Problems with 2.0.30

Dan Merillat (dan@merillat.org)
Wed, 06 Aug 1997 01:17:17 -0400


And I don't see any diffs in any pre-patches that address this.

I can hard lock 2.0.30 (and 2.0.29) by doing heavy IO to a
block device hanging off a Future Domain 16xx controller (compiled
as a module) I.E. NFS exporting a ZIP drive or an external SCSI drive.

The problem is that it's a cli/while 1 type lock... no errors, no
debugging, just a solid halt. I can do this in single user, by mke2fs
on a 2gig external drive... (Actually, I usually do this in single user.
fscking 10 gig of drives with too many partitions isn't fun. Espically
in buffer-hungry .30 ;-)

Interestingly, I can run a 14 gig news server off 2.0.30 (+atime) without
causing buffer starvation. Wierd, eh? Of course, it has 128 meg memory,
so it has enough to fsck when it boots.

I can't init MD disks, but I think I saw a patch for that go by.
(instant OOPS + lock because md couldn't get enough memory) I'll have to
try that when I get some extra 2gig disks freed up. Hmm...

Network wise, I havn't seen any problems. A few spurrus syn flood
messages in syslog, but no strange hangs. I'll try pre-3 to see if that
fixes the FD lockups, but I don't see anything that might affect it in
there.

--Dan