On Friday 14 December 2007 2:25:12 pm Christopher J. PeBenito wrote: > On Thu, 2007-12-13 at 10:45 -0500, Paul Moore wrote: > > On Thursday 13 December 2007 9:12:08 am Christopher J. PeBenito wrote: > > > On Wed, 2007-12-12 at 15:18 -0500, Paul Moore wrote: > > > > Assuming labeled networking is enabled, a forwarded packet would > > > > hit four checks: > > > > > > > > # inbound checks > > > > allow netif_t peer_t:peer ingress; > > > > allow netnode_t peer_t:peer ingress; > > > > # outbound checks > > > > allow netif_t peer_t:peer egress; > > > > allow netnode_t peer_t:peer egress; > > > > > > This helps. But this seems to be for the old networking, how does it > > > work with the secmark stuff? > > > > It doesn't work with the SECMARK stuff, or rather it works in parallel > > with the SECMARK stuff. We've debated integrating the peer labeling > > protocols (labeled IPsec, NetLabel) with the SECMARK mechanism many > > times but in the end we always end up deciding it doesn't make sense. > > So, with compat_net off, you'd still need the above policy, not the > packet type against the peer type? The compat_net setting has no effect on the peer object class permissions. The only thing that will cause the any of the access controls above (the ones I listed) to go into affect is the netpeer policy capability bit/flag. > ... not this: > > allow ssh_client_packet_t peer_t:peer egress; There should never be a check between the iptables/secmark label, "ssh_client_packet_t", and the network peer label, "peer_t". Ever. Well, unless I screwed up the code somewhere ;) -- paul moore linux security @ hp -- This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to majordomo@xxxxxxxxxxxxx with the words "unsubscribe selinux" without quotes as the message.
