Re: [PATCH 09/10] cfq-iosched: Uses its own open-coded rcu_barrier.

On Wed, Jun 24, 2009 at 08:42:37AM +0200, Jens Axboe wrote:

On Tue, Jun 23 2009, Jesper Dangaard Brouer wrote:

quoted

This module cfq-iosched, has discovered the value of waiting for
call_rcu() completion, but its has its own open-coded implementation
of rcu_barrier(), which I don't think is 'strong' enough.

This patch only leaves a comment for the maintainers to consider.

We need a stronger primitive and rcu_barrier(), since we also need to
wait for the rcu calls to even be scheduled. So I don't think the below
can be improved, it's already fine.

It is indeed important to first prevent new call_rcu() instances from
being invoked, and only then invoke rcu_barrier().

							Thanx, Paul

quoted

Signed-off-by: Jesper Dangaard Brouer <redacted>
---

 block/cfq-iosched.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 833ec18..c15555b 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c

@@ -2657,6 +2657,12 @@ static void __exit cfq_exit(void)
 	/*
 	 * this also protects us from entering cfq_slab_kill() with
 	 * pending RCU callbacks
+	 *
+	 * hawk-4UpuNZONu4c@public.gmane.org 2009-06-18: Maintainer please consider using
+	 *  rcu_barrier() instead of this open-coded wait for
+	 *  completion implementation.  I think it provides a better
+	 *  guarantee that all CPUs are finished, although
+	 *  elv_ioc_count_read() do consider all CPUs.
 	 */
 	if (elv_ioc_count_read(ioc_count))
 		wait_for_completion(&all_gone);

-- 
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html