I've visited Toshiba again today, and booted -current boot floppies as of
yesterday on their 4G-RAM 4*Xeon 550MHz machine. The only modification to
the floppies was that the kernel had DDB_ONPANIC=1, but that wasn't even
necessary - the machine booted straight into sysinst!

Dmesg output can be found at
http://www.feyrer.de/Misc/4G-toshiba/1/dmesg.out, /kern/physmem is at
http://www.feyrer.de/Misc/4G-toshiba/1/kern_physmem (in case this
matters).

Winner? Almost.
Besides the lot of RAM & extra CPUs, the machines is equipped with a token
ring card, and three Symbios Logic SCSI controllers - one 53c810 with only
a CDROM connected, and two 53c896, with one of them containing a working
SCO Unix system on ID #0. I don't know if it matters, the BIOS
memory check prints 4193280k RAM found (plus 640k), and 2048k Cache SDRAM.

Letting aside the fact that the total memory printed is slightly off,
there are still some problems. The biggest one is that the machine seems
to recognize a "wd0" when exchanging the harddisk at SCSI ID #0 (replacing
SCO with a scratch disk) at one of the two 53c896 controllers (note the
ST506?), and then gives timeouts from that.

See http://www.feyrer.de/Misc/4G-toshiba/2/dmesg.out for /kern/msgbuf of
that kernel.

I was told the machine has no IDE disks in it, so I have no clue where
the wd0 comes from.

Also, I don't see anything that looks like the token ring card and esp.
the two 53c896 SCSI controllers, configured or not. The CDROM on the
53c810 is found correctly.

I've checked that the INSTALL kernel enables ncr* and all PCI bridges, so
I don't relaly know how to go from there.

Any ideas? Does anyone see anything that could hide the two 53c896
controllers? Maybe the fact that the machine has several CPUs and no IRQ
routing code is used makes it not detect several devices? Do we support
the 53c896? Or is this still in Manuel's queue? 

BTW the system was at least stable enough to run sysinst once it's up (and
over the timeouts), and I was able to more /kern/msgbuf and poke around a
bit, so installing a system on the scratch disk I got would be the next
stop. Unfortunately I cannot connect that disk to the 53c810.


 - Hubert

-- 
NetBSD - because Unix isn't just #include <linux.h>, i386, ELF, ...!