From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeff Garzik Subject: Re: [PATCH RFC]: napi_struct V5 Date: Thu, 09 Aug 2007 01:32:24 -0400 Message-ID: <46BAA6E8.4050107@garzik.org> References: <1186587154.5155.43.camel@localhost> <20070808.212353.105403708.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: hadi@cyberus.ca, xma@us.ibm.com, netdev@vger.kernel.org, netdev-owner@vger.kernel.org, rusty@rustcorp.com.au, shemminger@linux-foundation.org To: David Miller , rdreier@cisco.com Return-path: Received: from srv5.dvmed.net ([207.36.208.214]:44064 "EHLO mail.dvmed.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754917AbXHIFcl (ORCPT ); Thu, 9 Aug 2007 01:32:41 -0400 In-Reply-To: <20070808.212353.105403708.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org David Miller wrote: > From: jamal > Date: Wed, 08 Aug 2007 11:32:34 -0400 > >> Think of a box where you have other network interfaces, the way you >> are implementing currently implies you are going to be very unfair to >> the other interfaces on the box. > > This was the point I was trying to make the other day. Agreed. That's one of the big selling points of NAPI when I talk to people -- the entire system works towards a single equilibrium, when multiple interfaces are under load. And conversely, without NAPI, the driver has no knowledge of conditions outside its limited view, and resource imbalances inevitably ensue. Particularly so for infiniband, 10gb, etc. where the natural motivation of the driver writer appears to trend towards "performance even at the expense of other system entities." Jeff