Control-c not work under openssh?

[Irving Popovetsky]

I am seeing the exact same thing, on all of the versions of Solaris
(2.6,2.7,2.8) that we have running here.  No problems on my Linux or *BSD
boxen. 

To clarify, when I ssh to a box running OpenSSH (any ssh client will do),
and then from the OpenSSH box attempt to ssh to anywhere else,  it will
not catch SIGINT properly.

I think I've narrowed it down to some possibly faulty SIGINT catching in
readpass.c, and am toying with some ideas.   But I haven't figured
anything out yet. 

I have truss output if anyone wants it.

-Irving



On Mon, Jul 17, 2000 at 04:28:57PM -0700, Garrick James wrote:
> Sorry this reply is pieced together wierd--I had to cut and paste from the
> web archives.
> 
> Anyway, I am seeing this same control-C problem, also.  Was there ever
> any fix found?
> 
> My details:
> 
> When making ssh connections to openssh-2.1.1p4 servers running on Solaris
> 2.6, Control-C no longer causes an interrupt.  The stty settings are
> reported as the same that are reported when just using telnet to the box
> (and control-c works then).
> 
> Strangely enough, it seems to be somewhat dependant on the client being
> used.  When the client is openssh-2.1.1p4 (on Solaris 2.6), the problem
> exists (that is, control-c does not work).  When the client is Tera Term
> Pro with the ssh1 add-on (Win32 client), the problem does NOT exist (that
> is, control-c works fine).  When the client is SecureCRT 3.1 (Win32
> client), the problem DOES exist.  When the client is openssh-2.1.1p4 (on
> Solaris 2.6) and the server is openssh-2.1.1p4 (on Linux), the problem
> does NOT exist.  Strange matrix, eh?
> 
> When the problem does exist, 'su - username' seems to fix it (regardless
> of the client :).  Needless to say, it is very annoying to have to su
> everytime I log in just to get control-c working.
> 
> Anyway, has anyone found the reason for this problem?  Anybody have an
> idea how to fix it?
> 
> -Garrick James
> 
> 
> ----- Previous Message(s) in Thread -----
> 
> >>>"JH" == John Horne <J.Horne at plymouth.ac.uk> writes:
> 
>  JH> I'm using openSSH 2.1.1p2 from my Linux PC to Solaris 7 and 8 systems
> (using
>  JH> 2.1.1p1). No problem, except that if I run a program on the Sun,
> having used
>  JH> slogin to connect to it, and then try and issue a control-c to break
> into
>  JH> the program it seems to be ignored. I can control-z and then kill the
> ...
> 
> Hi,
> 
> I got the same problem and want to add some details that might help to
> track down the problem.
> 
> We have a heterogeneous environment of Solaris 2.6 and 8 and Linux, most
> using ssh 1.x and some OpenSSH 2.1.1p1. After some testing, we are
> certain that the problem is with the OpenSSH daemon. The ctrl-c is
> ignored whenever the target host runs this sshd on Solaris 8 (didn't try
> 2.6; no problem on Linux). The problem exists, if the user has a Bourne
> shell or a bash. However, if you invoke (in a broken tty/shell) either
> /bin/su or zsh, ctrl-c works again. The ssty -a output in a broken and
> not-broken tty are the same.
> 
> <rough guess> 
> sshd might be doing something wrong when initializing the
> pty/tty. A truss/strace comparison of Solaris 8 and Linux shows that the
> Linux sshd does some more ioctls on the tty during login. Maybe su and
> zsh repair the problem by default but bash and sh don't.  
> </rough guess>
> 
> Hope that helps someone who know more about this stuff ... ;-)
> 
> Cheers, Nils
> 
>