[pulseaudio-discuss] [PATCH] loopback: Fix crash if Max Request change at unload

[Tanu Kaskinen]

On Fri, 2012-03-23 at 16:19 +0100, David Henningsson wrote:
> On 03/23/2012 02:47 PM, Dalleau, Frederic wrote:
> > Hi David,
> >
> >> For reference, the stack trace is here:
> >>
> >> https://bugs.launchpad.net/bugs/946400
> >
> > It is likely that the issues related : here is the backtrace I got
> > after applying your patch (without mine)
> >
> > #0  0xb7e9b520 in adjust_rates (u=0x80b3668) at modules/module-loopback.c:174
> > #1  0xb7e9d372 in sink_input_process_msg_cb (obj=0x80f2728, code=10,
> > data=0x0, offset=0, chunk=0x0)
> >      at modules/module-loopback.c:530
> > #2  0xb7c31f9b in pa_asyncmsgq_dispatch (object=0x80f2728, code=10,
> > userdata=0x0, offset=0, memchunk=0xbfffe96c)
> >      at pulsecore/asyncmsgq.c:322
> > #3  0xb7c320e1 in pa_asyncmsgq_flush (a=0x8100ea0, run=true) at
> > pulsecore/asyncmsgq.c:347
> > #4  0xb7c990fa in pa_thread_mq_done (q=0x80a7da0) at pulsecore/thread-mq.c:107
> > #5  0xb7ced418 in stop_thread (u=0x80a7d70) at
> > modules/bluetooth/module-bluetooth-device.c:2455
> > #6  0xb7cedf1f in card_set_profile (c=0x80a6a98, new_profile=0x80b1ed8)
> 
> This is puzzling. Exact what is it failing on and why? Could you set a 
> breakpoint right before the crash, and examine the relevant values?

At least in the situation where the
SINK_INPUT_MESSAGE_MAX_REQUEST_CHANGED message is processed after
module-loopback has been unloaded, adjust_rates() will crash if it's
called. Your patch sets adjust_time to zero at teardown, which reduces
the changes of adjust_rates() getting called, but when module-loopback
gets unloaded, that particular memory location may get overwritten with
a non-zero value.

-- 
Tanu