Re: ioctl issue scanning /24 networks

[Daniel Miller <bonsaiviking () gmail com>]

Nicolas,

Thanks for your bug report.

On Fri, Jun 6, 2014 at 4:25 AM, Nicolas Greneche <
nicolas.greneche () univ-paris13 fr> wrote:

> epoll_wait(4, {}, 128, 50)              = 0
> ioctl(3, TIOCGPGRP, [0])                = -1 ENOTTY (Inappropriate ioctl
> for device)
> epoll_wait(4, {}, 128, 50)              = 0
> ioctl(3, TIOCGPGRP, [0])                = -1 ENOTTY (Inappropriate ioctl
> for device)
> epoll_wait(4, {}, 128, 50)              = 0
> ioctl(3, TIOCGPGRP, [0])                = -1 ENOTTY (Inappropriate ioctl
> for device)
>
> Those messages loop forever. Nmap process never exit.
I'm not sure what's the core issue here, but I did look into the ioctl
calls, and they seem to be coming from tty_getchar in nmap_tty.cc. The call
to tcgetpgrp on fd 3 results in ENOTTY, which is ok because obviously the
process forked by python-nmap is not running in a TTY (a rare case of
ENOTTY actually meaning what it says!). We could try to avoid polling like
this when we can detect this situation, but as the comment in the function
states:

        // This is so that when the terminal has been disconnected, it will
be
        // reconnected when possible. If it slows things down, just remove
it

I think this covers the scenario of running Nmap in a GNU Screen session,
for instance, and then disconnecting the session. The process's output is
no longer a TTY, but when you reconnect, you will regain the ability to
send keyboard interaction messages.

Sorry for the long-winded explanation of part of your strace output, but I
wanted to make sure for myself that a recent change to our keyboard
interaction code was not responsible for your problem. It looks instead
like a Nsock issue, based on the epoll_wait calls.

If you can reliably cause the hang (especially if you can do it without
using python-nmap, but by running the command directly), please let us know
if it still hangs with "--nsock-engine select" and provide some Nmap
debugging output "-d" as well.

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