From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757207Ab0KOAt4 (ORCPT ); Sun, 14 Nov 2010 19:49:56 -0500 Received: from mail-ww0-f44.google.com ([74.125.82.44]:64675 "EHLO mail-ww0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755027Ab0KOAtx (ORCPT ); Sun, 14 Nov 2010 19:49:53 -0500 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=subject:from:to:cc:in-reply-to:references:content-type:date :message-id:mime-version:x-mailer:content-transfer-encoding; b=DnOP2YbOeaFDwd0Ll6lhC7sn6qdAvyNov0YLlHOsP+GnoYK4in/lVuMfyMqLljeBzJ kutZTAVH5jMOE94LqYHKK2m//rtzeFQenqIZDlOBHx3VCutmFX1oBXsbKDHQzJol8JMv kBCTfJDB4IkGZuwWDsr8XSeF+C/8fL/E9ZJNo= Subject: Re: [PATCH] Reduce number of pointer dereferences in IPv4 netfilter LOG module function dump_packet() From: Eric Dumazet To: Jesper Juhl Cc: Netfilter Core Team , netdev@vger.kernel.org, linux-kernel@vger.kernel.org, netfilter@vger.kernel.org, netfilter-devel@vger.kernel.org, "David S. Miller" , Rusty Russell In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" Date: Mon, 15 Nov 2010 01:49:40 +0100 Message-ID: <1289782180.2743.166.camel@edumazet-laptop> Mime-Version: 1.0 X-Mailer: Evolution 2.30.3 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Le dimanche 14 novembre 2010 à 22:35 +0100, Jesper Juhl a écrit : > By adding two pointer variables to > net/ipv4/netfilter/ipt_LOG.c::dump_packet() we can save 16 bytes of .text > and 9 pointer dereferences. > > before this patch we did 20 pointer dereferences and had this object file > size: > text data bss dec hex filename > 6233 600 3080 9913 26b9 net/ipv4/netfilter/ipt_LOG.o > > after this patch we do just 11 pointer dereferences and have this object > file size: > text data bss dec hex filename > 6217 600 3080 9897 26a9 net/ipv4/netfilter/ipt_LOG.o > > > Please Cc me on replies. > > > Signed-off-by: Jesper Juhl > --- > ipt_LOG.c | 16 ++++++++++------ > 1 file changed, 10 insertions(+), 6 deletions(-) > > diff --git a/net/ipv4/netfilter/ipt_LOG.c b/net/ipv4/netfilter/ipt_LOG.c > index 72ffc8f..02a92de 100644 > --- a/net/ipv4/netfilter/ipt_LOG.c > +++ b/net/ipv4/netfilter/ipt_LOG.c > @@ -39,6 +39,8 @@ static void dump_packet(struct sbuff *m, > struct iphdr _iph; > const struct iphdr *ih; > unsigned int logflags; > + struct sock *sk; > + struct socket *sk_socket; > > if (info->type == NF_LOG_TYPE_LOG) > logflags = info->u.log.logflags; > @@ -335,13 +337,15 @@ static void dump_packet(struct sbuff *m, > } > > /* Max length: 15 "UID=4294967295 " */ > - if ((logflags & IPT_LOG_UID) && !iphoff && skb->sk) { > - read_lock_bh(&skb->sk->sk_callback_lock); > - if (skb->sk->sk_socket && skb->sk->sk_socket->file) > + sk = skb->sk; > + sk_socket = sk->sk_socket; Really ? sk can be NULL, so you add a NULL dereference. > + if ((logflags & IPT_LOG_UID) && !iphoff && sk) { > + read_lock_bh(&sk->sk_callback_lock); > + if (sk_socket && sk_socket->file) > sb_add(m, "UID=%u GID=%u ", > - skb->sk->sk_socket->file->f_cred->fsuid, > - skb->sk->sk_socket->file->f_cred->fsgid); > - read_unlock_bh(&skb->sk->sk_callback_lock); > + sk_socket->file->f_cred->fsuid, > + sk_socket->file->f_cred->fsgid); > + read_unlock_bh(&sk->sk_callback_lock); > } > > /* Max length: 16 "MARK=0xFFFFFFFF " */ > > > Most of these "dereferences" are compiler optimized. You added a bug in your patch, and make ipt_LOG slower if rule is not asking for IPT_LOG_UID