From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexandru Dragoi Subject: Re: ipt_owner and ICMP Date: Sun, 19 Mar 2006 02:22:22 +0200 Message-ID: <441CA43E.1040706@zoomnet.ro> References: <004801c64aba$9e81b170$0301a8c0@asp> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <004801c64aba$9e81b170$0301a8c0@asp> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: netfilter-bounces@lists.netfilter.org Errors-To: netfilter-bounces@lists.netfilter.org Content-Type: text/plain; charset="us-ascii" To: jay@esuna.co.uk Cc: netfilter@lists.netfilter.org jay@esuna.co.uk wrote: >Hi, > >I'm currently using the ipt_owner module to enforce stronger outgoing packet >filtering on certain daemons. I create a custom chain with the stronger >rules and use '-m owner' to jump packets into the chain. > >This works fine for UDP and TCP, but my outgoing ICMP packets never match >the rule. I understand why incoming ICMP should fail to match, but why are >outgoing packets missing the filter? > >Chain OUTPUT (policy ACCEPT) >target prot opt source destination >acctboth all -- anywhere anywhere >ACCEPT all -- anywhere anywhere >apache-output all -- anywhere anywhere OWNER UID >match > iptest > >Chain apache-output (1 references) >target prot opt source destination >DROP icmp -- anywhere anywhere > >(nothing in the acctboth chain causes a jump) > >Any ideas? > > > I think this is because icmp packets are just generated and sent away by some part of the kernel after it received a syscall from a program with uid 0 (only root can use icmp).