On 05/28/2010 06:59 AM, Masatake YAMATO wrote:
>>>> On 5/27/2010 at 04:26 AM, "Caplan, Michael"<[email protected]>   wrote:
>>>>> Is there a corosync dissector available for Windows based Wireshark?
>>>>
>>>> I suspect the short answer to your question is "no".  Work has been
>>>> done in this area, but this doesn't appear to have ever made it into
>>>> upstream Wireshark.  See:
>>>>
>>>>     
>>>> https://lists.linux-foundation.org/pipermail/openais/2009-February/010603.html
>>>>
>>>> There's an enhancement request for this filed upstream, with a patch
>>>> for totemnet and totemsrp support:
>>>>
>>>>     https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=3232
>>>>
>>>> However, this doesn't appear to have seen any action since May last
>>>> year.
>>>
>>> Yes. As explained in the comment in bugs.wireshark.org, I have to
>>> rewrite encryption/decryption code with using a library in wireshark;
>>> in my old patch, I wrote the code based on code in corosync. It is a
>>> bit
>>> hard work for me. About upper layer, there is a good news. These days
>>> we can write wireshark dissector in lua or python.
>>>
>>> BTW, Has the wire format of totemsrp changed since May last year?
>>>
>> no
>
> Steven, I have two more questions.
>
>
>      Merge trunk revision 2660:
>      r2660 | sdake | 2010-02-18 13:08:39 -0700 (Thu, 18 Feb 2010) | 3 lines
>
>      Patch to set unset value in token hold cancel structure as to not crash
>      wireshark.
>
>      Index: exec/totemsrp.c
>      ===================================================================
>      --- exec/totemsrp.c      (revision 2659)
>      +++ exec/totemsrp.c      (revision 2660)
>      @@ -2627,6 +2627,7 @@
>            */
>          token_hold_cancel.header.type = MESSAGE_TYPE_TOKEN_HOLD_CANCEL;
>          token_hold_cancel.header.endian_detector = ENDIAN_LOCAL;
>      +        token_hold_cancel.header.encapsulated = 0;
>       token_hold_cancel.header.nodeid = instance->my_id.addr[0].nodeid;
>       memcpy (&token_hold_cancel.ring_id,&instance->my_ring_id,
>               sizeof (struct memb_ring_id));
>
> 1. It seems that you used wireshark with my patch.
>     Did my patch work well?
>
>     I don't well about corosync yet. (I started writing the patch to learn
>     the protocol.) So I don't convince myself that my patch covers whole the
>     protcol. If you have something unsatisfied with my patch, please let me
>     know.
>
>     This question is applicable to the other corosync/openais experts.
>     Any comments are welcome.
>
> 2. In the revision 2660 of corosync, you wrote "to not crash wireshark".
>     If wireshark crashes, it is a bug of my side. I'd like to fix it anyway.
>
>     I've looked at `encapsulated' field related code in my patch but I
>     cannot find my mistake. So the question is "did wireshark really crash?"
>
>     I guess, the wireshark didn't crash but could not recognize the corosync
>     protocol; it just report it as UDP diagram.
>
>

I don't have alot of details; HJ Lee in the community reported the 
problem and submitted the patch.  I have not tried wireshark although it 
would be great if the wireshark code was merged, even without encryption 
support, upstream.  This would aid people in debugging.

Regards
-steve

> Regards,
> Masatake YAMATO

_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais

Reply via email to