This is not related to OpenSolaris, except to the extent that the same
symptoms most likely would happen on an OS release if the same problem
occurred.  I'm looking for someone who understands the depths of the IP
stack... I'm happy to take this off-list and post a summary if it's
solved and the solution is relevant to the list.

On bootup of this Solaris 10 03/05 system, the network/physical service
comes up in maintenance, with errors in
/etc/svc/volatile/network-physical:default.log
configuring IPv4 interfaces: ce0.
ifconfig: socket: lo0: Address family not supported by protocol family
ifconfig: socket: lo0: Address family not supported by protocol family

and errors in network-loopback:default.log
/sbin/routeadm: unable to open /dev/ip: Protocol not supported

trussing "ifconfig -a4uD" gives me some systemcall failures related to
socket calls:
so_socket(PF_INET, SOCK_DGRAM, IPPROTO_IP, "", SOV_XPG4_2) Err#124
EAFNOSUPPORT
so_socket(PF_INET, SOCK_DGRAM, IPPROTO_IP, "/dev/udp", SOV_XPG4_2) = 3

So there's something apparently going on under the default socket path
(in XPG4 compatibility mode) that's failing, that is not failing when
it's being passed through the UDP socket path.

Interestingly, if I boot the system with a reconfiguration reboot, the
network/physical service fails initially but can be cleared by svcadm.
However in this state, all route commands hang in an uninterruptable
(even by kill -9) state.

So I'm guessing there's something corrupted in the code path for
sockets, but I'm not really sure where to go to track this down.  Any
suggestions?

Thanks,
--Joe
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to