[En-Nut-Discussion] Nutcracker

[Harald Kipp]

Bengan,

I strongly disbelieve, that this is a stack
problem. We are using UDP in some systems,
which are running 24h/7d without problems.

We also made several tests with large packets
and packet storms without being able to crash
the system.

Are you using version 3.3.2 (or at least 3.3.1)?

Are you using the latest CVS? This hadn't been
tested extensively.

I assume, that the problem is caused by a bug
in your application code. Are you running out
of memory?

_BUT_, everything is possible and you may indeed
have discovered a bug. Of course I'm very interested
in your modifications. If you send them to me,
I can add them to the Ethernut project webpage.

Thanks,

Harald


At 09:24 12.01.2004 +0100, you wrote:
>I experienced crashes in Nut/OS under following circumstances:
>
>1. 100 ingoing UDP pkts/s. Packetsize ~100B.
>2. 100 outgoing UDP broadcasts/s. Packetsize ~100B.
>3. 1 TCP/IP session, mostly inactive.
>4. 1 ping session, 1 ICMP/s.
>
>This crashes system after at most some minutes.
>I found the problem is related to RTL8019 overflow, so I modified this to be
>much faster.
>After that system is stable but sometimes stops communicating. After
>stopping and restarting
>ping session system is running normal again.
>
>Any clue?
>Do you want motifications for RTL driver?
>
>/bengan