From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexei Starovoitov Subject: Re: [PATCH net-next] net: sched: run ingress qdisc without locks Date: Fri, 01 May 2015 10:32:27 -0700 Message-ID: <5543B8AB.8000303@plumgrid.com> References: <1430450047-5671-1-git-send-email-ast@plumgrid.com> <55436CB2.9070506@mojatatu.com> <554387FB.5040008@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Cc: Daniel Borkmann , netdev@vger.kernel.org To: John Fastabend , Jamal Hadi Salim , "David S. Miller" Return-path: Received: from mail-ie0-f170.google.com ([209.85.223.170]:36510 "EHLO mail-ie0-f170.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751193AbbEARca (ORCPT ); Fri, 1 May 2015 13:32:30 -0400 Received: by iebrs15 with SMTP id rs15so93659247ieb.3 for ; Fri, 01 May 2015 10:32:30 -0700 (PDT) In-Reply-To: <554387FB.5040008@intel.com> Sender: netdev-owner@vger.kernel.org List-ID: On 5/1/15 7:04 AM, John Fastabend wrote: > > Well there is still the TX side, although not as obvious because > of mq and mqprio. Hopefully by say next LPC we can remove the > qdisc lock there in many scenarios. Also the netif_tx_lock is > there as well but many (most?) of the 10G/40G/etc devices can > create per core descriptor rings. I'll start prodding away on that > side shortly I suppose. that will be massive! 'per core descriptor ring' sounds like new driver API. Looking forward to your work :)