In article <Pine.LNX.4.43.0302262253450.7617-100000@pilchuck.reedmedia.net>,
Jeremy C. Reed <reed@reedmedia.net> wrote:
>Christos Zoulas suggested:
>> # Avoid pcmcia bus space conflicts with the default io space
>> # allocation on 12-bit wide busses (base 0x400 size 0xfff).
>> options PCIC_ISA_ALLOC_IOBASE=0x300
>> options PCIC_ISA_ALLOC_IOSIZE=0x0ff
>
>So I read about this in the pcmcia(4) and options(4) man pages.
>
>And I implemented that and my new kernel did:
>
> ne2 at pcmcia1 function 0 port 0x300-0x30f port 0x310-ox31f
> ne2: Linksys EtherFast PCM100 Ethernet
> ne2: Ethernet address 01:d4:ff:03:00:20
>
>But still different hardware address. And UDP and ICMP (dns and pings)
>work, but TCP (ftp and http) times out (which was same as before).
>
>It seems strange to me to have two port ranges. I am still researching
>this.
>
>The working Linux:
>>> eth0: NE2000 (DL10022 rev 05): io 0x300, irq 3, hw_addr 00:04:5A:08:78:E1
>
>The working OpenBSD:
>>>  ne3 at pcmcia1 function 0 "Linksys, EtherFast 10/100 Integrated PC Card
>>>  (PCM100), Ver 1.0" port 0x340/32: irq 9
>>>  ne3: address 00:04:5a:08:78:e1
>>>  nsphynter0 at ne3 phy 5: DP83815 10/100 integrated, rev. 3
>
>Thank you, Christos, for the reply.
>

Looks like the OpenBSD driver is if_ne.c is much more intelligent handling
these cards. You should pick up what they do from there.

christos