From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Leblond Subject: Re: [PATCH RFC 3/3] NFQUEUE: add --queue-cpu-fanout parameter Date: Sat, 23 Mar 2013 22:53:49 +0100 Message-ID: <1364075629.4349.11.camel@tiger2> References: <20130319141442.150005875@eitzenberger.org> <20130319141606.304161536@eitzenberger.org> <1363703652.30419.11.camel@tiger2> <20130323195256.GN13505@imap.eitzenberger.org> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netfilter-devel To: Holger Eitzenberger Return-path: Received: from ks28632.kimsufi.com ([91.121.96.152]:33282 "EHLO ks28632.kimsufi.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752081Ab3CWVx6 (ORCPT ); Sat, 23 Mar 2013 17:53:58 -0400 In-Reply-To: <20130323195256.GN13505@imap.eitzenberger.org> Sender: netfilter-devel-owner@vger.kernel.org List-ID: Hi, On Sat, 2013-03-23 at 20:52 +0100, Holger Eitzenberger wrote: > Hi Eric, > > > If possible, it could be interesting to be able to setup the balance > > parameter by using an option in the same way fail-open option: > > uint32_t flags = NFQA_CFG_F_FAIL_OPEN; > > uint32_t mask = NFQA_CFG_F_FAIL_OPEN; > > int r = nfq_set_queue_flags(qh, mask, flags); > > This way, it is possible to tune the system without changing the > > ruleset. > > Not sure how the FAIL_OPEN relates to the CPU fanout. Do you want to > setup the CPU fanout (on, off) per queue? You are right on a major point. FAIL_OPEN is definitely per-queue and CPU fanout is related to a fanout group. So we will have to have the whole mechanism bind to a queue, warn the kernel that it is in a fanout group... That could be interesting as we will be able to tune the setting from the application but this seems to be a lot of work too. And I don't see any functional benefit omit this tuning from the application. So, if you don't find any other one, forgot my request for now :) BR, -- Eric Leblond Blog: https://home.regit.org/