Changes and New Features

Category

Description

Rev 23.10.50000 (DRV 23.10.26252)

Health Syndrome, DDR

Added a health syndrome indicating that a hardware failure has occurred.

The following is the health syndrome message: PCI data poisoned error has been received while fetching ICM (synd = 18).

Link Speed Detection and Report

Added support for detecting and reporting Link Speed of 800G, especially for OSFP cables in the PDDR log.

VM RoCEv2 Traffic Restriction

Limited VM RoCEv2 traffic to a specific IPv6 source address. The feature can be controlled via mlx5cmd using the new subcommand "-RoceRestrict".

Additionally, the following new counters were added for the dropped packets by this feature in the "Mellanox WinOF-2 VF Port Traffic" counters:

  • RoCE Restrict Packets Discarded

  • RoCE Restrict Bytes Discarded

For further information, see RoCE Restrict Configuration Utility and Mellanox WinOF-2 VF Port Traffic.

Counters

Added new RDMA VF diagnostic counters. These counters are disabled by default, to enable them use the EnableVFRdmaCounters key.

For further information, see Mellanox WinOF-2 VF Diagnostics.

NicHealthMonitor Utility

Added a new utility to estimate the driver and the firmware health by analyzing diagnostic counters and checking the event log for events logged by the driver.

For further information see NicHealthMonitor Utility.

Bug Fixes

See Bug Fixes.

© Copyright 2023, NVIDIA. Last updated on Nov 3, 2023.