From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: [RFC] add rtnl semaphore to linux-atm Date: Wed, 1 Oct 2003 06:14:26 -0700 Sender: netdev-bounce@oss.sgi.com Message-ID: <20031001061426.0b67a235.davem@redhat.com> References: <20031001054226.126cea7b.davem@redhat.com> <200310011307.h91D7jkT004153@ginger.cmf.nrl.navy.mil> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: netdev@oss.sgi.com Return-path: To: chas williams In-Reply-To: <200310011307.h91D7jkT004153@ginger.cmf.nrl.navy.mil> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Wed, 01 Oct 2003 09:07:45 -0400 chas williams wrote: > i had initially written it that way but remembered at one point i > was going to use the rtnl semaphore to handle this problem. any > opinions on what is 'better'? Blocking all network configuration operations (even ones not for your subsystem) is a little bit anti-social in SMP cases. If you take the rwlock as a reader, you only interfere with a very minute class of network configuration code paths (those that need to take the rwlock in question as a writer). For example, if you use the rwlock-as-reader approach, someone doing IPV4 routing table updates (ie. routing daemon changing a couple thousand routes after a BGP flap) won't be perturbed while the ATM operation is in progress.