Apply_binaries failing

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
1

Hello I have downloaded the JP 6.1 L4T 36.4.0 BSP: R36.4.0

I am trying to flash on the basic file system but I am running into errors with the apply_binaries.sh script.

My steps are:

  1. Untar the BSP and cd into the Linux_for_Tegra dir
  2. Generate the NVIDIA basic root file system of Ubuntu 22.04 LTS (Jammy Jellyfish):
    sudo $L4T/tools/samplefs/nv_build_samplefs.sh --abi aarch64 --distro ubuntu --flavor basic --version jammy

This completes with no errors:

  1. Assemble the ROOTFS that was just generated:
sudo rm -r $L4T/rootfs/*
sudo tar xvpf $L4T/tools/samplefs/sample_fs.tbz2 -C $L4T/rootfs/
sudo $L4T/tools/l4t_flash_prerequisites.sh
sudo $L4T/apply_binaries.sh

The apply binary script encounters a lot of dpkg errors while processing the packages:

Errors were encountered while processing:
nvidia-l4t-init
nvidia-l4t-jetsonpower-gui-tools
nvidia-l4t-nvfancontrol
nvidia-l4t-nvpmodel
nvidia-l4t-nvpmodel-gui-tools
nvidia-l4t-weston
nvidia-l4t-x11
nvidia-l4t-xusb-firmware
nvidia-l4t-kernel
nvidia-l4t-kernel-dtbs
nvidia-l4t-kernel-headers
nvidia-l4t-kernel-oot-headers
nvidia-l4t-kernel-oot-modules
nvidia-l4t-bootloader
nvidia-l4t-3d-core
nvidia-l4t-gbm
nvidia-l4t-graphics-demos
nvidia-l4t-initrd
nvidia-l4t-jetson-io
nvidia-l4t-display-kernel

My host machine is running Ubuntu 22.04.5 LTS. It starts to go wrong when it gets to nvidia-l4t-init with the error:
depends on efibootmgr; however: Package efibootmgr is not installed.
However my system does have it installed:

image
image941×118 3.63 KB

Any help would be much appreciated. I have attached the full command output of the apply_binaries.sh script to this post.
apply_binaries_errors.txt (21.4 KB)

2

efibootmgr is missing from nvubuntu-jammy-basic-aarch64-packages

but it is included in nvubuntu-jammy-minimal-aarch64-packages and nvubuntu-jammy-desktop-aarch64-packages. Is this a bug/typo?

3

nvme-cli and libseat1 are also missing

4

Hope I am not reading your post wrong, I am making assumptions about the endpoint.
So far, what I have discovered with their system is this.

You can burn the SD card image using pi-imager and it runs.
Just started to use sdkmanager, it works so much better, cuda is in the path, etc.
Also built a yocto core-image-weston, it has not been tested yet.

5

Hi @foxsquirrel1, thanks for your input, for us we have to use the Nvidia flashing tools that come with there JP6.1.0 BSP and we desire to build our rootfs manually as we want the basic file system. I think there nvubuntu-jammy-basic-aarch64-packages is missing packages for the apply_binaries.sh script to work.

6

The sync script is missing from the download for the bsp so you do not get everything. Used sdkmanager and copied that to a isolated directory to work out of. That is missing the the setup for the build env. Did build a kernel and dtb but it was a headache. The DTC is even missing from the tree so I just repointed to the local dtc for that. It was also pointing to my host headers and that had to be repointed to jetson.

Not sure what is going on with this, if 6.1 is not ready they should state that as fact.

All we are doing now is wasting our time trying to get the gpio control on the 40 pin headers.

I am downloading the jetpack 5.0 and will see if that will allow us gpio control. If not, our backup plan is build an I/O card and just use a big box board with a GPU card. That is how I did it 30 years ago (ISA bus days) and it worked just fine. These boards are very nice but we are getting tired of the BS.

7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.