From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pablo Neira Ayuso Subject: Re: [ULOGD PATCH 08/14] Add state option to NFLOG input plugin. Date: Wed, 16 Apr 2008 13:39:18 +0200 Message-ID: <4805E566.7010902@netfilter.org> References: <1206289522679-git-send-email-eric@inl.fr> <12062895233806-git-send-email-eric@inl.fr> <47F79683.8070801@netfilter.org> <20080405191512.GJ5333@bayen.regit.org> <47FAB4C4.8050908@netfilter.org> <20080410045648.GA5526@bayen.regit.org> <20080413070324.GK20469@bayen.regit.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit To: Eric Leblond , netfilter-devel@vger.kernel.org Return-path: Received: from mail.us.es ([193.147.175.20]:53984 "EHLO us.es" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1752063AbYDPLj0 (ORCPT ); Wed, 16 Apr 2008 07:39:26 -0400 In-Reply-To: <20080413070324.GK20469@bayen.regit.org> Sender: netfilter-devel-owner@vger.kernel.org List-ID: Eric Leblond wrote: > Hello, > > On Thursday, 2008 April 10 at 6:56:48 +0200, Eric Leblond wrote: >> Hello, >> >> On Tuesday, 2008 April 8 at 1:56:52 +0200, Pablo Neira Ayuso wrote: >>> Eric Leblond wrote: >>>> On Saturday, 2008 April 5 at 17:10:59 +0200, Pablo Neira Ayuso wrote: >>>>> Eric Leblond wrote: >>>> 'decision' keyword is not really a good choice. >>> I see, however, why isn't --nflog-prefix enough to label the logs from >>> iptables? >> Yes, but it forces user to define a system to be able to know if >> decision is DROP or ACCEPT. For example, all prefix have to be labelled >> like [DA]:$MYSTRING. Well, it works but it overload prefix which is less >> human readable. Furthermore, it steals some bits in the prefix field >> which is in ULOG rather small. > > Not receiving a response to this mail make me thing about what I will > need to do to have the same feature without the 'state' flag: > > A standard logging packet logging table contains packet that have been > dropped or accepted. One of the most relevant question for an > administrator is: "Which packet have been dropped with this criteria ?". > Thus, we need an easy and *fast* mean to differentiate dropped and > aceepted packet. IMHO, a 'state' flag is one of the easiest and cleanest > solution. The computation overload is really small compares to a parsing > of the prefix. Furthermore, this approach is not disruptive with > firewall rules generator who already have their log prefix. I see. It makes sense. Please, could you resend this patch using 'label' instead of 'state' which seems more to me more appropriate for this feature? -- "Los honestos son inadaptados sociales" -- Les Luthiers