DOCA3.3 run DOCA-RDMA Sample :doca_rdma_write_requester/responser on the same BF3,but UAR alloc failed

I run doca-rdma sample on BF3 ARM, DOCA 3.3

I fount that the first start progress was started well,but the second progress started fail for UAR alloc failed .

The logs here :

./doca_rdma_write_responder -d mlx5_4 -cm
[2025-12-13 16:06:37:000552][4275341696][DOCA][INF][CORE][doca_log.cpp:900] DOCA version 3.3.0109
[2025-12-13 16:06:37:000636][4275341696][DOCA][INF][RDMA_WRITE_RESPONDER][rdma_write_responder_main.c:70][main] Starting the sample
[2025-12-13 16:06:37:007435][4275341696][DOCA][WRN][CORE][doca_mmap.cpp:2100] Mmap 0xb23222b79900: Memory range isn’t aligned to 64B - addr=0xb23222b788a0. For best performance using CPU memory, align address to 64B (cache-line size). For best performance using GPU memory, align address to 64KB (page size)
[2025-12-13 16:06:37:015699][4275341696][DOCA][INF][RDMA_WRITE_RESPONDER][rdma_write_responder_sample.c:213][rdma_write_responder_state_change_callback] RDMA context is running
[2025-12-13 16:06:37:015723][4275341696][DOCA][INF][RDMA][rdma_common.c:1190][rdma_cm_connect] Using RDMA_CM to setup RDMA connection
[2025-12-13 16:06:37:015729][4275341696][DOCA][INF][RDMA][rdma_common.c:1205][rdma_cm_connect] -----------------------------------------------
[2025-12-13 16:06:37:015731][4275341696][DOCA][INF][RDMA][rdma_common.c:1206][rdma_cm_connect] RDMA_CM connection params:
[2025-12-13 16:06:37:015740][4275341696][DOCA][INF][RDMA][rdma_common.c:1207][rdma_cm_connect] – Connection Role: Server
[2025-12-13 16:06:37:015745][4275341696][DOCA][INF][RDMA][rdma_common.c:1208][rdma_cm_connect] – Addr_type : 0
[2025-12-13 16:06:37:015753][4275341696][DOCA][INF][RDMA][rdma_common.c:1209][rdma_cm_connect] – Addr: NULL
[2025-12-13 16:06:37:015757][4275341696][DOCA][INF][RDMA][rdma_common.c:1210][rdma_cm_connect] – Port: 13579
[2025-12-13 16:06:37:015761][4275341696][DOCA][INF][RDMA][rdma_common.c:1211][rdma_cm_connect] – Num_connections: 1
[2025-12-13 16:06:37:015765][4275341696][DOCA][INF][RDMA][rdma_common.c:1212][rdma_cm_connect] -----------------------------------------------
[2025-12-13 16:06:37:015770][4275341696][DOCA][INF][RDMA][rdma_common.c:1218][rdma_cm_connect] Server calling doca_rdma_start_listen_to_port
[2025-12-13 16:06:37:015850][4275341696][DOCA][INF][RDMA][rdma_common.c:1248][rdma_cm_connect] [Server] started connection successfully
[2025-12-13 16:06:37:015859][4275341696][DOCA][INF][RDMA_WRITE_RESPONDER][rdma_write_responder_sample.c:221][rdma_write_responder_state_change_callback] RDMA context finished initialization

but:

./doca_rdma_write_requester -d mlx5_4 -cm -sa 2.2.2.2
[2025-12-13 16:07:43:594809][276531584][DOCA][INF][CORE][doca_log.cpp:900] DOCA version 3.3.0109
[2025-12-13 16:07:43:594896][276531584][DOCA][INF][RDMA_WRITE_REQUESTER][rdma_write_requester_main.c:70][main] Starting the sample
[2025-12-13 16:07:43:612150][276531584][DOCA][ERR][RDMA][rdma_ctx_cpu.cpp:181] RDMA 0xf7fb0fd50010: Failed to create UAR. err=DOCA_ERROR_DRIVER
[2025-12-13 16:07:43:612176][276531584][DOCA][ERR][RDMA][rdma_ctx_cpu.cpp:705] RDMA 0xf7fb0fd50010: Failed to start ctx: Failed to create connection objects. err=DOCA_ERROR_DRIVER
[2025-12-13 16:07:43:612247][276531584][DOCA][ERR][CORE][doca_pe.cpp:1119] Progress engine 0xb1da743fcec0: Failed to start context=0xf7fb0fd50010. err=DOCA_ERROR_DRIVER
[2025-12-13 16:07:43:612255][276531584][DOCA][ERR][RDMA_WRITE_REQUESTER][rdma_write_requester_sample.c:475][rdma_write_requester] Failed to start RDMA context: DOCA Driver call failure
[2025-12-13 16:07:43:616605][276531584][DOCA][ERR][RDMA_WRITE_REQUESTER][rdma_write_requester_main.c:103][main] rdma_write_requester() failed: DOCA Driver call failure
[2025-12-13 16:07:43:616630][276531584][DOCA][INF][RDMA_WRITE_REQUESTER][rdma_write_requester_main.c:115][main] Sample finished with errors

Here mlx5_4 is a SF, and the parent PF’s config is :

ubuntu@localhost:/home/build/req$ sudo mlxconfig -d /dev/mst/mt41692_pciconf0 q|grep BAR
PF_BAR2_SIZE 8
NON_PREFETCHABLE_PF_BAR False(0)
PF_BAR2_ENABLE True(1)
PF_LOG_BAR_SIZE 8
VF_LOG_BAR_SIZE 1
MEMIC_BAR_SIZE 0
BAR_PAGE_ALIGNMENT DEVICE_DEFAULT(0)
PF_SF_BAR_SIZE 10

ubuntu@localhost:/home/build/req$ sudo mlxconfig -d /dev/mst/mt41692_pciconf0.1 q |grep BAR
PF_BAR2_SIZE 8
NON_PREFETCHABLE_PF_BAR False(0)
PF_BAR2_ENABLE True(1)
PF_LOG_BAR_SIZE 8
VF_LOG_BAR_SIZE 1
MEMIC_BAR_SIZE 0
BAR_PAGE_ALIGNMENT DEVICE_DEFAULT(0)
PF_SF_BAR_SIZE 0
ubuntu@localhost:/home/build/req$ lspci -vvvs 03:00.0
03:00.0 Ethernet controller: Mellanox Technologies MT43244 BlueField-3 integrated ConnectX-7 network controller (rev 01)
Subsystem: Mellanox Technologies MT43244 BlueField-3 integrated ConnectX-7 network controller
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- SERR- <PERR- INTx-
Latency: 0
Interrupt: pin A routed to IRQ 35
IOMMU group: 7
Region 0: Memory at 800000000000 (64-bit, prefetchable) [size=32M]
Region 2: Memory at 800004000000 (64-bit, prefetchable) [size=8M]
Expansion ROM at 7fff00000000 [disabled] [size=1M]
Capabilities:
Kernel driver in use: mlx5_core
Kernel modules: mlx5_core

Q1: Why are there two dev under /dev/mst?

ll /dev/mst/mt41692_pciconf0
mt41692_pciconf0 mt41692_pciconf0.1

Q2: How to fix the secend progress start problem ?

Hello @shan.lijie ,

Thank you for posting your query on our community.

Q1: The two entries under /dev/mst are expected and correspond to the dual-port / multi-function device, where each function is exposed separately.

Q2: Please ensure that the SF is configured as required for DOCA RDMA samples.

Refer to the documentation here:
https://docs.nvidia.com/doca/sdk/doca-rdma/index.html#src-3984217536_id-.DOCARDMAv3.1.0-ConfiguringScalableFunctionsforDOCARDMASamples

If the issue persists, submit a support ticket for further troubleshooting. It can be opened by emailing " Networking-support@nvidia.com ". Please note that an active support contract would be required for the same. For contracts information, please feel free to reach out to our contracts team at " Networking-Contracts@nvidia.com "

Thanks,

Bhargavi