[alsa-devel] BUG: soft lockup in snd_virmidi_output_trigger

[Takashi Iwai]

On Fri, 27 Jul 2018 07:51:04 +0200,
Dae R. Jeong wrote:
> 
> On Fri, Jul 27, 2018 at 07:20:30AM +0200, Takashi Iwai wrote:
> 
> I think there is a mistake on the reported-by tag.
> The reported-by tag (e4c8abb920ef...) is for WARNING in port_delete.
> I guess the tag below is correct.
> Reported-by: syzbot+619d9f40141d826b097e at syzkaller.appspotmail.com
> Could you please make sure that the reported-by tag is correct?

Indeed, corrected now.  Thanks.

> By the way, I can't reproduce the crash WARNING in port_delete so
> far. I feel sad for saying this..
> I really want to reproduce the crash and thus, to help to fix the
> crash. If there is any progress on it, I will let you know
> immediately.

This looks like a sheer race condition.
Thanks for your help in anyway!


Takashi

> 
> Best regards,
> Dae R. Jeong
> 
> > On Fri, 27 Jul 2018 06:13:22 +0200,
> > Dae R. Jeong wrote:
> > > 
> > > I tested it and it worked.
> > > Thanks a lot!
> > 
> > Good to hear.  Below is the final patch with a proper comment (and
> > with syzbot reported-by, too) I'm going to queue to sound.git tree.
> > 
> > 
> > thanks,
> > 
> > Takashi
> > 
> > -- 8< --
> > 
> > From: Takashi Iwai <tiwai at suse.de>
> > Subject: [PATCH v2] ALSA: virmidi: Fix too long output trigger loop
> > 
> > The virmidi output trigger tries to parse the all available bytes and
> > process sequencer events as much as possible.  In a normal situation,
> > this is supposed to be relatively short, but a program may give a huge
> > buffer and it'll take a long time in a single spin lock, which may
> > eventually lead to a soft lockup.
> > 
> > This patch simply adds a workaround, a cond_resched() call in the loop
> > if applicable.  A better solution would be to move the event processor
> > into a work, but let's put a duct-tape quickly at first.
> > 
> > Reported-and-tested-by: Dae R. Jeong <threeearcat at gmail.com>
> > Reported-by: syzbot+e4c8abb920efa77bace9 at syzkaller.appspotmail.com
> > Cc: <stable at vger.kernel.org>
> > Signed-off-by: Takashi Iwai <tiwai at suse.de>
> > ---
> >  sound/core/seq/seq_virmidi.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> > 
> > diff --git a/sound/core/seq/seq_virmidi.c b/sound/core/seq/seq_virmidi.c
> > index 289ae6bb81d9..8ebbca554e99 100644
> > --- a/sound/core/seq/seq_virmidi.c
> > +++ b/sound/core/seq/seq_virmidi.c
> > @@ -163,6 +163,7 @@ static void snd_virmidi_output_trigger(struct snd_rawmidi_substream *substream,
> >  	int count, res;
> >  	unsigned char buf[32], *pbuf;
> >  	unsigned long flags;
> > +	bool check_resched = !in_atomic();
> >  
> >  	if (up) {
> >  		vmidi->trigger = 1;
> > @@ -200,6 +201,15 @@ static void snd_virmidi_output_trigger(struct snd_rawmidi_substream *substream,
> >  					vmidi->event.type = SNDRV_SEQ_EVENT_NONE;
> >  				}
> >  			}
> > +			if (!check_resched)
> > +				continue;
> > +			/* do temporary unlock & cond_resched() for avoiding
> > +			 * CPU soft lockup, which may happen via a write from
> > +			 * a huge rawmidi buffer
> > +			 */
> > +			spin_unlock_irqrestore(&substream->runtime->lock, flags);
> > +			cond_resched();
> > +			spin_lock_irqsave(&substream->runtime->lock, flags);
> >  		}
> >  	out:
> >  		spin_unlock_irqrestore(&substream->runtime->lock, flags);
> > -- 
> > 2.18.0
> > 
>