rcu: Fix holdoff for accelerated GPs for last non-dynticked CPU
authorPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Sat, 27 Feb 2010 22:53:07 +0000 (14:53 -0800)
committerIngo Molnar <mingo@elte.hu>
Sun, 28 Feb 2010 08:17:42 +0000 (09:17 +0100)
Make the holdoff only happen when the full number of attempts
have been made.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: laijs@cn.fujitsu.com
Cc: dipankar@in.ibm.com
Cc: mathieu.desnoyers@polymtl.ca
Cc: josh@joshtriplett.org
Cc: dvhltc@us.ibm.com
Cc: niv@us.ibm.com
Cc: peterz@infradead.org
Cc: rostedt@goodmis.org
Cc: Valdis.Kletnieks@vt.edu
Cc: dhowells@redhat.com
LKML-Reference: <1267311188-16603-1-git-send-email-paulmck@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/rcutree_plugin.h

index 464ad2c..79b53bd 100644 (file)
@@ -1010,6 +1010,10 @@ int rcu_needs_cpu(int cpu)
        int c = 0;
        int thatcpu;
 
+       /* Check for being in the holdoff period. */
+       if (per_cpu(rcu_dyntick_holdoff, cpu) == jiffies)
+               return rcu_needs_cpu_quick_check(cpu);
+
        /* Don't bother unless we are the last non-dyntick-idle CPU. */
        for_each_cpu_not(thatcpu, nohz_cpu_mask)
                if (thatcpu != cpu) {
@@ -1041,10 +1045,8 @@ int rcu_needs_cpu(int cpu)
        }
 
        /* If RCU callbacks are still pending, RCU still needs this CPU. */
-       if (c) {
+       if (c)
                raise_softirq(RCU_SOFTIRQ);
-               per_cpu(rcu_dyntick_holdoff, cpu) = jiffies;
-       }
        return c;
 }