From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jamal Hadi Salim Subject: Re: [NET]: Prevent multiple qdisc runs Date: Sat, 24 Jun 2006 09:50:39 -0400 Message-ID: <1151157039.6716.71.camel@jzny2> References: <20060619134227.GA16662@gondor.apana.org.au> <1150727009.5815.72.camel@jzny2> <20060619142928.GA17191@gondor.apana.org.au> <1150727810.5815.78.camel@jzny2> <20060619223326.GA20354@gondor.apana.org.au> <1150814526.5270.64.camel@jzny2> <20060620235207.GA31493@gondor.apana.org.au> <1151004682.5392.97.camel@jzny2> <20060622224302.GA10136@gondor.apana.org.au> <1151023937.5099.40.camel@jzny2> <20060623033525.GA11751@gondor.apana.org.au> Reply-To: hadi@znyx.com Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, "David S. Miller" , Robert Olsson Return-path: Received: from mailhub.znyx.com ([208.2.156.141]:41745 "EHLO mailhub.znyx.com") by vger.kernel.org with ESMTP id S1750713AbWFXNvV (ORCPT ); Sat, 24 Jun 2006 09:51:21 -0400 To: Herbert Xu In-Reply-To: <20060623033525.GA11751@gondor.apana.org.au> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Fri, 2006-23-06 at 13:35 +1000, Herbert Xu wrote: > On Thu, Jun 22, 2006 at 08:52:17PM -0400, jamal wrote: > > > > It does feel like the qdisc_is_running though is now a replacement > > for the need for dev->txlock which existed to protect multi-cpus from > > entering the device transmit path. Is that unintended side effect? > > i.e why would dev->txlock be needed anymore in that path? > > It's not totally redundant yet since you can set tx_queue_len to zero. > It also still protects against the asynchronous paths that take xmit_lock. The tx_timeout? > However, it wouldn't be a bad idea to see if there is a way to reduce the > number of locks required on the xmit path to one. It also seems to be there - dont know why it took me so long to see this. I am actually gut-feeling now that we may get better performance with LLTX drivers with this approach;-> talk about a flip-flop. The qualification for LLTX seems to be for a driver to have a private tx lock which is the case of about every ethernet driver out there. cheers, jamal