CUDA installation failing while downgrading to Drive OS 6.0.5 with SDK Manager

Please provide the following info (tick the boxes after creating this topic):
Software Version
DRIVE OS 6.0.8.1
DRIVE OS 6.0.6
DRIVE OS 6.0.5
DRIVE OS 6.0.4 (rev. 1)
DRIVE OS 6.0.4 SDK
other

Target Operating System
Linux
QNX
other

Hardware Platform
DRIVE AGX Orin Developer Kit (940-63710-0010-300)
DRIVE AGX Orin Developer Kit (940-63710-0010-200)
DRIVE AGX Orin Developer Kit (940-63710-0010-100)
DRIVE AGX Orin Developer Kit (940-63710-0010-D00)
DRIVE AGX Orin Developer Kit (940-63710-0010-C00)
DRIVE AGX Orin Developer Kit (not sure its number)
other

SDK Manager Version
1.9.3.10904
other

Host Machine Version
native Ubuntu Linux 20.04 Host installed with SDK Manager
native Ubuntu Linux 20.04 Host installed with DRIVE OS Docker Containers
native Ubuntu Linux 18.04 Host installed with DRIVE OS Docker Containers
other

I want to downgrade my Drive AGX Orinkit from DriveOS 6.0.8.1 (docker flashing) to use the SDK flashing with DriveOS 6.0.5. However, I am unable to flash the software because of the error here:

16:22:45 ERROR: CUDA on Host - host: E: Version '11.4.18*' for 'cuda-toolkit-11-4' was not found
16:22:45 ERROR: CUDA on Host - host: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo -E apt-get -y --allow-downgrades --allow-downgrades install cuda-toolkit-11-4=11.4.18* ; [error]: exit status 100; [deb_path]: /home/nitest/Downloads/nvidia/sdkm_downloads/cuda-repo-ubuntu2004-11-4-local_11.4.18-470.141.03-1_amd64.deb
16:22:45 ERROR: CUDA on Host - host: command terminated with error
16:22:45 SUMMARY: CUDA on Host - host: Failed to install debian repository file [host]
16:28:40 ERROR: CUDA on Host - host: E: Version '11.4.18*' for 'cuda-toolkit-11-4' was not found
16:28:40 ERROR: CUDA on Host - host: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo -E apt-get -y --allow-downgrades --allow-downgrades install cuda-toolkit-11-4=11.4.18* ; [error]: exit status 100; [deb_path]: /home/nitest/Downloads/nvidia/sdkm_downloads/cuda-repo-ubuntu2004-11-4-local_11.4.18-470.141.03-1_amd64.deb
16:28:40 ERROR: CUDA on Host - host: command terminated with error
16:28:40 SUMMARY: CUDA on Host - host: Failed to install debian repository file [host]

Dear @chris.hoffman,
I see the direct migration from 6.0.8.1 → 6.0.5 is not valid per Requirements for Your Development Environment | NVIDIA Docs.
May I know why you want to move back to 6.0.5?

@SivaRamaKrishnaNV, thanks for the feedback. As a company we have the same Drive AGX Orin kit in different regions, and one of the regions already got the camera working with their kit. Unfortunately, their Drive OS is 6.0.5 and I am trying to downgrade so that there is scope for reuse here.

Is there any way this downgrade can work? It will be really helpful in this case.

@SivaRamaKrishnaNV, just to clarify - I am getting this CUDA error in the installation process on the SDK manager itself, even before flashing.


What camera module is used? Is it listed on DRIVE AGX Orin Sensors & Accessories | NVIDIA Developer. Can you provide more details about the specific issues you’re encountering with the camera module on 6.0.8.1?

Please refer to the following documents to flash using the docker method:

@VickNV I tried using an existing camera we had which is the LI-IMX490-GMSL2-120H with 6.0.8.1. Before ordering a new camera, we want to reuse the other region’s work and for that we need to downgrade. At the time of first time flashing, the SDK manager didn’t work for our kit, so the only option then was to flash 6.0.8.1 using the docker.

Per valid migration paths table, the recommended path is 6.0.8.1 → 6. 0.6(flash target using NGC docker) → 6.0.5.

Just to confirm, did you install any DRIVE OS versions on host already using sdkmanager or do you have any CUDA version installed on host already. If so, it could lead to errors when installing DRIVE OS on host. Ideally, we expect them to be removed before installing using sdkmanager.

However, we recommend to check flashing using docker as Vick suggested to avoid host environment related issues.

6.0.8.1 docker image : nvcr.io/drive/driveos-sdk/drive-agx-orin-linux-aarch64-sdk-build-x86:6.0.8.1-0006
6.0.6 docker Image : nvcr.io/drive/driveos-sdk/drive-agx-orin-linux-aarch64-sdk-build-x86:6.0.6.0-0004
6.0.5 docker Image : nvcr.io/drive/driveos-sdk/drive-agx-orin-linux-aarch64-sdk-build-x86:6.0.5.0-0003

I used NGC docker to flash 6.0.8.1 (tag: latest ) and its currently already on the Drive Orin kit. I just removed all downloaded sdk versions from the host using the Uninstall option on SDK manager. Would I still have issues with CUDA in this case?

Also, based on @VickNV’s suggestion I just downloaded the 6.0.5 docker image. Are you now recommending I flash 6.0.6 with docker and then move to 6.0.5 with docker again?

Dear @chris.hoffman,
If you remove previous DRIVEOS versions and cuda from host, I expect the sdkmanager to install DRIVEOS on host without any issue. But let’s try to flash using docker first.
Please check flashing using 6.0.5 docker directly and let us know if you see any issue. I have not tested DRIVE OS 6.0.8.1-> 6.0.5 using docker.

@SivaRamaKrishnaNV I used the docker to flash 6.0.5 and got the output below:


I used minicom for terminal emulation and accepted EULA, setup usernames and got the output in the attached
minicom_log_6.0.5.txt (13.5 KB)
. Is this something that needs to be fixed?

Time taken from Error Reporting to SEH: 59 microseconds
DemoApp: ErrCode-0x18550080 ReptrId-0x8001 ErrAttr-0x0
EPS TimeStamp: 0xe9f752a
��verify_rate_range: FMON_DSC: rate 3515625 below min��Error Report to CcplexApp failed 
Error Report to CcplexApp fail�� 18750000
fmon_update_config: FMON_DSC: detected fault 0x80
verify_rate_rang��ed 
��e: FMON_DSC: rate 3515625 below min 18750000
fmon_update_config: FMON_DSC: detected fault 0x80
��
Time taken from Error Reporting to SEH: 60 microseconds
DemoApp: ErrCode-0x18550080 ReptrId-0x8001 ErrAttr-0x0
EPS TimeStamp: 0x1111876a

Time taken from Error Reporting to SEH: 59 microseconds
DemoApp: ErrCode-0x18550080 ReptrId-0x8001 ErrAttr-0x0
EPS TimeStamp: 0x11120922

Am I all set here with the flashing of Drive OS 6.0.5, is there anything else that needs to be verified?

Dear @chris.hoffman,
Flashing seems to be successful. Were you able to login to target? I see tegra ubuntu login screen in the log
tegra-ubuntu login: ��verify_rate_range: FMON_DSC: rate 12800000 below min 18750000

press ENTER and see if you can see login screen to enter user credentials?

@SivaRamaKrishnaNV, yes I was able to login to the system with the credentials I set. Thanks.

Another question I had was with respect to the supported camera & its drivers vs DriveOS version. What is the connection here? Can drivers not be built for later OS versions?

Please create another topic for your question about supported cameras and drivers. Thank you!

1 Like

edit: I was able to login to the system via minicom and via ssh.

What is the command to check the version of DriveOS on the Drive AGX orin?

on Orin:

$ cat /etc/nvidia/version-ubuntu-rootfs.txt

on Aurix:

NvShell>version
1 Like


Confirmed, thanks!

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