[lwip-users] Re: [lwip] delayed tcp_write() with ray stack interface ?

[Kieran Mansley]

On Wed, 27 Nov 2002, Jimi Simpson wrote:

> Hi again,
>
> When I perform a tcp_write() to a remote server I find that the actual
> <PSH, ACK> (observed using ethereal) does not occur until I receive some
> data into the client application. For you information I am confident
> that the tcp fast and slow timers are being called and when data is
> received into the client application I observe a successful transmission
> to the server and a consequent <ACK>. I have spent a fair few hours now
> trying to track down what I believe to be a problem, but I am now
> questioning my own sanity. Is it the case that in the lwip
> implementation that the tcp_write does not get processed until data is
> received or am I missing something else?

Could it be that the TCP window has been exceeded, and so the client can
not send more data until some of the stuff "in flight" has been
acknowledged.  This would result in writes stalling (I think) until some
data was received.

Kieran

[This message was sent through the lwip discussion list.]