Installing and Flashing DriveOS 6.0.4 Linux fails

Autonomous Vehicles DRIVE AGX Orin DRIVE AGX Orin General
1

Please provide the following info (tick the boxes after creating this topic):
Software Version
DRIVE OS 6.0.4 SDK

Target Operating System
Linux

Hardware Platform
DRIVE AGX Orin Developer Kit (not sure its number)

SDK Manager Version
1.8.2.10409

Host Machine Version
native Ubuntu Linux 20.04 Host installed with SDK Manager

Hello, i have a whole zoo of problems when trying to install and flash with SDK Manger DriveOS 6.0.4 Linux. Precondition with /usr/bin/python is satisfied.

PROBLEM 1: Login to developer.nvidia.com does not work behind corporate proxy

The login via browser works flawlessly, in SDK manager it shows “Login credentials retrieved successfully”, but then it takes long time and finally give the error message “Unable to connect to proxy server” although the proxy configuration is fine.

Therefore i downloaded the relevant files on a different PC (not behind proxy) and copied the files to the target PC which is connected to the orin eval board to continue in offline mode. All following issues are when proceeding in offline mode.

PROBLEM 2: installation of cuda and vulkan packages fails

SDK Manager recognizes the files placed in the download folder (see above) but then installation of some of them fails.

CUDA fails:

14:17:52 ERROR: CUDA on Host - host: E: Unable to locate package cuda-toolkit-11-4
14:17:52 INFO: CUDA on Host - host: [ Package Install Finished with Error ]
14:17:52 INFO: CUDA on Host - host: [host] [ 692.00 KB released. Disk Avail: 58.42 GB ]
14:17:52 INFO: CUDA on Host - host: [ NV_CUDA_HOST_COMP Install took 3s ]
14:17:52 ERROR: CUDA on Host - host: command terminated with error
14:17:52 SUMMARY: CUDA on Host - host: Unable to locate debian package

TensorRT x86 fails:

14:02:43 INFO: TensorRT x86 - host: verifying checksum of /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/nv-tensorrt-repo-ubuntu2004-cuda11.4-trt8.4.11.6-x86-host-ga-20220702_1-1_amd64.deb
14:02:43 INFO: TensorRT x86 - host: Found file /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/nv-tensorrt-repo-ubuntu2004-cuda11.4-trt8.4.11.6-x86-host-ga-20220702_1-1_amd64.deb with correct checksum, skip downloading.
14:17:39 INFO: TensorRT x86 - host: verifying checksum of /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/nv-tensorrt-repo-ubuntu2004-cuda11.4-trt8.4.11.6-x86-host-ga-20220702_1-1_amd64.deb
14:17:39 INFO: TensorRT x86 - host: Found file /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/nv-tensorrt-repo-ubuntu2004-cuda11.4-trt8.4.11.6-x86-host-ga-20220702_1-1_amd64.deb with correct checksum, skip downloading.
14:17:52 SUMMARY: TensorRT x86 - host: Depends on failed component

PROBLEM 3: OS image ready but flashing fails with “file not found”

14:17:48 INFO: Flash DRIVE OS - flash: cd driveinstaller && ./driveinstaller --installtype bindflash --boardtype ES --pdkinstalledpath /home/tje2abt/nvidia/nvidia_sdk/DRIVE_OS_6.0.4_SDK_Linux_DRIVE_AGX_ORIN/DRIVEOS --configjson ./manifests/install_pdk_config/av_orin_p3710_driveos_linux_tegraA_install_config.json --targetuser nvidia --port ${AURIX_PORT:-/dev/ttyACM1} --eula-accepted 1 --consoleout enable --logpath ~/.nvsdkm/

14:17:48 INFO: Flash DRIVE OS - flash: **********************
14:17:48 INFO: Flash DRIVE OS - flash: exec_command: /tmp/tmp_NV_DRIVE_FLASH_DRIVE_COMP.sh
14:17:48 INFO: Flash DRIVE OS - flash: The user `tje2abt' is already a member of `dialout'.
14:17:48 ERROR: Flash DRIVE OS - flash: sed: can't read ./driveinstaller/manifests/install_pdk_config/av_orin_p3710_driveos_linux_tegraA_install_config.json: No such file or directory
14:17:48 INFO: Flash DRIVE OS - flash: [ Component Install Finished with Error ]
14:17:48 INFO: Flash DRIVE OS - flash: [host] [ 20.00 KB used. Disk Avail: 58.42 GB ]
14:17:48 INFO: Flash DRIVE OS - flash: [ NV_DRIVE_FLASH_DRIVE_COMP Install took 1s ]
14:17:48 ERROR: Flash DRIVE OS - flash: command terminated with error
**14:17:48 SUMMARY: Flash DRIVE OS - flash: First Error: No such file or directory**

PROBLEM 4: installation of some packages requires interactive user-input for NV_WORKSPACE

I tried to export NV_WORKSPACE in .bashrc and .profile, but still the interactive request comes up (and of course in SDK manager i cannot type in anything to make it happy).

SDK manager lets a apt-get install producedure hanging waiting for user-input

tje2abt@ABTZ0O1F:~$ ps -aux | grep apt
root      832928  0.0  0.0  37108  5908 ?        S    10:13   0:00 sudo -E apt-get install -y --allow-downgrades /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/cuda-repo-ubuntu2004-11-4-local_11.4.17-470.129.06-1_amd64.deb
root      832930  0.1  0.2  97496 75892 ?        S    10:13   0:06 apt-get install -y --allow-downgrades /home/tje2abt/workspace/nvidia-driveos-6.0.4-linux-download/cuda-repo-ubuntu2004-11-4-local_11.4.17-470.129.06-1_amd64.deb
tje2abt   852171  0.0  0.0   9036   660 pts/1    S+   11:37   0:00 grep --color=auto apt

And when i kill this one and run

sudo dpkg --configure -a

It lets me input NV_WORKSPACE interactively and i can continue, but still CUDA is shown as not installed successfully.

SUMMARY:
I tried to cleanup everything with.

#!/bin/bash
sudo rm -rf $NV_WORKSPACE
mkdir $NV_WORKSPACE
rm -rf ~/nvidia/
rm -rf ~/.nvsdkm/
sudo -E apt-get -y --purge remove nv-driveos*

But again, i get in the same situation with problems above.

Attached you find the .nvsdkm folder after several attempts …
.nvsdkm.zip (88.4 KB)

2

Hi @janosch.steinhoff
It seems a few environment problems on your host systems.
Please try if DRIVE OS Docker Containers installation method mentioned in DRIVE OS 6.0 Installation Guide for NVIDIA Developers will be easier for your situation.

3

@janosch.steinhoff may I know if you are able to access the DRIVE AGX SDK Program files (e.g. locked documentation) from “behind your corporate proxy”? e.g. those with the locked symbol in NVIDIA DRIVE Documentation | NVIDIA Developer as below.

image

4

@janosch.steinhoff
The reflash issue was fixed in SDK Manager 1.8.3.
Please download the latest NVIDIA SDK Manager and try it. Thanks.

6

Hi @VickNV thanks a lot for your replies!

Regarding docker:
access to the containers in NGC was not working for me, but i now got response and workaround from our key-account manager, i will try asap

PROBLEM 1: Login to developer.nvidia.com does not work behind corporate proxy
I can access also documents with lock symbol from the documentation page. But i found that my browser uses a different proxy setting, so i need to further check on my end regarding this.

Problem 3: OS image ready but flashing fails with “file not found”
I tried the new version SDK Manager 1.8.3, thanks for the hint. This proceeds a little further but then gets stuck interactively asking for the sudo password (which it asked for already before). From the log i could see that it actually just wants to run /tmp/tmp_NV_DRIVE_FLASH_DRIVE_COMP.sh , so i did this manually.

Next it failed getting the board revision, this i was able to overcome by setting the board in recovery manually via the Aurix console.

Now i fail with
grafik

Any proposal how to continue with that? (Other than docker, which i plan to try anyhow)

7

.nvsdkm.zip (144.1 KB)

8

We will check the logs. In the meantime, please help see if DRIVE Orin DevKit SDK manager flashing failed to version DRIVE OS 6.0.4 - #24 by VickNV helps with this.

9

From the log, we could see that sudo password was provided automatically. However, we could not tell why it hang there. Does this issue always reproduce?

11:35:04.831 - info: NV_DRIVE_FLASH_DRIVE_COMP@DRIVE_AGX_ORIN: host sudo password for tje2abt:
11:35:04.834 - info: [HOST PASSWORD]
11:52:33.630 - info: NV_CUDA_HOST_COMP@host - download is paused

10

@janosch.steinhoff may I know the status of this?

11

Hello @VickNV sorry for the delay, holidays came inbetween.

However, we could not tell why it hang there. Does this issue always reproduce?

Yes, this does always happen.

In the meantime, please help see if DRIVE Orin DevKit SDK manager flashing failed to version DRIVE OS 6.0.4 - #24 by VickNV

I could not find any hints there which i didnt follow up anyhow in the meantime.

Finally i also got access to NGC (there seems to be a bug in the NGC website so the containers were not showing) and i was able to give the “docker way” a try.
It works.

So from my side interest in SDK-Manager is limited now that i have a working solution, but if you need me to try out something once again i am open to do so.

1 Like
12

Please help try on another host system at your convenience. Thanks.

13

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