>> Now I have another problem with a IBM Credit Card Ethernet Adapter.
>> 
>> Here is the Dmesg:
>> 
>> ...
>> 
>> I've modified if_ne_pcmcia.c(/sys/dev/pcmcia) to add my card's info as
>> 
>> /*
>>    { "National Semiconductor InfoMover",
>>      0x00a4, 0x0002, NULL, NULL, 0,
>>      0x0ff0, { 0x08, 0x00, 0x5a } },
>> */
>>    { "IBM CCAE",
>>      0x00a4, 0x0002, "IBM Corp.", "Ethernet", 0,
>>      0x0ff0, { 0x00, 0x04, 0xac } }, 
>> 
>> The "National Semecondution InfoMover" is commented, or the if_ne_pcmcia
>> will recognize my card as it first (because the same Manufactor code and
>> productor code), and complain : "0x00,0x04,0xac" should be "0x08,0x00,
>> 0x5a".
>> 
>> >From the dmesg's "ne0: where did the card go?" , I traced the ne2000.c
>> code(/sys/dev/ic), and found my card failed at:

The problem is that IBM and/or National Semiconductor are sharing the
same CIS, but are each allocating cards out of their own bit of the
ethenet MAC space.  This won't be too hard to fix, but for now, your
workaround is fine.

As for why the RST bit is clear in the ISR, I don't have any good
idea.  One very slim possibility is a timing bug I'm still tracking
down.  Perry, I'd run this by Jason and/or Charles.

		Marc