After spending a day and a half debugging the 1.4 kernel on my 4/260, I've
   come to the conclusion that nobody ever tried 1.4 on a Sun4 using a VME si
   device.

      si_attach() does not bootpath_store() before calling si_attach_common(),
   which, further down the line, causes dk_establish() to try to use the bp
   entry for the si, which results in a bogus SCSI target of -1 which causes
   a memory violation when it is used as an array index. After fixing that,
   I had to fix device_register(), which didn't recognize "si" as a controller
   and was nail_bootdev()ing it.

      It looks like the second problem would occur with the 4/3xx onboard si
   device and the 4/1xx sw as well, so maybe nobody ever tried 1.4 on a sun4
   of any kind.

      Anyway, after finally getting it to recognize and be happy with its
   boot device, it goes out to lunch trying to close the SCSI disk after sizing
   it, sitting forever in scsipi_wait_drain().

      I give up. I originally selected NetBSD because it has the best Sun-3
   support and I was hoping to have all my Suns run the same OS. But I don't
   have time for this kind of grief. If I wanted to write the OS myself, I'd
   be doing it for my Sun-2's.

      --James



I can't speak for the Sun 3, but the current si/sw driver on my 4/110 does
not function with versions around 1.4C.  I think that I filed a PR about
this [DMA problems].

Things did work pretty well in the 1.3 [or 1.2] days.  [I have not tried
with a more recent kernel].



Brad Spencer - brad@anduin.eldar.org   http://anduin.eldar.org
[finger brad@anduin.eldar.org for PGP public key]