Mlnx_qos set dcbx to firmware gives error, mlnx_qos PFC setting not persistent

Omniverse General Discussion
1

Hi,
I am using mlnx_qos 1.2 with connect x6.
My OS is ubuntu 20.04.1
NVIDIA-SMI 555.42.02, Driver Version: 555.42.02, CUDA Version: 12.5
mft 4.25.0-62

When I try to run sudo mlnx_qos -i cx6_0 --dcbx=fw, I get the following error.

Traceback (most recent call last):
File “/usr/bin/mlnx_qos”, line 396, in
ctrl.set_dcbx(0);
File “/usr/share/mlnx-tools/python/dcbnetlink.py”, line 204, in set_dcbx
self.check_err(m, DCB_ATTR_DCBX)
File “/usr/share/mlnx-tools/python/dcbnetlink.py”, line 164, in check_err
raise err
OSError: Netlink error: Bad value. see dmesg.

I am trying to do this because for some reason, after setting sudo mlnx_qos -i cx6_0 --pfc 1,1,1,1,1,1,1,1 and seeing that the PFC is enabled, after about 20seconds or so, running sudo mlnx_qos -i cx6_0 and I see PFC is disabled for all priorities. I was thinking maybe dcbx is messing with the setting so I tried making dcbx managed by firmware.
Any help with solving these problems would be appreciated.
Thanks
Ricky

2

Ricky,

Not sure if you got your issue resolved but I have been seeing this same problem with my CX-6 card. I seemed to have found a solution today and wanted to pass it along. I am not sure if the defaults have changed for CX-6 cards vs. CX-5 cards but I had to set the following card configuration parameters to allow the --dcbx=fw option to be accepted. Our CX-5 cards seemed to already have these parameters set but on our CX-6 cards they were not. After setting these parameters and rebooting, the “mlnx_qos -i --dcbx=fw” succeeded and reported the correct data.

mstconfig -d c5:00.0 set NUM_OF_VFS=8

mstconfig -d c5:00.0 set LLDP_NB_DCBX_P1=1

mstconfig -d c5:00.0 set LLDP_NB_DCBX_P2=1

  • substitute your device address for “c5:00.0” above.

Brian

1 Like
3

Thanks for the help Brian

4

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