Hi list,

since I have some time to spend between the years, here are some -current
data points wrt. SCSI and my dratted Fujitsu MO drive:


1) The 53c96 esp driver on a 33 MHz Q700

All is working well. :)

I.e., so far I have not seen any kernel crashes besides the usual unionfs
related ones (a 'make build' dies during the 'make includes' somewhere in
usr.bin). I have built two kernels - a 'make clean && make depend && make'
takes about 50 min vs. 2 1/2 hrs on an SE/30 - and copied a dir tree of
about 60 MB to MO while running the /etc/daily script. But then, without
any ethernet or serial interrupt related activities...

Anyway, this surely feels like a good piece of work.


2) The 53c80 ncrscsi driver on a SE/30

A few months ago, kernels with a ncrscsi driver used to die miserably in
/etc/rc with all kinds of strange errors; this has changed somewhat for me.
It appears that the ncrscsi driver is able to read from my root drive (a
Quantum LPS105S) just fine. I built a kernel without any problems (sources
on a MO mounted read-only to /mo, /mo union-mounted to /usr/src, then
/var/obj union-mounted on top of /usr/src) so the ncrscsi driver seems to
be happy with the Quantum TRB850 which holds /usr and /var.

It is definitely not happy with the LPS 105 when it comes to writing: I
copied the new kernel to /netbsd and rebooted, but the booter would not
accept the kernel. The last number displayed in the booter (symbols?) was
around 40 MBytes...
I booted with another kernel and copied the new kernel again, just to see
ls -l issue error messages about two kernel images. A reboot to single user
and an 'fsck -f /' showed that half a dozen files on / had gone down the
drain; luckily I had backups.

Still some way to go here...


Comments?


	hauke

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