Bug 80867

Summary:	Pulseaudio dies when using mmkbd-evdev and the input device is unplugged
Product:	PulseAudio	Reporter:	Stelios Bounanos <sb-fdbugs>
Component:	modules	Assignee:	pulseaudio-bugs
Status:	RESOLVED FIXED	QA Contact:	pulseaudio-bugs
Severity:	normal
Priority:	medium	CC:	lennart
Version:	unspecified
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:
Bug Depends on:
Bug Blocks:	75721

Description Stelios Bounanos 2014-07-03 18:34:44 UTC

With the mmkbd-evdev module enabled e.g. like so in ~/.pulse/default:

  load-module module-mmkbd-evdev device=/dev/input/by-id/usb-0430_Sun_USB_Keyboard-event-kbd

pulseaudio crashes when that keyboard is unplugged.  The following messages
are logged to syslog:

  Jul 3 19:21:17 hostname pulseaudio[11515]: [pulseaudio] module-mmkbd-evdev.c:
  Lost connection to evdev device.
  Jul 3 19:21:17 hostname pulseaudio[11515]: [pulseaudio] module-mmkbd-evdev.c:
  Assertion 'pa_close(u->fd) == 0' failed at modules/module-mmkbd-evdev.c:260,
  function module_mmkbd_evdev_LTX_pa__done(). Aborting.

I would guess that the fd has already been closed and close(2) returns EBADF
to pa_close().

Comment 1 Stelios Bounanos 2014-07-03 19:05:16 UTC

I should have mentioned that this is pulseaudio 5.0 on an up to date
Debian/unstable system running kernel 3.14.

Comment 2 Tanu Kaskinen 2014-07-07 12:14:02 UTC

6.0 blocker. Shouldn't be hard to fix...

Comment 3 David Henningsson 2014-09-02 09:26:32 UTC

(In reply to comment #2)
> 6.0 blocker. Shouldn't be hard to fix...

I had a look but didn't find anything obvious. My guess is that the kernel reports something strange on close, such as -ENODEV or something, instead of just closing it. Stelios, can you check what error we get on pa_close?

Comment 4 Stelios Bounanos 2014-09-11 14:18:48 UTC

(In reply to comment #3)
> (In reply to comment #2)
> > 6.0 blocker. Shouldn't be hard to fix...
> 
> I had a look but didn't find anything obvious. My guess is that the kernel
> reports something strange on close, such as -ENODEV or something, instead of
> just closing it. Stelios, can you check what error we get on pa_close?

Yes, indeed, after pa_close returns -1, errno is set to ENODEV.

Comment 5 David Henningsson 2014-09-12 08:11:35 UTC

Thanks, patch posted here:

http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-September/021617.html

Comment 6 David Henningsson 2014-10-03 12:28:45 UTC

I pushed the patch now.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.