I have no problems in flashing a compiled kernel jetpack version 34.1 over to a TX2 NX on a custom carrier. My issue though is that once the flashing is done and I boot the system, it doesn’t look like the jetpack drivers and cuda are installed. When I try to install them with a terminal command eventually I run out of disk space. The only way I figured out how to do it is by moving the root folder over to an SD card.
I would much prefer if I could flash the TX2 NX with jetpack drivers and CUDA already installed. Is there a way to do this so I don’t have to install them after the SoM is flashed?
Normally if you use the SDK manager you have the option to do this, but I’m trying to just use the flash.sh script.
Instead of this you can do flashing directly on ton SD card itself where space is more.
After that when you install CUDa and related package using sudo apt-install nvidia-jetpack
Then it will get installed automatically under rootfs on the sd card itself.
I would prefer to have the OS run on the on-board emmc.
Looking to know how to have jetpack and cuda compiled with the OS, or if there’s a way to have the drivers on a host computer and install onto the SoM so less memory is used, or… if there’s a way to point the installation of the drivers to the emmc instead of current SD card?
I have the tools folder but there’s no backp-restore folder in there. Checking you link you gave it seems like I can flash directly to the SD card in our system which is good, but I feel like in the end it would end up to a similar thing as transferring and changing the root folder after the OS is installed into the SoM.
Yes, you can just use flash script instead of backup/restore script.
You can also use flash script to clone/flash the APP partition only to prepare the rootfs with all required packages installed.
Please refer to To clone a Jetson device and flash part in the link I shared.