Subject: Strange hanging problem
To: None <port-mac68k@NetBSD.ORG>
From: Monroe Williams <monroe@teleport.com>
List: port-mac68k
Date: 03/12/1996 16:54:00
About a month ago, my NetBSD/mac68k machine started hanging pretty
reliably during moderate disk activity.  Since it's not a production
machine (i.e. I don't do much work on it), I haven't taken the time to
completely isolate the problem, but I have spent some time poking
around in the kernel debugger and tried a number of different kernels.
I have a bunch of details on the state of the machine after the crash
(kernel traces, etc.) that I can post or mail out individually if it
seems pertinent.

I strongly suspect that there may be hardware problems involved.  I
noticed the problem soon after I lost the machine's internal drive to
sticktion, and moved the external Quantum Fireball 730 (with all the
NetBSD stuff on it) to the internal bay.  So far, the problem has only
manifested while building the kernel and various binaries from the
NetBSD source tree.  I have an active terminator attached with a couple
of activity lights on it (send and rcv, or something similar), and
whenever the hang happens one of the lights stays lit.  (During normal
operation, they usually appear to flash in sync with each other.)  Once
the machine has hung, I can get to the kernel debugger using the
interrupt button, and it always seems to be somewhere in the scsi
interrupt handling code (AFAICT).

I have reinstalled the world from one of the -current binary snapshots,
and have tried a bunch of different kernels, including the one from the
1.1 distribution, ones I built myself (between crashes), -current
snapshots from eskimo, and the one Allen built recently with linked
commands disabled.  They all have the problem sooner or later.

If anyone out there recognizes this problem as definitely hardware, bad
scsi bus karma, etc., please let me know.  (i.e. something like "oh,
bad SCSI termination would make it do that" would be a big help.)  It's
really annoying, and I'm not quite sure where to start looking.  If
kernel traces or other info that can be gathered with the kernel
debugger would be useful, I can provide such.

Thanks,
-- monroe
------------------------------------------------------------------------
Monroe Williams                                      monroe@teleport.com