Re: multi-threaded Nmap

[David Fifield <david () bamsoftware com>]

On Wed, Mar 23, 2011 at 08:30:53PM +0530, Kamal Banga wrote:
> I am a second year CSE student. I have seen that many-a-times nmap users
> have to do intense scans. I know CUDA and a bit about multithreaded
> programming. How about developing multithreaded code for nmap. And after all
> it will be a brand image for nmap! as now-a-days softwares use it. Like
> Matlab is available in both single and multi threaded applications.

That's an interesting idea. You would have to make a very convincing
case that it would be worth it, though. Not only Nmap but other
high-performance network software is written single-threaded with
asynchronous I/O. Nmap doesn't do very much that is CPU-bound, so a
potential benefit of threads (easily taking advantage of multiple
processors) is lost.

However I don't mean to discourage you. There may be a place for forms
or parallelism that we aren't using now. Maybe simple online hash
cracking for the future dns-nsec3-enum script
(https://secwiki.org/w/Nmap/Script_Ideas#dns-nsec3-enum)? You just have
to justify it in your proposal. I suggest that you read scan_engine.cc
and nsock/src/nsock_core.c to see examples of parallel programming
without threads.

David Fifield
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/