From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jesper Dangaard Brouer Subject: Re: [net-next PATCH] igb: update adapter stats when reading /proc/net/dev. Date: Thu, 07 Oct 2010 13:36:00 +0200 Message-ID: <1286451360.6992.32.camel@firesoul.comx.local> References: <20101005141833.20929.10943.stgit@localhost> <20101006.233611.68115178.davem@davemloft.net> <1286433848.9417.49.camel@edumazet-laptop> <20101006.234611.191419337.davem@davemloft.net> <1286446004.6992.7.camel@firesoul.comx.local> <1286447084.2912.24.camel@edumazet-laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , netdev@vger.kernel.org, jeffrey.t.kirsher@intel.com To: Eric Dumazet Return-path: Received: from lanfw001a.cxnet.dk ([87.72.215.196]:55668 "EHLO lanfw001a.cxnet.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760334Ab0JGLgF (ORCPT ); Thu, 7 Oct 2010 07:36:05 -0400 In-Reply-To: <1286447084.2912.24.camel@edumazet-laptop> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 2010-10-07 at 12:24 +0200, Eric Dumazet wrote: > Le jeudi 07 octobre 2010 =C3=A0 12:06 +0200, Jesper Dangaard Brouer a= =C3=A9crit : > > On Wed, 2010-10-06 at 23:46 -0700, David Miller wrote: > > > From: Eric Dumazet > > > Date: Thu, 07 Oct 2010 08:44:08 +0200 > > >=20 > > > > Le mercredi 06 octobre 2010 =C3=A0 23:36 -0700, David Miller a = =C3=A9crit : > > > >> From: Jesper Dangaard Brouer > > > >> Date: Tue, 05 Oct 2010 16:18:33 +0200 > > > >>=20 > > > >> > Network driver igb: Improve the accuracy of stats in /proc/n= et/dev, by > > > >> > updating the adapter stats when reading /proc/net/dev. Curr= ently the > > > >> > stats are updated by the watchdog timer every 2 sec, or when= getting > > > >> > stats via ethtool -S. > > > >> >=20 > > > >> > A number of userspace apps read these /proc/net/dev stats ev= ery second, > > > >> > e.g. ifstat, which then gives a perceived very bursty traffi= c pattern, > > > >> > which is actually false. > > > >> >=20 > > > >> > Signed-off-by: Jesper Dangaard Brouer > > > >>=20 > > > >> I assume that the Intel folks will take care of integrating th= is > > > >> now that the locking is fixed. > > > >=20 > > > > I integrated Jesper patch into my cumulative patch. > > >=20 > > > Ok. > >=20 > > I'm fine with this, as long as the commit message describe this cha= nge > > of accuracy of stats in /proc/net/dev. > >=20 > > Something like: > >=20 > > This patch also increase the accuracy of stats in /proc/net/dev, b= y > > updating the adapter stats when reading /proc/net/dev. Previously > > the stats were only updated by the watchdog timer every 2 sec, whi= ch > > resulted in false observations from userspace. > >=20 > >=20 >=20 > Well, its not exactly true :) >=20 > Previously, igb stats were updated : > - By watchdog timer, every 2 secs > - Every time an "ethtool -S ethX" was done >=20 > There is no general guarantee netdev stats are immediately available = to > user. I'm not talking general, just for this driver. And I'm not giving any guarantees, that is why I choose the wording "increase the accuracy" and not "provide accurate stats". > ndo_get_stats() is not allowed to sleep, (because of bonding...), so > drivers can not always provide accurate stats, if they need to make a > long transaction with hardware. >=20 > Other drivers do the same (provide hardware statistics), with about o= ne > second resolution. Yes, a lot of drivers don't provide accurate states. And one second resolution, as most drivers use, will usually be good enough for most admins. Our usage pattern is a bit different, as we are very interested in measuring bursty traffic. I'm explain you why during the Netfilter Workshop, if I get my talk about IPTV accepted ;-) I also use the increased accuracy, when running my pktgen testing scripts. > So the "resulted in false observations from userspace." is something > that might upset admins, but is not a hard requirement of netdev stat= s. Its definitly not a hard requirement, but lets fix it where we can. Then change the text: "resulted in false observations from userspace" to: "resulted in inaccurate observations from userspace" --=20 Med venlig hilsen / Best regards Jesper Brouer ComX Networks A/S Linux Network Kernel Developer Cand. Scient Datalog / MSc.CS Author of http://adsl-optimizer.dk LinkedIn: http://www.linkedin.com/in/brouer