Bug Fixes in This Version

5118076

Description: When multiple virtual machines are started simultaneously with SNAP virtio-blk SR-IOV VFs passed through to them, some VFs may fail to start or pass through successfully.

Keyword: Virtio-blk; SR-IOV; VF passthrough; VFIO; QEMU; multiple VMs

Detected in version: 3.2.2

4944595

Description: When multiple virtual machines are started simultaneously with SNAP virtio-blk SR-IOV VFs passed through to them, some VFs may fail to start or pass through successfully.

Keyword: Virtio-blk; SR-IOV; VF passthrough; VFIO; QEMU; multiple VMs

Detected in version: 3.2.2

4926867

Description: SNAP may crash while recovering a virtio-blk VF used by a VFIO passthrough VM.

Keyword: Virtio-blk; VFIO passthrough; VM

Detected in version: 3.2.2

4796353

Description: If a host server is shut down while a SNAP Virtio-blk Live Update is in progress, the host may fail to power back on.

Keyword: Virtio; LU; FLR

Detected in version: 3.2.2

5131191 / NVbug 6409233

Description: Fixed an ABI break and segmentation fault in mlx5dv_create_flow() that occurred when supplying multiple actions. This is resolved by introducing a new symbol for dynamic stride support, while maintaining backward compatibility for affected DOCA releases.

Keyword: ABI break

Reported in version: 3.2.2

5042525 / 5036525

Description: Fixed an issue on RPM-based systems running DKMS version 3.4.0 or later (e.g., installed via CUDA) where DOCA-Host DKMS packages failed to execute DKMS commands during post-installation. This failure, caused by the dkms executable moving from /usr/sbin/dkms to /usr/bin/dkms, previously resulted in all DOCA-Host kernel drivers and nvidia_peermem (from nvidia-kernel-open-dkms) becoming unavailable.

Keyword: DOCA-Host DKMS packages

Reported in version: 3.2.2

5000492 / 4850615

Description: Fixed an issue where Debian/Ubuntu systems installed an old version of the openibd init script instead of the latest one. The package now installs the correct, up-to-date openibd script, fixing boot behavior and ensuring proper module reload and IB interface creation.

Keyword: Debian/Ubuntu, openibd init

Reported in version: 3.2.2

5017387 / 4848424

Description: Fixed an issue where a BlueField-3 host SF could remain stuck in the inactive+attached state and fail to reactivate after host restart flows.

Keyword: inactive+attached state

Reported in version: 3.2.2

5004557 / 5001782

Description: Fixed a NULL pointer dereference in the mlx5_core driver that could cause a kernel panic and report PCIe completion-timeout errors on ConnectX-8 bare-metal systems.

Keyword: mlx5_core driver, kernel panic, PCIe completion-timeout errors

Reported in version: 3.0.0

4999890 / 4994944

Description: Fixed a race condition that occurred when ovs-vswitchd queried netdev statistics over netlink during NIC bringup. Previously, this race condition led to a kernel NULL pointer dereference in the mlx5_core driver (specifically within mlx5e_fold_sw_stats64).

Keyword: Race condition; ovs-vswitchd, netlink; NIC bringup

Reported in version: 2.9.4

4057748 / 5057168

Description: Fixed an issue where rdma_resolve_addr() returned an ENODEV error on an IPoIB interface when the destination was an IPv4 broadcast address.

Keyword: RDMA

Reported in version: 3.2.0

4962099 / 4956297

Description: Fixed a race in the ib_core kernel module during concurrent address resolution that could produce a zero destination MAC (DMAC), causing silent packet loss and RDMA retry-exceeded errors.

Keyword: ib_core kernel

Reported in version: 3.2.2

4989952

Description: When upgrading BlueField-3 from firmware version 4.7.1.4 to 4.13.0, an extra reboot may be required to apply the DDR speed setting of 5200 MT/s. This affects cards which derive their DDR speed from an EEPROM value

Keyword: DDR; upgrade; OPN

Reported in version: 4.13.2

5118076

Description: After installing BFB on BlueField-2 in DPU mode, the message DPU-BMC RF credentials not found appears in the RShim log buffer. During installation, the BFB installer cannot reach the BMC at 192.168.240.1:443, causing certificate upload to fail and leaving Redfish credentials unconfigured.

Keyword: BFB install; DPU mode; BMC; Redfish; BlueField-2; RShim

Reported in version: 4.13.2

5031933 / 4234285 / 5031924 / 4983438

Description: During BlueField boot, the BlueField BIOS/UEFI may fail to acquire a DHCP IP address from the BMC for Redfish communication. This would cause the BIOS to bypass the Redfish step and to ignore any configuration changes in Redfish.

Keyword: DHCP; BIOS; UEFI

Reported in version: 4.13.2

4892178

Description: Triggering a software reset on a BlueField-3 DPU in NIC mode while Host VF traffic is active can cause the NIC subsystem to crash and fail its automated health recovery.

Keyword: SW_RESET; NIC mode; host VF

Reported in version: 4.13.2

4926609

Description: Secure boot database handling causes unnecessary non-volatile storage checks during normal boot, adding about 20 seconds to startup on affected systems.

Keyword: Boot time; UEFI; BlueField-2

Reported in version: 4.13.2

5081115

Description: Due to hardware limitations, excessively polling RSHIM logs or Redfish task status during an active BFB installation can cause the RSHIM USB_TIMEOUT to exceed its 40-second limit, leading to degraded performance or task failure.

Keyword: RSHIM; USB_TIMEOUT; BFB installation

Reported in version: 4.13.2

4982738

Description: BlueField-3 PA0 cards may ship configured with either 5200 MT/s or 5600 MT/s DDR memory speed. The default DDR memory speed has been set to 5200 MT/s to ensure consistency.

Keyword: DDR; 5200; 5600; memory speed

Reported in version: 4.13.2

4546931

Description: Running mlxfwreset on BlueField-3 with Ubuntu 24.04 may trigger an ext4 journal warning call trace (ext4_journal_check_start+0xc8/0x120); no functional impact observed, safe to ignore.

Keyword: mlxfwreset; ext4; call trace

Reported in version: 4.13.2

4811361

Description: When using OobUpdate to install a firmware bundle on a BlueField DPU that has no Linux OS installed, the installation may wait unnecessarily for the DPU to reach the OS-running state, resulting in excessively long completion times.

Keyword: OobUpdate; installation time; no OS

Reported in version: 4.13.2

4976718

Description: If a BMC task such as a log dump is started after the BMC firmware update task has been initiated but before monitoring has attached to it, the monitoring logic may track the other task instead of the firmware update task, which can cause the BMC reboot to fail.

Reported in version: 25.10-LTSU2

4917779

Description: Initiating an Arm GracefulReset can cause the BMC's Redfish UpdateService to incorrectly report its state as UnavailableOffline.

Reported in version: 25.10-LTSU2

4944048

Description: When upgrading or downgrading between the 25.10-LTSU2 and 26.04 releases, repeated BMC reboots may, in rare cases, cause the profile-manager service to fail due to a malformed JSON file. This failure triggers a core dump of the service, populates core dump logs, and causes the golden-image service to become unresponsive.

Reported in version: 25.10-LTSU2

5096810

Description: The Processor.Reset Redfish action incorrectly requires only the Login privilege for POST requests to /redfish/v1/Systems/<str>/Processors/<str>/Actions/Processor.Reset. This action should require the ConfigureComponents privilege per the Redfish privilege registry, as the Login privilege is granted to all authenticated users, including those in the ReadOnly role.

Reported in version: 25.10-LTSU2

5071650

Description: In rare cases, the EROT version string may contain space characters instead of the actual version characters, making the version appear partial. Rebooting the BMC reloads the string correctly.

Reported in version: 25.10-LTSU2

5017603

Description: During BFB installations via Redfish, the task state may change to "Exception" before the specific error message is appended to the HTTP response payload. This results in incomplete error logs on the initial poll following a failure.

Reported in version: 25.10-LTSU2

4994932

Description: In rare cases, the PLDM firmware update process in NIC mode may fail during the activation phase due to a transient communication error on the device.

Reported in version: 25.10-LTSU2

4966479

Description: In rare cases, accessing the /redfish/v1/Managers/Bluefield_BMC endpoint may time out due to repeated IPMB communication failures, causing the endpoint to become unresponsive. Resetting EFI variables and power cycling the system resolves the issue.

Reported in version: 25.10-LTSU2

5071627

Description: On some Aztlan BlueField-2 cards, the BMC may report false VCORE and 3P3V_AUX voltage warnings during normal operation. These false-positive warnings could appear in the system event log (SEL) and as warning sensor health in Redfish and IPMItool, triggering unnecessary alerts without indicating an actual hardware fault. This issue was fixed by adjusting the warning thresholds for the VCORE and 3P3V_AUX sensors to align with normal operating ranges.

Reported in version: 25.10-LTSU2

5071048

Description: Enhanced the Arm-side port link status reporting to ensure accurate link status evaluations and prevent false configuration error alarms in the BMC SEL log during DPU OS reboots.

Reported in version: 25.10-LTSU2

5071040

Description: Following a DPU reset, the BaseMAC, BaseGUID, and Description fields may incorrectly return as empty within the redfish/v1/Systems/Bluefield/Oem/Nvidia schema response.

Reported in version: 25.10-LTSU2

5030888

Description: When running OobUpdate --show_all_version repeatedly, the BOARD and/or BSP fields may intermittently return empty values due to a parsing issue in the OobUpdate script. The underlying firmware data remains available on the device.

Reported in version: 25.10-LTSU2

5030870

Description: Following a firmware update via the OobUpdate script, the version reported for components such as UEFI may occasionally fail to reflect the newly installed version. This is a transient reading error, not an update failure.

Reported in version: 25.10-LTSU2

5030860

Description: In rare cases, after a firmware upgrade, doca-installer --show-running-fw may report N/A for the CEC version on a device. The actual firmware is correctly installed; this is a transient read failure.

Reported in version: 25.10-LTSU2

5017581

Description: When performing a BMC firmware update via update-multipart with an incorrect target, the TaskStatus field may intermittently return OK instead of Critical, even when the update fails due to no matching devices.

Reported in version: 25.10-LTSU2

4970600

Description: Accumulated CPER (Common Platform Error Records) logs in /var/cper can fill the BMC root partition, causing critical BMC services such as entity-manager to fail. This can result in SDR read errors and ipmitool sdr elist failures.

Reported in version: 25.10-LTSU2

4966467

Description: Event log warnings for PLDM_Sensor_1_100 may appear in the BMC event log when the NIC temperature exceeds 91°C. These warnings are benign and will be removed in a future release.

Reported in version: 25.10-LTSU2

4935831

Description: After deleting an entry in the BMC WebUI Event Log, the table may not immediately reflect the deletion. Reloading the page displays the correct updated log.

Reported in version: 25.10-LTSU2

4885311

Description: In rare cases (approximately 1%), BMC sensors and SDR data may not be available within 2 minutes of a BMC reboot, due to additional initialization tasks introduced in recent versions.

Reported in version: 25.10-LTSU2

5096997 / 5079086

Description: Resolved an issue where plug and unplug operations failed to complete when the host was powered off.

Keywords: Plug and unplug operations

Detected in version:

32.43.2402

Fixed in Release:

32.47.3576

4895260 / 4683339

Description: Fixed an issue where QPs established before loading DOCA PCC could exhibit inconsistent algorithm-selected behavior between ports in LAG mode after DOCA PCC is loaded.

Keywords: Congestion Control, DOCA PCC

Detected in version:

32.43.2402

Fixed in Release:

32.47.3576

4874765 / 4657767

Description: Fixed an issue where repeatedly writing NVCONFIG TLVs could cause excessive NV_DATA partition swaps during garbage collection. This rapid cycling could accelerate flash wear (end-of-life at 100,000 erases) and potentially render the device inoperable.

Keywords: NVCONFIG TLVs

Detected in version:

32.46.1006

Fixed in Release:

32.47.3576

4971728 / 4959302

Description: Fixed an issue where loading a hostless DPA application could cause an assertion failure (syndrome 0x8b76) during a Function Level Reset (FLR), resulting in a fatal firmware internal error.

Keywords: Hostless DPA, Assert 0x8b76

Detected in version:

32.47.2682

Fixed in Release:

32.47.3576

4977614 / 4976266

Description: Fixed an issue in legacy queue affinity where traffic always used only the bandwidth of a single physical port.

Keywords: Legacy queue affinity

Detected in version:

32.47.2682

Fixed in Release:

32.47.3576

4719936 / 4731957 / 4575884

Description: Fixed an issue where NV Config operations could hang or produce untracked register errors, and where flash operations longer than 2 seconds could trigger an unnecessary assert. Debug information has been added to help detect stuck NV Config operations and register access errors.

Keywords: Stability and diagnostic

Detected in version:

32.43.3608

Fixed in Release:

32.47.3576

4982511 / 4947006

Description: Fixed two rare race conditions in the PCC steering extension when the feature is enabled via path migration, empty RTT, or a custom header.

Keywords: PCC steering extension

Detected in version:

32.47.1088

Fixed in Release:

32.47.3576

4850603 / 4907678

Description: Fixed a rare race condition in the SUSPEND_VHCA command that could cause the command to hang.

Keywords: Live Migration

Detected in version:

32.47.1088

Fixed in Release:

32.47.3576