On Thu, Dec 04, 2025 at 07:49:51PM +0300, Alexander Mukhin wrote:
> >Synopsis: eigrpd: multiple issues
> >Category: user
> >Environment:
> System : OpenBSD 7.8
> Details : OpenBSD 7.8 (GENERIC) #54: Sun Oct 12 12:45:58 MDT 2025
>
> [email protected]:/usr/src/sys/arch/amd64/compile/GENERIC
>
> Architecture: OpenBSD.amd64
> Machine : amd64
> >Description:
> eigrpd: multiple issues
> >How-To-Repeat:
> At first, eigrp daemon exited with error just after start:
>
> # cat eigrpd.conf
> address-family ipv4 {
> autonomous-system 1 {
> interface vio1
> interface vio2
> }
> }
> # eigrpd -d -v -f eigrpd.conf
> startup
> ...
> fatal in eigrpe: in_cksum: packet too big
> ...
> waiting for children to terminate
> terminating
>
> I found that in_cksum() is called with uninitialized data
> because of swapped source and destination in ibuf_from_ibuf().
> I fixed that. Then the daemon started but was unable to make
> an adjacency with a Cisco router. Cisco complained on wrong
> checksum in incoming EIGRP packets. In fact, the checksums were
> in wrong byte order. I added a call to htons(). Then eigrpd and
> Cisco made an adjacency and exchanged routes.
>
> Unfortunately, redistribution of static routes didn't work.
> It looks like children processes get configuration with
> empty redist_list despite of that list being correctly filled
> in the master process. I was unable to fix this issue.
> >Fix:
> (partial fix)
> --- packet.c
> +++ packet.c
> @@ -175,7 +175,7 @@ send_packet(struct eigrp_iface *ei, struct nbr *nbr,
> uint32_t flags,
> rtp_ack_stop_timer(nbr);
> }
>
> - ibuf_from_ibuf(buf, &ebuf);
> + ibuf_from_ibuf(&ebuf, buf);
Doh! I think that's the 2nd ibuf_from_ibuf I messed up.
> if (ibuf_get(&ebuf, &eigrp_hdr, sizeof(eigrp_hdr)) == -1)
> fatalx("send_packet: get hdr failed");
>
> @@ -189,7 +189,7 @@ send_packet(struct eigrp_iface *ei, struct nbr *nbr,
> uint32_t flags,
> if (ibuf_set_n16(buf, offsetof(struct eigrp_hdr, chksum), 0) == -1)
> fatalx("send_packet: set of chksum failed");
> if (ibuf_set_n16(buf, offsetof(struct eigrp_hdr, chksum),
> - in_cksum(ibuf_data(buf), ibuf_size(buf))) == -1)
> + htons(in_cksum(ibuf_data(buf), ibuf_size(buf)))) == -1)
Here you could just use ibuf_set_h16() instead of ibuf_set_n16() to
prevent the value to be double converted.
> fatalx("send_packet: set of chksum failed");
>
> /* log packet being sent */
>
Thanks for finding these issues. It is tricky for me to test eigrpd since
I have no cisco gear to test against.
--
:wq Claudio
