Ticket #506 (closed defect: elsewhere)

Opened 3 years ago

Last modified 3 years ago

Crash when snd_pcm_open is called from a thread with "pulse" as playback

Reported by: castagne Owned by: lennart
Milestone: Component: module-alsa-*
Keywords: thread alsa Cc:

Description

On Fedora10, up to date, systematic crash using alsa lib.

pulseaudio 0.9.14 installed

Whenever

snd_pcm_open

has been called from a pthread using "pulse" as playback then the program crashes when the thread terminates.

Crash in nptl_deallocate_tsd

I think this may really relate to pulseaudio

Sample code (very simple) included. Back trace : Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0xb7fdbb90 (LWP 15226)] 0x0517a300 in ?? () (gdb) where #0 0x0517a300 in ?? () #1 0x00fa45ca in nptl_deallocate_tsd () at pthread_create.c:154 #2 start_thread (arg=0xb7fdbb90) at pthread_create.c:304 #3 0x05b9304e in clone () at ../sysdeps/unix/sysv/linux/i386/clone.S:130

Attachments

main.cpp (2.7 kB) - added by castagne 3 years ago.
Simple sample code demonstarting the issue

Change History

Changed 3 years ago by castagne

Simple sample code demonstarting the issue

Changed 3 years ago by lennart

Hmm, I cannot reproduce this here. The test passes fine both with and without PA running. Any hint how I can convince it to be triggered more reliably?

Changed 3 years ago by lennart

I tried to reproduce this on both x86 and x86-64 with no luck. I tried both the java example from:

https://bugzilla.redhat.com/show_bug.cgi?id=477203

And your C++ example from:

https://bugzilla.redhat.com/show_bug.cgi?id=482797

I did test this against a glibc from Rawhide though. Any chance you could try to rebuild that glibc for F10 and check if the problem persists? I am tempted to point to glibc for this issue given that this smells like a TLS issue.

Changed 3 years ago by castagne

Hi there.

Good news,

I upgraded yesterday my 3 machines, and the bug I had with the sample code main.cpp disappeared....

Well, really dunno where it originated.

However, the original bug I had found with portaudio is still there. Perhaps we can close this issue on pulseaudio, and focus on this older bug.

It is reported there : https://bugzilla.redhat.com/show_bug.cgi?id=482797

Changed 3 years ago by lennart

  • milestone 0.9.15 deleted

Changed 3 years ago by lennart

  • status changed from new to closed
  • resolution set to elsewhere

Seems to be a problem in portaudio. Further discussion should happen on that rhbz bug.

Note: See TracTickets for help on using tickets.