[OpenAFS-port-darwin] callbacks hanging on Intel Macs?

[Nicholas Riley]

We've been having some issues with files mysteriously not appearing
recently.  It turns out to be pretty reproducible on our Intel Macs
with 1.4.2fc4, and not reproducible on corresponding PowerPC Macs (or
our other-OS machines, but they're running different client versions).
The server in question is running 1.4.2fc4 on Solaris 10, and I've
been able to reproduce it with a couple of different volumes on that
server.

Compare (in all cases, the file doesn't exist to start with):

intel% touch foo
ppc% ls foo
foo

with:

ppc% touch bar
touch: bar: No such file or directory
# should this happen?
ppc% touch bar
intel% ls bar
ls: bar: No such file or directory
% fs flushv
% ls bar
bar

We've also been seeing some mysterious hangs on the Intel Macs, often
on login, which I now think might be related to this problem:

ppc% touch baz
intel% time touch baz
touch baz  0.00s user 0.02s system 0% cpu 7:15.30 total

If I run cmdebug against the machine while it's hanging like that, I see:

** Cache entry @ 0x3b2eb7e4 for 89.536871243.1.1 [acm.uiuc.edu]
    locks: (writer_waiting, write_locked(pid:1262 at:135), 1 waiters)
           10240 bytes  DV       563000  refcnt     1
    callback 041e2784   expires 1160601804
    1 opens     0 writers
    volume root
    states (0x1), stat'd

It's possible to unhang the Intel Mac by doing a 'rm baz' on another
machine.

afsd is running on both PowerPC and Intel machines as:

root       491   0.0 -0.0    27616    272  ??  Us   10:15AM   0:00.03 /usr/sbin/afsd -afsdb -stat 10000 -dcache 2500 -daemons 5 -volumes 70 -dynroot -fakestat-all

Has anyone seen similar things?  Any suggestions on how to debug
further?  I'll be getting another Intel Mac to test with soon, but for
now we've only got two identically configured iMacs, so I'm not sure
if this is a local configuration problem or a bug.

Thanks,

-- 
Nicholas Riley <njriley@uiuc.edu> | <http://www.uiuc.edu/ph/www/njriley>