Chris Maynard
changed
bug 8629
| What |
Removed |
Added |
| Attachment #10719 is obsolete |
|
1
|
| Attachment #10719 Flags |
review_for_checkin?
|
|
| Attachment #10733 Flags |
|
review_for_checkin?
|
Comment # 10
on bug 8629
from Chris Maynard
Created attachment 10733 [details]
Patch to indicate 0x0000 as a bad checksum and allow reassembly for any
checksum.
It's not clear to me how others feel about this, but I was thinking that a bad
checksum shouldn't necessarily prevent reassembly from occurring as long as the
TCP preference to "Allow subdissector to reassemble TCP streams" is selected.
This patch explicitly indicates all bad checksums regardless of the value. It
also explicitly allows desegmentation, even if the checksum is bad.
For testing, I captured a short TCP session and began playing around. I have
several different variations of that TCP session where I intentionally
corrupted combinations of things like the TCP checksum, data in the payload and
a TCP header field. In all cases where the checksum was incorrect (regardless
of whether some data was corrupted), the stream still reassembled just fine.
Would that always be the case? I don't know. But if checksum offloading is
occurring, currently the streams can't be reassembled. With this patch, they
could be.
And of course, this patch will indicate bad checksums, even if the checksum is
0x0000, which I think should resolve the bug.
If anyone is interested in the capture files I used for testing, let me know
and I can attach them to this bug report.
You are receiving this mail because:
- You are watching all bug changes.