Erik E. Fair wrote:
> It was so many months ago that I installed on the Netra T1 AC200
> and the Sun Fire V120, that I'm don't recall if I booted from the
> NetBSD 3.1 CD-ROM, or if that didn't work and I set up an install
> by netbooting from my house NFS server (gotta love the Suns for
> supporting easy netboot from firmware since ... ever). The nice
> thing about netbooting is that if the INSTALL kernel doesn't work,
> you can always try GENERIC, and once you have one going, it's easy
> to use hot swap SCA SCSI disks for cloning more.

In my case, I *am* netbooting, since I'm too damn lazy to burn CDs when
the machine is connected to the 'net anyway.  GENERIC (at least the one
from -current, haven't gone back to the other versions), dies the same way.

BTW, the NetBSD ofwboot.net is *so* nice compared to what I remember it
being, and compared to the OpenBSD one (for example).  If it just zapped
all it's internal state between failed boot attempts, life would be
grand (it seems to remember state gathered from the bootp/dhcp server,
and in fact a second attempt at netbooting in the same instance of
ofwboot.net succeeded even though my DHCP server wasn't running due to a
dhpcd.conf typo).

> It is worth noting that I also tried this kernel on the Sun Blade
> 150 workstation (which is also apparently another variant of this
> same motherboard, with some additional devices (e.g. IEEE1394
> FireWire) configured in), and that failed to boot with the same
> sort of IDE missing interrupt notices and ultimately a freeze. I
> haven't tried booting that system since; busy with other things,
> and hoping someone else would fix the lossage.

Well, I'm going to dig into it some, since I'd like to run NetBSD on
these systems, rather than Sol10 or OpenBSD, and it works just fine on
the V100... not that I'm the best guy for the job, but I do have the
incentive ;)

The key difference between the V100 and AC200 hardware config appears to
be that on the V100 the PCI bus with all the built-in devices hangs
right off the psycho, whereas on the AC200 the IDE and SCSI controllers
are on subordinate buses (pci1 and pci2, respectively), which are behind
two Simba PCI bridges.  Maybe there's some missing interrupt swizzling
or un-swizzling or some extraneous interrupt mangling happening on the
Simba PCI bridges.

That, however, doesn't explain why it works on your AC200 and not on mine.

> We still have other work to do for complete support of these systems,
> in particular: [...stuff...]

That would be cake... all those same things aren't configured on the
V100 and that's happy as a pig in sh*t (well, modulo the crappy IDE
controller...  but it still works quite well).

I've just done a hand diff of your dmesg output and here are the key
differences (noting that I'm booting a 4.99.36 kernel, not 3.1_STABLE):

EEF> 2048MB memory
RKB<  512MB memory

EEF> IOTSB: 21a0000 to 2220000
RKB< IOTSB:  854000 to 8d4000

EEF> aceride0: bus-master DMA support present
EEF> aceride0: primary channel configured to native-PCI mode

(not in my config... I presume this is just due to version differences)

EEF> esiop0: interrupting at ivec 20
RKB< esiop0: interrupting at ivec 1820

EEF> esiop1: interrupting at ivec 20
RKB< esiop0: interrupting at ivec 1820

(snip'd Eric's BGE)

--rafal