L4T R23.2 for Jetson TX1 released

I have the same problem, so very eager to see a response to this post.

Sorry - this should be linked to a previous post by another member - but I can’t see how to do that.

I have the same problem so very eager to see a response to this post.

Eclectius,

Please try to manually install CUDA on host and see what’s the root cause:

  1. dpkg -i <JetPack_Dir>/jetpack_download/cuda-repo-ubuntu1404-7-0-local_7.0-73_amd64.deb;
  2. dpkg --add-architecture armhf;
  3. apt-get update;
  4. apt-get -y --force-yes install g+±arm-linux-gnueabihf gcc-arm-linux-gnueabihf cuda-toolkit-7-0 cuda-cross-armhf-7-0 libgomp1 libgomp1-armhf-cross libc6-armhf-cross libsfstdc++6-armhf-cross;

The reason for the 404 errors is explained in http://docs.nvidia.com/cuda/cuda-installation-guide-linux/#faq4
[i]Why do I see multiple “404 Not Found” errors when updating my repository meta-data on Ubuntu?

These errors occur after adding a foreign architecture because apt is attempting to query for each architecture within each repository listed in the system’s sources.list file. Repositories that do not host packages for the newly added architecture will present this error. While noisy, the error itself does no harm. Please see the Advanced Setup section for details on how to modify your sources.list file to prevent these errors.[/i]

I’m getting following error when installing JetPack 2.1 on my ubuntu 14.04 which already has CUDA 7.0-73 amd64 installed.
The installation stops at last line…
I tried removing all entries from /etc/apt/sources.list.d.
Any help urgently requested… I have a big presentation upcoming…

Thanks

W: Failed to fetch http://us.archive.ubuntu.com/ubuntu/dists/trusty/main/binary-armhf/Packages 404 Not Found [IP: 2001:67c:1562::19 80]

W: Failed to fetch http://us.archive.ubuntu.com/ubuntu/dists/trusty/restricted/binary-armhf/Packages 404 Not Found [IP: 2001:67c:1562::19 80]

W: Failed to fetch http://us.archive.ubuntu.com/ubuntu/dists/trusty/universe/binary-armhf/Packages 404 Not Found [IP: 2001:67c:1562::19 80]

W: Failed to fetch http://us.archive.ubuntu.com/ubuntu/dists/trusty/multiverse/binary-armhf/Packages 404 Not Found [IP: 2001:67c:1562::19 80]

W: Failed to fetch http://extras.ubuntu.com/ubuntu/dists/trusty/main/binary-armhf/Packages 404 Not Found [IP: 2001:67c:1360:8c01::23 80]

E: Some index files failed to download. They have been ignored, or old ones used instead.
Reading package lists…
Building dependency tree…
Reading state information…
E: Unable to locate package cuda-toolkit-7-0
E: Unable to locate package cuda-cross-armhf-7-0

BillMacc,

Please firstly remove all cuda packages from your host.

  1. use “dpkg -l | grep cuda” to find all existing CUDA packages on your host.
  2. remove all of them by “apt-get remove <cuda_package_name>” command. As you removed all entries from /etc/apt/sources.list.d/, you might encounter problem when using “apt-get remove” command. If this is the case, please use “dpkg --purge <cuda_package_name>” command instead.
  3. use “dpkg -l | grep cuda” to check that no CUDA packages remaining on your host.

Then re-install CUDA with steps in #23 and see if there is any problem.

Hi Edward,

Thanks for your response. I performed the 3 steps as you suggested and it showed no remaining CUDA packages. I then performed the 4 steps #23. When done manually gives same errors as the cuda_host_tx1.log file:

Reading package lists…
Building dependency tree…
Reading state information…
g+±arm-linux-gnueabihf is already the newest version.
gcc-arm-linux-gnueabihf is already the newest version.
libc6-armhf-cross is already the newest version.
libc6-armhf-cross set to manually installed.
libgomp1-armhf-cross is already the newest version.
libgomp1-armhf-cross set to manually installed.
libsfstdc++6-armhf-cross is already the newest version.
libsfstdc++6-armhf-cross set to manually installed.
libgomp1 is already the newest version.
libgomp1 set to manually installed.
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
cuda-toolkit-7-0 : Depends: cuda-samples-7-0 (= 7.0-73) but it is not going to be installed
Depends: cuda-documentation-7-0 (= 7.0-73) but it is not going to be installed
unity-control-center : Depends: libcheese-gtk23 (>= 3.4.0) but it is not going to be installed
Depends: libcheese7 (>= 3.0.1) but it is not going to be installed
E: Error, pkgProblemResolver::Resolve generated breaks, this may be caused by held packages.

I then tried manually installing each of those and they each fail due to dependency on each other???

Any ideas?

Thanks,

bill

Hi Bill,

I googled the error you met and found following bug report.
https://bugs.launchpad.net/ubuntu/+source/unity-control-center/+bug/1428972

Not sure if it helps.

Thanks,
Edward

Should this be updated to 24.1? https://developer.nvidia.com/embedded/linux-tegra

Also when can we expect CUDA x64? Thanks

Thanks for the inquiry, the page will be updated when we release the updated CUDA and JetPack packages next week.

Is there a workaround for this package dependency issue for now ?

Bill

I currently ran into the problem of being unable to install the Cuda 7.0 toolkit necessary to install all the files within L4T for R23.2.I had to install a fresh copy of L4T because the original eMMC memory ran out of space and thus had to resort to using a SSD connected to the TX1 Sata port. (The PCIe x4 port does not work for SSDs for some reason. I have asked a solution to this problem in a separate post) Going back, this installation error is mentioned further by user EdwardZhou, and cannot be simply fixed with manual installation. The four steps he suggests are:

  1. dpkg -i <JetPack_Dir>/jetpack_download/cuda-repo-ubuntu1404-7-0-local_7.0-73_amd64.deb;
  2. dpkg --add-architecture armhf;
  3. apt-get update;
  4. apt-get -y --force-yes install g+±arm-linux-gnueabihf gcc-arm-linux-gnueabihf cuda-toolkit-7-0 cuda-cross-armhf-7-0 libgomp1 libgomp1-armhf-cross libc6-armhf-cross libsfstdc++6-armhf-cross;
    However, at the first step, there is already an error saying the package architecture (amd64) does not match system (armhf). I ignored this warning and continued to steps 2, 3, 4. but Cuda 7.0 would not install. Help is greatly appreciated.

I’m wondering which machine you are installing CUDA to, and which one you are running dpkg on…PC versus JTX1. The reason for this question is that sometimes a machine has more than one architecture involved…first case for cross-tools for building a foreign architecture on a PC to later be copied to the other machine of the foreign architecture…the second case because the machine in question has an ability to support both architectures. This latter might apply on an ARMv8a 64-bit machine because of some compatibility with ARMv8 32-bit (and some ARMv7 32-bit if feature set is correct). On a JTX1 you can find its own native architecture (there by default), and possibly add some compatible 32-bit architecture (you might add this as a compatibility feature with older code), or you could use it as a cross-compiler (unlikely to ever happen, it would more likely be an academic use than practical, though it would work).

The amd64 would be for desktop PC (AMD is x86_64), not for any ARM CPU. On a desktop PC it would be quite plausible to have native 64-bit architecture, a 32-bit compatibility architecture, and multiple cross-compiler architectures not intended to directly run on the PC.

CUDA 7 on a desktop PC will not need any special architecture addition. Nor should a Jetson require additional architectures unless there is some sort of compatibility trick issue. Generally speaking, you need to install CUDA for PC on a PC, and CUDA for ARM on a Jetson. It looks as though CUDA for PC install is being attempted on Jetson. There is a possibility that whatever problems you have will be simply from mixing up the two packages, PC versus ARM.

As for the “–add-architecture armhf”, this would be 32-bit compatibility on a Jetson TX1 (this is native on a Jetson TK1). CUDA 7 is strictly 64-bit, and does not use this…the correct ARMv8a CUDA 7 will install on a JTX1 without a compatibility architecture, although there may be cases where working on 32/64-bit compatibility could use this for other reasons.

is there any release for fixing the usb webcam video capture issue with OpenCV4Tegra? keep getting

error: (-215) size.width>0 && size.height>0 in function imshow

Unable to stop the stream.: Device or resource busy
Unable to stop the stream.: Bad file descriptor
VIDIOC_STREAMON: Inappropriate ioctl for device
Cannot read a frame from video stream
Unable to stop the stream.: Inappropriate ioctl for device

Gary, FRC Team 4001

The new JetPack 2.2 has been posted, see here: https://devtalk.nvidia.com/default/topic/941913/jetson-tx1/jetpack-2-2-l4t-r24-1-64-bit-and-32-bit-cuda-released/

I believe updated OpenCV4Tegra included camera capture fixes. You can use JetPack to install it automatically or download it from here: https://developer.nvidia.com/embedded/linux-tegra

CUDA and Jetpack for R24.1 have now been posted, please see here —https://devtalk.nvidia.com/default/topic/941913/jetson-tx1/jetpack-2-2-l4t-r24-1-64-bit-and-32-bit-cuda-released/

Hi Everyone,

There is a good reason for not using this with Ubuntu 16.04:

E: Failed to fetch file:/var/cuda-repo-7-0-local/Release  No Hash entry in Release file /var/lib/apt/lists/partial/_var_cuda-repo-7-0-local_Release which is considered strong enough for security purposes

With luck someone attends to this soon!

-Mark

Something I only became aware of recently is that the CUDA .deb file not only adds a repository to find the CUDA stuff at, but that the repository is stored locally in “/var” (versus a remote server). I suspect that if there is some Debian package manager magic to forcibly use those CUDA packages without the hash entry, it would work fine. That hash entry is to determine if there was corruption and/or tampering…if you are convinced the file was downloaded correctly and that it came from a valid nVidia download page, then you might as well try to use it. How well it works with Ubuntu 16 is anyone’s guess…give it a shot and find out.

Hi
I have the similar problem in cuda, when install it in TX1 with JetPack L4T 2.2.1
Here is the error:
some packages could not be installed. this may mean that you have requested an impossible situation or if you are using the unstable distribution that some required packages have not yet been created or been moved out of incoming.
the following information may help to resolve the situation:
the following packages have unmet dependencies:
cuda-toolkit-7-0: depends:cuda-core-7-0(=7.0-76)but it is not going to be installed
depends:cuda-command-line-tools-7-0(=7.0-76)but it is not going to be installed
depends:cuda-samples-7-0(=7.0-76)but it is not going to be installed
depends:cuda-documentation-7-0(=7.0-76)but it is not going to be installed
libfreeimage-dev:depends:libfreeimage3(=3.15.4-3)but it is not going to be installed
libopenmpi-dev:depends:libopenmpi1.6(=1.6.5-8)but it is not going to be installed
depends:libibverbs-dev but it is not installable
depends:libhwloc-dev but it is not going to be installed
openmpi-bin:depends:libopenmpi1.6 but it is not going to be installed
connection to ‘tx1 ip’ closed.
dpkg-query:package’cuda-toolkit-7-0’ is not installed and no information is available
dpkg-query:package’libfreeimage-dev’ is not installed and no information is available
dpkg-query:package’libopenmpi-dev’ is not installed and no information is available
dpkg-query:package’openmpi-bin’ is not installed and no information is available
use dpkg --info(=dpkg-deb --info)to examine archive files,and dpkg --contents(=dpkg-deb --contents)to list their contents.
error:CUDA cannot be installed on device.This may be caused by other apt-get command running on device when installing CUDA.Please use apt-get command in a terminal to make sure following packages are installed correctly on device before continueing:
cuda-toolkit-7-0 libgomp1 libfreeimage-dev libopenmpi-dev openmpi-bin
After these packages are installed on device, press Enter key to continue