I've been -- and continue to be -- really busy with work, so I hope
you'll excuse my slow responses.

Alex Dumitriu <alexrd@earthlink.net> writes:
> For my own edification and education, what does "only supports memory mode"
> mean exactly? (If this is a can of worms you don't have the time or
> inclination to open, that's fine).

The cs8900 has several different possible modes of operation.

In one, which is normally used on PCs with "I/O" accesses, it provides a
relatively small number of ports and most of the internal registers
are accessed indirectly via index and data registers.  I.e. to access
register 'foo', you write 'foo' to the index register then read or
write the data register.

In the other, which is normally used on PCs with "memory" accesses, it
provides a larger range of ports and direct access to the internal
registers using them.  I.e., to access foo, you can do a normal memory
access (read or write, to pointer + offset) to the location where the
'foo' register lives in memory space.

The driver currently requires the latter, and from your ethernet's OFW
properties i'm not convinced that your hardware supports it.  Whether
or not the hardware supports it really is governed by the wiring of
the hardware.


> I have no serial port on this machine, but I'll copy what I need by hand. I
> assume it's not just the /vlbus/isa/ethernet@i300 section that is relevant.
> What else do I need?
> (FWIW, here's the ethernet stuff nonetheless):
> ****    /vlbus/isa/ethernet@i300
> interrupts  0000000b
>             00000003
> reg         00000001    00000300    00000010
> compatible    CS8900
> model         CS8900
> device_type     network
> name            ethernet

If the OFW node accurately describes the hardware -- and it probably
does -- that means that your chip only supports I/O-based accesses.

I.e., it'll need the kernel driver to be updated, in order to
function.  (It doesn't seem to have any DMA-related properties -- I
dunno if that's because DMA isn't hooked up, or because the node's
properties are incomplete.  I'd believe either.)

Unfortunately, I don't have the cycles to even attempt that at this
point.


I'd strongly recommend _against_ trying to update the OFW in this
machine; based on the information you've presented, i'd guess that the
hardware design _is_ slightly different than the Shark's.  That's not
surprising, since we published the schematics and companies building
boards were certainly free to modify them if they chose.

It sounds like you've had some issues in trying to upgrade the
firmware, too, which leads me to believe that there are even more
hardware differences.  (of course, substitution of flash parts
shouldn't be that big a deal... except when programming them.  8-)


> BTW Chris, Romanian?

Nope.  8-)



chris