Re: [PATCH, RFC 07/30] [media] radio-cadet: avoid interruptible_sleep_on race

[Arnd Bergmann]

On Friday 17 January 2014, Hans Verkuil wrote:
> > @@ -323,25 +324,32 @@ static ssize_t cadet_read(struct file *file, char __user *data, size_t count, lo
> >       struct cadet *dev = video_drvdata(file);
> >       unsigned char readbuf[RDS_BUFFER];
> >       int i = 0;
> > +     DEFINE_WAIT(wait);
> >  
> >       mutex_lock(&dev->lock);
> >       if (dev->rdsstat == 0)
> >               cadet_start_rds(dev);
> > -     if (dev->rdsin == dev->rdsout) {
> > +     while (1) {
> > +             prepare_to_wait(&dev->read_queue, &wait, TASK_INTERRUPTIBLE);
> > +             if (dev->rdsin != dev->rdsout)
> > +                     break;
> > +
> >               if (file->f_flags & O_NONBLOCK) {
> >                       i = -EWOULDBLOCK;
> >                       goto unlock;
> >               }
> >               mutex_unlock(&dev->lock);
> > -             interruptible_sleep_on(&dev->read_queue);
> > +             schedule();
> >               mutex_lock(&dev->lock);
> >       }
> > +
> 
> This seems overly complicated. Isn't it enough to replace interruptible_sleep_on
> by 'wait_event_interruptible(&dev->read_queue, dev->rdsin != dev->rdsout);'?
> 
> Or am I missing something subtle?

The existing code sleeps with &dev->lock released because the cadet_handler()
function needs to grab (and release) the same lock before it can wake up
the reader thread.

Doing the simple wait_event_interruptible() would result in a deadlock here.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/