Re: [lwip-users] Lwip 1.3.0 Running out of TCP PCB's fast

[Jonathan Larmour]

Pettinato, Jim wrote:
> This is due to pcbs being placed in a TIME_WAIT state; i.e. when you
> close a connection they're not really freed yet, they're hanging around
> waiting for enough time to pass to make sure no stray packets come in
> and screw up a future connection. Note that the default 'enough time' is
> like two hours, so an automatically refreshing web page will eat up pcbs
> very quickly. (This is TCP standard behavior as defined in the RFQs,
> though it makes little sense in today's high-speed networks.)
> 
> Even though the error count gets incremented, lwIP will (upon finding no
> free pcbs) allocate a pcb from the TIME_WAIT list and use it, so this
> may not actually be a problem for your implementation.

Ahh, while I knew about TIME_WAIT and lwIP reusing them, I didn't remember
it would increment the error count anyway, and I assumed B B was saying an
allocation had failed, corroborated by the error count. So yes, hopefully B
B was only speculating and not reporting a problem.

Jifl
-- 
eCosCentric Limited      http://www.eCosCentric.com/     The eCos experts
Barnwell House, Barnwell Drive, Cambridge, UK.       Tel: +44 1223 245571
Registered in England and Wales: Reg No 4422071.
------["Si fractum non sit, noli id reficere"]------       Opinions==mine