Bug Fixes in this Firmware Version | ConnectX-7 Firmware RN

Internal Ref.	Issue
5177191	Description: Resolved an issue with PLDM T5 component reporting. ConnectX devices now expose NIC firmware, NVIDIA NVConfig, and OEM NVConfig as independently updatable components. Optional NVConfig components may report an empty or 0 version and are shown only with PLDM T5 versions above 1.0.0.
	Keywords: PLDM
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4851684	Description: Resolved an issue where, under high-stress LRO traffic, an internal FIFO could enter a hardware deadlock state, preventing the NIC from receiving additional traffic. The fix detects and releases the deadlock if it occurs, restoring the NIC to a functional state.
	Keywords: LRO traffic
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
5174504 / 5140787	Description: Resolved an issue where intensive parallel QP INIT2RTR / DESTROY operations could incur high operation latency and reduce connection establishment rates when PCC was enabled.
	Keywords: PCC, INIT2RTR, DOCA PCC
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4947006 / 4980956 / 4982511	Description: Resolved two rare race conditions in the PCC steering extension when the feature is enabled through path migration, empty RTT, or a custom header. Under rare timing conditions during QP creation or destruction, these issues could cause traffic disruption or resource leaks.
	Keywords: PCC steering extension
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4948901	Description: Resolved an issue where the PLDM firmware-update flow used incorrect instance_id tracking for retries, which could allow repeated old chunk responses to be accepted as new data. The fix stores the actual transmitted request instance_id and uses it consistently during retries, preventing unexpected stale responses from being accepted.
	Keywords: PLDM firmware-update flow
	Detected in version: 28.46.3048
	Fixed in Release: 28.50.1002
5166648	Description: Resolved an issue involving an uninitialized stack read and an inverted EtherType check in the RX parser.
	Keywords: RX parser
	Detected in version: 28.47.1026
	Fixed in Release: 28.50.1002
4937640	Description: Resolved an issue where a PPCC mlxreg read operation could succeed when executed on a PF that is not the port owner, which could have security implications.
	Keywords: PCC, PPCC
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4517149	Description: Improved FLR and suspend/resume handling when VQoS was never enabled. In this case, there is no need to wait for all three iterations, since the wait was originally added to allow the SQ to be flushed from the distribution list. If VQoS was never enabled, there is nothing to clean up.
	Keywords: FLR
	Detected in version: 28.46.1006
	Fixed in Release: 28.50.1002
5180465	Description: Resolved an issue where, after an unexpected test termination while the max-entry-count ICMD was active, basic_debug could incorrectly report PSA error -30 because the expected steering_entry_count_limit event left a non-zero diagnostic context.
	Keywords: max-entry-count ICMD
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
5154717 / 5152076 / 5154806 / NVbug 6422398	Description: Resolved an issue where, when a transceiver module held its I2C management bus line stuck (SDA or SCL), the firmware detected the stuck bus and reset the module. After the reset, the module remained in low-power mode and its datapath was not re-established, preventing the link from coming up. On a shared module, all associated ports could drop simultaneously.
	Keywords: Cables
	Detected in version: 28.46.3048
	Fixed in Release: 28.50.1002
4923806 / 4867111 / 4927875	Description: Resolved an issue that could cause RoCE communication failures between specific NIC pairs after repeated queue pair (QP) create/destroy cycles. In the affected cases, traffic could enter a bad state and fail with transport retry counter exceeded errors. This fix prevents the NIC from entering that condition.
	Keywords: RDMA; connection failure
	Detected in version: 28.46.3048
	Fixed in Release: 28.50.1002
5093695	Description: Resolved an issue where using mlxconfig to configure all modules with the 0xff module split value could reset the modules to the default configuration.
	Keywords: mlxconfig
	Detected in version: 28.48.1000
	Fixed in Release: 28.50.1002
5090639	Description: Resolved an issue in single-domain mode where an FLR on a GVMI did not clear the per-domain usage bits in `SP_core_diag_multi_domain->pcie_link_latency_max_read` and `SP_core_diag_multi_domain->pcie_link_latency_min_read`. As a result, when the next GVMI took ownership and attempted to configure these counters, it could receive `DIAG_DATA_LIST_WRITE_STATUS_NO_RESOURCES`.
	Keywords: PCIe Link Latency Arch counters
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4704203 / 4704271 / 4986813 / NVbug 5621969	Description: Resulved a constant packet drop issue that caused DeepEP test case traffic to hang. The fix updates the STE search flow to run under a locking mechanism, preventing access to invalid STE entries that could result in packet drops.
	Keywords: DeepEP
	Detected in version: 28.46.1006
	Fixed in Release: 28.50.1002
5077326 / 4902171	Description: Resolved an issue where a slow module response to the NCSI GetModuleSerialData command could incorrectly trigger a dead IRISC watchdog assert. The firmware now handles this waiting state correctly, so slow module responses no longer cause a false watchdog hang indication.
	Keywords: NCSI
	Detected in version: 28.43.1014
	Fixed in Release: 28.50.1002
5018241	Description: Resolved an issue where firmware could raise FATAL_ERR_HW_INDICATED with ext_synd 0x0063 during boot due to a misaligned MACsec ASO ICM base address. The ASO core hardware alignment check was triggered during the FLR-cleanup ICM sweep that runs at boot under SR-IOV with DPU-style INIs. The assert occurred only under this specific configuration combination, where latent issues in the ASO ICM resource table and libfhi static configuration aligned.
	Keywords: FATAL_ERR_HW_INDICATED, FLR, SR-IOV
	Detected in version: 28.49.1014
	Fixed in Release: 28.50.1002
4890697	Description: Resolved an issue where setting the SRQ limit during traffic over multiple QPs on the same SRQ could lead to stuck slices and an RX hang. The SRQ limit update mechanism was changed to avoid canceling end-to-end credit requests from slices.
	Keywords: SRQ
	Detected in version: 28.48.1000
	Fixed in Release: 28.50.1002

Last updated: September 02, 2026