Re: [PATCH] posix-timer: fix deletion race

[Jeremy Katz]

On Tue, 24 Jul 2007, Oleg Nesterov wrote:

> On 07/23, Jeremy Katz wrote:
> > On Fri, 20 Jul 2007, Oleg Nesterov wrote:
> >
> > > I still can't believe we have a double-free problem, this looks imposiible.
> > > Do you see the
> > >
> > > 	"idr_remove called for id=%d which is not allocated.\n"
> > >
> > > in syslog?
> >
> > No.  I also added some accounting with atomic counters, and don't see
> > evidence of a second call to release_posix_timer.
>
> Interesting. Could you show the patch? Where does sys_timer_create() set
> counter == 1?

--- kernel/posix-timers.c.old      2007-07-24 11:21:29.000000000 -0700
+++ kernel/posix-timers.c       2007-07-20 15:49:51.000000000 -0700
@@ -409,13 +409,17 @@ static struct k_itimer * alloc_posix_tim
 #define IT_ID_NOT_SET  0
 static void release_posix_timer(struct k_itimer *tmr, int it_id_set)
 {
+       atomic_inc(&tmr->freecount);
@@ -443,6 +447,7 @@ sys_timer_create(const clockid_t which_c
        new_timer = alloc_posix_timer();
        if (unlikely(!new_timer))
                return -EAGAIN;
+       atomic_set(&new_timer->freecount, 0);

--- include/linux/posix-timers.h.old  2007-07-24 11:32:36.000000000 -0700
+++ include/linux/posix-timers.h      2007-07-20 15:15:44.000000000 -0700
@@ -62,6 +62,7 @@ struct k_itimer {
                        unsigned long expires;
                } mmtimer;
        } it;
+       atomic_t freecount;
 };

 struct k_clock {


Examination was done through a kgdb session when it caught the bug. 
freecount always had the expected value, 1 during a delete operation, 
and 0 otherwise.

> > > Could you try the patch below? Perhaps we have some wierd problem with
> > > ->sigq corruption.
> >
> > Tried, with apparent effect.
>
> You mean, "null pointer dereference" in release_posix_timer() ? Or it was
> BUG_ON(timr && !timr->sigq) in lock_timer() ?

Sorry.  That should have been "without apparent effect".

> I assume you didn't apply any other patches except Thomas's, yes?

Alone, as well as with Thomas' patch. Neither way changed the behavior.

Jeremy
-
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/