Invite ideas on implementing a packet generation tool for linux...

Hi All,

We are working on a idea of writing a linux based packet generation
tool. Some tools like like pktgen, libnet library etc. are already
available. We don't want to reinvent the wheel, but, trying to make a
generic and powerful packet generation tool which will support various
protocols. Most importantly that will be an open source. Your ideas on
that are most welcome.

Really appreciate a storming debate on this.

Thanks
-paul

On 7 Dec 2005 05:54:13 -0800, paul <(E-Mail Removed)> wrote:
> Hi All,
>
> We are working on a idea of writing a linux based packet generation
> tool. Some tools like like pktgen, libnet library etc. are already
> available. We don't want to reinvent the wheel, but, trying to make a
> generic and powerful packet generation tool which will support various
> protocols. Most importantly that will be an open source. Your ideas on
> that are most welcome.

Take a look at hping, nemesis, rain, sendip, etc. One or more of them
probably already do what you want, in which case there is no need to
invent anything. Hping3 has a TCL interpreter embedded into it, so
you can do fairly sophisticated automation with it if you need to.


Mike

--
Michael Zawrotny
Institute of Molecular Biophysics
Florida State University | email: (E-Mail Removed)
Tallahassee, FL 32306-4380 | phone: (850) 644-0069

> Take a look at hping, nemesis, rain, sendip, etc. One or more of them
> probably already do what you want, in which case there is no need to
> invent anything. Hping3 has a TCL interpreter embedded into it, so
> you can do fairly sophisticated automation with it if you need to.

Do they get timing right? Precise timing (well below the kernel
scheduling resolution) can be a major problem. You can't get this right
without some level of kernel assistance.

Allen McIntosh <(E-Mail Removed)> wrote:
> > Take a look at hping, nemesis, rain, sendip, etc.
>
> Do they get timing right? Precise timing (well below the kernel
> scheduling resolution) can be a major problem. You can't get this right
> without some level of kernel assistance.

Other than protocol stack development, I can't see why one would need
that level of resolution. For anything involving an actual remote
system, the vagaries of network latency, remote system load and
routing would generally dominate the low order bits of a high
resolution timer. There are a few situations where that isn't true,
like the SSL timing attack from a while ago...

In any case, the OP wanted to generate arbitrary network packets using
multiple protocols, which is what those tools do.


Mike

--
Michael Zawrotny
Institute of Molecular Biophysics
Florida State University | email: (E-Mail Removed)
Tallahassee, FL 32306-4380 | phone: (850) 644-0069

My idea is to develop a single tool which will be comparable with
IXIA...In linux box the bottleneck can be number of packets per
second...i want to implement it using c...right now we are just
exploring and have no idea how to implement it at kernel level...we
want it to be efficient, so that people can find it useful...any design
idea to get a start will be really helpful.

BTW i have read the following thread but could not understand how the
calcutaion is done and PCI bus speed related concepts...any help in
this matter will be appreciated.

http://groups.google.co.in/group/com...522d006427c642