Re: NSE Socket Operation on a non-socket

[David Fifield <david () bamsoftware com>]

On Tue, Jul 21, 2009 at 01:32:11PM -0600, David Fifield wrote:
> On Sat, Jul 18, 2009 at 11:05:09PM -0600, David Fifield wrote:
> > On Sat, Jul 18, 2009 at 04:54:41AM +0100, jah wrote:
> > > I've been struggling with an error which I can reliably reproduce with
> > > the following (r14408) nmap command against a windows machine with tcp
> > > ports 445 and 3389 open:
> > >
> > > nmap -sSV -p 445,3389 -PN -n --script banner,smb-enum-shares -d <target>
> > >
> > > The error looks like this:
> > >
> > > NSOCK (-1950392.4020s) nsock_loop error 10038: An operation was
> > > attempted on something that is not a socket.
> > > NSE: Script Engine Scan Aborted.
> > > An error was thrown by the engine: a fatal error occurred in nsock_loop
> > > stack traceback:
> > >         [C]: ?
> > >         [C]: in function 'nsock_loop'
> > >         C:\Program Files\Nmap\nse_main.lua:605: in function 'run'
> > >         C:\Program Files\Nmap\nse_main.lua:734: in function <C:\Program Files\Nmap\nse_main.lua:693>
> > >         [C]: ?
> >
> > I got this error too on Windows a few days ago while running an NSE
> > benchmark test for Patrick. It happened every time with
>
> I think I have found the cause of this problem. This too is related to
> the change to counting the number of pending reads and writes on sockets
> to control the select sets.
>
> Re: Ncat with ssl using 100% cpu (PATCH)
> http://seclists.org/nmap-dev/2009/q3/0019.html
>
> Re: Assertion failure in nsock_core
> http://seclists.org/nmap-dev/2009/q3/0128.html
>
> The error happens when handle_connect_result is called for an SSL
> connection. At this point the socket's read and write counts are
> decremented (to zero). Usually after the TCP connection,
> handle_connect_result needs to be called again to handle an
> SSL_ERROR_WANT_READ. In this case the read count is incremented (to
> one). It is normally decremented to zero after that, but in case of a
> connection timeout it is not. So the socket descriptor remains in the
> select set and causes the error when selected on.

The fix is committed in r14489.

David Fifield

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