From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1764312AbXKNGRB (ORCPT ); Wed, 14 Nov 2007 01:17:01 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1760488AbXKNGQv (ORCPT ); Wed, 14 Nov 2007 01:16:51 -0500 Received: from smtp105.mail.mud.yahoo.com ([209.191.85.215]:30700 "HELO smtp105.mail.mud.yahoo.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1750727AbXKNGQu (ORCPT ); Wed, 14 Nov 2007 01:16:50 -0500 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com.au; h=Received:X-YMail-OSG:From:To:Subject:Date:User-Agent:Cc:References:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding:Content-Disposition:Message-Id; b=yeMeHbBlSBVPtBgwxhGld8M6Uyg2K/e7KUjQjyM9C3ROgLJgXuJ6L82Of1N9lbDWDB0EyDn8yT4gxAW6LGuO9jGflr39huak4jeE7tLleb3F09pm/K+xHMGtZ0p7uPN2xm1ybB/PmLzpDIvLyGt4SpK8wMFe1N8ySb+eu2mYAqU= ; X-YMail-OSG: CK4n.3AVM1k80puSAQ.owsJvAg9pvTfY6SWwObdFbWfTfi0EfBZ9FBG3Ro.pnie0HsNKreon6A-- From: Nick Piggin To: David Miller Subject: Re: 2.6.24-rc2: Network commit causes SLUB performance regression with tbench Date: Wed, 14 Nov 2007 05:14:27 +1100 User-Agent: KMail/1.9.5 Cc: clameter@sgi.com, netdev@vger.kernel.org, herbert@gondor.apana.org.au, linux-kernel@vger.kernel.org References: <200711132241.59074.nickpiggin@yahoo.com.au> <200711140436.24492.nickpiggin@yahoo.com.au> <20071113.221242.153352652.davem@davemloft.net> In-Reply-To: <20071113.221242.153352652.davem@davemloft.net> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200711140514.28159.nickpiggin@yahoo.com.au> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wednesday 14 November 2007 17:12, David Miller wrote: > From: Nick Piggin > Date: Wed, 14 Nov 2007 04:36:24 +1100 > > > On Wednesday 14 November 2007 12:58, David Miller wrote: > > > I suspect the issue is about having a huge skb->data linear area for > > > TCP sends over loopback. We're likely getting a much smaller > > > skb->data linear data area after the patch in question, the rest using > > > the sk_buff scatterlist pages which are a little bit more expensive to > > > process. > > > > It didn't seem to be noticeable at 1 client. Unless scatterlist > > processing is going to cause cacheline bouncing, I don't see why this > > hurts more as you add CPUs? > > Is your test system using HIGHMEM? > > That's one thing the page vector in the sk_buff can do a lot, > kmaps. No, it's an x86-64, so no highmem. What's also interesting is that SLAB apparently doesn't have this condition. The first thing that sprung to mind is that SLAB caches order > 0 allocations, while SLUB does not. However if anything, that should actually favour the SLUB numbers if network is avoiding order > 0 allocations. I'm doing some oprofile runs now to see if I can get any more info.