From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [RFC] net: remove busylock Date: Thu, 19 May 2016 12:35:23 -0700 Message-ID: <1463686523.18194.232.camel@edumazet-glaptop3.roam.corp.google.com> References: <1463677716.18194.203.camel@edumazet-glaptop3.roam.corp.google.com> <1463684190.18194.228.camel@edumazet-glaptop3.roam.corp.google.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev , Alexander Duyck To: Alexander Duyck Return-path: Received: from mail-pa0-f66.google.com ([209.85.220.66]:33872 "EHLO mail-pa0-f66.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753910AbcESTf1 (ORCPT ); Thu, 19 May 2016 15:35:27 -0400 Received: by mail-pa0-f66.google.com with SMTP id yl2so8726267pac.1 for ; Thu, 19 May 2016 12:35:26 -0700 (PDT) In-Reply-To: <1463684190.18194.228.camel@edumazet-glaptop3.roam.corp.google.com> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 2016-05-19 at 11:56 -0700, Eric Dumazet wrote: > Removing busylock helped in all cases I tested. (at least on x86 as > David pointed out) > > As I said, we need to revisit busylock now that spinlocks are different. > > In one case (20 concurrent UDP netperf), I even got a 500 % increase. > > With busylock : > > lpaa5:~# sar -n DEV 4 4|grep eth0 > Average: eth0 12.19 112797.12 1.95 37672.28 0.00 0.00 0.69 > Hmpf, my sysctl logic was inverted. Really these results made little sense. Sorry for the noise. At least we have 8% confirmed gain with this stuff ;) > Presumably it would tremendously help if the actual kfree_skb() > was done after qdisc lock is released, ie not from the qdisc->enqueue() > method. > This part is still valid. We could have a per cpu storage of one skb pointer, so that we do not have to change all ->enqueue() prototypes.