nvpmodel systemd service crashes

[jjannik]

Im running kirkstone with L4T35.3.1. Since I changed to a read-only filesystem the nvpmodel.service will fail (around 50%fail). Im wondering if this is critical or not. Thanks for any help. The gpu seeems to work in ether way (cuda-sample/devicequery)

Here the output after a failure

root@agx-orin:~# /usr/sbin/nvpmodel -f /etc/nvpmodel.conf --verbose
NVPM VERB: Config file: /etc/nvpmodel.conf
NVPM VERB: parsing done for /etc/nvpmodel.conf
NVPM VERB: set power mode as MODE_30W(2) in /etc/nvpmodel.conf.
NVPM ERROR: Error writing 240 to /sys/devices/gpu.0/tpc_pg_mask: 16
NVPM WARN: Reboot required for changing to this power mode: 2
NVPM WARN: DO YOU WANT TO REBOOT NOW? enter YES/yes to confirm:

root@agx-orin:~# cat /var/lib/nvpmodel/status
pmode:0002

Im aware of this Nvidia link and the infos about tpc_pg_mask however Im not sure it this also true for meta-tegra.

[madisongh]

Yes, the same applies to Yocto builds - if you want to change the nvpmodel setting you do have to reboot, and you'll get that same message.

If nvpmodel.service is getting that error at boot time, then something else might be wrong, too. There were some commits that went in back in July to fix some startup ordering issues that could cause spurious failures, so make sure you're up-to-date with at least commits 7479f93 and 5b426d3.

[jjannik]

Thank you so much! That seems to solve the issue. Sorry for the overhead.