On Fri, 2017-01-27 at 22:08 +0900, Takashi Sakamoto wrote:
> Hi Tanu and Klaus,
>
> On Jan 27 2016 20:26, Tanu Kaskinen wrote:
> > On Tue, 2017-01-24 at 11:35 -0800, Klaus Badelt wrote:
> > > The 32 channel limit excludes use of audio interfaces like for example (my)
> > > RME HDSPe MADI (64x64), all other MADI interfaces (incl. RME MADIface), RME
> > > Fireface UFX+ (94x94), MOTU 1248 (32x34), MOTU 112D (112), Focusrite Red
> > > (64x64), ProTools hardward, and more. In general, we're excluding many pro
> > > audio and prosumer audio interfaces.
> > >
> > > A workaround could be to limit use of first 32 channels instead of crashing
> > > with assert on startup. Happy to work on a patch if you see this as a
> > > potential path.
> >
> > What's the assertion? At the very least the crash should be fixed, even
> > if it doesn't make the sound card work.
>
> I face the assertion in my current work for ALSA driver for MOTU
> FireWire units. One of my test device supports 34ch capture. When
> logging in desktop environment which starts pulseaudio process, I can
> see its abort due to below assertion.
>
> pulseaudio[18757]: [pulseaudio] channelmap.c: Assertion
> 'pa_channels_valid(channels)' failed at pulse/channelmap.c:401, function
> pa_cannel_map_init_extend(). Aborting.
>
> Below is the stacktrace.
>
> == Stacktrace =================================
> #0 __GI_raise (sig=sig at entry=6) at ../sysdeps/unix/sysv/linux/raise.c:58
> set = {__val = {0, 0, 0, 0, 0, 0, 4672924419136, 146028888064,
> 34, 189408057797700, 105961138159616, 24671, 4672924419136, 0, 0, 0}}
> pid = <optimized out>
> tid = <optimized out>
> #1 0x00007f50383433ea in __GI_abort () at abort.c:89
> save_stage = 2
> act = {__sigaction_handler = {sa_handler = 0x3000000028,
> sa_sigaction = 0x3000000028}, sa_mask = {__val = {0, 140732661601104,
> 4035625362880256, 140732661601088, 0, 139982539889656, 94743048657840,
> 139982546343824, 0, 0, 140732661601104, 0, 139982517379468,
> 189408057753607, 139981574111266, 140732661600784}}, sa_flags =
> 965515017, sa_restorer = 0x7f503990903e}
> sigs = {__val = {32, 0 <repeats 15 times>}}
> #2 0x00007f50398c9712 in pa_channel_map_init_extend () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-9.0.so
> No symbol table info available.
> #3 0x00007f502ac6f058 in pa_alsa_open_by_device_string () from
> /usr/lib/pulse-9.0/modules/libalsa-util.so
> No symbol table info available.
> #4 0x00007f502ac6f12c in pa_alsa_open_by_template () from
> /usr/lib/pulse-9.0/modules/libalsa-util.so
> No symbol table info available.
> #5 0x00007f502ac78ae1 in ?? () from
> /usr/lib/pulse-9.0/modules/libalsa-util.so
> No symbol table info available.
> #6 0x00007f502ac82161 in pa_alsa_profile_set_probe () from
> /usr/lib/pulse-9.0/modules/libalsa-util.so
> No symbol table info available.
> #7 0x00007f502b1b00ee in module_alsa_card_LTX_pa__init () from
> /usr/lib/pulse-9.0/modules/module-alsa-card.so
> No symbol table info available.
> #8 0x00007f5039b593b8 in pa_module_load () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecore-9.0.so
> No symbol table info available.
> #9 0x00007f502b3b6703 in ?? () from
> /usr/lib/pulse-9.0/modules/module-udev-detect.so
> No symbol table info available.
> #10 0x00007f502b3b7a0b in module_udev_detect_LTX_pa__init () from
> /usr/lib/pulse-9.0/modules/module-udev-detect.so
> No symbol table info available.
> #11 0x00007f5039b593b8 in pa_module_load () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecore-9.0.so
> No symbol table info available.
> #12 0x00007f5039b48381 in ?? () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecore-9.0.so
> No symbol table info available.
> #13 0x00007f5039b4e71c in pa_cli_command_execute_line_stateful () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecore-9.0.so
> No symbol table info available.
> #14 0x00007f5039b4ef31 in pa_cli_command_execute_file_stream () from
> /usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecore-9.0.so
> No symbol table info available.
> #15 0x0000562b1430c242 in main ()
> No symbol table info available.
>
> We can see SIGABRT is generated in below line:
> https://cgit.freedesktop.org/pulseaudio/pulseaudio/tree/src/pulse/channelmap.c#n401
Thanks for the stacktrace! I submitted a potential fix, I hope you can
test it and confirm that PulseAudio doesn't crash any more.
--
Tanu
https://www.patreon.com/tanuk
