Skip to contentSkip to conversation

sched_exit: move the callsite to do_exit()

PATCH New Oleg Nesterov 1 participant 1 message Jun 22, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 2/2 sched_exit: move the callsite to do_exit()

sched_exit() is called by release_task(). If the task auto-reaps itself this call happens a bit too early (the task still uses cpu after that). If the task goes to TASK_ZOMBIE this call is unpredictably delayed.

I think it is better to do sched_exit() right before the last schedule().

We can use read_lock_rcu() instead of write_lock(tasklist). In that case it is possible that sched_exit() changes ->time_slice/->sleep_avg of the already dead ->parent, but I think this is tolerable.

kernel/sched.c renamed from kernel/sched.c~MOVE +5 -1
@@ -1479,10 +1479,13 @@ void fastcall wake_up_new_task(task_t *p
  */
 void fastcall sched_exit(task_t *p)
 {
-	task_t *parent = p->parent;
+	task_t *parent;
 	unsigned long flags;
 	runqueue_t *rq;
 
+	rcu_read_lock();
+	parent = p->real_parent;
+
 	/*
 	 * If the child was a (relative-) CPU hog then decrease
 	 * the sleep_avg of the parent as well.
@@ -1498,6 +1501,7 @@ void fastcall sched_exit(task_t *p)
 		(EXIT_WEIGHT + 1) * EXIT_WEIGHT + p->sleep_avg /
 		(EXIT_WEIGHT + 1);
 	task_rq_unlock(rq, &flags);
+	rcu_read_unlock();
 }
 
 /**
kernel/exit.c renamed from kernel/exit.c~MOVE +2 -1
@@ -170,7 +170,6 @@ repeat:
 		zap_leader = (leader->exit_signal == -1);
 	}
 
-	sched_exit(p);
 	write_unlock_irq(&tasklist_lock);
 	spin_unlock(&p->proc_lock);
 	proc_pid_flush(proc_dentry);
@@ -940,6 +939,8 @@ fastcall NORET_TYPE void do_exit(long co
 	if (tsk->splice_pipe)
 		__free_pipe_info(tsk->splice_pipe);
 
+	sched_exit(tsk);
+
 	/* PF_DEAD causes final put_task_struct after we schedule. */
 	preempt_disable();
 	BUG_ON(tsk->flags & PF_DEAD);