From: Oleg Nesterov on
On 05/14, Stanislaw Gruszka wrote:
>
> On Fri, 14 May 2010 18:03:57 +0200
> Oleg Nesterov <oleg(a)redhat.com> wrote:
>
> > On 05/14, Andrey Vagin wrote:
> > >
> > > @@ -613,6 +613,8 @@ SYSCALL_DEFINE3(timer_create, const clockid_t, which_clock,
> > > * and may cease to exist at any time. Don't use or modify
> > > * new_timer after the unlock call.
> > > */
> > > +out_cleanup:
> > > + CLOCK_DISPATCH(new_timer->it_clock, timer_cleanup, (new_timer));
> >
> > But at first glance you are right, posix_cpu_timer_create() does
> > get_task_struct(it.cpu.task).
>
> If I understand problem correctly, seems to be fine to move
> CLOCK_DISPATCH(which_clock, timer_create, (new_timer));
> after all possible EFAULT errors and solve leak without creating
> new timer_cleanup() callback.

I thought about this too, we are doing copy_to_user(created_timer_id)
"in advance" anyway. Probably we can move all this code block

new_timer->it_id = (timer_t) new_timer_id;
new_timer->it_clock = which_clock;
new_timer->it_overrun = -1;
error = CLOCK_DISPATCH(which_clock, timer_create, (new_timer));
if (error)
goto out;

down, right before we take ->siglock.

But I don't understand the change in posix_cpu_timer_del() from 1/2.


Otoh, currently "The next step is hard to back out if there is an error"
comment is not right, release_posix_timer() does put_pid(). We can
move copy_to_user(created_timer_id) down after "if (timer_event_spec)"
block too. (but before CLOCK_DISPATCH(), of course).

Andrey, what do you think?

Oleg.

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