From: Eric Dumazet <dada1@cosmosbay.com>
Date: Wed, 4 Mar 2009 20:11:56 +0000 (-0800)
Subject: rcu: increment quiescent state counter in ksoftirqd()
X-Git-Tag: v2.6.29-rc8~56^2
X-Git-Url: https://git.openpandora.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=64ca5ab913f1594ef316556e65f5eae63ff50cee;p=pandora-kernel.git

rcu: increment quiescent state counter in ksoftirqd()

If a machine is flooded by network frames, a cpu can loop
100% of its time inside ksoftirqd() without calling schedule().
This can delay RCU grace period to insane values.

Adding rcu_qsctr_inc() call in ksoftirqd() solves this problem.

Paul: "This regression was a result of the recent change from
"schedule()" to "cond_resched()", which got rid of that quiescent
state in the common case where a reschedule is not needed".

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---

diff --git a/kernel/softirq.c b/kernel/softirq.c
index bdbe9de9cd8d..9041ea7948fe 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -626,6 +626,7 @@ static int ksoftirqd(void * __bind_cpu)
 			preempt_enable_no_resched();
 			cond_resched();
 			preempt_disable();
+			rcu_qsctr_inc((long)__bind_cpu);
 		}
 		preempt_enable();
 		set_current_state(TASK_INTERRUPTIBLE);