From owner-freebsd-net@FreeBSD.ORG  Tue Sep 30 18:01:02 2014
Return-Path: <owner-freebsd-net@FreeBSD.ORG>
Delivered-To: net@freebsd.org
Received: from mx1.freebsd.org (mx1.freebsd.org
 [IPv6:2001:1900:2254:206a::19:1])
 (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits))
 (No client certificate requested)
 by hub.freebsd.org (Postfix) with ESMTPS id BFBFBBD1;
 Tue, 30 Sep 2014 18:01:02 +0000 (UTC)
Received: from bigwig.baldwin.cx (bigwig.baldwin.cx [IPv6:2001:470:1f11:75::1])
 (using TLSv1 with cipher DHE-RSA-CAMELLIA256-SHA (256/256 bits))
 (Client did not present a certificate)
 by mx1.freebsd.org (Postfix) with ESMTPS id 9B5981DD;
 Tue, 30 Sep 2014 18:01:02 +0000 (UTC)
Received: from jhbbsd.localnet (unknown [209.249.190.124])
 by bigwig.baldwin.cx (Postfix) with ESMTPSA id 5622EB9CF;
 Tue, 30 Sep 2014 14:01:01 -0400 (EDT)
From: John Baldwin <jhb@freebsd.org>
To: net@freebsd.org
Subject: [PATCH] Only lock send buffer in sopoll() if needed
Date: Tue, 30 Sep 2014 14:00:09 -0400
User-Agent: KMail/1.13.5 (FreeBSD/8.4-CBSD-20140415; KDE/4.5.5; amd64; ; )
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="us-ascii"
Content-Transfer-Encoding: 7bit
Message-Id: <201409301400.09999.jhb@freebsd.org>
X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.2.7
 (bigwig.baldwin.cx); Tue, 30 Sep 2014 14:01:01 -0400 (EDT)
Cc: Robert Watson <rwatson@freebsd.org>
X-BeenThere: freebsd-net@freebsd.org
X-Mailman-Version: 2.1.18-1
Precedence: list
List-Id: Networking and TCP/IP with FreeBSD <freebsd-net.freebsd.org>
List-Unsubscribe: <http://lists.freebsd.org/mailman/options/freebsd-net>,
 <mailto:freebsd-net-request@freebsd.org?subject=unsubscribe>
List-Archive: <http://lists.freebsd.org/pipermail/freebsd-net/>
List-Post: <mailto:freebsd-net@freebsd.org>
List-Help: <mailto:freebsd-net-request@freebsd.org?subject=help>
List-Subscribe: <http://lists.freebsd.org/mailman/listinfo/freebsd-net>,
 <mailto:freebsd-net-request@freebsd.org?subject=subscribe>
X-List-Received-Date: Tue, 30 Sep 2014 18:01:02 -0000

Right now sopoll() always locks both socket buffers.  The receive socket 
buffer lock is always needed, but the send socket buffer lock is only needed 
while polling for writing (there is a potential test of SBS_CANTSENDMORE 
without the lock, but I think this might be ok).  What do folks think?

Index: uipc_socket.c
===================================================================
--- uipc_socket.c       (revision 272305)
+++ uipc_socket.c       (working copy)
@@ -3003,7 +3003,12 @@ sopoll_generic(struct socket *so, int events, stru
 {
        int revents = 0;
 
-       SOCKBUF_LOCK(&so->so_snd);
+       if (events & (POLLOUT | POLLWRNORM))
+               SOCKBUF_LOCK(&so->so_snd);
+       /*
+        * The so_rcv lock doubles as SOCK_LOCK so it it is needed for
+        * all requests.
+        */
        SOCKBUF_LOCK(&so->so_rcv);
        if (events & (POLLIN | POLLRDNORM))
                if (soreadabledata(so))
@@ -3038,7 +3043,8 @@ sopoll_generic(struct socket *so, int events, stru
        }
 
        SOCKBUF_UNLOCK(&so->so_rcv);
-       SOCKBUF_UNLOCK(&so->so_snd);
+       if (events & (POLLOUT | POLLWRNORM))
+               SOCKBUF_UNLOCK(&so->so_snd);
        return (revents);
 }

-- 
John Baldwin