From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752673AbbKKNE1 (ORCPT ); Wed, 11 Nov 2015 08:04:27 -0500 Received: from unicorn.mansr.com ([81.2.72.234]:44104 "EHLO unicorn.mansr.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752432AbbKKNEK convert rfc822-to-8bit (ORCPT ); Wed, 11 Nov 2015 08:04:10 -0500 From: =?iso-8859-1?Q?M=E5ns_Rullg=E5rd?= To: David Miller Cc: romieu@fr.zoreil.com, linux-kernel@vger.kernel.org, netdev@vger.kernel.org, slash.tmp@free.fr Subject: Re: [PATCH v5] net: ethernet: add driver for Aurora VLSI NB8800 Ethernet controller References: <1447172063-27234-1-git-send-email-mans@mansr.com> <20151110233448.GA8646@electric-eye.fr.zoreil.com> <20151110.211121.422193831316296487.davem@davemloft.net> Date: Wed, 11 Nov 2015 13:04:07 +0000 In-Reply-To: (=?iso-8859-1?Q?=22M=E5n?= =?iso-8859-1?Q?s_Rullg=E5rd=22's?= message of "Wed, 11 Nov 2015 12:22:28 +0000") Message-ID: User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.5 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Måns Rullgård writes: > David Miller writes: > >> From: Måns Rullgård >> Date: Wed, 11 Nov 2015 00:40:09 +0000 >> >>> When the DMA complete interrupt arrives, the next chain should be >>> kicked off as quickly as possible, and I don't see why that would >>> benefit from being done in napi context. >> >> NAPI isn't about low latency, it's about fairness and interrupt >> mitigation. >> >> You probably don't even realize that all of the TX SKB freeing you do >> in the hardware interrupt handler end up being actually processed by a >> scheduled software interrupt anyways. >> >> So you are gaining almost nothing by not doing TX completion in NAPI >> context, whereas by doing so you would be gaining a lot including >> more simplified locking or even the ability to do no locking at all. > > TX completion is separate from restarting the DMA, and moving that to > NAPI may well be a good idea. Should I simply napi_schedule() if the > hardware indicates TX is complete and do the cleanup in the NAPI poll > function? I tried that, and throughput (as measured by iperf3) dropped by 2%. Maybe I did something wrong. -- Måns Rullgård mans@mansr.com