[PATCH] Fix RCU race in access of nohz_cpu_mask

Accessing nohz_cpu_mask before incrementing rcp->cur is racy.  It can cause
tickless idle CPUs to be included in rsp->cpumask, which will extend
graceperiods unnecessarily.

Fix this race.  It has been tested using extensions to RCU torture module
that forces various CPUs to become idle.

Signed-off-by: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Dipankar Sarma <dipankar@in.ibm.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c
index f45b917..48d3bce 100644 (file)
--- a/kernel/rcupdate.c
+++ b/kernel/rcupdate.c
@@ -257,15 +257,23 @@ static void rcu_start_batch(struct rcu_ctrlblk *rcp, struct rcu_state *rsp,
 
        if (rcp->next_pending &&
                        rcp->completed == rcp->cur) {
-               /* Can't change, since spin lock held. */
-               cpus_andnot(rsp->cpumask, cpu_online_map, nohz_cpu_mask);
-
                rcp->next_pending = 0;
-               /* next_pending == 0 must be visible in __rcu_process_callbacks()
-                * before it can see new value of cur.
+               /*
+                * next_pending == 0 must be visible in
+                * __rcu_process_callbacks() before it can see new value of cur.
                 */
                smp_wmb();
                rcp->cur++;
+
+               /*
+                * Accessing nohz_cpu_mask before incrementing rcp->cur needs a
+                * Barrier  Otherwise it can cause tickless idle CPUs to be
+                * included in rsp->cpumask, which will extend graceperiods
+                * unnecessarily.
+                */
+               smp_mb();
+               cpus_andnot(rsp->cpumask, cpu_online_map, nohz_cpu_mask);
+
        }
 }