[Click] a bug in tcprewriter.cc:apply_sack() and/or
tcprewriter.cc:apply()?
Eddie Kohler
kohler at CS.UCLA.EDU
Sat Oct 1 21:38:58 EDT 2005
AAAAAAAAAAAARGH
TCPRewriter::TCPMapping::apply_sack()'s option parsing code had no
cases for options other than NOP, EOL, and SACK.
That meant that if for example there was a SACKOK option it would
infinite loop.
SO STUPID!!!!!
Weidong, the fix is checked in (and a diff appended). Let me know if
this seems to help.
Thanks for reporting the bug!
Eddie
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tcprw.patch
Type: application/octet-stream
Size: 616 bytes
Desc: not available
Url : https://amsterdam.lcs.mit.edu/pipermail/click/attachments/20051001/6d4a2510/tcprw.obj
-------------- next part --------------
On Sep 30, 2005, at 9:52 AM, Weidong Cui wrote:
> Hi Eddie,
>
>
>> (1) Do you have a trace file + config that can reliably replicate
>> the problem?
>>
>
> I don't have them. I run it on a real honeyfarm testbed in ICSI.
> But, I can give you my source code and click conf file if you want
> to somehow reproduce it by yourself. What my code does is
> basically to ack all tcp syns, check the first data packet (if
> any), and then set up another tcp connection to the real server,
> finally work as a tcp proxy (using tcprewriter).
>
>
>> (2) Do you know what the value of 'len' was in
>> TCPRewriter::apply_sack()? I wonder if it was less than 0. That
>> would cause some serious issues.
>>
>
> I think the 'len' is 40, which is weird because there shouldn't be
> any TCP options (as I explained in my last email).
>
> Please let me know if I can help on anything else.
>
> Thanks,
> Weidong
>
> --
> _____________________________________________________
> Weidong Cui
> Ph.D. Candidate wdc at EECS.Berkeley.EDU
> EECS, UC Berkeley http://www.cs.berkeley.edu/~wdc/
>
More information about the click
mailing list