[tip: core/rcu] rcu: Optimize and protect atomic_cmpxchg() loop

[tip: core/rcu] rcu: Optimize and protect atomic_cmpxchg() loop

[tip-bot2 for Paul E. McKenney]

The following commit has been merged into the core/rcu branch of tip:

Commit-ID:     faa059c397dec8a452c79e9dba64419113ea64e2
Gitweb:        https://git.kernel.org/tip/faa059c397dec8a452c79e9dba64419113ea64e2
Author:        Paul E. McKenney <paulmck@kernel.org>
AuthorDate:    Mon, 03 Feb 2020 14:20:00 -08:00
Committer:     Paul E. McKenney <paulmck@kernel.org>
CommitterDate: Thu, 20 Feb 2020 15:58:23 -08:00

rcu: Optimize and protect atomic_cmpxchg() loop

This commit reworks the atomic_cmpxchg() loop in rcu_eqs_special_set()
to do only the initial read from the current CPU's rcu_data structure's
->dynticks field explicitly.  On subsequent passes, this value is instead
retained from the failing atomic_cmpxchg() operation.

Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 kernel/rcu/tree.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 4a4a975..6c62481 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -342,14 +342,17 @@ bool rcu_eqs_special_set(int cpu)
 {
 	int old;
 	int new;
+	int new_old;
 	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
 
+	new_old = atomic_read(&rdp->dynticks);
 	do {
-		old = atomic_read(&rdp->dynticks);
+		old = new_old;
 		if (old & RCU_DYNTICK_CTRL_CTR)
 			return false;
 		new = old | RCU_DYNTICK_CTRL_MASK;
-	} while (atomic_cmpxchg(&rdp->dynticks, old, new) != old);
+		new_old = atomic_cmpxchg(&rdp->dynticks, old, new);
+	} while (new_old != old);
 	return true;
 }