I only exchanged the SD card while keeping the QSPI. I am not aware that I changed the QSPI partition table or any files it installs by updating my SD card previously with a custom OS.
Boot the system and see the status of the nv-l4t-bootloader-config service
Is there anything I could have modified inside the L4T directory that could affect the QSPI flash?
I think I did not modify anything, the jetson-xavier-nx-devkit-qspi.conf looks like this:
Okay, so from my side it looks like I have a 100% unmodified devkit OS and QSPI with the issue.
Unless you think anything else could have an effect on the QSPI flashing.
One more thing to ask here. If you prepare another empty sdcard, plugged to the board and flash with jetson-xavier-nx-devkit.conf. Remove the card after flashed and replace with the current “custom rootfs” sdcard, will the issue still there?
Both show this behavior.
You’d like me to flash an unmodified OS with the jetson-xavier-nx-devkit.conf? I’d have to do some work to flash the normal sample rootFS with the flash.sh as I never added that workflow to my scripts.
Would it alternatively be possible to flash with the SDKManager to produce the same result?
Otherwise if you wanted me to flash my custom rootFS, I already did that because I started from the “Custom” QSPI + Custom RootFS and switched to the “Custom” QSPI + RootFS from the 5.0.2 Image. After that I flashed only the QSPI on the board to see if it makes any difference. I say “Custom” because I am pretty sure that the QSPI that got flashed with my Custom RootFS was not modified from the one which would normally be flashed.
I am sorry for the late response. I’ve set up a PC with Ubuntu 18.04, installed the SDKManager and flashed the 5.0.2 Image. Unfortunately the result is completely different from the SD Card Image.
If I plug no display the system hangs at “please wait for the auto setup to complete”. If I plug a display it hangs at “please complete system configuration setup on desktop to proceed”…but the display remains black. Actually during the flashing in SDKmanager I selected to use a preselected username and password and skip the configuration process.
Anyway, I removed the not working SD Card from SDKManager and added my SD card with the custom OS:
Apr 21 12:56:10 localhost systemd[1]: Started Configure QSPI bootloader service.
Apr 21 12:56:10 localhost nv-l4t-bootloader-config.sh[2246]: Restore link: APP to /dev/mmcblk0p1
Apr 21 12:56:10 localhost nv-l4t-bootloader-config.sh[2246]: Restore link: kernel-dtb_b to /dev/mmcblk0p14
Apr 21 12:56:10 localhost nv-l4t-bootloader-config.sh[2246]: Restore link: recovery-dtb to /dev/mmcblk0p17
Apr 21 12:56:11 localhost nv-l4t-bootloader-config.sh[2207]: 3668-100---1--jetson-xavier-nx-devkit-
Apr 21 12:56:11 localhost nv-l4t-bootloader-config.sh[2207]: Info: Spec variable TegraPlatformSpec is not found.
Apr 21 12:56:11 localhost nv-l4t-bootloader-config.sh[2207]: Info: Write TegraPlatformSpec with 3668-200-0000-H.0-1-2-jetson-xavier-nx-devkit-.
Apr 21 12:56:11 localhost nv-l4t-bootloader-config.sh[2207]: Info: Spec variable TegraPlatformCompatSpec is not found.
Apr 21 12:56:11 localhost nv-l4t-bootloader-config.sh[2207]: Info: Write TegraPlatformCompatSpec with 3668-100---1--jetson-xavier-nx-devkit-.
Apr 21 12:56:12 localhost nv-l4t-bootloader-config.sh[2207]: Error. Cannot get valid version information
Apr 21 12:56:12 localhost nv-l4t-bootloader-config.sh[2207]: Exiting ...
Apr 21 12:56:12 localhost systemd[1]: nv-l4t-bootloader-config.service: Main process exited, code=exited, status=1/FAILURE
Apr 21 12:56:12 localhost systemd[1]: nv-l4t-bootloader-config.service: Failed with result 'exit-code'.
I am quite sure that you’ll be able to reproduce the issue.
Well as I said, after flashing with SDK Manager the rootfs which was flashed by SDK Manager was not working as it got stuck during the oem-config step. So I can’t tell if the SDK Manager rootfs would have shown the issue.
What I can tell is that with the QSPI from SDK Manager and my custom rootfs the issue shows.
It also shows with the official SD Card Image. So I don’t see any chance that you will not observe it when you try it yourself.
Maybe someone else who uses 5.0.2 can check with “systemctl” if the nv-l4t-bootloader-config service fails?