On Sun, Apr 24, 2011 at 12:49:46PM +0100, Stuart Henderson wrote:
> On 2011/04/24 10:42, Claudio Jeker wrote:
> > On Sat, Apr 23, 2011 at 08:58:12PM +0000, Stuart Henderson wrote:
> > > On 2011-04-23, Henning Brauer <lists-openbsdpo...@bsws.de> wrote:
> > > > * Stuart Henderson <s...@spacehopper.org> [2011-04-23 14:41]:
> > > >> Should we do this or should we rethink allowing sockets to be
> > > >> bound to broadcast IP addresses?
> > > >
> > > > i have no idea. and that seems to be common for all of us. i really
> > > > dunno. anybody?
> > > >
> > >
> > > Unless there are objections I'll commit my samba diff tomorrow,
> > > it can always come out again if anything else changes.
> >
> > What is this socket used for? You can not send a packet out of that
> > socket. I guess it is a UDP socket that should pickup the broadcast hellos
> > that flood windows infested networks.
>
> Yes, exactly.
>
> > Our network stack will deliver broadcast datagram to all IP RAW and all
> > UDP dgram sockets that match and are bound to 0.0.0.0 or ::.
> >
> > /*
> > * Deliver a multicast or broadcast datagram to *all* sockets
> > * for which the local and remote addresses and ports match
> > * those of the incoming datagram. This allows more than
> > * one process to receive multi/broadcasts on the same port.
> >
> > Now there is this bit of code in the loop as well:
> > if (inp->inp_laddr.s_addr != INADDR_ANY) {
> > if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
> > continue;
> > }
> >
> > So it may make sense to bind to a specific broadcast addr for SOCK_DGRAM
> > and SOCK_RAW AF_INET socks. In that case you only receive packets that are
> > sent to the network specific broadcast (this does not work for
> > packets sent to 255.255.255.255 on that LAN segment)
>
> That would look something like this horror..
>
> Index: in_pcb.c
> ===================================================================
> RCS file: /cvs/src/sys/netinet/in_pcb.c,v
> retrieving revision 1.116
> diff -u -p -r1.116 in_pcb.c
> --- in_pcb.c 19 Apr 2011 03:47:29 -0000 1.116
> +++ in_pcb.c 24 Apr 2011 11:31:56 -0000
> @@ -276,8 +276,12 @@ in_pcbbind(v, nam, p)
> } else if (sin->sin_addr.s_addr != INADDR_ANY) {
> sin->sin_port = 0; /* yech... */
> if (!(so->so_options & SO_BINDANY) &&
> - in_iawithaddr(sin->sin_addr,
> - inp->inp_rtableid) == NULL)
> + (ifa_ifwithaddr(sintosa(&sin),
> + inp->inp_rtableid) == NULL) &&
> + (so->so_type != SOCK_DGRAM) &&
> + (so->so_type != SOCK_RAW) &&
> + (in_iawithaddr(sin->sin_addr,
> + inp->inp_rtableid) == NULL))
> return (EADDRNOTAVAIL);
> }
> if (lport) {
I would prefer to split this monstrosity into multiple if statements.
Since this is totaly unreadable but I guess we want something like this.
We could call in_broadcast(sin->sin_addr, NULL) instead of in_iawithaddr()
to make the code more clear.
--
:wq Claudio
