Hello,

Have you (or anyone else) looked at this since March? I'm having some
problems that *might* be related. What I have is a 559-based cardbus
card that I have trouble getting up to speed in outgoing mode. Running
ttcp from another host to my laptop works fine (I get a few MB/s w/o
tuning), but ttcp from my laptop yields << 1MB/s.

It works w/o problem if I change from 100 to 10 Mb/s (I get about
9.5Mb/s bandwidth), as does ping -s 1400 -f (in 100 Mb/s mode, also
about 1MB/s). What doesn't work is TCP to this machine in 100Mb/s
mode.

If I patch myself into the irq-vector, I can see that the card doesn't
produce any interrupts for a many packets.  What I see is a burst of
packets, and then a long delay, and then more packets.

I can think of the following things not working:

* the card can be broken in some way (not very likely since it works
  fine in 10Mb/s mode, and windows doesn't seem to have any problem
  with it)

* the card fails to generate an interrupt for some packets (a
  hardware is also unlikely, but I guess it could be a software
  problem)

* the interrupt is for some reason not passed through to the os
  (since my setup involves the more or less experimental cardbus
  patches, this might not be completely unlikely)

* the os drops the interrupt on the floor (also very unlikely)

One of the reasons I think my problem might be cardbus software rather
than something fxp specific is that the modem shows similar behaviour
(although I really haven't looked into that problem) - it seems to
loose packets when running ppp.

Did you have the same problems in both directions? The `hang' you
talked about, was it a real hang, or just the result of a very lossy
tcp connection?

/Johan