or modifying
SDRs a system integrator can change the population of sensors within the constraint of the total available sensors
built into the hardware. (I.e. you can’t ‘create’ a sensor that doesn’t pre-exist in hardware).
SDRs are kept in a single, centralized Sensor Data Record Repository to simplify the ability for out-of-band
applications to get information about the platform management subsystem. This eliminates the need for out-of-band
applications, which may be over slow transports, to perform discovery actions. It also is a better mechanism to
ensure that the information actually represents what’s supposed to be in the system, instead of just what was
discovered.
27.1 SDR Repository Device
The SDR Repository Device is a logical device that accepts and responds to SDR Repository commands. The
SDR Repository Device isolated from most aspects of the data that is in an SDR. The SDR Device manages
SDRs but does not interpret them or take action on the record contents. The exceptions to this is a small set of
fixed fields that are used to identify the record and the record type. These fields are contained in the Record
Header area of the Sensor Data Record.
Another important set of fields are those that are identified as the Record Key fields. The combined information in
these fields uniquely identifies the record contents. The Record Key fields are used for record content
identification, while the Record ID field is used for record access. For example, a given instance of a sensor will
always have the same Record Key information. The Record ID field, however can vary with time as records are
added to and removed from the SDR Repository.
The present specification only allows one SDR Repository device per system. For host systems that incorporate a
BMC, the SDR Repository is implemented via the BMC. For peripheral chassis that use the ICMB, the device
holding the SDR Repository is specified by the Chassis Capabilities command (refer to the Intelligent Chassis
Management Bus Bridge specification).
317
Intelligent Platform Management Interface Specification
27.2 Modal and Non-modal SDR Repositories
There are two possible SDR Repository implementations: modal and non-modal. A non-modal SDR Repository
can be written to at any time. Writing to the SDR does not impact the operation of other commands in the
management controller.
A modal SDR Repository is only updated when the controller is in an SDR Repository update mode. This
provision is made to allow SDR information to be kept in non-volatile storage devices that may require lengthy
write operations, or interfere with other controller operations when updated. For example, this could allow the
SDR Repository to be stored in a FLASH device that also holds a portion of management controller code. A
modal SDR Repository implementation would allow the functions associated with that code to be temporarily
unavailable during the update process.
An implementation that provides Modal SDR Repository Updates is not required to support non-modal SDR
updates. Generic SDR update software should first issue a Get SDR Repository Info command to determine which
type of update is supported. If the command returns ‘unspecified’, update software should first try a modal update
by issuing an Enter SDR Update Mode command. If that command is accepted, it should perform the update in
SDR Update Mode. If the command is not accepted, it should then attempt to perform a non-modal update.
27.2.1 Command Support while in SDR Repository Update Mode
The controller is only required to support a subset of its normal commands while it is in SDR Repository
Update Mode. A completion code of D0h must be returned as the response to any commands that are rejected
because the controller is in update mode. The list of commands that must be supported after entering SDR
Repository Update mode are listed in the following table. Detailed information is provided in following
sections.
The update mode commands must be supported via the system interface to the BMC. If the controller provides
an IPMB, it is recommended, but not mandatory, that the IPMB must remain active in SDR Repository Update
mode.
Table 27-1, Mandatory SDR Update Mode Commands
Command
Get Device ID
[2]
Get SDR
[1][2]
Add SDR
[1][2]
Partial Add SDR
[2]
Clear SDR Repository
[2]
Exit SDR Repository Update Mode
1.
2.
Section
17.1
27.12
27.13
27.14
27.16
27.20
Either Add SDR or Partial Add SDR must be provided. Providing both is optional.
These commands are only accepted from the System Interface if SDR Repository
Update Mode was entered via the System Interface, or are only accepted from
the device that put the controller into SDR Repository Update mode. Other
devices that try to issue these commands will receive a completion code
indicating that SDR Repository Update is in progress. Reservation is not required
for executing these commands in SDR Repository Update mode.
27.3 Populating the SDR Repository
Most systems are fundamentally static with respect to their platform management configuration once the system
integrator has put the system together. Thus, the typical model for the SDR Repository is that it is manually
updated using a utility or other piece of software if the platform management configuration is changed in the
field.
318
Intelligent Platform Management Interface Specification
For example, suppose a system could be upgraded to accept a new RAID backplane that had extra fans and
temperature sensors. Part of the upgrade process would be to run a utility, supplied by the system integrator, that
updated the SDR Repository with the new SDRs.
27.3.1 SDR Repository Updating
An SDR Repository implementation is not required to implement the Delete SDR command. This means that
random updates of individual records is not supported. In this case, updating the SDR Repository requires
reading out the SDR Repository, updating the copy, clearing the SDR Repository, and writing the updated
records in. Note that this approach works for all implementations, and helps avoid potential issues with
fragmentation of the SDR Repository.
27.4 Discovering Management Controllers and Device SDRs
IPMI includes the capability for allowing system software to discover management controllers. The responsibility
of detecting and integrating new devices is left to system software. This is done to avoid placing additional
complexity in BMC firmware, and to allow the discovery and integration policy to be more flexible and
sophisticated.
A system can be created that allows new management controllers and SDRs to automatically be discovered and
integrated into the SDR Repository. The following steps outline this process:
1.
System management software uses the Broadcast Get Device ID command to discover all management
controllers on the IPMB. It does this by repeatedly issuing the Broadcast Get Device ID, incrementing the
second byte in the message to select different management controller slave addresses. The software only
needs to go through the slave addresses that are assignable to IPMB devices (refer to the IPMB Address
Allocation specification.) System management software can go through this process when it initializes, or,
preferably, run this as a ‘background’ process that scans for new devices during run-time.
2.
System management software reads the SDRs and gets a list of the known management controllers from the
Management Controller Device Locator records. For each discovered device, system management software
checks to see if the device is one of the know devices or not. If the device has a corresponding Management
Controller Confirmation record, this record can be used to verify that a different type or instance of controller
didn’t wind up at the address of a previously present controller.
3.
For each newly discovered device, system management software would typically prompt the system user for
whether the device should be integrated or not. (For ‘missing’ devices, the system user would be notified of
the change). If the device supports Device SDRs, system management software would be able to read the
SDRs from the device and write them to the SDR Repository. If the device didn’t include Device SDRs, the
software would likely prompt the user for update software supplied by the system integrator or device
provider. Note that the management controllers now include information such as the manufacturer ID, that
can be an aid to creating useful prompts for this kind of information.
27.5 Reading the SDR Repository
An application that retrieves records from the SDR Repository must first read them out sequentially. This is
accomplished by using the Get SDR command to retrieve the first SDR of the desired type. The response to this
command returns the requested record and the Record ID of the next SDR in sequence in the repository. Note that
Record IDs are not required to be sequential or consecutive. Applications should not assume that SDR Record
IDs will follow any particular numeric ordering.
The application retrieves succeeding records by issuing a Get SDR command using the ‘next’ Record ID that was
returned with the response of the previous Get SDR command. This is continued until the ‘End of Records’ ID is
encountered.
319
Intelligent Platform Management Interface Specification
Once the application has read out the desired records, it can then randomly access the records according to their
Record ID. An application that seeks to access records randomly must save a data structure that retains the Record
Key information according to Record ID.
Since it is possible for Record IDs to change with time, it is important for applications to first verify that the
Record Key information matches up with the retrieved record. If the Record Key information doesn’t match, then
the Record ID is no longer valid for that Record Key, and the SDR Records must again be accessed sequentially
until the record that matches the Record Key is located.
An application can also tell whether records have changed by examining the ‘most recent addition’ timestamp
using the Get SDR Repository Info command.
If record information has changed, an application does not need to list out the entire contents of all records. The
Get SDR command allows a partial read of the SDR. Thus, an application can search for a given Record Key by
just retrieving that portion of the record.
27.6 Sensor Initialization Agent
The Sensor Initialization Agent is not a logical device, but rather a collection of functions and services that are
specific to handling SDR information. Unlike the SDR Repository Device, the Sensor Initialization Agent works
directly with the content of SDRs, in particular, the Sensor Data Records and Device Locator Records.
The Initialization Agent utilizes the SDR information for sensor and IPMB Device initialization during system
startup. The Initialization Agent knows how to interpret Sensor Data Records and is directed by the ‘init required’
fields to load thresholds to sensors that have the ‘threshold initialization required’ bit set in their SDR record.
Other bits in the record direct the agent to enable sensors/devices that come up with sensors and/or events
disabled.
The Initialization Agent Function normally runs whenever the system powers up, and upon system Hard Resets.
This ensures that the sensor subsystem and threshold values will be re-initialized in response to 'push-button'
hardware resets. It is also recommended that the Initialization Agent function run when the BMC first receives
standby power.
Note that in systems that implement power-management, System Management Software may need to take
additional steps to restore intermediate settings after the system has ‘woken up’.
27.6.1 System Support Requirements for the Initialization Agent
The BMC requires information about when the system has been powered up, hard reset, or warm ‘ctrl-alt-del’
reset. This information is needed to trigger the Initialization Agent function. The mechanism for accomplishing
this is implementation-dependent. Two common ways to provide this information are via hardware signals to
the BMC, or via a BMC-specific application command from BIOS. A combination of the two can also be used.
For example, a hardware signals could be used to indicate when the system is hard-reset, while a command
from BIOS could indicate warm ‘ctrl-alt-del’ resets.
27.6.2 IPMI and ACPI Interaction
The Initialization Agent restores ‘power-on default’ threshold values and event enable settings. In order to
provide consistent operation, the initialization agent takes the same actions on ‘warm’ (e.g. ctrl-alt-del) resets.
In a system that has ACPI, the platform management subsystem cannot generally distinguish between power-up
from an S5 ‘OFF’ state and power-up from an S4 ‘Suspend-to-disk’ sleep state. When the system wakes from
an S4 state, system management software should recognize this condition so that it can restore any ‘volatile’
settings that it may have gotten reset by the Initialization Agent.
320
Intelligent Platform Management Interface Specification
For other sleep states (S1-S3), the management controllers should retain their settings and the Initialization
Agent should not be run on wake. If a management controller (other than the BMC) gets powered down in S1S3, that controller is responsible for retaining the last settings that were written to it by system software.
System management software should also be aware of ACPI interaction with the watchdog timer. The
watchdog timer does not automatically stop counting down when the system enters an S1-S3 sleep state. If the
watchdog timer is being used as an OS Watchdog, system management software should use support in the
operating system to schedule a ‘wake event’ such that system management software can run and reload the
timer before it expires. Alternatively, system management software could shut down the timer upon receiving a
notification of entry into a sleep state, but that would reduce the value of using a watchdog timer to monitor OS
or system software health.
27.6.3 Recommended Initialization Agent Steps
1.
Initialize any BMC internal functions that are required by BIOS during POST.
2.
Disable the Event Receiver function for events received from any interface but the system interface, or from
BMC internal sensors that require initialization. The BMC should accept event messages from BIOS while the
initialization agent is running. The implementation may elect to accept BMC internal event messages from
sensors that do not require initialization. It is recommended that any events related to the initialization agent
operation are logged during the initialization agent process - but they may be collected and logged at its
conclusion.
3.
Scan the SDR repository for Management Controller Device Locator records. Collect a list of the addresses of
management controllers that require initialization. (A field in the Management Controller Device Locator
record indicates whether the management controller requires initialization, and if so, whether event messaging
should be enabled after the controller has been initialized.) This list should include the BMC itself.
4.
For each Management Controller in the list, turn off Event Generation by using the Set Event Receiver
command to set the Event Receiver. If the Management Controller does not respond to the Set Event Receiver
command, take it off the list.
a)
Scan the SDR Repository for Type 01h & Type 02h SDRs. For each encountered:
b) Check the Device Owner ID to see if the sensor belongs to the BMC or one of the other management
controllers in the list. If it does not, go on to the next record.
c)
It is possible that a management controller may have other actions that it takes on an event, thus it is
important to disable event scanning before setting thresholds and hysteresis. Check the Sensor Capabilities
field to see if per-sensor or per-threshold/per-state disable is supported. If it is, then use the Set Sensor
Event Enable command to disable scanning and event messages per the SDR.
d) Set the sensor type, sensor thresholds, and hysteresis as directed by the SDR using the Set Sensor Type, Set
Sensor Thresholds, and Set Sensor Hysteresis commands.
e)
Use the Set Sensor Event Enable command to enable scanning and event generation per the SDR. Go on to
next SDR.
5.
Enable the BMC Event Receiver function for the IPMB and other interfaces.
6.
For each management controller in the list, enable event message generation or leave it disabled (A field in the
Management Controller Device Locator record indicates whether event messaging should be enabled after the
controller has been initialized.)
321
Intelligent Platform Management Interface Specification
27.7 SDR Repository Device Commands
The following sections describe the commands that an SDR Repository Device provides for accessing the SDR
Repository.
The commands are designed to simplify the SDR Repository device’s implementation by ‘pushing back’
intelligence to higher-level software where possible. The SDR Repository device is not intended to be a
‘database’ engine. Thus, the SDR access commands do not include automatic search functions. It is recommended
that an application read the SDR Repository into a RAM buffer and work from that copy (keeping track of the
SDR Timestamp to check for possible changes to the SDR Repository). The general procedure for reading SDRs
from the SDR Repository is described under the Get SDR command.
As with Event Messages, it is also the intent that the commands are designed so that the SDR Repository Device
is isolated from needing to know the content and format of the SDR records themselves.
Refer to Appendix G - Command Assignments for the specification of the Network Function and Command
(CMD) values and privilege levels for these commands.
Table 27-2, SDR Repository Device Commands
Command
Get SDR Repository Info
Get SDR Repository Allocation Info
Reserve SDR Repository
Get SDR
Add SDR
Partial Add SDR
Delete SDR
Clear SDR Repository
Get SDR Repository Time
Set SDR Repository Time
Enter SDR Repository Update Mode
Exit SDR Repository Update Mode
Run Initialization Agent
1.
2.
3.
4.
5.
Section
O/M
27.9
27.10
27.11
27.12
27.13
27.14
27.15
27.16
27.17
27.18
27.19
27.20
27.21
M
O
M
[5]
M
[1]
M
[1][5]
M
[5]
O
[5]
M
[2]
O/M
[2]
O/M
[3]
O
[3]
O
[4]
O
Either Add SDR or Partial Add SDR command must be provided via the system interface.
Providing both via the system interface is optional. For the IPMB, the Add SDR and Partial
Add SDR commands are optional.
If the SEL Device and SDR Repository Device are implemented in separate controllers,
then both these commands are Mandatory for the SDR Repository Device. If the SDR
Repository Device shares the same controller as the SEL Device (This is normally indicated
in the IPM Device Support field of the Get Device ID command), then the SDR device uses
the SEL Device’s Timestamp Clock. In this case, the Get SDR Repository Time command
is optional, and the Set SDR Repository Time command is not used.
Support for both these commands is mandatory for a modal SDR Repository. The Enter
SDR Repository Update Mode command is mandatory when in ‘operational’ mode, while
the Exit SDR Repository Update Mode is mandatory when in ‘update’ mode.
Highly recommended. This supports utilities that can update the SDRs during run-time.
Without this, a system reset will need to be performed to cause the initialization agent to
run.
A reservation field of 0000h is passed to these commands when in SDR Repository Update
Mode.
27.8 SDR ‘Record IDs’
In order to generalize SDR access, Sensor Data Records are accessed using a ‘Record ID’ number. There are a
fixed number of possible Record IDs for a given implementation of the SDR Repository.
322
Intelligent Platform Management Interface Specification
The most common implementation of ‘Record IDs’ is as a value that translates directly to an ‘index’ or ‘offset’
into the SDR Repository. However, it is also possible for an implementation to provide a level of indirection, and
implement Record IDs as ‘handles’ to the Sensor Data Records.
Record ID values may be ‘recycled’. That is, the Record ID of a previously deleted SDR can be used as the
Record ID for a new SDR. The requirement is that, at any given time, the Record IDs are unique for all SDRs in
the repository.
Record IDs can be reassigned by the SDR Repository Device as needed when records are added or deleted. An
application that uses a Record ID to directly access a record should always verify that the retrieved record
information matches up with the ID information (slave address, LUN, sensor ID, etc.) of the desired sensor. An
application that finds that the SDR at a given ‘Record ID’ has moved will need to re-enumerate the SDRs by
listing them out using a series of Get SDR commands. Note that it is not necessary to read out the full record data
to see if the Record ID for a particular record has changed. Software can determine whether a given record has
been given a different Record ID by examining just the SDR’s header and record key bytes.
323
Intelligent Platform Management Interface Specification
27.9 Get SDR Repository Info Command
This command returns the SDR command version for the SDR Repository. It also returns a timestamp for when
the last ADD, DELETE, or CLEAR occurred. The Most Recent Addition timestamp field returns the timestamp
for the last addition operation, while the Most Recent Erase field returns the timestamp for the last delete or clear
operation.
These timestamps are independent of timestamps that may be returned by other commands, such as those returned
by the Get SEL Info command. The timestamp reflects when the most recent SDR Repository add or erase
occurred, not when the last add or erase occurred on the physical storage device.
For example, the SDR Repository Info Most Recent Addition timestamp would reflect the last time a new record
was added to the SDR Repository. The SDR Repository’s most recent addition timestamp is always independent
of the most recent addition time for the SEL - even if the SEL and SDR Repository are implemented in the same
physical storage device.
Table 27-3, Get SDR Repository Info Command
byte
Request Data
Response Data
-
1
Completion Code
2
3
SDR Version - version number of the SDR command set for the SDR Device.
51h for this specification. (BCD encoded with bits 7:4 holding the Least
Significant digit of the revision and bits 3:0 holding the Most Significant bits.)
Record count LS Byte - number of records in the SDR Repository
4
Record count MS Byte - number of records in the SDR Repository
5:6
7:10
11:14
15
324
data field
-
Free Space in bytes, LS Byte first. 0000h indicates ‘full’, FFFEh indicates
64KB-2 or more available. FFFFh indicates ‘unspecified’.
Most recent addition timestamp. LS byte first.
Most recent erase (delete or clear) timestamp. LS byte first.
Operation Support
[7] - Overflow Flag. 1=SDR could not be written due to lack of space in the
SDR Repository.
[6:5] - 00b = modal/non-modal SDR Repository Update operation unspecified
01b = non-modal SDR Repository Update operation supported
10b = modal SDR Repository Update operation supported
11b = both modal and non-modal SDR Repository Update supported
[4] - reserved. Write as 0b
[3] - 1b=Delete SDR command supported
[2] - 1b=Partial Add SDR command supported
[1] - 1b=Reserve SDR Repository command supported
[0] - 1b=Get SDR Repository Allocation Information command supported
Intelligent Platform Management Interface Specification
27.10 Get SDR Repository Allocation Info Command
Returns the number of possible allocation units, the amount of usable free space (in allocation units), the
allocation unit size (in bytes), and the size of the largest contiguous free region (in allocation units). The
‘allocation unit size’ is the number of bytes in which storage is allocated. For example, if a 20-byte record is to be
added, and the SDR Repository has a 16-byte allocation unit size, then the record would take up 32-bytes of
storage.
The SDR Repository implementation shall, at a minimum, provide an allocation unit size of ≥16 bytes and a
“maximum record size” supporting a record of ≥ 64 bytes.
Software should assume an allocation unit size of 16-bytes if this command is not implemented.
Table 27-4, Get SDR Repository Allocation Info Command
byte
Request Data
Response Data
-
data field
-
1
Completion Code
2
3
Number of possible allocation units, LS Byte
Number of possible allocation units, MS Bytes
This number indicates whether the total number of possible allocation units is
equal to, or some number less than the log size divided by the allocation unit
size.
0000h indicates ‘unspecified’.
Allocation unit size in bytes. 0000h indicates ‘unspecified’.
4
5
6
7
8
9
10
Number of free allocation units, LS Byte
Number of free allocation units, MS Byte
Largest free block in allocation units, LS Byte
Largest free block in allocation units, MS Byte
Maximum record size in allocation units.
27.11 Reserve SDR Repository Command
This command is used to set the present ‘owner’ of the repository, as identified by the ‘Software ID’ or by the
Requester’s Slave Address from the command. The reservation process provides a limited amount of protection
on repository access from the IPMB when records are being deleted or incrementally read.
The Reserve SDR Repository command is provided to help prevent deleting the wrong record when doing
deletes, and to prevent receiving incorrect data when doing incremental reads.
The Reserve SDR Repository command does NOT guarantee access to the SDR Repository. That is, the case
exists that a pair of requesters could vie for access to the SDR in such a manner that they alternately cancel the
reservation that is held by the other - effectively ‘deadlocking’ each other.
A ‘Reservation ID’ value is returned in response to this command. This value is required in other requests, such
as the ‘Delete SDR’ command. These commands will not execute unless the correct Reservation ID value is
provided.
The Reservation ID is used in the following manner. Suppose an application wishes to delete a particular record.
The application would first ‘reserve’ the repository by issuing a Reserve SDR Repository command. The
application would then read the header and key information from the record to verify that it has the correct
Record ID for the record. Assuming this is correct, the application would then issue a Delete SDR command using
the Reservation ID and Record ID as parameters.
If an event had occurred that changed the Record IDs after the header and key information was read but before
the Delete SDR command, the Delete SDR command could be issued with the Record ID for the wrong record.
However, events that change Record IDs for any existing records cause the present Reservation ID to be
325
Intelligent Platform Management Interface Specification
‘canceled’. This prevents software from using an out-of-date Record ID to access a record. For example, it would
prevent the Delete SDR command from executing and deleting the wrong record in case a given Record ID was
reassigned to a different record.
Table 27-5, Reserve SDR Repository Command
byte
Request Data
Response Data
data field
-
-
1
Completion Code
2
Reservation ID, LS Byte
3
Reservation ID, MS Byte
27.11.1 Reservation Restricted Commands
A Requester must issue a ‘Reserve SDR Repository’ command prior to issuing any of the following SDR
Repository commands. Note that the ‘Reserve SDR Repository’ command only needs to be reissued if the
reservation is canceled. These commands shall be rejected if the Requester’s reservation has been canceled.
•
Delete SDR command
•
Clear SDR Repository command
•
Get SDR command (if a partial read)
•
Partial Add SDR command
If the given reservation has been canceled, a ‘reservation canceled’ completion code shall be returned in the
response to the above commands. This is explained further in the next section.
Note that since Record IDs could change between offset 0 ‘Gets’ of a given record, it is the responsibility of the
device accessing the repository to verify that the retrieved record information matches up with the ID
information (slave address, LUN, sensor ID, etc.) of the desired sensor.
27.11.2 Reservation Cancellation
The SDR Repository Device shall automatically cancel the present SDR Repository reservation after any of the
following events occur:
•
An SDR record is added using the Add SDR command such that other Record IDs change. As a
simplification, an implementation is allowed to cancel the reservation on any SDR record add.
•
An SDR record is deleted such that other Record IDs change. As a simplification, an implementation
is allowed to cancel the reservation on any SDR record deletion.
•
The SDR Repository is cleared.
•
The SDR Repository Device is reset (via hardware or Cold Reset command)
•
A new ‘Reserve SDR Repository’ command is received.
An error completion code will be returned if an attempt is made to execute a command that requires a
reservation ID, but the reservation ID used is not valid or current.
27.12 Get SDR Command
Returns the sensor record specified by ‘Record ID’. The command also accepts a ‘byte range’ specification that
allows just a selected portion of the record to be retrieved (incremental read). The Requester must first reserve the
326
Intelligent Platform Management Interface Specification
SDR Repository using the ‘Reserve SDR Repository’ command in order for an incremental read to an offset other
than 0000h to be accepted. (It is also recommended that an application use the Get SDR Repository Info command
to verify the version of the SDR Repository before it sends any other SDR Repository commands. This is
important since the SDR Repository command format and operation can change between versions.)
If ‘Record ID’ is specified as 0000h, this command returns the Record Header for the ‘first’ SDR in the
repository. FFFFh specifies that the ‘last’ SDR in the repository should be listed. If ‘Record ID’ is non-zero, the
command returns the information from the matching record, and the Record ID for the next SDR in the
repository.
An application that wishes to retrieve the full set of SDR Records must first issue the Get SDR starting with
0000h as the Record ID to get the first record. The Next Record ID is extracted from the response and this is then
used as the Record ID in a Get SDR request to get the next record. This is repeated until the ‘Last Record ID’
value (FFFFh) is returned in the ‘Next Record ID’ field of the response.
A partial read from offset 0000h into the record can be used to extract the header and associated ‘Key Fields’ for
the specified Sensor Data Record in the SDR Repository. An application can use the command in this manner to
get a list of what records are in the SDR and to identify the instances of each type. It can also be used to search
for an particular sensor record.
Note: to support future extensions, applications should check the SDR Version byte prior to interpreting any of
the data that follows.
The application issuing ‘Get SDR’ commands with a non-zero value for the Offset into record field must first
reserve the SDR Repository by issuing a ‘Reserve SDR Repository’ command.
If you issue a Get SDR command (storage 23h) with a 'bytes to read' size of 'FFh' - meaning 'read entire record'. A
value of 'FFh' will cause an error in most cases, since SDRs are bigger than the buffer sizes for the typical system
interface implementation. The controller therefore returns an error completion code if the number of record bytes
exceeds the maximum transfer length for the interface. The completion code CAh that indicates that the number
of requested bytes cannot be returned. Returning this code is recommended, although a controller could also
return an 'FFh' completion code. In either case, the algorithm for handling this situation is to "default to using
partial reads if the 'read entire record' operation fails" (that is, if you get a non-zero completion code).
Table 27-6, Get SDR Command
Request Data
Response Data
byte
1
data field
Reservation ID. LS Byte. Only required for partial reads with a nonzero ‘Offset into record’ field. Use 0000h for reservation ID
otherwise.
2
Reservation ID. MS Byte.
3
Record ID of record to Get, LS Byte
4
Record ID of record to Get, MS Byte
5
Offset into record
6
Bytes to read. FFh means read entire record.
1
Completion Code
2
Record ID for next record, LS Byte
3
Record ID for next record, MS Byte
4:3+N Record Data
327
Intelligent Platform Management Interface Specification
27.13 Add SDR Command
This command adds the specified sensor record to the SDR Repository and returns its ‘Record ID’. The data
passed in the request must contain the SDR data in its entirety.
Table 27-7, Add SDR Command
Request Data
Response Data
byte
1:N
1
2
3
data field
SDR Data
Completion Code
Record ID for added record, LS Byte
Record ID for added record, MS Byte
27.14 Partial Add SDR Command
This command is a version of the Add SDR command that allows the record to be incrementally added to the
repository. The Partial Add SDR command must be preceded by a ‘Reserve SDR Repository’ command. The first
partial add must be to offset 0000h, and partial adds must be done sequentially, with no gaps or overlap between
the adds.
The add must be completed before any of its contents can be retrieved from the SDR Repository. If the
reservation is canceled before the add is completed, the information is discarded and the add must be redone
starting at offset 0000h.
Software should assume an allocation unit size of 16-bytes if the Get SDR Allocation Info command is not
supported.
Table 27-8, Partial Add SDR Command
Request Data
byte
1
2
3
4
5
6
Response Data
7:N
1
2
3
328
data field
Reservation ID, LS Byte.
Reservation ID, MS Byte.
Record ID, LS Byte for continuing partial add. Use 0000h for Record
ID otherwise.
Record ID, MS Byte for continuing partial add. Use 0000h for
Record ID otherwise.
Offset into record.
In progress.
[7:4] - reserved
[3:0] - in progress
0h = partial add in progress.
1h = last record data being transferred with this request
SDR Record Data
Completion Code. Generic, plus following command-specific:
80h = Record rejected due to mismatch between record length in
header data and number of bytes written. (Verifying the
length is an optional operation for the management
controller)
Record ID for added record, LS Byte
Record ID for added record, MS Byte
Intelligent Platform Management Interface Specification
27.15 Delete SDR Command
Deletes the sensor record specified by ‘Record ID’. The Requester’s ID and the ‘Reservation ID’ must also match
the present ‘owner’ of the SDR Repository.
Table 27-9, Delete SDR Command
Request Data
Response Data
byte
1
2
3
4
1
2
3
data field
Reservation ID LS Byte
Reservation ID MS Byte
Record ID of record to delete, LS Byte
Record ID of record to delete, MS Byte
Completion Code
Record ID for deleted record, LS Byte
Record ID for deleted record, MS Byte
27.16 Clear SDR Repository Command
Clears all records from the SDR Repository and reinitializes the SDR Repository ‘subsystem’. Mainly a
development and production aid, use of this command should be generally avoided in utilities and system
management software. The Requester’s ID and Reservation ID information must also match the present ‘owner’
of the SDR Repository.
Table 27-10, Clear SDR Repository Command
Request Data
Response Data
byte
1
2
3
4
5
6
1
2
data field
Reservation ID. LS Byte.
Reservation ID. MS Byte.
‘C’ (43h)
‘L’ (4Ch)
‘R’ (52h)
AAh = initiate erase.
00h = get erasure status.
Completion Code
Erasure progress.
[7:4] - reserved
[3:0] - erasure in progress
0h = erasure in progress.
1h = erase completed.
329
Intelligent Platform Management Interface Specification
27.17 Get SDR Repository Time Command
This command returns the time from the SDR Repository Device. This time is used by the SDR Repository
Device for tracking when changes to the SDR Repository have been made. The time keeping format is specified
in Section 31, Timestamp FormatTimestamp Format.
A device that contains both a logical SEL device and an SDR Repository device can elect to implement just a
single Timestamp Clock, in which case, the Set SDR Repository command shall not be used. Instead, the Set SEL
Time command will be used for setting the time, and the Get SDR Repository Time and Get SEL Time commands
shall effectively return the same time values.
Table 27-11, Get SDR Repository Time Command
Request Data
Response Data
byte
1
2:5
data field
Completion Code
Time in four-byte format. LS byte first.
27.18 Set SDR Repository Time Command
This command initializes the time in the SDR Repository Device. This time is used by the SDR Device for
tracking when SDR Repository changes have been made. The time keeping format is specified in Section 31,
Timestamp Format.
A device that contains both a logical SEL device and an SDR Repository device can elect to implement just a
single Timestamp Clock, in which case, the Set SDR Repository command shall not be used. Instead, the Set SEL
Time command will be used for setting the time, and the Get SDR Repository Time and Get SEL Time commands
shall effectively return the same time values.
Table 27-12, Set SDR Repository Time Command
Request Data
Response Data
byte
1:4
1
data field
Time in four-byte format. LS byte first.
Completion Code
27.19 Enter SDR Repository Update Mode Command
Table 27-13, Enter SDR Repository Update Mode Command
Request Data
Response Data
330
byte
1
data field
Completion Code
Intelligent Platform Management Interface Specification
27.20 Exit SDR Repository Update Mode Command
Table 27-14, Exit SDR Repository Update Mode Command
Request Data
Response Data
byte
1
data field
Completion Code
27.21 Run Initialization Agent Command
This command can be used to cause the Initialization Agent to run. The command can be used to check the status
of the Initialization Agent as well.
Table 27-15, Run Initialization Agent
Request Data
Response Data
byte
1
1
2
data field
[7:1] - reserved
[0] - 1b = run initialization agent
0b = get status of initialization agent process
Completion Code
[7:1] reserved
[0] - 1b = initialization completed
0b = initialization in progress
331
Intelligent Platform Management Interface Specification
28. FRU Inventory Device Commands
The following sections describe the FRU (Field Replaceable Unit) Inventory Device format and access commands.
The FRU Inventory data contains information such as the serial number, part number, asset tag, and short
descriptive string for the FRU. The contents of a FRU Inventory Record are specified in the Platform Management
FRU Information Storage Definition.
The FRU Inventory Device is a ‘logical’ device. It is not necessarily implemented as a separate physical device,
though it can be. For example, the device that contains the SDR Repository Device also typically also holds ‘FRU
Inventory’ information for the main system board and chassis. On the other hand, there may be a separate FRU
Inventory device that provides access to the FRU information for a replaceable module such as a Memory Module.
Refer to Appendix G - Command Assignments for the specification of the Network Function and Command (CMD)
values and privilege levels for these commands.
Table 28-1, FRU Inventory Device Commands
Command
Section
O/M
28.1
28.2
28.3
M
M
M
Get FRU Inventory Area Info
Read FRU Data
Write FRU Data
O/M = Option/Mandatory for FRU Inventory Devices.
28.1 Get FRU Inventory Area Info Command
Returns overall the size of the FRU Inventory Area in this device, in bytes.
Table 28-2, Get FRU Inventory Area Info Command
Request Data
Response Data
332
byte
1
1
2
3
4
data field
FRU Device ID. FFh = reserved.
Completion Code
FRU Inventory area size in bytes, LS Byte
FRU Inventory area size in bytes, MS Byte
[7:1] - reserved
[0]
0b = Device is accessed by bytes, 1b = Device is accessed by words
Intelligent Platform Management Interface Specification
28.2 Read FRU Data Command
The command returns the specified data from the FRU Inventory Info area. This is effectively a ‘low level’ direct
interface to a non-volatile storage area. This means that the interface does not interpret or check any semantics or
formatting for the data being accessed. The offset used in this command is a ‘logical’ offset that may or may not
correspond to the physical address used in device that provides the non-volatile storage. For example, FRU
information could be kept in FLASH at physical address 1234h, however offset 0000h would still be used with
this command to access the start of the FRU information. IPMI FRU device data (devices that are formatted per
[FRU]) as well as processor and DIMM FRU data always starts from offset 0000h unless otherwise noted.
Note that while the offsets are 16-bit values, allowing FRU devices of up to 64K words, the count to read, count
returned, and count written fields are only 8-bits. This is in recognition of the limitations on the sizes of
messages. For example, as of this writing, IPMB messages are limited to 32-bytes total.
Table 28-3, Read FRU Data Command
Request Data
Response Data
byte
1
2
3
4
1
data field
FRU Device ID. FFh = reserved.
FRU Inventory Offset to read, LS Byte
FRU Inventory Offset to read, MS Byte
Offset is in bytes or words per device access type returned in the
Get FRU Inventory Area Info command.
Count to read --- count is ‘1’ based
Completion code. Generic, plus following command specific:
81h = FRU device busy. The requested cannot be completed
because the implementation of the logical FRU device is in a
state where the FRU information is temporarily unavailable.
This could be due to a condition such as a loss of arbitration
if the FRU is implemented as a device on a shared bus.
Software can elect to retry the operation after at least 30
milliseconds if this code is returned. Note that it is highly
recommended that management controllers incorporate builtin retry mechanisms. Generic IPMI software cannot be relied
upon to take advantage of this completion code.
2
Count returned --- count is ‘1’ based
3:2+N Requested data
28.3 Write FRU Data Command
The command writes the specified byte or word to the FRU Inventory Info area. This is a ‘low level’ direct
interface to a non-volatile storage area. This means that the interface does not interpret or check any semantics or
formatting for the data being written. The offset used in this command is a ‘logical’ offset that may or may not
correspond to the physical address used in device that provides the non-volatile storage. For example, FRU
information could be kept in FLASH at physical address 1234h, however offset 0000h would still be used with
this command to access the start of the FRU information. IPMI FRU device data (devices that are formatted per
[FRU]) as well as processor and DIMM FRU data always starts from offset 0000h unless otherwise noted.
Updating the FRU Inventory Data is presumed to be a system level, privileged operation. There is no requirement
for devices implementing this command to provide mechanisms for rolling back the FRU Inventory Area in the
case of incomplete or incorrect writes.
333
Intelligent Platform Management Interface Specification
Table 28-4, Write FRU Data Command
Request Data
Response Data
byte
1
2
3
4:3+N
1
2
334
data field
FRU Device ID. FFh = reserved.
FRU Inventory Offset to write, LS Byte
FRU Inventory Offset to write, MS Byte
Data to write
Completion code. Generic, plus following command specific:
80h = write-protected offset. Cannot complete write because one or
more bytes of FRU data are to a write-protected offset in the
FRU device. Note that an implementation may have allowed
a ‘partial write’ of the data to occur.
81h = FRU device busy. Refer to the preceding table for the Read
FRU Command for the description of this completion code.
Count written --- count is ‘1’ based
Intelligent Platform Management Interface Specification
29. Sensor Device Commands
The following table summarizes the commands that apply to a logical Sensor Device. Refer to Appendix G Command Assignments for the specification of the Network Function and Command (CMD) values and privilege
levels for these commands.
Table 29-1, Sensor Device Commands
Command
Notes:
Section
O/M
Get Device ID
Cold Reset
Warm Reset
Get Self Test Results
Manufacturing Test Mode On
Broadcast Get Device ID
reserved
Device Specific Commands
17.1
17.2
17.3
17.4
17.5
17.6
-
M
O
[3]
O
[3]
M
O
M
-
Get Device SDR Info
Get Device SDR
Reserve Device SDR Repository
Get Sensor Reading Factors
Set Sensor Hysteresis
Get Sensor Hysteresis
Set Sensor Threshold
Get Sensor Threshold
Set Sensor Event Enable
Get Sensor Event Enable
Re-arm Sensor Events
Get Sensor Event Status
reserved
Get Sensor Reading
Set Sensor Type
Get Sensor Type
29.2
29.3
29.4
29.5
29.6
29.7
29.8
29.9
29.10
29.11
29.12
29.13
29.14
29.15
29.16
O
[5]
O
[5]
O
[2]
O
O
O
O
[4]
O
O
[4]
O
[3]
O
O
M
O
[4]
O
Set Event Receiver
Get Event Receiver
Platform Event (a.k.a. Event Message)
23.1
23.2
23.3
M
[1]
M
[1]
M
1. 2. 3. 4.5.-
[1]
Mandatory for Event Message Generators only
Mandatory for Non-linear Sensors
Mandatory for manual re-arm Sensors
Mandatory if corresponding ‘Set’ command is implemented.
Mandatory per information returned in Get Device SDR Info
335
Intelligent Platform Management Interface Specification
29.1 Static and Dynamic Sensor Devices
Static Sensor Devices are defined as sensors that have their Sensor Data Records added to the SDR Repository
when the device is configured into the system. This is normally done either as part of the manufacturing of the
system, or via a separate utility when they are added to or deleted from the system configuration.
Dynamic Sensor Devices rely on being discovered by responding to a Broadcast Get Device ID formatted to their
slave address. (The IPMB format of this message is identical to that for a Get Device ID Request message that has
the entire message prefixed with the I2C broadcast slave address. [00h]) Once discovered, dynamic sensor devices
can be queried for their sensor population via the Get Device SDR Info command.
29.2 Get Device SDR Info Command
This command returns general information about the collection of sensors in a Dynamic Sensor Device.
Note: If the command is issued with no parameter for the request, the Device Sensor information is LUN based.
That is, it is returned individually for each LUN. E.g.. a device could implement four sensors under one LUN, and
twelve under another. The SDR Info does not return the aggregate of the sensor information. Rather, separate
‘Get Device SDR Info’ commands need to be issued to each LUN. The ‘Device LUNs’ field is provided in the
response to support this.
Software should assume an allocation unit size of 16-bytes if this command is not implementedfor device SDRs.
Table 29-2, Get Device SDR Info Command
Request Data
Response Data
336
-(1)
1
2
Operation (optional)
-[7:1] - reserved
[0] 1b = Get SDR count. This returns the total number of SDRs in the
device.
0b = Get Sensor count. This returns the number of sensors
implemented on LUN this command was addressed to.
Completion Code
For Operation = “Get Sensor Count” (or if byte 1 not present in request):
Number of sensors in device for LUN this command was addressed to.
For Operation = “Get SDR Count”:
Total Number of SDRs in the device.
3
Flags:
Dynamic population
[7] - 0b = static sensor population. The number of sensors handled by this
device is fixed, and a query shall return records for all sensors.
1b = dynamic sensor population. This device may have its sensor
population vary during ‘run time’ (defined as any time other that
when an install operation is in progress).
Reserved
[6:4] - reserved
Device LUNs
[3] - 1b = LUN 3 has sensors
[2] - 1b = LUN 2 has sensors
[1] - 1b = LUN 1 has sensors
[0] - 1b = LUN 0 has sensors
4:7
Sensor Population Change Indicator. LS byte first.
Four byte timestamp, or counter. Updated or incremented each time the
sensor population changes. This field is not provided if the flags indicate a
static sensor population.
Intelligent Platform Management Interface Specification
29.3 Get Device SDR Command
The ‘Get Device SDR’ command allows the population of sensors for the given LUN in the device to be listed
and the associated SDR information for those sensors sensors for a Sensor Device (typically implemented in a
satellite management controller) to be returned. The Get Device SDR Command can return any type of SDR, not
just Types 01h and 02h. This is an optional command for Static Sensor Devices, and mandatory for Dynamic
Sensor Devices. The format and action of this command is similar to that for the ‘Get SDR’ command for SDR
Repository Devices.
A Sensor Device shall always utilize the same sensor number for a particular sensor. This is mandatory to keep
System Event Log information consistent.
Sensor Devices that support the ‘Get Device SDR’ command return SDR Records that match the SDR Repository
formats. See section 37, Sensor Data Record FormatsSensor Data Record Formats.
The ‘Get Device SDR’ command includes a Reservation ID that is used to notify the Requester that a record may
have changed during the process of a multi-part read. See 27.11,Reserve SDR RepositoryReserve SDR Repository,
for more information on the function and use of the Reservation ID field.
Table 29-3, Get Device SDR Command
Request Data
1
Reservation ID. LS Byte. Only required for partial reads with a non-zero
‘Offset into record’ field. Use 0000h for reservation ID otherwise.
Reservation ID. MS Byte.
Response Data
2
3
4
5
6
1
Record ID of record to Get, LS Byte. 0000h returns the first record.
Record ID of record to Get, MS Byte
Offset into record
Bytes to read. FFh means read entire record.
Completion Code. Generic, plus following command specific:
80h = record changed. This status is returned if any of the record contents
have been altered since the last time the Requester issued the request
with 00h for the ‘Offset into SDR’ field.
Record ID for next record, LS Byt