On Tue, 28 Apr 1998, Allen Briggs wrote:

> > I've had a different problem with one drive that started about the same
> > time that "(quick)" started showing up in dmesg. Any access to the drive
> > (dd, disklabel) drops into the debugger, lately only after printing all
> > the requested output to the console.
> 
> Really?  This is the first I've heard about any such thing.  Can you
> send me what it prints as it goes into the debugger?  Presumably you get
> more than just (eg):
> 
> # disklabel sd3
> ddb>
> 
> There should be a panic or a diagnostic message of some sort.

Yes, of course. First of all, you get whatever output you ask for: an
entire disklabel, dump, dd, whatever. I tried

# dd if=/dev/sd2 of=/dev/null count=1000

and it accessed the drive for quit a while before it finally panicked.
Here what I got just now after `disklabel sd2'

[some detail omitted]

disklabel: super block size 0
vm_fault(0xf1688, 854000, 1,0) -> 1
   type 8, code [mmu,,ssw]. 1505
trap type 8, code 0x1505, v=0x854c14
Kernel: MMU fault trap
pid=795, pc=000035BA, ...

[more detail omitted]

panic: MMU fault
Stopped at _Debugger+0x6: unlk a6
db> t
_Debugger [omitted]
_panic
_trap
faultstkadj(2, 0, 0) + 0
fault() + c

Now I tried

> ps

and the machine froze solid, requiring a press of the "panic button" in
the back. Typically if I try to step through, it will dump core and
reboot, or I can just type reboot, and it will do that immediately.

> > My question is, is there a simple way to disable (quick)?  All I have
in
> > my kernel config is
> 
> In esp.c, set quick to 0 and skip the checks where it would get set to 1.
> I'd much rather try to fix it, though, than disable it.  It at least
> doubles the transfer speed.

I might try that, just for grins. I have only Mac Partitions on that drive
now, anyway. Here are the relavant snippets from dmesg:

NetBSD 1.3E (FB) #28: Sun Apr 26 22:55:48 CDT 1998
    fredb@fb.sa.enteract.com:/usr/src/sys/arch/mac68k/compile/FB
Apple Macintosh Quadra 630  (68040)

[It's actually a Performa 636CD, + a  real 040 chip.]

esp0 at obio0 (quick): address 0x906000: NCR53C96, 16MHz, SCSI ID 7
scsibus0 at esp0: 8 targets
cd0 at scsibus0 targ 3 lun 0: <MATSHITA, CD-ROM CR-8004A, 2.0a> SCSI2
5/cdrom removable
sd0 at scsibus0 targ 4 lun 0: <QUANTUM, FIREBALL1080S, 1Q09> SCSI2
0/direct fixed
sd0: 1042MB, 3835 cyl, 4 head, 139 sec, 512 bytes/sect x 2134305 sectors
sd1 at scsibus0 targ 5 lun 0: <NOMAI, MCD 540I, 5.55> SCSI2 0/direct
removable
sd1: 515MB, 4144 cyl, 2 head, 127 sec, 512 bytes/sect x 1055416 sectors
sd2 at scsibus0 targ 6 lun 0: <QUANTUM, FIREBALL_TM3200S, 300X> SCSI2
0/direct fixed
sd2: 3067MB, 6810 cyl, 5 head, 184 sec, 512 bytes/sect x 6281856 sectors

The bad boy is actually an Arriva brand 3.2G. I have rearranged the SCSI
cables, experimented with the termination on the internal CD-ROM, and
reformatted everything several times in the past year or so. Makes no
difference. It has also made no difference when there was a NetBSD
partition on that drive, or no partititions, or if it was the only
external drive. (Not all at once. I tried to boot from it as the only SCSI
drive.) I recently consolidated partitions and deleted the FWB bogus zero
length partitions by hand, with Norton's, so that disklabel now won't
complain about having more than eight, again to no avail. Everything is
hunky-dory under MacOS 8.1, except that, well, it's not NetBSD :).