[En-Nut-Discussion] TCP stops working after some time

[Harald Kipp]

Ummmphhh!!! Well, I had to find out, that the latest resets were
caused by our unstable test hardware setup.

What made me wonder was the fact, that they appeared around
or even inside the trace output, regardless where they were
placed. It's now obvious: The CPU spent most time in these
routines, so the hardware failure probably occured during
debug output. That had been very confusing while searching
for a software problem, which was actually a problem caused
by the hardware.

The assumption, that something might be wrong with processing
segments received in advance couldn't be confirmed. We are
now dropping every 10th segments, which causes a large amount
of packets received in advance. The stack seems to handle this
without problems.

Thus, it looks like the latest changes committed to CVS HEAD
did indeed fix the initial problem. Can anyone else confirm
this?

Harald