From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: UDP is bypassing qdisc statistics .... Date: Tue, 01 Sep 2009 16:58:15 +0200 Message-ID: <4A9D3687.2020101@gmail.com> References: <20090901063726.GA5222@ff.dom.local> <4A9D2DBF.8030200@trash.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Christoph Lameter , Mark Smith , Jarek Poplawski , netdev@vger.kernel.org, davem@linux-foundation.org To: Patrick McHardy Return-path: Received: from gw1.cosmosbay.com ([212.99.114.194]:35364 "EHLO gw1.cosmosbay.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754753AbZIAO6W (ORCPT ); Tue, 1 Sep 2009 10:58:22 -0400 In-Reply-To: <4A9D2DBF.8030200@trash.net> Sender: netdev-owner@vger.kernel.org List-ID: Patrick McHardy a =E9crit : > Christoph Lameter wrote: >> Eric: This occurs with a Broadcom driver (bnx2). There is only one >> network device up and the IP statistics are correctly incremented. >=20 > That explains it. The bnx2 driver uses multiple TX queues, but > tc_dump_qdisc() only dumps the statistics from queue number 0. >=20 > Not sure whether we should sum them up or dump the statistics > from each queue (or both). Summing them up avoids userspace > visible changes when drivers are converted to multiq. Ok mystery solved, partially at least :) Hmm, my bnx2 still displays correct info here (BCM5708S) ; so maybe Chr= istoph has a real multiqueue NIC adapter ? I thought he mentioned a mono queue NIC e= arlier. Dumping stats for each queue would be better for admin point of view IM= HO (restricted to real_num_tx_queues I guess, since bnx2 allocates 8 queue= s per device.