Re: [GIT PULL rcu/next] rcu commits for 2.6.40

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Fri, May 13, 2011 at 03:12:18PM +0200, Ingo Molnar wrote:
> 
> * Ingo Molnar <mingo@elte.hu> wrote:
> 
> > I started bisecting this, and the two relevant endpoints:
> > 
> >   bad: 11c476f: net,rcu: convert call_rcu(prl_entry_destroy_rcu) to kfree
> >  good: 0ee5623f: Linux 2.6.39-rc6
> > 
> > very clearly indicate that this is an RCU regression.
> 
> This might be the same one Yinghai found:
> 
>  e59fb3120bec: rcu: Decrease memory-barrier usage based on semi-formal proof
> 
> So with the config i sent it's definitely reproducible.
> 
> At first sight couldnt this be related not to barriers, but to not setting 
> need_resched() like we did before?

Thank you both!!!  I had inspected the commit, but missed the fact that
the new version refuses to call set_need_resched() if irqs are enabled.  :-(
The following (untested) patch restores the set_need_resched() operation.

Does this help?

							Thanx, Paul

------------------------------------------------------------------------

rcu: Reschedule from dyntick even if irqs are enabled.

Commit e59fb3120bec (rcu: Decrease memory-barrier usage based on
semi-formal proof) attempted to optimize by invoking set_need_resched()
only when irqs are disabled.  This patch restores the prior behavior
of invoking set_need_resched() even if irqs are enabled, but leaves
the memory-barrier-usage changes untouched.

Reported-by: Yinghai Lu <yinghai@kernel.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index 5b0b482..8c490ef 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -335,10 +335,9 @@ void rcu_enter_nohz(void)
 	local_irq_restore(flags);
 
 	/* If the interrupt queued a callback, get out of dyntick mode. */
-	if (in_irq() &&
-	    (__get_cpu_var(rcu_sched_data).nxtlist ||
-	     __get_cpu_var(rcu_bh_data).nxtlist ||
-	     rcu_preempt_needs_cpu(smp_processor_id())))
+	if (__get_cpu_var(rcu_sched_data).nxtlist ||
+	    __get_cpu_var(rcu_bh_data).nxtlist ||
+	    rcu_preempt_needs_cpu(smp_processor_id()))
 		set_need_resched();
 }