Bug Fixes in This Version

4366800

Description: Creating a packet filtering rule with match criteria on certain unsupported fields (e.g., specific VXLAN-GPE flags) does not return an error as expected.

Keyword: Packet filtering; DOCA Flow

Detected in version: 3.2.0

4802407

Description: DOCA Flow does not support masked matching of the VXLAN Reserved 1 (RSVD1) field.

Keyword: VXLAN RSVD1; Masked Matching; Hardware Offloading

Detected in version: 3.2.0

4802408

Description: DOCA Flow does not support matching the ICMPv6 Identifier field in Echo Request and Echo Reply packets.

Keyword: ICMPv6; echo request/reply; packet offloading

Detected in version: 3.2.0

4636365

Description: The names of several telemetry counters and parameters in the Multi-Rack Connectivity (MRC) service are non-intuitive or ambiguous.

Keyword: MRC Telemetry; counters; EV probes; ZTR RTTCC

Detected in version: 3.2.0

4742653

Description: DOCA Argus (APSH) fails to enumerate Virtual Memory Areas (VMA) on Arm64 architectures.

Keyword: Compatibility; enumeration; mm_struct

Detected in version: 3.2.0

4704261

Description: OVS user space is not supported on RHEL 10.0.

Keyword: Open vSwitch; DPDK EAL Panic; ovs-vsctl

Detected in version: 3.2.0

4823654

Description: When running the NVQual test suite on Socket Direct (SD) systems using synthetic mode, the tool fails to correctly identify the BDF (Bus:Device.Function) of the upstream PCIe ports.

Keyword: NVQual; Socket Direct; Synthetic Mode; PCIe BDF

Detected in version:

4687096

Description: The doca_upf_accel application is not supported on GB300 (ConnectX-8) systems, returning a Failed to allocate dynamic extension table data error.

Keyword: GB300; SEGV; Dynamic Extension Table

Detected in version: 3.2.0

4739125

Description: The doca_gpunetio_verbs_write_bw sample fails with a DOCA_ERROR_DRIVER error.

Keyword: GPUNetIO; sample

Detected in version: 3.1.0

4564535

Description: The DPU may fail to boot successfully after performing a firmware reset operation (mlxfwreset) following a software or firmware upgrade.

Keyword: Hang; unresponsive; boot

Detected in version: 2.9.4

4564535

Description: Under high flow-insertion rates on BlueField-3, pipe resize operations may time out (exceeding 10 seconds). This can cause OVS threads (ovs_rcu) to block, indicating a stall in the hardware offload process.

Keyword: OVS; Pipe Resize; Timeout

Detected in version: 3.1.0

4564538

Description: A pipe resize operation on BlueField-3, often triggered by dynamic flow limits, may fail. This can result in memory allocation errors and severe delays (10+ seconds), leading to a loss of network connectivity, such as BGP peering sessions dropping.

Keyword: Pipe Resize; Flow Limit; Memory Allocation

Detected in version: 3.1.0

4621473

Description: A race condition during flow deletion may cause application instability or incorrect counter values.

Keyword: Flow; delete

Detected in version: 3.1.0

4366800

Description: Creating a packet filtering rule with match criteria on certain unsupported fields (e.g., specific VXLAN-GPE flags) does not return an error as expected.

Keyword: Packet filtering; DOCA Flow

Detected in version: 3.1.0

4627850

Description: The RTC timestamp is not correctly parsed in the doca_telemetry_diag sample.

Keyword: RTC; PTP; DOCA Telemetry

Detected in version: 3.1.0

4548383

Description: The doca_flow_header_ib_bth structure includes fields for which match and action operations are not supported. Supported fields:

• Match and action: opcode, dest_qp

• Match only: AckReq (in flags1)

• Action only: psn

Keyword: DOCA Flow

Detected in version: 3.1.0

4560966

Description: Enabling the RX Rate flag in the ZTR RTTCC algorithm causes compilation to fail because the private DOCA PCC device header is not included.

Keyword: DOCA PCC; Zero-touch RoCE; round-trip time congestion control

Detected in version: 3.1.0

4223392

Description: CT miss pipe receives configured IP version only.

Keyword: DOCA Flow

Detected in version: 3.1.0

4834414

Description: Fixed an issue on RHEL systems where installing chkconfig after mlnx-ofa_kernel (pulled in by packages such as doca-ofed and doca-host) could fail during installation. In this sequence, chkconfig may be installed indirectly as a dependency of a package that places init scripts under /etc/init.d, but the install would error out because /etc/init.d already existed as a directory (whereas chkconfig expects to manage it as a symlink). This fix prevents the /etc/init.d directory/symlink conflict that caused the RPM unpacking failure.

Keyword: Installation

Detected in version: 3.2.1

4743484 / 4563750

Description: Fixed an issue where, in LAG mode, a firmware reset race could cause the driver to hang during unload while waiting for UMR completion. The fix ensures sys_error notifications are registered early and kept through device unregister so error events reach the bond device during teardown.

Keyword: LAG mode

Detected in version: 3.2.0

4858587 / 4858692 / 4858693 / 4858694

Description: Fixed soft lockup warnings that could occur while releasing large UMEMs by periodically yielding the CPU during the release flow.

Keyword: Soft lockup warnings

Detected in version: 3.2.0

4742666 / 4070154 / 4237797 / NVbug 5672643

Description: Fixed a kernel crash that could occur if the devlink userspace process was terminated while switching the device to switchdev mode. In this scenario, the kernel could dereference invalid state and trigger an oops, potentially causing immediate node failure on GB300 systems.

Keyword: devlink, switchdev mode

Detected in version: 3.2.0

4426799 / 4491971 / NVbug 5253854

Description: Fixed an issue where devlink operations were not synchronized with fw_reset, which could lead to various failures. Added a mechanism to coordinate fw_reset with devlink operations.

Keyword: devlink, fw_reset

Detected in version: 3.2.0

4775049 / 4540881 / 47779851 / 47779858 / 4830934 / 4865514

Description: Fixed an issue where the driver could incorrectly initialize polling-only CQs (e.g., ASO CQs), allowing them to receive EQ interrupts in rare cases. The CQ initialization flow was updated to prevent interrupts for polling-only CQs.

Keyword: CQs, EQ

Detected in version: 3.2.0

4802194

Description: Fixed an overflow issue where the max-rate value was stored in a unsigned 8-bit integer type , causing values above 255 to wrap (e.g., 257 became 1).

Keyword: max-rate value

Detected in version: 3.2.0

4830982

4748884

Description: Fixed an issue that prevented using more than 8 cores to set SF CPU affinity. This restriction has now been removed.

Keyword: SF CPU affinity

Detected in version: 3.2.0

4845039

Description: MPI-dependent DOCA applications failed to compile on OpenEuler due to missing development tools.

Keywords: OpenEuler; OpenMPI; compilation

Detected in version: 4.13.0

4845039

Description: A kernel call trace could occur during BFB updates if the host-side ConnectX has a pending firmware upgrade.

Keywords: Kernel

Detected in version: 4.13.0

4829571

Description: The DELETE key does not remove characters following the cursor in both UEFI Shell and GRUB environments.

Keywords: UEFI Shell; GRUB

Detected in version: 4.13.0

4764682

Description: mlnx-sf fails to apply -r parameter (disable RoCE for SF) as it was removed from mlxdevm.

Keywords: mlnx-sf; RoCE

Detected in version: 4.13.0

4824650

Description: BFB installer (bfb-install) does not display error and warning messages during installation.

Keywords: Bfb-install; warning messages; logging

Detected in version: 4.13.0

4693948

4782014

Description: Installation of Ubuntu 24.04 on BlueField-3 eMMC devices is not supported in the 3.2.0 BFB release.

Keywords: Installation; Ubuntu

Detected in version: 4.13.0

4826162

Description: On systems equipped with multiple BlueField-3 DPUs, the doca-installer --compare command may intermittently fail to retrieve the CEC firmware version from one of the cards.

Reported in version: 25.10

4786813

Description: Fixed an issue where the DPA kernel used unsafe ICM access during process creation/modification, which could cause the DPA kernel to hang during FLR.

Keywords: DPA kernel, FLR

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4804664 / 4806969

Description: Fixed an issue in the User Debugger “query caps” where it returned only the number of capabilities, not the capability bitmap.

Keywords: User Debugger “query caps”

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4346373 / 4073037 / 4371711 / NBbug 5150535

Description: Fixed an issue where, after an unexpected cable removal, the current-limiter enable could assert unexpectedly, and a subsequent module removal could trigger an error message. The BlueField-3 PRS indication was updated to use Power Good instead of Over-Current.

Keywords: virtio emulation, live migration, live update

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4819986 / 4827433 / 4832172 / 4832178

Description: Fixed an issue where the server could occasionally become stuck during live migration or while updating the Virtio emulation device.

Keywords: virtio emulation, live migration, live update

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4484662

Description: Fixed an issue where mlxlink reported 0 values for SNR (media and host) due to incorrect local port mapping in firmware and an incorrect page number used by MFT.

Keywords: mlxlink

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4450570 / 4780432 / 4780433

Description: Fixed an issue where the root complex sent MCTP-over-PCI messages before a BDF was assigned, causing responses to be sent with BDF 0. The fix ensures that MCTP messages routed by ID are ignored until a valid BDF is assigned.

Keywords: MCTP-over-PCI, BDF, MCTP messages

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4809134 / 4824635

Description: Fixed an issue where the steering tables were not updated after enabling partial Spectrum-X capabilities (BTH.AR) via LLPD.

Keywords: Steering tables, LLDP

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4773333 / 4711949

Description: Fixed an issue where, in older versions and under specific configurations, the MEMIC BAR size could be unaligned to 64 bytes. The software now treats the address as aligned down to a 64-byte boundary.

Keywords: MEMIC BAR size

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4785853 / 4653684 / 4796563 / 4797987

Description: Fixed an issue where the virtio software controller could encounter an infinite FLR loop if the host shut down during the software live-upgrade procedure in configurations with separate power supplies (i.e., the DPU remains powered when the host is powered off).

Keywords: Virtio, Software Live Upgrade, FLR

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4794198 / 4796352 / 4797985

Description: Fixed an issue where, in host-aware mode, the firmware did not block the hotplug FSM transition from HOTPLUG to HOTUNPLUG_PREPARE, which could cause a hotplug device to enter an unexpected state.

Keywords: Hotunplug, Virtio

Detected in version: 32.47.1026

Fixed in Release: 32.48.1000

4450570 / 4780432 / 4780433

Description: Fixed an issue where the root complex sent MCTP-over-PCI messages before a BDF was assigned, causing responses to be sent with BDF 0. The fix ensures that MCTP messages routed by ID are ignored until a valid BDF is assigned.

Keywords: MCTP-over-PCI, BDF, MCTP messages

Discovered in Version: 24.47.1026

Fixed in Release: 24.48.1000