Those are the different address/data width combinations that are
supported by the kernel and hardware. The VMEbus supports D8 and
D24, but your board doesn't. VME64 also supports A64/D64 and others.

The BGNT and IACK lines daisy-chain through the backplane. All of the
jumpers should be installed unless the board in that slot can be
a bus master or can generate an interrupt. The VMEbus supports
four levels of Bus Request/Bus Grant. Sun only supports one level,
I believe BG3.

If you have a standard VMEbus backplane then you can put the
SCSI/Ethernet board next to the CPU and forget the jumpers.
If you leave the SCSI/Ethernet board where it is then install jumpers
where you remove boards.

NetBSD expects the boards at standard addresses. I will give 1.4.1
Miniroot and DISKLESS a try tonight on my 3E.


At 09:35 PM 10/19/99 +0100, Ian Wells wrote:
>Michael Thompson writes:
> > In PCspeak the plug-and-play BIOS enumerates the PCI bus when it
> > pokes at known addresses to see if a board is installed and to
> > figure out what is installed there.
> > 
> > A UNIX kernel does the same thing when it boots by looking for
> > onboard I/O and VMEbus I/O at known addresses.
> > 
> > It is also called a discovery process.
>
>The relevant messages are:
>
>vme0 at mainbus0: (A16/D16)
>vme1 at mainbus0: (A16/D32)
>vme2 at mainbus0: (A24/D16)
>vme3 at mainbus0: (A24/D32)
>vme4 at mainbus0: (A32/D16)
>vme5 at mainbus0: (A32/D32)
>
>But, looking at this and thinking back over what I've heard about VME, 
>I would guess these are the VME address spaces that devices can appear 
>in and I've just misinterpreted them, haven't I...
>
>Anyway, the 'spurious level 2 interrupt' answer would appear to have
>to do with the VME backplane's jumpers.  I now have a vague idea about 
>what they do, and I assume that one of the cards I removed should have 
>been replaced by a jumper.  Replacing the original cards makes it work 
>with the SunOS 3.5 install.
>
>By the way, whoever said I should have a P2 bridge between the memory
>and the processor is wrong.  In fact, the memory is there for the use
>of the dedicated hardware, and is used over the VME bus.  It's not
>part of the Sun's core memory.
>
>Anyway: does NetBSD support a Sun3/E SCSI controller by default?  The
>address is 310000, which, admittedly, isn't the same as usual, but is
>exactly right for a Sun3/E, according to my manual.  If it doesn't,
>then it would be since if the sun3 page mentioned it.  I should point
>out I haven't tried a sun3 + netbsd boot yet, so I can't say if the
>jumper/card fix has changed matters.
>
>
>-- 
>Ian.
>
>

Michael Thompson
E-Mail: M_Thompson@IDS.net