Wireshark-dev: [Wireshark-dev] Need advice on modifying tvb

Date: Tue, 24 Nov 2009 14:42:19 -0500
I am trying to rewrite an existing dissector for a proprietary protocol that, in fact, is only a slight variation on a standard protocol that is supported by a builtin Wireshark dissector.

The proprietary frame begins with some special fields, followed by a normal frame of the standard protocol BUT the checksum at the end of the normal frame is recalculated to reflect the extra bytes at the beginning.  So while I can easily write a small dissector that parses the initial extra fields (and have done so), I cannot simply pass the rest of the buffer to the builtin dissector since then the checksum will be wrong.

My question is:  What is the "right" way to fix that checksum in Wireshark?

I've experimented with several strategies.  The tvb_composite functions would appear to be ideal, but I can't get them to work for some reason.  So at the moment I use tvb_memdup to put the data for the normal frame into an array, fix the bytes of the checksum, and then use tvb_new_real_data to create a new tvb to pass to the builtin dissector.   That strategy mostly works: the builtin dissector dissects the right fields and doesn't complain about the checksum, but the display isn't quite right.  When I select a field in the display tree, the wrong bytes are highlighted.  I could fix that by figuring out what internal field of the tvb (or packet_info) needs to be tweaked... but the fact that I would have to do that tells me that maybe I'm not going about this the right way.  Generally whenever I find myself needing to work around the Wireshark API, it means I'm using it wrong.  ;)

Those of you who are experienced Wireshark developers, what would you suggest?

Thanks,
b.