Orin AGX external USB drive boot flashing issuue

[sjelavic123]

Bought Orin AGX at Amazon but having issues with the small 64GB drive size and speed. Tried to use a 1TB external USB drive and boot from it. Having persistent issue flashing it.
Used the following procedure from here:
https://docs.nvidia.com/sdk-manager/install-with-sdkm-jetson/index.html
The host is a Linux machine running Ubuntu 2004. The SK manager recognizes Orin and reports that it is in recovery mode.

Step1

Looks OK
Step 2

Step 3
Problem:

Did the Skip
Step 4 Installation Failed

The log is available here
SDKM_logs_JetPack_6.1_Linux_for_Jetson_AGX_Orin_64GB_2024-12-13_07-24-01.zip (794.7 KB)
I also tried the command line flash but ran into a number of issues. Will provide the details later. Is there a well documented procedure for flashing the supported external drive THAT ACTUALLY WORKS.

[carolyuu]

Hi,
Here are some suggestions for the common issues:

1. Performance

Please run the below command before benchmarking deep learning use case:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

2. Installation

Installation guide of deep learning frameworks on Jetson:

• TensorFlow: Installing TensorFlow for Jetson Platform - NVIDIA Docs
• PyTorch: Installing PyTorch for Jetson Platform - NVIDIA Docs
  We also have containers that have frameworks preinstalled:
  Data Science, Machine Learning, AI, HPC Containers | NVIDIA NGC

3. Tutorial

Startup deep learning tutorial:

• Jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson
• TensorRT sample: Jetson/L4T/TRT Customized Example - eLinux.org

4. Report issue

If these suggestions don’t help and you want to report an issue to us, please attach the model, command/step, and the customized app (if any) with us to reproduce locally.

Thanks!

[sjelavic123]

Thank you.
My issue however is that the eMMC SW is corrupted and will not boot in normal mode. It does boot into the recovery mode. I know that because the SDK manager connects to it and reports that it is in recovery mode.
So what is need is the following:

1. Flash internal memory (eMMC) with the default software that will get me into the normal mode.
2. After that works, I would like to be able to boot from larger and faster external USB drive.

[jose.morera]

Hello @sjelavic123,

Were you able to successfully flash the device, and now you’re experiencing issues with the installation?

You can try accessing your Orin AGX Developer Kit and running the following commands to check if you can install the SDK components:

sudo apt update
sudo apt install nvidia-jetpack

I’m looking forward to your response.

Jose Morera
Embedded Software Engineer at RidgeRun
Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com
Website: www.ridgerun.com

[sjelavic123]

Hi Jose,
The original product came from Amazon and worked ok with the software running on the internal 64GB memory.
I am running some robotics apps that use Yolo11n and Image recognition with ONNX runtime. Was not happy with the performance
and started running out of the internal eMMC.
So tried to install the linux boot on a USB SSD 1TB using the linux ubuntu 2004. That linux machine is connected to Orin via a USB.
Used Nvidia SDK manager to flash the external drive following the Nvidia procedure. The flash failed.
After that my Orin would not boot in normal mode and I cannot log into it.
If I put it into recovery mode though and connect the linux host via a usb the host SK manager still sees.
So I thought I would reflash the eMMC so that I can log into the Orin machine but the flash fails.
The flash to the internal eMMC log was attached to my earlier post if you could log into it.
But for now I cannot use any of the commands that you suggested on the Orin because I cannot log into it.

[jose.morera]

Thank you for the explanation. Initially, before noticing the poor performance, were you able to successfully flash the eMMC using the same host machine?

[sjelavic123]

No, never successfully flashed it before.

[whitesscott]

Here’s the documents that tell how to flash a usb drive

[sjelavic123]

Thank you Scott,
According to the document flashing can be done with the drive connected to Orin. For that eMMC boot has to be working. Mine does not. It can only go into recovery mode.
The document also says that alternatively, the dive can be connected to the host. But for that setup proper flashed eMMC is also required.
So, I need to flash the eMMC first and that is what I tried with the SDK manager running on the host and with the target in recovery mode. The host recognizes the target and reports that that the target is in recovery node. But flashing fails. I included the log in the earlier message.

[sjelavic123]

OK managed to get the eMMC boot working. Updated the Jetpack on the target as per Jose’s instructions and that all works ok now.
But now I would like to flash the SSD drive and the instructions in the document are confusing.

https://docs.nvidia.com/jetson/archives/r36.4/DeveloperGuide/SD/FlashingSupport.html?highlight=flashing%20emmc#flashing-by-using-a-convenient-script

I assume the ./nvsdkmanager_flash.sh script is run on the host. The documentation should bee very specific where the command is run. I also assume that the target is in recovery mode as stated earlier in the document.

Do I have to specify the --storage option. The target is the standard AGX with eMMC. The doc says that the eMMC will be also modified. I think that is how I messed up my eMMC the first time I tried this procedure.
So in summary:
Target in recovery
Run sudo ./nvsdkmanager_flash.sh --storage nvme1n1p1
Reboot target when done
Is this correct?

[sjelavic123]

The procedure in my previous message did not work.
Here is the log.
flash_ext_drive_failure_log.txt (3.3 KB)

Should I follow the procedure suggested by whitesscott instead i.e. connect the drive to the host and run the l4t_initrd_flash.sh script? Should the target be in normal mode for that or recovery mode?

[DavidDDD]

Hi,

You could refer to this doc to manually flash

• Navigate to the directory ~/nvidia/nvidia_sdk/JetPack_6.1_Linux_JETSON_AGX_ORIN_TARGETS/Linux_for_Tegra/ and run these commands:

  sudo ./apply_binaries.sh
  sudo ./tools/l4t_flash_prerequisites.sh
  

• Verify that your device is in recovery mode by this section.
• And follow the section to do related flash command.

Also some questions to confirm:

• Is your host Ubuntu 20.04 running on a physical machine or a virtual machine(e.g. VMware, WSL, VirtualBox)

Thanks

[sjelavic123]

Thank you for the info.
The first command
sudo ./apply_binaries.sh
worked ok
The second command
sudo ./tools/l4t_flash_prerequisites.sh
Gave me the following errors:
singature_error_log.txt (2.4 KB)

[DavidDDD]

Hi,

It might be the similar issue discussed in this

Invalid public key for CUDA apt repository

@bassemfg Those are all local repos, installed in /var/, the ephemeral GPG pubkey for each is located in its directory, for example ls /var/cudnn-local-repo-ubuntu2204-8.9.3.28/cudnn-local-7F7A158C-keyring.gpg As prompted during installation of the local repo, copy the keyring to /usr/share/keyrings/ (see: NVIDIA CUDA Installation Guide for Linux)

Thanks

[sjelavic123]

OK that worked. Thank you DavidDDD,
Do you plan to fix the SDK Manager so that this backdoor process does not have to be used?
In summary this is the procedure that I used.
flashing_procedure_updated.txt (2.0 KB)
This issue can be closed