rcu/tree: Call setschedule() gp ktread to SCHED_FIFO outside of atomic region

sched_setscheduler() needs to acquire cpuset_rwsem, but it is currently
called from an invalid (atomic) context by rcu_spawn_gp_kthread().

Fix that by simply moving sched_setscheduler_nocheck() call outside of
the atomic region, as it doesn't actually require to be guarded by
rcu_node lock.

Suggested-by: Peter Zijlstra <[email protected]>
Tested-by: Dietmar Eggemann <[email protected]>
Signed-off-by: Juri Lelli <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>

1 files changed, 3 insertions, 3 deletions

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index a14e5fbbea46..eb764c24bc4d 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3234,13 +3234,13 @@ static int __init rcu_spawn_gp_kthread(void)
 	t = kthread_create(rcu_gp_kthread, NULL, "%s", rcu_state.name);
 	if (WARN_ONCE(IS_ERR(t), "%s: Could not start grace-period kthread, OOM is now expected behavior\n", __func__))
 		return 0;
+	if (kthread_prio)
+		sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
 	rnp = rcu_get_root();
 	raw_spin_lock_irqsave_rcu_node(rnp, flags);
 	rcu_state.gp_kthread = t;
-	if (kthread_prio) {
+	if (kthread_prio)
 		sp.sched_priority = kthread_prio;
-		sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
-	}
 	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
 	wake_up_process(t);
 	rcu_spawn_nocb_kthreads();