[pulseaudio-tickets] [Bug 55701] New: module-rescue-streams with LADSPA filter chain causes infinite recursion crash on PA shutdown

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=55701

          Priority: medium
            Bug ID: 55701
                CC: lennart at poettering.net
          Assignee: pulseaudio-bugs at lists.freedesktop.org
           Summary: module-rescue-streams with LADSPA filter chain causes
                    infinite recursion crash on PA shutdown
        QA Contact: pulseaudio-bugs at lists.freedesktop.org
          Severity: normal
    Classification: Unclassified
                OS: Linux (All)
          Reporter: tomastrnka at gmx.com
          Hardware: x86-64 (AMD64)
            Status: NEW
           Version: unspecified
         Component: modules
           Product: PulseAudio

PulseAudio 1.1 (tested with Fedora package pulseaudio-1.1-9.fc17.x86_64)
crashes on every daemon shutdown in my configuration consisting of an
module-alsa-sink (denoted as A) and two chained module-ladspa-sinks (L1, L2):

L2 -> L1 -> A

default sink = A

If module-rescue-streams is enabled, the following happens on daemon shutdown
(even with PA completely idle and no clients connected):

1) A is being unloaded
2) module-rescue-streams kicks in and tries to find a new home for the output
stream of L1
3) find_evacuation_sink() iterates over all sinks and finds L2 as a suitable
candidate
4) module-rescue-streams thus moves L1's output (L1out) to L2, creating a cycle
5) pa_sink_process_msg() processes the PA_SINK_MESSAGE_FINISH_MOVE message and
calls attach() on the pa_sink_input being moved (L1out). This points to
sink_input_attach_cb() of module-ladspa-sink.c
6) sink_input_attach_cb(L1out) calls pa_sink_attach_within_thread(L1out->sink),
where L1out->sink == L1
7) pa_sink_attach_within_thread(L2) iterates over L2's inputs and calls
attach() on each of them, including L2out. Again, this means
sink_input_attach_cb(L2out) is called.
8) sink_input_attach_cb(L2out) calls pa_sink_attach_within_thread(L2)
9) pa_sink_attach_within_thread(L2) calls attach() on L2's inputs:
sink_input_attach_cb(L1out)
10) Go to 6) and repeat until the stack overflows and PA is killed by SEGV.

I'm not sure how to fix this properly since I have only limited understanding
on the internals of PA. It would probably be an overkill for
module-rescue-streams to do full cycle detection somehow. Possibly it would be
best to just fix find_evacuation_sink() to ignore "filter-like" sinks if
there's a way how to recognise them easily (ignoring just module-ladspa-sinks
seems too hackish to me).

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/pulseaudio-bugs/attachments/20121006/275e23e4/attachment.html>