2008-03-05 21:54:42

by Oleg Nesterov

Subject: [PATCH -mm 2/2] send_sigqueue: don't forget about handle_stop_signal()

(on top of Pavel's signals-consolidate-send_sigqueue-and-send_group_sigqueue.patch)

send_group_sigqueue() calls handle_stop_signal(), send_sigqueue() doesn't.
This is not consistent and in fact I'd say this is (minor) bug.

Move handle_stop_signal() from send_group_sigqueue() to do_send_sigqueue(),
the latter is called by send_sigqueue() too.

Signed-off-by: Oleg Nesterov <[email protected]>

--- 25/kernel/signal.c~2_SS_HSS 2008-03-06 00:35:35.000000000 +0300
+++ 25/kernel/signal.c 2008-03-06 00:43:50.000000000 +0300
@@ -1274,8 +1274,10 @@ void sigqueue_free(struct sigqueue *q)
}

static int do_send_sigqueue(int sig, struct sigqueue *q, struct task_struct *t,
- struct sigpending *pending)
+ struct sigpending *pending)
{
+ handle_stop_signal(sig, t);
+
if (unlikely(!list_empty(&q->list))) {
/*
* If an SI_TIMER entry is already queue just increment
@@ -1335,7 +1337,6 @@ send_group_sigqueue(int sig, struct sigq
read_lock(&tasklist_lock);
/* Since it_lock is held, p->sighand cannot be NULL. */
spin_lock_irqsave(&p->sighand->siglock, flags);
- handle_stop_signal(sig, p);

ret = do_send_sigqueue(sig, q, p, &p->signal->shared_pending);

2008-03-05 23:53:25

by Oleg Nesterov

Subject: Re: [PATCH -mm 2/2] send_sigqueue: don't forget about handle_stop_signal()

On 03/06, Oleg Nesterov wrote:
>
> (on top of Pavel's signals-consolidate-send_sigqueue-and-send_group_sigqueue.patch)
>
> send_group_sigqueue() calls handle_stop_signal(), send_sigqueue() doesn't.
> This is not consistent and in fact I'd say this is (minor) bug.
>
> Move handle_stop_signal() from send_group_sigqueue() to do_send_sigqueue(),
> the latter is called by send_sigqueue() too.

Sorry for the noise, but I forgot to mention that this patch of course depends
on signals-re-assign-cld_continued-notification-from-the-sender-to-reciever.patch

Oleg.

2008-03-06 11:06:25

by Roland McGrath

Subject: Re: [PATCH -mm 2/2] send_sigqueue: don't forget about handle_stop_signal()

> send_group_sigqueue() calls handle_stop_signal(), send_sigqueue() doesn't.
> This is not consistent and in fact I'd say this is (minor) bug.

Yes, it was wrong. I think I noticed it long ago and never bothered with
it because send_sigqueue is nearly unused. SIGEV_THREAD_ID is in practice
only used internally by libc (timer_create) to implement SIGEV_THREAD, and
it never uses any of the signals handle_stop_signal checks for.

Your fix is good and should go in.
I'd bet noone will ever notice the difference now or ever.


Thanks,
Roland