[Bioclusters] OS X -> Linux NFS mount problems

[Josh Goodman]

	Has anyone experienced NFS problems when mounting a directory from
OS X to a linux box?  We have a G5 server running 10.3.5 and exporting a
disk that resides on our xServe raid array.  The raid is connected to the
G5 via fibre.  Our linux boxes are SuSe 9.1 and RedHat AS3.  In cases
where the directory contains a thousand or more directories and files,
doing an ls -lR on the linux box in the mounted directory causes ls to
enter into an infinite loop.  As it loops it continues to consume memory
until there is no more left and it promptly dies saying memory exhausted.
Directories with smaller numbers of files are fine.

	An ethereal log of tcp traffic shows NFS calls going back and
forth with no obvious problems that I can see (not that I am an expert at
analyzing traffic).  An strace of the ls spits out this over and over
again

lstat64("./ana/ana00193.html", {st_mode=S_IFREG|0664, st_size=11422, ...})
= 0
getxattr("./ana/ana00193.html", "system.posix_acl_access", (nil), 0) = -1
EOPNOTSUPP (Operation not supported)

The file name changes as the ls chugs on but it eventually gets stuck
looping over files in a directory where ls has hung.

Our solaris 8 box and another G5 xserve are happily mounting and reading
from the same mount point without any trouble.  After talking on the phone
with Apple they found a few indications of other people having similar
problems and raised its importance a bit higher but they offered no
solution or workaround.

There is a bug report that is possibly related here:
http://sources.redhat.com/bugzilla/show_bug.cgi?id=109

Things I have tried include forcing nfs over tcp, using nolock, turning
posix on/off, adjusting name lengths, and r/w sizes.  Nothing works.

Since I know several folks on this list have OS X servers I thought I
might get lucky and find someone who ran into this and was more
successful than I at finding the solution.

Thanks,
Josh Goodman

Department of Biology
Indiana University