Mlx5: missing registers BAR error code -19

Infrastructure & Networking Software And Drivers Network Management Products
1

After endlessly troubleshooting I am resorting to the manufacturer in order to resolve an issue with Mellanox MT27800 Family [ConnectX-5] Drivers not performing properly.

I have updated all firmware to the latest available especially BIOS:
A47 v2.72 (04/20/2023)

BIOS is NOT in safe mode.

I installed a fresh ISO from Ubuntu for LTE version 22.04.2 running kernel version 5.15.0-72-generic
5.15.0-72-generic #79-Ubuntu SMP Wed Apr 19 08:22:18 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

I checked that the network adaptors are being seen:

lshw -C network

  *-network:0 UNCLAIMED
       description: Ethernet controller
       product: MT27800 Family [ConnectX-5]
       vendor: Mellanox Technologies
       physical id: 0
       bus info: pci@0000:a6:00.0
       version: 00
       width: 64 bits
       clock: 33MHz
       capabilities: pciexpress vpd msix pm cap_list
       configuration: latency=0
  *-network:1 UNCLAIMED
       description: Ethernet controller
       product: MT27800 Family [ConnectX-5]
       vendor: Mellanox Technologies
       physical id: 0.1
       bus info: pci@0000:a6:00.1
       version: 00
       width: 64 bits
       clock: 33MHz
       capabilities: pciexpress vpd msix pm cap_list
       configuration: latency=0

lspci -v | grep Mellanox

a6:00.0 Ethernet controller: Mellanox Technologies MT27800 Family [ConnectX-5]
        Subsystem: Mellanox Technologies MT27800 Family [ConnectX-5]
a6:00.1 Ethernet controller: Mellanox Technologies MT27800 Family [ConnectX-5]
        Subsystem: Mellanox Technologies MT27800 Family [ConnectX-5]

Installation succeeds for the driver software located here:
Linux Ethernet Drivers - MLNX_EN (nvidia.com)

Yet I get this error when restarting the NIC driver…

image
image984×109 22.7 KB

mlx5_pci_init:1117: error requesting BARs, aborting
mlx5_pci_init failed with error code -19

The NICs do not show up when using ‘ip a’

I’m not sure what else could possibly be wrong. Please assist.

2

Please set the pci=realloc=off kernel parameter in Linux, and let us know the status.

1 Like
3

This worked, in order to build intuition, can you explain how you knew to try this?

4

The issue happens when installing 2 ConnectX adapters on 2 slots on the same server with Ubuntu OS.

Ubuntu OS cannot allocate enough memory UAR BAR to the second interface (adapter). Therfore, The OS boots with only one adapter (from the first slot).

5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.