From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Hemminger <shemminger@vyatta.com>
Subject: Re: [PATCH] netfilter: use per-cpu spinlock rather than RCU
Date: Mon, 13 Apr 2009 11:11:06 -0700
Message-ID: <20090413111106.718888ca@nehalam>
References: <20090411174801.GG6822@linux.vnet.ibm.com>
	<18913.53699.544083.320542@cargo.ozlabs.ibm.com>
	<20090412173108.GO6822@linux.vnet.ibm.com>
	<20090412.181330.23529546.davem@davemloft.net>
	<20090413040413.GQ6822@linux.vnet.ibm.com>
	<20090413095309.631cf395@nehalam>
	<49E37908.2080903@cosmosbay.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: paulmck@linux.vnet.ibm.com, David Miller <davem@davemloft.net>,
	paulus@samba.org, mingo@elte.hu, torvalds@linux-foundation.org,
	laijs@cn.fujitsu.com, jeff.chua.linux@gmail.com,
	jengelh@medozas.de, kaber@trash.net, r000n@r000n.net,
	linux-kernel@vger.kernel.org, netfilter-devel@vger.kernel.org,
	netdev@vger.kernel.org, benh@kernel.crashing.org
To: Eric Dumazet <dada1@cosmosbay.com>
Return-path: <netdev-owner@vger.kernel.org>
In-Reply-To: <49E37908.2080903@cosmosbay.com>
Sender: netdev-owner@vger.kernel.org
List-Id: netfilter-devel.vger.kernel.org

On Mon, 13 Apr 2009 19:40:24 +0200
Eric Dumazet <dada1@cosmosbay.com> wrote:

> Stephen Hemminger a =C3=A9crit :
> > This is an alternative version of ip/ip6/arp tables locking using
> > per-cpu locks.  This avoids the overhead of synchronize_net() durin=
g
> > update but still removes the expensive rwlock in earlier versions.
> >=20
> > The idea for this came from an earlier version done by Eric Duzamet=
=2E
> > Locking is done per-cpu, the fast path locks on the current cpu
> > and updates counters.  The slow case involves acquiring the locks o=
n
> > all cpu's.
> >=20
> > The mutex that was added for 2.6.30 in xt_table is unnecessary sinc=
e
> > there already is a mutex for xt[af].mutex that is held.
> >=20
> > Tested basic functionality (add/remove/list), but don't have test c=
ases
> > for stress, ip6tables or arptables.
> >=20
> > Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
>=20
> Patch seems good to me, but apparently xt_replace_table()
> misses the "acquiring the locks on all cpus" you mentioned in ChangeL=
og ?

It happens in get_counters already.

> I am still off-computers until tomorrow so cannot provide a patch for=
 this, sorry.
>=20
> Some form of
>=20
> local_bh_disable();
> for_each_possible_cpu(cpu)
> 	spin_lock(&per_cpu(ip_tables_lock, cpu));
>=20
> oldinfo =3D private;
> /* do the substitution */
> table->private =3D newinfo;
> newinfo->initial_entries =3D oldinfo->initial_entries;
>=20
> for_each_possible_cpu(cpu)
> 	spin_unlock(&per_cpu(ip_tables_lock, cpu));
> local_bh_enable();
>=20
>=20
> But I wonder if this could hit a limit of max spinlocks held by this =
cpu, say on a 4096 cpu machine ?
>=20
>=20
>=20