Re: [PATCH] skip increamenting nr for TASK_UNINTERRUPTIBLE

[Oleg Nesterov <oleg@redhat.com>]

On 12/21, Linus Torvalds wrote:
>
> Again, adding Oleg to the cc. And I don't think this is correct,

me too, but I can't reply right now, will do tomorrow,


> > In coredump case, where thread_1 faults while thread_2 is in
> > TASK_UNINTERRUPTIBLE state, it cannot handle the SIGKILL.

Yes, and we have to wait. We can not simply ignore its state.
Not to mention, every another_task->state check is racy.

> > Thus the process hangs on event.
> > The coredump routine freezes until the thread state is
> > uninterruptible.
> >
> > Solution: Continue for coredump, without waiting for uninterruptible
> >  thread, as it will get killed as soon as it returns from
> >  uninterruptible state.
> >  Therefore do not increament thread count for threads with
> >  TASK_UNINTERRUPTIBLE.
> >
> > Signed-off-by: Ajeet Yadav <ajeet.yadav.77@gmail.com>
> > Signed-off-by: Vaibhav Shinde <v.bhav.shinde@gmail.com>
> > ---
> >  fs/coredump.c |    3 ++-
> >  1 file changed, 2 insertions(+), 1 deletions(-)
> >
> > diff --git a/fs/coredump.c b/fs/coredump.c
> > index 447b02c..54b0664 100644
> > --- a/fs/coredump.c
> > +++ b/fs/coredump.c
> > @@ -281,7 +281,8 @@ static int zap_process(struct task_struct *start,
> > int exit_code)
> >   if (t != current && t->mm) {
> >   sigaddset(&t->pending.signal, SIGKILL);
> >   signal_wake_up(t, 1);
> > - nr++;
> > + if(!(t->state & TASK_UNINTERRUPTIBLE))
> > + nr++;
> >   }
> >   } while_each_thread(start, t);
> >
> > --
> > 1.7.9.5