Re: [lwip-users] tcp_abandon

[Bill Auerbach]

>> Ok then, why isn't it public?  How do I abandon a tcp_pcb without any
>> further activity on it and releasing it ASAP.  This is a real-time
>> embedded system - loss of link is fatal.
>
>I guess because it hasn't been needed, yet. Why do your need to abandon
>a pcb ASAP if you lost the link? I would have thought you had plenty of
>time to do that until the link returns...
>
>Simon

If data gets slowed enough for whatever reason and it's not at the device
when it needs to be, the system has to be stopped.  Usually this happens
when the user tries to run too fast even without system bottlenecks (like
from his system setup, the PC or the network).  In this case we stop because
he's run out of data (Underflow).  We are in very noisy places with multiple
large motors, heaters, UV lights and we do get flaky behavior at times.  But
we don't want flakiness where a short loss of link recovers and sometimes it
doesn't underflow and sometimes it does underflow.  We have to debug
completely remotely and the errors we can catch and control we have to.
With loss of link being fatal, we stop but we want the PC to reconnect as
quickly as possible and not leave anything taking up resources from the
previous connection. On the next connect I pass back that the previous
connection lost link so at least I can report why the system was stopped.
If we have flaky wiring or a flaky switch causing loss of link we want to
know about it. Once we had to replace a switch that was causing problems.

Thanks,
Bill