From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH] [NET] Size listen hash tables using backlog hint Date: Thu, 19 Oct 2006 01:41:20 -0700 (PDT) Message-ID: <20061019.014120.78710373.davem@davemloft.net> References: <45371C8D.20603@cosmosbay.com> <20061018.235747.74746971.davem@davemloft.net> <200610191029.00720.dada1@cosmosbay.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org Return-path: Received: from dsl027-180-168.sfo1.dsl.speakeasy.net ([216.27.180.168]:7870 "EHLO sunset.davemloft.net") by vger.kernel.org with ESMTP id S1030322AbWJSIlV (ORCPT ); Thu, 19 Oct 2006 04:41:21 -0400 To: dada1@cosmosbay.com In-Reply-To: <200610191029.00720.dada1@cosmosbay.com> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org From: Eric Dumazet Date: Thu, 19 Oct 2006 10:29:00 +0200 > Cons : > TLB cost For those hot x86 and x86_64 cpus you tend to be using, this particular cost is relatively small. :-) It's effectively like another memory reference in the worst case, in the best case it's "free". > With actual somaxconn=128 limit, my patch ends in allocating less > ram (half of a page) than current x86_64 kernel (2 pages) Understood. But the issue is that there are greater security implications than before when increasing this sysctl. To be honest, it's probably water under the bridge, because if you can stuff up SOMAXCONN number of sockets into the system per listening socket which is a lot more than the hash table eats up. :-)