From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: ipt_LOG and friends. Question Date: Thu, 24 Aug 2006 18:14:53 +0200 Message-ID: <44EDD07D.1020805@trash.net> References: <20060824140320.GJ31235@kriss.csbnet.se> <20060824151923.GK31235@kriss.csbnet.se> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Cc: netfilter-devel@lists.netfilter.org Return-path: To: Joakim Axelsson In-Reply-To: <20060824151923.GK31235@kriss.csbnet.se> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: netfilter-devel-bounces@lists.netfilter.org Errors-To: netfilter-devel-bounces@lists.netfilter.org List-Id: netfilter-devel.vger.kernel.org Joakim Axelsson wrote: > The default 29-32 chars as a prefix is far too little for me. Needs to > change to 128. But again i see some code allocating this on the stack. It > this a problem? We can't change it due to binary compatibility. In your local tree you can simply increase both sizes (not on the stack of course). > The code says that a maximum of 830 chars can be used (ICMP-pcaket with > one recursion level). Are you sure? That sounds a bit high. > 128 char prefix and some more on IN, OUT, PHYSIN/OUT > and MAC fields will give an overall figure of around 1024 chars a good > buffer length. Anything above that can probably be cut. It won't happen > alot. > > Another idea of not allocating on the stack nor using GFP_ATOMIC is to have > a global buffer per CPU. Is this a good idea with the hot adding and > removing of CPUs? Are linux kernel guarantied to continue on the same CPU if > an interupt occurs and so on? That seems overkill, one global buffer protected by a lock will do fine.