- IPMI - Intelligent Platform Management Interface Specification Second Generation

- IPMI - Intelligent Platform Management Interface Specification Second Generation
- IPMI Intelligent Platform Management
Interface Specification
Second Generation
v2.0
Document Revision 1.0
February 12, 2004
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
1/27/04
See v1.5
spec
2/12/04
1.5
1.5
1.1
1.2
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
Markup updated per errata document version 5
IPMI 1.5 updated per errata document version 5
2.0
1.0
IPMI Second Generation document. Initial release.
Copyright © 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.
P
P
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.
P
P
P
P
P
P
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 Second Generation (v2.0),
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.
U
U
U
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.
U
U
U
U
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.
U
U
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.
U
U
5. Title. Title or the right to possess Confidential Information as between the parties will remain in the disclosing party.
U
U
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.
U
U
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.
U
U
iii
Intelligent Platform Management Interface Specification
Table of Contents
1.
TU
Introduction .......................................................................................................................................... 1
1.1 Audience .........................................................................................................................................................1
1.2 Reference Documents .....................................................................................................................................1
1.3 Conventions and Terminology ........................................................................................................................5
1.4 Background - Architectural Goals...................................................................................................................6
1.5 New for IPMI v1.5 ..........................................................................................................................................7
1.6 New for IPMI v2.0 ..........................................................................................................................................9
1.7 IPMI Overview..............................................................................................................................................11
1.7.1 Intelligent Platform Management...........................................................................................................11
1.7.2 IPMI Relationship to other Management Standards ..............................................................................11
1.7.3 Management Controllers and the IPMB.................................................................................................12
1.7.4 IPMI Messaging .....................................................................................................................................13
1.7.5 Sensor Model .........................................................................................................................................13
1.7.6 System Event Log and Event Messages .................................................................................................13
1.7.7 Sensor Data Records & Capabilities Commands ...................................................................................14
1.7.8 Initialization Agent ................................................................................................................................15
1.7.9 Sensor Data Record Repository .............................................................................................................15
1.7.10 Private Management Busses ..................................................................................................................15
1.7.11 FRU Information ....................................................................................................................................15
1.7.12 FRU Devices ..........................................................................................................................................16
1.7.13 Entity Association Records ....................................................................................................................16
1.7.14 Linkage between Events and FRU Information .....................................................................................16
1.7.15 Differentiation and Feature Extensibility ...............................................................................................16
1.7.16 System Interfaces ...................................................................................................................................17
1.7.17 Other Messaging Interfaces...................................................................................................................17
1.7.18 Serial/Modem Interface ..........................................................................................................................17
1.7.19 LAN Interface ........................................................................................................................................18
1.7.19.1 Payloads ............................................................................................................................................18
1.7.20 Serial Over LAN (SOL) .........................................................................................................................18
1.7.21 IPMI and ASF ........................................................................................................................................19
1.7.22 LAN Alerting .........................................................................................................................................19
1.7.23 Serial/Modem Alerting and Paging ........................................................................................................19
1.7.24 Platform Event Filtering (PEF) ..............................................................................................................20
1.7.25 Call Down Lists and Alert Policies ........................................................................................................20
1.7.26 Channel Model, Authentication, Sessions, and Users ............................................................................20
1.7.27 Standardized Watchdog Timer ...............................................................................................................21
1.7.28 Standardized POH Counter ....................................................................................................................21
1.7.29 Firmware Firewall ..................................................................................................................................21
1.7.30 Command and Function Discovery ........................................................................................................21
1.7.31 IPMI Hardware Components .................................................................................................................22
1.7.32 Configuration Interfaces ........................................................................................................................22
1.8 IPMI and BIOS .............................................................................................................................................22
1.9 System Management Software (SMS) ..........................................................................................................23
1.10 SMI Handler ..................................................................................................................................................23
1.11 Overview of Changes from IPMI v1.0 ..........................................................................................................25
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
TU
UT
TU
UT
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
Baseboard Management Controller (BMC) ..................................................................................... 29
3.1 Required BMC Functions..............................................................................................................................32
TU
UT
UT
TU
UT
General Mgmt. Controller Required Functions............................................................................... 35
UT
iv
UT
UT
Logical Management Device Types ................................................................................................. 26
4.
UT
UT
TU
TU
5.
UT
UT
UT
TU
UT
TU
UT
TU
TU
3.
TU
UT
UT
UT
UT
UT
TU
UT
TU
TU
2.
TU
TU
TU
UT
Message Interface Description ......................................................................................................... 37
TU
UT
Intelligent Platform Management Interface Specification
5.1 Network Function Codes...............................................................................................................................37
5.2 Completion Codes .........................................................................................................................................39
5.3 Completion Code Requirements ...................................................................................................................41
5.3.1 Response Field Truncation on non-zero Generic Completion Codes ....................................................41
5.3.2 Summary of Completion Code Use .......................................................................................................41
5.4 Sensor Owner Identification..........................................................................................................................42
5.5 Software IDs (SWIDs) ..................................................................................................................................42
5.6 Isolation from Message Content....................................................................................................................43
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
6.
TU
UT
UT
UT
UT
UT
UT
UT
IPMI Messaging Interfaces ................................................................................................................ 44
6.1 Terminology ..................................................................................................................................................44
6.2 Channel Model ..............................................................................................................................................44
6.3 Channel Numbers ..........................................................................................................................................45
6.4 Channel Protocol Type ..................................................................................................................................45
6.5 Channel Medium Type ..................................................................................................................................46
6.6 Channel Access Modes .................................................................................................................................47
6.7 Logical Channels ...........................................................................................................................................48
6.8 Channel Privilege Levels ..............................................................................................................................48
6.9 Users & Password Support............................................................................................................................49
6.9.1 ‘Anonymous Login’ Convention ...........................................................................................................49
6.9.2 Anonymous Login Status .......................................................................................................................49
6.10 System Interface Messaging..........................................................................................................................50
6.10.1 BMC Channels and Receive Message Queue ........................................................................................50
6.10.2 Event Message Buffer ............................................................................................................................50
6.11 System Interface Discovery and Multiple Interfaces ....................................................................................51
6.12 IPMI Sessions................................................................................................................................................52
6.12.1 Session-less Connections .......................................................................................................................52
6.12.2 Single-session Connections....................................................................................................................52
6.12.3 Multi-session Connections .....................................................................................................................52
6.12.4 Per-Message and User Level Authentication Disables ..........................................................................53
6.12.5 Link Authentication ...............................................................................................................................53
6.12.6 Summary of Connection Characteristics ................................................................................................54
6.12.7 Session Activation and IPMI Challenge-Response ................................................................................55
6.12.8 Session Sequence Numbers....................................................................................................................56
6.12.9 Session Sequence Number Generation ..................................................................................................56
6.12.10 Inbound Session Sequence Number Tracking and Handling .................................................................56
6.12.11 Out-of-order Packet Handling ................................................................................................................57
6.12.12 Outbound Session Sequence Number Tracking and Handling ..............................................................57
6.12.13 Session Inactivity Timeouts ...................................................................................................................57
6.12.13.1 Avoiding ‘Slot Stealing’ ..............................................................................................................58
6.12.14 Additional Session Specifications and Characteristics ..........................................................................58
6.13 BMC Message Bridging ................................................................................................................................59
6.13.1 BMC LUN 10b Routing .........................................................................................................................60
6.13.2 Send Message Command From System Interface ..................................................................................60
6.13.3 Send Message Command with Response Tracking ...............................................................................61
6.13.4 Bridged Request Example ......................................................................................................................62
6.14 Message Size & Private Bus Transaction Size Requirements .......................................................................64
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
UT
UT
TU
UT
UT
TU
TU
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
8.
UT
UT
UT
UT
UT
TU
UT
IPMB Interface .................................................................................................................................... 67
7.1 IPMB Access via Master Write-Read command...........................................................................................67
7.2 BMC IPMB LUNs ........................................................................................................................................67
7.3 Sending Messages to IPMB from System Software......................................................................................67
7.4 Sending IPMB Messages to System Software ..............................................................................................68
7.5 Testing for Event Message Buffer Support ...................................................................................................69
UT
TU
UT
UT
UT
7.
TU
UT
TU
TU
TU
TU
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
ICMB Interface .................................................................................................................................... 71
TU
UT
v
Intelligent Platform Management Interface Specification
8.1 Virtual ICMB Bridge Device ........................................................................................................................71
8.2 ICMB Bridge Commands in BMC using Channels ......................................................................................71
8.2.1 ICMB Bridging from System Interface to Remote IPMB using Channels ............................................71
8.2.2 ICMB Bridging from Local IPMB to Remote IPMB using Channels ...................................................72
UT
TU
TU
UT
TU
9.
TU
TU
UT
UT
TU
UT
TU
TU
UT
TU
UT
UT
Keyboard Controller Style (KCS) Interface ..................................................................................... 75
9.1 KCS Interface/BMC LUNs ...........................................................................................................................75
9.2 KCS Interface-BMC Request Message Format ............................................................................................75
9.3 BMC-KCS Interface Response Message Format ..........................................................................................76
9.4 Logging Events from System Software via KCS Interface ...........................................................................76
9.5 KCS Interface Registers ................................................................................................................................76
9.6 KCS Interface Control Codes ........................................................................................................................78
9.7 Status Register ...............................................................................................................................................78
9.7.1 SMS_ATN Flag Usage ..........................................................................................................................79
9.8 Command Register ........................................................................................................................................80
9.9 Data Registers ...............................................................................................................................................80
9.10 KCS Control Codes .......................................................................................................................................80
9.11 Performing KCS Interface Message Transfers ..............................................................................................80
9.12 KCS Communication and Non-communication Interrupts ...........................................................................81
9.13 Physical Interrupt Line Sharing.....................................................................................................................82
9.14 Additional Specifications for the KCS interface ...........................................................................................82
9.15 KCS Flow Diagrams .....................................................................................................................................83
9.16 Write Processing Summary ...........................................................................................................................87
9.17 Read Processing Summary ............................................................................................................................87
9.18 Error Processing Summary............................................................................................................................87
9.19 Interrupting Messages in Progress ................................................................................................................88
9.20 KCS Driver Design Recommendations .........................................................................................................88
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
TU
UT
UT
TU
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
10. SMIC Interface .................................................................................................................................... 90
10.1 SMS Transfer Streams ..................................................................................................................................90
10.2 SMIC Communication Register Overview ...................................................................................................90
10.3 SMIC/BMC Message Interface Registers .....................................................................................................91
10.3.1 Flags Register .........................................................................................................................................91
10.3.2 Control/Status Register ..........................................................................................................................92
10.3.2.1 Control and Status Codes..................................................................................................................92
10.3.3 Data Register ..........................................................................................................................................93
10.4 Performing a single SMIC/BMC Transaction ...............................................................................................93
10.5 Performing a SMIC/BMC Message Transfer ................................................................................................94
10.6 Interrupting Streams in Progress ...................................................................................................................94
10.7 Stream Switching ..........................................................................................................................................95
10.8 DATA_RDY Flag Handling .........................................................................................................................95
10.9 SMIC Control and Status Code Ranges ........................................................................................................96
10.10 SMIC SMS Stream Control Codes ................................................................................................................97
10.11 SMIC SMS Stream Status Codes ..................................................................................................................98
10.12 SMIC Messaging ...........................................................................................................................................99
10.13 SMIC/BMC LUNs ........................................................................................................................................99
10.14 SMIC-BMC Request Message Format..........................................................................................................99
10.15 BMC-SMIC Response Message Format .....................................................................................................100
10.16 Logging Events from System Software via SMIC ......................................................................................100
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
TU
UT
TU
TU
UT
TU
UT
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
11. Block Transfer (BT) Interface ......................................................................................................... 102
11.1 BT Interface-BMC Request Message Format .............................................................................................102
11.2 BMC-BT Interface Response Message Format ...........................................................................................103
11.3 Using the Seq Field .....................................................................................................................................103
11.4 Response Expiration Handling ....................................................................................................................104
11.5 Logging Events from System Software via BT Interface ...........................................................................104
TU
UT
vi
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
Intelligent Platform Management Interface Specification
11.6 Host to BMC Interface ................................................................................................................................105
11.6.1 BT Host Interface Registers .................................................................................................................105
11.6.2 BT BMC to Host Buffer (BMC2HOST) ..............................................................................................105
11.6.3 BT Host to BMC Buffer (HOST2BMC) ..............................................................................................105
11.6.4 BT Control Register (BT_CTRL) ........................................................................................................106
11.6.5 BT Interrupt Mask Register (INTMASK)............................................................................................108
11.7 Communication Protocol.............................................................................................................................109
11.8 Host and BMC Busy States .........................................................................................................................110
11.9 Host Command Power-On/Reset States ......................................................................................................110
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
12. SMBus System Interface (SSIF) ..................................................................................................... 112
12.1 Single Threaded Interface ...........................................................................................................................112
12.2 Single-part Write .........................................................................................................................................112
12.3 Multi-part Write ..........................................................................................................................................113
12.3.1 Error conditions for Multi-part Writes .................................................................................................113
12.4 Single-part Read Transaction ......................................................................................................................114
12.5 Multi-part Read Transactions ......................................................................................................................114
12.6 Retention of Output Data ............................................................................................................................115
12.7 SMBAlert Signal Handling .........................................................................................................................116
12.7.1 Enabling/disabling SSIF SMBAlert .....................................................................................................116
12.8 Polling for output data .................................................................................................................................116
12.9 SMBus NACKs and Error Recovery ..........................................................................................................116
12.10 PEC Handling..............................................................................................................................................116
12.11 SMBus Timeout and Hang Handling ..........................................................................................................117
12.12 Discovering SSIF ........................................................................................................................................117
12.13 SSIF Support Requirements for IPMI v1.5-only BMCs .............................................................................119
12.14 SSIF Support Requirements for IPMI v2.0 & Later BMCs ........................................................................119
12.15 Summary of SMBus Commands Values for SSIF ......................................................................................119
12.16 SSIF IPMI Commands ................................................................................................................................120
12.17 SSIF Timing ................................................................................................................................................120
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
TU
UT
TU
TU
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
13. IPMI LAN Interface ........................................................................................................................... 122
13.1 RMCP ..........................................................................................................................................................123
13.1.1 ASF Messages in RMCP......................................................................................................................123
13.1.2 RMCP Port Numbers ...........................................................................................................................124
13.1.3 RMCP Message Format .......................................................................................................................125
13.2 Required ASF/RMCP Messages for IPMI-over-LAN ................................................................................125
13.2.1 RMCP ACK Messages.........................................................................................................................126
13.2.2 RMCP ACK Handling .........................................................................................................................127
13.2.3 RMCP/ASF Presence Ping Message ....................................................................................................127
13.2.4 RMCP/ASF Pong Message (Ping Response) .......................................................................................128
13.3 RMCP+ .......................................................................................................................................................128
13.4 BMC Support Requirements for v1.5 and v2.0/RMCP+ Protocols ............................................................129
13.4.1 Session-less Command Support ...........................................................................................................129
13.5 IPMI Messages Encapsulation Under RMCP .............................................................................................130
13.5.1 RMCP/ASF and IPMI Byte Order .......................................................................................................130
13.6 Example IPMI over LAN Packet ................................................................................................................131
13.7 VLAN Support ............................................................................................................................................134
13.8 IPMI LAN Message Format ........................................................................................................................134
13.9 LAN Alerting ..............................................................................................................................................135
13.10 IPMI LAN Configuration ............................................................................................................................136
13.10.1 IP and MAC Address Configuration ....................................................................................................136
13.10.2 ‘Teamed’ and Fail-over LAN Channels ...............................................................................................136
13.11 ARP Handling and Gratuitous ARP ............................................................................................................136
13.11.1 OS-Absent problems with ARP ...........................................................................................................137
13.11.2 Resolving ARP issues ..........................................................................................................................137
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
TU
UT
UT
UT
UT
TU
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
UT
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
TU
UT
UT
TU
TU
UT
TU
TU
TU
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
UT
vii
Intelligent Platform Management Interface Specification
13.11.3 BMC-generated ARPs..........................................................................................................................138
13.12 Retaining IP Addresses in a DHCP Environment .......................................................................................138
13.12.1 Resolving DHCP issues .......................................................................................................................138
13.13 Discovering Support For IPMI over IP Connections ..................................................................................139
13.14 IPMI v1.5 LAN Session Activation ............................................................................................................139
13.15 IPMI v2.0/RMCP+ Session Activation .......................................................................................................141
13.16 RMCP+ Session Termination......................................................................................................................142
13.17 RMCP+ Open Session Request ...................................................................................................................142
13.18 RMCP+ Open Session Response ................................................................................................................144
13.19 RAKP Messages..........................................................................................................................................145
13.20 RAKP Message 1 ........................................................................................................................................145
13.21 RAKP Message 2 ........................................................................................................................................147
13.22 RAKP Message 3 ........................................................................................................................................148
13.23 RAKP Message 4 ........................................................................................................................................149
13.24 RMCP+ and RAKP Message Status Codes ................................................................................................150
13.25 Differences between v1.5 and v2.0/RMCP+ Sessions ................................................................................150
13.26 IPMI v2.0 RMCP+ Payload Types .............................................................................................................151
13.27 Payloads and Payload Type Numbers .........................................................................................................151
13.27.1 IPMI Message Payloads and IPMI Commands....................................................................................152
13.27.2 OEM Payload Type Handles ................................................................................................................152
13.27.3 Payload Type Numbers ........................................................................................................................153
13.28 Authentication, Integrity, and Confidentiality Algorithm Numbers ...........................................................153
13.28.1 RAKP-HMAC-SHA1 Authentication Algorithm ................................................................................153
13.28.2 RAKP-none Authentication Algorithm................................................................................................154
13.28.3 RAKP-HMAC-MD5 Algorithm ..........................................................................................................154
13.28.4 Integrity Algorithms .............................................................................................................................154
13.28.5 Confidentiality (Encryption) Algorithms .............................................................................................155
13.29 AES-CBC-128 Encrypted Payload Format .................................................................................................155
13.29.1 Generating the Initialization Vector .....................................................................................................155
13.29.2 Encryption with AES ...........................................................................................................................156
13.29.3 CBC (Cipher Block Chaining) .............................................................................................................157
13.30 xRC4 Encrypted Payload Format ................................................................................................................157
13.30.1 Generating the xRC4 Initialization Vector...........................................................................................157
13.30.2 Initializing the xRC4 State Machines...................................................................................................158
13.31 RMCP+ Authenticated Key-Exchange Protocol (RAKP)...........................................................................160
13.32 Generating Additional Keying Material ......................................................................................................162
13.33 Setting User Passwords and Keys ...............................................................................................................162
13.34 Random Number Generation ......................................................................................................................163
13.34.1 Random Number Key ..........................................................................................................................163
13.34.2 Random Number Generator Counters ..................................................................................................163
13.34.3 Random Number Generator Operation ................................................................................................163
TU
TU
UT
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
TU
UT
UT
UT
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
TU
TU
TU
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
14. IPMI Serial/Modem Interface ........................................................................................................... 164
14.1 Serial/Modem Capabilities ..........................................................................................................................164
14.2 Connection Modes ......................................................................................................................................164
14.2.1 PPP/UDP Proxy Operation ..................................................................................................................165
14.2.2 Asynchronous Communication Parameters .........................................................................................165
14.2.3 Serial Port Sharing ...............................................................................................................................166
14.2.4 Serial Port Switching ...........................................................................................................................167
14.2.5 Access Modes ......................................................................................................................................167
14.2.6 Console Redirection with Serial Port Sharing......................................................................................167
14.2.6.1 Detecting Who Answered The Phone.............................................................................................168
14.2.6.2 Connecting to the BMC ..................................................................................................................168
14.2.6.3 Connecting to the Console Redirection ..........................................................................................169
14.2.6.4 Directing the Connection After Power Up / Reset..........................................................................169
TU
UT
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
viii
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
Intelligent Platform Management Interface Specification
14.2.6.5 Interaction with Microsoft ‘Headless’ Operation ...........................................................................169
14.2.6.6 Pre-boot Only Mode .......................................................................................................................169
14.2.6.7 Always Available Mode .................................................................................................................170
14.2.6.8 Shared Mode ...................................................................................................................................170
14.2.7 Serial Port Sharing Access Characteristics ..........................................................................................170
14.2.8 Serial Port Sharing Hardware Implementation Notes ..........................................................................172
14.2.9 Connection Mode Auto-detect .............................................................................................................173
14.2.10 Modem-specific Options ......................................................................................................................175
14.2.11 Modem Activation ...............................................................................................................................175
14.3 Serial/Modem Connection Active (Ping) Message .....................................................................................176
14.3.1 Serial/Modem Connection Active Message Parameters ......................................................................177
14.3.2 Mux Switch Coordination ....................................................................................................................177
14.3.3 Receive During Ping ............................................................................................................................177
14.3.4 Application Handling of the Serial/Modem Connection Active Message ...........................................177
14.4 Basic Mode..................................................................................................................................................178
14.4.1 Basic Mode Packet Framing ................................................................................................................178
14.4.2 Data Byte Escaping ..............................................................................................................................178
14.4.3 Message Fields .....................................................................................................................................179
14.4.4 Message Retries ...................................................................................................................................180
14.4.5 Packet Handshake ................................................................................................................................180
14.5 PPP/UDP Mode ...........................................................................................................................................181
14.5.1 PPP/UDP Mode Sessions .....................................................................................................................181
14.5.2 PPP Frame Format ...............................................................................................................................181
14.5.3 PPP Frame Implementation Requirements...........................................................................................181
14.5.4 Link Control Protocol (LCP) packets ..................................................................................................182
14.5.5 Configuration Requests ........................................................................................................................182
14.5.6 Maximum Receive Unit Handling .......................................................................................................184
14.5.7 Protocol Field Compression Handling .................................................................................................184
14.5.8 Address & Control Field Compression Handling ................................................................................184
14.5.9 IPMI/RMCP Message Format in PPP Frame .......................................................................................185
14.5.10 Example of IPMI Frame with Field Compression ...............................................................................186
14.5.11 Frame Data Encoding...........................................................................................................................186
14.5.12 Escaping Algorithm .............................................................................................................................186
14.5.13 Escaped Character Handling ................................................................................................................186
14.5.14 Asynch Control Character Maps (ACCM)...........................................................................................186
14.5.15 IP Network Protocol Negotiation (IPCP) .............................................................................................187
14.5.16 CHAP Operation in PPP Mode ............................................................................................................188
14.6 Serial/Modem Callback ...............................................................................................................................189
14.6.1 Callback Control Protocol (CBCP) Support ........................................................................................189
14.6.1.1 CBCP Address Type and Dial String Characters ...........................................................................190
14.7 Terminal Mode ............................................................................................................................................190
14.7.1 Terminal Mode Versus Basic Mode Differences .................................................................................191
14.7.2 Terminal Mode Message Format .........................................................................................................191
14.7.3 IPMI Message Data ..............................................................................................................................191
14.7.4 Terminal Mode IPMI Message Bridging .............................................................................................193
14.7.5 Sending Messages to SMS ...................................................................................................................193
14.7.6 Sending Messages to Other Media .......................................................................................................194
14.7.7 Terminal Mode Packet Handshake ......................................................................................................195
14.7.8 Terminal Mode ASCII Text Commands ..............................................................................................195
14.7.9 Terminal Mode Text Command and IPMI Message Examples ...........................................................198
14.8 Terminal Mode Line Editing .......................................................................................................................198
14.9 Terminal Mode Input Restrictions ..............................................................................................................199
14.10 Page Blackout Interval ................................................................................................................................199
14.11 Dial Paging ..................................................................................................................................................199
14.11.1 Alert Strings for Dial Paging ...............................................................................................................200
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
TU
UT
TU
UT
TU
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
TU
TU
TU
TU
TU
UT
UT
UT
TU
TU
UT
UT
TU
UT
TU
UT
UT
UT
UT
UT
TU
UT
ix
Intelligent Platform Management Interface Specification
14.11.2 Dialing Digits .......................................................................................................................................200
14.11.3 <Enter> Character (control-M) ............................................................................................................200
14.11.4 Long Pause Character (control-L) ........................................................................................................200
14.11.5 Empty (delimiter) Character (FFh) ......................................................................................................200
14.11.6 ‘Null’ Terminator Character (00h) .......................................................................................................200
14.12 TAP Paging .................................................................................................................................................201
14.12.1 TAP Escaping (data transparency) .......................................................................................................202
14.12.2 TAP Checksum ....................................................................................................................................202
14.12.3 TAP Response Codes ...........................................................................................................................202
14.12.4 TAP Page Success Criteria...................................................................................................................202
14.13 PPP Alerting ................................................................................................................................................203
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
TU
TU
UT
UT
UT
UT
UT
UT
TU
UT
15. Serial Over LAN................................................................................................................................ 204
15.1 System Serial Controller Requirements ......................................................................................................204
15.2 SOL and Serial Port Sharing .......................................................................................................................204
15.3 SOL Operation Overview ...........................................................................................................................205
15.4 SOL Security ...............................................................................................................................................206
15.5 SOL Sequence Numbers .............................................................................................................................206
15.6 Flow Control ...............................................................................................................................................206
15.7 Bit Rate Handling ........................................................................................................................................206
15.8 Volatile and Non-volatile SOL Configuration Parameters .........................................................................206
15.9 SOL Payload Data Format ..........................................................................................................................207
15.10 Activating SOL using RMCP+ Authentication ...........................................................................................209
15.11 SOL Packet Acknowledge and Retries .......................................................................................................209
15.12 SOL Interaction with Windows.NET Escape Sequences ............................................................................211
15.13 SOL Payload Activated with Serial Port Sharing .......................................................................................212
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
16. Event Messages ............................................................................................................................... 213
16.1 Critical Events and System Event Log Restrictions ....................................................................................213
16.2 Event Receiver Handling of Event Messages .............................................................................................214
16.3 IPMB Seq Field use in Event Messages ......................................................................................................215
16.4 Event Status, Event Conditions, and Present State .....................................................................................216
16.5 System Software use of Sensor Scanning bits & Entity Info ......................................................................216
16.6 Re-arming ....................................................................................................................................................217
16.6.1 ‘Global’ Re-arm ...................................................................................................................................217
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
TU
UT
TU
UT
17. ‘Platform Event Filtering (PEF) ....................................................................................................... 219
17.1 Alert Policies ...............................................................................................................................................219
17.2 Deferred Alerts ............................................................................................................................................219
17.3 PEF Postpone Timer....................................................................................................................................219
17.4 PEF Startup Delay .......................................................................................................................................220
17.4.1 Last Processed Event Tracking ............................................................................................................220
17.5 Event Processing When The SEL Is Full ....................................................................................................220
17.6 PEF Actions ................................................................................................................................................221
17.7 Event Filter Table ........................................................................................................................................221
17.8 Event Data 1 Event Offset Mask .................................................................................................................224
17.9 Using the Mask and Compare Fields ..........................................................................................................224
17.10 Mask and Compare Field Examples ............................................................................................................224
17.11 Alert Policy Table .......................................................................................................................................225
17.12 Alert Testing................................................................................................................................................227
17.13 Alert Processing ..........................................................................................................................................227
17.13.1 Alert Processing after Power Loss .......................................................................................................227
17.13.2 Processing non-Alert Actions after Power Loss ..................................................................................227
17.13.3 Alert Processing when IPMI Messaging is in Progress .......................................................................227
17.13.4 Sending Multiple Alerts On One Call ..................................................................................................228
17.13.5 Serial/Modem Alert Processing ...........................................................................................................228
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
TU
UT
UT
x
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
TU
TU
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
Intelligent Platform Management Interface Specification
17.14 PEF and Alert Handling Example ...............................................................................................................229
17.15 Event Filter, Policy, Destination, and String Relationships ........................................................................230
17.16 Populating a PET.........................................................................................................................................231
17.16.1 OEM Custom Fields and Text Alert Strings for IPMI v1.5 PET .........................................................233
17.17 PEF Performance Target .............................................................................................................................233
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
TU
UT
TU
TU
UT
UT
18. Firmware Firewall & Command Discovery .................................................................................... 235
TU
UT
TU
UT
19. Command Specification Information ............................................................................................. 237
19.1 Specification of Completion Codes .............................................................................................................237
19.2 Handling ‘Reserved’ Bits and Fields ..........................................................................................................237
19.3 Logical Unit Numbers (LUNs) for Commands ...........................................................................................237
19.4 Command Table Notation ...........................................................................................................................237
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
20. IPM Device “Global” Commands ................................................................................................... 239
20.1 Get Device ID Command ............................................................................................................................240
20.2 Cold Reset Command..................................................................................................................................243
20.3 Warm Reset Command ...............................................................................................................................243
20.4 Get Self Test Results Command .................................................................................................................244
20.5 Manufacturing Test On Command ..............................................................................................................244
20.6 Set ACPI Power State Command ................................................................................................................245
20.7 Get ACPI Power State Command ...............................................................................................................247
20.8 Get Device GUID Command ......................................................................................................................248
20.9 Broadcast ‘Get Device ID’ ..........................................................................................................................249
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
21. Firmware Firewall & Command Discovery Commands ............................................................... 250
21.1 Completion Codes with Firmware Firewall ................................................................................................250
21.2 Get NetFn Support Command .....................................................................................................................251
21.3 Get Command Support Command ..............................................................................................................252
21.4 Get Command Sub-function Support Command ........................................................................................253
21.5 Get Configurable Commands Command ....................................................................................................254
21.6 Get Configurable Command Sub-functions Command...............................................................................255
21.7 Set Command Enables Command ...............................................................................................................256
21.8 Get Command Enables Command ..............................................................................................................257
21.9 Set Configurable Command Sub-function Enables Command ...................................................................258
21.10 Get Configurable Command Sub-function Enables Command ..................................................................259
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
22. IPMI Messaging Support Commands ............................................................................................ 260
22.1 Set BMC Global Enables Command ...........................................................................................................261
22.2 Get BMC Global Enables Command ..........................................................................................................261
22.3 Clear Message Flags Command ..................................................................................................................262
22.4 Get Message Flags Command .....................................................................................................................262
22.5 Enable Message Channel Receive Command .............................................................................................263
22.6 Get Message Command...............................................................................................................................263
22.7 Send Message Command ............................................................................................................................266
22.8 Read Event Message Buffer Command ......................................................................................................269
22.9 Get System Interface Capabilities Command ..............................................................................................269
22.10 Get BT Interface Capabilities Command ....................................................................................................271
22.11 Master Write-Read Command.....................................................................................................................272
22.12 Session Header Fields .................................................................................................................................272
22.13 Get Channel Authentication Capabilities Command...................................................................................273
22.14 Get System GUID Command ......................................................................................................................276
22.15 Get Channel Cipher Suites Command.........................................................................................................276
22.15.1 Cipher Suite Records ...........................................................................................................................277
22.15.2 Cipher Suite IDs ...................................................................................................................................278
22.16 Get Session Challenge Command ...............................................................................................................279
22.17 Activate Session Command.........................................................................................................................280
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
TU
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
UT
UT
UT
xi
Intelligent Platform Management Interface Specification
22.17.1 AuthCode Algorithms ..........................................................................................................................283
22.18 Set Session Privilege Level Command........................................................................................................283
22.19 Close Session Command .............................................................................................................................284
22.20 Get Session Info Command.........................................................................................................................284
22.21 Get AuthCode Command ............................................................................................................................286
22.22 Set Channel Access Command....................................................................................................................288
22.23 Get Channel Access Command ...................................................................................................................290
22.24 Get Channel Info Command .......................................................................................................................291
22.25 Set Channel Security Keys Command ........................................................................................................292
22.26 Set User Access Command .........................................................................................................................293
22.27 Get User Access Command .........................................................................................................................295
22.28 Set User Name Command ...........................................................................................................................296
22.29 Get User Name Command ..........................................................................................................................296
22.30 Set User Password Command .....................................................................................................................297
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
23. IPMI LAN Commands....................................................................................................................... 299
23.1 Set LAN Configuration Parameters Command ...........................................................................................299
23.2 Get LAN Configuration Parameters Command ..........................................................................................300
23.3 Suspend BMC ARPs Command..................................................................................................................305
23.4 Get IP/UDP/RMCP Statistics Command ....................................................................................................307
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
24. RMCP+ Support and Payload Commands .................................................................................... 308
24.1 Activate Payload Command ........................................................................................................................308
24.2 Deactivate Payload Command ....................................................................................................................310
24.3 Suspend/Resume Payload Encryption Command .......................................................................................311
24.4 Get Payload Activation Status Command ...................................................................................................312
24.5 Get Payload Instance Info Command ..........................................................................................................313
24.6 Set User Payload Access Command ...........................................................................................................314
24.7 Get User Payload Access Command ...........................................................................................................315
24.8 Get Channel Payload Support Command ....................................................................................................315
24.9 Get Channel Payload Version Command ....................................................................................................316
24.10 Get Channel OEM Payload Info Command ................................................................................................317
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
25. IPMI Serial/Modem Commands ...................................................................................................... 318
25.1 Set Serial/Modem Configuration Command ...............................................................................................318
25.2 Get Serial/Modem Configuration Command ..............................................................................................319
25.3 Set Serial/Modem Mux Command ..............................................................................................................336
25.4 Get TAP Response Codes Command ..........................................................................................................337
25.5 Set PPP UDP Proxy Transmit Data Command ...........................................................................................338
25.6 Get PPP UDP Proxy Transmit Data Command ..........................................................................................338
25.7 Send PPP UDP Proxy Packet Command ....................................................................................................339
25.8 Get PPP UDP Proxy Receive Data Command ............................................................................................340
25.9 Serial/Modem Connection Active (Ping) Command...................................................................................341
25.10 Callback Command .....................................................................................................................................341
25.11 Set User Callback Options Command .........................................................................................................342
25.12 Get User Callback Options Command ........................................................................................................343
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
UT
26. SOL Commands ............................................................................................................................... 344
26.1 SOL Activating Command ..........................................................................................................................344
26.2 Set SOL Configuration Parameters Command............................................................................................345
26.3 Get SOL Configuration Parameters Command ...........................................................................................345
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
27. BMC Watchdog Timer Commands ................................................................................................. 349
27.1 Watchdog Timer Actions ............................................................................................................................349
27.2 Watchdog Timer Use Field and Expiration Flags .......................................................................................349
27.2.1 Using the Timer Use field and Expiration flags ...................................................................................350
27.3 Watchdog Timer Event Logging .................................................................................................................350
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
TU
TU
xii
UT
UT
TU
TU
UT
UT
Intelligent Platform Management Interface Specification
27.4 Pre-timeout Interrupt ...................................................................................................................................350
27.4.1 Pre-timeout Interrupt Support Detection ..............................................................................................350
27.4.2 BIOS Support for Watchdog Timer .....................................................................................................351
27.5 Reset Watchdog Timer Command ..............................................................................................................351
27.6 Set Watchdog Timer Command ..................................................................................................................351
27.7 Get Watchdog Timer Command .................................................................................................................353
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
28. Chassis Commands ........................................................................................................................ 355
28.1 Get Chassis Capabilities Command ............................................................................................................355
28.2 Get Chassis Status Command......................................................................................................................357
28.3 Chassis Control Command ..........................................................................................................................358
28.4 Chassis Reset Command .............................................................................................................................358
28.5 Chassis Identify Command .........................................................................................................................359
28.6 Set Front Panel Enables...............................................................................................................................359
28.7 Set Chassis Capabilities Command .............................................................................................................359
28.8 Set Power Restore Policy Command...........................................................................................................360
28.9 Set Power Cycle Interval .............................................................................................................................361
28.10 Remote Access Boot control .......................................................................................................................361
28.11 Get System Restart Cause Command ..........................................................................................................362
28.12 Set System Boot Options Command ...........................................................................................................362
28.13 Get System Boot Options Command ..........................................................................................................363
28.14 Get POH Counter Command.......................................................................................................................368
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
29. Event Commands............................................................................................................................. 369
29.1 Set Event Receiver Command .....................................................................................................................369
29.2 Get Event Receiver Command ....................................................................................................................370
29.3 Platform Event Message Command ............................................................................................................370
29.4 Event Request Message Fields ....................................................................................................................371
29.5 IPMB Event Message Formats ....................................................................................................................371
29.6 System Interface Event Request Message Format ......................................................................................371
29.7 Event Data Field Formats ............................................................................................................................373
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
30. PEF and Alerting Commands ......................................................................................................... 375
30.1 Get PEF Capabilities Command..................................................................................................................375
30.2 Arm PEF Postpone Timer Command ..........................................................................................................375
30.3 Set PEF Configuration Parameters Command ............................................................................................376
30.4 Get PEF Configuration Parameters Command............................................................................................377
30.5 Set Last Processed Event ID Command ......................................................................................................382
30.6 Get Last Processed Event ID Command .....................................................................................................382
30.7 Alert Immediate Command .........................................................................................................................382
30.8 PET Acknowledge Command .....................................................................................................................384
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
31. System Event Log (SEL) Commands............................................................................................. 385
31.1 SEL Device Commands ..............................................................................................................................385
31.2 Get SEL Info Command ..............................................................................................................................386
31.3 Get SEL Allocation Info Command ............................................................................................................387
31.4 Reserve SEL Command ..............................................................................................................................387
31.4.1 Reservation Restricted Commands ......................................................................................................388
31.4.2 Reservation Cancellation .....................................................................................................................388
31.5 Get SEL Entry Command............................................................................................................................389
31.6 Add SEL Entry Command ..........................................................................................................................389
31.6.1 SEL Record Type Ranges ....................................................................................................................390
31.7 Partial Add SEL Entry Command ...............................................................................................................391
31.8 Delete SEL Entry Command .......................................................................................................................391
31.9 Clear SEL Command ..................................................................................................................................392
31.10 Get SEL Time Command ............................................................................................................................392
31.11 Set SEL Time Command .............................................................................................................................392
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
UT
xiii
Intelligent Platform Management Interface Specification
31.12 Get Auxiliary Log Status Command ...........................................................................................................393
31.13 Set Auxiliary Log Status Command ............................................................................................................394
TU
UT
TU
TU
UT
TU
UT
UT
32. SEL Record Formats ....................................................................................................................... 395
32.1 SEL Event Records .....................................................................................................................................395
32.2 OEM SEL Record - Type C0h-DFh ............................................................................................................396
32.3 OEM SEL Record - Type E0h-FFh.............................................................................................................396
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
33. SDR Repository................................................................................................................................ 397
33.1 SDR Repository Device ..............................................................................................................................397
33.2 Modal and Non-modal SDR Repositories ...................................................................................................398
33.2.1 Command Support while in SDR Repository Update Mode ...............................................................398
33.3 Populating the SDR Repository ..................................................................................................................398
33.3.1 SDR Repository Updating ...................................................................................................................399
33.4 Discovering Management Controllers and Device SDRs ...........................................................................399
33.5 Reading the SDR Repository ......................................................................................................................399
33.6 Sensor Initialization Agent ..........................................................................................................................400
33.6.1 System Support Requirements for the Initialization Agent..................................................................400
33.6.2 IPMI and ACPI Interaction ..................................................................................................................400
33.6.3 Recommended Initialization Agent Steps ............................................................................................401
33.7 SDR Repository Device Commands ...........................................................................................................402
33.8 SDR ‘Record IDs’ .......................................................................................................................................402
33.9 Get SDR Repository Info Command...........................................................................................................404
33.10 Get SDR Repository Allocation Info Command .........................................................................................405
33.11 Reserve SDR Repository Command ...........................................................................................................405
33.11.1 Reservation Restricted Commands ......................................................................................................406
33.11.2 Reservation Cancellation .....................................................................................................................406
33.12 Get SDR Command .....................................................................................................................................406
33.13 Add SDR Command....................................................................................................................................408
33.14 Partial Add SDR Command ........................................................................................................................408
33.15 Delete SDR Command ................................................................................................................................409
33.16 Clear SDR Repository Command ...............................................................................................................409
33.17 Get SDR Repository Time Command .........................................................................................................410
33.18 Set SDR Repository Time Command..........................................................................................................410
33.19 Enter SDR Repository Update Mode Command.........................................................................................410
33.20 Exit SDR Repository Update Mode Command...........................................................................................411
33.21 Run Initialization Agent Command ............................................................................................................411
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
UT
UT
TU
UT
TU
TU
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
UT
TU
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
UT
TU
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
UT
34. FRU Inventory Device Commands ................................................................................................. 412
34.1 Get FRU Inventory Area Info Command ....................................................................................................412
34.2 Read FRU Data Command ..........................................................................................................................413
34.3 Write FRU Data Command .........................................................................................................................413
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
35. Sensor Device Commands ............................................................................................................. 416
35.1 Static and Dynamic Sensor Devices ............................................................................................................417
35.2 Get Device SDR Info Command .................................................................................................................417
35.3 Get Device SDR Command ........................................................................................................................418
35.4 Reserve Device SDR Repository Command ...............................................................................................420
35.5 Get Sensor Reading Factors Command.......................................................................................................421
35.6 Set Sensor Hysteresis Command .................................................................................................................421
35.7 Get Sensor Hysteresis Command ................................................................................................................422
35.8 Set Sensor Thresholds Command................................................................................................................422
35.9 Get Sensor Thresholds Command ...............................................................................................................423
35.10 Set Sensor Event Enable Command ............................................................................................................424
35.11 Get Sensor Event Enable Command ...........................................................................................................426
35.12 Re-arm Sensor Events Command................................................................................................................427
35.13 Get Sensor Event Status Command.............................................................................................................429
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
xiv
UT
UT
UT
UT
Intelligent Platform Management Interface Specification
35.13.1 Response According to Sensor Type ...................................................................................................429
35.13.2 Hysteresis and Event Status .................................................................................................................430
35.13.3 High-going versus Low-going Threshold Events ................................................................................430
35.13.4 Get Sensor Event Status Command Format .........................................................................................431
35.14 Get Sensor Reading Command ...................................................................................................................434
35.15 Set Sensor Type Command .........................................................................................................................435
35.16 Get Sensor Type Command ........................................................................................................................436
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
UT
UT
36. Sensor Types and Data Conversion .............................................................................................. 437
36.1 Linear and Linearized Sensors ....................................................................................................................437
36.2 Non-Linear Sensors .....................................................................................................................................437
36.3 Sensor Reading Conversion Formula ..........................................................................................................438
36.4 Resolution, Tolerance and Accuracy...........................................................................................................438
36.4.1 Tolerance..............................................................................................................................................438
36.4.2 Resolution ............................................................................................................................................438
36.4.2.1 Resolution for Non-linear & Linearizable Sensors.........................................................................439
36.4.2.2 Offset Constant Relationship to Resolution....................................................................................439
36.5 Management Software, SDRs, and Sensor Display ....................................................................................439
36.5.1 Software Display of Threshold Settings ..............................................................................................439
36.5.2 Notes on Displaying Sensor Readings & Thresholds ..........................................................................440
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
UT
TU
UT
TU
TU
UT
TU
UT
UT
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
UT
37. Timestamp Format ........................................................................................................................... 442
37.1 Special Timestamp values ...........................................................................................................................442
TU
UT
TU
UT
TU
UT
TU
UT
38. Accessing FRU Devices .................................................................................................................. 443
TU
UT
TU
UT
39. Using Entity IDs ............................................................................................................................... 445
39.1 System- and Device-relative Entity Instance Values ..................................................................................445
39.2 Restrictions on Using Device-relative Entity Instance Values....................................................................446
39.3 Sensor-to-FRU Association.........................................................................................................................446
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
40. Handling Sensor Associations ....................................................................................................... 447
40.1 Entity Presence ............................................................................................................................................447
40.2 Software detection of Entities .....................................................................................................................447
40.3 Using Entity Association Records...............................................................................................................448
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
41. Sensor & Event Message Codes .................................................................................................... 450
41.1 Sensor Type Code .......................................................................................................................................450
41.2 Event/Reading Type Code ..........................................................................................................................450
41.3 SDR Specification of Event Types..............................................................................................................451
41.4 SDR Specification of Reading Types..........................................................................................................451
41.5 Use of Codes in Event Messages ................................................................................................................452
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
UT
UT
42. Sensor and Event Code Tables ...................................................................................................... 454
42.1 Event/Reading Type Codes .........................................................................................................................454
42.2 Sensor Type Codes and Data ......................................................................................................................458
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
UT
43. Sensor Data Record Formats ......................................................................................................... 470
43.1 SDR Type 01h, Full Sensor Record ............................................................................................................471
43.2 SDR Type 02h, Compact Sensor Record ....................................................................................................478
43.3 SDR Type 03h, Event-Only Record ............................................................................................................484
43.4 SDR Type 08h - Entity Association Record ...............................................................................................486
43.5 SDR Type 09h - Device-relative Entity Association Record ......................................................................488
43.6 SDR Type 0Ah:0Fh - Reserved Records ....................................................................................................489
43.7 SDR Type 10h - Generic Device Locator Record .......................................................................................490
43.8 SDR Type 11h - FRU Device Locator Record............................................................................................491
43.9 SDR Type 12h - Management Controller Device Locator Record .............................................................493
43.10 SDR Type 13h - Management Controller Confirmation Record ................................................................495
43.11 SDR Type 14h - BMC Message Channel Info Record ...............................................................................496
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
UT
UT
UT
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
UT
TU
UT
UT
xv
Intelligent Platform Management Interface Specification
43.12
43.13
43.14
43.15
43.16
43.17
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
SDR Type C0h - OEM Record ...................................................................................................................498
Device Type Codes .....................................................................................................................................499
Entity IDs ....................................................................................................................................................500
Type/Length Byte Format ...........................................................................................................................501
6-bit ASCII Packing Example .....................................................................................................................502
Sensor Unit Type Codes..............................................................................................................................503
TU
UT
UT
TU
UT
TU
UT
TU
UT
TU
UT
44. Examples .......................................................................................................................................... 504
44.1 Processor Sensor with Sensor-specific States & Event Generation ............................................................504
44.2 Processor Sensor with Generic States & Event Generation ........................................................................506
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
UT
UT
Appendix A - Previous Sequence Number Tracking ........................................................................... 507
TU
UT
Appendix B - Example PEF Mask Compare Algorithm ....................................................................... 509
TU
UT
Appendix C1 - Locating IPMI System Interfaces via SM BIOS Tables ............................................... 511
C1-1 IPMI Device Information - BMC Interface .................................................................................................512
C1-1.1 Interface Type ......................................................................................................................................512
C1-1.2 IPMI Specification Revision Field .......................................................................................................512
C1-1.3 I2C Slave Address Field .......................................................................................................................512
C1-1.4 NV Storage Device Address Field .......................................................................................................513
C1-1.5 Base Address Field...............................................................................................................................513
C1-1.6 Base Address Modifier Field ...............................................................................................................513
C1-1.7 System Interface Register Alignment...................................................................................................513
C1-1.7.1 Byte-spaced I/O Address Examples ...............................................................................................513
C1-1.7.2 32-bit Spaced I/O Address Examples .............................................................................................513
C1-1.7.3 Memory-mapped Base Address ......................................................................................................514
C1-1.7.4 Interrupt Info Field .........................................................................................................................514
C1-1.8 Interrupt Number Field ........................................................................................................................514
TU
UT
TU
UT
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UP
UP
UT
UT
UT
UT
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
TU
TU
TU
UT
UT
UT
UT
UT
TU
UT
Appendix C2 - Locating IPMI System Interfaces on PCI ..................................................................... 515
TU
UT
Appendix C3 - Locating IPMI System Interfaces with ACPI ............................................................... 517
C3-1 SPMI Description Table and ACPI Control Methods .................................................................................517
C3-2 Locating IPMI System Interfaces in ACPI Name Space.............................................................................519
C3-3 Example IPMI Definition ASL Code ..........................................................................................................521
Example 1: SMIC Interface in I/O Space ..........................................................................................................521
Example 2: KCS Interface in 64-bit Address Space ..........................................................................................521
Example 3: SMIC Interface in I/O Space ..........................................................................................................522
Example 4: SSIF Interface .................................................................................................................................523
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
TU
UT
TU
UT
TU
UT
TU
UT
Appendix D - Determining Message Size Requirements .................................................................... 525
TU
UT
Appendix E - Terminal Mode Grammar................................................................................................. 527
E-1 Notation .......................................................................................................................................................527
E-2 Grammar for Terminal Mode Input.............................................................................................................527
E-3 Grammar for Terminal Mode Output ..........................................................................................................528
TU
UT
TU
UT
TU
TU
UT
TU
TU
UT
TU
UT
UT
UT
Appendix F - TAP Flow Summary ......................................................................................................... 531
TU
UT
Appendix G - Command Assignments ................................................................................................. 535
TU
UT
Appendix H - Sub-function Assignments ............................................................................................. 541
TU
UT
xvi
Intelligent Platform Management Interface Specification
List of Figures
Figure 1-1, IPMI and the Management Software Stack ..............................................................................................11
Figure 1-2, IPMI Block Diagram ................................................................................................................................12
Figure 2-1, Intelligent Platform Management Logical Devices ..................................................................................28
Figure 6-1, Session Activation ....................................................................................................................................55
Figure 6-2, LAN to IPMB Bridged Request Example ................................................................................................63
Figure 7-1, IPMB Request sent using Send Message Command ................................................................................68
Figure 7-2, Send Message Command Response..........................................................................................................68
Figure 7-3, Response for Set Event Receiver in Receive Message Queue..................................................................69
Figure 7-4, Get Message Command Response............................................................................................................69
Figure 9-1, KCS Interface/BMC Request Message Format ........................................................................................75
Figure 9-2, KCS Interface/BMC Response Message Format......................................................................................76
Figure 9-3, KCS Interface Event Request Message Format........................................................................................76
Figure 9-4, KCS Interface Event Response Message Format .....................................................................................76
Figure 9-5, KCS Interface Registers............................................................................................................................78
Figure 9-6, KCS Interface SMS to BMC Write Transfer Flow Chart.........................................................................84
Figure 9-7, KCS Interface BMC to SMS Read Transfer Flow Chart..........................................................................85
Figure 9-8, Aborting KCS Transactions in-progress and/or Retrieving KCS Error Status .........................................86
Figure 10-1, SMIC/BMC Interface Registers..............................................................................................................91
Figure 10-2, SMIC/BMC Request Message Format ...................................................................................................99
Figure 10-3, SMIC/BMC Response Message Format ...............................................................................................100
Figure 10-4, SMIC Event Request Message Format .................................................................................................100
Figure 10-5, SMIC Event Response Message Format...............................................................................................100
Figure 11-1, BT Interface/BMC Request Message Format.......................................................................................102
Figure 11-2, BT Interface/BMC Response Message Format ....................................................................................103
Figure 11-3, BT Interface Event Request Message Format.......................................................................................104
Figure 11-4, BT Interface Event Response Message Format ....................................................................................104
Figure 11-5, BT_CTRL Register format ...................................................................................................................106
Figure 11-6, BT_INTMASK Register format ...........................................................................................................108
Figure 13-1, Embedded LAN Controller Implementation.........................................................................................122
Figure 13-2, PCI Management Bus Implementation .................................................................................................123
Figure 13-3, IPMI LAN Packet Layering..................................................................................................................130
Figure 13-4, IPMI LAN Message Formats................................................................................................................135
Figure 13-5, LAN Session Startup ............................................................................................................................141
Figure 14-1, Serial Port Sharing Logical Diagram....................................................................................................166
Figure 14-2, Basic Mode Message Fields .................................................................................................................179
Figure 14-3, PPP Frame Format ................................................................................................................................181
Figure 14-4, Configure-Request, -Ack, -Nak, -Reject Packet Format ......................................................................182
Figure 14-5, IPMI Message in PPP Frame Format....................................................................................................185
Figure 14-6, IP Frame with Field Compression ........................................................................................................186
Figure 14-7, Terminal Mode Request to BMC..........................................................................................................192
Figure 14-8, Terminal Mode Response from BMC...................................................................................................192
Figure 14-9, Terminal Mode Request to SMS...........................................................................................................194
Figure 14-10, Terminal Mode Response from SMS .................................................................................................194
Figure 14-11, Send Message Command for Bridged Request...................................................................................194
Figure 14-12, Response to Send Message Command for Bridged Request ..............................................................194
Figure 14-13, Bridged Response to Remote Console................................................................................................194
Figure 15-1, SOL with Serial Port Sharing ...............................................................................................................205
Figure 17-1, Alert Processing Example.....................................................................................................................230
xvii
Intelligent Platform Management Interface Specification
Figure 17-2, Event Filter, Alert Policy, and Alert Destination, & String Relationships ...........................................231
Figure 20-1, Broadcast Get Device ID Request Message .........................................................................................249
Figure 29-1, IPMB Event Request Message Format .................................................................................................371
Figure 29-2, Example SMIC Event Request Message Format ..................................................................................372
Figure 35-1, High-Going and Low-Going Event Assertion/Deassertion Points .......................................................431
Figure 39-1, Sensor to FRU Lookup .........................................................................................................................446
Figure 43-1, 6-bit Packed ASCII Example................................................................................................................502
Figure B-1, Example Event Data Comparison Algorithm.........................................................................................509
Figure D-1, SMBus Write-Block by Master Write-Read through KCS/SMIC .........................................................525
Figure D-2, Master Write-Read Response via KCS/SMIC ......................................................................................525
Figure D-3, Get Message Response via KCS/SMIC ................................................................................................525
Figure D-4, Master Write-Read Request via LAN/PPP ...........................................................................................526
Figure D-5 Master Write-Read Response via LAN/PPP..........................................................................................526
Figure D-6, Master Write-Read Response via LAN/PPP.........................................................................................526
xviii
Intelligent Platform Management Interface Specification
List of Tables
Table 1-1, Glossary .......................................................................................................................................................5
Table 3-1, Required BMC Functions ..........................................................................................................................32
Table 5-1, Network Function Codes ...........................................................................................................................38
Table 5-2, Completion Codes ......................................................................................................................................40
Table 5-3, Sensor Owner ID and Sensor Number Field Definitions ...........................................................................42
Table 5-4, System Software IDs ..................................................................................................................................43
Table 6-1, Channel Number Assignments ...................................................................................................................45
Table 6-2, Channel Protocol Type Numbers ...............................................................................................................45
Table 6-3, Channel Medium Type Numbers ...............................................................................................................46
Table 6-4, Channel Access Modes ..............................................................................................................................47
Table 6-5, Channel Privilege Levels ...........................................................................................................................48
Table 6-6, Session-less , Single-session and Multi-session Characteristics ................................................................54
Table 6-7, Default Session Inactivity Timeout Intervals .............................................................................................58
Table 6-8, Message Bridging Mechanism by Source and Destination ........................................................................62
Table 6-9, IPMI Message and IPMB / Private Bus Transaction Size Requirements...................................................65
Table 7-1, BMC IPMB LUNs .....................................................................................................................................67
Table 8-1, System Interface Request For Delivering Remote IPMB Request via ICMB............................................72
Table 8-2, Send Message Response ............................................................................................................................72
Table 8-3, IPMB Request For Delivering Remote IPMB Request via ICMB.............................................................73
Table 8-4, Send Message Response ............................................................................................................................73
Table 8-5, IPMB Response For Remote IPMB Request Delivered via ICMB............................................................73
Table 9-1, KCS Interface Status Register Bits ............................................................................................................79
Table 9-2, KCS Interface State Bits ............................................................................................................................79
Table 9-3, KCS Interface Control Codes.....................................................................................................................80
Table 9-4, KCS Interface Status Codes .......................................................................................................................80
Table 10-1, SMIC Flags Register Bits.........................................................................................................................92
Table 10-2, SMS Transfer Stream control codes.........................................................................................................97
Table 10-3, SMS Transfer Stream Status Codes .........................................................................................................98
Table 11-1, BT Interface Registers............................................................................................................................105
Table 11-2, BT_CTRL Register Bit Definitions .......................................................................................................106
Table 11-3, BT_INTMASK Register Bit Definitions ...............................................................................................108
Table 11-4, BT Interface Write Transfer ...................................................................................................................109
Table 11-5, BT Interface Read Transfer ....................................................................................................................110
Table 12-1, BMC Single-part Write ..........................................................................................................................113
Table 12-2, BMC Multi-part Write Start ...................................................................................................................114
Table 12-3, BMC Multi-part Write Middle ...............................................................................................................114
Table 12-4, BMC Multi-part Write End ....................................................................................................................114
Table 12-5, BMC Single-part Read ...........................................................................................................................114
Table 12-6, BMC Multi-part Read Start ....................................................................................................................115
Table 12-7, BMC Multi-part Read Middle ................................................................................................................115
Table 12-8, BMC Multi-part Read Retry ..................................................................................................................115
Table 12-9, BMC Multi-part Read End .....................................................................................................................115
Table 12-10, Summary of SMBus Commands for SSIF ...........................................................................................120
Table 12-11, SSIF Commands...................................................................................................................................120
Table 12-12, SSIF Timing Specifications .................................................................................................................120
Table 13-1, RMCP Port Numbers .............................................................................................................................124
Table 13-2, RMCP Message Format .........................................................................................................................125
Table 13-3, Message Type Determination Under RMCP..........................................................................................125
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
UT
UT
UT
TU
TU
TU
TU
UT
xix
Intelligent Platform Management Interface Specification
Table 13-4, ASF/RMCP Messages for IPMI-over-LAN ...........................................................................................126
Table 13-5, RMCP ACK Message Fields..................................................................................................................126
Table 13-6, RMCP Packet Fields for ASF Presence Ping Message (Ping Request) .................................................127
Table 13-7, RMCP Packet Fields for ASF Presence Pong Message (Ping Response) ..............................................128
Table 13-8, RMCP/RMCP+ Packet Format for IPMI via Ethernet ...........................................................................131
Table 13-9, RMCP+ Open Session Request ..............................................................................................................143
Table 13-10, RMCP+ Open Session Response .........................................................................................................144
Table 13-11, RAKP Message 1 .................................................................................................................................146
Table 13-12, RAKP Message 2 .................................................................................................................................147
Table 13-13, RAKP Message 3 .................................................................................................................................148
Table 13-14, RAKP Message 4 .................................................................................................................................149
Table 13-15, RMCP+ and RAKP Message Status Codes .........................................................................................150
Table 13-16, Payload Type Numbers ........................................................................................................................153
Table 13-17, Authentication Algorithm Numbers .....................................................................................................153
Table 13-18, Integrity Algorithm Numbers ...............................................................................................................154
Table 13-19, Confidentiality Algorithm Numbers ....................................................................................................155
Table 13-20, AES-CBC Encrypted Payload Fields ...................................................................................................155
Table 13-21, xRC4 Encrypted Payload Fields ..........................................................................................................157
Table 14-1, Serial Port Switching Triggers ...............................................................................................................167
Table 14-2, Serial Port Sharing Access Characteristics ............................................................................................170
Table 14-3, Auto-Connection Mode Patterns ............................................................................................................174
Table 14-4, Modem String Summary ........................................................................................................................175
Table 14-5, Basic Mode Special Characters ..............................................................................................................178
Table 14-6, BASIC MODE Data Byte Escape Encoding..........................................................................................178
Table 14-7, LCP Code Fields ....................................................................................................................................182
Table 14-8, Overview of PPP Configure-Ack, -Nak, & -Reject Packet Use ............................................................182
Table 14-9, PPP Link Configuration Option Support Requirements ........................................................................183
Table 14-10, Default Escaped Characters .................................................................................................................186
Table 14-11, CBCP Callback Number Options.........................................................................................................190
Table 14-12, Terminal Mode Message Bridge Field .................................................................................................193
Table 14-13, Terminal Mode Text Commands .........................................................................................................195
Table 14-14, Terminal Mode Examples ....................................................................................................................198
Table 14-15, TAP Escaping ......................................................................................................................................202
Table 14-16, TAP Success Codes .............................................................................................................................202
Table 15-1, Mux Settings ..........................................................................................................................................204
Table 15-2, SOL Payload Data Format .....................................................................................................................207
Table 15-3, Remote Console to BMC SOL Packet Handling ...................................................................................210
Table 15-4, Set Serial/Modem Mux Command Operation while SOL Active .........................................................212
Table 16-1, Event Message Reception ......................................................................................................................214
Table 17-1, PEF Action Priorities .............................................................................................................................221
Table 17-2, Event Filter Table Entry .........................................................................................................................222
Table 17-3, Comparison-type Selection according to Compare Field bits ................................................................224
Table 17-4, Alert Policy Table Entry ........................................................................................................................226
Table 17-5, Serial/Modem Alert Destination Priorities .............................................................................................228
Table 17-6, PET Specific Trap Fields .......................................................................................................................231
Table 17-7 - PET Variable Bindings Field ................................................................................................................232
Table 17-8, IPMI PET Multirecord Field Format .....................................................................................................233
Table 20-1, IPM Device ‘Global’ Commands ...........................................................................................................239
Table 20-2, Get Device ID Command .......................................................................................................................240
Table 20-3, Cold Reset Command.............................................................................................................................243
Table 20-4, Warm Reset Command ..........................................................................................................................243
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
xx
Intelligent Platform Management Interface Specification
Table 20-5, Get Self Test Results Command ............................................................................................................244
Table 20-6, Manufacturing Test On ..........................................................................................................................245
Table 20-7, Set ACPI Power State Command ...........................................................................................................246
Table 20-8, Get ACPI Power State Command ..........................................................................................................247
Table 20-9, Get Device GUID Command .................................................................................................................248
Table 20-10, GUID Format .......................................................................................................................................248
Table 21-1, Firmware Firewall Commands ...............................................................................................................250
Table 21-2, Get NetFn Support Command ................................................................................................................251
Table 21-3, Get Command Support Command .........................................................................................................252
Table 21-4, Get Command Sub-function Support Command ...................................................................................253
Table 21-5, Get Configurable Commands Command ...............................................................................................254
Table 21-6, Get Configurable Command Sub-functions Command .........................................................................255
Table 21-7, Set Command Enables Command ..........................................................................................................256
Table 21-8, Get Command Enables Command .........................................................................................................257
Table 21-9, Set Configurable Command Sub-function Enables Command ..............................................................258
Table 21-10, Get Configurable Command Sub-function Enables Command ...........................................................259
Table 22-1, IPMI Messaging Support Commands ....................................................................................................260
Table 22-2, Set BMC Global Enables Command ......................................................................................................261
Table 22-3, Get BMC Global Enables Command .....................................................................................................262
Table 22-4, Clear Message Flags Command .............................................................................................................262
Table 22-5, Get Message Flags Command ................................................................................................................263
Table 22-6, Enable Message Channel Receive Command ........................................................................................263
Table 22-7, Get Message Command .........................................................................................................................265
Table 22-8, Get Message Data Fields ........................................................................................................................266
Table 22-9, Send Message Command .......................................................................................................................267
Table 22-10, Message Data for Send Message Command ........................................................................................268
Table 22-11, Read Event Message Buffer Command ...............................................................................................269
Table 22-12, Get System Interface Capabilities Command.......................................................................................270
Table 22-13, Get BT Interface Capabilities Command .............................................................................................271
Table 22-14, Master Write-Read Command .............................................................................................................272
Table 22-15, Get Channel Authentication Capabilities Command ...........................................................................274
Table 22-16, Get System GUID Command ...............................................................................................................276
Table 22-17, Get Channel Cipher Suites Command..................................................................................................277
Table 22-18, Cipher Suite Record Format.................................................................................................................278
Table 22-19, Cipher Suite IDs ...................................................................................................................................279
Table 22-20, Get Session Challenge Command ........................................................................................................280
Table 22-21, Activate Session Command .................................................................................................................281
Table 22-22, AuthCode Algorithms ..........................................................................................................................283
Table 22-23, Set Session Privilege Level Command ................................................................................................284
Table 22-24, Close Session Command ......................................................................................................................284
Table 22-25, Get Session Info Command .................................................................................................................285
Table 22-26, Get AuthCode Command .....................................................................................................................287
Table 22-27, Set Channel Access Command ............................................................................................................288
Table 22-28, Get Channel Access Command ............................................................................................................290
Table 22-29, Get Channel Info Command ................................................................................................................291
Table 22-30, Set Channel Security Keys Command .................................................................................................292
Table 22-31, Set User Access Command ..................................................................................................................294
Table 22-32, Get User Access Command..................................................................................................................295
Table 22-33, Set User Name Command ....................................................................................................................296
Table 22-34, Get User Name Command ...................................................................................................................296
Table 22-35, Set User Password Command ..............................................................................................................297
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
UT
TU
TU
UT
UT
TU
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
UT
UT
TU
TU
TU
TU
UT
xxi
Intelligent Platform Management Interface Specification
Table 23-1, IPMI LAN Commands ...........................................................................................................................299
Table 23-2, Set LAN Configuration Parameters Command ......................................................................................299
Table 23-3, Get LAN Configuration Parameters Command .....................................................................................300
Table 23-4, LAN Configuration Parameters .............................................................................................................300
Table 23-5, Suspend BMC ARPs Command ............................................................................................................306
Table 23-6, Get IP/UDP/RMCP Statistics Command ...............................................................................................307
Table 24-1, RMCP+ Support and Payload Commands .............................................................................................308
Table 24-2, Activate Payload Command ...................................................................................................................309
Table 24-3, Deactivate Payload Command ...............................................................................................................311
Table 24-4, Payload-specific Encryption Behavior ...................................................................................................311
Table 24-5, Suspend/Resume Payload Encryption Command ..................................................................................312
Table 24-6, Get Payload Activation Status Command ..............................................................................................313
Table 24-7, Get Payload Instance Info Command.....................................................................................................313
Table 24-8, Set User Payload Access Command ......................................................................................................314
Table 24-9, Get User Payload Access Command ......................................................................................................315
Table 24-10, Get Channel Payload Support Command.............................................................................................316
Table 24-11, Get Channel Payload Version Command.............................................................................................317
Table 24-12, Get Channel OEM Payload Info Command .........................................................................................317
Table 25-1, IPMI Serial/Modem Commands ............................................................................................................318
Table 25-2, Set Serial/Modem Configuration Command ..........................................................................................318
Table 25-3, Get Serial/Modem Configuration Command .........................................................................................319
Table 25-4, Serial/Modem Configuration Parameters ...............................................................................................320
Table 25-5, Set Serial/Modem Mux Command .........................................................................................................336
Table 25-6, Get TAP Response Codes Command.....................................................................................................337
Table 25-7, Set PPP UDP Proxy Transmit Data Command ......................................................................................338
Table 25-8, Get PPP UDP Proxy Transmit Data Command .....................................................................................338
Table 25-9, Send PPP UDP Proxy Packet Command ...............................................................................................339
Table 25-10, Get PPP UDP Proxy Receive Data Command .....................................................................................340
Table 25-11, Serial/Modem Connection Active Command ......................................................................................341
Table 25-12, Callback Command ..............................................................................................................................341
Table 25-13, Set User Callback Options Command ..................................................................................................342
Table 25-14, Get User Callback Options Command .................................................................................................343
Table 26-1, SOL Commands .....................................................................................................................................344
Table 26-2, SOL Activating Command .....................................................................................................................344
Table 26-3, Set SOL Configuration Parameters Command ......................................................................................345
Table 26-4, Get SOL Configuration Parameters Command ......................................................................................345
Table 26-5, SOL Configuration Parameters ..............................................................................................................346
Table 27-1, BMC Watchdog Timer Commands ........................................................................................................349
Table 27-2, Reset Watchdog Timer Command .........................................................................................................351
Table 27-3, Set Watchdog Timer Command .............................................................................................................352
Table 27-4, Get Watchdog Timer Command ............................................................................................................353
Table 28-1, Chassis Commands ................................................................................................................................355
Table 28-2, Get Chassis Capabilities Command .......................................................................................................356
Table 28-3, Get Chassis Status Command ................................................................................................................357
Table 28-4, Chassis Control Command .....................................................................................................................358
Table 28-5, Chassis Reset Command ........................................................................................................................358
Table 28-6, Chassis Identify Command ....................................................................................................................359
Table 28-7, Set Front Panel Button Enables Command ............................................................................................359
Table 28-8, Set Chassis Capabilities Command ........................................................................................................360
Table 28-9, Set Power Restore Policy Command......................................................................................................360
Table 28-10, Set Power Cycle Interval Command ....................................................................................................361
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
xxii
Intelligent Platform Management Interface Specification
Table 28-11, Get System Restart Cause Command ...................................................................................................362
Table 28-12, Set System Boot Options Command ....................................................................................................363
Table 28-13, Get System Boot Options Command ...................................................................................................363
Table 28-14, Boot Option Parameters .......................................................................................................................364
Table 28-15, Get POH Counter Command................................................................................................................368
Table 29-1, Event Commands ...................................................................................................................................369
Table 29-2, Set Event Receiver .................................................................................................................................369
Table 29-3, Get Event Receiver Command ...............................................................................................................370
Table 29-4, Platform Event (Event Message) Command ..........................................................................................370
Table 29-5, Event Request Message Fields ...............................................................................................................371
Table 29-6, Event Request Message Event Data Field Contents ..............................................................................373
Table 30-1, PEF and Alerting Commands.................................................................................................................375
Table 30-2, Get PEF Capabilities Command.............................................................................................................375
Table 30-3, Arm PEF Postpone Timer Command .....................................................................................................376
Table 30-4, Set PEF Configuration Parameters Command .......................................................................................376
Table 30-5, Get PEF Configuration Parameters Command ......................................................................................377
Table 30-6, PEF Configuration Parameters ...............................................................................................................377
Table 30-7, Set Last Processed Event ID Command .................................................................................................382
Table 30-8, Get Last Processed Event ID Command ................................................................................................382
Table 30-9, Alert Immediate Command ....................................................................................................................383
Table 30-10, PET Acknowledge Command ..............................................................................................................384
Table 31-1, SEL Device Commands .........................................................................................................................385
Table 31-2, Get SEL Info Command.........................................................................................................................386
Table 31-3, Get SEL Allocation Info Command .......................................................................................................387
Table 31-4, Reserve SEL Command .........................................................................................................................388
Table 31-5, Get SEL Entry ........................................................................................................................................389
Table 31-6, Add SEL Entry .......................................................................................................................................390
Table 31-7, Partial Add SEL Entry Command ..........................................................................................................391
Table 31-8, Delete SEL Entry ...................................................................................................................................391
Table 31-9, Clear SEL ...............................................................................................................................................392
Table 31-10, Get SEL Time Command .....................................................................................................................392
Table 31-11, Set SEL Time Command......................................................................................................................392
Table 31-12, Get Auxiliary Log Status Command ....................................................................................................393
Table 31-13, Set Auxiliary Log Status Command.....................................................................................................394
Table 32-1, SEL Event Records ................................................................................................................................395
Table 32-2, OEM SEL Record (Type C0h-DFh) ......................................................................................................396
Table 32-3, OEM SEL Record (Type E0h-FFh) .......................................................................................................396
Table 33-1, Mandatory SDR Update Mode Commands............................................................................................398
Table 33-2, SDR Repository Device Commands ......................................................................................................402
Table 33-3, Get SDR Repository Info Command .....................................................................................................404
Table 33-4, Get SDR Repository Allocation Info Command ....................................................................................405
Table 33-5, Reserve SDR Repository Command ......................................................................................................406
Table 33-6, Get SDR Command................................................................................................................................407
Table 33-7, Add SDR Command ..............................................................................................................................408
Table 33-8, Partial Add SDR Command ...................................................................................................................408
Table 33-9, Delete SDR Command ...........................................................................................................................409
Table 33-10, Clear SDR Repository Command ........................................................................................................409
Table 33-11, Get SDR Repository Time Command ..................................................................................................410
Table 33-12, Set SDR Repository Time Command ..................................................................................................410
Table 33-13, Enter SDR Repository Update Mode Command .................................................................................410
Table 33-14, Exit SDR Repository Update Mode Command ...................................................................................411
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
UT
TU
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
UT
UT
TU
TU
TU
UT
UT
UT
TU
TU
TU
UT
UT
TU
TU
UT
UT
TU
TU
UT
xxiii
Intelligent Platform Management Interface Specification
Table 33-15, Run Initialization Agent .......................................................................................................................411
Table 34-1, FRU Inventory Device Commands ........................................................................................................412
Table 34-2, Get FRU Inventory Area Info Command ...............................................................................................412
Table 34-3, Read FRU Data Command .....................................................................................................................413
Table 34-4, Write FRU Data Command ....................................................................................................................414
Table 35-1, Sensor Device Commands .....................................................................................................................416
Table 35-2, Get Device SDR Info Command ............................................................................................................417
Table 35-3, Get Device SDR Command ...................................................................................................................420
Table 35-4, Reserve Device SDR Repository ...........................................................................................................420
Table 35-5, Get Sensor Reading Factors Command .................................................................................................421
Table 35-6, Set Sensor Hysteresis .............................................................................................................................422
Table 35-7, Get Sensor Hysteresis ............................................................................................................................422
Table 35-8, Set Sensor Thresholds ............................................................................................................................422
Table 35-9, Get Sensor Thresholds ...........................................................................................................................423
Table 35-10, Set Sensor Event Enable ......................................................................................................................424
Table 35-11, Get Sensor Event Enable......................................................................................................................426
Table 35-12, Re-arm Sensor Events ..........................................................................................................................428
Table 35-13, Get Sensor Event Status Response Overview ......................................................................................430
Table 35-14, Get Sensor Event Status Command .....................................................................................................431
Table 35-15, Get Sensor Reading Command ............................................................................................................434
Table 35-16, Set Sensor Type Command ..................................................................................................................435
Table 35-17, Get Sensor Type ...................................................................................................................................436
Table 38-1, FRU Device Locator Field Usage ..........................................................................................................444
Table 39-1, System and Device-Relative Entity Instance Values .............................................................................445
Table 42-1, Event/Reading Type Code Ranges.........................................................................................................455
Table 42-2, Generic Event/Reading Type Codes ......................................................................................................455
Table 42-3, Sensor Type Codes.................................................................................................................................458
Table 43-1, Full Sensor Record - SDR Type 01h......................................................................................................471
Table 43-2, Compact Sensor Record - SDR Type 02h ..............................................................................................478
Table 43-3, Event-Only Sensor Record - SDR Type 03h .........................................................................................484
Table 43-4, Entity Association Record - SDR Type 08h ..........................................................................................487
Table 43-5, Device-relative Entity Association Record - SDR Type 09h .................................................................488
Table 43-6, Generic Device Locator Record - SDR Type 10h..................................................................................490
Table 43-7, FRU Device Locator Record - SDR Type 11h ......................................................................................491
Table 43-8, Management Controller Device Locator - SDR 12h ..............................................................................493
Table 43-9, Management Controller Confirmation Record - SDR Type 13h ...........................................................495
Table 43-10, BMC Message Channel Info Record - SDR Type 14h ........................................................................496
Table 43-11, OEM Record - SDR Type C0h ............................................................................................................498
Table 43-12, IPMB/I2C Device Type Codes .............................................................................................................499
Table 43-13, Entity ID Codes ....................................................................................................................................500
Table 43-14, 6-bit ASCII definition ..........................................................................................................................502
Table 43-15, Sensor Unit Type Codes.......................................................................................................................503
Table 44-1, Example discrete Processor sensor with Sensor-specific states & event generation .............................505
Table 44-2, Example discrete Processor sensor with Generic states & event generation .........................................506
Table C1-44-3, SM BIOS IPMI Device Information Record....................................................................................511
Table C1-44-4, Interface Type field values ...............................................................................................................512
Table C1-44-5, Byte-aligned I/O Mapped Register Address examples ....................................................................513
Table C1-44-6, 32-bit aligned I/O Mapped Register Address examples ...................................................................513
Table C2-44-7, PCI Class Codes for IPMI ................................................................................................................515
Table C3-44-8, Service Processor Management Interface Description Table Format ..............................................517
Table C3-44-9, IPMI Device Object Control Methods .............................................................................................520
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UP
TU
UP
UT
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
TU
UT
xxiv
Intelligent Platform Management Interface Specification
Table 44-10, TAP Flow Summary ............................................................................................................................531
Table 44-11, Command Number Assignments and Privilege Levels ........................................................................536
Table H-44-12, Sub-function Number Assignments .................................................................................................541
TU
UT
UT
TU
TU
UT
xxv
Intelligent Platform Management Interface Specification
xxvi
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.
P
P
The implementation of certain aspects of IPMI may require access to other specifications and documents that are not
part Refer to the Reference 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.
1.2 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/
HTU
UTH
1
Intelligent Platform Management Interface Specification
[ACPI 2.0]
Advanced Configuration and Power Interface Specification, Revision 2.0c, August 25, 2003.
©1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003 Compaq Computer Corporation, Intel
Corporation, Microsoft Corporation, Phoenix Technologies Ltd., Toshiba Corporation.
http://www.acpi.info
HTU
[AES]
UTH
Advanced Encryption Standard, FIPS 197, November 2001.
http://csrc.nist.gov/publications/fips/fips197/fips-197.pdf
HTU
[ADDR]
UTH
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
P
HTU
[ASF]
Alert Standard Format v1.0 Specification, ©2001, Distributed Management Task Force.
http://www.dmtf.org
HTU
[ASF 2.0]
P
UTH
UTH
Alert Standard Format (ASF) Specification Version 2.0, 23 April 2003, ©2000-2003, Distributed
Management Task Force, Inc. http://www.dmtf.org
HTU
UTH
[BR1]
Entity Authentication and Key Distribution, Bellare and Rogaway, 1993.
[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
HTU
[FRU]
UTH
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
HTU
[I2C]
P
P
UTH
The I2C Bus And How To Use It, ©1995, Philips Semiconductors. This document provides the
timing and electrical specifications for I2C busses.
P
P
P
[ICMB]
P
Intelligent Chassis Management Bus Bridge Specification v1.0, rev. 1.3, © 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
HTU
UTH
[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
[MODES]
Recommendation for Block Cipher Modes of Operation: Methods and Techniques, NIST Special
Publication 800-38A, December 2001.
http://csrc.nist.gov/publications/nistpubs/800-38a/sp800-38a.pdf
HTU
UTH
[MSFT EMS] Building Hardware and Firmware to Complement Microsoft Windows Headless Operation, Version
1.00, July 16, 2002. http://www.microsoft.com/whdc/hwdev/platform/server/headless/
IPMB. http://developer.intel.com/design/servers/ipmi
HTU
HTU
UTH
UTH
[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.
2
Intelligent Platform Management Interface Specification
[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.
[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.
[RFC2104]
RFC 2104, HMAC: Keyed-Hashing for Message Authentication, H. Krawczyk, IBM, M. Bellare,
UCSD, R. Canetti, IBM, February 1997.
http://www.ietf.org/rfc/rfc2104.txt
HTU
UTH
[RFC2153]
RFC 2153, PPP Vendor Extensions, Simpson, W., Daydreamer, May 1997.
[RFC2404]
RFC 2404, The Use of HMAC-SHA-1-96 within ESP and AH, C. Madson, Cisco Systems Inc., R.
Glenn, NIST, November 1998.
http://www.ietf.org/rfc/rfc2404.txt
HTU
UTH
[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
[SHA-1]
NIST, FIPS PUB 180-2: Secure Hash Standard, August 2002.
http://csrc.nist.gov/publications/fips/fips180-2/fips180-2.pdf
HTU
UTH
[SMBIOS]
System Management BIOS Specification, Version 2.3.1, © 1997, 1999 American Megatrends Inc.,
Award Software International, Compaq Computer Corporation, Dell Computer Corporation,
Hewlett-Packard 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
HTU
UTH
HU
UH
3
Intelligent Platform Management Interface Specification
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.
4
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
P
P
ICMB
IERR
IPM
IPMB
P
ISA
KB
LUN
P
P
P
5
Intelligent Platform Management Interface Specification
NMI
MD2
MD5
Payload
PEF
PERR
PET
POST
Re-arm
SDR
SEL
SERR
SMI
SMIC
SMM
SMS
Soft Reset
Word
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.
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.
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.
–
6
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
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:
•
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
7
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 (Baseboard Management Controller). 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.
8
Intelligent Platform Management Interface Specification
1.6 New for IPMI v2.0
IPMI v1.5 is an extension of the v1.5 specification. IPMI v2.0 adds new capabilities and incorporates learnings,
feedback, and features gathered from industry review and experiences deploying IPMI v1.5 enabled systems. The
following summarizes the most significant new features for IPMI v2.0:
Enhanced Authentication
Extensions to the protocols for IPMI over IP, collectively referred to as “RMCP+”, support new
algorithms that provide more robust key exchange process for establishing sessions and authenticating
users. These steps more closely align with those used for the DMTF ASF 2.0 specification (see
[ASF2.0]), making it simpler to create applications that can connect to both ASF and IPMI-based
system.
VLAN Support
Configuration options have been added to support IEEE 802.1q VLAN (virtual LAN) headers for
IPMI over IP sessions on IEEE 802.3 Ethernet. VLAN works with VLAN-aware routers and switches
to allow a physical network to be partitioned into ‘virtual’ networks where a group of devices on
different physical LAN segments which can communicate with each other as if they were all on the
same physical LAN segment. This can be used to isolate classes of network membership at the
Ethernet Packet level rather than at the IP level, as might be done with a router. This can be used to set
up a ‘management VLAN’ where only devices that are members of that VLAN will receive packets
related to management, and, conversely, will be isolated from the need to process network traffic for
other VLANs.
Serial Over LAN (SOL)
Serial Over LAN provides a mechanism that enables the serial controller of a managed system to be
redirected over an IPMI session over IP. This enables remote console applications to provide access to
text-based interfaces for BIOS, utilities, operating systems, and applications while simultaneously
providing access to IPMI platform management functions. SOL is implemented as a payload type
under the new payload capability in RMCP+.
Payloads
RMCP+ adds the ability to enable IPMI over IP sessions to other types of traffic in addition to IPMI
messages. This includes both standard payload types defined in the IPMI specification (such as SOL),
and OEM ‘value-added’ payload types.
Encryption Support
IPMI messages and other payloads carried over RMCP+ can be encrypted. This enables confidential
remote configuration of parameters such as user passwords and transfer of sensitive payload data over
SOL.
Extended User Login Options
New options support “Role Only” logins for simple environments where it is desirable to just enable
logins according to a given privilege level, without the need to assign or configure usernames. Support
for “two key” logins enables a BMC to be configured for a very robust environment, where both a
user-specific and BMC-specific key are required to connect to a given BMC.
Firmware Firewall
Firmware Firewall is the name for a collection of commands that enable a BMC implementation to
restrict the ability to execute certain commands or functions from a given interface. This can be used
to protect against operations that errant or malicious software may use to affect the managed system or
9
Intelligent Platform Management Interface Specification
other systems. For example, this enables a BMC to block the ability for local software to send a
Chassis Control command to reset another blade in a modular server implementation where BMCs on
individual blades share a common management bus across the blade backplane. Firmware Firewall
includes a set of commands that enable software to discover which commands and functions are
present and enabled on a given management controller. These commands can be used by themselves to
provide a more efficient way for software and conformance tests to discover which features are
available.
SMBus System Interface (SSIF)
The SMBus System Interface (SSIF) is a new, low pin-count, option for the hardware interface that
provides local access to the BMC via a connection to the system’s SMBus host controller. SSIF helps
support lower-cost BMC implementations by enabling an interface that can be used on low-cost
microcontrollers in low pin-count packages.
10
Intelligent Platform Management Interface Specification
1.7 IPMI Overview
This section presents an overview of IPMI and its main elements and characteristics.
1.7.1
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.
1.7.2
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 1-1, below.
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
11
Intelligent Platform Management Interface Specification
1.7.3
Management Controllers and the IPMB
Figure 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.
P
P
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
e.g. Voltages, Temperatures,
Fans, Power & Reset control, etc.
Serial
Controller
Motherboard
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
P
12
P
P
P
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.
1.7.4
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 Diagram,
The same core of IPMI messages is available over every IPMI-specified 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.
P
1.7.5
P
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.
1.7.6
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
13
Intelligent Platform Management Interface Specification
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.
1.7.7
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.
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.
14
Intelligent Platform Management Interface Specification
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).
1.7.8
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
33.6, Sensor Initialization Agent, for details on the initialization agent process.
1.7.9
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.
1.7.10 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.
P
P
P
P
P
P
1.7.11 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.
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.
15
Intelligent Platform Management Interface Specification
1.7.12 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).
P
P
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.)
P
P
P
P
P
P
1.7.13 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.
1.7.14 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.
1.7.15 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.
16
Intelligent Platform Management Interface Specification
1.7.16 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:
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.
1.7.17 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.
1.7.18 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.
17
Intelligent Platform Management Interface Specification
•
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
1.7.19 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.
IPMI v2.0 defines an extended packet format and capabilities that are collectively referred to as “RMCP+”.
RMCP+ is actually defined under the IPMI-specific portion of an RMCP packet. RMCP+ utilizes
authentication algorithms that are more closely aligned with the mechanisms used for the ASF 2.0 specification.
In addition, RMCP+ adds data confidentiality (encryption) and a ‘payloads’ capability.
1.7.19.1
Payloads
“Payloads” are a capability specified for RMCP+ that enable an IPMI session to carry types of traffic that are
in addition to IPMI Messages. Payloads can be ‘standard’ (defined in the IPMI specifications) or ‘OEM’
(specified by an OEM or other organization). Standard payload types include IPMI Messages, messages for
session setup under RMCP+, and the payload for the “Serial Over LAN” capability introduced in IPMI v2.0.
A BMC implementation can allow a payload to be activated (launched) the same IPMI session that a remote
user connected to the BMC over, or the BMC can require that the remote console establish a separate session
for the payload. This enables an implementation to off-load the payload processing to another device, if
desired.
1.7.20 Serial Over LAN (SOL)
Serial Over LAN (SOL) is the name for the redirection of baseboard serial controller traffic over an IPMI
session. This can be used to enable asynchronous serial-based OS and pre-OS communication over a
connection to the BMC. SOL is specified in Section 15, Serial Over LAN. SOL provides can be used to provide
a user at a remote console a means of interacting with serial text-based interfaces such as operating system
command-line interfaces, serial redirected BIOS interfaces, and serial text-based applications over and IPMI
LAN session. A single remote console application can use SOL to simultaneously provide LAN access to IPMI
platform management and serial text redirection under a unified user interface. SOL is implemented as a
payload type under the IPMI v2.0 “RMCP+” protocol. Access privileges for SOL are managed under the same
user configuration interfaces that are used for IPMI management. This simplifies the creation of configuration
software, remote management applications, and cross-platform configuration utilities.
18
Intelligent Platform Management Interface Specification
1.7.21 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,
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).
1.7.22 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.
1.7.23 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.
19
Intelligent Platform Management Interface Specification
•
PPP Alert: The BMC connects to a remote LAN via PPP and delivers a PET trap to a specified IP address.
1.7.24 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.
1.7.25 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.
1.7.26 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.
20
Intelligent Platform Management Interface Specification
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.
1.7.27 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.
1.7.28 Standardized POH Counter
This is an optional counter to return a counter value proportional to the system operating (S0) power-on hours.
1.7.29 Firmware Firewall
“Firmware Firewall” is the name for a capability that is primarily provided to enable a BMC implementation to
block certain configuration, messaging, and write operations from being done from the System Interface. This is
done primarily to provide a way prevent local software from unintentionally or maliciously setting values or
performing actions that could affect multiple nodes in a modular “blade” chassis, but can be used in BMC
implementations in ‘standalone’ systems as well. Firmware Firewall provides mechanisms that enable the BMC to
block IPMI commands and functions from being accessed from a given interface, and a set of “command
discovery” commands that let software discovery and configure which commands and functions are available.
1.7.30 Command and Function Discovery
The ‘command discovery’ commands that support Firmware Firewall can be implemented separately as a way of
enabling software to more efficiently discover command support, and as a way to assist automated ‘conformance
testing’ of IPMI implementations. Without the command discovery commands, IPMI utilizes several mechanisms
that software can use to determine what commands and functions are supported on a given management
controller. Some commands are simply mandatory, other commands and IPMI functions are discoverable via the
Sensor Data Records, ‘capabilities’ commands, or bit fields in responses. Remaining functions are discovered by
a ‘test for support’ approach - where software trys issuing the command to see if it is implemented or not. Support
for some functions is also implied by whether or not other commands are present because they’re part of a set.
For example, if a “Set Configuration Parameters” command is supported, then it can be inferred that the
corresponding “Get Configuration Parameters” will also be supported. The command discovery commands,
however, enable a BMC to optionally provide a ‘centralized’ way of reporting command and function support,
rather than the ‘distributed’ and ‘test based’ mechanism that is the default.
21
Intelligent Platform Management Interface Specification
1.7.31 IPMI Hardware Components
IPMI provides very few specifications for 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.
1.7.32 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.
1.8 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
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
22
Intelligent Platform Management Interface Specification
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.
1.9 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:
•
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.
1.10 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
23
Intelligent Platform Management Interface Specification
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.
24
Intelligent Platform Management Interface Specification
1.11 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 accommodate 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.
•
The read/write FRU commands are the same.
P
P
P
P
25
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.
P
P
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.
26
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.
27
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
Privately
Managed
Non-Volatile
Storage
Baseboard
Serial Controller
BASEBOARD MANAGEMENT
CONTROLLER
SYSTEM MESSAGE
INTERFACE
SYSTEM - BMC INTERFACE
SYSTEM MESSAGE
INTERFACE
SYSTEM SOFTWARE
MESSAGE HANDLER
28
BIOS
SYSTEM
MANAGEMENT S/W
SMI HANDLER
OS
Serial
Serial Port Sharing
Logic
SERIAL
INTERFACE
MESSAGE
HANDLER
LAN
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
29
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.
30
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.
31
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
32
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.
33
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
34
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.
35
Intelligent Platform Management Interface Specification
36
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 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.
37
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
38
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])
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 - PCI Industrial Computer Manufacturer’s Group.
(www.picmg.com)
01h DMTF Pre-OS Working Group ASF Specification
(www.dmtf.org)
all other Reserved
HTU
HTU
UTH
2Eh-2Fh
30h-3Fh
*
**
OEM/Group
Controllerspecific
OEM/Group
OEM/NonIPMI group
Requests and
Response
-
UTH
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
39
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
00h
C0h
C1h
C2h
C3h
C4h
C5h
C6h
C7h
C8h
C9h
CAh
CBh
CCh
CDh
CEh
CFh
D0h
D1h
D2h
D3h
D4h
D5h
D6h
FFh
40
Definition
GENERIC COMPLETION CODES 00h, C0h-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 due to insufficient privilege level or other securitybased restriction (e.g. disabled for ‘firmware firewall’).
Cannot execute command. Command, or request parameter(s), not supported in
present state.
Cannot execute command. Parameter is illegal because command sub-function
has been disabled or is unavailable (e.g. disabled for ‘firmware firewall’).
Unspecified error.
DEVICE-SPECIFIC (OEM) CODES 01h-7Eh
01h-7Eh
Device specific (OEM) completion codes. This range is used for commandspecific 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
Standard command-specific codes. This range is reserved for commandspecific completion codes for commands specified in this document.
Intelligent Platform Management Interface Specification
Code
all other
Definition
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.
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.
New for IPMI v2.0 Controllers and software that handle IPMI commands: The D4h completion code has been
extended to indicate that an insufficient privilege level or command restriction due to Firmware Firewall was the
reason a command could not be accessed. Similarly, a D6h completion code has been added to indicate that a
particular sub-function could not accessed due to Firmware Firewall.
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.
•
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.
41
Intelligent Platform Management Interface Specification
•
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.
•
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.
P
P
P
P
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
42
Intelligent Platform Management Interface Specification
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.
Table 5-4, System Software IDs
1
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
P
P
Resultant 8-bit value
1
P
P
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.
43
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.
44
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 Type number is now used to explicitly indicate the class of the media. Both these values are
used in the Get 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
Description
reserved
Used for IPMB, serial/modem Basic Mode, and LAN
ICMB v1.0 - See Section 8, ICMB Interface.
reserved
IPMI on SMBus 1.x - 2.x [1]
Request = [rsSA, Netfn(even)/rsLUN, 00h, rqSA, rqSeq/rqLUN, CMD[2], <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)
Interface
SMIC System Interface Format - See Section 10, SMIC Interface
BT System Interface Format, IPMI v1.0 - See Section 11, Block Transfer (BT)
Interface
P
P
P
05h
KCS
06h
07h
SMIC
BT-10
P
45
Intelligent Platform Management Interface Specification
08h
BT-15
09h
1Ch-1Fh
all other
TMode
n/a
1.
2.
BT System Interface Format, IPMI v1.5 - See Section 11, Block Transfer (BT)
Interface
Terminal Mode - See Section 14.7, Terminal 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
46
Description
reserved
IPMB (I2C)
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
P
P
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 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 Access command. The Set 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 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 14-2, 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 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.
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
47
Intelligent Platform Management Interface Specification
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 14-2, 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 Levels, lists the
currently defined privilege levels. The Set Channel Access command is used to set the maximum privilege level
limit for a channel. The Set 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
48
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
level.
Intelligent Platform Management Interface Specification
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.
49
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.
50
Intelligent Platform Management Interface Specification
6.11 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 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 PCI.
•
If the SPMI Table is not present, the driver should look for the SMBIOS Type 38 table (See
51
Intelligent Platform Management Interface Specification
Appendix 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.
6.12 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.
An IPMI messaging connection to the BMC fits one of three classifications, session-less, single-session, or multisession.
6.12.1 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.
6.12.2 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.
6.12.3 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
52
Intelligent Platform Management Interface Specification
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.
6.12.4 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
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.
6.12.5 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.
53
Intelligent Platform Management Interface Specification
6.12.6 Summary of Connection Characteristics
The following table summarizes the key characteristics that differentiate session-less, single-session, and multisession connections:
P
54
P
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
Intelligent Platform Management Interface Specification
6.12.7 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:
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
55
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 6.12.4, Per-Message and User
Level Authentication Disables.
6.12.8 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.
6.12.9 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.
6.12.10 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
56
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).
6.12.11 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 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.
6.12.12 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.
6.12.13 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.
57
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
6.12.13.1
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.)
6.12.14 Additional Session Specifications and Characteristics
58
•
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 6.13, BMC 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.
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.
•
An Administrator can force sessions on any channel to be terminated.
6.13 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.
59
Intelligent Platform Management Interface Specification
6.13.1 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.
6.13.2 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:
60
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
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.
6.13.3 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.
61
Intelligent Platform Management Interface Specification
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
Request from LAN to Serial
Response from Serial to LAN
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
Send Message
BMC LUN 00b
yes
yes
yes
yes
6.13.4 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.
62
Intelligent Platform Management Interface Specification
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.
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
Channel Channel ID
Address / Channel Responder's Channel Channel
RqSeq# Number Handle
Address
SWID
Number
RsLUN
RqA
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
NetFn/
RsLUN
=S/E,
00b
h
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
63
Intelligent Platform Management Interface Specification
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.
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.
U
U
6.14 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
64
Intelligent Platform Management Interface Specification
Appendix 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.)
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.
P
Private Bus Output
P
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.
An implementation will typically implement a private bus using an
actual I2C 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
P
P
P
LAN/PPP Input
P
65
Intelligent Platform Management Interface Specification
LAN/PPP Output
66
Required:
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.
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
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.
P
P
P
P
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
67
Intelligent Platform Management Interface Specification
system interface (see Table 29-2, 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.
68
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.
69
Intelligent Platform Management Interface Specification
70
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
71
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.
72
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
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
73
Intelligent Platform Management Interface Specification
74
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:
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.
75
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
76
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
77
Intelligent Platform Management Interface Specification
Appendix 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.
78
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
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
condition exists[2]. 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
P
1
IBF
0
OBF
1.
2.
[1]
Description
P
P
R/O
R/O
R/O
R/O
R/O
P
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
Interrupts, 9.13, Physical Interrupt Line Sharing, and 9.14, Additional
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/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
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.
79
Intelligent Platform Management Interface Specification
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
Interrupts, 9.13, Physical Interrupt Line Sharing, and 9.14, Additional 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
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
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
80
Intelligent Platform Management Interface Specification
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.
81
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-
82
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.
83
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?
84
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.
85
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
86
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.
87
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.
88
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.
89
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.
90
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.
91
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 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
92
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 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.
93
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.
94
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.
95
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.
96
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
97
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
98
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.
99
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)
100
00
Command
(02h = Platform Event)
Completion Code
Event Data 3
Intelligent Platform Management Interface Specification
101
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.
U
U
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.
T
T
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:
102
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
103
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
104
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.
105
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
106
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.
U
U
U
U
U
U
U
U
U
U
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.
U
U
U
U
* R=read; W=write; S=set; C=clear
107
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.
108
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
109
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).
110
Intelligent Platform Management Interface Specification
111
Intelligent Platform Management Interface Specification
12. SMBus System Interface (SSIF)
The SMBus System Interface (SSIF) defines a SMBus-based system interface to the BMC. Unlike the other system
interface definitions (e.g. KCS), SSIF does not specify a set of registers that is I/O or memory mapped into the host
processor’s space. SSIF assumes the existence of a SMBus host controller in the system. The host-side register
interface for SMBus host controllers is not standardized. Therefore, in order for system software to utilize this
interface, a host controller-specific driver for the given operating system is required.
SSIF encapsulates IPMI messages and transfers them between the host controller and BMC using the SMBus
“Write Block” and “Read Block” protocols. With SSIF, the BMC is always accessed as a slave device on SMBus.
The host controller masters the to write data to the BMC. When the BMC has data for the host, it asserts the
SMBAlert to the host controller to signal that data is available. Software then directs the host controller to master
the bus and perform a SMBus Read Block transaction to ‘pull’ the data from the BMC.
Thus, the SMBus System Interface requires that that host controller support the SMBus “SMBAlert” signal. This
signal is used as an interrupt to the host controller that indicates that the BMC has data available that is ready to be
retrieved. SSIF also allows the BMC to be polled for data.
A standard SMBus transaction is limited to transferring 32 data bytes. Some IPMI messages can take more than 32
bytes. Therefore, the SSIF definition includes optional support for using more than one SMBus transaction to move
data to and from the BMC in order to support IPMI messages that are longer than 32 bytes.
The SSIF can optionally use the SMBus PEC (Packet Error Check) for data integrity. This is an 8-bit CRC on the
SMBus transaction data. It is highly recommended that PEC be used in implementations where there may be
electrical noise or where there may be other masters on the bus besides the SMBus host controller. PEC should be
considered mandatory in any implementation where there could be devices that are “hot-plugged” or removed from
the bus during SMBus transactions with the BMC.
The SSIF uses two types of transactions for read and write operations, “single-part” and “multi-part”. Single part
transactions are used when the entire IPMI message content can fit within the 32-byte maximum data portion of and
SMBus Write- or Read-Block protocol transfer. Multi-part transactions are used when more than 32-bytes of IPMI
message data need to be transferred across the system interface.
12.1 Single Threaded Interface
Like the KCS interface, the SSIF Interface is only specified as a ‘Single Threaded Interface’ for standard IPMI
commands. That is, the BMC implementation is not expected to process more than one IPMI request at a time.
While an implementation is allowed to have a degree of ‘command queuing’, for standard IPMI messages the SSIF
lacks a ‘Seq’ field that software can use to match up particular instances of requests with responses.
It is possible that a driver or software that issues a request (writes to the BMC) before the response for a previous
command has been returned could get the response for the earlier command before getting the response to the
present request, or possibly will only get one of the expected responses. Therefore, generic management software or
drivers for SSIF should take care to avoid issuing new requests before prior requests have been completed, and
software should always check fields in the response (e.g. NetFn/LUN and Command) to verify a given response
matches up with a request.
12.2 Single-part Write
The Single-part write is a SMBus transaction that can transfer IPMI messages up to 32 bytes in length. The
following table shows the format of this transaction. The values in parentheses indicate the number of bits for the
particular field when the given field is not 8-bits. Only the address and data portions of the SMBus transactions are
shown. SMBus START and STOP conditions and ACK/NACK bits are left out for simplification. The length field
112
Intelligent Platform Management Interface Specification
provides the count of data bytes of IPMI message content, up to 32 bytes. [PEC] indicates the optional presence of
an SMBus PEC (packet error code) byte. This byte is NOT included in the byte count provided in the length field.
Table 12-1, BMC Single-part Write
Slave Address
(7)
R/W=0
SMBus CMD
(1)
= 02h
IPMI CMD
IPMI Data
(0 or more
bytes)
Length
NetFn
(6)
LUN
(2)
[PEC]
12.3 Multi-part Write
A multi-part write is used when more than 32-bytes of IPMI message data need to be written to the BMC. This
requires two or more SMBus Write-Block transactions, consisting of either a “Start” transaction followed by an
“End” transaction, or a Start transaction, followed by one or more “Middle” transactions, and then an End
transaction.
The first part of the IPMI message is written using the Start transaction. Since Multi-part writes are for the purpose
of transferring IPMI message data that, the Start transaction must always move 32-bytes of data; therefore the value
of the length byte for the Start transaction is always 20h.
The combination of a Start transaction followed by an End transaction can transfer up to 63 bytes of IPMI message.
The Middle transaction is available when there is a need to transfer an IPMI message of greater than 63 bytes. As of
this writing, there are no standard IPMI messages to the BMC that are longer than 63 bytes. Therefore, the ‘middle’
transaction is defined solely as needed by any OEM/group network functions (network function codes 2Ch:3Fh) in
the particular BMC implementation.
There is no specified limit to the number of ‘middle’ transactions that can occur in a transfer. As many ‘middle’
transactions as needed can be used to move the desired amount of data. Note, however, that since the interface is
‘single threaded’ normal IPMI messaging will be unavailable until such transfers have completed. Note, however,
that the maximum message size returned by the Get SSIF Interface Capabilities command is 255 bytes.
It is required that all multi-part write transfers end with an “End” transaction. Middle transactions must move 32bytes of data, therefore the value of the length byte for Middle transactions is always 20h.
The End transaction is used for the last portion of message data that is written to the BMC. It indicates to the BMC
that the message data transfer has completed and the BMC can process the message. The number of message data
bytes in the End transaction can range from 1 to 32 bytes.
Note that the SMBus specification does not allow the length (byte count) in the Write-Block protocol to be zero.
Therefore, it is illegal to have the last Middle transaction in the sequence carry 32-bytes and have a length of ‘0’ in
the End transaction. Software that uses the Middle transaction should take care to correctly handle the cases where
the number of IPMI message bytes is an exact multiple of 32.
12.3.1 Error conditions for Multi-part Writes
It is possible that out-of-order operations may occur in the course of restarting systems or loading and unloading
software. For example, the BMC could have just received a Middle transaction when a system restart cause the next
operation to be a Start transaction.
113
Intelligent Platform Management Interface Specification
•
The BMC shall discard any multi-part write data it has received if a Start transaction is received prior to
receiving a complete End transaction.
•
If the BMC receives an incorrect length (not = 20h) in a Start or Middle transaction, it shall discard any
received data that is received until the next Start transaction is received.
Table 12-2, BMC Multi-part Write Start
Slave Address
(7)
R/W=0 SMBus CMD
(1)
= 06h
IPMI CMD
IPMI Data
Length
NetFn
=20h
(6)
[PEC]
LUN
(2)
Table 12-3, BMC Multi-part Write Middle
Slave Address
(7)
R/W=0
(1)
SMBus CMD
= 07h
Length
=20h
IPMI
Data
[PEC]
Table 12-4, BMC Multi-part Write End
Slave Address
(7)
R/W=0
(1)
SMBus CMD
= 07h
Length
IPMI Data
[PEC]
12.4 Single-part Read Transaction
The following table illustrates the format of a SMBus Read Block protocol for a Single-part Read transaction for
SSIF. This transaction is primarily used for retrieving IPMI response data from the BMC.
Table 12-5, BMC Single-part Read
Slave Address
(7)
R/W = 0
(1)
SMBus CMD
= 03h
IPMI CMD
Slave Address
(7)
Completion
Code*
IPMI Data
R/W=1
(1)
Length
NetFn
(6)
LUN
(2)
[PEC]
* = present in standard IPMI response messages
12.5 Multi-part Read Transactions
The following table illustrates the format of a SMBus Read Block protocols for Multi-part Read transactions for
SSIF. There are four different transactions that can be used: Multi-part Read Start, Multi-part Read Middle, Multipart Read Retry, and Multi-part Read End.
The Read Start and Read End transactions are used together when 33 to 61 bytes of IPMI message data must be read
from the BMC. The Read Start transfers the first 30 bytes of IPMI message data, the Read End transfers the
remaining 3 to 31 bytes. (The Read Start uses a special pattern of 00h,01h as the first two SMBus data bytes in the
Read Block. Since the Read Block can carry a maximum of 32 read data bytes, the Read Start carries 30 bytes of
IPMI message data. The Read End transaction includes a 1-byte constant “FFh” as the first SMBus data byte in the
Read Block. Therefore, the Read Middle transaction can carry up to 31 bytes of IPMI message data. )
If more than 61 bytes must be read, one or more Read Middle transactions are also used.
The Read Middle includes a 1-byte Block Number field as the first byte of SMBus data in the Read Block. Thus,
each Read Middle transaction can carry up to 31 bytes of IPMI message data.
114
Intelligent Platform Management Interface Specification
The Multi-part Read Retry is used to retry ‘middle’ blocks of data in a multi-part read transaction that uses Read
Middle transactions. This is described in more detail below.
The following tables illustrate the SMBus protocol formats for the multi-part read transactions. All multi-part read
transactions follow the SMBus Read Block protocol, with the exception of the Multi-part Read Retry transaction
which uses the Write Block protocol.
Table 12-6, BMC Multi-part Read Start
Slave Address
R/W = 0
(7)
(1)
NetFn
LUN
(6)
(2)
SMBus CMD
Slave Address
1
Length
00h
= 03h
(7)
IPMI CMD Completion
IPMI Data
[PEC]
Code*
01h
Table 12-7, BMC Multi-part Read Middle
Slave Address
(7)
R/W = 0
(1)
SMBus CMD
= 09h
IPMI Data
Slave Address
(7)
[PEC]
1
Length
Block
number
Where: block number is a number that is incremented, starting with 0, for each new block of message data returned
using the Middle transaction. Block Number FFh is reserved for the Read End transaction.
The multi-part read retry transaction is a Write-Block transaction that tells the BMC to return the given middle
Block Number on the next multi-part read Middle transaction. This SMBus CMD is only required when multi-part
read Middle transactions are implemented.
Table 12-8, BMC Multi-part Read Retry
Slave Address
(7)
R/W = 0
SMBus CMD
(1)
= 0Ah
IPMI Data
[PEC]
Length
=1
Block
number
The multi-part read End transaction is a Read-Block transaction that concludes a multi-part read operation.
Table 12-9, BMC Multi-part Read End
Slave Address
(7)
R/W = 0
(1)
SMBus CMD
= 09h
IPMI Data
Slave Address
(7)
[PEC]
1
Length
FFh
12.6 Retention of Output Data
A BMC that implements PEC must retain previous output message data until the occurrence of a valid Write Start
transaction, at which time the output message data is cleared. This behavior is needed to better support PEC. If the
BMC automatically discarded data as it was read out, there might be no way to recover the message data if the PEC
indicated the data was corrupted. However, with this provision, system software can retry the read transaction that
had the error.
The BMC will return the retained data if the Single- and Multi-part Read Start transactions are retried prior to the
next valid write Start transaction. To re-read a given block of Middle data in a multi-part read, the block number
must first be written to the BMC using the Multi-part Read Retry transaction. The next Multi-part Read Middle
transaction will then return that block number, and any subsequent Multi-part Read Middle transactions will
increment the block number and return following blocks.
115
Intelligent Platform Management Interface Specification
12.7 SMBAlert Signal Handling
The SMBAlert signal is automatically cleared by the BMC the first time a Single-part Read Start or multi-part Read
Start is used to read a given set of data. SMBAlert will not be asserted by the BMC again until a new instance of
data or status is available.
A BMC is allowed to implement OEM functions that can assert SMBAlert. However, since such functions could
interfere with the operation of a generic driver for the system interface, they must require being enabled by an OSresident driver or software that is explicitly aware of those features. Furthermore, non-SSIF features that assert
SMBAlert from the BMC must automatically become disabled prior to OS-load if the system is restarted (warm or
cold reset).
A BMC that supports SMBAlert being disabled will start up with SMBAlert disabled by default. A driver will need
to explicitly enable SMBAlert. The BMC will return SMBAlert to the disabled state on system power-up and resets.
12.7.1 Enabling/disabling SSIF SMBAlert
The “Enable Receive Message Queue Full Interrupt” bit in the Set Global Enables command is used to
enable/disable SMBAlert. Note that an implementation is allowed to have SMBAlert always enabled, however it is
highly recommended that enable/disable control be implemented.
12.8 Polling for output data
If SMBAlert is disabled, software can poll for output data by issuing Read Start transactions until data is returned. If
there is no data available, the BMC will NACK the read portion of the SMBus transfer.
12.9 SMBus NACKs and Error Recovery
The BMC can NACK the SMBus host controller if it is not ready to accept a new transaction. This could occur if
write transactions follow too closely together, for example. (See Section 12.17,SSIF Timing) Typically, this will be
exhibited by the BMC NACK’ing its slave address. In some cases the BMC may NACK a SMBus data byte that is
being written to it. This can occur if software attempts to write more data bytes to the BMC than it can handle (for
example, in a multi-part write), or if some internal state change caused the BMC to need to reset its internal
operation.
If the BMC NACKs a single part transaction, software can simply retry it. If a ‘middle’ or ‘end’ transaction is
NACK’d, software should not retry the particular but should restart the multi-part read or write from the beginning
Start transaction for the transfer.
12.10 PEC Handling
[SMBus] allows a slave that receives a PEC to check the PEC and NACK the byte that carried the PEC value if the
PEC is incorrect. Accomplishing this may require special hardware in order to generate the NACK without
significant SMBus clock stretching. In order to avoid this requirement, a BMC implementation is allowed to always
ACK the PEC. A BMC that receives an invalid PEC shall drop the data for the write transaction and any further
transactions (read or write) until the next valid read or write Start transaction is received.
A BMC that implements PEC will automatically start using PEC in read transactions after it receives any SSIF
single-part write or multi-part write Start transaction that includes a valid PEC. The BMC shall cease using PEC in
read transactions after it receives any SSIF single-part write or multi-part write Start transaction that does not
include a PEC byte. (A BMC detects PEC by noting that it has received one more byte in the SMBus Write-Block
transaction than was indicated by the length byte. If this occurs, it assumes that the additional byte was the PEC byte
and then checks it for validity.)
116
Intelligent Platform Management Interface Specification
12.11 SMBus Timeout and Hang Handling
[SMBUS] provides an option for devices to drop off the bus and go back to waiting for the SMBus START
condition if the SMBus clock is held low for greater than 25 milliseconds. There is no requirement for BMCs to
implement this option.
Per [SMBUS] the BMC must synch up to a SMBus START or STOP condition, regardless of which SMBus clock
the condition occurs on. In order for a master to place a START or STOP condition on the bus, there must be no
other party driving the SMBus data line low during the SMBus clock. It is possible on SMBus that a missed clock
or incorrectly terminated transfer could leave a slave device that is being read in the state where it is outputting a ‘0’
data bit on the bus, waiting for the master to continue to clock the bus for the next bit(s). In this condition, the bus is
in a state where START and STOP conditions cannot be generated by the master because the slave is holding the
data line low.
The BMC must allow the bus master to resynch the bus by allowing the master to clock the BMC until the master
can issue a START or STOP condition. This means a that a BMC should ‘drop off the bus’ and let its data and clock
lines go high (un-driven) if it gets clocked for returning more data than it has available.
12.12 Discovering SSIF
The recommended SMBus slave address for the BMC is address 20h (0010_000x binary). The SSIF Interface can
be located at alternative addresses depending on the implementation. An Intel architecture compatible system
implementation that supports SMBIOS must include an SMBIOS “Type 38” record to support system management
software discovery of the existence and slave address of the SSIF (see
117
Intelligent Platform Management Interface Specification
Appendix C1 - Locating IPMI System Interfaces via SM BIOS Tables).
In addition, systems that support ACPI should also provide an SPMI table for the interface (see
118
Intelligent Platform Management Interface Specification
Appendix C3 - Locating IPMI System Interfaces with ACPI)
12.13 SSIF Support Requirements for IPMI v1.5-only BMCs
The SSIF can be used with BMCs that implement just IPMI v1.5 commands. If the BMC uses SSIF and reports
itself as an IPMI v1.5 BMC, then only the single-part write and single-part read transactions are required to be
supported.
Note that since single-part transactions only support IPMI messages up to 32 bytes, it limits the ability to transfer
full-sized IPMI messages between the System Interface and other channels, such as IPMB. This is because
transferring a message to another channel requires the message to be encapsulated in the data portion of a Send
Message, Master Write-Read, or Get Message command. For example, the size of the response message for
retrieving a full-size (32-byte) IPMB message using the Get Message command is 36 bytes. The largest IPMB
message that could be obtained using Get Message with a single-part read would thus be limited be 28 bytes rather
than 32. This can potentially cause problems with accessing satellite controllers on IPMB.
It is highly recommended that multi-part writes and reads are implemented if SSIF is retro-fitted to an IPMI v1.5
implementation, as described for IPMI v2.0 implementations, below.
12.14 SSIF Support Requirements for IPMI v2.0 & Later BMCs
If a BMC reports itself as conformant with IPMI v2.0 (or later), then the BMC must support multi-part writes and
reads for IPMI messages if the BMC supports messaging between system software and other channels (e.g.
implements serial or LAN channels, IPMB or PCI SMBus, etc.). This is because such messaging requires the ability
to transfer IPMI message that are >32 bytes in order to fully support messages to another channel.
The IPMB and PCI SMBus channels also need to support the Master Write-Read command as an alternative
mechanism for delivering IPMI messages to their respective busses, and, in addition, it is recommended that PCI
SMBus supports using the Master Write-Read command for performing full-size SMBus protocol operations.
The following items are required of SSIF implementations on IPMI v2.0 or later BMCs:
•
If the BMC implements any channels other than the system interface, it must implement multi-part writes
and reads to enable accepting a 40 byte IPMI input message size, minimum, and support a 38 byte IPMI
output message size, minimum. See Table 6-9, IPMI Message and IPMB / Private Bus Transaction Size
Requirements and refer to Appendix D, Message Size Requirements for more information.
•
In addition, if the PCI SMBus is supported, the SSIF must support using the Master Write-Read command
to execute all SMBus protocols (with and without PEC) on the target bus, including full-size SMBus
Write-Block, Read-Block, and Block Write-Read Process Call.
12.15 Summary of SMBus Commands Values for SSIF
The following table summarizes the allocation of SMBus commands for SSIF. Note that there are command values
that are reserved for future definition by the IPMI specifications.
119
Intelligent Platform Management Interface Specification
Table 12-10, Summary of SMBus Commands for SSIF
Operation
BMC Single-part Write
BMC Multi-part Write
Start - first part
Middle part(s) if any
End - last part
BMC Single-part Read
BMC Multi-part Read
Start - first part
Middle part(s) if any
End - last part
Retry
Reserved
SMBus
CMD
02h
Protocol
Write Block
06h
07h
08h
03h
Write Block
Write Block
Write Block
Read Block
03h
09h
09h
0Ah
0Bh-17h
OEM
00h,01h
All other
Read Block, first two data bytes after length = [0x01,0x00]
Read Block, first data byte after length = 0x00
Read Block, first data byte after length = 0x01
Write Block, first data byte after length = block number
Reserved. (Reserved CMD values apply to any SMBus
protocol that uses a CMD byte.)
Available for OEM use
12.16 SSIF IPMI Commands
The following sections list the IPMI commands used with the SSIF. See Table 22-1, IPMI Messaging Support
Commands for Optional/Mandatory usage of this command with SSIF.
Table 12-11, SSIF Commands
Section
Defined
Command
Get System Interface Capabilities
22.9
12.17 SSIF Timing
The following table lists the recommended timing specifications on SMBus for a BMC implementing the SSIF.
Note that this timing can be dependent on the performance of the SMBus host controller used in the system.
Specifications are given for a SMBus operating at 100 kbps.
Table 12-12, SSIF Timing Specifications
Internal Timing Specifications
min
max
Overall Message Duration
Time-out waiting for bus free
Time-out waiting for a response, internal
Time between Event Message Requests
Request-to-Response time
T1
T2
T3
T4
T5
60 ms
[1]
60 ms
5 ms
-
Number of Request retries
Time between Request retries
Number of Event Message Request retries
Overall Message Byte Duration
C1
T6
C2
T7
5
60 ms
3
per SMBus
spec
20 ms
[1]
T6max
T6maxT1max[1]
3ms
250 ms
10
3 ms
P
P
P
120
P
[2]
P
P
P
P
This interval is measured from the end of the
request transmission through the end of response
transmission. (SMBus STOP to SMBus STOP)
recommended
recommended
Intelligent Platform Management Interface Specification
SM Bus Clock Low hold
T8
per SMBus
spec
3 ms
recommended
The BMC should avoid stretching the clock more
than 3ms at a time. The BMC must tolerate the
clock being stretched up to the maximum value
specified by the SMBus specification.
Notes:
1.
Unless otherwise specified, this timing applies to the mandatory and optional commands specified in the Intelligent Platform
Management Interface Specification. For controller-specific Application and Firmware commands, the Responder should
attempt to meet this specification. In cases that it cannot, the interface specification for the Responder must clearly specify the
‘Request to Response’ time that was implemented. Because timing can vary according to command and controller,
communication routines should be designed to support response timeouts and retry counts accordingly.
2.
This is a recommended value only. The protocol does not require that non-Event Message requests be retried. The
implementation of retries and the number used is based on the application’s requirements for message delivery.
121
Intelligent Platform Management Interface Specification
13. 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 13-1, and the second
is using a LAN controller on an add-in card, as shown in Figure 13-2.
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 13-1 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 13-1, 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 13-2 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.
122
Intelligent Platform Management Interface Specification
Figure 13-2, PCI Management Bus Implementation
LAN
Controller A
LAN
Controller B
Add-in Card
PCI
SMBus
IPMB
BMC
PCI
System Bus
13.1 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 v1.5 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.
13.1.1 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.
123
Intelligent Platform Management Interface Specification
13.1.2 RMCP Port Numbers
RMCP uses two well-known ports under UDP. The following table describes these ports and summarizes their
use.
Table 13-1, 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.
124
Intelligent Platform Management Interface Specification
13.1.3 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 13-2, 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 13-3, 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
13.2 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.
125
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 13-4, 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 13.2.1, RMCP ACK Messages, and
13.2.2, RMCP 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)
13.2.1 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 13-5, 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
126
Intelligent Platform Management Interface Specification
13.2.2 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.
13.2.3 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 13-6, RMCP Packet Fields for ASF Presence Ping Message (Ping Request)
size in
bytes
Field
UDP Header
Source Port
Destination Port
UDP Length
UDP Checksum
Version
Reserved
RMCP Sequence Number
RMCP Header
2
2
2
2
1
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 13.2.1, RMCP
ACK Messages, and 13.2.2, RMCP 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
P
Class of Message
IANA Enterprise Number
Message Type
Message Tag
ASF Message
Reserved
Data Length
1.
1
4
1
1
1
1
P
Some systems may not generate RMCP ACKs even if requested. Software should be designed to
handle this occurrence.
127
Intelligent Platform Management Interface Specification
13.2.4 RMCP/ASF Pong Message (Ping Response)
This message must be returned from the managed system in response to the Presence Ping message.
Table 13-7, RMCP Packet Fields for ASF Presence Pong Message (Ping Response)
size in
bytes
Field
UDP Header
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
RMCP Header
ASF Message
OEM-defined
2
2
2
2
1
1
1
1
4
1
1
1
1
4
4
Supported Entities
1
Supported Interactions
1
Value
26Fh
from Ping request
per UDP
per UDP
6 = RMCP Version 1.0
00h
FFh for IPMI[2]
06h = ASF
4542 = ASF IANA
40h = Presence Pong
from Ping request
00h
16 (10h)
P
P
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
[7]
Set to 1b if RMCP security
extensions are supported[1]
[6:0] Reserved for future definition by
ASF specification, set to 0000000b
Reserved for future definition by ASF
specification, set to 00 00 00 00 00 00h
P
Reserved
1.
6
P
IPMI v1.5 and IPMI v2.0/RMCP+ do not use RMCP security extensions specified in [ASF 2.0], thus
this bit will typically be 0bs. It’s possible a BMC implementation could also support ASF 2.0
messages, in which case this bit could be set to indicate those extensions are supported for ASF
messages that would utilize them.
13.3 RMCP+
RMCP+ is the name used in this specification for an enhanced protocol for transferring IPMI messages and other
types of payloads to an IPMI-based BMC over IP. RMCP+ uses RMCP overall packet format, but defines
extensions to the fields defined under the IPMI Message Class data that is carried within the RMCP Packet. These
extensions support enhanced authentication, encryption, discovery, and the ability to carry additional types of
traffic (“payloads”) in addition to IPMImessages over an IPMI session, whereas v1.5 was only specified for
128
Intelligent Platform Management Interface Specification
carrying IPMI messages. Since RMCP+ is specified under the IPMI message class in RMCP, RMCP+ packets are
conformant with the RMCP specification at the RMCP packet level.
RMCP+ includes:
•
Support for multiple payload types over an IPMI session. These include both ‘standard’ payloads (such
as the payload for the “Serial Over LAN” capability defined in this specification) and ‘OEM’ payloads.
Payload support enables types of traffic other than IPMI messages to be simultaneously carried over an
IPMI session. The specification also allows a separate session to be established for carrying payloads.
•
Enhanced user authentication algorithms. RMCP+ includes more robust session set-up and key handling
algorithms than those for IPMI over LAN in IPMI v1.5.
•
Encryption support. IPMI messages and other payloads can be encrypted under an IPMI session. This
enables confidentiality for remote operations such as setting user passwords and for SOL.
•
Better alignment with the ASF 2.0. RMCP+ follows many of the packet format and authentication
elements defined for RMCP (Remote Management Control Protocol) as specified in the Distributed
Management Task Force “ASF 2.0” specification. (See [ASF 2.0] in the Reference Documents section.) The
session establishment messages and packet formats vary slightly from their ASF 2.0 counterparts, but are
close enough to make it straightforward to create remote management applications that can support both
ASF 2.0 and IPMI v2.0 -based remote management connections.
•
Supports Encrypted/Unencrypted and Authenticated/Unauthenticated Traffic on Single Connection.
Encryption and Authentication are handled as the “IPMI Message Class” level. This means Authenticated
and Encrypted sessions can be established on any UDP port, including port 26Fh. (This is different from
ASF 2.0 which requires using a different port for authenticated traffic than the port used for unauthenticated
messages.) IPMI allows a BMC to be configured so that authentication and encryption are only utilized
when the payload or privilege level of operation requires it. This eliminates the need to have all traffic
authenticated or encrypted on a connection, when only a small portion of the traffic may require that level
of security. This can provide a a significant performance benefit when using inexpensive microcontrollers
for BMCs.
13.4 BMC Support Requirements for v1.5 and v2.0/RMCP+ Protocols
An IPMI v2.0 conformant BMC (a BMC that reports v2.0 as the IPMI version in the Get Device ID command)
that supports RMCP+ for IPMI messaging and standard payloads is required to simultaneously support IPMI v1.5
Packet formats. For a given BMC, Users must be equally configurable for IPMI v2.0 or v1.5 IPMI Messaging
access.
IPMI v2.0-specific capabilities related to payloads (e.g. Serial Over LAN) are only available over IPMI
v2.0/RMCP+ sessions. Otherwise, unless specified, new IPMI v2.0 commands and command extensions are also
available under IPMI v1.5 sessions, provided the user has appropriate privilege.
13.4.1 Session-less Command Support
IPMI supports certain commands that can be delivered to the BMC without having to first establish a session.
For backward compatibility and interoperability, the BMC shall always accept IPMI v1.5 formatted packets for
messages that are accepted outside of a session. I.e. session-less commands that are common to both IPMI v1.5
and IPMI v2.0/RMCP+ must be accepted in either format. For commands that are only used with IPMI
v2.0/RMCP+, a BMC may elect to only accept those commands in IPMI v2.0/RMCP+ packets. Within a
session, a BMC only accepts packets that are formatted for the type of session (v2.0 or v1.5) that was activated.
129
Intelligent Platform Management Interface Specification
13.5 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 13-3, IPMI LAN Packet Layering
Ethernet Framing
MAC Address
IP/UDP
IP Address, RMCP Port #
RMCP message
Class=IPMI
RMCP Sequence# = FFh
IPMI v1.5 or IPMI v2.0+
Session Wrapper
IPMI Message
NetFn
LUN
Seq#
CMD
Data
13.5.1 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.
130
Intelligent Platform Management Interface Specification
13.6 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.
The table includes a set of columns under the label “Format”. The first column shows the format used for IPMI
over LAN as originally defined in the IPMI v1.5 specification, the second column shows the format for
authenticated packets delivered using RMCP per [ASF 2.0]. This column is shown for reference only. The third
column shows the format of packets for RMCP+.
Table 13-8, RMCP/RMCP+ Packet Format for IPMI via Ethernet
Format
RMCP /
IPMI 1.5
26Fh
Field
802.1q[10]
P
P
IP Header
Destination Address
6
Source Address
6
TPI
2
VLAN TAG - user priority
3-bits
VLAN TAG - CFI
1-bit
VLAN TAG - VLAN ID
Frame Type
Version
Header Length
12-bits
2
4-bits
4-bits
ASF
RMCP
298h
RMCP /
IPMI 2.0
“RMCP+”
Value
26Fh
Dest. MAC Address for
802.3
Source MAC Address for
802.3
Tag Protocol Identifier
=8100h
[11]
User Priority
default = 000b, configurable
Canonical Format Indicator
= 0b
0’s = no VLAN
P
P
0800h
4h for IPv4
5h
(length of IP header in units of 4bytes)
Precedence
Service Type (Type of Service)
3-bits
4-bits
reserved
Total Length
Identification[5]
Flags
1-bit
2
2
3-bits
P
UDP Header
P
Fragment Offset
Time-to-Live
Protocol
Header Checksum
Source IP Address
Destination IP Address
Source Port
Destination Port
UDP Length
13-bits
1
1
2
4
4
2
2
2
[4]
000b
P
P
[4]
1000b
(minimize delay)
0b
P
[5]
note
P
P
[6]
010b
(don’t fragment)
[7]
0_0000_0000_0000b
P
P
P
P
[3]
40h
P
P
11h
(28 bytes)
26Fh, 298h
131
Intelligent Platform Management Interface Specification
UDP Checksum
RSP Header
RMCP Header
IPMI Session
Header
Session ID
Session Sequence #
Version
Reserved
RMCP Seq #
Class of Message
Auth Type / Format
(36 bytes) Some payloads
may not use the UDP
checksum, in which case
this field gets set to 0’s. The
receiver must accept the
packet and ignore the
checksum field in this case.
2
1
1
1
1
1
4
4
1
1
1
1
06h (ASF 2.0)
1
1
1
1
1
Payload Type
1
OEM IANA
4
[2]
FFh for IPMI
P
[7:4] - reserved
[3:0] - Authentication Type /
Format
0h = none
1h = MD2
2h = MD5
3h = reserved
4h = straight password /
key
5h = OEM proprietary
6h = Format = RMCP+
(IPMI v2.0 only)
all other = reserved
Payload Type
[7] - 0b = payload is
unencrypted
1b = payload is
encrypted
[6] - 0b = payload is
unauthenticated (no
AuthCode field)
1b = payload is
authenticated
(AuthCode field is
present)
[5:0] = payload type. See
Table 13-16, Payload
Type Numbers.
This field is only present
when Payload Type = 02h
(OEM Explicit)
byte 1:3 - OEM IANA
byte 4 - reserved
This field is only present
when Payload Type = 02h
(OEM Explicit). The
definition and values of this
field are specified by the
company or body identified
by the OEM IANA field.
[8]
note Session ID is
0000_0000h for messages
that are sent ‘outside’ of a
session.
U
U
U
132
OEM Payload ID
2
IPMI v2.0 RMCP+ Session
ID
4
P
07h for IPMI
P
U
P
Intelligent Platform Management Interface Specification
IPMI Payload
Session Sequence Number
4
IPMI v1.5 Session ID
4
Msg. Auth. Code Code
(AuthCode)
(not present when
Authentication Type set to
‘none’.)
IPMI Msg/Payload length
16
1
4
2
Confidentiality Header
Payload Data
var
var
var
Confidentiality Trailer
IPMI
Session
Trailer[9] /
RSP Trailer
P
2
var
var
Integrity PAD
var
var
P
Pad Length
1
Next Header
1
AuthCode
var
(Integrity Data)
var
[8]
note For IPMI v2.0
“RMCP+” there are separate
sequence numbers tracked
for authenticated and
unauthenticated packets.
0000_0000h is used for
packets that are sent
‘outside’ of a session.
[8]
note Session ID is
0000_0000h for messages
that are sent ‘outside’ of a
session.
P
P
P
P
Payload length in bytes.
1-based.
For encrypted payloads,
based on encryption type for
given payload. The
confidentiality header is not
encrypted.
For IPMI v2.0: IPMI, SOL,
KVM, etc. per Payload Type
field.
For encrypted payloads,
based on encryption type for
given payload. The
confidentiality trailer is
typically encrypted along
with the Payload Data.
Added as needed to cause
the number of bytes in the
data range covered by the
AuthCode (Integrity Data)
field to be a multiple of 4
bytes (DWORD). If present,
each Integrity Pad byte is
set to FFh.
indicates how many pad
bytes were added so that
the amount of non-pad data
can be determined.
Reserved in IPMI v2.0. Set
to Always = 07h for this
specification.
For IPMI v1.5 this field is as
specified by Auth Type.
For IPMI v2.0 (RMCP+) if
this field is present, then it is
calculated according to the
Integrity Algorithm that was
negotiated during the
session open process. See
Table 13-18, Integrity
Algorithm Numbers.
Legacy PAD[1]
P
MAC level
CRC
P
This field is absent when the
packet is unauthenticated.
legacy PAD not needed for
IPMI v2.0
1
4
4
4
133
Intelligent Platform Management Interface Specification
1. 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.
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 (0000_0000h) for the Session ID and/or Session Sequence Number (null Session ID, null Session
Sequence Number) are special values only used for messages and commands that can be executed prior to
establishing a session, e.g. Get System GUID, Get Channel Authentication Capabilities, Get Session Challenge,
and RAKP messages. When the Session ID is 0000_0000h, the Sequence Number field is ignored, however the
Session Sequence Number should still be set to 0000_0000h. In addition, for IPMI v2.0 RMCP+ packets, unless
otherwise indicated bits 7:6 in the payload type field must also indicate that the packet is both unauthenticated and
unencrypted. Note that the IPMI v1.5 Activate Session uses a null (all 0’s) Session Sequence Number before a
session is activated, but does not use a null Session ID. Instead, it uses the Temporary Session ID given by the
BMC in the response to the Get Session Challenge command.
9. For IPMI v2.0 RMCP+ packets, the IPMI Session Trailer is absent whenever the Session ID is 0000_0000h, or
whenever bit 6 in the payload type field indicates the packet is unauthenticated
10. Four bytes only present for IEEE 802.1q “VLAN” formatted packets
11. The use and interpretation of this number is defined in ISO/IEC 15802-3.
13.7 VLAN Support
VLAN support is optional, but recommended, for BMC access via IPMI 1.5 packet and IPMI v2.0 packet
formats. A BMC that supports VLAN on a channel is required to support it for both packet formats. A BMC is
not required to support VLAN equally across all LAN channels. This is allowed because some LAN connections
may not have hardware that supports VLAN with a LAN connection to the BMC.
When a VLAN ID is configured into the LAN parameters, the BMC will only accept packets with that VLAN
tag. This includes all RMCP and RMCP+ packets as well as DHCP and ARP packets. Conversely, all BMCgenerated packets will include the given VLAN tag.
13.8 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.
P
P
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:
134
Intelligent Platform Management Interface Specification
Figure 13-4, 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
cmd
completion code
data
LUN
netFn
rq
rqLUN
rqAddr
rqSeq
rs
rsLUN
rsAddr
Seq
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.
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.
13.9 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.
135
Intelligent Platform Management Interface Specification
13.10 IPMI LAN Configuration
This section provides background information on certain configuration options that are available for LAN
channels and how they’re used.
13.10.1 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.
13.10.2 ‘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.
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.
13.11 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.
136
Intelligent Platform Management Interface Specification
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.
13.11.1 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.
13.11.2 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.
•
Increase ARP Cache expiration intervals in routers and applications.
•
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.
137
Intelligent Platform Management Interface Specification
13.11.3 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 23.3, Suspend BMC ARPs Command for more information.
13.12 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
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.
13.12.1 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.
138
Intelligent Platform Management Interface Specification
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.
13.13 Discovering Support For IPMI over IP Connections
There are two mechanisms for discovering whether a given system (IP Address) supports IPMI v1.5 and/or IPMI
v2.0 connections. The first is the ‘RMCP Ping discovery’ mechanism where the BMC returns whether IPMI is
supported via the supported entities field of the Ping response (a.k.a. the ‘Pong’ message). The BMC can
optionally return values in the Ping response indicating what level of IPMI connection(s) (IPMI v1.5 and/or IPMI
v2.0) it supports.
The second mechanism is an ‘IPMI command discovery’ mechanism where the remote console discovers that the
system supports IPMI by issuing a Get Channel Authentication Capabilities command to determine support for
IPMI v1.5/IPMI v2.0 connections. BMCs that support IPMI v2.0/RMCP+ must support the Get Channel
Authentication Capabilities command in both the IPMI v1.5 and IPMI v2.0 packet formats. It is recommended
that a remote console use the IPMI v1.5 formats until it has confirmed IPMI v2.0 support.
When the remote console decides to connect to the discovered system, it can use the Get Channel Authentication
Capabilities and (for IPMI v2.0/RMCP+) the Get Channel Cipher Suites commands to determine which
authentication, integrity, and confidentiality algorithms can be used for establishing the connection.
13.14 IPMI v1.5 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 6.12.7, Session Activation and IPMI Challenge-Response. Also refer to sections 6.9, Users & Password
Support, and 6.12.3, Multi-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
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.
139
Intelligent Platform Management Interface Specification
5.
140
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.
Intelligent Platform Management Interface Specification
Figure 13-5, LAN Session Startup
Remote Console
Managed System
Discovery
RMCP Ping
RMCP "Pong"
(Ping response)
Get Channel Authentication
Capabilities, Rq
(Console requests information on what
authentication algorithms must be used to
connect at a given maximum privilege level,
e.g. Operator)
SessionID=0, Sess. seq# = 0,
Auth. type =None, AuthCode =
not present [Requested maximum
privilege level = Operator]
SessionID=0, Sess. seq# = 0,
Auth. type =None, AuthCode =
not present [Auth Type(s) = MD2,
MD5]
Get Channel Authentication
Capabilities, Rs
(BMC returns which authentication algorithms can
be used for connecting at requested maximum
privilege level [e.g. MD2 & MD5])
Get Session Challenge, Rq
SessionID=0, Sess. seq# = 0,
Auth. type =None, AuthCode =
not present [username, Auth Type
= MD2]
Activation
(Console requests a challenge for given
user and using MD2 authentication type)
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
(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
13.15 IPMI v2.0/RMCP+ Session Activation
This section describes the process that is used for authenticating the user’s credentials and establishing an IPMI
session using RMCP+. The messages for RMCP+ are specified under the IPMI Message class (07h) for RMCP.
The payload type field in the RMCP+ packet format is used to identify the messages that are used for activating a
session under RMCP+. The following messages are used to activate a session using RMCP+:
141
Intelligent Platform Management Interface Specification
Get Channel Authentication Capabilities request / response
This message exchange provides a way for a remote console to discover what IPMI version is supported.
I.e. whether or not the BMC supports the IPMI v2.0 / RMCP+ packet format. It also provides information
that the remote console can use to determine whether anonymous, “one-key”, or “two-key” logins are used.
This information can guide a remote console in how it presents queries to users for username and password
information. This is a ‘session-less’ command that the BMC accepts in both IPMI v1.5 and v2.0/RMCP+
packet formats.
RMCP+ Open Session Request, RMCP+ Open Session Response
The RMCP+ Open Session request and response messages are used to enable a remote console to discover
what Cipher Suite(s) can be used for establishing a session at a requested maximum privilege level. These
messages are also used for transferring the sessions IDs that the remote console and BMC wish to for the
session once it’s been activated, and to track each party during the exchange of messages used for
establishing the session.
RAKP Message 1, RAKP Message 2
These messages are used to exchange random number and identification information between the BMC
and the remote console that are, in effect, mutual challenges for a challenge/response. (Unlike IPMI v1.5,
the v2.0/RMCP+ challenge/response is symmetric. I.e. the remote console and BMC both issues
challenges, and both need to provide valid responses for the session to be activated.)
The remote console request (RAKP Message 1) passes a random number and username/privilege
information that the BMC will later use to ‘sign’ a response message based on key information associated
with the user and the Authentication Algorithm negotiated in the Open Session Request/Response
exchange. The BMC responds with RAKP Message 2 and passes a random number an GUID (globally
unique ID) for the managed system that the remote console uses according the Authentication Algorithm to
sign a response back to the BMC.
RAKP Message 3, RAKP Message 4
The session activation process is completed by the remote console and BMC exchanging messages that are
signed according to the Authentication Algorithm that was negotiated, and the parameters that were passed
in the earlier messages. RAKP Message 3 is the signed message from the remote console to the BMC.
After receiving RAKP Message 3, the BMC returns RAKP Message 4 - a signed message from BMC to
the remote console.
The RMCP+ and RAKP Messages are specified in detail later in this section.
13.16 RMCP+ Session Termination
The following actions can terminate a session:
•
The Close Session command
•
Session Inactivity Timeout (See Per IPMI v1.5 section 6.11.13, Session Inactivity Timeouts)
Terminating a session causes any payloads that were activated under that session to be automatically deactivated.
Terminating one session will not cause other sessions to terminate. If multiple sessions are opened by a remote
console they will need to be terminated individually.
13.17 RMCP+ Open Session Request
The remote console sends this RMCP+ message to the managed system to open a protected session. The client
responds with an RMCP+ Open Session Response message. Following the Remote Console Session ID field, this
message contains one or more Authentication Payload proposals, one or more Integrity Payload proposals, and
142
Intelligent Platform Management Interface Specification
one or more Confidentiality Payload proposals. If the RMCP+ Open Session Request is accepted, the BMC has
found a Cipher Suite that matches up with the one or more combinations of the algorithm proposals.
The following table defines the RMCP+ packet fields for this message.
Table 13-9, RMCP+ Open Session Request
byte
U
U
data field
U
U
Payload Type = RMCP+ Open Session Request
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
Message Tag - Selected by remote console. Used by remote console to help match
responses up with requests. In this case, the corresponding Open Session Response
that is returned by the BMC. The BMC can use this value to help differentiate retried
messages from new messages from the remote console.
Requested Maximum Privilege Level (Role)
[7:4] - Reserved for future definition by this specification, set to 0h
[3:0] - Requested Maximum Privilege Level (Role).
0h = Highest level matching proposed algorithms.
BMC will pick the Cipher Suite returned in the RMCP+ Open Session
Response by checking the algorithms proposed in the RMCP+ Open
Session Request against the Cipher Suites available for each privilege
level, starting with the “OEM Proprietary level” and progressing to lower
privilege levels until a match is found. The resultant match results in an
‘effective’ maximum privilege level for the session. The resultant level is
returned in the RMCP+ Open Session Response.
1h = CALLBACK level
2h = USER level
3h = OPERATOR level
4h = ADMINISTRATOR level
5h = OEM Proprietary level
reserved - write as 00_00h
IPMI Session Header
U
U
U
U
U
IPMI Payload
U
1
U
2
U
U
3:4
5:8
U
U
U
Remote Console Session ID. Selected by the remote console to identify packets that
are received for the given session by the remote console
Authentication Payload. Identifies the authentication type that the managed system
wants to use for the session.
byte 1 - Payload Type
00h = authentication algorithm
byte 2:3 - reserved = 0000h
byte 4 - Payload Length in bytes (1-based). The total length in bytes of the payload
including the header (= 08h for this specification).
00h = Null field (“wildcard”). BMC picks algorithm based on Requested Maximum
Privilege Level and that matches with the proposed Integrity and Confidentiality
payloads. If the Requested Maximum Privilege Level is ‘unspecified’ the BMC
picks algorithm based on the Integrity and Confidentiality algorithm proposals
starting from the highest privilege level until a match is found.
byte 5 - Authentication Algorithm
[7:6] - reserved
[5:0] - Authentication Algorithm (See Table 13-17, Authentication Algorithm
Numbers)
byte 6:8 - reserved
Integrity Payload. Identifies the integrity type that the managed system wants to use for
the session.
byte 1 - Payload Type
01h = integrity algorithm
byte 2:3 - reserved = 0000h
byte 4 - Payload Length in bytes (1-based). The total length in bytes of the payload
including the header (= 08h for this specification).
00h = Null field (“wildcard”). BMC picks algorithm based on Requested Maximum
Privilege Level and that matches with the proposed Authentication and
Confidentiality payloads. If the Requested Maximum Privilege Level is
‘unspecified’ the BMC picks algorithm based on the Authentication and
Confidentiality algorithm proposals starting from the highest privilege level until
U
9:16
U
U
U
U
U
U
U
U
U
U
U
U
U
17:24
U
U
U
U
U
U
U
U
143
Intelligent Platform Management Interface Specification
a match is found.
byte 5 - Integrity Algorithm
[7:6] - reserved
[5:0] - Integrity Algorithm (See Table 13-18, Integrity Algorithm Numbers)
byte 6:8 - reserved
Confidentiality Payload. Defined confidentiality algorithms are:
byte 1 - Payload Type
02h = confidentiality algorithm
byte 2:3 - reserved = 0000h
byte 4 - Payload Length in bytes (1-based). The total length in bytes of the payload
including the header (= 08h for this specification).
00h = Null field (“wildcard”). BMC picks algorithm based on Requested Maximum
Privilege Level and that matches with the proposed Authentication and Integrity
payloads. If the Requested Maximum Privilege Level is ‘unspecified’ the BMC
picks algorithm based on the Authentication and Integrity algorithm proposals
starting from the highest privilege level until a match is found.
byte 5 - Confidentiality Algorithm
[7:6] - reserved
[5:0] - Confidentiality Algorithm (See Table 13-19, Confidentiality Algorithm
Numbers)
byte 6:8 - reserved
U
U
U
U
25:32
U
U
U
U
U
U
U
U
U
U
U
U
U
13.18 RMCP+ Open Session Response
A managed system sends this RMCP+ message to the management console in response to an RMCP+ Open
Session Request message. Following the Status Code, Mgmt Console and Managed System Session ID fields, this
message contains a single Authentication payload, a single Integrity payload, and a single Confidentiality payload.
These payloads represent the proposals that the managed system selected from the list offered by the management
console.
The following table defines the RMCP+ packet fields for this message.
Table 13-10, RMCP+ Open Session Response
byte
data field
Payload Type = RMCP+ Open Session Response
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
Message Tag - The BMC returns the Message Tag value that was passed by the remote
console in the Open Session Request message.
RMCP+ Status Code - Identifies the status of the previous message. If the previous message
generated an error, then only the Status Code, Reserved, and Remote Console Session ID
fields are returned. See Table 13-15, RMCP+ and RAKP Message Status Codes. The session is
then immediately closed at the BMC.
Maximum Privilege Level (Role) - Indicates the Maximum Privilege Level allowed for the session
based on the security algorithms that were proposed in the RMCP+ Open Session Request.
[7:4] - Reserved for future definition by this specification, set to 0h
[3:0] - Requested Maximum Privilege Level (Role).
0h = unspecified (returned with error completion code).
1h = CALLBACK level
2h = USER level
3h = OPERATOR level
4h = ADMINISTRATOR level
5h = OEM Proprietary level
reserved - write as 00h
IPMI Session Header
U
U
U
U
U
IPMI Payload
1
2
3
4
5:8
9:12
144
U
U
U
U
U
U
U
U
U
Remote Console Session ID The Remote Console Session ID specified by RMCP+ Open
Session Request message associated with this response.
Managed System Session ID The Session ID selected by the Managed System for this new
session. A null Session ID (All 0’s) is not valid in this context.
U
U
U
U
Intelligent Platform Management Interface Specification
13:20
21:28
29:36
Authentication Payload This payload defines the authentication algorithm proposal selected by
the Managed System to be used for this session (see Table 13-9, RMCP+ Open Session
Request for the definition of this payload). A single algorithm will be returned. The ‘Null field’ is
not allowed.
Integrity Payload This payload defines the integrity algorithm proposal selected by the Managed
System to be used for this session (see Table 13-9, RMCP+ Open Session Request for the
definition of this payload). A single algorithm will be returned. The ‘Null field’ is not allowed.
Confidentiality Payload This payload defines the confidentiality algorithm proposal selected by
the Managed System to be used for this session (see Table 13-9, RMCP+ Open Session
Request for the definition of this payload). A single algorithm will be returned. The ‘Null field’ is
not allowed.
U
U
U
U
U
U
13.19 RAKP Messages
RAKP Messages are transferred in the payload portion of an IPMI over LAN packet with a Format field set to
“RMCP+”. The Payload Type field indicates which RAKP message is held in the IPMI Payload portion of the
packet.
13.20 RAKP Message 1
A remote console sends this RAKP message to the managed system to begin the session authentication process.
The remote console selects a Remote Console Random Number, a Maximum Requested Privilege Level (Role),
and an optional User Name and sends them to the managed system along with the Managed System Session ID
specified by the client on the previous RMCP+ Open Session Response.
Upon receiving RAKP Message 1, the managed system verifies that the message contains an active Managed
System Session ID and that a session can be created using the given user information and security algorithm
proposal. The managed system responds with an RAKP Message 2.
145
Intelligent Platform Management Interface Specification
The format of the IPMI Session Header, Payload, and Session Trailer for the RAKP Message 1 is shown in the
following table.
Table 13-11, RAKP Message 1
byte
data field
Payload Type = RAKP Message 1
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
Message Tag - Selected by remote console. Used by remote console to help match responses
up with requests. In this case, the corresponding RAKP Message 2 that is returned by the BMC.
The BMC can use this value to help differentiate retried messages from new messages from the
remote console.
reserved - write as 00_00_00h
Managed System Session ID
The Managed System’s Session ID for this session, returned by the Managed System on the
previous RMCP+ Open Session Response message.
Remote Console Random Number
Random number selected by the Remote Console
Requested Maximum Privilege Level (Role)
[7:5] - Reserved for future definition by this specification, set to 000b
[4] 0b = Username/Privilege lookup. Both the Requested Privilege Level and User Name
are used to look up password/key. The BMC will search the user entries starting
with USER ID 1 and use the first entry that matches the specified user name and
has a Maximum Privilege Level that matches the Requested Privilege Level. This
can be used in combination with ‘null’ user names to enable a “role only” login with
a password that is just associated with the requested privilege level.
IPMI Session
Header
U
U
U
U
U
IPMI Payload
U
1
U
2:4
5:8
U
U
U
U
9:24
U
25
U
U
U
U
1b = Name-only lookup. User Name alone is used to look up password/key. Privilege
Level field acts as a ‘Maximum Requested Privilege Level’ ala IPMI v1.5. The rules
for privilege level handling are summarized as follows:
U
U
• If the Requested Privilege Level is greater than the privilege limit for the
channel/user, the user will be allowed to connect but will be restricted to the
channel/user privilege limit that was configured for the user.
• If the Requested Privilege Level is less than the channel/user privilege limit, the
user will be allowed to connect and Request Privilege Level will become the
effective privilege limit for the user. I.e. the user will not be able to raise their
privilege level higher than the Requested Privilege Level.
[3:0] - Requested Maximum Privilege Level
0h = reserved
1h = CALLBACK level
2h = USER level
3h = OPERATOR level
4h = ADMINISTRATOR level
5h = OEM Proprietary level
reserved - write as 00_00h
26:27
User Name Length
28
U
29:44
146
00h No name present
01h-10h Name length
11h-FFh Reserved for future definition by this specification
User Name ASCII character Name that the user at the Remote Console wishes to assume for
this session. No NULL characters (00h) are allowed in the name. Sixteen-bytes, max.
U
U
Intelligent Platform Management Interface Specification
13.21 RAKP Message 2
The managed system sends this RAKP message to a remote console in response to the receipt of an RAKP
Message 1. Once RAKP Message 1 has been validated the managed system selects a Managed System Random
Number and computes a Key Exchange Authentication Code over the values specified by the RAKP algorithm.
The managed system sends those values along with the Managed System Globally Unique ID (GUID) and the
Remote Console Session ID (sent by the console on the previous RMCP+ Open Session Request) to the remote
console.
Upon receiving RAKP Message 2, the remote console verifies that the Remote Console Session ID is active and
that the Managed System GUID matches the managed system that the remote console has associated with the
session. The remote console then validates the AuthCode (Integrity Check Value) and responds with an RAKP
Message 3.
The format of an RAKP Message 2 message’s Data section is as follows:
Table 13-12, RAKP Message 2
byte
data field
Payload Type = RAKP Message 2
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
IPMI Session Header
U
U
U
U
U
IPMI Payload
U
Message Tag - The BMC returns the Message Tag value that was passed by the
remote console in RAKP Message 1.
1
2
U
U
RMCP+ Status Code - Identifies the status of the previous message. If the previous
message generated an error, then only the Completion Code, Reserved, and
Remote Console Session ID fields are returned. If the Remote Console Session ID
field is indeterminate (as would be the case if the Managed System Session ID in
RAKP Message 1 were invalid) then the Remote Console Session ID field will be
set to all zeros.
U
U
On error, the remote console can attempt to correct the error and send a new RAKP
Message 1. Note that the remote console must change the Message Tag value to
ensure the BMC sees the message as a new message and not as a retry.
See Table 13-15, RMCP+ and RAKP Message Status Codes for the status codes
defined for this message.
3:4
Reserved - write as 00_00h.
5:8
Remote Console Session ID - The Remote Console Session ID specified by the
RMCP+ Open Session Request message associated with this response.
9:24
Managed System Random Number - Random number selected by the managed
system.
25:40
Managed System GUID - The Globally Unique ID (GUID) of the Managed System.
This value is specified by the client system’s SMBIOS implementation.
41:N
Key Exchange Authentication Code
An integrity check value over the relevant items specified by the RAKP algorithm for
RAKP Message 2. The size of this field depends on the specific Authentication
Algorithm (e.g. for RAKP-HMAC-SHA1 the Authentication Algorithm would be
HMAC using SHA1 to generate a 20-byte authentication code) that was identified in
the RMCP+ Open Session Response. This field may be 0-bytes (absent) for some
algorithms (e.g. RAKP-none).
147
Intelligent Platform Management Interface Specification
13.22 RAKP Message 3
A remote console sends this RAKP message to a managed system in response to the receipt of an RAKP Message
2. Once it validates RAKP Message 2, the remote console creates a Session Integrity Key using the values
specified by the RAKP algorithm. The remote console then computes an Integrity Check Value over the values
specified by the RAKP algorithm, and sends that along with the Managed System Session ID (sent by the
managed system on the previous RMCP+ Open Session Response message) to the managed system.
After receiving RAKP Message 3, the managed system verifies that the Managed System Session ID is active and
then validates the Integrity Check Value. If the Integrity Check Value is valid, the managed system creates a
Session Integrity Key using the values specified by the RAKP algorithm. With the shared Session Integrity Key in
place, integrity protected messages can now be exchanged between the remote console and the managed system.
The format of an RAKP Message 3 message’s Data section is as follows:
Table 13-13, RAKP Message 3
byte
data field
Payload Type = RAKP Message 3
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
IPMI Session Header
U
U
U
U
U
IPMI Payload
U
Message Tag - Selected by remote console. Used by remote console to help match
responses up with requests. In this case, the corresponding RAKP Message 4 that is
returned by the BMC. The BMC can use this value to help differentiate retried
messages from new messages from the remote console.
1
2
U
U
RMCP+ Status Code Identifies the status of the previous message. If the previous
message generated an error, then only the Completion Code, Reserved, and
Managed System Session ID fields are returned.
U
U
If the BMC receives an error from the remote console, it will immediately close the
session if the Key Exchange Authentication code (below) is valid.
3:4
5:8
9:N
148
See Table 13-15, RMCP+ and RAKP Message Status Codes for the status codes
defined for this message.
Reserved - write as 00_00h.
Managed System Session ID
The Managed System’s Session ID for this session, returned by the managed system
on the previous RMCP+ Open Session Response message.
Key Exchange Authentication Code
An integrity check value over the relevant items specified by the RAKP algorithm for
Message 3. The size of this field depends on the specific Authentication Algorithm
that was selected when the session was created. This field may be 0 bytes (absent)
for some algorithms (e.g. RAKP-none).
Intelligent Platform Management Interface Specification
13.23 RAKP Message 4
A managed client sends this RAKP message to a management console in response to the receipt of an RAKP
Message 3. Once RAKP Message 3 has been validated, the managed client computes an Integrity Check Value
over the values specified by the RAKP algorithm. The managed client then sends the Mgmt Console Session ID
and the Integrity Check Value to the management console.
Upon receiving RAKP Message 4, the management console verifies that the Mgmt Console Session ID is active
and then validates the Integrity Check Value.
The format of an RAKP Message 4 message’s Data section is as follows:
Table 13-14, RAKP Message 4
byte
IPMI Session Header
data field
Payload Type = RAKP Message 4
Session ID = 00_00_00_00h
Session Sequence Number = 00_00_00_00h
U
U
U
U
U
IPMI Payload
1
U
Message Tag - The BMC returns the Message Tag value that was passed by the
U
U
remote console in RAKP Message 3.
2
3:4
RMCP+ Status Code - Identifies the status of the previous message. If the
previous message generated an error, then only the Status Code,
Reserved, and Management Console Session ID fields are returned. See
2.1.3.6.1 for the status codes defined for this message.
U
U
Reserved - Reserved for future definition by this specification set to
000000h.
U
5:8
U
9:N
Mgmt Console Session ID The Mgmt Console Session ID specified by the
RMCP+ Open Session Request (83h) message associated with this
response.
U
U
Integrity Check Value An integrity check value over the relevant items
specified by the RAKP algorithm for Message 4. The size of this field
depends on the specific algorithm that was selected when the session was
created. This field may be 0 bytes (absent) for some authentication
algorithms (e.g. RAKP-none)
U
U
149
Intelligent Platform Management Interface Specification
13.24 RMCP+ and RAKP Message Status Codes
The table below lists the status codes for specific RMCP+ and RAKP messages.
Table 13-15, RMCP+ and RAKP Message Status Codes
Status
Code
Description
Message
RMCP+
Open
Session
Response
00h
01h
02h
03h
04h
05h
06h
07h
08h
09h
0Ah
0Bh
0Ch
0Dh
0Eh
0Fh
10h
11h
12h
13h-FFh
No errors
Insufficient resources to create a
session
Invalid Session ID
Invalid payload type
Invalid authentication algorithm
Invalid integrity algorithm
No matching authentication payload
No matching integrity payload
Inactive Session ID
Invalid role
Unauthorized role or privilege level
requested
Insufficient resources to create a
session at the requested role
Invalid name length
Unauthorized name
Unauthorized GUID. (GUID that BMC
submitted in RAKP Message 2 was not
accepted by remote console)
Invalid integrity check value
Invalid confidentiality algorithm
No Cipher Suite match with proposed
security algorithms
Illegal or unrecognized parameter
Reserved for future definition by this
specification
RAKP
Msg 2
RAKP
Msg 3
RAKP
Msg 4
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
13.25 Differences between v1.5 and v2.0/RMCP+ Sessions
The following presents an overview of some notable differences as well as similarities between IPMI v1.5 and
RMCP+ user setup and session activation mechanisms:
150
•
IPMI v1.5 had a hook that would allow packets under a session to have different authentication
signatures than the type of signature that was negotiated to open the session. This hook was to include
a ‘Authentication Type’ field that specified the type of authentication on a per-packet basis. This
capability was not used in the specification. Thus, to simplify things a packet can only have two types
of authentication: the type of authentication selected in the “Open Session” command or “none” - thus
the ‘Authentication Type’ field is deleted and instead the presence or absence of the “Integrity Data”
field is used to indicate whether a given packet in the session is authenticated or not.
•
IPMI v1.5 uses a single challenge-response mechanism for user authentication (the BMC issues a
challenge, and the remote console must issue a response). IPMI v2.0/RMCP+ uses a symmetric
Intelligent Platform Management Interface Specification
challenge where both the BMC and Remote Console issue challenges, and both the BMC and Remote
Console must return correct responses for the session to be activated.
•
ASF 2.0 authentication defined ‘Roles’ such as User and Administrator, where a key (password) was
associated with each role. IPMI v1.5 authentication associated a key with each User Name where a
‘privilege level’ (such as User or Administrator) was configured for each user name. These two
approaches are both available in IPMI v2.0/RMCP+. The BMC can be configured with ‘null’ user
names, whereby key lookup is done based on ‘privilege level only’, or with non-null user names,
where the key lookup for the session is determined according to the user name.
•
IPMI v1.5 uses a single, common Session ID that identifies the session to the BMC and remote
console. IPMI v2.0/RMCP+ allows the BMC and remote console to both pick Session IDs that identify
their incoming traffic for the session.
•
IPMI v1.5 uses a single key (the user key/password) that is used both for authentication and in
integrity (AuthCode) calculations. IPMI v2.0/RMCP+ can be configured to use a single key (“onekey”) login where the user key is used both for authentication and to generate a Session Integrity Key
that is used in integrity (AuthCode) calculations, or a “two key” login where the user key is used for
authentication, and a separate “BMC key”, KG, is used to create the Session Integrity Key that is used
in integrity (AuthCode) calculations.
13.26 IPMI v2.0 RMCP+ Payload Types
The Payload Type field in the IPMI v2.0/RMCP+ packet carries a Payload Type Number that identifies what type
of payload field is being carried in that particular packet.
The Payload Type Numbers (also referred to as payload types) are classified into three main categories: Standard
Payload Types - used to identify payloads that are specified by the IPMI specifications, Session Setup Payload
Types - used to identify payloads that are for session setup messages specified by the IPMI specifications, and
OEM Payload Types that are used to identify payloads that are specified by a given OEM.
The following table lists the assignment and ranges of the Payload Type Numbers. The complete identification of
an OEM Payload is given by the combination of a three-byte IANA ID for the OEM, a reserved byte, plus a twobyte OEM Payload ID that is assigned and defined by the given OEM. These can either be carried explicitly in
each packet (adding six bytes of overhead) or an application can elect to use an OEM Payload Type Handle. The
OEM Payload Type Handle in the Payload Type provides a value that represents a particular OEM IANA and
OEM Payload ID on a system. The Get Channel Payload Support command is used discover what OEM Payloads
(if any) are used on the managed system, and to associate the OEM Payload Type number with the OEM IANA
ID and Payload ID.
OEM Payload Handle assignments can vary from system to system. For example, “OEM0” on one system may
not correspond to the same type of OEM Payload as “OEM0” on another system. Software that uses OEM ayload
Handles must not assume that a given OEM Payload Handle number will correspond to a particular OEM IANA
and Payload ID combination across multiple systems. Software must use the Get Channel Payload Support
command to discover the relationship.
Associated with each payload type is a format version number that provides information on the backward
compatibility with different versions of the payload type. See Section 24.9, Get Channel Payload Version
Command, for more information.
13.27 Payloads and Payload Type Numbers
Payload Type Numbers are used in the “payload type” field of an IPMI v2.0/RMCP+ packet to identify what’s
being carried in the data portion of the packet. This data can be categorized into the following types of content:
Standard payload types that specify commands and data content for messages and protocols defined in this
151
Intelligent Platform Management Interface Specification
specification. Session Setup payload types that are used for messages for session startup and remote access
authentication algoritihms defined in this specification, and OEM Payload Type Handles that are used to identify
OEM-specified data that is carried in an IPMI v2.0/RMCP+ packet.
The Get Channel Payload Support command returns which standard payload type numbers and OEM payload
type handles are available on a given channel of a BMC.
13.27.1 IPMI Message Payloads and IPMI Commands
IPMI Message Payloads are always accepted over any IPMI session, because they are used for IPMI commands
that are used for managing sessions. Thus, the IPMI payload type does not need to be explicitly enabled, and
cannot be disabled via the Activate and Deactivate Payload commands, respectively.
T
However, while the IPMI Message payload type is accepted, specific IPMI commands may not be accepted.
For example, the Set User Access command determines whether a given user can execute IPMI commands that
are not specific to managing a session or to specific to a particular payload type. For example, if IPMI
Messaging is disabled for a user, but the user is enabled for activating the SOL payload type, then IPMI
commands associated with SOL and session management, such as Get SOL Configuration Parameters and
Close Session are available, but generic IPMI commands such as Get SEL Time are unavailable on the SOL
Payload session.
13.27.2 OEM Payload Type Handles
OEM Payload Type Handles are a specific numeric range of values that can be carried in the payload type field
of an IPMI v2.0/RMCP+ packet. These values do not explicitly specify a type of OEM payload, but instead are
“handles” that are used to identify and access an OEM payload type on a given implementation or instance of a
BMC. OEM Payload Types are actually specified by the combination of an OEM IANA and an OEM-specified
Payload ID number. The OEM Payload Type Handle can be used in the Get Channel OEM Payload Info
command can be used to look up the OEM IANA and OEM Payload ID associated with a particular payload
type number.
152
Intelligent Platform Management Interface Specification
13.27.3 Payload Type Numbers
The following table defines the payload type numbers and ranges for OEM Payload Type Handles.
Table 13-16, Payload Type Numbers
number[1]
P
0h
1h
2h
10h
11h
12h
13h
14h
15h
20h-27h
1.
P
type
Standard Payload Types
IPMI Message
SOL (serial over LAN)
OEM Explicit
(When this payload type appears in the network
header, it indicates that the packet includes explicit
OEM IANA and OEM Payload ID fields that identify
the payload type, instead of using an OEM Payload
Type Handle to identify the type of payload
contained in the packet. When used, this option
adds 6 bytes to the overhead of the packet.)
Session Setup Payload Types
RMCP+ Open Session Request
RMCP+ Open Session Response
RAKP Message 1
RAKP Message 2
RAKP Message 3
RAKP Message 4
OEM Payload Type Handles
Handle values for OEM payloads OEM0 through
OEM7, respectively.
major format
version
minor format
version
1h
1h
OEM specified
according to OEM
IANA and OEM
Payload ID.
0h
0h
OEM specified
according to OEM
IANA and OEM
Payload ID.
1h
1h
1h
1h
1h
1h
0h
0h
0h
0h
0h
0h
OEM specified
according to OEM
IANA and OEM
Payload ID.
OEM specified
according to OEM
IANA and OEM
Payload ID.
all other
reserved
The payload type number is a 6-bits (00h-3Fh).
13.28 Authentication, Integrity, and Confidentiality Algorithm Numbers
The Authentication Algorithm Number specifies the type of authentication “handshake” process that is used and
identifies any particular variations of hashing or signature algorithm that is used as part of the process.
Table 13-17, Authentication Algorithm Numbers
number*
*
type
Mandatory /
Optional1
TP
PT
00h
RAKP-none
M
01h
RAKP-HMAC-SHA1
M
02h
RAKP-HMAC-MD5
O
C0h-FFh
OEM
O
all other
reserved
The number range is limited to six (6) bits (00h-3Fh)
13.28.1 RAKP-HMAC-SHA1 Authentication Algorithm
RAKP-HMAC-SHA1 specifies the use of RAKP messages for the key exchange portion of establishing the
session, and that HMAC-SHA1 is used to create the Key Exchange 20-byte Authentication Code fields in
TP
1
PT
Mandatory/Optional is with respect to BMC support. It is recommended that remote consoles support all specified algorithms in
order to support maximum number of BMC implementations.
153
Intelligent Platform Management Interface Specification
RAKP Message 2 and RAKP Message 3. HMAC-SHA1-96 (per RFC2404) is used for generating a 12-byte
Integrity Check Value field for RAKP Message 4.
13.28.2 RAKP-none Authentication Algorithm
RAKP-none uses the same steps and messages as RAKP-HMAC-SHA1, but the Key Exchange Authentication
Code field in RAKP Message 2 and RAKP Message 3 and the Integrity Check Value field in RAKP Message 4
are absent since they are not used. The RAKP steps establish Session IDs and privilege level for the given
username/role. A BMC implementation can be configured with a null username that has a null (all 0’s)
password. A BMC configured this way, and using the RAKP-none Authentication Algorithm, provides a way
to enable access the BMC without requiring a username and password.
13.28.3 RAKP-HMAC-MD5 Algorithm
This authentication algorithm operates the same way as RAKP-HMAC-SHA1 except that the HMAC with
MD5 is (per [RFC2104] is used for RAKP authentication operations in place of SHA-1. Thus, the Key
Exchange Authentication Code fields in RAKP Message 3 and RAKP Message 4 and the Integrity Check Value
field in RAKP Message 4 are all 16-byte fields (128-bit MD5). Since MD5 requires fewer computational steps
than SHA-1, this option can be used to offer a quicker session activation, particularly on management
controllers that have limited computational resources.
13.28.4 Integrity Algorithms
The Integrity Algorithm Number specifies the algorithm used to generate the contents for the AuthCode
“signature” field that accompanies authenticated IPMI v2.0/RMCP+ messages once the session has been
established.
Unless otherwise specified, the integrity algorithm is applied to the packet data starting with the
AuthType/Format field up to and including the field that immediately precedes the AuthCode field itself.
HMAC-SHA1-96 and HMAC-MD5-128 utilize the Session Integrity Key as the key for use in HMAC. For
“two key” logins, Kg is used in the creation of SIK, for “one key” logins, the user’s key is used.
MD5-128 uses a straight MD5 signature with the user’s key information appended at the beginning and the end
of the packet data as: AuthCode = MD5(password + AuthType/Format + … + Next_Header + password)
Table 13-18, Integrity Algorithm Numbers
number*
type
Mandatory /
Optional
00h
none
M
01h
HMAC-SHA1-96
M
02h
HMAC-MD5-128
O
03h
MD5-128
O
C0h-FFh
OEM
O
all other
reserved
* The number range is limited to six (6) bits (00h-3Fh)
154
Intelligent Platform Management Interface Specification
13.28.5 Confidentiality (Encryption) Algorithms
The Confidentiality Algorithm Number specifies the encryption/decryption algorithm field that is used for
encrypted payload data under the session. The ‘encrypted’ bit in the payload type field being set identifies
packets with payloads that include data that is encrypted per this specification. When payload data is encrypted,
there may be additional “Confidentiality Header” and/or “Confidentiality Trailer” fields that are included within
the payload. The size and definition of those fields is specific to the particular confidentiality algorithm.
Table 13-19, Confidentiality Algorithm Numbers
number*
Mandatory /
Optional
type
00h
01h
none
AES-CBC-128
(See Section 13.29, AES-CBC-128 Encrypted Payload Format, for more information)
02h
xRC4-128
(See Section 13.30, xRC4 Encrypted Payload Format, for more information)
03h
xRC4-40
(See Section 13.30, xRC4 Encrypted Payload Format, for more information)
30-3Fh
OEM
all other
reserved
* The number range is limited to six (6) bits (00h-3Fh)
M
M
O
O
O
-
13.29 AES-CBC-128 Encrypted Payload Format
The following table summarizes the contents of the IPMI Payload when AES-CBC encryption is used.
Table 13-20, AES-CBC Encrypted Payload Fields
Field
Confidentiality
Header
Size
16
Sub field
Initialization Vector
Payload Data
Confidentiality
Trailer
variable
variable
Payload Data
Confidentiality Pad
1
Confidentiality Pad Length
Description
For the AES algorithm in CBC mode, this field must be a
16-byte random value generated uniquely for each
message (packet).
Added to the Data field to be encrypted (including the
Confidentiality Pad Length field) so that they have a
length that is a multiple of the block size of algorithm
being used. For the AES algorithm, the block size is 16
bytes.
Defines the number of Confidentiality Pad bytes present
in the message. For the AES algorithm, this number will
range from 0 to 15 bytes. This field is mandatory. If no
Confidentiality Pad bytes are required, the Confidentiality
Pad Length field is set to 00h. If present, the value of
the first byte of Confidentiality Pad shall be one (01h) and
all subsequent bytes shall have a monotonically
increasing value (e.g., 02h, 03h, 04h, etc). The receiver,
as an additional check for proper decryption, shall check
the value of each byte of Confidentiality Pad. Some
messages may not require padding if the messages
already provide the necessary alignment.
13.29.1 Generating the Initialization Vector
The initialization vector (IV) should be unpredictable. In particular, for any given plaintext, it must not be
possible to predict what the next IV will be from the last IV. For AES-CBC-128, the IV is recommended to be a
155
Intelligent Platform Management Interface Specification
16-byte random number generated by a high quality random number generation process. See Section 13.34,
Random Number Generation.
13.29.2 Encryption with AES
AES-128 uses a 128-bit Cipher Key. The Cipher Key is the first 128-bits of key “K2”, K2 is generated from
the Session Integrity Key (SIK) that was created during session activation. See Section 0,
156
Intelligent Platform Management Interface Specification
RAKP Message 3 and Section 13.32, Generating Additional Keying Material.
Once the Cipher Key has been generated it is used to encrypt the payload data. The payload data is padded to
make it an integral numbers of blocks in length (a block is 16 bytes for AES). The payload is then encrypted
one block at a time from the lowest data offset to the highest using Cipher_Key as specified in [AES].
13.29.3 CBC (Cipher Block Chaining)
When CBC is used, before a block of payload data is encrypted it is first exclusive-ORed with the previous
ciphertext block (or in the case of the first block, with the initialization vector). For AES, this means that the
each 16-byte block of plaintext payload data is exclusive-ORed with the previous 16-bytes of encrypted data
before being encrypted. See [MODES] for information on CBC.
For AEC-CBC-128 encrypted payloads under IPMI v2.0 RMCP+, CBC does not span between packets, it only
applies to blocks within a packet. Instead, each individual packet is encrypted using a different initialization
vector. Thus, packets can be decrypted even if an intermediate block is lost.
13.30 xRC4 Encrypted Payload Format
The following applies to both xRC4-128 and xRC4-40 encryption. The difference between the two has to do with
the size of the key value used to initialize the algorithm. xRC4-128 uses a 128-bit key, and xRC4-40 uses a 40-bit
key. The generation of the initialization key is described in the last paragraphs of this section.
Table 13-21, xRC4 Encrypted Payload Fields
Field
Confidentiality
Header
Size
4
16
Payload Data
Confidentiality
Trailer
variable
0
Sub field
Data offset
Initialization Vector
Payload Data
none
Description
This value advances ‘N’ counts for every N-bytes of new
payload data that is encrypted. The value for the first packet of
payload data is 0000_0000h. If the first packet contains 12 bytes
of payload data, the data offset for the second packet will be 12
(0Ch). If the second packet contained 8 bytes of payload data,
the offset for the third packet will be 20 (14h), and so on. The
xRC4 algorithm operates in a manner similar to a large pseudorandom number generator. Therefore, decryption can handle
missed packets by advancing the state machine by the number
of steps to
The Initialization Vector is a 128-bit random number that is used
in conjunction key information for the session to initialize the
state machine for xRC4. The Initialization Vector is only passed
when the xRC4 state machine is initialized or is reinitialized
(data offset = 0000_0000h). This field is absent when the data
offset is non-zero.
xRC4 does not add use a confidentiality trailer.
13.30.1 Generating the xRC4 Initialization Vector
The initialization vector (IV) should be unpredictable. In particular, for any given plaintext, it must not be
possible to predict what the next IV will be from the last IV. For xRC4, the IV is recommended to be a 16-byte
random number generated by a high quality random number generation process. See Section 13.34, Random
Number Generation.
157
Intelligent Platform Management Interface Specification
13.30.2 Initializing the xRC4 State Machines
There are two xRC4 State Machines that are maintained by the BMC for each xRC4 encrypted payload stream.
One is used for BMC-to-Remote Console encryption, and the other for Remote Console-to-BMC decryption.
These shall be referred to as the BMC Encryption and BMC Decryption state machines, respectively.
The BMC is responsible for creating the Initialization Vector used for initializing the BMC Encryption state
machine. The remote console is responsible for generating the Initialization Vector for the BMC Decryption
state machine. The BMC initializes the BMC Encryption state machine for the first encrypted packet it
generates for the payload, and re-initializes if the remote console requests it via the Suspend/Resume Payload
Encryption command.
The BMC initializes the BMC Decryption State machine whenever it receives an encrypted packet that has a
value of 0000_0000h for the data offset. These packets will contain and Initialization Vector that was generated
by the remote console.
The state machines for both the BMC and remote console are initialized using the same algorithm. This
algorithm creates a key using a combination of the Initialization Vector and the first 128-bits of key “K2” to
initialize the state table for xRC4. (K2 is generated from the Session Integrity Key (SIK) that was created
during session activation. See Section 0,
158
Intelligent Platform Management Interface Specification
RAKP Message 3, and Section 13.32, Generating Additional Keying Material) This key is then fed into the
xRC4 algorithm to initialize a 256-byte state table.
The xRC4 key (KRC) is generated using the combination of K2 and the Initialization Vector as:
KRC = MD5(K2, IV)
Where:
MD5 =
MD5 algorithm applied to the concatenation of K2 and IV.
K2 =
128-bit key generated from Session Integrity Key as described in Section 0,
159
Intelligent Platform Management Interface Specification
RAKP Message 3 and Section 13.32, Generating Additional Keying Material.
IV =
Initialization Vector. A 128-bit random number
For xRC4 using a 128-bit key, all bits of KRC are used for initialization. For xRC4 using a 40-bit key, only the
most significant forty bits of KRC are used.
13.31 RMCP+ Authenticated Key-Exchange Protocol (RAKP)
RMCP+ can support a number of different authentication and key exchange protocols during its Creation (session
activation) phase. For this specification, the mandatory-to-implement authentication and key exchange protocol
is the RMCP+ Authenticated Key-Exchange Protocol (RAKP). RAKP (defined below) was developed based on
the Authenticated Key Exchange Protocol (AKEP) defined by Bellare and Rogaway in [BR1].
RAKP-HMAC-SHA1 uses pre-shared symmetric keys and HMAC-based integrity algorithms to mutually
authenticate a remote console to a given managed system and to generate pair-wise unique symmetric keying
material that can be used with a number of integrity and confidentiality algorithms to provide protection for
RMCP messages. For this specification, RAKP shall use the HMAC-SHA1 integrity algorithm defined in
[RFC2104] and the HMAC-SHA1-96 integrity algorithm defined in [RFC2404].
RAKP also supports the concept of remote console user “roles” and optionally “usernames” (e.g. operator “x” or
administrator “y”), which are established by RAKP when a session is created.
Examples of behavior that can be controlled include the roles that the managed system can use to establish
sessions (e.g. operator-only sessions) and the roles and names (optional) allowed to execute each RMCP message
the managed system might receive during a given session.
Before a given managed system’s RMCP implementation can become operational, it must be configured with
various RMCP-related parameters. This includes installing user passwords (keys) and usernames, setting up
access rights for the individual users, and configuring which Cipher Suites are used for authenticated and/or
encrypted transfers with the managed system.
The managed system can be configured with keys for each username, or ‘null’ usernames can be used, in which
case the key is associated solely with a given privilege level (role). The different user keys are specified using the
notation K[UID], where UID represents the User ID number that is used in the user-specific configuration
commands in IPMI.
The user keys are ‘shared secrets’ between the BMC and the remote console(s). RAKP/RMCP+ does not include
a secure, confidential mechanism for installing and distributing user keys between BMCs and remote consoles.
The installation and distribution of user keys can typically be accomplished with a software utility that uses OSprovided mechanisms for the secure transfer of keys. If authentication and encryption are available, an ‘Admin’
level user can use IPMI commands such as Set User Password for remotely updating and configuring user key
and privilege information over an authenticated and confidential session to the BMC.
An additional key, KG, is used for key generation operations. KG functions essentially as a key for the overall
BMC, and is thus also referred to as the “BMC Key”. A user needs to know both KG and a user password (key
KUID) to establish a session, unless the channel is configured with a ‘null’ KG, in which case the user key (KUID)
is used in place of KG in the algorithms. The scope of these keys (whether they are shared by multiple managed
systems and the remote console or are pair-wise unique for each managed system and the remote console) is a
local policy issue that is determined by the equipment owner at the time of installation. Setting keys is described
further in Section 13.33, Setting User Passwords and Keys.
Once this and other necessary RMCP-related data is installed in the managed system and the managed system is
initialized, the remote console can initiate sessions with the managed system. Following the exchange of RMCP
Presence Ping/Pong and RMCP+ Open Session Request/Response messages (exchanging Session IDs and
160
Intelligent Platform Management Interface Specification
selecting RAKP for use), the remote console starts the RAKP protocol. First, the remote console selects a random
number, RM, a requested role, RoleM, a user name length, ULengthM, a user name (optional - denoted by < >
below), UNameM, and the managed system’s Session ID, SIDC, and sends them to the managed system as
Message 1.
Message 1: Remote Console -► Managed System
SIDC, RM, RoleM, ULengthM, < UNameM >
After receiving Message 1, the managed system verifies that the value SIDC is active and that a session can be
created using RoleM, ULengthM, and (optional), UNameM for the given selections for security algorithms.
If the request is valid, the managed system then selects a random number, RC, and sends to the remote console as
Message 2 the values SIDM, RC, and GUIDC as well as the HMAC per [RFC2404] of the values (SIDM, SIDC,
RM, RC, GUIDC, RoleM, ULengthM, < UNameM >) generated using key K[UID] selected by the given
username, UNameM, and role, RoleM.
Message 2: Managed System -► Remote Console
SIDM, RC, GUIDC,
HMACK[UID] (SIDM, SIDC, RM, RC, GUIDC, RoleM, ULengthM, < UNameM >)
Where:
Parameter
SIDM
SIDC
RM
RC
GUIDC
RoleM
B
B
B
B
B
B
B
B
B
B
B
B
bytes
4
4
16
16
16
1
ULengthM
1
UNameM
1
B
B
B
B
Name
Remote_Console_Session_ID
Managed_System_Session_ID
Remote Console Random_Number
Managed System Random Number
Managed_System_GUID
Requested Privilege Level (Role) (this is the entire byte
holding the Requested Privilege Level field)
User Name Length byte (number of bytes of UNameM = 0
for ‘null’ username)
User Name bytes (absent for ‘null’ username)
After receiving RAKP Message 2, the remote console verifies that the value SIDM is active and that GUIDC
matches the managed system that the remote console is expecting to communicate with. The remote console then
validates the HMAC. If the HMAC is valid, the remote console creates the Session Integrity Key (SIK) by
generating an HMAC per [RFC2104] of the concatenation of RM, RC, RoleM, ULengthM, and (optional)
UNameM using key KG (note - no truncation).
SIK = HMACKG (RM | RC | RoleM | ULengthM | < UNameM >)
Then the remote console sends to the managed system as Message 3 the value SIDC and the HMAC per
[RFC2404] of the values (RC, SIDM, RoleM, ULengthM, < UNameM >) generated using key K[UID] selected by
the username, UNameM, and role RoleM.
Message 3: Remote Console -► Managed System
SIDC, HMACK[UID] (RC, SIDM, RoleM, ULengthM, < UNameM >)
161
Intelligent Platform Management Interface Specification
After receiving Message 3, the managed system verifies that the value SIDC is active and then validates the
HMAC. If the HMAC is valid, the managed system creates the SIK by generating an HMAC per [RFC2104] of
the concatenation of RM, RC, RoleM, ULengthM, and (optional) UNameM using key KG (note - no truncation).
SIK = HMACKG (RM | RC | RoleM | ULengthM | < UNameM >)
The managed system then sends to the management console as Message 4 the values SIDM, and the HMAC per
[RFC2404] of the values (RM, SIDC, GUIDC) generated using key SIK. The managed system then transitions
into the Message Transfer session state,
Message 4: Managed System -► Mgmt Console
SIDM, HMACSIK (RM, SIDC, GUIDC)
After receiving Message 4, the management console verifies that the value SIDM is active and then validates the
HMAC. If the HMAC is valid, the management console has verification that mutual authentication with the
managed system was successful and that the same pair-wise unique SIK was successfully generated on both ends
of the connection. The management console then transitions into the Message Transfer session state (the session
is now active and, if authentication or authentication/encryption have been enabled, the transfer of authenticated
and authenticated/encrypted payloads can commence).
The same RAKP steps are followed for session activation even if the Cipher Suite indicates that there are no
integrity or encryption algorithms required for the session.
13.32 Generating Additional Keying Material
Because this specification supports both integrity and confidentiality services for a session, RSP needs more
keying material than can be provided by the session integrity key, SIK, alone. As a result, all keying material for
the RSP integrity and confidentiality algorithms will be generated by processing a pre-defined set of constants
using HMAC per [RFC2104], keyed by SIK.
K1 = HMACSIK (const 1)
K2 = HMACSIK (const 2)
K3 = HMACSIK (const 3)
These constants are constructed using a hexadecimal octet value repeated up to the HMAC block size in length
starting with the constant 01h. This mechanism can be used to derive up to 255 HMAC-block-length pieces of
keying material from a single SIK. For the mandatory-to-implement integrity and confidentiality algorithms
defined in this specification, processing the first two (2) constants will generate the require amount of keying
material.
Const 1
Const 2
Const 3
= 0x01010101010101010101
= 0x02020202020202020202
= 0x03030303030303030303
.
.
Const 255 = 0xFFFFFFFFFFFFFFFFFFFF
01010101010101010101
02020202020202020202
03030303030303030303
FFFFFFFFFFFFFFFFFFFF
13.33 Setting User Passwords and Keys
User passwords (keys) are set using the Set User Password command. KG is set using the Set Security Keys
command. The Set Security Keys command allows a different Kg to be used on each channel. To enable the
162
Intelligent Platform Management Interface Specification
option of a ‘single key’ login on a given channel, the convention is to configure KG to a reserved value of all 0’s.
The BMC will then use the value KUID in place of KG. Software that allows a user to login to a BMC by entering
a username, password, and BMC Key should send values of all 0’s to the BMC when the user does not provide
explicit values for those fields. The Get Channel Authentication Capabilities command can help a remote console
application tailor any data entry queries for username and password information to what is in use on a given
BMC.
13.34 Random Number Generation
Algorithms for authentication (data integrity) and confidentiality (for Initialization Vectors) depend on "quality"
random numbers for their security. Quality in this context means that the numbers must be random in a
cryptographic sense (i.e., they must be genuinely unpredictable). To ensure that a baseline-level of quality
random numbers are provided for remote consoles and managed systems, this specification recommends the
following algorithm be used for random number generation for use in authentication and data integrity algorithms
if no other higher-quality source of random numbers is available (e.g., a hardware random number generator).
13.34.1 Random Number Key
Each BMC is configured with a 160-bit key, KR, which is unique for each managed system. The Set Security
Keys command can be used for setting this key. Note that to avoid the possibility of run-time software changing
this key, the BMC includes an option for the key to be locked so that run-time software cannot change the key.
13.34.2 Random Number Generator Counters
The managed system maintains two (2) 32 bit counters, CP and CQ. CP is used to count the number of device
power cycles and its value is saved in non-volatile storage. CQ is used to count the number of random number
generation requests per power cycle and is volatile. Whenever a new KR is installed in non-volatile storage, the
counters are reset to zero (0). Once initialized, CP is incremented by one (1) after each power cycle and its new
value is again saved in non-volatile storage. CQ is reset to zero (0) on each power cycle (i.e., its value is not
saved across power cycles).
13.34.3 Random Number Generator Operation
The managed system creates a random number by generating an HMAC per [RFC2104] of the concatenation of
CP and CQ using key KR using the SHA1 hash function. The output of this generator is a 160-bit pseudorandom number. Uses that require fewer bits can draw the required number of bits from the 160-bit value. Since
the value is random, it shouldn’t matter which bits are used. Most implementations will simply take either the
most-significant ‘N’ bits or least-significant, whichever is most convenient.
Random Number = HMACKR (CP | CQ)
CQ is incremented by one (1) after each random number generation request. After each power cycle, the value
of CQ is reset to zero (0) (i.e., its value is not saved across power cycles). If during a given power cycle, CQ
rolls-over to zero, the managed system must increment CP by one (1) and save its new value back into nonvolatile storage.
163
Intelligent Platform Management Interface Specification
14. 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.
14.1 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.
14.2 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.
164
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
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.
14.2.1 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.
14.2.2 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:
•
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).
165
Intelligent Platform Management Interface Specification
14.2.3 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 14-1, 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 14-1 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 14-1, 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)
166
Intelligent Platform Management Interface Specification
14.2.4 Serial Port Switching
The following can cause a switch of the serial port:
Table 14-1, Serial Port Switching Triggers
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])
Baseboard to BMC
[1]
disable
P
yes
yes
no
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
P
Notes
<ESC>Q
<ESC>(
Pattern is also used for
Connection Mode Auto-detect
capability. See 14.2.9,
Connection Mode Auto-detect.
Requires Basic Mode to be
enabled. Pattern is also used for
Connection Mode Auto-detect
capability. See 14.2.9,
Connection Mode Auto-detect.
Requires PPP mode to be
enabled. Pattern is also used for
Connection Mode Auto-detect
capability. See 14.2.9,
Connection Mode Auto-detect.
yes
In Modem Connect mode only.
See 14.2.11, Modem 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’.
14.2.5 Access Modes
BMC channels used for serial/modem access can be configured for several Access Modes using the Set
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 Modes for more
information.
14.2.6 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
167
Intelligent Platform Management Interface Specification
starting point. The actual specification of the BMC with BIOS console redirection is outside the scope of this
specification.
14.2.6.1
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.
14.2.6.2
Connecting to the BMC
The remote console can cause the connection to be switched back to the BMC by the mechanisms listed in
Error! Reference source not found.. 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.
168
Intelligent Platform Management Interface Specification
14.2.6.3
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.
14.2.6.4
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.
14.2.6.5
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.
14.2.6.6
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
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
169
Intelligent Platform Management Interface Specification
Access command to determine that the BMC is configured to operate in Pre-boot Only mode for the serial
connection.
14.2.6.7
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.
14.2.6.8
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.
14.2.7 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 14-2, 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
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.
U
U
U
disabled
any
except
‘disabled’
U
U
U
U
170
U
Intelligent Platform Management Interface Specification
enabled
disabled
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 = If 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. BMC will take over mux if Ring Time expires while Ring is
being detected via the RI signal. If system power is on, the 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 connect with the modem 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.
• 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.
U
U
U
enabled
pre-boot
U
U
U
U
U
U
U
U
enabled
always
available
U
U
U
U
U
U
U
U
U
171
Intelligent Platform Management Interface Specification
enabled
•
•
•
shared
U
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 = If 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. BMC will take over mux if Ring Time expires while Ring is
being detected via the RI signal. If system power is on, the 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 connect with the modem 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.
U
U
U
U
U
U
U
14.2.8 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
172
•
The BMC needs the ability to monitor the DCD and RI signals from the serial connector in order to
detect incoming modem calls.
•
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.
Intelligent Platform Management Interface Specification
•
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.
14.2.9 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.
•
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
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.
173
Intelligent Platform Management Interface Specification
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 14-3, 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
174
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
is terminated or the data pattern for Basic Mode or PPP Mode
IPMI-RMCP packet is detected.
Intelligent Platform Management Interface Specification
14.2.10 Modem-specific Options
The serial/modem configuration parameters (set using the Set 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 14-4, 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.
14.2.11 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 parameters2 and Section 14.2.7, Serial Port Sharing Access
Characteristics. After getting the connection, if DCD is already asserted, the BMC will monitor the incoming data
TP
TP
2
PT
PT
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.
175
Intelligent Platform Management Interface Specification
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.
14.3 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.
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.
176
Intelligent Platform Management Interface Specification
14.3.1 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.
14.3.2 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.
14.3.3 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.
14.3.4 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
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.
177
Intelligent Platform Management Interface Specification
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.
14.4 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.
14.4.1 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 14-5, Basic Mode Special Characters
Description
Value
Start Character
Stop Character
Packet Handshake Character
Data Escape Character
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 14.4.5, Packet Handshake.
P
P
14.4.2 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.
Table 14-6, 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
178
Intelligent Platform Management Interface Specification
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.
14.4.3 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 14-2, 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
netFn
rq
rqLUN
rqAddr
rqSeq
rs
rsLUN
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.
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
179
Intelligent Platform Management Interface Specification
rsAddr
Seq
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.
14.4.4 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.
14.4.5 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.
180
Intelligent Platform Management Interface Specification
14.5 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.
14.5.1 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.
14.5.2 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 14-3, 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
14.5.3 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 packet3.
P
P
Support for the 16-bit (2-byte) FCS (frame check sequence) is mandatory.
TP
3
PT
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.
181
Intelligent Platform Management Interface Specification
14.5.4 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 14-7, 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
14.5.5 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 14-8, 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 14-4, 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 14-7, 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.
182
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 14-9, PPP Link Configuration Option Support Requirements
ID
1
2
3
Type
Maximum Receive Unit
Asynch Control
Character Map
Authentication Protocol
Len
4
Data
bytes 1:2 - 2-byte value indicating
request
U
U
Note that per PPP an implementation must
accept at least 1500 bytes. See Section 14.5.6,
Maximum Receive Unit Handling for more
information.
Request and Response: Optional
≥4
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
U
U
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
U
4
Quality Protocol
7
8
1.
Magic Number
≥4
6
Request and Response: based on firmware
support for PAP, CHAP, and MS-CHAP
U
U
U
U
5
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.
Request and Response: Optional
U
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.
U
U
183
Intelligent Platform Management Interface Specification
14.5.6 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.
14.5.7 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.
14.5.8 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
184
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.
14.5.9 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 14-5, 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 13.8, IPMI LAN
Value
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
P
P
P
P
P
P
7Eh
[1]
FFh
[1]
03h
P
P
P
P
0021 = IPv4
11h
26Fh
[3]
FFh for IPMI
07h for IPMI
P
P
[5]
note
[5]
note
P
P
P
P
1
varies
Message Format
PPP Frame
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.
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
P
P
185
Intelligent Platform Management Interface Specification
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.
14.5.10 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 14-6, IP Frame with Field Compression
Flag
(7Eh)
Protocol
(21h = IPv4)
UDP/RMCP/IPMI Packet data
FCS
2 bytes
Flag
(7Eh)
14.5.11 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.
14.5.12 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.
14.5.13 Escaped Character Handling
By default, the following characters are escaped:
Table 14-10, 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 14.5.14, 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.
14.5.14 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
186
Intelligent Platform Management Interface Specification
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 14.5.13, Escaped 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 14-10, 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.
14.5.15 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.
•
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 25-4, 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
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.
187
Intelligent Platform Management Interface Specification
•
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.
14.5.16 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:
188
•
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 25-4, Serial/Modem Configuration Parameters.
•
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.
Intelligent Platform Management Interface Specification
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.
14.6 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 6.12.13, Session 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.
14.6.1 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.
189
Intelligent Platform Management Interface Specification
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 14-11, 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.
14.6.1.1
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 14.11.1, Alert 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.
14.7 Terminal Mode
Terminal Mode is an operating mode of a serial/modem channel used for the following purposes:
•
190
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.
Intelligent Platform Management Interface Specification
•
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.
14.7.1 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
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 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.
14.7.2 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.
14.7.3 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
191
Intelligent Platform Management Interface Specification
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 14-7, 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 14-8, Terminal Mode Response from BMC
Byte 1
NetFn (odd) / rsLUN=00b (BMC)
192
Byte 2
rqSeq/Bridge=00b (BMC)
Byte 3
Cmd
Byte 4
Completion Code
Byte 5:N
Data
Intelligent Platform Management Interface Specification
14.7.4 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 14-12, Terminal Mode Message Bridge Field
Message
Direction
(to BMC)
Bridge
Field
00b
Request/
Response
Request
00b
Response
Out
00b
Request
In
LUN
00b,
01b,
11b
00b,
01b,
11b
10b
00b
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
In
Message Interpretation
Remote Console request to BMC functionality
Message is a request from the remote console to the BMC
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.
14.7.5 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.
193
Intelligent Platform Management Interface Specification
Figure 14-9, 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 14-10, 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
14.7.6 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 14-11, 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.4
TP
PT
Figure 14-12, 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 14-13, 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.
TP
4
PT
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.
194
Intelligent Platform Management Interface Specification
14.7.7 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.
14.7.8 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 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 14-13, 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
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 28-14, 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
returns the data from the command in hex-ascii format, with a maximum of four hex-
195
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
ascii 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 25-4, 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 25-4, 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 25-4, 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>
PowerSystem:xx<output termination sequence>
Cooling:xx<output termination sequence>
Drives:xx<output termination sequence>
Security:xx<output termination sequence>
196
Intelligent Platform Management Interface Specification
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
(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.
197
Intelligent Platform Management Interface Specification
14.7.9 Terminal Mode Text Command and IPMI Message Examples
The following table presents some examples of terminal mode commands and IPMI messages.
Table 14-14, 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
14.8 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.
198
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.
14.9 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 6.14, Message 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.
14.10 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.
14.11 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.
199
Intelligent Platform Management Interface Specification
14.11.1 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
14.11.2 Dialing Digits
Per [TIA-602] the dialing digits consist of the ASCII characters 0..9, *, #, A, B, C, and D.
14.11.3 <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.
14.11.4 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.
14.11.5 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.
14.11.6 ‘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.
200
Intelligent Platform Management Interface Specification
14.12 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)
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.
201
Intelligent Platform Management Interface Specification
Appendix F - TAP Flow Summary, presents an additional overview and implementation notes for TAP paging via
a BMC.
14.12.1 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 14-15, 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.
14.12.2 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.
14.12.3 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 Command as an aid to TAP
connection setup and debugging.
14.12.4 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 14-16, TAP Success Codes
code
211
213
202
TAP Definition
Page(s) Sent Successfully
Message accepted - held for deferred delivery.
Intelligent Platform Management Interface Specification
14.13 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.
203
Intelligent Platform Management Interface Specification
15. Serial Over LAN
Serial over LAN (SOL) is the specification of packet formats and protocols for transmitting serial data over LAN
using IPMI over LAN packets. The typical goal of this capability is to redirect the traffic to/from a local
asynchronous serial controller interface. This enables communication over LAN with local software that only
understands how to communicate through a local serial controller. This can be used for implementing a virtual
remote serial terminal for enabling the user or remote software interaction with serial-based interfaces for operating
systems (e.g. “command-line” interfaces and Linux console) and management services (e.g. Microsoft’s serialbased Emergency Management Services (EMS)).
15.1 System Serial Controller Requirements
The IPMI specifications do not set any mandatory requirement on the system-side interface for the serial
controller used with Serial over LAN. A “16550” serial controller register set interface is expected to be the most
common implementation. Note that other specifications, such as EMS, may have requirements for using a 16550
or 16450 register set). It is required that the system serial controller makes the functions of the RS-232 serial
hardware handshake signals (RTS, CTS, DCD/DSR, DTR) available to the BMC. These do not need to be
physical external signals as long as the BMC has the ability to perform the same flow control and set the same
serial controller status that would be set as if the serial controller were connected to an external terminal via RS232.
15.2 SOL and Serial Port Sharing
Serial-over-LAN has been designed to be able to work alongside the Serial Port Sharing capability of IPMI. This
supports an implementation where traffic to/from the serial controller interface on the baseboard can be routed either
to the BMC, SOL, or to the system’s serial port connector. To accomplish this, the specifications define commands to
control the routing of the serial stream using the serial multiplexer logic, also referred to as the ‘mux’. Table 15-1, Mux
Settings describes the basic connections that support Serial-over-LAN. When Serial Port Sharing is used with SOL,
the mux has three settings:
Table 15-1, Mux Settings
Setting
System
Description
The mux connects the serial connector on the chassis to the system serial controller. The RxD line is
routed so the BMC can snoop incoming serial traffic for escape sequences and character patterns.
BMC
The mux connects the serial connector on the chassis to the BMC. The mux circuitry also places the
hardware handshake lines to the baseboard serial controller into a steady state.
SOL
The mux is set to connect the baseboard serial controller directly to the BMC.
Note that there is no requirement that SOL is used with serial port sharing. SOL can be implemented with a dedicated
serial controller interface where the interface only provides traffic for SOL connections. Serial Port Sharing can be
implemented on a separate port if desired, or not at all.
The IPMI specifications do not specify the hardware design and implementation of SOL or Serial Port Sharing. Figure
15-1, SOL with Serial Port Sharing, presents an example block diagram for the purposes of illustrating the concept of
SOL when used with Serial Port Sharing.
The example figure shows the signal routing when the mux is set to ‘SOL’. The bold lines represent the flow of data.
The interconnections and blocks shown are to illustrate the functional relationships between the system management
elements, and do not map directly to the exact circuit implementation of the architecture.
When the mux is set to ‘SOL’, the serial connection at the back of the box is isolated from both the baseboard serial
controller and the BMC and the serial connector cannot be used for communicating with the BMC or the system.
204
Intelligent Platform Management Interface Specification
However, when SOL is not in use, the BMC can allow the serial connector to be used for functions such as the
communicating with the BMC via IPMI over Serial, or as a regular serial port.
Refer to Section 15.12, SOL Interaction with Windows.NET Escape Sequences, and Section 15.13, SOL Payload
Activated with Serial Port Sharing for additional information on the interaction between SOL and Serial Port Sharing.
Figure 15-1, SOL with Serial Port Sharing
TxD
RxD
Control
MODEM
Serial Port Connector
(E.g. 'COM2')
Serial Port Transceivers
LAN
control
switching
LAN Traffic
to/from BMC
BMC
TxD
TxD
RxD
TxD
control
switching
UART
SMBus
BMC
RxD
RxD
Control
LAN Traffic
to/from System
Baseboard
Management
Controller
(BMC)
Packet
Routing
Network Interface Chip
(LAN Controller)
System Interface
System Bus
System Bus
Baseboard Serial Controller
System Bus
(e.g. ISA or PCI)
15.3 SOL Operation Overview
SOL operation is conceptually straightforward. A remote management application can establish an IPMI-overLAN session with the BMC. Once the session is established, the remote console can request that the SOL be
activated. If SOL is used with Serial Port Sharing, this causes the BMC to set the mux to ‘SOL’.
From this point, any outgoing characters from the baseboard serial controller are assembled into packets by the
BMC and sent to the remote console over the LAN. Conversely, in-bound LAN packets carrying characters for
205
Intelligent Platform Management Interface Specification
the system serial controller have their character data extracted by the BMC and delivered to the baseboard serial
controller.
The SOL character data is carried as SOL Messages that are carried in UDP datagrams. The packet format is that
for IPMI v2.0 RMCP+ with the Payload Type set to “SOL”. The SOL Payload includes fields that are used for
supporting acknowledge and retries of SOL messages, and for supporting functions such as flushing buffers or
temporarily suspending serial traffic using flow control.
15.4 SOL Security
Authentication and encryption for SOL are handled at the RMCP+ level.
15.5 SOL Sequence Numbers
At the session level, SOL Payloads share the session sequence numbers for authenticated and unauthenticated
packets with other packets under the IPMI session. At the payload level, SOL packets include their own message
sequence numbers that are used for tracking missing and retried SOL messages.
15.6 Flow Control
Flow Control is used to help ensure that serial data is not lost because of differences in the throughput between
the serial controller interface and the network interface. When SOL is used with a physical serial controller, flow
control on the serial controller side is accomplished by the BMC controlling and monitoring the hardware
handshake signal lines (RTS, CTS, DCD/DSR, DTR) on the serial controller.
Some implementations may have the serial controller function integrated into the BMC or another device. Such
implementations may not have physical hardware flow control lines, but there must be internal control capabilities
that accomplish the same thing.
Flow control on the network side is handled by use of acknowledges (ACKs) and negative-acknowledges
(NACKs) that indicate whether the BMC is ready to accept more data or not.
15.7 Bit Rate Handling
Some implementations of SOL will connect pins for a serial interface on the BMC to the pins for the serial
interface of a serial controller for the system. For standard microcontroller serial interfaces, the BMC would need
to know the bit rate setting of the motherboard serial controller in order for the BMC’s serial controller to be in
synch. Thus, there is a non-volatile configuration setting for setting the bit rate for SOL in the BMC.
Other implementations may incorporate an embedded serial controller in the BMC or may have hardware
mechanisms that allow the BMC to get the bit rate setting directly from the system serial controller. In this case,
the non-volatile configuration bit rate setting is not used.
15.8 Volatile and Non-volatile SOL Configuration Parameters
SOL configuration parameters may be volatile, non-volatile, or have both volatile and non-volatile settings.
Unless otherwise specified, the volatile settings are copied from the non-volatile settings when the BMC first
initializes. Subsequently, the volatile settings are restored whenever the payload is first activated.
Unless otherwise specified, changes to volatile parameters take effect immediately (within normal command
processing time, typically ~30 milliseconds) and for the duration that the payload is activated. For example,
206
Intelligent Platform Management Interface Specification
changing the bit rate of SOL using the non-volatile setting will cause the BMC to immediately change its bit rate
setting.
It is desirable that some volatile settings, such as bit rate, take effect before communication proceeds on the
channel. To help support this, the Activate Payload command includes an auxiliary parameter that enables the
remote console to direct the BMC to leave CTS and DCD/DSR deasserted after the payload has been activated.
Assuming the baseboard serial controller is paying attention to hardware handshake, this enables the remote
console to hold off character transmission from the baseboard until it has changed volatile settings.
15.9 SOL Payload Data Format
Table 15-2, SOL Payload Data Format, specifies the fields that make up the SOL Payload in an RMCP+ packet.
Table 15-2, SOL Payload Data Format
Field
Packet Sequence
Number
Size
1
Packet ACK /NACK
Sequence Number
1
Accepted Character
count
1
Description
Sequence Number for this packet.
Sequence numbers must be non-zero. Multiple outstanding sequence numbers are
not supported in this version of the specification. Retried packets use the same
sequence number as the first packet.
[7:4]
Reserved
[3:0]
Packet sequence number. 0h = ACK-Only packet.
Sequence Number for packet being ACK’d or NACK’d.
[7:4]
Reserved. Write as 0h.
(Future spec may use this to specify a range of packets being acknowledged)
[3:0]: Packet sequence number being ACK’d/NACK’d.
0h: Informational packet. No request packet being ACK’d or NACK’d.
Accepted Character Count 1-based.
This field indicates the number of characters accepted from the packet, if any.
00h = Packet received but no data bytes accepted.
U
U
For BMC-to-Console:
In order to improve throughput, the BMC is allowed to append additional characters
to a packet when it resends it. To support this, the remote console must accept
retried packets (packets with the same packet sequence number) and check to see if
the packet contains additional data. If the packet does contain addition data, the
remote console should accept the data and acknowledge the packet using the packet
sequence number and return the count of the number of characters that it had
received.
The console must either accept all packet data sent to it or NACK the entire packet. It
is not allowed to accept partial packet data.
For Console-to-BMC:
The BMC is allowed to accept partial packet data by NACK’ing the packet and
returning a character offset that is less than the data length sent by the remote
console. The remote console would then send the next packet starting with the first
byte of data that the BMC rejected. Retried packets from the remote console are
unchanged from the original packet. The remote console is not allowed to append
additional data to retried packets. This eliminates the need for the BMC to check the
content of packets with duplicate packet sequence numbers.
207
Intelligent Platform Management Interface Specification
Field
Operation / Status
208
Size
1
Description
BMC to Remote Console:
Remote Console to BMC:
Note: Operations are executed before
Operations are executed before
character data is transferred.
character data is transferred.
[7] reserved
[7] reserved
[6] ACK/NACK
[6]
1b: NACK. Packet is being NACK’d by
1b: Packet is being NACK’d. The
the remote console.
BMC is unable to accept all
character data from packet.
0b: ACK. Packet is being ACK’d by the
Note: Operation field is still
remote console.
accepted even if packet is
[5]
Ring/WOR
NACK’d.
Assert RI (may not be supported on all
0b: ACK. BMC ready to accept next
implementations) - Goal is to allow this
packet of character data.
to be used for generating a WOR.
[1]
[5]
[4] Break
A NACK packet with this status will
1b: Generate BREAK (300 ms,
be automatically sent one time after
nominal)
this bit changes state. (Whenever
[3]
CTS Pause
the system enters or leaves a power
state where character transfers to
1b: Deassert CTS (clear to send) to the
the system serial controller are
baseboard serial controller. This is
possible)
the default state when SOL is
deactivated.
A NACK packet with “Character
transfer is unavailable” status will
0b: Let BMC control CTS.
also be sent for each character
[2] Drop DCD/DSR
transfer request from the remote
1b: Deassert DCD/DSR to baseboard
console when the system is in a
serial controller
powered-down or sleep state.
0b:
Assert
DCD/DSR to baseboard
1b: Character transfer is unavailable
serial
controller.
because system is in a powereddown or sleep state.
[1] Flush Inbound
0b: SOL character transfer is
1b: Drop (flush) data from remote
available.
console to BMC [not including data
[2]
carried in this packet, if any]
[4]
[0]
Flush
Outbound
A NACK packet with this status will
be automatically sent one sent once,
1b: Flush Outbound Character Data
just before the BMC deactivates SOL
(flush data from BMC to remote
because of a front panel powerconsole)
button or a reset.
1b: SOL is deactivated/deactivating.
[Remote console can use this to
tell if SOL was deactivated by
some other party, or by local
pushbutton reset or power on/off].
0b: SOL is active.
[3] Transmit Overrun
1b: characters were dropped
between transmitting this packet
and the previous packet, because
the system did not pay attention
to hardware flow control.
0b: no characters were lost between
this packet and the preceding
packet.
[2] Break
1b: A break condition from the
P
P
P
P
Intelligent Platform Management Interface Specification
Field
Character Data
Size
Description
system has been detected. The
BMC will generate this only on
one packet at the start of the
break.
0b: no break detected
[1:0] Reserved
varies
Data length, in bytes, is equal to the IPMI Message/SOL Message Length field in the
Session Header, minus the bytes for the Packet Sequence Number, through and
including the Operation Field.
1.
If the system is powered down or in certain sleep states, the baseboard serial controller will not be available for transferring
characters. When the BMC receives data from the remote console, and the system is powered down or sleeping, the BMC can
use this bit to return status to indicate to the remote console why the characters it received may not be able to be transferred to
the system. Note that this is an ‘advisory’ bit. The BMC will still attempt to put characters into the system serial controller in
case that the serial controller is configured to wake the system under this condition. It is mandatory that the BMC returns this bit
when the system is powered down (S4/S5). The BMC may not be able to differentiate other system sleep states, in which case
the bit should be returned as 0b (transfers available). If the SOL payload is launched over a separate or dedicated session, the
device managing that session may not be able to tell when the system is powered down. In this case, the ‘character transfer is
unavailable’ function of this bit is optional, but recommended.
2.
The BMC issues this status when the payload is deactivated for due to the following conditions: a. The payload is deactivated
because of a manual power down or reset. (An implementation is recommended to have a local manual “pushbutton” reset or
power-off to deactivate an SOL payload. This is provided as a way of terminating remote control connections for local
servicing.) b. The SOL payload is deactivated via the Deactivate Payload or Close Session commands. These command may
have been issued from another session. (System software operating through the system interface, and users with Admin
privilege have the ability to issue the Deactivate Payload and Close Session commands to other sessions). If the SOL payload
is launched over a separate or dedicated session, the device managing that session may not be able to tell whether the system
is being locally powered on/off or reset. In this case, the ‘SOL de-activating’ function of this bit on local power and reset
transitions is optional, but recommended.
15.10 Activating SOL using RMCP+ Authentication
To use SOL a remote console or remote application must first establish an IPMI Session with the BMC. This is
accomplished by sending the specified IPMI and RMCP+ RMCP+/RAKP messages to the BMC with the
appropriate user name, role, and password/key information. If the remote console plans to use encryption with
SOL, the console must also negotiate an encryption algorithm at the time that the IPMI session is established.
Once the IPMI session has been established, the Get Channel Payload Support command can be used to retrieve
the present availability of SOL and the version of SOL.
If SOL is not already active on another session, the next step is to issue the Activate Payload command. The
command will return information about what serial input and output data buffer sizes are available on the BMC as
well as the UDP port number over which SOL packets can be transferred.
The port that was used for establishing the IPMI session may not be the same port number that SOL is available
over. Some implementations transfer SOL payloads available over a separate UDP port in order to provide better
performance. If the Activate Payload command returns a port number that is different than the port number that
was used to establish the IPMI session, the remote console must establish a separate IPMI Session to the specified
port number using the same IP Address, username and password/key information that was used to establish the
IPMI session.
Note that if a second session has already been established on that port number for a different payload type, the
existing session can also be used for SOL payloads, provided that the session was established at a privilege level
that matches the privilege level and authentication required for SOL. Otherwise, the remote console will need to
close that session and re-establish it at the necessary privilege level.
15.11 SOL Packet Acknowledge and Retries
A packet acknowledge is of one of two types:
209
Intelligent Platform Management Interface Specification
• An ACK, indicating that the packet has been received and all its data has been accepted
• A NACK, indicating that the packet was received but some or all of the data could not be accepted
To improve efficiency, the packet acknowledgment information can be carried in a packet that also carries the
SOL character data. Conversely, a packet can be an ACK-only packet that carries ACK or NACK information,
but no data. However, packets are only acknowledged if they contain character data. Therefore, ACK-only
packets themselves are not acknowledged.
Except for ACK-only packets, the remote console must acknowledge each SOL packet that it receives. If the
BMC does not receive an ACK packet within a timeout interval, the BMC will resend (retry) the packet. The
number of retries and the amount of time between retries are configurable through the SOL Configuration
Parameters. Once the number of retries has been met the BMC will drop the packet and the data will be lost.
If the remote console wants to temporarily stop the BMC from accepting characters from the system, it should use
the “CTS Pause” bit in the control/status byte. Whether this stops the system from transmitting is dependent on
whether the system software pays attention to the CTS (hardware handshake) status or not. If the system
continues to send characters to the BMC, the BMC will attempt to transmit them to the remote console.
It is possible that additional characters could be received from the system serial controller while the BMC was
waiting for a retry timeout. In order to improve throughput, the BMC is allowed the option of appending
additional characters to a packet whenever it resends it. To support this, the remote console must accept retried
packets (packets with the same packet sequence number) and check to see if the packet contains additional data.
If the packet does contain addition data, the remote console should accept the data and acknowledge the packet
using the packet sequence number and character offset value that it had received.
Table 15-3, Remote Console to BMC SOL Packet Handling
ACK/
NACK
ACK
ACK
NACK
Packet fields from remote console:
ACK’d/NACK’d
Accepted
Seq#
Character Count
Seq # from BMC data
Non-zero
packet
matches character
count from BMC
Seq # from BMC data
packet
Seq # from BMC data
packet
Non-zero
less than character
count from BMC
0
Function / BMC Action
“Completion ACK”
BMC processing for specified packet is done.
A packet from the remote console with an ACK and an
Accepted Character Count for the full amount of data
for the packet indicates that the remote console has
successfully accepted the packet.
“Partial ACK”
BMC immediately retransmits specified packet
It is possible that the remote console may have missed
a BMC retry where the BMC had appended more data
to the packet (retry intervals should be configured to
avoid this scenario). If the BMC receives an ACK for
less than the last amount of transmitted data, the BMC
will cease appending data to the packet and will
retransmit the packet until it receives an ACK from the
remote console with an Accepted Character Count for
the full amount of packet data.
“Suspend NACK”
BMC stops sending specified packet.
The remote console would use the Suspend NACK if it
were temporarily out of buffer space for characters
already queued up in the BMC and did not want those
characters to get dropped. The BMC stops sending /
retrying specified packet and waits until it gets either a
“Partial ACK”, “Completion ACK”, “Resume ACK” or a
Flush Outbound operation from the remote console.
The BMC will deassert CTS when it gets near running
out of buffer space for characters from the system. If
210
Intelligent Platform Management Interface Specification
ACK
ACK
Seq # from BMC data
packet
0h
0
0
characters continue to come in (CTS is ignored by the
system) a transmit overrun condition can occur.
“Resume ACK”
BMC immediately retransmits specified packet.
Cancels a “Suspend NACK”.
A packet from the remote console with an ACK and an
Accepted Character Count of zero (0) bytes will cause
the BMC to immediately retransmit the packet with the
corresponding sequence number to the remote
console. This can be used as a way to get the BMC to
restart transmission after a Suspend NACK from the
remote console.
“Control-only Packet”
BMC performs operation specified in the control/status
field.
For FLUSH operation:
A packet from the remote console with a “Flush
Outbound” operation will cause the BMC to cease any
retries in progress and the BMC will start accumulating
character data anew. The remote console can use this
as a recovery mechanism if it gets ‘lost’ in the
sequence from the BMC.
See packet format table for info on other functions.
15.12 SOL Interaction with Windows.NET Escape Sequences
The Microsoft .NET Emergency Management Services specification (See [MSFT EMS]) defines certain character
sequences for performing the following operations:
System Hard Reset
Invoke Service Processor (i.e. switch to BMC)
Exit Service Processor (with optional prompt)
Exit Service Processor (without confirmation prompt)
<ESC>R<ESC>r<ESC>R
<ESC>(
Q
<ESC>Q
The specification also requires that a switch to the Service Processor is acknowledged by sending an <ESC>*
sequence to the remote console:
Acknowledge Switch to Service Processor
<ESC>*
Typically, the input escape sequences would be received by the BMC over the serial/modem connection. In this
case, the sequences for Invoking and Exiting the service processor would control the serial/modem mux setting
associated with serial port sharing. However, there are separate streams for SOL and BMC traffic, so unlike serial
port sharing there is no need for a mux to switch between console traffic to the system serial controller and traffic
to the BMC.
Therefore, since you can always send commands to the BMC as IPMI Messages, the BMC itself does not snoop
for or handle Windows .NET Escape sequences in the SOL character data. However, a remote console
application can emulate support for the Windows .NET headless sequences by filtering for the Windows .NET
escape sequences prior to placing the data in an SOL packet. If the Windows .NET escape sequences are detected,
the remote console can then take the appropriate actions. For example, if the Reset escape sequence is detected,
the remote console would send the IPMI command for a system reset.
211
Intelligent Platform Management Interface Specification
15.13 SOL Payload Activated with Serial Port Sharing
The following lists the behavior of certain IPMI commands while SOL is activated. This applies only when the
SOL is being used in conjunction with Serial Port Sharing.
•
The Set Channel Access command is accepted for the serial channel, but while SOL is activated the
Channel Access Mode is forced to ‘disabled’. Attempting to change the access mode while SOL is
activated will result in a D5h completion code. The access mode shall be saved when SOL is activated, and
restored when SOL is deactivated.
•
The Set Serial/Modem Configuration Parameters and Get Serial/Modem Configuration Parameters
commands are accepted for the serial/modem channel and have no special behavior while SOL is activated.
•
The UDP proxy commands, Set PPP UDP Proxy Transmit Data, Get PPP UDP Proxy Transmit Data,
Send PPP UDP Proxy Packet, Get PPP UDP Proxy Receive Data, will receive a D5h error completion
code while SOL is activated.
•
The Callback command will receive a D5h error completion code when the command is targeted to the
serial/modem channel being used for SOL and SOL is activated. The Set User Callback Options and Get
User Callback Options commands will be accepted.
•
The Send Message Command will receive a D5h error completion code when the command is targeted to
the serial/modem channel being used for SOL and SOL is activated.
•
The Alert Immediate Command will receive a D5h error completion code when the command is targeted to
the serial/modem channel being used for SOL and SOL is activated.
•
While SOL is activated, the Set Serial/Modem Mux command will respond to the requested operations as
follows:
Table 15-4, Set Serial/Modem Mux Command Operation while SOL Active
212
Operation
0h = get present mux setting/status only
Response
Accepted. BMC returns that mux is set to ‘System’
1h = request switch of mux to system
Accepted
2h = request switch of mux to BMC
Rejected (see response data for command)
3h = force switch of mux to system
Allowed
4h = force switch of mux to BMC
BMC returns D5h completion code
5h = block requests to switch mux to system
BMC returns D5h completion code
6h = allow requests to switch mux to system
Accepted
7h = block requests to switch mux to BMC
BMC returns D5h completion code
8h = allow requests to switch mux to BMC
Accepted
Intelligent Platform Management Interface Specification
16. 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.
16.1 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.
213
Intelligent Platform Management Interface Specification
IPMB Events
PCI Mgmt. Bus
events
System Interface
SMS Events
BIOS Events
Table 16-1, 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.
16.2 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.
214
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.
16.3 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.
215
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.
16.4 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 42.1, Event/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 16.6, Rearming, below.
16.5 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
216
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.
16.6 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. A reading/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.
16.6.1 ‘Global’ Re-arm
A device that receives a Set Event Receiver command shall ‘re-arm’ event generation for all its internal sensors.
217
Intelligent Platform Management Interface Specification
218
Intelligent Platform Management Interface Specification
17. ‘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.
17.1 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.
17.2 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.
17.3 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.
219
Intelligent Platform Management Interface Specification
17.4 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 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.
17.4.1 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.
17.5 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.
220
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.
17.6 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 17-1, 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
17.11, Alert Policy Table.
(optional) Priority determined by OEM.
17.7 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.
221
Intelligent Platform Management Interface Specification
Table 17-2, 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
222
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] 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] 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
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
Channel Number / LUN to match. FFh = match any see section 32, SEL
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 17.8, Event 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 17.8, Event 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 17.8, Event 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.
U
U
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
223
Intelligent Platform Management Interface Specification
17.8 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.
17.9 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 17-3, 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
17.10 Mask and Compare Field Examples
The following examples show how the fields are used. See
224
Intelligent Platform Management Interface Specification
Appendix B - Example PEF Mask Compare Algorithm for example matching algorithm.
Example 1:
AND Mask:
0000 0110
Compare 1:
Compare 2:
1111 1111
0000 0110
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
U
U
U
U
U
U
U
U
U
U
U
U
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.
17.11 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.
225
Intelligent Platform Management Interface Specification
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.
Table 17-4, 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.
226
Intelligent Platform Management Interface Specification
17.12 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.
17.13 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).
17.13.1 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.
17.13.2 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.
17.13.3 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.
227
Intelligent Platform Management Interface Specification
17.13.4 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.
•
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.
17.13.5 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 17-5, 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:
228
•
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.
Intelligent Platform Management Interface Specification
•
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.
•
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 14.10, Page Blackout Interval.
17.14 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.
229
Intelligent Platform Management Interface Specification
Figure 17-1, 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
Policy Number
Event Filter Table
Alert String Key
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
17.15 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.
230
Intelligent Platform Management Interface Specification
Figure 17-2, 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"
17.16 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 17-6, PET Specific Trap Fields
PET Field
Event Sensor Type
Event Type
Event Offset
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
231
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 17-7 - 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
232
byte
00_0001b
00_0010b
00_0100b
00_1000b a.k.a. ‘warning’
01_0000b
10_0000b
In IPMI this holds an ID (I2C 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.
P
P
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.
17.16.1 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 17-8, 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
17.17 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
233
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.
234
Intelligent Platform Management Interface Specification
18. Firmware Firewall & Command Discovery
IPMI v1.5 and earlier specifications allow almost any supported IPMI operation to be accomplished from the
system interface. This includes writing SDRs, configuring alerts, setting thresholds, sending messages to other
media, and other actions that could be used to ‘spoof’ a non-existent system failure. In a standalone system, this is
normally an acceptable risk since errant local software would typically only affect the system it was running on.
Some blade chassis implementations have a central management module (CMM) that aggregates and acts on
information from BMCs that reside on individul compute blades. In this environment, a malicious piece of
software could potentially spoof a hazardous voltage or thermal condition that causes the CMM to shutdown the
entire chassis. Or, if a shared management bus is used, software running on one blade could potentially send IPMI
messages that could shut down, reset, or reconfigure management on other blades.
However, it takes more than just blocking those types of commands or operations. The IPMI specifications
needed to provide a way to tell software that commands that would otherwise be mandatory have been
intentionlly made unavailable for protection purposes and are not unavailable because of an error.
Configuration of Firmware Firewall capabilities is supported by commands that allow software to enable/disable
individual commands and command sub-functions, and to discover which particular commands and command
sub-functions can be configured on a given implementation. These ‘command support discovery’ commands can
implemented without requiring the enable/disable commands of Firmware Firewall in order to provide a common
mechanism for discovering which IPMI commands and functions a given management controller supports.
Firmware Firewall capabilities can be extended to include interfaces (channels) other than the system interface,
such as the IPMB. This can be used to prevent add-in cards or other parties from performing certain IPMI
functions.
The Firmware Firewall capability does not affect the operation of user and channel privileges. That is, if a
command requires Admin privilege level to be executed, it will still require Admin privilege if enabled by
Firmware Firewall.
The Firmware Firewall and command discovery commands include the following:
1.
Commands for the general discovery of supported commands and sub-functions on a given interface a.k.a. “command discovery commands”:
Command
Get NetFn Support
Get Command Support
Get Command Sub-function Support
2.
Commands to discovery what commands and functions can be enabled/disabled - a.k.a. “configurable
command discovery” commands:
Command
Get Configurable Commands
Get Configurable Command Sub-functions
3.
Section
21.2
21.3
21.4
Section
21.5
21.6
Commands that are used to enable/disable any configurable commands or sub-functions:
Command
Set Command Enables
Get Command Enables
Set Command Sub-function Enables
Get Command Sub-function Enables
Section
21.7
21.8
21.9
21.10
235
Intelligent Platform Management Interface Specification
236
Intelligent Platform Management Interface Specification
19. Command Specification Information
This section provides specifications for elements that apply to all requests and responses presented later in this
document.
19.1 Specification of Completion Codes
Completion codes are specified in Section 5.2, Completion Codes. Additional command-specific 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.
19.2 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.
19.3 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.
19.4 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
237
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.
238
Intelligent Platform Management Interface Specification
20. 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
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands. O/M = Optional/Mandatory.
Table 20-1, 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
20.1
20.2
20.3
20.4
20.5
20.6
20.7
20.8
M
O[1]
O
M
O
O
O[3]
O
20.9
M[2]
P
P
P
P
Broadcast Commands
Broadcast ‘Get Device ID’
[1]
[2]
[3]
P
P
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.
239
Intelligent Platform Management Interface Specification
20.1 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 20-2, 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
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.
U
7
8:10
11:12
240
U
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 20.8, Get 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
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.
U
U
241
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.
U
U
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
242
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.
Intelligent Platform Management Interface Specification
20.2 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 20-3, 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.
20.3 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 20-4, Warm Reset Command
Request Data
Response Data
byte
1
data field
Completion Code
243
Intelligent Platform Management Interface Specification
20.4 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 20-5, 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
20.5 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.
244
Intelligent Platform Management Interface Specification
Table 20-6, Manufacturing Test On
Request Data
Response Data
byte
1:N
1
data field
device specific parameters. See text.
Completion Code
20.6 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.
245
Intelligent Platform Management Interface Specification
Table 20-7, Set ACPI Power State Command
Request Data
byte
1
2
Response Data
246
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.
Intelligent Platform Management Interface Specification
20.7 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 20-8, Get ACPI Power State Command
Request Data
Response Data
byte
1
2
3
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.
247
Intelligent Platform Management Interface Specification
20.8 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 20-9, 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 20-10, GUID Format
GUID byte
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
248
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
Intelligent Platform Management Interface Specification
20.9 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 20-1, 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.
P
P
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.
249
Intelligent Platform Management Interface Specification
21. Firmware Firewall & Command Discovery
Commands
The following sections provide the specifications for the commands that support the Firmware Firewall capability.
Table 21-1, Firmware Firewall Commands
Command
Get NetFn Support
Get Command Support
Get Command Sub-function Support
Get Configurable Commands
Get Configurable Command Sub-functions
Set Command Enables
Get Command Enables
Set Command Sub-function Enables
Get Command Sub-function Enables
1.
2.
3.
Section
Defined
O/M
21.1
21.3
21.4
21.5
21.6
21.7
21.8
21.9
21.10
O[1,3]
O[1,3]
O[1,3]
O[2]
O[2]
O
O[2]
O[2]
O[2]
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
Mandatory on any channel/interface to the BMC on which a typically mandatory
command can be or is disabled for firmware firewall purposes.
Mandatory on any channel/interface to the BMC on which the Set Command
Enables command is implemented. The Set Command Enables, Get Command
Enables, Set Command Sub-function Enables, and Get Command Sub-function
Enables commands must be implemented as a set.
The Get NetFn Support, Get Command Support, and Get Command Sub-function
Support commands must be implemented as a set.
21.1 Completion Codes with Firmware Firewall
When Firmware Firewall is used, the “D4h” completion code should be returned for any commands that are not
available because of a security-based restriction. Commands that would normally be available (mandatory
commands, or optional commands that are mandatory because they are required by another command or as part of
an optional feature) but are un-available because they have been disabled via the Set Command Enables command
should return D4h if an attempt is made to execute them. Similarly, a new completion code, “D6h”, can be used
to indicate that a normally available command sub-function cannot be configured due to a security-based
restriction.
250
Intelligent Platform Management Interface Specification
21.2 Get NetFn Support Command
This command returns which NetFn and LUNs support commands on a given channel. Since command support
can vary by channel, the Channel Number parameter is part of the request.
Table 21-2, Get NetFn Support Command
IPMI Request Data
1
IPMI Response Data
1
2
3:18
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
Completion Code
LUN support
[7:6] - LUN 3 (11b) support
00b = no commands supported on LUN 3 (11b)
01b = commands follow base IPMI specification. Commands exist on
LUN, but no special restriction of command functions.
Comands follow standard Optional/Mandatory specifications.
10b = commands exist on LUN, but some commands/operations may
be restricted by firewall configuration.
11b = reserved
[5:4] - LUN 2 (10b) support
Note that a BMC uses LUN 10b for message bridging. The message
bridging capability is enabled/disabled by enabling/disabling the Send
Message command.
00b = no commands supported on LUN 2 (10b)
01b = commands follow base IPMI specification. Commands exist on
LUN, but no special restriction of command functions.
Comands follow standard Optional/Mandatory specifications.
10b = commands exist on LUN, but some commands/operations may
be restricted by firewall configuration.
11b = reserved
[3:2] - LUN 1 (01b) support
[1:0] - LUN 0 (00b) support
There are 32 possible Network Function (NetFn) pairs. The following bytes are
treated as bitfields where each bit indicates the support for a given Network
Function pair. Thus, it takes 4 bytes to fully list support for NetFn values under
a given LUN. Since there are four possible LUNs for a management controller,
a total of 16 bytes will return the settings for all four possible LUNs. 0b =
NetFn pair is not used, 1b = NetFn pair is used
byte 1, bit 0 corresponds to NetFn pair 0h,1h for LUN 00b
byte 1, bit 7 corresponds to NetFn pair Eh,Fh for LUN 00b
…
byte 16, bit 0 corresponds to NetFn pair 30h, 31h for LUN 11b
byte 16, bit 7 corresponds to NetFn pair 3Eh, 3Fh for LUN 11b
251
Intelligent Platform Management Interface Specification
21.3 Get Command Support Command
This command provides a way to get a list of which command values are available on a given channel for a given
Network Function and LUN. System software can iterate using this command to determine what commands are
supported on a given management controller. The command returns information for 128 command values at a
time. Thus, two iterations of the command are required to get all information for a given channel/NetFn/LUN
combination.
Table 21-3, Get Command Support Command
IPMI Request Data
1
2
3
IPMI Response Data
1
2:17
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - Operation
00b = return support mask for commands 00h through 7Fh.
01b = return support mask for commands 80h through FFh.
10b, 11b = reserved.
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
Completion Code
Support Mask
These sixteen bytes form a 128-bit bitfield where each bit indicates
support for a particular command value under the given NetFn.
For each bit in the bitfield:
0b = indicates the command is available
1b = indicates the command not available
Depending on the value of the “Operation” parameter passed in the
request:
byte 1, bit 0 corresponds to command 00h or command 80h
byte 1, bit 7 corresponds to command 07h or command 87h
…
byte 16, bit 0 correspond to command 78h or command F8h
byte 16, bit 7 corresponds to command 7Fh or command FFh
252
Intelligent Platform Management Interface Specification
21.4 Get Command Sub-function Support Command
This command is used to determine what sub-functions of a given command are presently available on the given
channel. The command can also be used to find out specifically which version of the specification was used to
define the command operation. The latest version of the specification that the command conforms to is what is
used. For example, a command that is conformant with both the IPMI v1.5 and v2.0 versions of the specification
shall be identified as and IPMI v2.0 command.
Table 21-4, Get Command Sub-function Support Command
IPMI Request Data
1
2
3
4
IPMI Response Data
1
2
3
4
(5:8)
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - reserved
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] - reserved
[1:0] - LUN
[7:0] - CMD. Command number to return command sub-function information
for.
Completion Code
Specification Type / Errata
[7:4] - Specification Type
0h = IPMI, 1h = IPMB, 2h = ICMB, all other = reserved
[3:0] - Errata Version
This field returns the errata document version that was used in
defining the command’s operation. For IPMI specifications, this is the
revision number of the IPMI errata that goes with the specification
version and revision, below. The latest errata of the specification to
which the command is conformant should be used. Use 0h if there is
no errata document available at the time the command was defined.
Specification Version
This field returns the specification Version, in BCD format for which
the command was specified. Bits 7:4 hold the most significant digit of
the version, while bits 3:0 hold the least significant bits, e.g. a value of
20h indicates version 2.0. The latest version number for the
specification that the command is conformant to should be used.
Specification Revision
This field returns the specification Revision, in BCD format for which
the command was specified. Bits 7:4 hold the most significant digit of
the version, while bits 3:0 hold the least significant bits, e.g. a value of
10h indicates version 1.0. The latest revision number for the
specification that the command is conformant to should be used.
Support Mask 1 (ls-byte first)
These sixteen bits form a bitfield where each bit indicates support for
a particular sub-function for the given command. The bit offset
corresponds to the number of the sub-function.
1b
0b
indicates that a mandatory sub-function or option is unavailable.
indicates that a mandatory sub-function or option is available. 0b is
also used when a given offset is undefined. Thus, a command that
implements all functions (mandatory and optional) will return all 0’s for
the bitfield. See Table H-44-12, Sub-function Number Assignments.
[31] [30] …
[1] [0] -
bit for sub-function 31.
bit for sub-function 30.
bit for sub-function 1.
bit for sub-function 0.
253
Intelligent Platform Management Interface Specification
21.5 Get Configurable Commands Command
This command returns the IPMI commands that can be enabled/disabled via the Set Command Enables command
for a given channel/NetFn/LUN.
Table 21-5, Get Configurable Commands Command
IPMI Request Data
1
2
3
IPMI Response Data
1
2:17
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - Operation
00b = return support mask for commands 00h through 7Fh.
01b = return support mask for commands 80h through FFh.
10b, 11b = reserved.
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
Completion Code
Support Mask
These sixteen bytes form a 128-bit bitfield where each bit indicates
enable/disable support for a particular command value under the given
NetFn.
For each bit in the bitfield:
0b = indicates the command value is not configurable
1b = indicates the command can be enabled/disabled
Depending on the value of the “Set Selector” parameter passed in the
request:
byte 1, bit 0 corresponds to command 00h or command 80h
byte 1, bit 7 corresponds to command 07h or command 87h
…
byte 16, bit 0 correspond to command 78h or command F8h
byte 16, bit 7 corresponds to command 7Fh or command FFh
254
Intelligent Platform Management Interface Specification
21.6 Get Configurable Command Sub-functions Command
This command enables software to discover which command sub-functions can be enabled/disabled via the Set
Command Sub-function Enables command.
Table 21-6, Get Configurable Command Sub-functions Command
IPMI Request Data
1
2
3
4
IPMI Response Data
1
2:5
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - reserved
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
[7:0] - CMD. Command number to return command sub-function information
for.
Completion Code
Support Mask (ls-byte first)
These sixteen bits form a bitfield where each bit indicates support for a
particular sub-function for the given command. The bit offset
corresponds to the number of the sub-function. See Table H-44-12, Subfunction Number Assignments.
1b
0b
[31] [30] …
[1] [0] -
indicates that the sub-function can be enabled/disabled.
indicates that the sub-function is not configurable, or is unavailable. 0b is
also used for unspecified/reserved sub-function numbers.
bit for sub-function 31.
bit for sub-function 30.
bit for sub-function 1.
bit for sub-function 0.
255
Intelligent Platform Management Interface Specification
21.7 Set Command Enables Command
This command enables software to enable/disable commands for a given channel/netFn/LUN. The command sets
the enables/disables for a large group of commands simultaneously. Therefore, software must perform a readmodify-write operation to change a single command setting or any subset of the group. This can be accomplished
by using the Get Command Enables command to get the present setting, then ‘OR-ing’ or ‘AND-ing’ in the
desired change, and using the Set Command Enables command to set the change into the management controller.
It is highly recommended that the implementation takes steps to prevent the Set Command Enables command
from being used to disable itself. The Set Command Enables command should always be an ‘un-configurable’
command on at least one channel into the BMC.
Table 21-7, Set Command Enables Command
IPMI Request Data
1
2
3
4:18
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - Operation. The enable/disable settings are non-volatile. The
management controller must reject all new settings (must not change
present settings) if there is any error in the command (non-zero
completion code returned).
00b = Set enable/disables for commands 00h through 7Fh.
01b = Set enables/disables for commands 80h through FFh.
10b, 11b = reserved.
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
Enable/Disable Mask
These sixteen bytes form a 128-bit bitfield where each bit controls the
enable/disable of a particular command value under the given NetFn.
For each bit in the bitfield:
0b = disables the command
1b = enables the command
Note that if a bit position corresponds to a command that is not
configurable, the BMC will return an error if an attempt is made to
change the enabled/disabled state for that command. I.e. if the bit is
fixed at 0b, and error will be generated if an attempt is made to set it to
1b, and vice versa. Software can use the Get Configurable Commands
command and the Get Command Enables command together to process
the bits for this command to ensure setting the correct state.
IPMI Response Data
256
1
Depending on the value of the “Set Selector” parameter passed in the
request:
byte 1, bit 0 corresponds to command 00h or command 80h
byte 1, bit 7 corresponds to command 07h or command 87h
…
byte 16, bit 0 correspond to command 78h or command F8h
byte 16, bit 7 corresponds to command 7Fh or command FFh
Completion Code
Generic, plus following command-specific codes:
80h = attempt to enable an unsupported or un-configurable command.
Intelligent Platform Management Interface Specification
21.8 Get Command Enables Command
This command enables software to determine which commands are enabled/disabled for a given
channel/netFn/LUN.
Table 21-8, Get Command Enables Command
IPMI Request Data
1
2
3
IPMI Response Data
1
2:18
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - Operation
00b = Get enable/disables for commands 00h through 7Fh.
01b = Get enables/disables for commands 80h through FFh.
10b, 11b = reserved.
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
Completion Code
Enable/Disable Mask
These sixteen bytes form a 128-bit bitfield where each bit returns the
enable/disable of a particular command value under the given NetFn. If
a command is not supported at all, a 0b will be returned.
For each bit in the bitfield:
0b = command is disabled or not supported
1b = command is enabled
Software can use the Get Command Support command to determine
which are supported, and the Get Configurable Commands command to
determine which commands are configurable.
Depending on the value of the “Operation” parameter passed in the
request:
byte 1, bit 0 corresponds to command 00h or command 80h
byte 1, bit 7 corresponds to command 07h or command 87h
…
byte 16, bit 0 correspond to command 78h or command F8h
byte 16, bit 7 corresponds to command 7Fh or command FFh
257
Intelligent Platform Management Interface Specification
21.9 Set Configurable Command Sub-function Enables Command
This command is used for enabling/disabling configurable sub-functions for the given command.
Table 21-9, Set Configurable Command Sub-function Enables Command
IPMI Request Data
1
2
3
4
5:8
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - reserved
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
[7:0] - CMD. Command number to set command sub-function enables for.
Sub-Function Enables (ls-byte first). The enable/disable settings are nonvolatile and take effect on successful completion of the command. The
management controller must reject all new settings (must not change
present settings) if there is any error in the command (non-zero
completion code returned).
These sixteen bits form a bitfield where each bit indicates support for a
particular sub-function for the given command. The bit offset
corresponds to the number of the sub-function.
1b
0b
IPMI Response Data
258
1
enables the sub-function
disables the sub-function. 0b is also used for un-configurable/reserved
sub-function numbers. See Table H-44-12, Sub-function Number
Assignments.
[31] - bit for sub-function 31.
[30] - bit for sub-function 30.
…
[1] bit for sub-function 1.
[0] bit for sub-function 0.
Completion Code
Generic, plus following command-specific completion codes:
80h = attempt to enable an unsupported or un-configurable sub-function.
Intelligent Platform Management Interface Specification
21.10 Get Configurable Command Sub-function Enables Command
This command enables software to determine which sub-functions are enabled/disabled for a given command on
the specified channel/netFn/LUN.
Table 21-10, Get Configurable Command Sub-function Enables Command
IPMI Request Data
1
2
3
4
IPMI Response Data
1
5:6
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
[7:6] - reserved
[5:0] - NetFn. Network function code to look up command support for. The
management controller will return the same values for odd or even
NetFn values. I.e. the value for bit [0] is ignored.
[7:2] reserved
[1:0] LUN
[7:0] - CMD. Command number to set command sub-function enables for.
Completion Code
Generic, plus following command-specific completion codes:
80h = attempt to enable an unsupported or un-configurable sub-function.
Sub-Function Enables (ls-byte first)
These sixteen bits form a bitfield where each bit indicates support for a
particular sub-function for the given command. The bit offset
corresponds to the number of the sub-function. See Table H-44-12, Subfunction Number Assignments.
1b
0b
sub-function is enabled
sub-function is disabled or is un-configurable/reserved.
[15] - bit for sub-function 15.
[14] - bit for sub-function 14.
…
[1] - bit for sub-function 1.
[0] - bit for sub-function 0.
259
Intelligent Platform Management Interface Specification
22. 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
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 22-1, IPMI Messaging Support Commands
Command
Section
Defined
O/M
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 System Interface Capabilities
Get BT Interface Capabilities
Master Write-Read
Get System GUID
Get Channel Authentication Capabilities
Get Channel Cipher Suites
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 Channel Security Keys
Set User Access Command
Get User Access Command
Set User Name
Get User Name Command
Set User Password Command
22.1
22.2
22.3
22.4
22.5
22.6
22.7
22.8
22.9
22.10
22.11
22.14
22.13
22.15
22.15.1
22.17
22.18
22.19
22.20
22.21
22.22
22.23
22.24
22.25
22.26
22.27
22.28
22.29
22.30
M
M
M
M
O
M[1]
M[1]
O
O[7]
M[2]
M[6]
O[5]
O[3]
O[8]
O[4]
O[4]
O[4]
O[4]
O[4]
O
O[4]
O[4]
O[4]
O[8]
O[4]
O[4]
[5]
O
O[4]
O[4]
1.
2.
3.
260
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.
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
Intelligent Platform Management Interface Specification
4.
5.
6.
7.
8.
Mandatory if session-based channels are supported
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.
Mandatory for IPMI v2.0 or later implementations of SSIF, and for any SSIF
implementation if the BMC supports multi-part writes and reads. Recommended
but not mandatory for KCS implementations.
Mandatory if IPMI v2.0/RMCP+ session-based channels are implemented.
22.1 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 22-2, 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.
22.2 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.
261
Intelligent Platform Management Interface Specification
Table 22-3, 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.)
22.3 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 22-4, 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.
22.4 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.
262
Intelligent Platform Management Interface Specification
Table 22-5, 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
22.5 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 22-6, 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
22.6 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.
263
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.
264
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 22-7, 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.
265
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 22-8, Get Message Data Fields
1
Originating Channel Type
IPMB (I2C)
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
P
P
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 Channels
See Section 8.2, ICMB 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.
22.7 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 6.13, BMC Message Bridging, for information on
how the Send Message command is used.
For IPMI v2.0 the Send Message command has been updated to include the ability to indicate whether a
message must be sent authenticated or with encryption (for target channels on which authentication and/or
encryption are supported and configured).
266
Intelligent Platform Management Interface Specification
Table 22-9, Send Message Command
byte
Request Data
data field
1
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 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 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
[5]
1b = Send message with encryption. BMC will return an error completion code if this encryption is
unavailable.
0b = Encryption not required. The message will be sent unencrypted if that option is available
under the given session. Otherwise, the message will be sent encrypted.
[4]
1b = Send message with authentication. BMC will return an error completion code if this
authentication is unavailable.
0b = Authentication not required. Note behavior is dependent on whether authentication is used
is depending on whether the target channel is running an IPMI v1.5 or IPMI v2.0/RMCP+
session, as follows:
IPMI v1.5 sessions will default to sending the message with authentication if that option is
available for the session.
U
U
IPMI v2.0/RMCP+ sessions will send the message unauthenticated if that option is
available under the session. Otherwise, the message will be sent with authentication.
U
[3:0] channel number to send message to.
Message Data. Format dependent on target channel type. See Table 22-10, 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.
2:N
Response Data
U
1
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
U
P
P
267
Intelligent Platform Management Interface Specification
(2:N)
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.
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 22-10, Message Data for Send Message Command
1
Target Channel Type
IPMB (I2C)
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
P
P
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 Channels
See Section 8.2, ICMB Bridge Commands in BMC
using Channels
RQ: Session Handle[1], rsSWID, netFn/rsLUN, chk1,
rqSWID or rqSA, rqSeq/rqLUN, cmd, <data>,
chk2
RS: Session Handle[1], 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
P
P
P
1.
268
5
Asynch. Serial/Modem
(RS-232)
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 (Basic
Mode, Terminal
Mode, and PPP
Mode)
IPMB
IPMI-SMBus
n/a
n/a
P
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
22.8 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 22-11, 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 32-1, 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.
22.9 Get System Interface Capabilities Command
This command can be used to determine whether the SSIF supports multi-part transactions, and what size of IPMI
messages can be transferred. The Get System Interface Capabilities command is mandatory for BMCs that
implement multi-part writes or reads. Thus, software can assume that if the Get System Interface Capabilities
command is not implemented, the interface only supports single-part writes and reads.
269
Intelligent Platform Management Interface Specification
Table 22-12, Get System Interface Capabilities Command
Request Data
Response Data
byte
1
1
2
3
4
5
3
270
data field
System Interface Type
[7:4] - reserved
[3:0] - System Interface Type (For BT use the Get BT Interface Capabilities
command)
0h = SSIF
1h = KCS
2h = SMIC
all other = reserved
Completion Code
Reserved. Returned as 00h.
For System Interface Type = SSIF:
[7:6] - Transaction support
00b = only single-part reads/writes supported.
01b = multi-part reads/writes supported. Start and End
transactions only.
10b = multi-part reads/writes supported. Start, Middle, and
End transactions supported.
11b = reserved.
[5:4] - reserved.
[3] PEC support.
1b = implements PEC. BMC will start using PEC in read
transactions after it receives any SSIF write transaction
that includes a valid PEC. The BMC ceases using PEC if
it receives an SSIF write transaction that does not
include PEC.
0b = does not support PEC. Note that a BMC implementation
may reject write transactions that include a PEC byte.
[2:0] - SSIF Version
000b = version 1 (version defined in this specification).
Input message size in bytes. (1 based.)
Number of bytes of IPMI message data that the BMC can accept.
This number does not include slave address, SMBus length,
PEC, or SMBus CMD bytes, just the IPMI message data. A BMC
that just supports single-part writes would return 32 (20h) for this
value. A BMC that supports multi-part Start and End would return
a value from 33 to 64. A BMC that supports multi-part with Middle
transactions would return a value from 65 to 255.
Output message size in bytes. (1 based.)
Maximum number of bytes of IPMI message data that can be
read from the BMC. This number does not include slave address,
SMBus length, PEC, SMBus CMD bytes, special bytes (such as
the special bytes following the length byte in the multi-part read
middle and end transactions) just the IPMI message data. A BMC
that just supports single-part reads would return 20h (32) for this
value. A BMC that supports multi-part Start and End would return
a value from 33 to 62 (the reason this is 62 instead of 64 is that
there are two special bytes after the length byte.) A BMC that
supports multi-part with Middle transactions would return a value
from 63 to 255.
For System Interface Type = KCS
[7:3] - reserved
[2:0] - System Interface Version
000b = version 1 (conformant with KCS interface defined in
this specification).
Intelligent Platform Management Interface Specification
4
Input maximum message size in bytes. (1 based.)
Largest number of bytes that can be transferred in a KCS
FFh means 255 or more.
22.10 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 22-13, Get BT Interface Capabilities Command
Request Data
Response Data
byte
-
data field
-
1
Completion Code
2
Number of outstanding requests supported (1 based. 0 illegal)
Input (request) buffer size in bytes. (1 based.)
Output (response) buffer 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
271
Intelligent Platform Management Interface Specification
22.11 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.
P
P
Table 22-14, 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.
P
P
22.12 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 13-8, RMCP/RMCP+ Packet
272
Intelligent Platform Management Interface Specification
Format 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.
22.13 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.
BMC implementations of IP-based channels must support the Get Channel Authentication Capabilities Command
using the IPMI v1.5 packet format. BMCs that support IPMI v2.0 RMCP+ must also support the command using
the IPMI v2.0/RMCP+ format.
The Get Channel Authentication Capabilities command can also be used as a no-op ‘Ping’ to keep a session from
timing out.
273
Intelligent Platform Management Interface Specification
Table 22-15, Get Channel Authentication Capabilities Command
Session Request Data
IPMI Request Data
authentication type = NONE / payload type = IPMI Message
session seq# = null (0’s)
Session ID = null (0’s)
AuthCode = NOT PRESENT
1
2
Session Response Data
IPMI Response Data
authentication type = NONE / payload type = IPMI Message
session seq# = null (0’s)
Session ID = null (0’s)
AuthCode = NOT PRESENT
1
2
3
274
Channel Number
[7] - 1b = get IPMI v2.0+ extended data.
0b = Backward compatible with IPMI v1.5. Result data only returns bytes
1:9, bit [7] of byte 3 (Authentication Type Support) returns as 0b,
byte 5 returns 00h.
[6:4] - 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
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] - 1b = IPMI v2.0+ extended capabilities available. See Extended
Capabilities field, below.
0b = IPMI v1.5 support only.
[6] - reserved
[5:0] - IPMI v1.5 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
Intelligent Platform Management Interface Specification
4
[7:6] - reserved
[5] - KG status (two-key login status). Applies to v2.0/RMCP+ RAKP
Authentication only.
0b = KG is set to default (all 0’s). User key KUID will be used in place of
KG in RAKP. (Knowledge of KG not required for activating session.)
1b = KG is set to non-zero value. (Knowledge of both KG and user
password (if not anonymous login) required for activating session.)
Following bits apply to IPMI v1.5 and v2.0:
[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.
[2:0] - Anonymous Login status
This parameter returns values that tells the remote console whether
there are users on the system that have ‘null’ usernames. This can be
used to guide the way the remote console presents login options to the
user. (see IPMI v1.5 specification sections 6.9.1, ‘Anonymous Login’
Convention and 6.9.2, Anonymous Login )
[2] - 1b = Non-null usernames enabled. (One or more users are enabled
that have non-null usernames).
[1] - 1b = Null usernames enabled (One or more users that have a null
username, but non-null password, are presently enabled)
[0] - 1b = Anonymous Login enabled (A user that has a null username
and null password is presently enabled)
For IPMI v1.5: - reserved
5
U
U
For IPMI v2.0+: - Extended Capabilities
[7:2] - reserved
[1] - 1b = channel supports IPMI v2.0 connections.
[0] - 1b = channel supports IPMI v1.5 connections.
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.
U
6:8
9
U
275
Intelligent Platform Management Interface Specification
22.14 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 22-16, 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 20-10, GUID Format.
22.15 Get Channel Cipher Suites Command
This command can be executed prior to establishing a session with the BMC. The command is used to look up
what authentication, integrity, and authentication algorithms are supported. The algorithms are used in
combination as ‘Cipher Suites’. This command only applies to implementations that support IPMI v2.0/RMCP+
sessions.
The data is accessed 16-bytes at a time starting from List Index field value of 0 in the request and then repeating
the request incrementing the List Index field each time until fewer than 16-bytes of algorithm data (or no
algorithm data) is returned in the response, or the maximum List Index value has been reached.
A given Cipher Suite may only be available for establishing a session at a particular maximum privilege level or
lower. For example, a Cipher Suite that has a privilege level of ‘Admin’ can therefore be used for any privilege
level, while a privilege level of User can only be used for establish sessions with a Maximum Requested Privilege
Level of User or Callback.
Because the authentication algorithm specifies the steps for authenticating the user, it is a necessary part of
session establishment. Therefore an authentication algorithm number is required for all Cipher Suites. It is
possible that a given Cipher Suite may not specify use of an integrity or confidentiality algorithm. If the Cipher
Suite has integrity and/or confidentiality of 'none', then all the same steps for establishing a session are used (open
session request/response, RAKP messages) - but the integrity (AuthCode) and confidentiality fields will be absent
in packets for that are sent under the session.
276
Intelligent Platform Management Interface Specification
Table 22-17, Get Channel Cipher Suites Command
IPMI Request Data
1
2
3
IPMI Response Data
1
2
(3:18)
Channel Number
[7:4] - reserved
[3:0] - channel number.
0h-7h, Fh = channel numbers
Eh = retrieve information for channel this request was issued on.
Payload Type.
[7:6] - reserved
[5:0] - Payload Type number
Typically 00h (IPMI).
The Payload Type number is used to look up the Security Algorithm support
when establishing a separate session for a given payload type.
List Index.
[7] - 1b = list algorithms by Cipher Suite
0b = list supported algorithms
[6] - reserved
[5:0] - List index (00h-3Fh). 0h selects the first set of 16, 1h selects the next
set of 16, and so on.
00h = Get first set of algorithm numbers. The BMC returns sixteen (16)
bytes at a time per index, starting from index 00h, until the list data
is exhausted, at which point it will 0 bytes or <16 bytes of list data.
Completion Code
Channel Number
Channel number that the Authentication Algorithms are 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.
Cipher Suite Record data bytes, per Table 22-18, Cipher Suite Record
Format. Record data is ‘packed’; there are no pad bytes between records. It is
possible that record data will span across multiple List Index values.
The BMC returns sixteen (16) bytes at a time per index, starting from index
00h, until the list data is exhausted, at which point it will 0 bytes or <16 bytes
of list data.
22.15.1 Cipher Suite Records
The data from the Get Channel Cipher Suites command is issued as Cipher Suite records. Tag bits are used to
delimit different fields in the record. Each record starts off with a “Start Of Record” byte. This byte can be 30h or
31h, indicating that the Start Of Record byte is followed either by an Cipher Suite ID, or by a OEM Cipher Suite
ID plus OEM IANA.
Following the header bytes are algorithm number bytes for the different algorithms that form the Cipher Suite.
Each byte is tagged with the type of algorithm the number is for. Cipher Suite records are required to list
algorithms in the order: Authentication Algorithm number first, Integrity Algorithm numbers next, and
Confidentiality Algorithm numbers last.
277
Intelligent Platform Management Interface Specification
If more than one algorithm of a given type is listed in the Cipher Suite Record, then any one of the algorithms can
be used in combination with the other types. For example, if a Cipher Suite response returns both MD5 and MD2
as Authentication and Integrity algorithms, and xRC4 for confidentiality, then the allowed combinations are
[MD2, MD2, xRC4], [MD2, MD5, xRC4], [MD5, MD2, xRC4], and [MD5, MD5, xRC4]. I.e. a remote console
can negotiate for those combinations when establishing a session.
Table 22-18, Cipher Suite Record Format
size
2 or 5
Tag bits
[7:6]
11b
Tag bits
[5:0]
Start Of Record
U
[5:0] = 00_000b. Start of record followed by Cypher Suite ID tag.
Data following tag byte:
Byte 1 - Cipher Suite ID
This value is used a numeric way of identifying the Cipher Suite on the platform. It’s used
in commands and configuration parameters that enable and disable Cipher Suites. See
Table 22-19, Cipher Suite IDs.
U
U
[5:0] = 00_001b. Start of record followed by Cipher Suite ID + OEM IANA tag.
Data following tag byte:
Byte 1 - OEM Cipher Suite ID
See Table 22-19, Cipher Suite IDs.
U
Byte 2:4 - OEM IANA
Least significant byte first. 3-byte IANA for the OEM or body that defined the Cipher Suite.
[5:0] = Authentication Algorithm Number.
A Cipher Suite is only allowed to utilize one Authentication algorithm. See Table 13-17,
Authentication Algorithm Numbers
[5:0] = Integrity Algorithm Number(s). See Table 13-18, Integrity Algorithm Numbers
[5:0] = Confidentiality Algorithm Number(s). See Table 13-19, Confidentiality Algorithm
Numbers
U
1
00b
var
var
01b
10b
U
U
U
U
U
U
22.15.2 Cipher Suite IDs
The following table provides the number ranges and assignments for Cipher Suite IDs. The Cipher Suite ID
values are used as a way to identify different Cipher Suites in configuration parameters and IPMI commands.
The OEM IDs do not correspond to a particular Cipher Suite, but are handles that can be used to identify the
Cipher Suite on a particular implementation of a BMC. I.e. the OEM Cipher Suite corresponding to “80h” can
be different from one BMC to the next. These handles can, however, be used in configuration parameters and
commands the same way as the IPMI-defined Cipher Suite IDs.
The Get Channel Cipher Suites command will return the algorithms used to form a given Cipher Suite (those
numbers can then be used by a remote console in the commands for establishing a session). For OEM defined
Cipher Suites, the Get Channel Cipher Suites command will also return the IANA for the OEM or body that
defined the Cipher Suite.
278
Intelligent Platform Management Interface Specification
Table 22-19, Cipher Suite IDs
ID
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
80hBFh
C0hFFh
characteristics
“straight
password"
S
S, A
S, A, E
S, A, E
S, A, E
S
S, A
S, A, E
S, A, E
S, A, E
S, A
S, A, E
S, A, E
S, A, E
OEM specified
reserved
Authentication
Algorithm
RAKP-none
Integrity
Algorithm(s)
None
Confidentiality
Algorithm(s)
None
01h, 00h, 00h
01h, 01h, 00h
01h, 01h, 01h
01h, 01h, 02h
01h, 01h, 03h
02h, 00h, 00h
02h, 02h, 00h
02h, 02h, 01h
02h, 02h, 02h
02h, 02h, 03h
02h, 03h, 00h
02h, 03h, 01h
02h, 03h, 02h
02h, 03h, 03h
OEM specified
RAKP-HMAC-SHA1
None
HMAC-SHA1-96
RAKP-HMAC-MD5
None
HMAC-MD5-128
None
None
AES-CBC-128
xRC4-128
xRC4-40
None
None
AES-CBC-128
xRC4-128
xRC4-40
None
AES-CBC-128
xRC4-128
xRC4-40
OEM specified
-
-
Cipher Suite
00h, 00h, 00h
MD5-128
OEM specified
OEM specified
-
-
Key:
S = authenticated session setup (correct role, username and password/key required to establish session)
A = authenticated payload data supported.
E = authentication and encrypted payload data supported
22.16 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.
279
Intelligent Platform Management Interface Specification
Table 22-20, 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
22.17 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.
280
Intelligent Platform Management Interface Specification
Table 22-21, 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 22.17.1,
AuthCode 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 6.12.4, Per-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 22.17.1, AuthCode 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.
281
Intelligent Platform Management Interface Specification
Session Response Data
IPMI Response Data
1
2
3:6
7:10
11
282
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 22.17.1,
AuthCode 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 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
22.17.1 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.
•
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.
Table 22-22, 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
AuthCode = H(password + Session ID[1] + IPMI Message data + session_seq# + password)
AuthCode = H(password + Session ID[1] + IPMI Message data + session_seq# +
password)
Get AuthCode AuthCode carried in IPMI message data, per command description
See description of Get AuthCode command.
P
P
P
P
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.
22.18 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.
283
Intelligent Platform Management Interface Specification
Table 22-23, 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
22.19 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 22-24, Close Session Command
Request Data
Response Data
byte
1:4
1
data field
Session ID. For IPMI v2.0/RMCP+ this is the Managed System Session ID
value that was generated by the BMC, not the ID from the remote
console.
Completion Code
87h = invalid Session ID in request
22.20 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.
For IPMI v2.0, a previously reserved field has been defined to hold a value indicating whether a session operating
on a channel of Channel Type = 802.3 LAN is presently using IPMI v1.5 or v2.0/RMCP+ protocols.
284
Intelligent Platform Management Interface Specification
Table 22-25, 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
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. For IPMI
v2.0/RMCP+ this is the Session ID value that was generated by the BMC, not
the ID from the remote console.
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.
[7:4] - Session protocol auxiliary data
For Channel Type = 802.3 LAN:
0h = IPMI v1.5
1h = IPMI v2.0/RMCP+
Channel that session was activated over.
[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 UDP packet that held the Activate Session command that activated
the session (for IPMI v1.5 packets) or that was used for in the packet for
RAKP Message 3 (for IPMI v2.0 / RMCP+ packets).
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.
285
Intelligent Platform Management Interface Specification
10:13
14:15
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.
22.21 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. Table 22-22, 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.
•
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)
B
•
286
B
Managed system accepts data if software agents finds that K2 = K2BMC.
B
B
Intelligent Platform Management Interface Specification
Table 22-26, Get AuthCode Command
IPMI Request Data
byte
1
data field
[7:6] - Authentication Type / Integrity Algorithm Number
00b = IPMI v1.5 AuthCode Algorithms
01b = IPMI v2.0/RMCP+ Algorithm Number
For [7:6] = 00b, IPMI v1.5 AuthCode Number:
[5:4] - reserved
[3:0] - hash type
0h = reserved
1h = MD2
2h = MD5
3h = reserved
4h = Reserved (change from IPMI v1.5). This shall result in an error
completion code.
5h = OEM proprietary
all other = reserved
2
3
IPMI Response Data
4:19
1
For [7:6] = 01b, IPMI v2.0/RMCP+ Integrity Algorithm Number
[5:0] - Integrity Algorithm Number. See Table 13-18, Integrity Algorithm
Numbers. The “none” Integrity Number (0) is illegal and shall result in an
error completion code.
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
data to hash (must be 16 bytes)
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.
For IPMI v1.5 AuthCode Number:
2:17
AuthCode = See 22.17.1, AuthCode Algorithms.
2:21
Resultant hash, per selected Integrity algorithm. Up to 20 bytes.
For IPMI v2.0 Integrity Algorithm Number
287
Intelligent Platform Management Interface Specification
22.22 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 22-27, 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.
288
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
289
Intelligent Platform Management Interface Specification
22.23 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 22-28, Get Channel Access Command
byte
Request Data
1
2
Response Data
1
2
3
290
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.
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
Intelligent Platform Management Interface Specification
22.24 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 22-29, 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
Numbers
Channel Protocol Type:
[7:5] - reserved
[4:0] - 5-bit Channel IPMI Messaging Protocol Type per Table 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.
291
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
U
U
U
U
For OEM channel types:
byte 1:2 = OEM specified per OEM identified by Vendor ID field.
U
U
All other channel types:
byte 1:2 = reserved.
U
U
U
22.25 Set Channel Security Keys Command
The Set Channel Security Keys command provides a standardized interface for initializing system unique keys
that are used for the pseudo-random number generator key (KR) and the key-generation key (KG) used for
RMCP+. Implementing the ability to set Kr is optional. The command is provided mainly to offer a common
interface for BMCs that are not pre-configured with a KR values, or which may need their KR values to be
restored if they are lost due to a data corruption or firmware update.
The command includes a mechanism that allows specified keys to be ‘locked’. Once locked, the key value cannot
be read back or rewritten via standard IPMI commands. It is possible, however, that a firmware update or reinstallation procedure may cause the keys to be cleared or unlocked. Software utilities responsible for BMC initial
installation and setup should check to see whether keys have been locked and if not, should initialize them
appropriately and lock them.
If this command is not supported, it indicates that the keys are either permanently pre-configured, or that they are
only configurable via an OEM/BMC-specific mechanism.
Table 22-30, Set Channel Security Keys Command
byte
Request Data
1
2
data field
Channel Number
[7:4] - reserved
[3:0] - Channel Number
Operation
[7:2] - reserved
[1:0] - Operation
00b = read key
BMC returns value of specified key, provided key has not yet been
locked. Some BMCs may allow the key to be re-written if it does
not match the expected value. Other BMCs may only allow one
‘set’ operation. If the key value has not yet been initialized, the
BMC will return 0’s for the key value. Utility software responsible
for BMC installation and initial setup can use this Operation to
also check to see whether keys have been initialized and locked.
01b = set key
BMC writes given key value to non-volatile storage.
10b = lock key
292
Intelligent Platform Management Interface Specification
BMC locks out modification or reading the key value. Once a key
has been locked, it is not cannot be rewritten or read via IPMI
specified commands.
3
(4:M)
Response Data
1
2
(3:N)
11b = reserved
Key ID
[7:0] - key ID.
00h = RMCP+ “KR” key. The “KR” key is used as a unique value for random
number generation. Note: A BMC implementation is allowed to share a
single KR value across all channels. A utility can set KR and lock it for
one channel, and then verify it has been set and locked for any other
channels by using this command to read the key from other channels
and checking the ‘lock status’ field to see if it matches and is unlocked.
01h = RMCP+ “KG” key. “KG” key acts as a value that is used for key
exchange for the overall channel. This key is not lockable in order to
enable a password/key configuration utility to set its value. This value is
used in conjunction with the user key values (passwords) in RAKPHMAC-SHA1 and RAKP-HMAC-MD5 authentication. I.e. the remote
console needs to have a-priori knowledge of both this key value and
the user password setting, in order to establish a session.
all other = reserved
Key value. Value for specified key. Used for “set” Operation only. Otherwise,
this field is not used in the request. The BMC will ignore any bytes following
the ‘Key ID’ byte.
Completion Code. Generic, plus following command-specific completion
codes:
80h = Cannot perform set / confirm. Key is locked (mandatory)
81h = insufficient key bytes
82h = too many key bytes
83h = key value does not meet criteria for specified type of key
84h = KR is not used. BMC uses a random number generation approach
that does not require a KR value.
7:2 - reserved.
1:0 - lock status
00b = key is not lockable.
01b = key is locked.
10b = key is unlocked.
11b = reserved
Key value.
The BMC returns the specified key value when the Operation is set to
“confirm key”. Otherwise, the BMC returns no additional bytes past the
completion code.
22.26 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.
293
Intelligent Platform Management Interface Specification
Table 22-31, Set User Access Command
byte
Request Data
1
2
3
(4)
294
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. In this case, “IPMI Messaging” refers to the ability to
execute generic IPMI commands that are not associated with a
particular payload type. For example, if IPMI Messaging is disabled for
a user, but that user is enabled for activating the SOL payload type,
then IPMI commands associated with SOL and session management,
such as Get SOL Configuration Parameters and Close Session are
available, but generic IPMI commands such as Get SEL Time are
unavailable.)
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.
Intelligent Platform Management Interface Specification
Response Data
1
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.
22.27 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 22-32, 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)
295
Intelligent Platform Management Interface Specification
22.28 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 Support for more information.
Table 22-33, 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.
22.29 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 22-34, Get User Name Command
byte
Request Data
Response Data
296
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).
Intelligent Platform Management Interface Specification
22.30 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 or 20-byte (for IPMI v2.0/RMCP+) ‘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 is done, any 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 22-35, Set User Password Command
byte
Request Data
1
data field
User ID.
For IPMI v2.0, the BMC shall support 20-byte passwords (keys) for all
supported user IDs that have configurable passwords. The BMC shall
maintain an internal tag that indicates whether the password was set as
a 16-byte or as a 20-byte password.
A 16-byte password can be used in algorithms that call for a 20-byte
password. In this case, the 16-byte password is padded with 0’s to 20bytes.
The ‘test password’ operation shall return the ‘test failed’ error
completion code if an attempt is made to test a password that was
stored as a 20-byte password as a 16-byte password (per ‘password
size’ bit 7), and vice versa.
A password that has been stored as a 20-byte password cannot be used
for establishing an IPMI v1.5 session. If it is necessary to configure the
same password for both IPMI v2.0 and IPMI v1.5 access, it must be set
5
as a 16-byte password . The password will be padded with 0’s as
necessary for IPMI v2.0 / RMCP+ use.
TP
2
TP
5
PT
PT
The ‘test password’ operation can be used to determine whether a
password has been stored as 16-bytes or 20-bytes.
[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
Note that the same username is allowed to be used with different passwords on different channels. The BMC scans usernames
until it finds the first one that is enabled for a particular channel. Thus, it is possible for a BMC implementation to allow being
configured to allow a 20-byte password on one channel, and a 16-byte password on another channel for the same username.
This requires multiple user entries.
297
Intelligent Platform Management Interface Specification
Response Data
298
3:18
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.
For password size = 20 bytes:
3:22
20-byte Password data. This is a required, fixed length field when used for the
set- and 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 = (mandatory) password test failed. Password size correct, but
password data does not match stored value.
81h = (mandatory) password test failed. Wrong password size was used.
1
Intelligent Platform Management Interface Specification
23. 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
for the specification of the Network Function and Command (CMD) values and privilege levels for these
commands.
Table 23-1, 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
23.1
23.2
23.3
23.4
M[1]
M[1]
O[2]
O
P
P
P
P
P
P
Mandatory if LAN channel is supported.
Mandatory if BMC autonomously generates Gratuitous ARPs
23.1 Set LAN Configuration Parameters Command
This command is used for setting parameters such as the network addressing information required for IPMI LAN
operation.
Table 23-2, Set LAN Configuration Parameters Command
byte
Request Data
1
2
Response Data
data field
[7:4] - reserved
[3:0] - Channel number.
Parameter selector
3:N
Configuration parameter data, per Table 23-4, LAN Configuration Parameters
1
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
299
Intelligent Platform Management Interface Specification
23.2 Get LAN Configuration Parameters Command
This command is used for retrieving the configuration parameters from the Set LAN Configuration command.
Table 23-3, Get LAN Configuration Parameters Command
byte
Request Data
data field
1
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)
2
3
[7] -
4
Response Data
1
2
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 23-4, 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.
3:N
Table 23-4, LAN Configuration Parameters
Parameter
#
Set In Progress
(volatile)
0
300
Parameter Data (non-volatile unless otherwise noted)[1]
P
P
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
U
U
Intelligent Platform Management Interface Specification
Parameter
#
Parameter Data (non-volatile unless otherwise noted)[1]
Authentication Type
Support (Read Only)
1
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.
P
P
[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
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
byte 5: Authentication Type (s) for maximum privilege = OEM level
(format follows byte 1)
data 1:4 - IP Address
MS-byte first.
301
Intelligent Platform Management Interface Specification
Parameter
#
Parameter Data (non-volatile unless otherwise noted)[1]
IP Address Source
4
MAC Address
5
Subnet Mask
IPv4 Header Parameters
6
7
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.
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
(optional[2])
P
8
9
10
P
Gratuitous ARP
interval (optional)
11
Default Gateway
Address
12
Default Gateway MAC
Address
Backup Gateway
Address
13
Backup Gateway MAC
Address
Community String
15
302
14
16
P
P
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.
Intelligent Platform Management Interface Specification
Parameter
#
Number of Destinations
(Read Only)
17
Destination Type
(volatile / non-volatile see description)
18
Parameter Data (non-volatile unless otherwise noted)[1]
P
P
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 Immediate command.
[7:4] - reserved.
[3:0] - Number LAN Destinations. A count of 0h indicates LAN Alerting is not supported.
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 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.
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).
303
Intelligent Platform Management Interface Specification
Parameter
Destination Addresses
#
19
Parameter Data (non-volatile unless otherwise noted)[1]
P
P
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 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
data 4:7 - Alerting IP Address (MS-byte first)
data 8:13 - Alerting MAC Address (MS-byte first)
Following parameters are introduced with IPMI v2.0 / RMCP+
VLAN configuration can be used with IPMI v1.5 and IPMI v2.0sessions. Parameters labeled “RMCP+” are specific to
IPMI v2.0 implementations that implement IPMI v2.0 / RMCP+ sessions.
802.1q VLAN ID (12-bit)
20
data 1
[7:0] -
Least significant 8-bits of the VLAN ID. 00h if VLAN ID not used.
data 2
[7] -
VLAN ID enable.
0b = disabled, 1b = enabled. If enabled, the BMC will only accept packets for this
channel if they have 802.1q fields and their VLAN ID matches the VLAN ID value
given in this parameter.
802.1q VLAN Priority
RMCP+ Messaging
Cipher Suite Entry
Support
(Read Only)
RMCP+ Messaging
Cipher Suite Entries
(Read Only)
21
22
[6:4] -
reserved
[3:0] -
most significant four bits of the VLAN ID
data 1
[7:5] -
reserved
[2:0] -
Value for Priority field of 802.1q fields. Ignored when VLAN ID enable is 0b (disabled)
- See 802.1q VLAN ID parameter, above. Setting is network dependent. By default,
this should be set to 000b.
This parameter provides a count of the number (16 max.) of Cipher Suites available to be
enabled for use with IPMI Messaging on the given channel.
Software can find out what security algorithms are associated with given Cipher Suite ID by
using the Get Channel Cipher Suites command. In addition, there are Cipher Suite IDs
assigned for standard Cipher Suites (see Table 22-19, Cipher Suite IDs)
23
data 1
[7:5] - reserved
[4:0] - Cipher Suite Entry count. Number of Cipher Suite entries, 1-based, 10h max.
This parameter contains zero to sixteen (16) bytes of Cipher Suite IDs for Cipher Suites that
can be used for establishing an IPMI messaging session with the BMC. The number of
Cipher Suites that are supported is given in the preceding parameter.
data 1 - Reserved
data 2 - Cipher Suite ID entry A. data 3 - Cipher Suite ID entry B....
data 17 - Cipher Suite ID entry P.
304
Intelligent Platform Management Interface Specification
Parameter
RMCP+ Messaging
Cipher Suite Privilege
Levels
#
24
Parameter Data (non-volatile unless otherwise noted)[1]
P
P
This parameter allows the configuration of which privilege levels are associated with each
Cipher Suite. The total number of nibbles supported (zero to sixteen) matches the number of
fixed Cipher Suite IDs.
data 1 data 2 [7:4] [3:0] -
Destination Address
VLAN TAGs
25
Reserved
Maximum Privilege Level for 1st and 2nd Cipher Suites
Maximum Privilege Level for 2nd Cipher Suite
Maximum Privilege Level for 1st Cipher Suite
0h = Unspecified (given Cipher Suite is unused)
1h = Callback level
2h = User level
3h = Operator level
4h = Administrator level
5h = OEM Proprietary level
data 3 - Maximum Privilege Level for 3rd and 4th Cipher Suites
data 4 - Maximum Privilege Level for 5th and 6th Cipher Suites
…
data 9 - Maximum Privilege Level for 15th and 16th Cipher Suites
Sets/Gets the VLAN IDs (if any) addresses that a LAN alert can be sent to. This parameter is
not present if the Number of Destinations parameter is 0, or if the implementation does not
support the use of VLAN IDs. Otherwise, the number of VLAN TAG entries matches the
number of Alert Destinations.
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 Immediate command.
data 2 - Address Format
[7:4] Address Format.
0h = VLAN ID not used with this destination
1h = 802.1q VLAN TAG
[3:0] reserved
OEM Parameters
192
:
255
For Address Format = 1h:
data 3:4 - VLAN TAG
[7:0] VLAN ID, least-significant byte
[11:8] - VLAN ID, most-significant nibble
[12] CFI (Canonical Format Indicator. Set to 0b)
[15:13] - User priority (000b, typical)
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.
23.3 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.
305
Intelligent Platform Management Interface Specification
Table 23-5, 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.
306
Intelligent Platform Management Interface Specification
23.4 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 23-6, 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
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
307
Intelligent Platform Management Interface Specification
24. RMCP+ Support and Payload Commands
The following sections list the commands associated with discovering, enabling, and activating payloads under IPMI
v2.0/RMCP+. Also included in this section are updates and additions to IPMI commands to support IPMI v2.0/RMCP+
sessions, authentication, and configuration.
Table 24-1, RMCP+ Support and Payload Commands
Command
Activate Payload
Deactivate Payload
Suspend/Resume Payload Encryption Command
Get Payload Activation Status
Get Payload Instance Info
Set User Payload Access Command
Get User Payload Access Command
Get Channel Payload Support Command
Get Channel Payload Version Command
Get Channel OEM Payload Info Command
1.
2.
3.
Section
Defined
O/M
24.1
24.2
24.3
24.4
24.5
24.6
24.7
24.8
24.9
24.10
O[1]
O[1]
O[3]
O[1]
O[1]
O[1]
O[1]
O[1]
O[1]
O[1,2]
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
Mandatory if standard payloads other than IPMI Messaging are available on the
channel.
Mandatory if OEM payloads are available on the channel.
Mandatory for standard payloads per Table 24-4, Payload-specific Encryption
Behavior, when a channel supports Cipher Suites that allow a session to be
established with a standard Cipher Suite that supports encryption.
24.1 Activate Payload Command
This command is used for activating and deactivating a payload type under a given IPMI session. The ability to
execute this command is determined via the user’s privileges as assigned via the Set User Payload Access
command.
The Activate Payload command may return a port number that is separate from the port number for the session
that the command was issued under. In this case, the remote console must establish a session on the port number
that the Activate Session command returned. The remote console must then issue the Activate Payload command
on that port number in order to actually activate the payload. It is possible that the remote console already had a
session active on the given port number. If the privileges associated with that session are sufficient (this will
typically be the case unless the remote console activated the session at a privilege level that was lower than the
maximum level for the user) the remote console can re-use the existing session and just use the Activate Payload
command to activate the new payload type.
BMCs may have limited resources for handling multiple sessions. It is highly recommended that a remote console
avoids creating multiple sessions and shares sessions for multiple payloads whenever possible.
The Activate Payload Command is only accepted over a channel on which payloads can be activated. E.g. the
Activate Payload command cannot be executed from the IPMB.
308
Intelligent Platform Management Interface Specification
Table 24-2, Activate Payload Command
byte
Request Data
1
2
3:6
data field
[7:6] - reserved
[5:0] - payload type (See Table 13-16, Payload Type Numbers) IPMI
Message payloads do not need to be explicitly activated. A payload
that is required to be launched over a different port than that used to
establish the initial IPMI session is only required to support the IPMI
commands needed by the particular payload type.
Payload Instance
[7:4] - reserved
[3:0] - payload instance. 1-based. 0h = reserved.
Auxiliary Request Data. Additional payload-specific parameters to configure
behavior of the payload when it becomes activated. Ignored if no auxiliary
data is specified for given payload type.
For Payload Type = SOL:
byte 1
[7] - Encryption Activation
Note: the encryption algorithms specified in this document must be used
with Authentication. The BMC will return an error completion code if an
attempt is made to activate encryption without also activating
authentication.
1b: Activate payload with encryption. All SOL payload data from the
BMC will be encrypted, if encryption was negotiated at the time of
session activation.
0b: Activate payload without encryption. BMC will send all SOL
payload data unencrypted, if that option is allowed. (An SOL
configuration parameter allows a system to be configured to
require encryption for all SOL transfers)
[6] - Authentication Activation
1b: Activate payload with authentication. All SOL payload data from
the BMC will be authenticated, if authentication was negotiated at
the time of session activation.
0b: Activate payload without authentication. BMC will send all SOL
payload data unauthenticated, if that option is allowed. (An SOL
configuration parameter allows a system to be configured to
require authentication for all SOL transfers)
[5:4] - reserved
[3:2] - Shared Serial Alert Behavior
The following settings are determine what happens to serial alerts if IPMI
over Serial and SOL are sharing the same baseboard serial controller.
11b: Reserved
10b: Serial/modem alerts succeed while SOL active.
01b: Serial/modem alerts deferred while SOL active.
00b: Serial/modem alerts fail while SOL active.
[1] SOL startup handshake
0b: BMC asserts CTS and DCD/DSR to baseboard upon activation.
1b: CTS and DCD/SDR remain deasserted after activation. Remote
console must send an SOL Payload packet with control field
settings to assert CTS and DCD/DSR. (This enables the remote
console to first alter volatile configuration settings before
hardware handshake is released).
[0] reserved
byte 2:4 reserved - write a 00h
309
Intelligent Platform Management Interface Specification
Response Data
1
2:5
6:7
8:9
10:11
12:13
Completion Code
Generic plus the following command-specific completion codes:
80h: Payload already active on another session (required).
This will be returned any time an attempt is made to activate a payload
type when that type is already activated for another session, and when
the BMC only supports one instance of that payload type running at a
time.
81h: Payload type is disabled (optional). Given payload type is not
configured to be enabled for activation.
82h: Payload activation limit reached. Cannot activate given payload type
because the maximum number of simultaneous instances of that
payload type are already running.
83h: Cannot activate payload with encryption.
84h: Cannot activate payload without encryption. BMC requires encryption
for all payloads for given privilege level.
Auxiliary Response Data.
For Payload = SOL: reserved. Return as 00_00_00_00h.
Inbound Payload Size
Maximum size of payload data field from remote console to BMC. Excludes
size of confidentiality header and trailer fields, if any. 1-based.
Outbound Payload Size
Maximum size of payload data field from BMC to remote console. Excludes
size of confidentiality header and trailer fields, if any. 1-based.
Payload UDP Port Number
UDP port number that payload can be transferred over. If the port number is
same as the port that was used to establish the IPMI session, then SOL
payload transfers are now available under that IPMI session on that port.
Otherwise, the remote console will need to establish a separate IPMI Session
to the specified Port Number using the same IP Address, username and
password/key information that was used to establish the IPMI session. SOL
payload transfers will then be available over that session.
If the remote console already has an IPMI session established on that port for
a different payload type, the SOL payload type will now also be available over
that session - provided that the session was established at a privilege level
that matches the privilege level and authentication required for SOL.
Otherwise, the remote console will need to close that session and reestablish it at the necessary privilege level.
Payload VLAN Number - FFFFh if VLAN addressing is not used.
24.2 Deactivate Payload Command
This command is used to terminate use of a given payload on an IPMI session. This type of traffic then becomes
freed for activation by another session, or for possible re-activation under the present session. The Deactivate
Payload command does not cause the session to be terminated. The Close Session command should be used for
that purpose. A remote console application does not need to explicitly deactivate payload(s) prior to terminating a
session. When a session terminates all payloads that were active under that session are automatically deactivated
by the BMC.
310
Intelligent Platform Management Interface Specification
Table 24-3, Deactivate Payload Command
Request Data
byte
1
2
3:6
Response Data
1
data field
[7:6] - reserved
[5:0] - payload type (See Table 13-16, Payload Type Numbers)
Payload Instance
[7:4] - reserved
[3:0] - payload instance. 1-based. 0h = reserved.
Payload Auxiliary Data. Additional parameters to configure behavior of the
payload when it becomes deactivated. Ignored if no auxiliary data is specified
for given payload type.
For Payload Type = SOL: (no auxiliary data) write as 0000_0000h
Completion Code
Generic plus the following command-specific completion codes:
80h: Payload already deactivated.
81h: Payload type is disabled (optional). Given payload type is not
configured to be enabled for activation.
24.3 Suspend/Resume Payload Encryption Command
This command enables a remote console to control whether payload data from the BMC is sent encrypted or not.
Since encryption can be a significant burden on software, this command provides a mechanism to allow higher
performance by operating without encryption and only activating encryption when it is required for data
confidentiality. The command can also trigger a regeneration of the encryption Initialization Vector and reinitialization of the encryption state machine for algorithms such as xRC4 that use the same initialization vector
for multiple packets.
The extent at which this command can control encryption of data from the BMC is dependent on the payload
definition. Some payload definitions may use a mix of encrypted and unencrypted payload data transfers. For
example, a payload may implement a ‘request/response’ protocol, where the BMC would return an encrypted or
unencrypted response based on whether the request from the remote console was encrypted or unencrypted. In
this case, the command may only affect data that is autonomously generated by the BMC. Other payload
definitions may just use whatever encryption the session was activated with, and offer no ‘run-time’ control of
encryption/decryption, while other payload definitions may be ‘stream based’ where it is desirable for the remote
console to be able to select when payload data is from the BMC is encrypted or not.
The Suspend/Resume Payload Encryption command is only accepted from the channel that the payload was
activated on.
Table 24-4, Payload-specific Encryption Behavior
Payload Type = IPMI Messaging
•
Encrypted requests from the remote console will get encrypted responses from the BMC.
•
The Suspend/Resume Payload Encryption command controls whether asynchronous
(unrequested) messages from the BMC are encrypted or not.
•
PET Traps (which are actually separate from IPMI Messaging) are always sent unencrypted.
Payload Type = SOL
•
The SOL configuration parameters allow configuring the system to require that SOL data be
encrypted.
•
The BMC will transmit SOL payload data according to encryption settings that were selected when
the payload was activated unless over-ridden by SOL configuration parameters.
•
The Suspend/Resume Payload Encryption command controls whether SOL Payload data is
encrypted or not.
311
Intelligent Platform Management Interface Specification
Table 24-5, Suspend/Resume Payload Encryption Command
Request Data
byte
1
2
3
Response Data
1
data field
[7:6] - reserved
[5:0] - payload type (See Table 13-16, Payload Type Numbers)
Payload Instance
[7:4] - reserved
[3:0] - payload instance. 1-based. 0h = reserved.
[7:2] - reserved
[4:0] - Operation
2h = Regenerate initialization vector. For xRC4 encryption, this causes
the BMC to reinitialize the xRC4 state machine, reset the data
offset, and deliver a new Initialization Vector value in the next
encrypted packet it sends to the remote console. Because of
processing delays and potential tasks in progress, the remote
console may receive additional packets from the BMC that are
encrypted using the prior Initialization Vector before getting packets
that use the new IV.
1h = Resume/Start encryption on all transfers of specified payload data
from the BMC.
0h = Suspend encryption on all transfers of specified payload messages
from the BMC.
Completion Code
Generic plus the following command-specific completion codes:
80h: Operation not supported for given payload type.
81h: Operation not allowed under present configuration for given payload
type.
82h: Encryption is not available for session that payload type is active under.
83h: The payload instance is not presently active.
24.4 Get Payload Activation Status Command
This command returns how many instances of a given payload type are presently activated, and how many total
instances can be activated.
312
Intelligent Platform Management Interface Specification
Table 24-6, Get Payload Activation Status Command
byte
Request Data
1
Response Data
1
2
3
4
data field
Payload Type Number - Type number of the standard payload type or OEM
Payload Handle to retrieve status for.
See Table 13-16, Payload Type Numbers.
Completion Code
Instance capacity
[7:4] reserved.
[3:0] - Number of instances of given payload type that can be simultaneously
activated on BMC. 1-based. 0h = reserved.
[7] - 1b = instance 8 is activated.
0b = instance 8 is deactivated.
[6] - 1b = instance 7 is activated.
0b = instance 7 is deactivated.
…
[0] - 1b = instance 1 is activated.
0b = instance 1 is deactivated.
[7] - 1b = instance 16 is activated.
0b = instance 16 is deactivated.
[6] - 1b = instance 15 is activated.
0b = instance 15 is deactivated.
…
[0] - 1b = instance 9 is activated.
0b = instance 9 is deactivated.
24.5 Get Payload Instance Info Command
This command returns information about a specific instance of a payload type. It is primarily used by software
that may want to negotiate with an application that is presently using the given payload type. It accomplishes this
by using the Session ID returned from this command with the Get Session Info command to look up the
addressing information for the party that activated the payload. The application may then use that information to
establish a direct dialog with the application that presently ‘owns’ the payload (note that this inter-application
communication is not defined in the IPMI specifications).
Table 24-7, Get Payload Instance Info Command
byte
Request Data
1
2
Response Data
1
2:5
4:11
data field
Payload Type Number - Type number of the standard payload type or OEM
Payload Handle to retrieve status for.
See Table 13-16, Payload Type Numbers.
Payload Instance. 1-based. 0h = reserved.
Completion Code
Session ID - ID of session that instance is presently activated on. (The
Managed System Session ID that the BMC generated when the
session was activated). 00_00_00_00h if given instance is not
activated. Remote software can use this information with the Get
Session Info command to identify the remote console that presently is
using a given payload type.
Payload-specific information (8-bytes)
For Payload Type = SOL:
Byte 1: Port Number
A number representing the system serial port that is being redirected. 1based. 0h = unspecified. Used when more than one port can be
redirected on a system.
Byte 2:8 = reserved.
313
Intelligent Platform Management Interface Specification
24.6 Set User Payload Access Command
IPMI v2.0 introduces the Set User Payload Access and Get User Payload Access commands. These commands
can be thought of as extensions to the Set User Access and Get User Access commands, respectively. The Set
User Payload Access command controls whether the specified user has the ability to activate the specified
payload type on the given channel.
The Set User Payload Access command uses bitfields to allow a configuration utility to use a single command to
set enable/disable multiple payloads at a time. Standard payloads are set separately from OEM payload enables.
The command would be issued at least once with Standard payloads selected to set the configuration for Standard
payloads, and then at least once with OEM Payloads selected to set the configuration for OEM payloads.
Table 24-8, Set User Payload Access Command
byte
Request Data
1
2
3
4
5
6
Response Data
1
1.
314
data field
Channel Number
[7:4] - reserved
[3:0] - Channel Number
[7:6] - Operation
00b = ENABLE.
Writing a “1b” to enable/disable bit ENABLES corresponding
payload. Writing “0b” to bit causes no change to enabled/disabled
state
01b = DISABLE.
Writing a “1b” to bit DISABLES corresponding payload. Writing 0b
to bit causes no change to enabled/disabled state.
10b, 11b = reserved
[5:0] - User ID. 000000b = reserved.
Standard Payload enables 1
[7:2] - reserved for standard payloads 2-7 enable/disable bits
[1] standard payload 1 (SOL) enable/disable
[0] reserved. Note: IPMI Messsaging is enabled/disabled for users via the
Set User Access command.
Standard Payload Enables 2 - reserved
OEM Payload Enables 1
[7] - OEM Payload 7 enable/disable
[6] - OEM Payload 6 enable/disable
[5] - OEM Payload 5 enable/disable
[4] - OEM Payload 4 enable/disable
[3] - OEM Payload 3 enable/disable
[2] - OEM Payload 2 enable/disable
[1] - OEM Payload 1 enable/disable
[0] - OEM Payload 0 enable/disable
OEM Payload Enables 2 - 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 set using the Set Channel Access command and
provide notification of any mismatch.
The following commands remain available for payloads if IPMI Messaging Payload
type is disabled: Deactivate Payload, Suspend/Resume Payload Encryption (as
defined for given payload), Get Payload Activation Status, Get Channel Payload
Version Command, Get Channel OEM Payload Info (if implemented), Set Session
Privilege Level, and Close Session.
Intelligent Platform Management Interface Specification
24.7 Get User Payload Access Command
The Get User Payload Access command returns the user payload enable settings that were set using the Set User
Payload Access command.
Table 24-9, Get User Payload Access Command
byte
Request Data
1
2
Response Data
1
2
3
4
5
data field
Channel Number
[7:4] - reserved
[3:0] - Channel Number
User ID
Standard/OEM Select
[7:6] - reserved
User ID
[5:0] - User ID. 000000b = reserved
Completion Code
Standard Payload enables 1
[7:2] - reserved for standard payloads 2-7 enabled/disabled state
[1] 1b = standard payload 1 enabled (SOL)
0b = standard payload 1 disabled
[0] reserved
Standard Payload Enables 2 - reserved
OEM Payload Enables 1. For each bit:
1b = payload enabled
0b = payload disabled
[7] - OEM Payload 7 enabled/disabled
[6] - OEM Payload 6 enabled/disabled
[5] - OEM Payload 5 enabled/disabled
[4] - OEM Payload 4 enabled/disabled
[3] - OEM Payload 3 enabled/disabled
[2] - OEM Payload 2 enabled/disabled
[1] - OEM Payload 1 enabled/disabled
[0] - OEM Payload 0 enabled/disabled
OEM Payload Enables 2 - reserved
24.8 Get Channel Payload Support Command
This command enables local and remote console software to determine what payloads are enabled on the given
BMC. The command returns a bitfield indicating which Payload Type numbers can be activated on the given
channel.
315
Intelligent Platform Management Interface Specification
Table 24-10, Get Channel Payload Support Command
byte
Request Data
1
Response Data
1
2
3
4
5
6
7
8:9
data field
Channel Number
[7:4] - reserved
[3:0] - Channel Number
Completion Code
[7] = Standard payload type #7 supported
…
[0] = Standard payload type #0 supported
[7] = Standard payload type #15 (0Fh) supported
…
[0] = Standard payload type #8 supported
[7] = Session Setup payload type #7 supported
…
[0] = Session Setup payload type #0 supported
[7] = Session Setup payload type #15 (0Fh) supported
…
[0] = Session Setup payload type #8 supported
[7] = Payload type 27h (OEM7) used
…
[0] = Payload type 20h (OEM0) used
[7] = Payload type 2Fh (OEM15) used
…
[0] = Payload type 28h (OEM8) used
reserved. Return as 0000h
24.9 Get Channel Payload Version Command
This command returns version information for the given payload type. The version number has major and minor
parts. The major part of the version should only increment when there are significant changes to the payload
format, commands, or payload-specific protocols that break backward compatibility with earlier versions. The
minor part of the version increments when there are extensions to the payload format that are significant but are
backwards compatible with earlier versions under the same major version number. An example of a major change
would be a change to the payload activation process that would prevent earlier applications from activating the
given payload type. An example of a minor format version change would be the definition of commands for new
functions that did not exist under the previous format, but if unused, do not interfere with the operation of older
applications.
316
Intelligent Platform Management Interface Specification
Table 24-11, Get Channel Pa