In message <200505020332.j423Wafe000376@morgart.com>,
"William S. Morgart" writes:

>I have looked at the forcedeth driver but decided that Quinton's work
>on FreeBSD was easier to port initially - I had nVidia hardware that I
>wanted to get NetBSD fully operational on. Now looking at the
>forcedeth driver is possible since I have networking up.

Sure, that makes sense to me.

[...]


>I don't know ... the rlphy.c file is what is in FreeBSD and I only
>changed what was necessary to get it to work in NetBSD ... So checking
>if the parent is an rl device could be completely useless in
>NetBSD. Its also code that's not reached in the rlphy_match function
>since the check for realtek 8201 phy at the beginning should always
>match for the nvidia hardware thus the other two checks in that
>function could be deleted.

I'd have to look closer, then. But if you want to import it, and
displace/replace the existing rtk PHY support, then the new driver
should handle all cases cleanly.

>Jonathan> Also, if we did import this, we should rip out the
>Jonathan> corresponding code from the 81x9 driver (we don't want to
>Jonathan> duplicate it).
>
>It would probably be more appropriate to add to the 81x9 phy driver
>support for the 8139 ... but I haven't looked at whether that is
>sensible or not.


Again, I'd have to look closer, and compare to what Bill Paul did in
FreeBSD. I re-checked the FreeBSD CVSweb logs right after your initial
message; it looks like FreeBSD split out the 8139?() rtk PHY into a
separate driver from what corresponds to our sys/dev/ic/r*81x9.c.