]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
rcu: handle quiescent states for PREEMPT_RCU=n, PREEMPT_COUNT=y
authorAnkur Arora <ankur.a.arora@oracle.com>
Fri, 13 Dec 2024 04:06:56 +0000 (20:06 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 29 May 2025 09:13:39 +0000 (11:13 +0200)
[ Upstream commit 83b28cfe796464ebbde1cf7916c126da6d572685 ]

With PREEMPT_RCU=n, cond_resched() provides urgently needed quiescent
states for read-side critical sections via rcu_all_qs().
One reason why this was needed: lacking preempt-count, the tick
handler has no way of knowing whether it is executing in a
read-side critical section or not.

With (PREEMPT_LAZY=y, PREEMPT_DYNAMIC=n), we get (PREEMPT_COUNT=y,
PREEMPT_RCU=n). In this configuration cond_resched() is a stub and
does not provide quiescent states via rcu_all_qs().
(PREEMPT_RCU=y provides this information via rcu_read_unlock() and
its nesting counter.)

So, use the availability of preempt_count() to report quiescent states
in rcu_flavor_sched_clock_irq().

Suggested-by: Paul E. McKenney <paulmck@kernel.org>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
kernel/rcu/tree_plugin.h

index 3600152b858e88284a92607a4c09e8706899c36f..4328ff3252a354147e5052602945d99f23fe5fec 100644 (file)
@@ -975,13 +975,16 @@ static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
  */
 static void rcu_flavor_sched_clock_irq(int user)
 {
-       if (user || rcu_is_cpu_rrupt_from_idle()) {
+       if (user || rcu_is_cpu_rrupt_from_idle() ||
+            (IS_ENABLED(CONFIG_PREEMPT_COUNT) &&
+             (preempt_count() == HARDIRQ_OFFSET))) {
 
                /*
                 * Get here if this CPU took its interrupt from user
-                * mode or from the idle loop, and if this is not a
-                * nested interrupt.  In this case, the CPU is in
-                * a quiescent state, so note it.
+                * mode, from the idle loop without this being a nested
+                * interrupt, or while not holding the task preempt count
+                * (with PREEMPT_COUNT=y). In this case, the CPU is in a
+                * quiescent state, so note it.
                 *
                 * No memory barrier is required here because rcu_qs()
                 * references only CPU-local variables that other CPUs