[PATCH 4/3] thread_group_cputime: move a couple of callsites outside of ->siglock

[PATCH 4/3] thread_group_cputime: move a couple of callsites outside of ->siglock

[Oleg Nesterov]

->siglock buys nothing for thread_group_cputime() in do_sys_times() and
wait_task_zombie() (which btw takes the unrelated parent's ->siglock).
Actually I think do_sys_times() doesn't need ->siglock at all.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

--- K-28/kernel/exit.c~SYS_TIMES_NO_SIGLOCK	2008-11-10 15:07:22.000000000 +0100
+++ K-28/kernel/exit.c	2008-11-17 02:02:12.000000000 +0100
@@ -1330,10 +1330,10 @@ static int wait_task_zombie(struct task_
 		 * group, which consolidates times for all threads in the
 		 * group including the group leader.
 		 */
+		thread_group_cputime(p, &cputime);
 		spin_lock_irq(&p->parent->sighand->siglock);
 		psig = p->parent->signal;
 		sig = p->signal;
-		thread_group_cputime(p, &cputime);
 		psig->cutime =
 			cputime_add(psig->cutime,
 			cputime_add(cputime.utime,
--- K-28/kernel/sys.c~SYS_TIMES_NO_SIGLOCK	2008-11-06 19:11:02.000000000 +0100
+++ K-28/kernel/sys.c	2008-11-17 01:56:17.000000000 +0100
@@ -858,8 +858,8 @@ void do_sys_times(struct tms *tms)
 	struct task_cputime cputime;
 	cputime_t cutime, cstime;
 
-	spin_lock_irq(&current->sighand->siglock);
 	thread_group_cputime(current, &cputime);
+	spin_lock_irq(&current->sighand->siglock);
 	cutime = current->signal->cutime;
 	cstime = current->signal->cstime;
 	spin_unlock_irq(&current->sighand->siglock);

Re: [PATCH 4/3] thread_group_cputime: move a couple of callsites outside of ->siglock

[Ingo Molnar]

* Oleg Nesterov <oleg@redhat.com> wrote:

> ->siglock buys nothing for thread_group_cputime() in do_sys_times() and
> wait_task_zombie() (which btw takes the unrelated parent's ->siglock).
> Actually I think do_sys_times() doesn't need ->siglock at all.
> 
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>

applied your changes to tip/core/signal:

2b5fe6d: thread_group_cputime: move a couple of callsites outside of ->siglock
ce39447: thread_group_cputime: kill the bogus ->signal != NULL check
74fcd52: account_steal_time: kill the unneeded account_group_system_time()

(merged the sched/urgent fixlet before them)

thanks Oleg!

	Ingo