Hi David,
Not at the moment, no. Anyone else?
I discovered some #if’d-out debugging code in tcp_analyse_sequence_number(), which seems to reflect the problem:
analyze_sequence numbers frame:1 FWD list lastflags:0x0000 base_seq:0: nextseq:0 lastack:0 REV list lastflags:0x0000 base_seq:0 nextseq:0 lastack:0 analyze_sequence numbers frame:2 FWD list lastflags:0x0000 base_seq:0: nextseq:0 lastack:0 REV list lastflags:0x0000 base_seq:3871803454 nextseq:3871803455 lastack:0 Frame:1 Seq:3871803454 Nextseq:3871803455 analyze_sequence numbers frame:3 FWD list lastflags:0x0000 base_seq:3871803454: nextseq:3871803455 lastack:3871803455 REV list lastflags:0x0000 base_seq:3273800524 nextseq:3273800525 lastack:0 Frame:2 Seq:3273800524 Nextseq:3273800525 analyze_sequence numbers frame:4 FWD list lastflags:0x0000 base_seq:3871803454: nextseq:3871803455 lastack:3871803455 REV list lastflags:0x0000 base_seq:3273800524 nextseq:3273800525 lastack:3273800525 analyze_sequence numbers frame:5 FWD list lastflags:0x0000 base_seq:3273800524: nextseq:3273800525 lastack:3273800525 REV list lastflags:0x0000 base_seq:3871803454 nextseq:3871803504 lastack:3871803455 Frame:4 Seq:3871803455 Nextseq:3871803504 analyze_sequence numbers frame:6 FWD list lastflags:0x0000 base_seq:3273800524: nextseq:3273800525 lastack:3273800525 REV list lastflags:0x0000 base_seq:3871803454 nextseq:3871803504 lastack:3871803504 analyze_sequence numbers frame:7 FWD list lastflags:0x0000 base_seq:3871803454: nextseq:3871803504 lastack:3871803504 REV list lastflags:0x0000 base_seq:3273800524 nextseq:3273800529 lastack:3273800525 Frame:6 Seq:3273800525 Nextseq:3273800529 analyze_sequence numbers frame:8 FWD list lastflags:0x0000 base_seq:3871803454: nextseq:3871803504 lastack:3871803504 REV list lastflags:0x0000 base_seq:3273800524 nextseq:3273800529 lastack:3273800529 analyze_sequence numbers frame:9 FWD list lastflags:0x0000 base_seq:3273800524: nextseq:3273800529 lastack:3273800529 REV list lastflags:0x0000 base_seq:3871803454 nextseq:3871803553 lastack:3871803504 Frame:8 Seq:3871803504 Nextseq:3871803553 analyze_sequence numbers frame:10 FWD list lastflags:0x0000 base_seq:0: nextseq:0 lastack:0 REV list lastflags:0x0000 base_seq:0 nextseq:0 lastack:0 analyze_sequence numbers frame:11 FWD list lastflags:0x0000 base_seq:3273800561: nextseq:0 lastack:3273800562 REV list lastflags:0x0000 base_seq:3871803552 nextseq:3871803553 lastack:0 analyze_sequence numbers frame:12 FWD list lastflags:0x0000 base_seq:3871803552: nextseq:3871803553 lastack:3871803553 REV list lastflags:0x0000 base_seq:3273800561 nextseq:3273800575 lastack:3273800562 Frame:11 Seq:3273800562 Nextseq:3273800575
The sequence numbers in frame 10 appear to have been completely reset.
The bogus (relative) sequence number displayed for frame #9 (not reflected here — the absolute values, and packet bytes themselves, appear to be fine) is a result of a bad value for tcpd->fwd->base_seq during the calculations, bearing no resemblance to the initial sequence numbers for either direction’s flow. I haven’t figured out where that’s coming from yet.
d On 2 Dec 2018, at 23:36, David Arnold < davida@xxxxxxxxx> wrote:
Hi Jaap,
Thanks for looking into this.
The problem with frame 9 appears to be the result of a change to use ws_strtoi32() to convert a string with trailing whitespace. A very quick workaround of that (just supplying an end pointer) avoids the reported error, but doesn’t avoid the TCP sequence number corruption.
Still investigating; any further suggestions?
d
On 30 Nov 2018, at 01:43, Jaap Keuter <jaap.keuter@xxxxxxxxx> wrote:
Your frame 9 dissection errors out (as malformed), which probably trips up the TCP dissector as well, not allowing it to do all it’s work after the payload dissector is done.
Thanks,
Jaap
On 29 Nov 2018, at 13:34, David Arnold <davida@xxxxxxxxx> wrote:
Hi all,
I’ve discovered an odd issue with my dissector, and I’d really appreciate some debugging pointers.
I have a capture file (attached) which, when viewed without any explicit decoding, looks just fine — in particular, all the TCP seq/ack numbers appear reasonable, and don’t flag any errors. When I set the “Decode As …” option to “SoupBinTCP” (the appropriate protocol), I start to see some errors with the TCP sequence numbers.
Specifically, the reported (relative) sequence numbers are fine for the first 8 packets in the capture, but on the 9th packet, the *reported* value is screwy, and all subsequent packets are therefore messed up too. The bogus reported value is not reflected in in the shown packet bytes, which look consistent with other packets.
I’m testing using a recent clone of Git master, but have also reproduced the problem on v2.1.0 (which I had installed on a handy machine), so it’s not a new problem.
Any suggestions for what might be going wrong much appreciated.
Thanks in advance,
d
<tradenow.pcap>
___________________________________________________________________________
___________________________________________________________________________Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>Archives: https://www.wireshark.org/lists/wireshark-devUnsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
|