linux-kernel - Re: Sound+USB: deadlock problem

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <s5hy5i5iz97.wl%tiwai@suse.de>
Date:	Tue, 13 Nov 2012 11:16:36 +0100
From:	Takashi Iwai <tiwai@...e.de>
To:	Jiri Slaby <jslaby@...e.cz>
Cc:	Jaroslav Kysela <perex@...ex.cz>, alsa-devel@...a-project.org,
	LKML <linux-kernel@...r.kernel.org>,
	USB list <linux-usb@...r.kernel.org>,
	Jiri Slaby <jirislaby@...il.com>
Subject: Re: Sound+USB: deadlock problem

At Tue, 13 Nov 2012 10:39:14 +0100,
Jiri Slaby wrote:
> 
> On 11/13/2012 09:45 AM, Takashi Iwai wrote:
> > At Tue, 13 Nov 2012 09:30:10 +0100,
> > Jiri Slaby wrote:
> >>
> >> On 11/13/2012 09:28 AM, Takashi Iwai wrote:
> >>> At Tue, 13 Nov 2012 09:26:17 +0100,
> >>> Jiri Slaby wrote:
> >>>>
> >>>> Hello,
> >>>>
> >>>> I've just plugged+unplugged+plugged my USB audio card and the audio
> >>>> subsystem got stuck:
> >>>
> >>> Which kernel?  If it's older than 3.7-rc5, could you try 3.7-rc5?
> >>
> >> Oh, forgot to mention:
> >> 3.7.0-rc5-next-20121112_64+
> > 
> > Through a quick glance, one of mutex locks seems stuck.
> > Just a blind short -- does the patch below cure the problem?
> 
> Yes, this helps.
> 
> Adding a LOCKDEP splash _before_ applying the patch:

Thanks, this helps much for understanding the cause.
It's a new rwsem in the usb-audio driver that deadlocks.

Although the previous patch fixes the deadlock it opens the race
again. 

Could you try the patch below instead?  The rwsem above doesn't have
to be taken for the whole operation, but just needs to protect for
turning on the flag.  The patch below should suffice for protecting
races but avoids the deadlock.


Takashi

---
diff --git a/sound/usb/card.c b/sound/usb/card.c
index 282f0fc..dbf7999 100644
--- a/sound/usb/card.c
+++ b/sound/usb/card.c
@@ -559,9 +559,11 @@ static void snd_usb_audio_disconnect(struct usb_device *dev,
 		return;
 
 	card = chip->card;
-	mutex_lock(&register_mutex);
 	down_write(&chip->shutdown_rwsem);
 	chip->shutdown = 1;
+	up_write(&chip->shutdown_rwsem);
+
+	mutex_lock(&register_mutex);
 	chip->num_interfaces--;
 	if (chip->num_interfaces <= 0) {
 		snd_card_disconnect(card);
@@ -582,11 +584,9 @@ static void snd_usb_audio_disconnect(struct usb_device *dev,
 			snd_usb_mixer_disconnect(p);
 		}
 		usb_chip[chip->index] = NULL;
-		up_write(&chip->shutdown_rwsem);
 		mutex_unlock(&register_mutex);
 		snd_card_free_when_closed(card);
 	} else {
-		up_write(&chip->shutdown_rwsem);
 		mutex_unlock(&register_mutex);
 	}
 }
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/