From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bill Fink Subject: Re: [PATCH] net: add ability to clear stats via ethtool - e1000/pcnet32 Date: Fri, 30 May 2008 15:12:50 -0400 Message-ID: <20080530151250.b44a119a.billfink@mindspring.com> References: <482DA5B6.1020606@sngx.net> <482DB46A.8020103@cosmosbay.com> <482EF192.4070707@sngx.net> <482F5113.5090703@cosmosbay.com> <482F610D.2080108@sngx.net> <20080518003104.GK28241@solarflare.com> <482FBA09.80201@sngx.net> <483E0AAE.2020107@sngx.net> <20080528221118.63da4092.akpm@linux-foundation.org> <483EA2D1.8050603@sngx.net> <20080529154525.3916c7b5@core> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: James Cammarata , Andrew Morton , linux-kernel@vger.kernel.org, Linux Netdev List To: Alan Cox Return-path: Received: from elasmtp-masked.atl.sa.earthlink.net ([209.86.89.68]:49494 "EHLO elasmtp-masked.atl.sa.earthlink.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752009AbYE3TNF (ORCPT ); Fri, 30 May 2008 15:13:05 -0400 In-Reply-To: <20080529154525.3916c7b5@core> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 29 May 2008, Alan Cox wrote: > > > Many of the kernel's accounting accumulators cannot be reset. We > > > handle that in userspace tools by using subtraction. > > > > I don't think that should preclude the ability to reset these, unless it is > > shown that it would break something very badly. > > For one quite a few of the network cards keep the stats in hardware and > don't neccessarily have a way to reset them. In other cases there is a > mix of OS accumulated stats bulk updated by overflow events on the device > itself. > > Its a lot of complexity, and changes all over the places for the sake of > a trivial userspace change. I would suggest you instead write a quick bit > of perl or python that fetches the stats and then updates every second > with the changes. When diagnosing network problems, the ability to zero counters is a major aid in diagnosis. Writing scripts is not a general solution since often several systems are involved and it's not simple to do this via a script. Saving stats output and running beforeafter on a number of systems is a royal pain when troubleshooting. I like the idea someone else had. where the clear stats actually did a checkpoint of the stats, and then future get stats would return the diff between the checkpointed and current values (perhaps there could also be a mechanism to still get the absolute values if desired). -Bill