GNOME Bugzilla – Bug 537539
[pulse] Rhythmbox starts using 100% of CPU time when pulseaudio gets killed
Last modified: 2009-01-29 13:32:22 UTC
Moved from http://pulseaudio.org/ticket/188: When rhythmbox is playing through GStreamer and PulseAudio, and PulseAudio gets killed, Rhythmbox will start eating 100% of CPU time and memory usage will grow until the machine runs out of memory. I submitted this first to the GNOME/Rhythmbox bugzilla as I thought this was a Rhythmbox problem, but there they state that this would be a gst-pulse problem. Can you comment on this? The Rhythmbox bug report: http://bugzilla.gnome.org/show_bug.cgi?id=503126 ----------------------------------------------------------------------------- 03/04/08 12:30:23 changed by coling Hi Fred, I'm not seeing 100% CPU but I do get a deadlock in rhythmbox - e.g. it totally freezes. As pointed out by CJ on the pa ML, this deadlock I'm seeing is possibly due to #251. 03/04/08 12:32:36 changed by coling ----------------------------------------------------------------------------- I should clarify, that I have seem other apps exhibit high CPU and memory eating characteristics when killing the pa server or just the stream. This usually happens for me when the underlying ALSA driver can no longer output and the gstreamer streams somehow just freeze in pa and sit there. I can usually trigger this specific behaviour when suspending/resuming (about one in three times).
Yesterday I ran into a very very similar bug where totem started to spin 100% CPU with uncontrolled VIRT growth. This happened immediately when I typed "killall pulseaudio". However, generally what happens when I play something in totem using pulse-backend and then kill pulse is that totem GUI freezes with an empty error messagebox (no CPU spin and no VIRT growth) so there is clearly some particular condition which put be met for the "killall pulseaudio" to go into CPU spin mode. That said, I was able to get a lot of info out of gdb attached to totem when it was spinning and that information was sufficient for wim to fix that bug. My totem+pulse CPU spin was reported here: http://bugzilla.gnome.org/show_bug.cgi?id=554771
*** Bug 525169 has been marked as a duplicate of this bug. ***
This has been fixed in GIT and 0.10.14 will contain the fix.