At 18:53 Uhr +0100 12.12.1998, Space Case wrote:
>On Dec 12,  5:07am, Frederick Bruckman wrote:
>>On Fri, 11 Dec 1998, Space Case wrote:
>>> The reason I ask, is that my "new" Q650, running GENERIC-88 (built
>>> October 26) is handling two disks just fine...
>>
>>Lucky you! The problem with my Q630 started early in September. Does
>>anyone know what's different about a Q650?
>
>Just as an update, I'm now running a kernel built yesterday from Dec. 10
>sources, still accessing two disks with no problems.

I'll try that (just downloading; link is pretty bad from here).

>Please refresh my memory...  When did this lossage start?  Is it limited
>to the Q700?

Well... I have seen variants of this problem (kernel panics during the
first time a disk is accessed) ever since I got my 230M MO drive, which was
about three years ago. Since some time after 1.2, the first access to that
MO drive panics the kernel on a 53c80 based machine 4 out of five times,
even if I bring down the machine to single user. Somewhere after 1.3, i
gave up entirely on using the MO with 53c80 machines (IIci, IIsi, SE/30).
Instead, I hooked it to the Q700 (53c96 based) which was a bit more
tolerant, and nfs-mounted the media from there.

I recall problems with first access to hard disks, too, but far less often.

When I got larger disks, which had their own set of problems
(mac68k/{5174,5214}), I put the MO with all its problems away.

Other PRs to look up in this (esp) context are mac68k/6411 (Frederick
Bruckman) and mac68k/6549 (SUNAGAWA Keiki). I agree with SUNAGAWA Keiki's
analysis that crashes become more frequent with increasing interrupt rate,
esp. tty and net interrupts.

Sigh. As I see it, we have two people on mac68k who are capable of tracking
this can of worms down, and they lack the time, the hardware on which the
panics happen, or both. I do not see this thing getting fixed anytime soon,
and I still find the SCSI framework to be way over my head.

	hauke


--
"It's never straight up and down"     (DEVO)