Re: 1.7.7: after upgrade lost ability to login via ssh

[Gerry Reno <greno@verizon.net>]

On 02/09/2011 08:04 PM, Gerry Reno wrote:
> On 02/09/2011 07:21 PM, Gerry Reno wrote:
>   
>> On 02/09/2011 06:43 PM, Larry Hall (Cygwin) wrote:
>>   
>>     
>>> On 2/9/2011 5:56 PM, Gerry Reno wrote:
>>>     
>>>       
>>>> On 02/09/2011 05:35 PM, Larry Hall (Cygwin) wrote:
>>>>       
>>>>         
>>>>> On 2/9/2011 5:07 PM, Gerry Reno wrote:
>>>>>         
>>>>>           
>>>>>> On 02/09/2011 04:56 PM, Gerry Reno wrote:
>>>>>>           
>>>>>>             
>>>>>>> On 02/08/2011 11:07 PM, Larry Hall (Cygwin) wrote:
>>>>>>>
>>>>>>>             
>>>>>>>               
>>>>>>>> On 2/8/2011 9:14 PM, Gerry Reno wrote:
>>>>>>>>
>>>>>>>>               
>>>>>>>>                 
>>>>>>>>> Something else I just discovered after upgrading to 1.7.7 is that
>>>>>>>>> I now
>>>>>>>>> have lost the ability to login via ssh.
>>>>>>>>>
>>>>>>>>> I have OpenSSH installed and running sshd as a service.  Both
>>>>>>>>> password
>>>>>>>>> and keys accepted.  But now neither means will work.
>>>>>>>>>
>>>>>>>>>        # ssh -i keypair1.pem   Administrator@MACHINE_IP
>>>>>>>>>        Last login: Fri Feb  4 17:19:26 2011 from
>>>>>>>>> LOCAL_CLIENT_MACHINE
>>>>>>>>>        Connection to MACHINE_IP closed.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> So I increased verbosity but did not see anything obvious.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>        # ssh -v -i keypair1.pem   Administrator@MACHINE_IP
>>>>>>>>>        OpenSSH_5.2p1, OpenSSL 0.9.8k-fips 25 Mar 2009
>>>>>>>>>        debug1: Reading configuration data /etc/ssh/ssh_config
>>>>>>>>>        debug1: Applying options for *
>>>>>>>>>        debug1: Connecting to MACHINE_IP [MACHINE_IP] port 22.
>>>>>>>>>        debug1: Connection established.
>>>>>>>>>        debug1: permanently_set_uid: 0/0
>>>>>>>>>        debug1: identity file keypair1.pem type -1
>>>>>>>>>        debug1: Remote protocol version 2.0, remote software version
>>>>>>>>> OpenSSH_5.8
>>>>>>>>>
>>>>>>>>>                 
>>>>>>>>>                   
>>>>>>>> Does reverting OpenSSH to 5.7 make a difference?
>>>>>>>>
>>>>>>>>
>>>>>>>>               
>>>>>>>>                 
>>>>>>> Downgraded to 5.7:
>>>>>>>
>>>>>>>       bash-4.1$ sshd --version
>>>>>>>       sshd: unknown option -- -
>>>>>>>       OpenSSH_5.7p1, OpenSSL 0.9.8r 8 Feb 2011
>>>>>>>
>>>>>>>
>>>>>>>             
>>>>>>>               
>>>>>>>>  From client:
>>>>>>>>               
>>>>>>>>                 
>>>>>>>       ssh -i keypair1.pem   Administrator@MACHINE_IP
>>>>>>>       Last login: Wed Feb  9 12:54:08 2011 from LOCAL_CLIENT_IP
>>>>>>>       Connection to MACHINE_IP closed.
>>>>>>>
>>>>>>>
>>>>>>> Nope.  Still have the same problem.  Connection is made but
>>>>>>> immediately
>>>>>>> closes.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>             
>>>>>>>               
>>>>>> I'm suspecting this is related to running Cygwin 1.7.
>>>>>>
>>>>>> In looking back though some notes I started having bash shell problems
>>>>>> after upgrading from 1.5 to 1.7.
>>>>>>
>>>>>> Now on 1.7 if I try to run bash as a login shell it just gets "Bad
>>>>>> address" or segfault errors and immediately exits the shell which also
>>>>>> probably affects 'ssh'.
>>>>>>
>>>>>> I don't remember having any bash problems when I was running Cygwin
>>>>>> 1.5
>>>>>> on this machine.  My notes reflect screen copies showing bash able to
>>>>>> run as a login shell without any problem.
>>>>>>           
>>>>>>             
>>>>> Yep, that's the way we all run by default (see cygwin.bat).  I agree
>>>>> that if you're having problems getting bash to behave, it's best to
>>>>> focus
>>>>> on that issue first.  Your ssh problems may just be another symptom of
>>>>> the same thing.  How about sending cygcheck output
>>>>> (<http://cygwin.com/problems.html>)?  There may be something helpful in
>>>>> that which someone on the list might pick up on.
>>>>>
>>>>>         
>>>>>           
>>>> Ok, ran a new cygcheck and attached it.
>>>>       
>>>>         
>>> OK, thanks.  What went wrong with the first installation?
>>>
>>> I notice that this is using TS.  Can you try experiment with this machine
>>> locally?  Or perhaps just try:
>>>
>>> <http://cygwin.com/faq-nochunks.html#faq.setup.setup-fails-on-ts>
>>>
>>>     
>>>       
>> I reduced DEP down to just Windows executables and dlls and then rebooted.
>>
>> And it actually seemed to make the problem worse:
>>
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash-4.1$ (for p in $(ls /etc/profile.d/*.sh);do . $p;done)
>>     bash: /etc/profile.d/lapack0.sh: Bad address
>>     bash-4.1$
>>
>>
>> So DEP in is play here but sort of inverse from what I'd expect.  There
>> was no switch now to totally disable it.  I guess they want you to
>> fiddle with the registry to turn it all the way off.
>>
>>
>>   
>>     
> I tried reinstalling bash and coreutils which installed ok but both
> their postinstall scripts have an abnormal exit 128 which is exactly
> what I was seeing previously.
>
>
>   
So I try running lines from these postinstall scripts manually.

In /etc/postinstall/bash.sh this code snippet kills everything:

    bash-4.1$ result=0
    bash-4.1$ test $result = 0
          2 [main] bash 2484 sig_send: wait for sig_complete event
    failed, signal -3
    4, rc -1, Win32 error 6
    <SHELL IS DEAD>

From MSDN:

    ERROR_INVALID_HANDLE    The handle is invalid.
    6 (0x6)


So what is this?

    Bad compilation?
    Stack corruption?



Regards,
Gerry

   




--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple