Bug Fixes History

NVIDIA ConnectX-5 Adapter Cards Firmware Release Notes v16.35.3502 LTS

The following table provides a list of bugs fixed in previous versions. For a list of bug fixed from the current version, see Bug Fixes.

Internal Ref.

Issues

3199650

Description: A physical link failure between switches while a SHARP job is running and utilizing the link can cause one of the switches to become invalid for further SHARP jobs. This can result in either a "No resource" response for new SHARP job requests or in jobs getting stuck.
The bug fix requires SHARP version 3.2.

Keywords: SHARP

Discovered in Version: 31.2010.4010

Fixed in Version: 31.2010.4102

3245821

Description: In case of an AR group table set request, the ARN mask is flushed for group that has an active pFRN timer.

Keywords: PFRN

Discovered in Version: 31.2010.4010

Fixed in Version: 31.2010.4102

3253717

Description: mask_force_clear_timeout timer in pFRN feature was not functional (the mask was not cleared when the timer expired).

Keywords: PFRN

Discovered in Version: 31.2010.4010

Fixed in Version: 31.2010.4102

3242209

Description: Set PFRN mad did not return error on wrong inputs in mask_clear_timer and mask_force_clear_timer fields.

Keywords: PFRN

Discovered in Version: 31.2010.4010

Fixed in Version: 31.2010.4102

3143685

Description: The switch does not return SN or PN when trying to call via mlxlink or ibdiagnet.

Keywords: Transceiver

Discovered in Version: 31.2010.2300

Fixed in Version: 31.2010.4010

3174239

Description: On rare occasions, traps were not properly repressed, which caused redundant traps to be sent multiple times.

Keywords: Traps

Discovered in Version: 31.2010.3118

Fixed in Version: 31.2010.4010

3002314

Description: On rare occasion, when port is configured to mloop toggle may cause link to not rise.

Keywords: Optic in Mloop

Discovered in Version: 31.2010.2110

Fixed in Version: 31.2010.3118

3127727

Description: On rare occasion, when egress port is split to two, the egress port may get stuck due to wrong Fast Path configuration.

Keywords: Switch Hang, Fast Path, Split

Discovered in Version: 31.2010.3004

Fixed in Version: 31.2010.3118

3082569

Description: In some traffic patterns involving small packets, the PortRcvErrors counter may mistakenly count events of local physical errors due to an internal flow in the hardware that involves link packets.

Keywords: Counters

Discovered in Version: 31.2010.2246

Fixed in Version: 31.2010.3004

3085427

Description: On rare occasions, SHARP semaphore may remain locked on a port following an event of a port link down or an application crash.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.3004

3011581

Description: On rare occasions, job failures with SharpError trap may be experienced as a result of previous jobs that have failed.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.3004

3000602

Description: After disconnecting MMS4X00-NL* cable and connecting Ultron cable to the same port, ports fails to link up.

Keywords: Cables

Discovered in Version: 31.2010.2110

Fixed in Version: 31.2010.2300

3060122

Description: In the event of link fault of a link between root switch and non-root switch during the run of a job, the next job run on the non-root switch may fail.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.2300

2923464

Description: When using MMS4X00-NL Optical module, on rare occasions port that is in NDR speed may get stuck and stay in Polling state.

Keywords: NDR, Optical Module

Discovered in Version: 31.2010.1404

Fixed in Version: 31.2010.2246

2859363

Description: When using NVIDIA Quantum-2 systems in Auto-Neg mode, NDR speed in one lane (1x) is not supported.

Keywords: Auto-Negotiation

Discovered in Version: 31.2010.1310

Fixed in Version: 31.2010.2246

3033131

Description: The number of flows changed from 2 to 1, as intended.

Keywords: SHARPv3

Discovered in Version: 31.2010.2110

Fixed in Version: 31.2010.2246

2972388

Description: Running of concurrent jobs may lead to states where jobs unexpectedly terminate or get stuck.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.2110

2982113

Description: On rare occasions, job resource cleanup may fail.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.2110

2971339

Description: During high load scenarios, performance degradation may be experienced.

Keywords: SHARPv3

Discovered in Version: 31.2010.2036

Fixed in Version: 31.2010.2110

2849215

Description: On NVIDIA Quantum-2 switches, when working with MFA7U10-H0xx cables, if one of the ports in a cage is disabled at the time of initialization by user configuration, reenabling the port will require toggling the link (i.e. enable → disable → enable).

Keywords: NVIDIA Quantum-2, Cables

Discovered in Version: 31.2010.1310

Fixed in Version: 31.2010.2036

2890632

Description: On NVIDIA Quantum-2 systems, changing the Optical module rate was not allowed.

Keywords: Optical Modules

Discovered in Version: 31.2010.1310

Fixed in Version: 31.2010.2036

2885798

Description: In NVIDIA Quantum-2 systems, effective errors may occur with short Copper cable MCP4Y10-N00B.

Workaround: N/A

Discovered in Version: 31.2010.1310

Fixed in Version: 31.2010.2036

2910161

Description: In auto-negotiation flow, using copper cables when toggling both port's sides may cause the port to get stuck on rare occasions.

Keywords: Auto-Negotiation, Copper Cables

Discovered in Version: 31.2010.1310

Fixed in Version: 31.2010.2036

© Copyright 2023, NVIDIA. Last updated on May 23, 2023.