From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from shards.monkeyblade.net ([184.105.139.130]:46788 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932140AbeB1Olt (ORCPT ); Wed, 28 Feb 2018 09:41:49 -0500 Date: Wed, 28 Feb 2018 09:41:47 -0500 (EST) Message-Id: <20180228.094147.1456557446129712432.davem@davemloft.net> To: xiangxia.m.yue@gmail.com Cc: netdev@vger.kernel.org Subject: Re: [PATCH] tcp: Support tcp socket allocated counter in namespace. From: David Miller In-Reply-To: References: <1518489840-27911-1-git-send-email-xiangxia.m.yue@gmail.com> <20180213.121946.1379751706930638939.davem@davemloft.net> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: netdev-owner@vger.kernel.org List-ID: From: Tonghao Zhang Date: Wed, 28 Feb 2018 21:01:52 +0800 >> The amount of new conditional tests in these fast paths are not >> justified for this new counter which is of debatable usefullness. > sorry, too late for reply. Did you mean this counter will affect performance ? > I tested the patch witch netperf. A single flow TCP session with no loss doesn't tell us much. I just know from how these things go that under load, and with all kinds of different flows with different characterstics, every conditional test in the fast paths matter. If you want to have a chance at your change getting accepted, running all kinds of benchmarks won't do it. Instead, please find a way to make your feature work without all of the newly added tests (whilst not adding another cost at the same time). Thank you.