]> www.infradead.org Git - users/hch/misc.git/commitdiff
rcu: Document that rcu_barrier() hurries lazy callbacks
authorPaul E. McKenney <paulmck@kernel.org>
Mon, 23 Jun 2025 19:39:55 +0000 (12:39 -0700)
committerPaul E. McKenney <paulmck@kernel.org>
Wed, 13 Aug 2025 22:04:06 +0000 (15:04 -0700)
This commit adds to the rcu_barrier() kerneldoc header stating that this
function hurries lazy callbacks and that it does not normally result in
additional RCU grace periods.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
kernel/rcu/tree.c

index 8eff357b0436be7a1dc2afeeeff8b1cffbeb21c7..1291e0761d70abf82602919e73e06a2bb4fdc9c2 100644 (file)
@@ -3800,6 +3800,11 @@ static void rcu_barrier_handler(void *cpu_in)
  * to complete.  For example, if there are no RCU callbacks queued anywhere
  * in the system, then rcu_barrier() is within its rights to return
  * immediately, without waiting for anything, much less an RCU grace period.
+ * In fact, rcu_barrier() will normally not result in any RCU grace periods
+ * beyond those that were already destined to be executed.
+ *
+ * In kernels built with CONFIG_RCU_LAZY=y, this function also hurries all
+ * pending lazy RCU callbacks.
  */
 void rcu_barrier(void)
 {