From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] TCP : keep copied_seq, rcv_wup and rcv_next together Date: Thu, 22 Feb 2007 11:43:32 +0100 Message-ID: <200702221143.32622.dada1@cosmosbay.com> References: <1171583205.29021.44.camel@localhost> <200702221122.02806.dada1@cosmosbay.com> <20070222.023220.92577095.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: David Miller Return-path: Received: from pfx2.jmh.fr ([194.153.89.55]:40373 "EHLO pfx2.jmh.fr" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750726AbXBVKnn (ORCPT ); Thu, 22 Feb 2007 05:43:43 -0500 In-Reply-To: <20070222.023220.92577095.davem@davemloft.net> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Thursday 22 February 2007 11:32, David Miller wrote: > From: Eric Dumazet > Date: Thu, 22 Feb 2007 11:22:02 +0100 > > > rcv_wup and copied_seq should be next to rcv_nxt field, to lower number > > of active cache lines in hot paths. (tcp_rcv_established(), tcp_poll(), > > ...) > > Please fixup the assignment order in tcp_create_openreq_child() else > we'll get a new store buffer stall during socket creation :-) OK :) I wonder if : newtp->snd_nxt = newtp->snd_una = newtp->snd_sml = treq->snt_isn + 1; is correct, since compiler will write snd_sml, then snd_una, and snd_nxt Maybe we should also reorder this too ? Eric