| Bug ID |
9470
|
| Summary |
Packet scoped "proto" data
|
| Classification |
Unclassified
|
| Product |
Wireshark
|
| Version |
SVN
|
| Hardware |
All
|
| URL |
http://www.wireshark.org/lists/wireshark-dev/201311/msg00170.html
|
| OS |
All
|
| Status |
CONFIRMED
|
| Severity |
Enhancement
|
| Priority |
Medium
|
| Component |
Dissection engine (libwireshark)
|
| Assignee |
[email protected]
|
| Reporter |
[email protected]
|
| CC |
[email protected]
|
| Attachment #12150 Flags |
review_for_checkin?
|
Created attachment 12150 [details]
Packe scoped proto data API
Build Information:
SVN r53516
--
Working off of
http://www.wireshark.org/lists/wireshark-dev/201311/msg00170.html
I created functionality that allows for packet scoped proto data. I'm not
really sure where this functionality should live, but since it was already with
the frame_data, I just appended the functionality.
I admit that its a bit of a hack to key off of allocator as to where the pool
is drawn from, but it does work. I've provided 2 patches - one with the new
functionality + my "test" dissector (ethertype). The other is the "compile
changes" needed for all dissectors to use the new API. I defaulted all to
file_scope() since thats how it behaved previously. If we keep the function
name/signature I figured the "hard part" of search/replace would be done.
With just the ethertype dissector converted there was a substantial memory
savings. I had a ~500000 frame capture (all ethernet packets) and it saved
~35Mb). All the work I had been doing to remove members from packet_info (to
reduce per packet memory usage) was actually making the memory situation worse
(although the code itself got better).
You are receiving this mail because:
- You are watching all bug changes.