This is the mail archive of the cygwin-xfree@cygwin.com mailing list for the Cygwin XFree86 project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

RE: new testcase: Re: Info on "Can't open display"




> I just did a bit of reading about aventail - it is the successor to
> AutoSOCKS. As such it will be hooking into TCP at a very low level.


Oh the crap....  This is like UnixLink from NetManage Sock I was using years
ago...
That was screwing up a lots of TCP/IP applications because it bascially
replaced the Microsoft TCP/IP with their and installed their own Socket.

Anyway, here is what is going on with Aventail.... which they calim to be
great Engineering invotaions by them... FROM my experience it is a total a
bad engineering invention.  From their White pages and Documents
(http://web.oxi.net/wingate/autosocksdata.htm)  Aventail replaced TCP/IP
with their own and install a SOCK, which forces all the application to
communicate through Aventail Socket, Aventail is active or not.

So Rob... do you really think you can fix it in Cygwin?  I guess it will be
a tough job.  They are not even using Microsoft WinSock, instead they are
forcing everyone to use their SOCKETS regardless of you run Aventail or not,
once you install it, you are hooked to their SWOCKET.  In my opinion it is a
CRAP.

Suhaib

>From their FACT SHEETS...

HOW AUTOSOCKS WORKS AutoSOCKS transparently makes Windows-based TCP/IP
applications SOCKS compliant. By implementing AutoSOCKS with a corporation's
existing TCP/IP applications, clients can communicate through the
SOCKS-enabled servers. 

AutoSOCKS works by intercepting WinSock communication requests issued by
applications and then processes the requests based on a set of rules. These
rules govern whether or not a WinSock request is redirected through a SOCKS
server. While communications within the local network may proceed unchanged,
all communication with external networks can be redirected through one or
many SOCKS servers, providing a single controlled access server for traffic
out to the Internet from your corporate desktops and access in from the
Internet for your mobile users. 


> 
> Can someone experiencing the problem with Aventail try the following:
> 
> start the X server, but no clients.
> start bash
> $ export DISPLAY=<insert ethernet or ppp adapter ip address here>:0.0
> then from the same bash session run vtwm or some window manager.
> 
> I'm guessing that that will work with Aventail.
> 
> === reasoning ===
> Looking at the AF_UNIX code in cygwin, all the source connections for
> AF_UNIX sockets are emulated by a normal socket connection made with the
> following settings...
> 
> sockaddr_in sin;
> sin.sin_family = AF_INET;
> sscanf (buf + strlen (SOCKET_COOKIE), "%hu %08x-%08x-%08x-%08x",
>         &sin.sin_port,
>         secret_ptr, secret_ptr + 1, secret_ptr + 2, secret_ptr + 3);
> sin.sin_port = htons (sin.sin_port);
> sin.sin_addr.s_addr = htonl (INADDR_LOOPBACK);
> *out = sin;
> *outlen = sizeof sin;
> ret = 1;
> 
> IE a fixed port (dependent on the contents of the socket file, and a
> fixed source address (127.0.0.1)...
> 
> If the above test works then we'll have a workaround for Aventail users.
> And hopefully someone with Aventail support can log a bug report with
> them.
> 
> If this is the problem, it isn't something that is appropriate to change
> in Cygwin. (AF_UNIX sockets are only allowed to operate on the local
> machine, and using 127.0.0.1 as the source port guarantees that that is
> the case).
> 
> Rob
> 
> ----- Original Message -----
> From: "Suhaib Siddiqi" <ssiddiqi@inspirepharm.com>
> To: <cygwin-xfree@cygwin.com>
> Sent: Saturday, June 16, 2001 10:06 PM
> Subject: RE: Info on "Can't open display"
> 
> 
> >
> >
> > > -----Original Message-----
> > > From: Christopher Faylor [mailto:cgf@redhat.com]
> > > Sent: Friday, June 15, 2001 10:10 PM
> > > To: cygwin-xfree@cygwin.com
> > > Subject: Re: Info on "Can't open display"
> > >
> > > On Fri, Jun 15, 2001 at 09:34:22PM -0400, Suhaib Siddiqi wrote:
> > > >
> > > >> You will lose fork() support for sockets if you use winsock32
> directly.
> > > >
> > > >
> > > >I know.  What I want to see if this display problem goes away with
> it.
> > > >I know it is not the best way.
> > > >
> > > >> I'm much more keen on identifying why the problem occurs - we can
> > > >> probably solve it quickly inside cygwin1.dll once we know that.
> > > >
> > > >Can anyone build a debug version of Cygwin1.dll with the Egor Duda
> patch
> > > he
> > > >posted this morning?  I do not have time to fool around with Cygwin
> code.
> > >
> > > I'm sure everyone knows this, but Egor's patch was just a temporary
> > > commenting out of some AF_UNIX security code.  It wasn't a real
> "fix".
> > > So, don't get too excited if things start working for you after
> applying
> > > the patch.
> >
> >
> > I know it was a temporary fix.  I man for checking if his Security
> check
> > For AF_UNIX is creating problem.  This is also for testing purposes.
> >
> > Suhaib
> >
> > >
> > > cgf
> >


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]