ALSA: usb-audio: Fix races at disconnection and PCM closing
authorTakashi Iwai <tiwai@suse.de>
Wed, 25 Jun 2014 12:24:47 +0000 (14:24 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 7 Jul 2014 01:57:29 +0000 (18:57 -0700)
commit9ef36a04f57e47715cdad2649daf2fe31d2e50c9
tree30fac39502c6316ea452f80bc729fb605dc717db
parent6a2e92a4a660b38fcf15a134ffc98e6b8526e577
ALSA: usb-audio: Fix races at disconnection and PCM closing

commit 92a586bdc06de6629dae1b357dac221253f55ff8 upstream.

When a USB-audio device is disconnected while PCM is still running, we
still see some race: the disconnect callback calls
snd_usb_endpoint_free() that calls release_urbs() and then kfree()
while a PCM stream would be closed at the same time and calls
stop_endpoints() that leads to wait_clear_urbs().  That is, the EP
object might be deallocated while a PCM stream is syncing with
wait_clear_urbs() with the same EP.

Basically calling multiple wait_clear_urbs() would work fine, also
calling wait_clear_urbs() and release_urbs() would work, too, as
wait_clear_urbs() just reads some fields in ep.  The problem is the
succeeding kfree() in snd_pcm_endpoint_free().

This patch moves out the EP deallocation into the later point, the
destructor callback.  At this stage, all PCMs must have been already
closed, so it's safe to free the objects.

Reported-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
sound/usb/card.c
sound/usb/endpoint.c
sound/usb/endpoint.h