Re: strange ne2000ish card on A1200

[is%netbsd.org@localhost]

Hi Karoly,

On Thu, Mar 26, 2020 at 01:53:13AM +0100, Karoly Balogh (Charlie/SGR) wrote:
> 
> On Mon, 23 Mar 2020, is%netbsd.org@localhost wrote:
> 
> > hehe. The reason is, that as far as I can tell, cnet.device doesn't do
> > any check on anything card-specific. If a card function type is provided
> > as network card, it believes it. If no card function type is provided,
> > it believes it is a network card, too.
> 
> Yeah. No huge surprise there. :)
> 
> > or add the three line marked with +, without the plusses, after line
> > 183 in netbsd-current; then recompile the kernel.
> 
> I tried it with netbsd-7, as that's what I had at hand. Is it a problem
> and I should try current? Sadly, it didn't work. The card is now bound as
> ne0, but then it cannot proceed further:
> 
> pcmcia0 at pccard0
> ne0 at pcmcia0 function 0: <PCMCIA, Fast-Ethernet, , >

That's good, so the match function works, and there isn't any invisible
goo in the CIS strings.

> ne0: where did the card go?

:-(

Does anything interesting happen if you take it out and put it in again?
(Maybe do such experiments in single user mode first, without r/w mounted
file systems.)

> I can try over the WE with netbsd-current, if necessary.

I think 7 vs. current shouldn' make a difference, but: must be 
latest 7 tree, *with* the 16bit patch for Amiga PCMCIA. Do you
have that? Your file

src/sys/arch/amiga/dev/gayle_pcmcia.c

should have the version

1.32 or later (-current a.k.a. -9.99.x, -10 or later)
1.31.22.1 or later (netbsd-9, -9.0 is not enough)
1.31.10.1 or later (netbsd-8, -8.1 is not enough)
1.29.4.3 or later (-7; the -7.2 is not enough)

Maybe PCMCIAVERBOSE would be a help here, too.

Else we can take a look at the card_info .bin file.

I also suspect we might need to set some flag for the 100mbit/s
variants (AXIS 99... chipsets) somehwere - I had hoped this would
be detected by querying the hardware. More driver code reading.

Regards,
	-is