Wireshark-dev: Re: [Wireshark-dev] Defect in reassembling TCP stream. Bug and Patch are availab

From: Pavel Karneliuk <Pavel_Karneliuk@xxxxxxxx>
Date: Fri, 28 Mar 2014 16:09:52 +0000

Yes, I am going to gerrit.

 

From: wireshark-dev-bounces@xxxxxxxxxxxxx [mailto:wireshark-dev-bounces@xxxxxxxxxxxxx] On Behalf Of Pascal Quantin
Sent: Friday, March 28, 2014 6:57 PM
To: Developer support list for Wireshark
Subject: Re: [Wireshark-dev] Defect in reassembling TCP stream. Bug and Patch are available on Bugzilla.

 


Le 28 mars 2014 16:52, "Pavel Karneliuk" <Pavel_Karneliuk@xxxxxxxx> a écrit :
>
> Hi Pascal,
>
> thank you for answer. I saw your commits to follow.c and I hoped for your reply.
>
> 450:if( newseq > seq[idx] ) {
>
> I think – Yes. It compares sequence numbers.
>
>
> 459: if ( current->data_len > new_pos ) {
> I am sure,  that – No. Because it compares length of data from fragment instead of sequence numbers.
Doh that's what happens when you reply without looking carefully at the code ;)
>
>
> There are some places in check_fragments() and reassemble_tcp() with a “naive” comparison of sequence numbers:
> 369: if( sequence < seq[src_index] ) {
>
>  
>
> I think, they should be replaced with macros from packet-tcp.h 51-55.  At least to be uniformly.

As Graham suggested, it would be great if you could submit a patch on gerrit against master branch. Would it be feasible?

Regards,
Pascal.