Re: Nmap on IP-over-InfiniBand interfaces - was: (lingering problem: route_dst_netlink: can't find interface)

[Djalal Harouni <tixxdz () opendz org>]

On Tue, Aug 28, 2012 at 01:22:07PM -0400, starlight.2012q3 () binnacle cx wrote:
> Thanks!  The patch works.  'nmap' starts
> up fine and functions as expected on Ethernet
> interfaces.  Shows the chopped MAC address
> for ib0 an ib1 with --iflist.
Ok, thanks.

> As expected, IB networks can be scanned
> with --send-ip.  Seems perfectly fine.
>
> --send-eth fails to detect other systems
> on the IB subnets.  Strangely it does
> work for the local IB interface.  Possibly
> that's a linux kernel bug since really
> it should not.  Or possibly since the
> frames are local the MAC layer is entirely
> ignored.
I don't know if we can send ethernet frames, and I would suspect that the
driver does some packet re-writing magic to handle 20bytes Hwaddr...

Ok at least it works with IP packets.

> Since use of IB is relatively rare at stable
> or declining levels (10G and 40G Ethernet
> work well) I'd say this degree of support is
> "good enough."  Perhaps 'nmap' can sense
> when it's asked to route over an IB
> interface (or a non-Ethernet interface)
> and automatically apply --send-ip.
Yes this can be used. Internally we flag some interfaces as ethernet,
p2p... and we adapt the code. Actually libdnet also does this, but at
some points it can report wrong types.

Anyway to be sure how to handle this I need to take another look at the
code.


David I'll let you handle this, in my opinion making libdnet/Nmap able
to handle upto 20bytes of Hwaddr is the best. Actually I think that Linux
kernel has room for 32bytes (I'm not sure). However, how much code needs
to be updated in this case ?

Otherwise and if we are not able to send ethernet frames, just report
6bytes! and fallback to --send-ip...


I've also changed the subject so others can find it.
(See previous mails of the same thread).

Thanks.

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