Re: [lwip-users] webserver problem

[address@hidden]

Wilson, Dave (Stellaris S/W) wrote:
> I took a look at the trace you posted. Aside from all the IP header checksum 
> errors in the requests sent from the browser (which appear to be ignored),

Such ignored bad checksums most often come from wireshark monitoring a 
(windows-) network card which has checksum-offloading enabled. If you 
would monitor the wire, the checksums would probably OK.

>   I don't see anything particularly strange. Perhaps someone better versed in 
> low level TCP protocol can offer more help, though.
>
>    The request for background.png is in packet number 207. The HTTP server 
> acknowledges this in packet 209 and sends 107 bytes of image data in packet 
> 210. The client acknowledges this in packet 223 and closes the connection.

Now this already is the problem: in HTTP/1.0 (which the browser says it 
is), the *server* closes the connection, not the client (since only the 
server knows when the file is finished). Since the client closed the 
connection, you can assume the browser detected some kind of problem and 
decided to not wait for data any longer.

>   I have no idea how large background.png is but I suspect the connection is 
> closed before all the data has been sent since WireShark reports that it can't 
> decode the PNG that was sent in response to the GET request.

It is closed too early, indeed, or else there would be a FIN from the 
server.
Also, what's strange is that the browser sends a RST packet in response 
to FIN from the server (packet #199).

And still another strange thing is that the server sends a RST after 
establishing a connection (packet #235). This and the fact the 
index.html is only sent in very small pieces leads me to think that you 
might be *very* limited in resources (i.e. running out of memory 
somewhere). You might want to increase memory settings or check the 
stats to allow more simultaneous connections and more data in flight.

Apart from that, I think the browser aborts receiving the 2 PNG files 
because it detected some kind of error in the HTML page: it aborts those 
connections directly after completely receiving the HTML. Maybe there's 
a format error in the HTML? At least, http://validator.w3.org/ reports a 
lot of errors, which could be the reason the browser aborted reception.

I would try to create a very small HTML that does not fail validation 
and if that works, fix the validation errors in your current page.

>    By the way, for faster response to Stellaris-related questions, you may be 
> better to post them to the Stellaris support forum on e2e.ti.com since this is 
> monitored by our support folks all the time. Only a few of us watch this 
> mailing list and questions sometimes slip by!
I'm happy to have these webserver-related questions on the list here, as 
(thanks to you:), the Stellaris-webserver is very much the same as lwIP's.

Simon