On Tue, Jun 13, 2017 at 12:01:17AM +0200, Thomas Gleixner wrote:
> On Mon, 12 Jun 2017, Andrei Vagin wrote:
> > > diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c
> > > index b53a0b5..88517dc 100644
> > > --- a/kernel/time/posix-timers.c
> > > +++ b/kernel/time/posix-timers.c
> > > @@ -828,6 +828,8 @@ SYSCALL_DEFINE4(timer_settime, timer_t, timer_id,
> > > int, flags,
> > > if (!timespec64_valid(&new_spec64.it_interval) ||
> > > !timespec64_valid(&new_spec64.it_value))
> > > return -EINVAL;
> > > + if (rtn)
> > > + memset(rtn, 0, sizeof(*rtn));
> >
> > Maybe we need to call memset after "retry:"?
>
> That would be counter productive.
>
> > common_timer_get() is called at the begining of common_timer_set(), then
> > common_timer_set() can return TIMER_RETRY. common_timer_get() will be
> > called again and some fields of rtn which have been touched first time
> > will not be touched.
> >
> > At the end, rtn will contain data from two executions of
> > common_timer_get().
>
> No. See the full code sequence:
>
> retry:
> timr = lock_timer(timer_id, &flag);
> if (!timr)
> return -EINVAL;
>
> kc = clockid_to_kclock(timr->it_clock);
> if (WARN_ON_ONCE(!kc || !kc->timer_set))
> error = -EINVAL;
> else
> error = kc->timer_set(timr, flags, &new_spec64, rtn);
>
> unlock_timer(timr, flag);
> if (error == TIMER_RETRY) {
> rtn = NULL; // We already got the old time...
> goto retry;
> }
>
> If you clear it after retry, you'll get all zeros in the retry case. Not
> what you really want.
Yes, you are right. Sorry for the noise.
>
> Thanks,
>
> tglx
