> You're right and I'd forgotten about that fix in the current driver
> we have, the 590 and 595 (and the eisa variants whose numbers I've
> forgotten) require the thresholds to be in longwords.

Oh, duh! I read this in the comments in the Linux driver a little
while back, and wondered how it could be backward compatible if
they did that. :-) (The reason they did this is that there's an
11-bit field that the packet length is put in to, and the newer
cards now support 4.5K FDDI-size packets.)

I'll try this out today.

In other news, my 3c59x documentation is supposed to be arriving
on Monday or Tuesday.  (I'm on holidays next week, so I'll have
lots of time to play, but I'll have to go back into the office if
I have to physically touch the hardware. This is a dilemma. :-))

The 3c90x documentation, however, is not available, and at this
point the 3Com person in charge of handing this stuff out thinks
the engineers aren't going to have it ready for her until the end
of the year. The reason that this is important is that the 3c900
uses a somewhat different (and faster) method of bus-master DMA
transfers, if you run it in DMA mode. I also hear that the next
stepping of the 3c900 chip will not support PIO mode any more,
which means we will have to use a different driver for it.  (The
Linux driver is extremely well documented, so I can give that a
close reading and see how much it gives me to go on.)

Of course, I don't know anything about PCI DMA, so now there's all
that to learn....

cjs

Curt Sampson    curt@portal.ca		Info at http://www.portal.ca/
Internet Portal Services, Inc.	
Vancouver, BC   (604) 257-9400		De gustibus, aut bene aut nihil.