exec: RT sub-thread can livelock and monopolize CPU on exec

de_thread() yields waiting for ->group_leader to be a zombie. This deadlocks
if an rt-prio execer shares the same cpu with ->group_leader. Change the code
to use ->group_exit_task/notify_count mechanics.

This patch certainly uglifies the code, perhaps someone can suggest something
better.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 7 insertions, 3 deletions

diff --git a/kernel/exit.c b/kernel/exit.c
index 25f6805be5fe..4c108df88a37 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -92,10 +92,9 @@ static void __exit_signal(struct task_struct *tsk)
                  * If there is any task waiting for the group exit
                  * then notify it:
                  */
-                if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
+                if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count)
                         wake_up_process(sig->group_exit_task);
-                        sig->group_exit_task = NULL;
-                }
+
                 if (tsk == sig->curr_target)
                         sig->curr_target = next_thread(tsk);
                 /*
@@ -827,6 +826,11 @@ static void exit_notify(struct task_struct *tsk)
                 state = EXIT_DEAD;
         tsk->exit_state = state;
 
+        if (thread_group_leader(tsk) &&
+            tsk->signal->notify_count < 0 &&
+            tsk->signal->group_exit_task)
+                wake_up_process(tsk->signal->group_exit_task);
+
         write_unlock_irq(&tasklist_lock);
 
         list_for_each_safe(_p, _n, &ptrace_dead) {