SDK Manager Connecting to Jetson TX2: Incorrect Username and Password

I am able to ping and ssh the Jetson TX2 with my username and password from my host Ubuntu machine. But When I tried to install SDK components (cuda) to the TX2 through the SDK manager, it always says my username and password are incorrect. I am able to ssh to the TX2 with the same username and password.

I have been stuck on this for a while. I simply need to get cuda installed, in order to get PyTorch on the TX2, and then to get libtorch working for my C++ neural network.

Hi xiao3,

Can you please package and attach the ~/.nvsdkm directory for more investigation?

This is the ~/.nv directory on the TX2. nv.zip (312.6 KB)

Hi xiao3,

Can you please package and attach the ~/.nvsdkm directory on host instead of ~/.nv directory on TX2?

This is the ~/.nvsdkm on the host nvsdkm.zip (371.4 KB)

Thanks xiao3.

SDK Manager tries to automatically fill in password according to prompt in terminal. It seems that the default password prompt is changed when setting up ssh connection to Jetson TX2. SDK Manager thought the prompt was sent from host thus entered host password. The default password prompt pattern is:
@<TX2_IP_Address>'s password:
Now the prompt is:
Password:

There are two ways to solve the problem:

  1. Reset the password prompt when ssh connectting to Jetson TX2.
  2. Temporarily change the TX2 password to be the same as host sudo password.

It looks like option 1 solve the problem for me! I tried option 2 first, but it didn’t work. Then I tried option 1 (with option 1 also done), it starts to work!

Thank you so much EdwardZhou! I have been stuck on this for a while and it finally works out!

However, while running installation on the TX2, OpenCV failed with

01:06:46 INFO: OpenCV on Target: conflicting packages - not installing libopencv-dev

01:06:46 INFO: OpenCV on Target: Errors were encountered while processing:

01:06:46 INFO: OpenCV on Target: /opt/nvidia/deb_repos/OpenCV-4.1.1-2-gd5a58aa75-aarch64-dev.deb

01:06:46 INFO: OpenCV on Target: Unhandled error when running sudo dpkg -i /opt/nvidia/deb_repos/OpenCV-4.1.1-2-gd5a58aa75-aarch64-dev.deb

01:06:46 INFO: OpenCV on Target: [ Package Install Finished with Error ]

01:06:46 INFO: OpenCV on Target: [ NV_L4T_OPENCV_TARGET_POST_INSTALL_COMP Install took 48s ]

01:06:46 ERROR: OpenCV on Target: [error]: dpkg install on target failed; [exec_command]: safe_apt_install “sudo dpkg -i /opt/nvidia/deb_repos/OpenCV-4.1.1-2-gd5a58aa75-aarch64-dev.deb” 5; sleep 5; ; [deb_path]: /home/xuesu/Downloads/nvidia/sdkm_downloads/OpenCV-4.1.1-2-gd5a58aa75-aarch64-dev.deb

01:06:46 ERROR: OpenCV on Target: command terminated with error

01:06:46 SUMMARY: OpenCV on Target: Failed to install debian package [target]

And VPI failed with

00:59:22 INFO: VPI on Target: verifying checksum of /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-lib-0.4.4-cuda10-aarch64-l4t.deb

00:59:22 INFO: VPI on Target: verifying checksum of /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-dev-0.4.4-cuda10-aarch64-l4t.deb

00:59:22 INFO: VPI on Target: verifying checksum of /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-samples-0.4.4-cuda10-aarch64-l4t.deb

00:59:22 INFO: VPI on Target: Found file /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-dev-0.4.4-cuda10-aarch64-l4t.deb with correct checksum, skip downloading.

00:59:23 INFO: VPI on Target: Found file /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-samples-0.4.4-cuda10-aarch64-l4t.deb with correct checksum, skip downloading.

00:59:23 INFO: VPI on Target: Found file /home/xuesu/Downloads/nvidia/sdkm_downloads/vpi-lib-0.4.4-cuda10-aarch64-l4t.deb with correct checksum, skip downloading.

01:06:46 SUMMARY: VPI on Target: Depends on failed component

Anyway to at least fix OpenCV?

Hi xiao3,

Each version of JetPack is shipped with a bundle of compatible softwares. Target components, like CUDA, OpenCV, etc, shipped in JetPack version A, may be incompatible to L4T OS shipped in JetPack version B. To install target components on it, please make sure the L4T OS on target is compatible. If not sure, please flash Jetson device with the same version of JetPack.