Bug#1007832: vulkan-tools: vulkaninfo,vkcube seg-faults on head-less cuda/gpu machine

[Alois Schlögl <alois.schloegl@ist.ac.at>]

Package: vulkan-tools
Version: 1.2.162.0+dfsg1-1
Severity: normal
X-Debbugs-Cc: alois.schloegl@ist.ac.at

Dear Maintainer,

*** Reporter, please consider answering these questions, where 
appropriate ***

   * What led up to the situation?

     Trying to run wine-staging on a GPU/cuda compute node failed 
because vulkan was not correctly configured.
     see also https://bugs.winehq.org/show_bug.cgi?id=52647
     In the course of investigation, I noticed that vulkaninfo did 
always seg-fault on headless cuda/GPU machines under Debian11.


   * What exactly did you do (or not do) that was effective (or
     ineffective)?

     I've tried to install and uninstall various graphics/vulkan driver.
     Testing a machine without GPU's did not show this issue.

     Here is the output when running the debugger on vulkaninfo

    $ gdb vulkaninfo
    For help, type "help".
    Type "apropos word" to search for commands related to "word"...
    Reading symbols from vulkaninfo...
    (No debugging symbols found in vulkaninfo)
    (gdb) r
    Starting program: /usr/bin/vulkaninfo
    [Thread debugging using libthread_db enabled]
    Using host libthread_db library 
"/lib/x86_64-linux-gnu/libthread_db.so.1".

    Program received signal SIGSEGV, Segmentation fault.
    0x000015554ceb23fc in ?? () from 
/usr/lib/x86_64-linux-gnu/libvulkan_intel.so
    (gdb) Quit
    quit)

     The output of these commands is attached

        $ vulkaninfo
        $ vkcube
        $ vkcubepp

        $ lspci|grep VGA
        $ nvidia-smi

        $ ls -altr  /usr/share/vulkan/icd.d/
        $ cat  /usr/share/vulkan/icd.d/*

        $ dpkg -l|grep "icd\|xorg\|nvid\|cuda\|mesa\|vulkan"

        $ glxinfo

   * What was the outcome of this action?

     vulkaninfo, vkcube always ended up in a seg-fault when running on 
Cuda/GPU machines.
     Doing this on machines without Nvidia GPU's does not show this issue.


   * What outcome did you expect instead?
     I understand that Cuda is proprietary technology and therefore 
this platform is difficult to support.

     I'd expect that vulkaninfo does not seg-fault, and finishes 
gracefully with a more or less meaning full error message.
     Ideally, I should indicate what is needed to fix this issue.

     Afterall, it should be possible to run vulkaninfo on a headless 
GPU/cuda machine, that should enable also
     runing wine-staging on that machine.


*** End of the template - remove these template lines ***


-- System Information:
Debian Release: 11.2
  APT prefers stable
  APT policy: (990, 'stable'), (500, 'stable-security')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 5.10.0-12-amd64 (SMP w/24 CPU threads)
Kernel taint flags: TAINT_PROPRIETARY_MODULE, TAINT_OOT_MODULE, 
TAINT_UNSIGNED_MODULE
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8), LANGUAGE 
not set
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled

Versions of packages vulkan-tools depends on:
ii  libc6               2.31-13+deb11u2
ii  libgcc-s1           10.2.1-6
ii  libstdc++6          10.2.1-6
ii  libvulkan1          1.2.162.0-1
ii  libwayland-client0  1.18.0-2~exp1.1
ii  libx11-6            2:1.7.2-1
ii  libxcb1             1.14-3

vulkan-tools recommends no packages.

vulkan-tools suggests no packages.

-- no debconf information