Release Notes v2.4-1.0.4

Release Notes v2.4-1.0.4
Mellanox OFED for Linux Release Notes
Rev 2.4-1.0.4
www.mellanox.com
Rev 2.4-1.0.4
NOTE:
THIS HARDWARE, SOFTWARE OR TEST SUITE PRODUCT (“PRODUCT(S)”) AND ITS RELATED
DOCUMENTATION ARE PROVIDED BY MELLANOX TECHNOLOGIES “AS-IS” WITH ALL FAULTS OF ANY
KIND AND SOLELY FOR THE PURPOSE OF AIDING THE CUSTOMER IN TESTING APPLICATIONS THAT USE
THE PRODUCTS IN DESIGNATED SOLUTIONS. THE CUSTOMER'S MANUFACTURING TEST ENVIRONMENT
HAS NOT MET THE STANDARDS SET BY MELLANOX TECHNOLOGIES TO FULLY QUALIFY THE PRODUCT(S)
AND/OR THE SYSTEM USING IT. THEREFORE, MELLANOX TECHNOLOGIES CANNOT AND DOES NOT
GUARANTEE OR WARRANT THAT THE PRODUCTS WILL OPERATE WITH THE HIGHEST QUALITY. ANY
EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT ARE DISCLAIMED.
IN NO EVENT SHALL MELLANOX BE LIABLE TO CUSTOMER OR ANY THIRD PARTIES FOR ANY DIRECT,
INDIRECT, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES OF ANY KIND (INCLUDING, BUT NOT
LIMITED TO, PAYMENT FOR PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY FROM THE USE OF THE PRODUCT(S) AND RELATED DOCUMENTATION EVEN IF
ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
Mellanox Technologies
350 Oakmead Parkway Suite 100
Sunnyvale, CA 94085
U.S.A.
www.mellanox.com
Tel: (408) 970-3400
Fax: (408) 970-3403
Mellanox Technologies, Ltd.
Hakidma 26
Ofer Industrial Park
Yokneam 2069200
Israel
www.mellanox.com
Tel: +972 (0)74 723 7200
Fax: +972 (0)4 959 3245
© Copyright 2015. Mellanox Technologies. All Rights Reserved.
Mellanox®, Mellanox logo, BridgeX®, ConnectX®, Connect-IB®, CoolBox®, CORE-Direct®, GPUDirect®, InfiniBridge®,
InfiniHost®, InfiniScale®, Kotura®, Kotura logo, MetroX®, MLNX-OS®, PhyX®, ScalableHPC®, SwitchX®, TestX®,
UFM®, Virtual Protocol Interconnect®, Voltaire® and Voltaire logo are registered trademarks of Mellanox Technologies,
Ltd.
ExtendX™, FabricIT™, FPGADirect™, HPC-X™, Mellanox CloudX™, Mellanox Open Ethernet™, Mellanox
PeerDirect™, Mellanox Virtual Modular Switch™, MetroDX™, NVMeDirect™, Switch-IB™, Unbreakable-Link™ are
trademarks of Mellanox Technologies, Ltd.
All other trademarks are property of their respective owners.
2
Mellanox Technologies
Rev 2.4-1.0.4
Table of Contents
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
List Of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Release Update History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Chapter 1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1
1.2
1.3
Main Features in This Release. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Content of Mellanox OFED for Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Supported Platforms and Operating Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Supported Hypervisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Supported Non-Linux Virtual Machines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4
1.5
1.6
1.7
Hardware and Software Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Supported HCAs Firmware Versions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
RoCE Modes Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Chapter 2 Changes in Rev 2.4-1.0.0 From Rev 2.3-2.0.5 . . . . . . . . . . . . . . . . . . . . . . . . 11
Chapter 3 Bug Fixes History. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Chapter 4 Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
4.14
4.15
4.16
4.17
4.18
4.19
4.20
4.21
4.22
4.23
4.24
IPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ethernet Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
General Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
VGT+ Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
eIPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
XRC Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ABI Compatibility Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
System Time Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ConnectX®-3 Adapter Cards Family Known Issues. . . . . . . . . . . . . . . . . . . . . .
Verbs Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Resiliency Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Driver Start Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Performance Tools Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Performance Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Connection Manager (CM) Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SR-IOV Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Port Type Management Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Flow Steering Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Quality of Service Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Installation Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Driver Upload Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
UEFI Secure Boot Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fork Support Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ISCSI over IPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
18
19
20
21
21
22
22
22
22
23
23
24
24
25
25
27
28
28
28
29
29
29
30
Mellanox Technologies
1
Rev 2.4-1.0.4
4.25
4.26
4.27
4.28
4.29
4.30
4.31
4.32
4.33
4.34
4.35
4.36
4.37
4.38
4.39
4.40
MLNX_OFED Sources Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
InfiniBand Utilities Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
mlx5 Driver Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ethernet Performance Counters Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . .
Uplinks Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tools Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Resources Limitation Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
RoCE Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Storage Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SRP Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SRP Interop Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
DDN Storage Fusion 10000 Target Known Issues . . . . . . . . . . . . . . . . . . . . . . .
Oracle Sun ZFS Storage 7420 Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . .
iSER Initiator Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iSER Target Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ZFS Appliance Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
30
30
31
31
31
32
32
34
34
34
34
34
35
35
35
Chapter 5 Change Log History. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Chapter 6 API Change Log History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2
Mellanox Technologies
Rev 2.4-1.0.4
List Of Tables
Table 1:
Release Update History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Table 2:
Mellanox OFED for Linux Software Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Table 3:
Supported Platforms and Operating Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Table 4:
Additional Software Packages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Table 5:
Supported HCAs Firmware Versions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Table 6:
MLNX_OFED Rev 2.4-1.0.4 Compatibility Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Table 7:
RoCE Modes Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Table 8:
Changes in v2.4-1.0.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Table 9:
Fixed Bugs List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Table 10:
IPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Table 11:
Ethernet Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Table 12:
General Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Table 13:
VGT+ Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Table 14:
eIPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Table 15:
XRC Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Table 16:
ABI Compatibility Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Table 17:
System Time Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Table 18:
ConnectX®-3 Adapter Cards Family Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Table 19:
Verbs Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Table 20:
Resiliency Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Table 21:
Driver Start Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Table 22:
Performance Tools Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Table 23:
Performance Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Table 24:
Connection Manager (CM) Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Table 25:
SR-IOV Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Table 26:
Port Type Management Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Table 27:
Flow Steering Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Table 28:
Quality of Service Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Table 29:
Installation Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Table 30:
Driver Upload Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 31:
UEFI Secure Boot Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 32:
Fork Support Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 33:
ISCSI over IPoIB Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Table 34:
MLNX_OFED Sources Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Table 35:
InfiniBand Utilities Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Mellanox Technologies
3
Rev 2.4-1.0.4
Table 36:
mlx5 Driver Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Table 37:
Ethernet Performance Counters Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Table 38:
Uplinks Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Table 39:
Tools Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Table 40:
Resources Limitation Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Table 41:
RoCE Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Table 42:
Storage Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Table 43:
SRP Known Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Table 44:
SRP Interop Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Table 45:
DDN Storage Fusion 10000 Target Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Table 46:
Oracle Sun ZFS Storage 7420 Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Table 47:
iSER Initiator Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Table 48:
iSER Target Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Table 49:
ZFS Appliance Known Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Table 50:
Change Log History. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Table 51:
API Change Log History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4
Mellanox Technologies
Rev 2.4-1.0.4
Release Update History
Table 1 - Release Update History
Release
2.4-1.0.4
Date
March 12 2015
Description
Initial Release
Mellanox Technologies
5
Rev 2.4-1.0.4
1
Overview
Overview
These are the release notes of Mellanox OFED for Linux Driver, Rev 2.4-1.0.4. Mellanox OFED
is a single Virtual Protocol Interconnect (VPI) software stack and operates across all Mellanox
network adapter solutions supporting the following uplinks to servers:
1.1
•
10, 20, 40 and 56 Gb/s InfiniBand (IB)
•
10, 40 and 561 Gb/s Ethernet
•
2.5 or 5.0 GT/s PCI Express 2.0
•
8 GT/s PCI Express 3.0
Main Features in This Release
MLNX_OFED Rev 2.4-1.0.4 provides the following features:
1.2
•
Verbs performance optimization
•
EQ pool management
•
MAC Forwarding DataBase (FDB) status reporting
•
Ethtool Settings
•
Adaptive Interrupt Moderation Scheme
•
RSS support for fragmented IP datagrams
•
iSER Target driver compatible with upstream Linux IO Target
Content of Mellanox OFED for Linux
Mellanox OFED for Linux software contains the following components:
Table 2 - Mellanox OFED for Linux Software Components
Components
OpenFabrics core and ULPs
OpenFabrics utilities
MPI
PGAS
HPC Acceleration packages
Description
•
•
•
•
•
•
•
•
•
•
•
•
•
•
InfiniBand and Ethernet HCA drivers (mlx4, mlx5)
core
Upper Layer Protocols: IPoIB, SRP, iSER and iSER Initiator and Target
OpenSM: IB Subnet Manager with Mellanox proprietary Adaptive
Routing
Diagnostic tools
Performance tests
OSU MPI (mvapich2-2.0) stack supporting the InfiniBand interface
Open MPI stack 1.6.5 and later supporting the InfiniBand interface
MPI benchmark tests (OSU benchmarks, Intel MPI benchmarks, Presta)
HPC-X OpenSHMEM v2.2 supporting InfiniBand, MXM and FCA
HPC-X UPC v2.2 supporting InfiniBand, MXM and FCA
Mellanox MXM v3.0 (p2p transport library acceleration over Infiniband)
Mellanox FCA v2.5 (MPI/PGAS collective operations acceleration
library over InfiniBand)
KNEM, Linux kernel module enabling high-performance intra-node
MPI/PGAS communication for large messages
1. 56 GbE is a Mellanox propriety link speed and can be achieved while connected to Mellanox SX10XX switch series
6
Mellanox Technologies
Rev 2.4-1.0.4
Table 2 - Mellanox OFED for Linux Software Components
Components
Extra packages
Description
•
•
•
ibutils2
ibdump
MFT
Sources of all software modules
(under conditions mentioned in
the modules' LICENSE files)
except for MFT, OpenSM
plugins, ibutils2, and ibdump
Documentation
1.3
Supported Platforms and Operating Systems
The following are the supported OSs in MLNX_OFED Rev 2.4-1.0.4:
Table 3 - Supported Platforms and Operating Systems
Operating System
RHEL/CentOS 6.2
RHEL/CentOS 6.3
RHEL/CentOS 6.4
RHEL/CentOS 6.5
RHEL/CentOS 6.6
RHEL/CentOS 7.0
SLES11 SP2
SLES11 SP3
SLES12
OEL 6.4
OEL 6.5
OEL 6.6
Citrix XenServer Host 6.2
Fedora 19
Fedora 21
Ubuntu 12.04.4
Ubuntu 14.04
Ubuntu 14.10
Debian 6.0.9
Debian 7.5
Debian 7.6
Platform
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64/PPC64
x86_64
x86_64
x86_64
x86_64
i686
x86_64
x86_64
x86_64
x86_64/PPC64le
x86_64/PPC64le
x86_64
x86_64
x86_64
kernel 3.10.28a
kernel 3.11.10a
kernel 3.14.3a
kernel 3.15a
kernel 3.16a
kernel 3.17a
Mellanox Technologies
7
Rev 2.4-1.0.4
Overview
a. This kernel is supported only when using the Operating Systems stated in the table above.
If you wish to install OFED on a different kernel, you need to create a new ISO image,
using mlnx_add_kernel_support.sh script.
See the MLNX_OFED User Guide for instructions.
Upgrading MLNX_OFED on your cluster requires upgrading all of its nodes to the newest
version as well.
1.3.1
Supported Hypervisors
The following are the supported hypervisors in MLNX_OFED Rev 2.4-1.0.4:
1.3.2
•
KVM: RedHat 6.4, 6.5, 7.0, Ubuntu 14.04, Sles11SP3
•
Xen4.2
•
XenServer6.2
Supported Non-Linux Virtual Machines
The following are the supported Non-Linux (InfiniBand only) Virtual Machines in
MLNX_OFED Rev 2.4-1.0.4:
1.4
•
Windows Server 2012 R2
•
Windows Server 20012
•
Windows Server 2008 R2
Hardware and Software Requirements
The following are the hardware and software requirements of MLNX_OFED Rev 2.4-1.0.4.
•
Linux operating system
•
Administrator privileges on your machine(s)
•
Disk Space: 1GB
For the OFED Distribution to compile on your machine, some software packages of your operating system (OS) distribution are required.
To install the additional packages, run the following commands per OS:
Table 4 - Additional Software Packages
Operating System
8
Required Packages Installation Command
RHEL/OEL/
Fedora
yum install perl pciutils python gcc-gfortran libxml2-python tcsh libnl.i686
libnl expat glib2 tcl libstdc++ bc tk gtk2 atk cairo numactl
XenServer
yum install perl pciutils python libxml2-python libnl expat glib2 tcl bc libstdc++ tk
Mellanox Technologies
Rev 2.4-1.0.4
Table 4 - Additional Software Packages
Operating System
1.5
Required Packages Installation Command
SLES 11 SP2
zypper install perl pciutils python libnl-32bit libxml2-python tcsh libnl libstdc++46 expat glib2 tcl bc tklibcurl4 gtk2 atk cairo
SLES 11 SP3
zypper install perl pciutils python libnl-32bit libxml2-python tcsh libstdc++43 libnl expat glib2 tcl bc tk libcurl4 gtk2 atk cairo
Ubuntu/Debian
apt-get install perl dpkg autotools-dev autoconf libtool automake1.10 automake m4 dkms debhelper tcl tcl8.4 chrpath swig graphviz tcl-dev tcl8.4-dev
tk-dev tk8.4-dev bison flex dpatch zlib1g-dev curl libcurl4-gnutls-dev
python-libxml2 libvirt-bin libvirt0 libnl-dev libglib2.0-dev libgfortran3
automake m4
Supported HCAs Firmware Versions
MLNX_OFED Rev 2.4-1.0.4 supports the following Mellanox network adapter cards firmware
versions:
Table 5 - Supported HCAs Firmware Versions
HCA
Recommended Firmware Rev.
Additional Firmware Rev. Supported
Connect-IB®
Rev 10.10.5020
Rev 10.10.4064
ConnectX®-3 Pro
Rev 2.33.5100
Rev 2.33.5000
ConnectX®-3
Rev 2.33.5100
Rev 2.33.5000
ConnectX®-2
Rev 2.9.1000
Rev 2.9.1000
For official firmware versions please see:
http://www.mellanox.com/content/pages.php?pg=firmware_download
1.6
Compatibility
MLNX_OFED Rev 2.4-1.0.4 is compatible with the following:
Table 6 - MLNX_OFED Rev 2.4-1.0.4 Compatibility Matrix
Mellanox Product
MLNX-OS®
Grid Director™
FabricIT™ EFM
FabricIT™ BXM
Unified Fabric Manager (UFM®)
MXM
HPC-X UPC
HPC-X OpenSHMEM
FCA
HPC-X MPI
Description/Version
MSX6036 w/w MLNX-OS® version 3.3.4304a
4036 w/w Grid Director™ version 3.9.1-985
IS5035 w/w FabricIT EFM version 1.1.3000
MBX5020 w/w FabricIT BXM version 2.1.2000
v4.8
v3.2
v2.18.0
v1.8.3
v2.5 and v3.1
v1.8.3
Mellanox Technologies
9
Rev 2.4-1.0.4
Overview
Table 6 - MLNX_OFED Rev 2.4-1.0.4 Compatibility Matrix
Mellanox Product
Description/Version
MVAPICH
v2.0
a. MLNX_OFED v2.3-1.0.1 was tested with this switch however, additional switches might be supported as
well.
1.7
RoCE Modes Matrix
The following is RoCE modes matrix:
Table 7 - RoCE Modes Matrix
Software Stack /
Inbox Distribution
MLNX_OFED
Kernel.org
RHEL
SLES
Ubuntu
10
RoCE IP Based (Layer 2)
Supported as of Version
2.1-x.x.x
3.14
6.6; 7.0
12
14.04
Mellanox Technologies
RoCEv2 (Layer 3)
Supported as of Version
2.3-x.x.x
Rev 2.4-1.0.4
2
Changes in Rev 2.4-1.0.0 From Rev 2.3-2.0.5
Table 8 - Changes in v2.4-1.0.0
Category
Bug Fixes
Description
See “Bug Fixes History” on page 12.
Mellanox Technologies
11
Rev 2.4-1.0.4
3
Bug Fixes History
Bug Fixes History
Table 9 lists the bugs fixed in this release.
Table 9 - Fixed Bugs List
#
Issue
Discovered
in Release
Fixed in
Release
1.
Security
CVE-2014-8159 Fix: Prevented integer overflow in 2.0-2.0.5
IB-core module during memory registration.
2.4-1.0.4
1.
mlx5_ib
Fixed the return value of max inline received size in 2.3-2.0.1
the created QP.
2.4-1.0.0
Resolved soft lock on massive amount of user mem- 2.3-2.0.1
ory registrations
2.4-1.0.0
2.
3.
InfiniBand Counters
Occasionally, port_rcv_data and
2.3-1.0.1
port_xmit_data counters may not function properly.
2.4-1.0.0
4.
mlx4_en
LRO fixes and improvements for jumbo MTU.
2.3-2.0.1
2.4-1.0.0
5.
Fixed a crash occurred when changing the number of 2.2-1.0.1
rings (ethtool set-channels) when interface connected to netconsole.
2.4-1.0.0
6.
Fixed ping issues with IP fragmented datagrams in
MTUs 1600-1700.
2.2-1.0.1
2.4-1.0.0
7.
The default priority to TC mapping assigns all prior- 2.3-1.0.1
ities to TC0. This configuration achieves fairness in
transmission between priorities but may cause undesirable PFC behavior where pause request for priority “n” affects all other priorities.
2.4-1.0.0
Fixed an issue related to large memory regions regis- 2.3-2.0.1
tration. The problem mainly occurred on PPC systems due to the large page size, and on non PPC
systems with large pages (contig pages).
2.3-2.0.5
Fixed an issue in verbs API: fallback to glibc on con- 2.3-2.0.1
tiguous memory allocation failure
2.3-2.0.5
8.
mlx5_ib
9.
12
Description
10.
IPoIB
Fixed a memory corruption issue in multi-core system due to intensive IPoIB transmit operation.
2.3-2.0.1
2.3-2.0.5
11.
IB MAD
Fixed an issue to prevent process starvation due to
MAD packet storm.
2.3-2.0.1
2.3-2.0.5
Mellanox Technologies
Rev 2.4-1.0.4
Table 9 - Fixed Bugs List
#
Issue
Fixed in
Release
Fixed an issue which prevented the spread of events 2.3-1.0.1
among the closet NUMA CPU when only a single
RX queue existed in the system.
2.3-2.0.0
13.
Returned the CQ to its original state (armed) to pre- 2.3-1.0.1
vent traffic from stopping
2.3-2.0.0
14.
Fixed a TX timeout issue in CM mode, which
occurred under heavy stress combined with ifup/
ifdown operation on the IPoIB interface.
2.1-1.0.0
2.3-2.0.0
12.
IPoIB
Discovered
in Release
Description
15.
mlx4_core
Fixed "sleeping while atomic" error occurred when 2.3-1.0.1
the driver ran many firmware commands simultaneously.
2.3-2.0.0
16.
mlx4_ib
Fixed an issue related to spreading of completion
2.1-1.0.0
queues among multiple MSI-X vectors to allow better utilization of multiple cores.
2.3-2.0.0
Fixed an issue that caused an application to fail
when attaching Shared Memory.
2.3-1.0.1
2.3-2.0.0
Fixed dmesg warnings: "NOHZ:
2.3-1.0.1
2.3-2.0.0
Fixed erratic report of hardware clock which caused 2.1-1.0.0
bad report of PTP hardware Time Stamping.
2.3-2.0.0
Fixed race when async events arrived during driver
load.
2.3-1.0.1
2.3-2.0.0
21.
Fixed race in mlx5_eq_int when events arrived
before eq->dev was set.
2.3-1.0.1
2.3-2.0.0
22.
Enabled all pending interrupt handlers completion
before freeing EQ memory.
2.3-1.0.1
2.3-2.0.0
17.
18.
mlx4_en
local_softirq_pending 08".
19.
20.
mlx5_core
23.
mlnx.conf
Defined mlnx.conf as a configuration file in mlnxofa_kernel RPM
2.1-1.0.0
2.3-2.0.0
24.
SR-IOV
Fixed counter index allocation for VFs which
enables Ethernet port statistics.
2.3-1.0.1
2.3-2.0.0
25.
iSER
Fixed iSER DIX sporadic false DIF errors caused in 2.3-1.0.1
large transfers when block merges were enabled.
2.3-2.0.0
26.
RoCE v2
RoCE v2 was non-functional on big Endian
machines.
2.3-1.0.1
2.3-2.0.0
27.
Verbs
Fixed registration memory failure when fork was
enabled and contiguous pages or ODP were used.
2.3-1.0.1
2.3-2.0.0
Mellanox Technologies
13
Rev 2.4-1.0.4
Bug Fixes History
Table 9 - Fixed Bugs List
#
Issue
Discovered
in Release
Fixed in
Release
28.
IPoIB
Changing the GUID of a specific SR-IOV guest after 2.1-1.0.0
the driver has been started, causes the ping to fail.
Hence, no traffic can go over that InfiniBand interface.
2.3-1.0.1
29.
XRC
XRC over ROCE in SR-IOV mode is not functional 2.0-3.1.0
2.2-1.0.1
30.
mlx4_en
Fixed wrong calculation of packet true-size reporting 2.1-1.0.0
in LRO flow.
2.2-1.0.1
31.
Fixed kernel panic on Debian-6.0.7 which occurred 2.1-1.0.0
when the number of TX channels was set above the
default value.
2.2-1.0.1
32.
Fixed a crash incidence which occurred when
enabling Ethernet Time-stamping and running
VLAN traffic.
2.0-2.0.5
2.2-1.0.1
33.
IB Core
Fixed the QP attribute mask upon smac resolving
2.1-1.0.0
2.1-1.0.6
34.
mlx5_ib
Fixed a send WQE overhead issue
2.1-1.0.0
2.1-1.0.6
35.
Fixed a NULL pointer de-reference on the debug
print
2.1-1.0.0
2.1-1.0.6
36.
Fixed arguments to kzalloc
2.1-1.0.0
2.1-1.0.6
37.
mlx4_core
Fixed the locks around completion handler
2.1-1.0.0
2.1-1.0.6
38.
mlx4_core
Restored port types as they were when recovering
from an internal error.
2.0-2.0.5
2.1-1.0.0
Added an N/A port type to support port_type_array 2.0-2.0.5
module param in an HCA with a single port
2.1-1.0.0
Fixed memory leak in SR-IOV flow.
2.0-2.0.5
2.0-3.0.0
Fixed communication channel being stuck
2.0-2.0.5
2.0-3.0.0
Fixed ALB bonding mode failure when enslaving
Mellanox interfaces
2.0-3.0.0
2.1-1.0.0
43.
Fixed leak of mapped memory
2.0-3.0.0
2.1-1.0.0
44.
Fixed TX timeout in Ethernet driver.
2.0-2.0.5
2.0-3.0.0
45.
Fixed ethtool stats report for Virtual Functions.
2.0-2.0.5
2.0-3.0.0
46.
Fixed an issue of VLAN traffic over Virtual Machine 2.0-2.0.5
in paravirtualized mode.
2.0-3.0.0
47.
Fixed ethtool operation crash while interface down. 2.0-2.0.5
2.0-3.0.0
39.
40.
SR-IOV
41.
42.
14
Description
mlx4_en
Mellanox Technologies
Rev 2.4-1.0.4
Table 9 - Fixed Bugs List
#
48.
49.
Issue
IPoIB
Discovered
in Release
Description
Fixed in
Release
Fixed memory leak in Connected mode.
2.0-2.0.5
2.0-3.0.0
Fixed an issue causing IPoIB to avoid pkey value 0
for child interfaces.
2.0-2.0.5
2.0-3.0.0
Mellanox Technologies
15
Rev 2.4-1.0.4
4
Known Issues
Known Issues
The following is a list of general limitations and known issues of the various components of this
Mellanox OFED for Linux release.
4.1
IPoIB Known Issues
Table 10 - IPoIB Known Issues
Index
1.
2.
3.
4.
5.
6.
7.
8.
Description
When user increases receive/send a buffer, it might
consume all the memory when few child's interfaces
are created.
The size of send queue in Connect-IB® cards cannot
exceed 1K.
In 32 bit devices, the maximum number of child
interfaces that can be created is 16. Creating more
that, might cause out-of-memory issues.
In RHEL7.0, the Network-Manager can detect when
the carrier of one of the IPoIB interfaces is OFF and
can decide to disable its IP address.
Workaround
-
-
Set “ignore-carrier” for the corresponding
device in NetworkManager.conf.
For further information, please refer to "man
NetworkManager.conf"
IPoIB interface does not function properly if a third party application changes the PKey table.
We recommend modifying PKey tables via
OpenSM.
Fallback to the primary slave of an IPoIB bond does not work with ARP monitoring. (https://bugs.openfabrics.org/show_bug.cgi?id=1990)
Out-of memory issue might occur due to overload of To calculate the allowed memory per each
interfaces created.
IPoIB interface check the following:
• Num-rings = min(num-cores-on-thatdevice, 16)
• Ring-size = 512 (by default, it is module
parameter)
• UD memory: 2 * num-rings * ring-size *
8K
• CM memory: ring-size * 64k
• Total memory = UD mem + CM mem
Connect-IB does not reach the bidirectional line rate Optimize the IPoIB performance in
Connect-IB:
cat /sys/class/net/<interface>/
device/local_cpus > /sys/class/net/
<interface>/queues/rx-0/rps_cpus
9.
10.
16
Set the CONNECTED_MODE=yes parameter
If the CONNECTED_MODE=no parameter is set to
“no” or missing from the ifcfg file for Connect-IB® in the ifcfg file for Connect-IB® interface.
IPoIB interface then the "service network
restart" will hang.
Joining a multicast group in the SM using the
RDMA_CM API requires IPoIB to first join the
broadcast group.
Mellanox Technologies
Rev 2.4-1.0.4
Table 10 - IPoIB Known Issues (Continued)
Index
11.
Description
Workaround
Whenever the IOMMU parameter is enabled in the To avoid such issue:
kernel it can decrease the number of child interfaces • Decrease the amount of the RX receive buffers (module parameter, the default is 512)
on the device according to resource limitation.
The driver will stuck after unknown amount of child • Decrease the number of RX rings (sys/fs or
ethtool in new kernels)
interfaces creation.
• Avoid using IOMMU if not required
For further information, please see:
For KVM users:
https://access.redhat.com/site/articles/66747
Run:
http://support.citrix.com/article/CTX136517
echo 1 > /sys/module/kvm/paramehttp://www.novell.com/support/kb/
ters/
doc.php?id=7012337
allow_unsafe_assigned_interrupts
https://bugzilla.redhat.com/
show_bug.cgi?id=1044595
To make this change persist across reboots, add
the following to the /etc/modprobe.d/
kvm.conf file (or create this file, if it does not
exist):
options kvm
allow_unsafe_assigned_interrupts=1
kernel parameters
12.
System might crash in skb_checksum_help()
Use UD mode in IPoIB
while performing TCP retransmit involving packets
with 64k packet size.
A similar out to the below will be printed:
kernel BUG at net/core/dev.c:1707!
invalid opcode: 0000 [#1] SMP
RIP: 0010:[<ffffffff81448988>]
skb_checksum_help+0x148/0x160
Call Trace:
<IRQ>
[<ffffffff81448d83>]
dev_hard_start_xmit+0x3e3/0x530
[<ffffffff8144c805>]
dev_queue_xmit+0x205/0x550
[<ffffffff8145247d>]
neigh_connected_output+0xbd/0x1
…
…
13.
When InfiniBand ports are removed from the host
(e.g when changing port type from IB to Eth or
removing a card from the PCI bus) the remaining
IPoIB interface might be renamed.
To avoid it and have persistent IPoIB network
devices names for ConnectX ports, add to the
/etc/udev/rules.d/70-persistentnet.rules file:
SUBSYSTEM=="net", ACTION=="add",
DRIVERS=="?*",
ATTR{address}=="*<Port GID>",
NAME="ibN"
14.
Where N is the IPoIB required interface index
After releasing a bond interface that contains IPoIB slaves, a call trace might be printed into the dmesg.
Mellanox Technologies
17
Rev 2.4-1.0.4
Known Issues
Table 10 - IPoIB Known Issues (Continued)
Index
Description
Workaround
IPoIB interfaces are loaded without an IP address on 1. Open the "/etc/wicked/common.xml" file.
SLES 12.
2. Change:
15.
"<use-nanny>false</use-nanny>"
to
"<use-nanny>true</use-nanny>"
3. Run:
# systemctl restart wickedd.service wicked
# ifup all
In RHEL7.0, running ifdown then ifup on an interface after changing CONNECTED_MODE in its
ifcfg file, will cause the interface bring up to fail.
Clone interfaces receive a duplicated IPv6 address
when a child interface with the same PKey (a.k.a
clone interface) is used for all the clones.
eth_ipoib module is not loaded after reloading the
ib_ipoib module.
16.
17.
18.
In Ubuntu and Debian, the default of the
recv_queue_size and send_queue_size is 128
according to the io_mmu issue.
19.
4.2
Reload the driver "/etc/init.d/openibd
restart" or reboot the system.
-
To restart the IPoIB driver, run "/etc/
init.d/openibd restart". Do not restart
it by manually restarting each module.
-
Ethernet Known Issues
Table 11 - Ethernet Known Issues
Index
1.
Description
Workaround
When creating more than 125 VLANs and SR-IOV mode is enabled, a kernel warning message will be
printed indicating that the native VLAN is created
but will not work with RoCE traffic.
kernel warning: mlx4_core 0000:07:00.0:
vhcr command ALLOC_RES (0xf00) slave:0
in_param 0x7e in_mod=0x107, op_mod=0x1
failed with error:0, status -28
2.
3.
4.
Kernel panic might occur during FIO splice in kernels
before 2.6.34-rc4.
baff42a net: Fix
oops from
tcp_collapse() when
using splice()
In PPC systems when QoS is enabled a harmless
Kernel DMA mapping error messages might appear
in kernel log (IOMMU related issue).
Transmit timeout might occur on RH6.3 as a result of lost interrupt (OS issue). In this case, the following message will be shown in dmesg:
do_IRQ: 0.203 No irq handler for vector
(irq -1)
18
Use kernel v2.6.34-rc4
which provides the following solution:
Mellanox Technologies
Rev 2.4-1.0.4
Table 11 - Ethernet Known Issues (Continued)
Index
5.
6.
7.
8.
Description
Mixing ETS and strict QoS policies for TCs in
40GbE ports may cause inaccurate results in bandwidth division among TCs.
Creating a VLAN with user priority >= 4 on
ConnectX®-2 HCA is not supported.
Affinity hints are not supported in Xen Hypervisor
(an irqblancer issue). This causes a non-optimal IRQ
affinity.
Reboot might hang in SR-IOV when using the
“probe_vf” parameter with many Virtual Functions. The following message is logged in the kernel
log:
Workaround
-
To overcome this issues, run:
set_irq_affinity.sh eth<x>
-
"waiting for eth to become free. Usage
count =1"
9.
10.
11.
12.
4.3
In ConnectX®-2, RoCE UD QP does not include
VLAN tags in the Ethernet header
VXLAN may not be functional when configured
over Linux bridge in RH7.0 or Ubuntu14.04.
The issue is within the bridge modules in those kernels. In Vanilla kernels above 3.16 issues is fixed.
In RH6.4, ping may not work over VLANs that are configured over Linux bridge when the bridge has a
mlx4_en interface attached to it.
The interfaces LRO needs to be set to "OFF" manu- Run: ethtool -K ethX lro off
ally when there is a bond configured on Mellanox
interfaces with a Bridge over that bond.
General Known Issues
Table 12 - General Known Issues
Index
1.
Description
Workaround
On ConnectX-2/ConnectX-3 Ethernet adapter cards, N/A.
Please use the GUID value returned by the fabthere is a mismatch between the GUID value
ric/driver utilities (not 0xfffff).
returned by firmware management tools and that
returned by fabric/driver utilities that read the GUID
via device firmware (e.g., using ibstat). Mlxburn/
flint return 0xffff as GUID while the utilities return a
value derived from the MAC address. For all driver/
firmware/software purposes, the latter value should
be used.
Mellanox Technologies
19
Rev 2.4-1.0.4
4.4
Known Issues
VGT+ Known Issues
Table 13 - VGT+ Known Issues
Index
1.
2.
3.
4.
Description
Before adding a VLAN on the VM, the parent inter- face should be brought up. Otherwise VLAN creation will fail and the following message will be
presented:
"Fail to register network rule."
Bringing down and up the parent interface with
VLANs configured over it, may result in traffic over
VLANs being lost.
On some of the OSes the callback
ndo_vlan_rx_add/kill_vid returns void, therefore ethX.Y is created. However, if VLAN Y is not
listed in the set of the allowed VLANs, no traffic
will pass.
When un-tagged traffic is not allowed, the below
message will appear on Dom0 after driver restart on
DomU:
mlx4_core 0000:16:00.0: vhcr command
ALLOC_RES (0xf00) slave:1 in_param 0x0
in_mod=0x207, op_mod=0x1 failed with
error:0, status -1
20
Mellanox Technologies
Workaround
Rev 2.4-1.0.4
4.5
eIPoIB Known Issues
Table 14 - eIPoIB Known Issues
Index
1.
Description
Workaround
On rare occasions, upon driver restart the following message is shown in the dmesg:
'cannot create duplicate filename '/
class/net/eth_ipoib_interfaces'
2.
3.
4.
5.
6.
7.
4.6
No indication is received when eIPoIB is non functional.
eIPoIB requires libvirtd, python
eIPoIB supports only active-backup mode for bonding.
eIPoIB supports only VLAN Switch Tagging (VST)
mode on guests.
IPv6 is currently not supported in eIPoIB
eIPoIB cannot run when Flow Steering is enabled
Run ‘ps -ef | grep ipoibd’ to verify its
functionality.
-
XRC Known Issues
Table 15 - XRC Known Issues
Index
1.
Description
Legacy API is deprecated, thus when recompiling
applications over MLNX_OFED v2.0-3.x.x, warnings such as the below are displayed.
Workaround
-
rdma.c:1699: warning:
'ibv_open_xrc_domain' is deprecated
(declared at /usr/include/infiniband/
ofa_verbs.h:72)
rdma.c:1706: warning:
'ibv_create_xrc_srq' is deprecated
(declared at /usr/include/infiniband/
ofa_verbs.h:89)
2.
3.
4.
These warnings can be safely ignored.
XRC is not functional in heterogeneous clusters
containing non Mellanox HCAs.
XRC options do not work when using qperf tool.
Use perftest instead
Out-of memory issue might occur due to overload of XRC receive QP with non zero receive queue size
created.
XRC QPs do not have receive queues.
Mellanox Technologies
21
Rev 2.4-1.0.4
4.7
Known Issues
ABI Compatibility Known Issues
Table 16 - ABI Compatibility Known Issues
Index
MLNX_OFED v2.3-1.0.1 is not ABI compatible
with previous MLNX_OFED/OFED versions.
1.
4.8
Description
Workaround
Recompile the application over the new
MLNX_OFED version
System Time Known Issues
Table 17 - System Time Known Issues
Index
Workaround
Loading the driver using the openibd script when no InfiniBand vendor module is selected (for example
mlx4_ib), may cause the execution of the
/sbin/start_udev’ script.
In RedHat 6.x and OEL6.x this may change the local
system time.
1.
4.9
Description
ConnectX®-3 Adapter Cards Family Known Issues
Table 18 - ConnectX®-3 Adapter Cards Family Known Issues
Index
Using RDMA READ with a higher value than 30
SGEs in the WR might lead to "local length error".
1.
4.10
Description
Workaround
Do not set the value of SGEs higher than 30
when RDMA READ is used.
Verbs Known Issues
Table 19 - Verbs Known Issues
Index
1.
2.
22
Description
Workaround
Using libnl1_1_3~26 or earlier, requires
ibv_create_ah protection by a lock for multithreaded applications.
In MLNX_OFED v2.4-1.0.0, if several CQEs are
When getting an event, poll the CQ until it is
received on a CQ, they will be coalesced and a user- empty.
space event will be triggered only once.
Mellanox Technologies
Rev 2.4-1.0.4
4.11
Resiliency Known Issues
Table 20 - Resiliency Known Issues
Index
1.
2.
3.
4.
5.
6.
4.12
Description
Reset Flow can run on XenServer 6 only after the
active user space applications running verbs are terminated.
SR-IOV non persistent configuration (such as VGT,
VST, Host assigned GUIDs, and QP0-enabled VFs)
may be lost upon Reset Flow.
Upon Reset Flow or after running restart driver,
Ethernet VLANs are lost.
Restarting the driver or running
connectx_port_config when Reset Flow is running might result in a kernel panic
Networking configuration (e.g. VLANs, IPv6)
should be statically defined in order to have them set
after Reset Flow as of after restart driver.
After recovering from an EEH event, mlx5_core/
mlx4_core unload may fail due to a bug in some kernel versions. The bug is fixed in Kernel 3.15
Workaround
-
Reset Admin configuration post Reset Flow
Reset the VLANs using the ifup command.
-
-
-
Driver Start Known Issues
Table 21 - Driver Start Known Issues
Index
1.
2.
3.
Description
"Out-of-memory" issues may rise during drivers
load depending on the values of the driver module
parameters set (e.g. log_num_cq).
When reloading/starting the driver using the /etc/
init.d/openibd the following messages are displayed if there is a third party RPM or driver
installed:
"Module mlx4_core does not belong to
MLNX_OFED"
or
"Module mlx4_core belong to <rpm name> which is
not a part of MLNX_OFED"
Occasionally, when trying to repetitively reload the
NES hardware driver on SLES11 SP2, a soft lockups
occurs that required reboot.
Workaround
-
Remove the third party RPM/non
MLNX_OFED drivers directory, run:
"depmod" and then rerun "/etc/init.d/
openibd restart"
-
Mellanox Technologies
23
Rev 2.4-1.0.4
Known Issues
Table 21 - Driver Start Known Issues (Continued)
Index
Description
Workaround
In ConnectX-2, if the driver load succeeds, the
informative message below is presented conveying
the below limitations:
• If port type is IB the number of maximum supported VLs is 4
• If port type is ETH then the maximum priority
for VLAN tagged is 3
4.
“mlx4_core 0000:0d:00.0: command
SET_PORT (0xc) failed:
in_param=0x120064000, in_mod=0x2,
op_mod=0x0, fw status = 0x40”
“openibd start” unloads kernel modules that were
loaded from initrd/initramfs upon boot.
This affects only kernel modules which come with
MLNX_OFED and are included in initrd/initramfs.
If a Lustre storage is used, it must be fully unloaded 1.Unmount any mounted Lustre storages:
# umount<lustre_mount_point>
before restarting the driver or rebooting the
2.Unload all Lustre modules:
machine, otherwise machine might get stuck/panic.
5.
6.
# lustre_rmmod
4.13
Performance Tools Known Issues
Table 22 - Performance Tools Known Issues
Index
perftest package in MLNX_OFED v2.2-1.0.1 and
onwards does not work with older versions of the
driver.
1.
4.14
Description
Workaround
-
Performance Known Issues
Table 23 - Performance Known Issues
Index
1.
2.
3.
24
Description
On machines with irqbalancer daemon turned off,
the default InfiniBand interrupts will be routed to a
single core which may cause overload and software/
hardware lockups.
Out of the box throughput performance in
Ubuntu14.04 is not optimal and may achieve results
below the line rate in 40GE link speed.
UDP receiver throughput may be lower then
expected, when running over mlx4_en Ethernet
driver.
This is caused by the adaptive interrupt moderation
routine, which sets high values of interrupt coalescing, causing the driver to process large number of
packets in the same interrupt, leading UDP to drop
packets due to overflow in its buffers.
Mellanox Technologies
Workaround
Execute the following script as root:
set_irq_affinity.sh <interface or
IB device> [2nd interface or IB
device]
For additional performance tuning, please refer
to Performance Tuning Guide.
Disable adaptive interrupt moderation and set
lower values for the interrupt coalescing manually.
ethtool -C <eth>X adaptive-rx off
rx-usecs 64 rx-frames 24
Values above may need tuning, depending the
system, configuration and link speed.
Rev 2.4-1.0.4
Table 23 - Performance Known Issues (Continued)
Index
4.
4.15
Description
Workaround
Performance degradation might occur when bonding Ethernet interfaces on Centos 6.5
Connection Manager (CM) Known Issues
Table 24 - Connection Manager (CM) Known Issues
Index
1.
4.16
Description
Workaround
When 2 different ports have identical GIDs, the CM All ports must have different GIDs.
might send its packets on the wrong port.
SR-IOV Known Issues
Table 25 - SR-IOV Known Issues
Index
1.
2.
3.
4.
5.
6.
7.
Description
When using legacy VMs with MLNX_OFED 2.x
hypervisor, you may need to set the
'enable_64b_cqe_eqe' parameter to zero on the
hypervisor.
It should be set in the same way that other module
parameters are set for mlx4_core at module load
time.
For example, add “options mlx4_core
enable_64b_cqe_eqe=0” as a line in the file /
etc/modprobe.d/mlx4_core.conf.
InfiniBand counters are not available in the VM.
mlx4_port1_mtu sysfs entry shows a wrong MTU
number in the VM.
When at least one port is configured as InfiniBand,
and the num_vfs is provided but the probe_vf is
not, HCA initialization fails.
When working with a bonding device to enslave the
Ethernet devices in active-backup mode and failover
MAC policy in a Virtual Machine (VM), establishment of RoCE connections may fail.
Attaching or detaching a Virtual Function on
SLES11 SP3 to a guest Virtual Machine while the
mlx4_core driver is loaded in the Virtual Machine
may cause a kernel panic in the hypervisor.
When detaching a VF without shutting down the
driver from a VM and reattaching it to another VM
with the same IP address for the Mellanox NIC,
RoCE connections will fail
Workaround
-
Use both the num_vfs and the probe_vf in
the modprobe line.
Unload the module mlx4_ib and reload it in the
VM.
Unload the mlx4_core module in the hypervisor before attaching or detaching a function to
or from the guest.
Shut down the driver in the VM before detaching the VF.
Mellanox Technologies
25
Rev 2.4-1.0.4
Known Issues
Table 25 - SR-IOV Known Issues (Continued)
Index
8.
9.
10.
11.
Description
Enabling SR-IOV requires appending the
“intel_iommu=on” option to the relevant OS in
file /boot/grub/grub.conf.
Without that SR-IOV cannot be loaded.
On various combinations of Hypervisor/OSes and
Guest/OSes, an issue might occur when attaching/
detaching VFs to a guest while that guest is up and
running.
When working with SR-IOV in Xen-4.2 virtualization platform, only the built-in xen_pciback driver
should be loaded.
The xen_pciback module in dom0 should not be
loaded, as loading them simultaneously may cause
interrupts loss and cause the driver to enter the reset
flow.
The known PCI BDFs for all VFs in kernel command line should be specified by adding xen-pci-
Workaround
-
Attach/detach VFs to/from a VM only while
that VM is down.
-
-
back.hide
12.
13.
14.
15.
16.
26
For further information, please refer to
http://wiki.xen.org/wiki/Xen_PCI_Passthrough
The qemu version (2.0) provided inbox with Ubuntu
14.04 does not work properly when more than 2
VMs are run over an Ubuntu 14.04 Hypervisor.
SR-IOV UD QPs are forced by the Hypervisor to
use the base GID (i.e., the GID that the VF sees in its
GID entry at its paravirtualized index 0). This is
needed for security, since UD QPs use Address Vectors, and any GID index may be placed in such a
vector, including indices not belonging to that VF.
Attempting to attach a PF to a VM when SR-IOV is
already enabled on that PF may result in a kernel
panic.
osmtest on the Hypervisor fails when SR-IOV is
enabled. However, only the test fails, OpenSM will
operate correctly with the host. The failure reason is
that if an mcg is already joined by the host, a subsequent join request for that group succeeds automatically (even if the join parameters in the request are
not correct). This success does no harm.
If a VM does not support PCI hot plug, detaching an
mlx4 VF and probing it to the hypervisor may cause
the hypervisor to crash.
Mellanox Technologies
-
-
-
-
-
Rev 2.4-1.0.4
4.17
Port Type Management Known Issues
Table 26 - Port Type Management Known Issues
Index
1.
2.
3.
4.
Description
OpenSM must be stopped prior to changing the port
protocol from InfiniBand to Ethernet.
After changing port type using
connectx_port_config interface ports’ names
can be changed. For example. ib1 -> ib0 if port1
changed to be Ethernet port and port2 left IB.
A working IP connectivity between the RoCE
devices is required when creating an address handle
or modifying a QP with an address vector.
IPv4 multicast over RoCE requires the MGID format to be as follow ::ffff:<Multicast IPv4
Workaround
Use udev rules for persistent naming configuration.
For further information, please refer to the
User Manual
-
-
Address>
5.
6
IP routable RoCE does not support Multicast Listener Discovery (MLD) therefore, multicast traffic
over IPv6 may not work as expected.
DIF: When running IO over FS over DM during
unstable ports, block layer BIOs merges may cause
false DIF error.
Mellanox Technologies
27
Rev 2.4-1.0.4
4.18
Known Issues
Flow Steering Known Issues
Table 27 - Flow Steering Known Issues
Index
Description
Workaround
1.
Flow Steering is disabled by default in firmware ver- To enable it, set the parameter below as follow:
log_num_mgm_entry_size should set
sion < 2.32.5100.
2.
IPv4 rule with source IP cannot be created in
SLES 11.x OSes.
RFS does not support UDP.
When working in DMFS:A0 mode and VM/hypervisor is MLNX_OFED 2.3-x.x.x, the second side
(hypervisor/VM respectively) should be
MLNX_OFED 2.3-x.x.x as well.
to -1
3.
4.
4.19
-
Quality of Service Known Issues
Table 28 - Quality of Service Known Issues
Index
Workaround
QoS is not supported in XenServer, Debian 6.0 and 6.2 with uek kernel
When QoS features are not supported by the kernel, mlnx_qos tool may crash.
1.
2.
4.20
Description
Installation Known Issues
Table 29 - Installation Known Issues
Index
1.
2.
3.
4.
5.
6.
28
Description
Workaround
When upgrading from an earlier Mellanox OFED
version, the installation script does not stop the earlier version prior to uninstalling it.
Upgrading from the previous OFED installation to
this release, does not unload the kernel module
ipoib_helper.
Installation using Yum does not update HCA firmware.
"--total-vfs <0-63>" installation parameter is
no longer supported
Stop the old OFED stack (/etc/init.d/
openibd stop) before upgrading to this new
version.
Reboot after installing the driver.
See “Updating Firmware After Installation” in
OFED User Manual
Use '--enable-sriov' installation parameter to burn firmware with SR-IOV support. The
number of virtual functions (VFs) will be set to
16.
For further information, please refer to the
User Manual.
When using bonding on Ubuntu OS, the "ifenslave" package must be installed.
On PPC systems, the ib_srp module is not installed If your system does not require the ibmvscsi
by default since it breaks the ibmvscsi module.
module, run the mlnxofedinstall script with the
"--with-srp" flag.
Mellanox Technologies
Rev 2.4-1.0.4
4.21
Driver Upload Known Issues
Table 30 - Driver Upload Known Issues
Index
1.
4.22
Description
Workaround
"openibd stop" can sometime fail with the error: Re-run "openibd stop"
Unloading ib_cm [FAILED]
ERROR: Module ib_cm is in use by
ib_ipoib
UEFI Secure Boot Known Issues
Table 31 - UEFI Secure Boot Known Issues
Index
1.
Description
Workaround
On RHEL7 and SLES12, the following error is dis- For further information, please refer to the
played in dmesg if the Mellanox's x.509 Public Key User Manual section "Enrolling Mellanox's
x.509 Public Key On your Systems".
is not added to the system:
[4671958.383506] Request for unknown
module key 'Mellanox Technologies signing key:
61feb074fc7292f958419386ffdd9d5ca999e4
03' err -11
2
3
4.23
This error can be safely ignored as long as Secure
Boot is disabled on the system.
Ubuntu12 requires update of user space
open-iscsi to v2.0.873
The initiator does not respect interface parameter
while logging in.
Configure each interface on a different subnet.
Fork Support Known Issues
Table 32 - Fork Support Known Issues
Index
1.
Description
Workaround
Fork support from kernel 2.6.12 and above is avail- able provided that applications do not use threads.
fork() is supported as long as the parent process
does not run before the child exits or calls exec().
The former can be achieved by calling
wait(childpid), and the latter can be achieved by
application specific means. The Posix system() call
is supported.
Mellanox Technologies
29
Rev 2.4-1.0.4
4.24
Known Issues
ISCSI over IPoIB Known Issues
Table 33 - ISCSI over IPoIB Known Issues
Index
Workaround
When working with ISCSI over IPoIB, LRO must be disabled (even if IPoIB is set to connected mode)
due to a a bug in older kernels which causes a kernel
panic.
1.
4.25
Description
MLNX_OFED Sources Known Issues
Table 34 - MLNX_OFED Sources Known Issues
Index
Workaround
MLNX_OFED includes the OFED source RPM
packages used as a build platform for kernel code
but does not include the sources of Mellanox proprietary packages.
1.
4.26
Description
InfiniBand Utilities Known Issues
Table 35 - InfiniBand Utilities Known Issues
Index
When running the ibdiagnet check nodes_info on
the fabric, a warning specifying that the card does
not support general info capabilities for all the
HCAs in the fabric will be displayed.
ibdump does not work when IPoIB device managed
Flow Steering is OFF and at least one of the ports is
configured as InfiniBand.
1.
2.
4.27
Description
Workaround
Run ibdiagnet --skip nodes_info
Enable IPoIB Flow Steering and restart the
driver.
For further information, please refer to
MLNX_OFED User Manual section Enable/
Disable Flow Steering.
mlx5 Driver Known Issues
Table 36 - mlx5 Driver Known Issues
Index
1.
2.
3.
30
Description
Atomic Operations in Connect-IB® are fully supported on big-endian machines (e.g. PPC). Their
support is limited on little-endian machines (e.g.
x86)
EEH events that arrive while the mlx5 driver is loading may cause the driver to hang.
The mlx5 driver can handle up to 5 EEH events per
hour.
Mellanox Technologies
Workaround
-
If more events are received, cold reboot the
machine.
Rev 2.4-1.0.4
4.28
Ethernet Performance Counters Known Issues
Table 37 - Ethernet Performance Counters Known Issues
Index
1.
2.
3.
4.29
Description
Workaround
In a system with more than 61 VFs, the 62nd VF and onwards is assigned with the SINKQP counter, and
as a result will have no statistics, and loopback prevention functionality for SINK counter.
Since each VF tries to allocate 2 more QP counter for its RoCE traffic statistics, in a system with less
than 61 VFs, if there is free resources it receives new
counter otherwise receives the default counter which
is shared with Ethernet. In this case RoCE statistics
is not available.
In ConnectX®-3, when we enable function-based
loopback prevention for Ethernet port by default
(i.e., based on the QP counter index), the dropped
self-loopback packets increase the IfRxErrorFrames/
Octets counters.
Uplinks Known Issues
Table 38 - Uplinks Known Issues
Index
1.
4.30
Description
Workaround
On rare occasions, ConnectX®-3 Pro adapter card Restart the driver
may fail to link up when performing parallel detect
to 40GbE.
Tools Known Issues
Table 39 - Tools Known Issues
Index
1.
Description
Workaround
Running ibdump in InfiniBand mode with firmware Run ibdump with firmware v2.33.5000 and
older than v2.33.5000, may cause the server to hang higher
due to a firmware issue.
Mellanox Technologies
31
Rev 2.4-1.0.4
4.31
Known Issues
Resources Limitation Known Issues
Table 40 - Resources Limitation Known Issues
Index
Description
Workaround
The device capabilities reported may not be reached as it depends on the system on which the device is
installed and whether the resource is allocated in the
kernel or the userspace.
mlx4_core can allocate up to 64 MSI-X vectors, an MSI-X vector per CPU.
Setting more IP addresses than the available GID
entries in the table results in failure and the
"update_gid_table error message is displayed:
1.
2.
3.
GID table of port 1 is full. Can't add
<address>" message.
Registering a large amount of Memory Regions
(MR) may fail because of DMA mapping issues on
RHEL 7.0.
Occasionally, a user process might experience some
memory shortage and not function properly due to
Linux kernel occupation of the system’s free memory for its internal cache.
4.
5.
-
To free memory to allow it to be allocated in a
user process, run the drop_caches procedure
below.
Performing the following steps will cause the
kernel to flush and free pages, dentries and
inodes caches from memory, causing that
memory to become free.
Note: As this is a non-destructive operation
and dirty objects are not freeable, run `sync'
first.
•
To free the pagecache:
echo 1 > /proc/sys/vm/
drop_caches
•
To free dentries and inodes:
echo 2 > /proc/sys/vm/
drop_caches
•
To free pagecache, dentries and inodes:
echo 3 > /proc/sys/vm/
drop_caches
4.32
RoCE Known Issues
Table 41 - RoCE Known Issues
Index
1.
2.
32
Description
Not configuring the Ethernet devices or independent Restart the driver
VMs with a unique IP address in the physical port,
may result in RoCE GID table corruption.
If RDMA_CM is not used for connection manage- ment, then the source and destination GIDs used to
modify a QP or create AH should be of the same
type - IPv4 or IPv6.
Mellanox Technologies
Workaround
Rev 2.4-1.0.4
Table 41 - RoCE Known Issues (Continued)
Index
3.
4.
5.
6.
7.
Description
On rare occasions, the driver reports a wrong GID
table (read from /sys/class/infiniband/mlx4_*/ports/
*/gids/*). This may cause communication problems.
MLNX_OFED v2.1-1.0.0 and onwards is not
interoperable with older versions of MLNX_OFED.
Since the number of GIDs per port is limited to 128,
there cannot be more than the allowed IP addresses
configured to Ethernet devices that are associated
with the port. Allowed number is:
• “127” for a single function machine
• “15” for a hypervisor in a multifunction
machine
• “(127-15)/n” for a guest in a multifunction
machine (where n is the number of virtual functions)
A working IP connectivity between the RoCE
devices is required when creating an address handle
or modifying a QP with an address vector.
IPv4 multicast over RoCE requires the MGID format to be as follow ::ffff:<Multicast IPv4
Workaround
-
-
-
-
Address>
8.
9.
10.
IP routable RoCE does not support Multicast Listener Discovery (MLD) therefore, multicast traffic
over IPv6 may not work as expected.
Using GID index 0 (the default GID) is possible
only if the matching IPv6 link local address is configured on the net device of the port. This behavior
is possible even though the default GID is configured regardless the presence of the IPv6 address.
Using IPv6 link local address (GID0) when VLANs are configured is not supported.
Mellanox Technologies
33
Rev 2.4-1.0.4
4.33
Known Issues
Storage Known Issues
Table 42 - Storage Known Issues
Index
Older versions of rescan_scsi_bus.sh may not
recognize some newly created LUNs.
1.
4.34
Description
Workaround
If encountering such issues, it is recommended
to use the '-c' flag.
SRP Known Issues
Table 43 - SRP Known Issues
Index
MLNX_OFED SRP installation breaks the
ibmvstgt and ibmvscsi symbol resolution in
RHEL7.0
1.
4.35
Description
Workaround
-
SRP Interop Known Issues
Table 44 - SRP Interop Known Issues
Index
Workaround
The driver is tested with Storage target vendors rec- ommendations for multipath.conf extensions (ZFS,
DDN, TMS, Nimbus, NetApp).
1.
4.36
Description
DDN Storage Fusion 10000 Target Known Issues
Table 45 - DDN Storage Fusion 10000 Target Known Issues
Index
Workaround
DDN does not accept non-default P_Key connection establishment.
1.
4.37
Description
Oracle Sun ZFS Storage 7420 Known Issues
Table 46 - Oracle Sun ZFS Storage 7420 Known Issues
Index
1.
34
Description
Ungraceful power cycle of an initiator connected
with Targets DDN, Nimbus, NetApp may result in
temporary "stale connection" messages when initiator reconnects.
Mellanox Technologies
Workaround
Rev 2.4-1.0.4
4.38
iSER Initiator Known Issues
Table 47 - iSER Initiator Known Issues
Index
1.
Description
On SLES OSs, the ib_iser module does not load
on boot.
Workaround
Add a dummy interface using iscsiadm:
• # iscsiadm -m iface -I ib_iser •
2
3
4
5
4.39
Ubuntu12 requires update of user space
open-iscsi to v2.0.873
The initiator does not respect interface parameter
while logging in.
iSCSID v2.0.873 can enter an endless loop on bind
error.
iSCSID may hang if target crashes during logout
sequence (reproducible with TCP).
o new
# iscsiadm -m iface -I ib_iser o update -n iface.transport_name
-v ib_iser
Configure each interface on a different subnet.
-
iSER Target Known Issues
Table 48 - iSER Target Known Issues
Index
1.
2
3
4
4.40
Description
Currently only the following OSs are supported:
RHEL/ContOS 7.0, SLES12, Ubuntu14.04.
Stress login/logout from multiple initiators may
cause iSER target to panic.
RHEL/CentOS 7.0: Discovery over RDMA is not
supported.
ib_isert is unavailable on custom kernels after running the mlnx_add_kernel_support.sh script.
Workaround
1.Add "isert=y" to the
mlnx_add_kernel_support.sh script
after "cat << EOF > ofed.conf".
2.Use the updated script to build MLNX_OFED
for the custom kernel.
ZFS Appliance Known Issues
Table 49 - ZFS Appliance Known Issues
Index
1.
Description
Connection establishment occurs twice which may
cause iSER to log a stack trace.
Workaround
-
Mellanox Technologies
35
Rev 2.4-1.0.4
5
Change Log History
Change Log History
Table 50 - Change Log History
Release
2.4-1.0.0
Category
mlx4_en netdevice Ethtool
Description
Added support for Ethtool speed control and advertised link mode.
Added ethtool txvlan control for setting ON/OFF hardware TX
VLAN insertion: ethtool -k txvlan [on/off]
Ethtool report on port parameters improvements.
Ethernet TX packet rate improvements.
RoCE
RoCE uses now all available EQs and not only the 3 legacy EQs.
InfiniBand
IRQ affinity hints are now set when working in InfiniBand mode.
Virtualization
VXLAN fixes and performance improvements.
libmlx4 & libmlx5
Improved message rate of short massages.
libmlx5
Added ConnectX®-4 device (4114) to the list of supported devices
(hca_table),
Storage
Added iSER Target driver.
Ethernet net-device
New adaptive interrupt moderation scheme to improve CPU utilization.
RSS support of fragmented IP datagram.
2.3-2.0.5
Connect-IB Virtual Function
Added Connect-IB Virtual Function to the list of supported
devices.
mlx5_core
Added the following files under /sys/class/infiniband/
mlx5_0/mr_cache/:
• rel_timeout: Defines the minimum allowed time between
•
2.3-2.0.1
36
the last MR creation to the first MR released from the cache.
When rel_timeout = -1, MRs are not released from the
cache
rel_imm: Triggers the immediate release of excess MRs from
the cache when set to 1. When all excess MRs are released
from the cache, rel_imm is reset back to 0.
Bug Fixes
See “Bug Fixes History” on page 12.
Bug Fixes
See “Bug Fixes History” on page 12.
Mellanox Technologies
Rev 2.4-1.0.4
Table 50 - Change Log History
Release
2.3-2.0.0
2.3-1.0.1
Category
Description
Connect-IB
Added Suspend to RAM (S3).
Reset Flow
Added Enhanced Error Handling for PCI (EEH), a recovery strategy for I/O errors that occur on the PCI bus.
Register Contiguous Pages
Added the option to ask for a specific address when the register
memory is using contiguous page.
mlx5_core
Moved the mr_cache subtree from debugfs to mlx5_ib while
preserving all its semantics.
InfiniBand Utilities
Updated the ibutils package. Added to the ibdiagnet tool the
"ibdiagnet2.mlnx_cntrs" option to enable reading of Mellanox
diagnostic counters.
Bug Fixes
See “Bug Fixes History” on page 12.
OpenSM
Added Routing Chains support with Minhop/UPDN/FTree/DOR/
Torus-2QoS
Added double failover elimination.
When the Master SM is turned down for some reason, the Standby
SM takes ownership over the fabric and remains the Master SM
even when the old Master SM is brought up, to avoid any unnecessary re-registrations in the fabric.
To enable this feature, set the “master_sm_priority” parameter
to be greater than the “sm_priority” parameter in all SMs in the
fabric. Once the Standby SM becomes the Master SM, its priority
becomes equal to the “master_sm_priority”. So that additional SM handover is avoided. Default value of the
master_sm_priority is 14.
To disable this feature, set the “master_sm_priority” in
opensm.conf to 0.
Added credit-loop free unicast/multicast updn/ftree routing
Added multithreaded Minhop/UPDN/DOR routing
RoCE
Added IP routable RoCE modes.
For further information, please refer to the MLNX_OFED User
Manual.
Installation
Added apt-get installation support.
Mellanox Technologies
37
Rev 2.4-1.0.4
Change Log History
Table 50 - Change Log History
Release
Category
Ethernet
Description
Added support for arbitrary UDP port for VXLAN.
From upstream 3.15-rc1 and onward, it is possible to use arbitrary
UDP port for VXLAN.
This feature requires firmware version 2.32.5100 or higher.
Additionally, the following kernel configuration option
CONFIG_MLX4_EN_VXLAN=y must be enabled.
MLNX_OFED no longer changes the OS sysctl TCP parameters.
Added Explicit Congestion Notification (ECN) support
Added Flow Steering: A0 simplified steering support
Added RoCE v2 support
2.3-1.0.1
InfiniBand
Network
Added Secure host to enable the device to protect itself and the
subnet from malicious software.
Added User-Mode Memory Registration (UMR) to enable the
usage of RDMA operations and to scatter the data at the remote
side through the definition of appropriate memory keys on the
remote side.
Added On-Demand-Paging (ODP), a technique to alleviate much
of the shortcomings of memory registration.
Added Masked Atomics operation support
Added Checksum offload for packets without L4 header support
Added Memory re-registration to allow the user to change attributes of the memory region.
Resiliency
Added Reset Flow for ConnectX®-3 (+SR-IOV) support.
SR-IOV
Added Virtual Guest Tagging (VGT+), an advanced mode of Virtual Guest Tagging (VGT), in which a VF is allowed to tag its own
packets as in VGT, but is still subject to an administrative VLAN
trunk policy.
Ethtool
Added Cable EEPROM reporting support
Disable/Enable ethernet RX VLAN tag striping offload via ethtool
128 Byte Completion Queue Entry (CQE)
Non-Linux Virtual
Machines
38
Mellanox Technologies
Added Windows Virtual Machine over Linux KVM Hypervisor
(SR-IOV with InfiniBand only) support
Rev 2.4-1.0.4
Table 50 - Change Log History
Release
Rev 2.2-1.0.1
Rev 2.2-1.0.1
Category
Description
mlnxofedinstall
32-bit libraries are no longer installed by default on 64-bit OS.
To install 32-bit libraries use the '--with-32bit' installation
parameter.
openibd
Added pre/post start/stop scripts support.
For further information, please refer to section “openibd Script” in
the MLNX_OFED User Manual.
Reset Flow
Reset Flow is not activated by default. It is controlled by the
mlx4_core'internal_err_reset' module parameter.
InfiniBand Core
Asymmetric MSI-X vectors allocation for the SR-IOV hypervisor
and guest instead of allocating 4 default MSI-X vectors. The maximum number of MSI-X vectors is num_cpu for port
ConnectX®-3 has 1024 MSI-X vectors, 28 MSI-X vectors are
reserved.
• Physical Function - gets the number of MSI-X vectors according to the pf_msix_table_size (multiple of 4 - 1) INI
parameter
• Virtual Functions – the remaining MSI-X vectors are spread
equally between all VFs, according to the num_vfs
mlx4_core module parameter
Ethernet
Ethernet VXLAN support for kernels 3.12.10 or higher
Power Management Quality of Service: when the traffic is active,
the Power Management QoS is enabled by disabling the CPU
states for maximum performance.
Ethernet PTP Hardware Clock support on kernels/OSes that support it
Verbs
Added additional experimental verbs interface.
This interface exposes new features which are not integrated yet in
to the upstream libibverbs. The Experimental API is an extended
API therefor, it is backward compatible, meaning old application
are not required to be recompiled to use MLNX-OFED v2.2-1.0.1.
Performance
Out of the box performance improvements:
• Use of affinity hints (based on NUMA node of the device) to
indicate the IRQ balancer daemon on the optimal IRQ affinity
• Improvement in buffers allocation schema (based on the hint
above)
• Improvement in the adaptive interrupt moderation algorithm
Mellanox Technologies
39
Rev 2.4-1.0.4
Change Log History
Table 50 - Change Log History
Release
Rev 2.1-1.0.6
Category
Description
IB Core
Added allocation success verification process to ib_alloc_device.
dapl
dapl is recompiled with no FCA support.
openibd
Added the ability to bring up child interfaces even if the parent's
ifcfg file is not configured.
libmlx4
Unmapped the hca_clock_page parameter from
mlx4_uninit_context.
scsi_transport_srp
scsi_transport_srp cannot be cleared up when rport reconnecting
fails.
mlnxofedinstall
Added support for the following parameters:
• '--umad-dev-na'
• '--without-<package>'
Rev 2.1-1.0.6
Content Packages
Updates
The following packages were updated:
• bupc to v2.2-407
• mstflint to v3.5.0-1.1.g76e4acf
• perftest to v2.0-0.76.gbf9a463
• hcoll to v2.0.472-1
• Openmpi to v1.6.5-440ad47
• dapl to v2.0.40
Rev 2.1-1.0.0
EoIB
EoIB is supported only in SLES11SP2 and RHEL6.4.
eIPoIB
eIPoIB is currently at GA level.
Connect-IB®
Added the ability to resize CQs.
IPoIB
Reusing DMA mapped SKB buffers: Performance improvements
when IOMMU is enabled.
mlnx_en
Added reporting autonegotiation support.
Added Transmit Packet Steering (XPS) support.
Added reporting 56Gbit/s link speed support.
Added Low Latency Socket (LLS) support.
Added check for dma_mapping errors.
eIPoIB
40
Mellanox Technologies
Added non-virtual environment support.
Rev 2.4-1.0.4
Table 50 - Change Log History
Release
Rev 2.0-3.0.0
Category
Description
Operating Systems
Additional OS support:
• SLES11SP3
• Fedora16, Fedora17
Drivers
Added Connect-IB™ support
Installation
Added ability to install MLNX_OFED with SR-IOV support.
Added Yum installation support
EoIB
EoIB (at beta level) is supported only in SLES11SP2 and RHEL6.4
mlx4_core
Modified module parameters to associate configuration values with
specific PCI devices identified by their bus/device/function value
format
mlx4_en
Reusing DMA mapped buffers: major performance improvements
when IOMMU is enabled
Added Port level QoS support
IPoIB
Reduced memory consumption
Limited the number TX and RX queues to 16
Default IPoIB mode is set to work in Datagram, except for Connect-IB™ adapter card which uses IPoIB with Connected mode as
default.
Rev 2.0-3.0.0
Storage
iSER (at GA level)
Mellanox Technologies
41
Rev 2.4-1.0.4
Change Log History
Table 50 - Change Log History
Release
Rev 2.0-2.0.5a
Category
Description
Virtualization
SR-IOV for both Ethernet and InfiniBand (at Beta level)
Ethernet Network
RoCE over SR-IOV (at Beta level)
eIPoIB to enable IPoIB in a Para-Virtualized environment (at
Alpha level)
Ethernet Performance Enhancements (NUMA related and others)
for 10G and 40G
Ethernet Time Stamping (at Beta level)
Flow Steering for Ethernet and InfiniBand. (at Beta level)
Raw Eth QPs:
• Checksum TX/RX
• Flow Steering
InfiniBand Network
Contiguous pages:
• Internal memory allocation improvements
• Register shared memory
• Control objects (QPs, CQs)
Installation
YUM update support
VMA
OFED_VMA integration to a single branch
Storage
iSER (at Beta level) and SRP
Operating Systems
Errata Kernel upgrade support
API
VERSION query API: library and headers
Counters
64bit wide counters (port xmit/recv data/packets unicast/mcast)
a. SR-IOV, Ethernet Time Stamping and Flow Steering are ConnectX®-3 HCA capability.
42
Mellanox Technologies
Rev 2.4-1.0.4
6
API Change Log History
Table 51 - API Change Log History
Release
2.4-1.0.0
Rev 2.3-1.0.1
Name
libibverbs
libibverbs
Description
Added the following verbs interfaces:
•
•
•
•
ibv_create_flow
ibv_destroy_flow
ibv_exp_use_priv_env
ibv_exp_setenv
•
ibv_exp_rereg_mr - Added new API for memory
region re-reintegration (For futher information, please
refer to MLNX_OFED User Manual)
Added to the experimental API ibv_exp_post_send
the following opcodes:
•
• IBV_EXP_WR_EXT_MASKED_ATOMIC_CMP_AND_SWP
• IBV_EXP_WR_EXT_MASKED_ATOMIC_FETCH_AND_ADD
• IBV_EXP_WR_NOP
and these completion opcodes:
• IBV_EXP_WC_MASKED_COMP_SWAP
• IBV_EXP_WC_MASKED_FETCH_ADD
Rev 2.2-1.0.1
libibverbs
The following verbs changed to align with upstream libibverbs:
• ibv_reg_mr - ibv_access_flags changed.
• ibv_post_send - opcodes and send flags
•
•
•
•
changed and wr fields removed (task, op, dc
and bind_mw)
ibv_query_device - capability flags
changed.
ibv_poll_cq - opcodes and wc flags changed.
ibv_modify_qp - mask bits changed
ibv_create_qp_ex - create_flags field
removed.
The following verbs removed to align with upstream libibverbs:
• ibv_bind_mw
• ibv_post_task
• ibv_query_values_ex
• ibv_query_device_ex
• ibv_poll_cq_ex
• ibv_reg_shared_mr_ex
• ibv_reg_shared_mr
• ibv_modify_cq
• ibv_create_cq_ex
• ibv_modify_qp_ex
Mellanox Technologies
43
Rev 2.4-1.0.4
API Change Log History
Table 51 - API Change Log History
Release
Rev 2.2-1.0.1
Rev 2.1-1.0.0
Name
Verbs Experimental API
Description
The following experimental verbs added (replacing the
removed extended verbs):
• ibv_exp_bind_mw
• ibv_exp_post_task
• ibv_exp_query_values
• ibv_exp_query_device
• ibv_exp_poll_cq
• ibv_exp_reg_shared_mr
• ibv_exp_modify_cq
• ibv_exp_create_cq
• ibv_exp_modify_qp
New experimental verbs:
• ibv_exp_arm_dct
• ibv_exp_query_port
• ibv_exp_create_flow
• ibv_exp_destroy_flow
• ibv_exp_post_send
• ibv_exp_reg_mr
• ibv_exp_get_provider_func
Dynamically Connected (DC) The following verbs were added:
• struct ibv_dct *ibv_exp_create_dct(struct
•
•
Verbs Extension API:
Verbs extension API defines
OFA APIs extension scheme
to detect ABI compatibility
and enable backward and forward compatibility support.
44
Mellanox Technologies
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
ibv_context *context, struct
ibv_exp_dct_init_attr *attr)
int ibv_exp_destroy_dct(struct ibv_dct
*dct)
int ibv_exp_query_dct(struct ibv_dct *dct,
struct ibv_exp_dct_attr *attr)
ibv_post_task
ibv_query_values_ex
ibv_query_device_ex
ibv_create_flow
ibv_destroy_flow
ibv_poll_cq_ex
ibv_reg_shared_mr_ex
ibv_open_xrcd
ibv_close_xrcd
ibv_modify_cq
ibv_create_srq_ex
ibv_get_srq_num
ibv_create_qp_ex
ibv_create_cq_ex
ibv_open_qp
ibv_modify_qp_ex
Rev 2.4-1.0.4
Table 51 - API Change Log History
Release
Rev 2.1-1.0.0
Rev 2.0-3.0.0
Rev 2.0-2.0.5
Name
Description
Verbs Experimental API:
Verbs experimental API
defines MLNX-OFED APIs
extension scheme which is
similar to the “Verbs extension API”. This extension
provides a way to introduce
new features before they are
integrated into the formal
OFA API and to the upstream
kernel and libs.
XRC
•
•
•
•
•
ibv_exp_create_qp
ibv_exp_query_device
ibv_exp_create_dct
ibv_exp_destroy_dct
ibv_exp_query_dct
Libibverbs - Raw QPs
QP types IBV_QPT_RAW_PACKET and IBV_QPT_RAW_ETH
are not supported
• Added Contiguous pages support
• Added function ibv_reg_shared_mr
• The enumeration IBV_M_WR_CALC was renamed to
The following verbs have become deprecated:
• struct ibv_xrc_domain *ibv_open_xrc_domain
• struct ibv_srq *ibv_create_xrc_srq
• int ibv_close_xrc_domain
• int ibv_create_xrc_rcv_qp
• int ibv_modify_xrc_rcv_qp
• int ibv_query_xrc_rcv_qp
• int ibv_reg_xrc_rcv_qp
• int ibv_unreg_xrc_rcv_qp
Libibverbs - Extended speeds • Missing the ext_active_speed attribute from the
struct ibv_port_attr
• Removed function ibv_ext_rate_to_int
• Added functions ibv_rate_to_mbps and
mbps_to_ibv_rate
Libibverbs - Contiguous
pages
Libmverbs
IBV_M_WR_CALC_SEND
•
•
•
•
The enumeration IBV_M_WR_WRITE_WITH_IMM was
added
In the structure ibv_m_send_wr, the union wr.send was
renamed to wr.calc_send and wr.rdma was added
The enumerations
IBV_M_WQE_CAP_CALC_RDMA_WRITE_WITH_IMM was
added
The following enumerations were renamed:
• From IBV_M_WQE_SQ_ENABLE_CAP to
IBV_M_WQE_CAP_SQ_ENABLE
• From IBV_M_WQE_RQ_ENABLE_CAP to
IBV_M_WQE_CAP_RQ_ENABLE
• From IBV_M_WQE_CQE_WAIT_CAP to
IBV_M_WQE_CAP_CQE_WAIT
• From IBV_M_WQE_CALC_CAP to
IBV_M_WQE_CAP_CALC_SEND
Mellanox Technologies
45
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement