Re: [lwip-users] Bug on lwIP TCP section (or HTTP client from contrib)?

[Enrico Murador - Research & Development - CET]

Dear Sergio,

I agree you can't see any problem from the logs, because in theory, that 401 header should be followed (at least in my case) by an HTTP message body, so an RST from the server would be expected. This somewhat "hides" the issue I'm trying to point out.
The TCP stack sends RST "instead of" sending the 401 HTTP message body, so there should be no unacked data at that point. As I wrote on my first email, looking with the debugger I've seen that TCP sends RST because there is only 1 byte of unacked data; if the RST had been sent because the client interrupted the transfer, I should have found N bytes of unacked data (where N = size of 401 message body) that should have triggered the RST reply.
At the time TCP decides to send the RST, that message body has not yet been sent, because the HTTP server called http_eof() (that in turn called tcp_rst()) before sending the last packet, to "add the FIN flag right into the last data segment", as you can read in the code.

P.S.
What wrong checksums? Are you referring to the capture file I posted on November 02, or to the last one I posted on November 11?

Thanks
Enrico

On 18/11/2015 14:14, Sergio R. Caprile wrote:

Sorry Enrico, I don't see any problem (other than the wrong checksums)
in your capture file. To me the sequence numbers are correct, and the
RST is correct as per the described client behaviour.
Ciao