On Sun, 2013-06-02 at 19:58 -0700, David Miller wrote:

From: Eric Dumazet <redacted>
Date: Sun, 02 Jun 2013 16:31:35 -0700

quoted

On Thu, 2013-05-30 at 12:25 +0400, Roman Gushchin wrote:

quoted

On 29.05.2013 23:06, Eric Dumazet wrote:

quoted

On Wed, 2013-05-29 at 14:09 +0400, Roman Gushchin wrote:

True, these lookup functions are usually structured the same around the
hlist_nulls_for_each_entry_rcu() loop.

A barrier() right before the loop seems to be a benefit, the size of
assembly code is reduced by 48 bytes.

And its one of the documented way to handle this kind of problems
(Documentation/atomic_ops.txt line 114)

I guess we should amend this documentation, eventually.

Thanks, please add you "Signed-off-by" if you agree with the patch.

Signed-off-by: Roman Gushchin <redacted>

Many thanks to you, Paul E. McKenney and David Laight for your
patches, help and participation in this discussion.

Thanks to you !

David, is there any problem with the patch ?

http://patchwork.ozlabs.org/patch/247360/ says "Not applicable", please
tell me what is needed to get it merged.

It's not a networking patch, it's a patch for generic RCU upstream.
So it either goes through Paul McKenney, or directly via Linus on
linux-kernel.

Well, whole discussion went on linux kernel, and you were the original
committer of this patch five years ago.

http://git.kernel.org/cgit/linux/kernel/git/davem/net-next.git/commit/?id=bbaffaca4810de1a25e32ecaf836eeaacc7a3d11

So please Paul or David make sure the patch goes in.

My only concern is that Paul seems quite busy these days.

Thanks