On Thu, 16 Mar 2000, Bob Nestor wrote:

> Jason R Thorpe  (thorpej@zembu.com) wrote:
> 
> >On Thu, Mar 16, 2000 at 06:31:21AM -0600, Bob Nestor wrote:
> >
> > > I've been running the latest 1.4.2 snapshot on my 68k Mac building 
> > > packages.  It's uptime is about 1 week but last nite it crashed with the 
> > > "lockmgr: locking against myself".  I thought that problem was fixed some 
> > > time ago.  Is it only in -current and not in 1.4.2, or am I wrong about 
> > > it being fixed?
> >
> >What is the call traceback?  Important to know who's calling lockmgr(), as
> >it's just a tool used by many other parts of the kernel.
> 
> I hit the reboot button before I was completely awake, but I'll copy it 
> down next time (if it happens again).  At first glance it looked like all 
> the other ones I'd seen previously on this port.

Coincidently, I've experienced a few similar crashes while building
packages these last two weeks. I've managed to collect a couple of
core dumps, but no useful stack traces. I viewed the traces before
rebooting, but didn't write them down: one ended in pkg_create, and
the other was in some other disk intensive executable.

Most of the evidence points to the SCSI hardware/cables. Two "panic:
lockmgr:..." during or shortly after heavy disk I/O (as above), one
"getblk: block size invariant failed" (with most tasks waiting on
"inode"), and some esp0 timeouts, also during heavy disk I/O (but not
always a crash there).

Got one MMU fault on start-up: never seen that before (heat?); and one
hard freeze earlier today, shortly after deleting my pkgobj tree. Now
it's been up for 12 hours, light use (X and some makes).

Like Bob, I have a Quadra 840 AV (128mb, 2disks: 500mb, 10gb). It was
remarkably stable during most of the 1.4.2 cycle, until recently.