Re: [Nmap, Ncat] Odd Nsock bug related to ENETUNREACH

[Henri Doreau <henri.doreau () gmail com>]

2014-07-28 20:29 GMT+02:00 Henri Doreau <henri.doreau () gmail com>:
> 2014-07-28 20:23 GMT+02:00 Jacek Wielemborek <d33tah () gmail com>:
> > 28.07.2014 20:19, Henri Doreau:
> > > 2014-07-28 15:42 GMT+02:00 Jacek Wielemborek <d33tah () gmail com>:
> > > > Just wanted to point out that nsock/include/nsock.h says:
> > > >
> > > > > /* EVENT CREATION FUNCTIONS
> > > > >  * ---
> > > > >  * These functions request asynchronous
> > > > >  * notification of completion of an event.  The handler will never be
> > > > >  * synchronously called back during the event creation call (that causes too
> > > > >  * many hard to debug errors and plus we don't want people to have to deal with
> > > > >  * callbacks until they actually call nsock_loop). */
> > > >
> > > > IIRC my IRC chat with Henri, this is not true anymore as ENETUNREACH
> > > > errors can cause callbacks to run synchronously.
> > > Hello,
> > >
> > > thanks, you're right, this comment isn't accurate anymore (actually it
> > > wasn't totally before anyway). Before I nuke it, do we agree that
> > > calling handlers from outside nsock_loop to report errors is
> > > acceptable? Jacek, you've become a heavy nsock user. What do you
> > > think?
> > >
> > > Regards
> >
> > I think that at the very least all ncat_connect_* calls should be
> > checked very carefully as they could have been written with the
> > assumption that the callbacks will be written when we call nsock_loop.
> >
> > Anyway, I think that it's very unintuitive and should be changed, but if
> > all the Nmap family programs are bug-free in this region, IMHO this is
> > acceptable (provided that we shout in the comments of ncat.h as well as
> > the relevant .c files about this behavior).
>
> Ok, then let me write a test patch to make nsock report errors only
> from within nsock_loop (except for IOD/NSP deletion, where I don't
> think it would make sense) so that we can see what we find best.
>
> Regards
>
> --
> Henri

There you go! It's a quickly written and almost untested PoC, but you
have the idea (and it seems to work as expected).

Regards

-- 
Henri

Attachment: nsock_deferred_errors.patch
Description:

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