Re: [ncat] Listening Unix domain socket and UDP dies on Connect

[Daniel Miller <bonsaiviking () gmail com>]

John,

I ran across your email again in my to-do folder, and discovered that this
is probably the same bug that we have been discussing recently. If this is
still affecting you, please weigh in over at http://issues.nmap.org/46

Dan

On Mon, Nov 3, 2014 at 8:52 AM, John Schwarz <jschwarz () redhat com> wrote:

> Daniel,
>
> Thanks a lot for responding. Answers inline.
>
> > On Oct28, 2014, at 01:34, Daniel Miller <bonsaiviking () gmail com> wrote:
> >
> > John,
> >
> > I'm sorry it has taken so long to get back to you on this.
> > Unfortunately, I'm not able to reproduce the error. Is it still
> > affecting you?
>
> I tried working on this feature again a few weeks back and still
> encountered
> the same problem. It will probably take me a while to rewrite the code as I
> reworked it (into a dummy python code which runs the server instead of
> ncat) :(
> > On Sun, Sep 21, 2014 at 12:00 PM, John Schwarz <jschwarz () redhat com>
> wrote:
> > > Hi guys,
> > >
> > > I'm encountering a really weird problem with the latest ncat version
> > > taken from your SVN sources, and hopefully you guys could help me (or
> > > otherwise confirm this is a bug?)
> > >
> > > I have an haproxy process which is configured to send logs to a specific
> > > unix domain socket located at some path at the HD. Prior to running the
> > > haproxy, I start a ncat process like so:
> > >
> > > ncat --unixsock <path> --output /tmp/logs --listen --unix -vvv
> >
> > I'm assuming you meant --udp and not --unix here
>
> Yeps. Sorry :)
> > > Once haproxy starts and sends the first log message, though, ncat quits
> > > and the printout in [1] is given. Using GDB, I've tracked the "connect:
> > > Invalid argument" to a call to Connect in ncat_listen.c:818 (function
> > > name: ncat_listen_dgram) which produce EINVAL. Furthermore, commenting
> > > out the Connect call and recompiling produces actual log messages being
> > > read by the ncat process.
> > >
> > > Do you guys have any ideas what am I doing wrong here?
> > >
> > > Much appreciated,
> > >
> > > [1]: http://pastebin.com/5xwdWZAE
> > > --
> > > John Schwarz,
> > > Software Engineer, Red Hat.
> >
> >
> > In order to better solve your problem, will you please provide some
> > information about your platform? I don't see EINVAL as a documented
> > error from connect in the Linux man pages, but I'd assume you are
> > running Red Hat. What kernel version are you using?
> I’m running Fedora 20, x86_64, kernel version 3.15.10-200.fc20.x86_64.
> Netcat’s version is taken from upstream (6.47SVN is what was running
> during the last attempt). What else can I provide you with?
> > Can you reproduce
> > the bug using ncat as the client as well, or is the crash specific to
> > the haproxy connection?
>
> Surprisingly, using netcat as a client as well as the server doesn’t cause
> the server to crash at all. How odd.
> > Of note, I am running across a bug where I can't send more than a
> > couple messages to a listening unixsock+udp ncat before it stops
> > receiving data. I'll continue to look into it.
> Sounds very similar to what I’m experiencing.
> > Dan
_______________________________________________
Sent through the dev mailing list
http://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/