From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [UNTESTED] gro: Add page frag support Date: Mon, 29 Dec 2008 23:07:03 -0800 (PST) Message-ID: <20081229.230703.265458431.davem@davemloft.net> References: <20081226225157.GA22866@gondor.apana.org.au> <20081229122445.GA11161@gondor.apana.org.au> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: herbert@gondor.apana.org.au Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:53544 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1751629AbYL3HHA (ORCPT ); Tue, 30 Dec 2008 02:07:00 -0500 In-Reply-To: <20081229122445.GA11161@gondor.apana.org.au> Sender: netdev-owner@vger.kernel.org List-ID: From: Herbert Xu Date: Mon, 29 Dec 2008 23:24:46 +1100 > On Sat, Dec 27, 2008 at 09:51:57AM +1100, Herbert Xu wrote: > > > > This patch is totally untested so please don't apply it. I need > > to hack the e1000e driver to not make skbs before I can test it. > > OK I've tested it with a hacked e1000e driver, which indeed revealed > a gaping hole in my design :) The pages interface needs to have some > extra fields for the checksum information. > > Here's a patch that actually works. > > gro: Use gso_size to store MSS > > In order to allow GRO packets without frag_list at all, we need to > store the MSS in the packet itself. The obvious place is gso_size. > The only thing to watch out for is if the packet ends up not being > GRO then we need to clear gso_size before pushing the packet into > the stack. ... > And the real thing. This work looks great to me Herbert. When you want me to apply this stuff, send them as two seperate changes with updated commit messages. Thanks a lot!