On 8 Jul 2003, Frank McConnell wrote:
I am pleased to be able to say that I no longer
remember the details,
but the problem was something like that the card had enough room in
its buffer to hold no more than one received packet. If you sent it
another one before the processor on its end got around to reading the
packet out of it, one packet or the other got dropped (I'm thinking
the one in the buffer got dropped so that the new one could be
buffered).
Indeed, the SEEQ chipset (...) could only do one thing at a time, so
they sized the buffers accordingly. The Etherlink II had more buffer
space, but was crippled in the same way: it was single-tasked.
If someone else out there remembers more, please do
feel free to jump in.
I dont wanna ;)
--fred (who co-wrote that damn driver in linux, AND debugged it to get
it a LITTLE bit working..)