Wireshark-dev: [Wireshark-dev] Changing dissector based on TCP options
From: Edwin Groothuis <wireshark@xxxxxxxxxxx>
Date: Mon, 30 Aug 2010 21:33:08 +1000
Hello, I have been fighting with find_conversation(), conversation_new() and conversation_set_dissector() to find a way to set change the TCP payload dissector based on the data in the TCP dissector. All coming from section 2.2, 2.3 and 2.4 from README.developer. Programming-wise it all works, output-wise it doesn't. And now I'm wondering if what I want is possible or not. Short version: If a certain TCP option exist, then the content of the TCP payload is not the "well-known" payload associated with the TCP port number. So even if the TCP port number is port 80, it still isn't HTTP traffic. How do you know? Because there is a TCP option which tells me "this is not HTTP, this is an internal protocol.". In dissect_tcpopt_ourinternal(), I have the following code. Shouldn't be too difficult: Get the dissector handle, find the conversation, set the new dissector for that conversation: conversation_t *conversation; static dissector_handle_t *ourinternal_handle = NULL; /* * We need to map this TCP session on our own dissector instead of what * Wireshark thinks runs on these ports - Edwin */ if (ourinternal_handle == NULL) { fprintf(stderr, "Finding ourinternal dissector\n"); ourinternal_handle = find_dissector("ourinternal"); } if (ourinternal_handle != NULL) { fprintf(stderr, "Setting ourinternal dissector? "); conversation = find_conversation(pinfo->fd->num, &pinfo->src, &pinfo->dst, pinfo->ipproto, pinfo->srcport, pinfo->destport, 0); if (conversation == NULL) { fprintf(stderr, "New conversation! "); conversation = conversation_new(pinfo->fd->num, &pinfo->src, &pinfo->dst, pinfo->ipproto, pinfo->srcport, pinfo->destport, 0); } if (conversation->dissector_handle != ourinternal_handle) { fprintf(stderr, "Setting to ourinternal dissector!\n"); conversation_set_dissector(conversation, ourinternal_handle); } else { fprintf(stderr, "Already done!\n"); } } And the result I get is: [~/wireshark] edwin@k7>./wireshark/tshark -nr test2.cap -c 10 1 0.000000 10.19.143.22 -> 10.28.7.99 TCP S, 1945 > 8080 [SYN] Seq=0 Win=8192 Len=0 MSS=1460 2 0.046720 10.28.7.99 -> 10.19.143.22 TCP SA, 8080 > 1945 [SYN, ACK] Seq=0 Ack=1 Win=8192 Len=0 MSS=1460 Finding ourinternal dissector Setting ourinternal dissector? New conversation! Setting to ourinternal dissector! 3 0.047463 10.19.143.22 -> 10.28.7.99 TCP [TCP Port numbers reused] TRPY, 1945 > 8080 [SYN] Seq=0 Win=5840 Len=0 MSS=1460 TSV=245836922 TSER=0 WS=2 Setting ourinternal dissector? Already done! 4 0.096437 10.28.7.99 -> 10.19.143.22 TCP TRPY, 8080 > 1945 [SYN, ACK] Seq=0 Ack=1 Win=5792 Len=0 MSS=1460 TSV=250693356 TSER=245836922 WS=2 Setting ourinternal dissector? Already done! 5 0.096463 10.19.143.22 -> 10.28.7.99 TCP TRPY, 1945 > 8080 [ACK] Seq=1 Ack=1 Win=5840 Len=0 TSV=245836971 TSER=250693356 Setting ourinternal dissector? Already done! 6 0.096532 10.19.143.22 -> 10.28.7.99 HTTP Continuation or non-HTTP traffic Setting ourinternal dissector? Already done! 7 0.096930 10.19.143.22 -> 10.28.7.99 HTTP Continuation or non-HTTP traffic Setting ourinternal dissector? Already done! The output of fprintf(stderr,) gets printed before packet, so during the parsing of packet 3 it comes in the dissect_tcpopt_ourinternal(), finds that there is no value for the dissector_handle is, gets it (subsequent calls to the dissect_tcpopt_ourinternal() show that it was assigned correctly). It find that it's a new conversation and assigns our internal dissector to it. The logging on packet 4 and following shows that the conversation is still valid, that the dissector handle is still set to our internal dissector. But then, packet 6 and 7 are still talking about HTTP Continuation and setting a breakpoint at the http_dissector show that this one is still called... So euhm... Is there a way to change the dissector of the TCP Payload based while you are still one level higher in the TCP dissector? Edwin -- Edwin Groothuis Website: http://www.mavetju.org/ edwin@xxxxxxxxxxx Weblog: http://www.mavetju.org/weblog/
- Follow-Ups:
- Re: [Wireshark-dev] Changing dissector based on TCP options
- From: James Deucker
- Re: [Wireshark-dev] Changing dissector based on TCP options
- From: didier
- Re: [Wireshark-dev] Changing dissector based on TCP options
- Prev by Date: [Wireshark-dev] buildbot failure in Wireshark (development) on Windows-7-x64
- Next by Date: [Wireshark-dev] buildbot failure in Wireshark (development) on Windows-7-x64
- Previous by thread: Re: [Wireshark-dev] FW: wireshark decodes xml files?
- Next by thread: Re: [Wireshark-dev] Changing dissector based on TCP options
- Index(es):