* For preemptible RCU it is sufficient to call rcu_read_unlock in order
* to exit the grace period. For classic RCU, a reschedule is required.
*/
-static void rcu_lock_break(struct task_struct *g, struct task_struct *t)
+static bool rcu_lock_break(struct task_struct *g, struct task_struct *t)
{
+ bool can_cont;
+
get_task_struct(g);
get_task_struct(t);
rcu_read_unlock();
cond_resched();
rcu_read_lock();
+ can_cont = pid_alive(g) && pid_alive(t);
put_task_struct(t);
put_task_struct(g);
+
+ return can_cont;
}
/*
goto unlock;
if (!--batch_count) {
batch_count = HUNG_TASK_BATCHING;
- rcu_lock_break(g, t);
- /* Exit if t or g was unhashed during refresh. */
- if (t->state == TASK_DEAD || g->state == TASK_DEAD)
+ if (!rcu_lock_break(g, t))
goto unlock;
}
/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */