I’m running Fedora 33 and just upgraded to Linux 5.9.16 and NVIDIA 460.27.04 drivers. I’m seeing occasional short hangs which are 100% reproducible when running cuda programs, including nvidia-smi.
If I time a run of nvidia-smi, it takes more than 8 seconds to run. A portion of the time while it is running, my mouse stops responding, but comes back intermittently, and then behaves normally after nvidia-smi completes.
If I run journalctl, I can see that I reliably get kernel warnings each time I run that look like
Dec 29 10:21:54 kernel: NVRM: GPU 0000:09:00.0: rm_init_adapter failed, device minor number 0
Dec 29 10:21:54 kernel: NVRM: GPU 0000:09:00.0: RmInitAdapter failed! (0x26:0x65:1290)
Dec 29 10:21:50 kernel: NVRM: GPU 0000:09:00.0: rm_init_adapter failed, device minor number 0
Dec 29 10:21:50 kernel: NVRM: GPU 0000:09:00.0: RmInitAdapter failed! (0x26:0x65:1290)
I’ve attached a bug report generated by the tool nvidia-bug-report.log.gz (673.0 KB)