In message <Pine.HPP.3.93l.961005191453.8025A-100000@unix26.andrew.cmu.edu>
Andrew Y Ng writes:
>
>well, the 3c590 driver will definitely freeze.  I was just *guessing* that
>all ep*, turned out that I was wrong.

The 3c595 and the 3c590 use the same driver. The 3Com documentation
claims they're substantially the same hardware (Vortex).   The same
is true of the EISA version.    (I don't know about the 3c900).

It's somewhat suggestive that this bug only seems to show up at
10Mbit operation, which is all I've seen reported, to date.

It could be  related to media-selection; or  perhaps an problem
when the board gets overrun with data.


>I have a 10BaseT connection. 
>> 	* Is port-i386 really the right place for bugs with PCI boards?
>
>hmmm... it's the driver that has problems, not the board, imho.

I didn't mean to pin the blame on the board versus the driver.
(I meant ``bugs with NetBSD using PCI boards'').

But either way, it's not really an i386 problem, which was my point.

If the 1.2/386 port had interrupt counters that worked, this
would be a *whole* lot easier to debug...