rcu: Make force_quiescent_state() start grace period if needed

Grace periods cannot be started while force_quiescent_state() is
active.  This is OK in that the affected CPUs will try again
later, but it does induce needless grace-period delays.  This
patch causes rcu_start_gp() to record a failed attempt to start
a grace period. When force_quiescent_state() prepares to return,
it then starts the grace period if there was such a failed
attempt.

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: <12626465501854-git-send-email->
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index d920285..55e8f6e 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -660,6 +660,8 @@
 	struct rcu_node *rnp = rcu_get_root(rsp);
 
 	if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {
+		if (cpu_needs_another_gp(rsp, rdp))
+			rsp->fqs_need_gp = 1;
 		if (rnp->completed == rsp->completed) {
 			spin_unlock_irqrestore(&rnp->lock, flags);
 			return;
@@ -1239,6 +1241,12 @@
 		break;
 	}
 	rsp->fqs_active = 0;
+	if (rsp->fqs_need_gp) {
+		spin_unlock(&rsp->fqslock); /* irqs remain disabled */
+		rsp->fqs_need_gp = 0;
+		rcu_start_gp(rsp, flags); /* releases rnp->lock */
+		return;
+	}
 	spin_unlock(&rnp->lock);  /* irqs remain disabled */
 unlock_fqs_ret:
 	spin_unlock_irqrestore(&rsp->fqslock, flags);