kern/113874: netatalk broken on big-endian systems
nathanw at ginger.rh.uchicago.edu
Wed Jun 20 03:30:03 UTC 2007
>Synopsis: netatalk broken on big-endian systems
>Arrival-Date: Wed Jun 20 03:30:02 GMT 2007
>Originator: Nathan Whitehorn
University of Chicago
FreeBSD sunrise.rh 7.0-CURRENT FreeBSD 7.0-CURRENT #7: Tue Jun 19 22:04:43 CDT 2007 root at mail.rh:/usr/obj/sparc64/usr/home/nathanw/src/sys/US3 sparc64
Same behavior on:
FreeBSD avila 7.0-CURRENT FreeBSD 7.0-CURRENT #4: Tue Jun 19 21:11:23 CDT 2007 root at mail.rh:/usr/obj/arm/usr/home/nathanw/src/sys/AVILA arm
I'm not entirely sure whether this is a kernel or netatalk port bug, but at least some of the problem is in the kernel AARP layer, and they have a close relationship.
The ports part:
On big-endian systems, atalkd will not cleanly restart. If it has previously run on an interface, it says something like this:
sunrise# /usr/local/etc/rc.d/netatalk restart
bge0: Address already in use
Can't configure multicast.
atalkd: zero interfaces, exiting.
This is with no remaining atalk address on the interface.
The kernel part:
The kernel seems incapable of correctly sending or receiving AARP packets. Aecho doesn't work, nbplkup doesn't work, and tcpdump gives a lot of these packets coming from the problem machine:
22:54:19.872587 aarp who-has 0.0.0 tell 0.0
Turning on NETATALKDEBUG, I get this at the console when these packets are sent (sending induced by aecho 65280.125, which is a mac. 65280.175 is the machine itself):
aarp: sending request for 65280.175
The machine also doesn't reply to AARP packets addressed to it.
More information about the freebsd-bugs