Re: [PATCH] hrtimer: make epoll_wait() use the hrtimer range feature

[Shawn Bohrer]

On Wed, Nov 24, 2010 at 03:33:02AM -0500, Mike Frysinger wrote:
> On Sun, Aug 8, 2010 at 18:45, Shawn Bohrer wrote:
> > @@ -1116,18 +1113,22 @@ static int ep_send_events(struct eventpoll *ep,
> > Âstatic int ep_poll(struct eventpoll *ep, struct epoll_event __user *events,
> > Â Â Â Â Â Â Â Â Â int maxevents, long timeout)
> > Â{
> > - Â Â Â int res, eavail;
> > + Â Â Â int res, eavail, timed_out = 0;
> > Â Â Â Âunsigned long flags;
> > - Â Â Â long jtimeout;
> > + Â Â Â long slack;
> > Â Â Â Âwait_queue_t wait;
> > -
> > - Â Â Â /*
> > - Â Â Â Â* Calculate the timeout by checking for the "infinite" value (-1)
> > - Â Â Â Â* and the overflow condition. The passed timeout is in milliseconds,
> > - Â Â Â Â* that why (t * HZ) / 1000.
> > - Â Â Â Â*/
> > - Â Â Â jtimeout = (timeout < 0 || timeout >= EP_MAX_MSTIMEO) ?
> > - Â Â Â Â Â Â Â MAX_SCHEDULE_TIMEOUT : (timeout * HZ + 999) / 1000;
> > + Â Â Â struct timespec end_time;
> > + Â Â Â ktime_t expires, *to = NULL;
> > +
> > + Â Â Â if (timeout > 0) {
> > + Â Â Â Â Â Â Â ktime_get_ts(&end_time);
> > + Â Â Â Â Â Â Â timespec_add_ns(&end_time, (u64)timeout * NSEC_PER_MSEC);
> > + Â Â Â Â Â Â Â slack = estimate_accuracy(&end_time);
> > + Â Â Â Â Â Â Â to = &expires;
> > + Â Â Â Â Â Â Â *to = timespec_to_ktime(end_time);
> > + Â Â Â } else if (timeout == 0) {
> > + Â Â Â Â Â Â Â timed_out = 1;
> > + Â Â Â }
> >
> > Âretry:
> > Â Â Â Âspin_lock_irqsave(&ep->lock, flags);
> > @@ -1149,7 +1150,7 @@ retry:
> > Â Â Â Â Â Â Â Â Â Â Â Â * to TASK_INTERRUPTIBLE before doing the checks.
> > Â Â Â Â Â Â Â Â Â Â Â Â */
> > Â Â Â Â Â Â Â Â Â Â Â Âset_current_state(TASK_INTERRUPTIBLE);
> > - Â Â Â Â Â Â Â Â Â Â Â if (!list_empty(&ep->rdllist) || !jtimeout)
> > + Â Â Â Â Â Â Â Â Â Â Â if (!list_empty(&ep->rdllist) || timed_out)
> > Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Âbreak;
> > Â Â Â Â Â Â Â Â Â Â Â Âif (signal_pending(current)) {
> > Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Âres = -EINTR;
> > @@ -1157,7 +1158,9 @@ retry:
> > Â Â Â Â Â Â Â Â Â Â Â Â}
> >
> > Â Â Â Â Â Â Â Â Â Â Â Âspin_unlock_irqrestore(&ep->lock, flags);
> > - Â Â Â Â Â Â Â Â Â Â Â jtimeout = schedule_timeout(jtimeout);
> > + Â Â Â Â Â Â Â Â Â Â Â if (!schedule_hrtimeout_range(to, slack, HRTIMER_MODE_ABS))
> > + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â timed_out = 1;
> > +
> > Â Â Â Â Â Â Â Â Â Â Â Âspin_lock_irqsave(&ep->lock, flags);
> > Â Â Â Â Â Â Â Â}
> > Â Â Â Â Â Â Â Â__remove_wait_queue(&ep->wq, &wait);
> 
> this code introduces a warning:
> fs/eventpoll.c: In function âep_pollâ:
> fs/eventpoll.c:1119: warning: âslackâ may be used uninitialized in this function
> 
> looks to me like you arent properly handling negative timeouts.
> certainly epoll_wait() passes the timeout value straight from
> userspace to ep_poll() without any error checking, so if userspace
> passes a negative timeout value, it looks like "slack" will be used
> uninitialized.

If a negative timeout is passed in then 'to' remains NULL.  When 'to
is NULL schedule_hrtimeout_range() has an infinite timeout and the
'slack' parameter is never used.  So technically everything should be
fine here.

Of course it would be safest and best to simply initialize slack to 0.
I can send a patch this evening with the fix.

--
Shawn
--
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/