Bug#816283: nvidia-cuda-toolkit: cuda broken after upgrade (x86_64, Jessie, GTX980)

Alois Schloegl schloegl at ist.ac.at
Mon Feb 29 12:38:57 UTC 2016


Package: nvidia-cuda-toolkit
Version: 7.0.28-4
Severity: important

Dear Maintainer,



I tried to upgrade the nvidia-drivers from 352.41-1
to  the current version "352.79-3" and nvidia-cuda-toolkit to a matching
version, cuda became unusable. Rebooting the machine did not help.

I used the following two tests.

First, nvidia-smi reports this error:

  # nvidia-smi
  Failed to initialize NVML: GPU access blocked by the operating system

Second, I do have a short cuda program from here
https://github.com/leschzinerlab/Motion-correction.git
and can be installed with the following commands

    git clone https://github.com/leschzinerlab/Motion-correction.git
    cd Motion-correction/motioncorr_v2.1/src/
    make -B
    ../bin/gpuinfo

The compilation works without problems, but running the command
     ../bin/gpuinfo
results in this message:

    Simple GPU info query.

    You have 32720 nVidia GPGPU.

    DeviceID Name                    Version Memory(Mb)
    Failed to get GPU info of Device #0
    Failed to get GPU info of Device #1
    Failed to get GPU info of Device #2
    Failed to get GPU info of Device #3
    ...
    Failed to get GPU info of Device #32568
    Failed to get GPU info of Device #32569
    Failed to get GPU info of Device #32570
    Failed to get GPU info of Device #32571


I tried a number of additional things, like using the packages in
jessie-backports, and  nvidia-cuda 7.0 from debian/unstable. None of
these attempts solved the issue.


Best,
   Alois



-- System Information:
Debian Release: 8.3
  APT prefers stable
  APT policy: (990, 'stable'), (500, 'unstable'), (500, 'testing')
Architecture: amd64 (x86_64)

Kernel: Linux 3.16.0-4-amd64 (SMP w/12 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)

Versions of packages nvidia-cuda-toolkit depends on:
ii  g++-4.8                          4.8.4-1
ii  g++-4.9                          4.9.2-10
ii  gcc-4.8                          4.8.4-1
ii  gcc-4.9                          4.9.2-10
ii  libc6                            2.19-18+deb8u3
ii  libgcc1                          1:5.3.1-8
ii  libnvvm3                         7.0.28-4
ii  libstdc++6                       5.3.1-8
ii  nvidia-cuda-dev                  7.0.28-4
ii  nvidia-profiler                  7.0.28-4
ii  ocl-icd-opencl-dev [opencl-dev]  2.2.3-1+deb8u1

Versions of packages nvidia-cuda-toolkit recommends:
ii  nvidia-cuda-doc         7.0.28-4
ii  nvidia-cuda-gdb         7.0.28-4
pn  nvidia-visual-profiler  <none>

Versions of packages nvidia-cuda-toolkit suggests:
pn  libcupti-dev   <none>
ii  nvidia-driver  352.79-4

-- no debconf information



More information about the pkg-nvidia-devel mailing list