On Thu, 2006-08-31 at 12:20 -0500, Klaus Weidner wrote:
> On Thu, Aug 31, 2006 at 08:57:05AM -0400, Paul Moore wrote:
> > Klaus Weidner wrote:
> > > I was a bit surprised that a "s2-s2" process can connect successfully to
> > > a "s3-s3" process, send it data, and select/poll(2) waiting for data.
> [...]
> > > It works as expected the other way around, the s3 process gets an
> > > immediate "connection refused" when trying to connect to the s2 process.
> >
> > This smells like a policy issue to me. Taking a quick look at the
> > reference policy mls constraints (this is from a svn snapshot which is
> > probably a month or two old, so it may have changed) I see the following
> > constraint (unrelated types removed for clarity):
>
> It's a bit more complex now:
>
> # the socket "read" ops (note the check is dominance of the low level)
> mlsconstrain { socket tcp_socket udp_socket rawip_socket netlink_socket
> packet_socket key_socket unix_stream_socket unix_dgram_socket
> netlink_route_socket netlink_firewall_socket netlink_tcpdiag_socket
> netlink_nflog_socket netlink_xfrm_socket netlink_selinux_socket
> netlink_audit_socket netlink_ip6fw_socket netlink_dnrt_socket } { read
> getattr listen accept getopt recvfrom recv_msg }
> (( l1 dom l2 ) or
> (( t1 == mlsnetreadtoclr ) and ( h1 dom l2 )) or
> ( t1 == mlsnetread ));
>
> # the socket "write" ops
> mlsconstrain { socket tcp_socket udp_socket rawip_socket netlink_socket
> packet_socket key_socket unix_stream_socket unix_dgram_socket
> netlink_route_socket netlink_firewall_socket netlink_tcpdiag_socket
> netlink_nflog_socket netlink_xfrm_socket netlink_selinux_socket
> netlink_audit_socket netlink_ip6fw_socket netlink_dnrt_socket } { write
> setattr relabelfrom connect setopt shutdown }
> ((( l1 dom l2 ) and ( l1 domby h2 )) or
> (( t1 == mlsnetwritetoclr ) and ( h1 dom l2 ) and ( l1 domby l2 )) or
> ( t1 == mlsnetwrite ));
>
> Based on that policy, I'd expect the constraint on "connect" to prevent a
> l1=s2,h1=s2 process from connecting to a l2=s2,h2=s2 port. Am I
> misunderstanding something or is the check mixed up?
connect is a process-to-socket check handled entirely in the socket
layer, i.e. can process A initiate a connection via socket S (where S is
its local endpoint, not the peer). It isn't a check on the peer.
recv_msg is likely the one of interest to you for NetLabel. You could
separate out the mls constraint on tcp_socket to require equivalence for
recv_msg in that case (overloading with old net controls is bad though).
If secid reconciliation progresses fully and secmark is enabled, then
packet recv would be the check of interest.
--
Stephen Smalley
National Security Agency
--
redhat-lspp mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/redhat-lspp
