arla-0.33 wedges NetBSD-1.4.1 machines

[Dr. Lex Wennmacher]

Hi,

we are experiencing problems with arla-0.33 on NetBSD-1.4.1. I have installed
the kth-krb4-1.0.1, arla-0.33, and xdm-krb4-3.3.6.1 from the NetBSD packages
collection.

Arla is used to act as an AFS client for the NetBSD/i386 machines. The home
directories reside in the AFS filesystem and are read protected
(system:anyuser l). When arla tries to access a user's home directory
without ticket, it tends to hang, instead of returning errors.

After some period of flawless operation the machine starts to wedge.
This happens usually, but not always, at night, when /etc/daily is run.
Eventually, no file access is possible, not even on local file systems.

Here is what *I think* happens: The command `calendar -a' is invoced from
/etc/daily. This command su's to every user and tries to read the user's
calendar file. Not having a ticket this fails, but instead of returning EPERM
the open() system call in calendar blocks.

>From that point on, subsequent file I/O starts to block and eventually not even
an `ls -l /' will return and the machine has to be rebooted.

I have debugging output available for the situation described above, but given
it's size (several MB) I don't include it in this e-mail. In any case, the last
messages that are logged before the machine is totally wedged are:

Jul  6 09:53:17 othello /netbsd: xfs_dnlc_lookup: v_id = 5860934
Jul  6 09:53:17 othello /netbsd: xfs_dnlc_lookup: calling cache_lookup:dvp =
fa0
19270, cnp = (.geo.uni-koeln.de/usr/wennmacher//bin, 17), flags = 84044
Jul  6 09:53:17 othello /netbsd: xfs_dnlc_lookup: cache_lookup returned. error
=
 0, *res = 0
Jul  6 09:53:17 othello /netbsd: xfs_message_rpc opcode = 4
Jul  6 09:53:17 othello /netbsd: xfs_alloc: xfs_allocs - xfs_frees 518
Jul  6 09:53:17 othello /netbsd: xfs_alloc: xfs_allocs - xfs_frees 519
Jul  6 09:53:17 othello /netbsd: xfs_alloc: xfs_allocs - xfs_frees 520
Jul  6 09:53:17 othello /netbsd: xfs_statfs: mp = f0566200, sbp = f0566240,
proc
 = fa0ba890
Jul  6 09:53:20 othello /netbsd: xfs_sync: mp = f0566200, waitfor = 2, cred =
f0
4e4480, proc = f9e62af8
Jul  6 09:53:20 othello /netbsd: xfs_root mp = f0566200
Jul  6 09:53:20 othello /netbsd: xfs_lock: fa019270, flags 0x10002
Jul  6 09:53:20 othello /netbsd: xfs_lock before: lk flags: 1024 share: 0 wait:
0 excl: 1 holder: 19089
Jul  6 09:53:50 othello /netbsd: xfs_sync: mp = f0566200, waitfor = 2, cred =
f0
4e4480, proc = f9e62af8
Jul  6 09:54:20 othello /netbsd: xfs_sync: mp = f0566200, waitfor = 2, cred =
f0
4e4480, proc = f9e62af8
Jul  6 09:56:20 othello last message repeated 4 times


Cheers

Lex