Dear Maintainer,
I was trying to collect some more information for bug #973811.
There it looks like openblas does its own memory management seems
therefore affected by some sandboxing disallowing a SYS_bind syscall.

With the test VM I used there I get with current testing this
output for a manually created test mkv with h264 video the
following output.

This is now kind of different, but might be caused by some
new package versions since last October.
Therefore I am not sure if this is really what I observe.

It seems like openblas tries the allocators in variable memoryalloc,
and if no one works calls the terminating null pointer with
the current version.

Kind regards,
Bernhard



# ffmpeg -f lavfi -i testsrc=duration=10:size=1280x720:rate=30 -vcodec h264 
-acodec libvorbis output.mkv

benutzer@debian:~$ totem-video-thumbnailer -v output.mkv tmp.png
TotemVideoThumbnailer-Message: 17:37:58.702: Initialised libraries, about to 
create video widget
TotemVideoThumbnailer-Message: 17:37:58.710: setting URI 
file:///home/benutzer/output.mkv
TotemVideoThumbnailer-Message: 17:37:58.710: Video widget created
TotemVideoThumbnailer-Message: 17:37:58.710: About to open video file
OpenBLAS blas_thread_init: pthread_create failed for thread 6 of 16: Die 
Ressource ist zur Zeit nicht verfügbar
OpenBLAS blas_thread_init: RLIMIT_NPROC 11759 current, 11759 max
Speicherzugriffsfehler (Speicherabzug geschrieben)

benutzer@debian:~$ coredumpctl -q list
TIME                            PID   UID   GID SIG COREFILE  EXE
Wed 2021-04-07 17:37:59 CEST   3087  1000  1000  11 present   
/usr/bin/totem-video-thumbnailer

benutzer@debian:~$ coredumpctl -q gdb 3087
           PID: 3087 (totem-video-thu)
           UID: 1000 (benutzer)
           GID: 1000 (benutzer)
        Signal: 11 (SEGV)
     Timestamp: Wed 2021-04-07 17:37:58 CEST (7min ago)
  Command Line: totem-video-thumbnailer -v output.mkv tmp.png
    Executable: /usr/bin/totem-video-thumbnailer
 Control Group: /user.slice/user-1000.slice/session-8.scope
          Unit: session-8.scope
         Slice: user-1000.slice
       Session: 8
     Owner UID: 1000 (benutzer)
       Boot ID: 505eeb0b4fe548338077ab20802215cc
    Machine ID: 33f18f39d2a9438eb75b0ed52848afcd
      Hostname: debian
       Storage: 
/var/lib/systemd/coredump/core.totem-video-thu.1000.505eeb0b4fe548338077ab20802215cc.3087.1617809878000000.zst
       Message: Process 3087 (totem-video-thu) of user 1000 dumped core.
Stack trace of thread 3096:
                #0  0x0000000000000000 n/a (n/a + 0x0)

...
Core was generated by `totem-video-thumbnailer -v output.mkv tmp.png'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x0000000000000000 in ?? ()
[Current thread is 1 (Thread 0x7f4ab3aa6700 (LWP 3096))]
(gdb) bt
#0  0x0000000000000000 in  ()
#1  0x00007f4ab5364709 in blas_memory_alloc (procpos=procpos@entry=2) at 
memory.c:2793
#2  0x00007f4ab5364f04 in blas_thread_server (arg=<optimized out>) at 
blas_server.c:366
#3  0x00007f4ade5deea7 in start_thread (arg=<optimized out>) at 
pthread_create.c:477
#4  0x00007f4adebdadef in clone () at 
../sysdeps/unix/sysv/linux/x86_64/clone.S:95
(gdb) cd 
/home/benutzer/source/libopenblas0-pthread/orig/openblas-0.3.13+ds/driver/others
Working directory 
/home/benutzer/source/libopenblas0-pthread/orig/openblas-0.3.13+ds/driver/others.
(gdb) up
#1  0x00007f4ab5364709 in blas_memory_alloc (procpos=procpos@entry=2) at 
memory.c:2793
2793            map_address = (*func)((void *)base_address);
(gdb) print memoryalloc
$1 = {0x7f4ab5364230 <alloc_mmap>, 0x7f4ab53641d0 <alloc_malloc>, 0x0}

--
debian-science-maintainers mailing list
[email protected]
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers

Reply via email to