Implement TCP server Automaton #3690
Conversation
As a child class of TCP_client, server automatically completes a TCP handshake when a client connects and is integrated with supersocket. Anything that could be removed due to parent class was removed. States were required to stay due to Automaton design.
Codecov Report
@@ Coverage Diff @@
## master #3690 +/- ##
==========================================
- Coverage 86.23% 86.16% -0.07%
==========================================
Files 296 296
Lines 66907 66987 +80
==========================================
+ Hits 57698 57722 +24
- Misses 9209 9265 +56
|
|
Thanks for your PR. Looks good, so far. Could you please provide some unit-tests? |
There was a problem hiding this comment.
Thanks a lot for the PR & your interest in Scapy !
This is an interesting topic and it's really nice to see what you did.
Unfortunately though, I believe it would need a bit of refactoring: we want to support multiple clients, so you probably would need to implement TCBs of some sort. There is a bit of refactoring to do :p
|
|
||
| @ATMT.action(incoming_data_received) | ||
| def receive_data(self, pkt): | ||
| super().receive_data(pkt) |
There was a problem hiding this comment.
This won't work on Python 2 (super(TCP_server, self)). Did you test it? same elsewhere
(sorry this should be the last release to support it :p)
| self.l4 = IP(src=ip) / TCP(sport=self.sport, flags=0, | ||
| seq=random.randrange(0, 2**32)) | ||
| self.src = self.l4.src | ||
| self.sack = self.l4[TCP].ack |
There was a problem hiding this comment.
This is a bit cheap: you handled only a single client at the time. I must say I envisioned something a bit more useful, handling multiple streams.
Did you have a look at Transmission Control Block?
|
I will continue this soon, been out of town for a bit so please don't close |
|
Still working on this @jabbate19 ? |
|
Closing for inactivity. |
As a child class of TCP_client, new TCP_server automatically completes a TCP handshake when a client connects and is integrated with supersocket. Inspiration was taken from #2409
Anything that could be removed due to parent class was removed. States were required to stay due to Automaton design.
While working on this, I found/fixed a bug within TCP_client that would cause the ACK number of a response from data to be wrong when a small amount of data was sent. Padding was being considered in the length for the ack number when it is not in the protocol.
Part of #399