Issues creating basic rootfs for Jetson Linux 36.2

Autonomous Machines Jetson & Embedded Systems Jetson Orin NX
4

The flashing seems to work with the patch mentioned above, but now I have problems with setting the power model and overclocking. I had a similar issue with the power model in version 35.4.1, though I found the missing package (bc) as explained here NVPM WARN: power mode is not set! And cannot set it using config file - #5 by hvbotten

Now it seems like there are different errors:

root@motus-tegra:~# sudo jetson_clocks
Error: nvidia-smi not found.

root@motus-tegra:~# nvpmodel -m 3 --verbose
NVPM VERB: Config file: /etc/nvpmodel.conf
NVPM VERB: parsing done for /etc/nvpmodel.conf
NVPM VERB: Using default settings.
NVPM VERB: set power mode as 25W(3) in /etc/nvpmodel.conf.
NVPM ERROR: Error opening /sys/devices/platform/gpu.0/tpc_pg_mask: 2
NVPM ERROR: optMask is 1, no request for power mode
NVPM VERB: Using default settings.

I thought nvidia-smi wasn’t supported for Tegra-based platforms, so why is jetson_clocks complaining about it missing? And what is happening with nvpmodel? The /etc/nvpmodel.conf is similar to the one presented here.

I’ll try to flash the samplefs (or desktop) and see if the problem is present there as well, but I would like some information about the errors above if possible.

For information this is the command I used to flash the device (after creating the default user), in case I’m doing something wrong

Linux_for_Tegra/tools/kernel_flash/l4t_initrd_flash.sh \
    --external-device nvme0n1p1 \
    -c Linux_for_Tegra/tools/kernel_flash/flash_l4t_external.xml \
    -p "-c Linux_for_Tegra/bootloader/generic/cfg/flash_t234_qspi.xml" \
    --showlogs \
    --network usb0 \
    jetson-orin-nano-devkit \
    internal

EDIT: Found out that i needed the nvidia-l4t-nvml package for nvidia-smi, though it’s is still not working:

root@motus-tegra:~# sudo jetson_clocks
Error: fail to do nvidia-smi operation. The exit code of nvidia-smi is 255
root@motus-tegra:~# nvidia-smi
Unable to determine the device handle for GPU00002:00:00.0: Unkown Error

Which makes sense as we have an internal GPU…

With the minimal rootfs these commands work, though I had to add some packages for it to flash properly as well (like matplotlib and python-tk, which are packages I really wish i didn’t have to add to a minimal rootfs :/ ).