From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sowmini Varadhan Subject: Re: [PATCH net-next v3 1/2] RDS: TCP: Add sysctl tunables for sndbuf/rcvbuf on rds-tcp socket Date: Wed, 16 Mar 2016 07:32:11 -0400 Message-ID: <20160316113211.GC21307@oracle.com> References: <5c18f6bdde99628e486d63a071c256d1be036c28.1458064902.git.sowmini.varadhan@oracle.com> <56E93596.90503@stressinduktion.org> <20160316111055.GB21307@oracle.com> <56E942F0.90201@stressinduktion.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: netdev@vger.kernel.org, davem@davemloft.net, santosh.shilimkar@oracle.com, eric.dumazet@gmail.com To: Hannes Frederic Sowa Return-path: Received: from aserp1040.oracle.com ([141.146.126.69]:36889 "EHLO aserp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965449AbcCPLcW (ORCPT ); Wed, 16 Mar 2016 07:32:22 -0400 Content-Disposition: inline In-Reply-To: <56E942F0.90201@stressinduktion.org> Sender: netdev-owner@vger.kernel.org List-ID: On (03/16/16 12:26), Hannes Frederic Sowa wrote: > > My hope was actually that by using the ->data pointer in netns you > don't need to provide the two functions, just simply use something > like the following for both cases. Ok, it would probably be less hacky to do it as you suggest. > Do I understand it correctly that all connections of a namespace > will be dropped if you modify those sysctls? yes. it's unfortunate, but afaict there's no other way to make them use the new vars. But as I said in the comments, an admin who goes around creating this churn is probably doing this very rarely, and for a good reason, and is fully aware of the cost. So there is some degree of human control possible. --Sowmini