Feng Xiagen wrote;

> 1.It cannot recognize the cable media at all.
> ...
> 2.IP setting action will shut you link state DOWN for ever

There are large number of ARM boards equipped with SMsC 91C111
or CS8900A in market.  They are claimed as a 'Non-PCI/Local Bus'
Ethernet controllers, however, they are rebadgged legacy ISA devices
in fact (imposing offset 0x300 as if ISA I/O space makes no sense
memory mapped device scheme, look at the number of 'address' pins).

The issue of 91C111 is the after-thought MII/internal PHY upgrade to
old ISA design HW.  I could manage to run 91C111 hooked to a certain
ARM9 SoC after enormous hardship.  If I had a chance to do it again, I
would choose LAN9118 which has re-designed modern FIFO architecture
from ground.  Tx logic is scattered mbuf friendly and Rx ptr can be
offset to absorb STRICT_ALIGNMENT.  Many of ARM9 SoC DMA circuits
look like each other and NetBSD would make a good fit. 

> at sm0 phy 0: Seeq 84220 10/100 media interface, rev. 0
> sqphy0: using Seeq 84220 isolate/reset hack
> sqphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto 

In your case it looks PHY is never controlled well.  Considerable
amount effort would be necessary around MII/PHY control; probing
pins to see _real_ electorical activities, or inserting ASSERTIONs make
sure SW logic does the right things as expected.

Toru Nishimura/ALKYL Technology