rocm-smi 4.1 returns error

[perestoronin]

python3 /opt/rocm/bin/rocm_smi.py
Failed to get "domain" properity from properties files for kfd node 1.
rsmi_init() failed
Exception caught: rsmi_init.
ERROR:root:ROCm SMI returned 8 (the expected value is 0)

but rocminfo work perfect

[ROCmSupport]

Hi @perestoronin
Thanks for reaching out.
In my case, rocm-smi shows information properly.
Can you please share us with below information.
Asic, OS, Kernel, ROCm# version and dmesg output.

[perestoronin]

Linux 5.4.110-gentoo-rt54 #1 SMP PREEMPT_RT Thu x86_64 AMD Phenom(tm) II X6 1100T

rocminfo https://gist.github.com/raw/37969bd15b49c281c4d3535e9791876a

python /opt/rocm-4.1.0/bin/rocm_smi.py
Failed to get "domain" properity from properties files for kfd node 1.
rsmi_init() failed
Exception caught: rsmi_init.
ERROR:root:ROCm SMI returned 8 (the expected value is 0)

dmesg https://gist.github.com/raw/800e51601e1aa6ccc0326abb44736890

cat /sys/class/kfd/kfd/topology/nodes/1/properties https://gist.github.com/raw/eaa36a7bcac19abc5711ecf8cb692e0a

[ROCmSupport]

Thanks for more information @perestoronin
I looked at the device properties and found that "domain" property is missed and so its throwing error.

But in my case, "domain 0" and so no issue observed.
Working more to gather more information.

[perestoronin]

I investigated, that domain introduced since kernel version 5.8.

@ROCmSupport what versions of kernel compiance with rocm 4.1.1 ? In docs for rocm obsolete information for requerement kernel version driver amdgpu ?

[perestoronin]

/opt/rocm/bin # python rocm_smi.py
GPU Temp AvgPwr SCLK MCLK Fan Perf PwrCap VRAM% GPU%
0 39.0c 7.0W 852Mhz 167Mhz 14.9% auto 220.0W 1% 6%

/opt/rocm/bin # uname -a
Linux 5.10.28-gentoo-rt36 #1 SMP PREEMPT_RT x86_64 AMD Phenom(tm) II X6 1100T

Аfter update kernel from deprecated 5.4 to actual for rocm-4.1.1 kernel-5.10 issue is resolved.

[ROCmSupport]

Thanks for upgrading kernel and closing this issue.
Thank you.