Re: many scripts with output cause abort

[Brandon Enright <bmenrigh () ucsd edu>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Mon, 16 Mar 2009 23:50:30 -0600 or thereabouts Patrick Donnelly
<batrick () batbytes com> wrote:

> On Mon, Mar 16, 2009 at 6:15 PM, Brandon Enright <bmenrigh () ucsd edu>
> wrote:
> > I've noticed that with a large --min-parallelism and a large
> > - --min-hostgroup NSE seems to overload Nsock and causes Nmap to
> > corrupt mammary.  David and I looked at it a little but never
> > really pursued it as a bug.
>
> min-hostgroup shouldn't affect the nsock exclusion mechanism which
> ensures only --max-parallelism OR 10 threads have open sockets. The
> problem is threads may then open as many sockets as they want. This
> was a design decision I made last summer due to deadlocks occurring
> when threads would compete for the maximum 10 sockets but some needed
> 2. It seemed reasonable since all scripts up to that point had at most
> 2 sockets open at any time. Of course, this may not actually be the
> problem you are experiencing? Is nsock allocating too many open
> sockets?

It's possible Nsock had too many open sockets.  I have my max
file handles set to 65535.  I don't know how many ephemeral ports Linux
has available by default.  I am recycling and reusing tw buckets to
help take the pressure off sockets.

If I run with a --min-hostgroup 2048 --min-parallelism 2048 Nsock blows
up when NSE starts working.  David and I ran into this while
troubleshooting something else so we didn't look at it too hard.

If you're interested I'll test tomorrow in your nse-lua-merge branch
and see what happens.

Brandon

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)

iEYEARECAAYFAkm/O8MACgkQqaGPzAsl94LULgCgqYC0T/z3KO2iGV9CGSxSegrg
UwUAoKszUfH+KrjIF9Zep50v6tfKoQU5
=oC2U
-----END PGP SIGNATURE-----

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