Hello folks.  I have a funny problem I've never seen before.
I'm running NetBSD/i386 with native endian filesystems and a 1.5R kernel
with 1.5 binaries.  All works well, except on one of the filesystems, where
/sbin/dump goes into an endless loop reading one of the blocks of the
filesystem.  Fsck on the filesystem shows no errors, and tar can
successfully read all files on the filesystem.  Running from the filesystem
causes no ill effects, and using tar to dump the contents of the filesystem
to another filesystem doesn't reproduce the symptom on the new filesystem.
All other filesystems on the system dump just fine.
	I'd like to send-pr this problem, but don't know how gnats will
enjoy an attachment containing the binary image of the bad filesystem,
since I'm sure that's what it will take to reproduce this problem on another
machine.  
	Anyone ever seen this problem before?  Anyone know the internals
of /sbin/dump well enough to tell me where I might set a break point that
would be useful if I watched this under gdb?  It gets to phase III, dumping
directories.  It looks to me like it gets stuck reading the beginning of
one directory over and over again.

Any suggestions greatly appreciated.
-thanks
-Brian