[lxc-devel] [Lxc-users] child setpgid [...] : No such process

[Daniel Lezcano]

lxc at zitta.fr wrote:
> Le 12/03/2010 14:15, Daniel Lezcano a écrit :
>   
>> lxc at zitta.fr wrote:
>>     
>>> Le 12/03/2010 13:51, Daniel Lezcano a écrit :
>>>  
>>>       
>>>> lxc at zitta.fr wrote:
>>>>    
>>>>         
>>>>> Le 11/03/2010 19:47, Daniel Lezcano a écrit :
>>>>>  
>>>>>      
>>>>>           
>>>>>> lxc at zitta.fr wrote:
>>>>>>           
>>>>>>             
>>>>>>> I created a new container (karmic), then I type any command there is
>>>>>>> curious message, but it works:
>>>>>>>                   
>>>>>>>               
>>>>>> Do you mean you created a system container with karmic inside ?
>>>>>>             
>>>>>>             
>>>>> Yes, I'm testing a new version of my provisioning scripts.
>>>>>  
>>>>>      
>>>>>           
>>>>>> Can you give the kernel version, the lxc version, the container
>>>>>> configuration and the command used to spawn the container ?
>>>>>>             
>>>>>>             
>>>>> config as attachment.
>>>>>
>>>>> black provisioning # uname -a
>>>>> Linux black 2.6.31-zen11-lxc-bt #1 ZEN SMP PREEMPT Tue Feb 23 09:13:02
>>>>> CET 2010 x86_64 Intel(R) Core(TM)2 Quad CPU Q9450 @ 2.66GHz
>>>>> GenuineIntel
>>>>> GNU/Linux
>>>>>
>>>>> black provisioning # eix -I lxc | grep Installed
>>>>>      Installed versions:  0.6.4-r2(22:25:37 04/01/2010)(doc -examples)
>>>>>
>>>>> Container started with : lxc-start -d -n mycontainer
>>>>>
>>>>> I access to it via ssh.
>>>>>
>>>>> Just a question, config file is used at once at create?
>>>>>  
>>>>>      
>>>>>           
>>>>>>> root at mycontainer:~# ls /
>>>>>>> -bash: child setpgid (28212 to 28212): No such process
>>>>>>> bin  boot  dev  etc  home  lib  lib64  media  mnt  opt  proc  root
>>>>>>> sbin  selinux  srv  sys  tmp  usr  var
>>>>>>>                 
>>>>>>>               
>>>> When you are in the container, can you give the ouput of:
>>>>
>>>>  echo $$
>>>>  ps axjf
>>>>
>>>>
>>>>
>>>>     
>>>>         
>>> yes, I can :
>>>
>>>     root at mycontainer:~# ls
>>>     -bash: child setpgid (1905 to 1905): No such process
>>>     root at mycontainer:~# echo $$
>>>     74
>>>     root at mycontainer:~# ps axjf
>>>     -bash: child setpgid (1907 to 1907): No such process
>>>      PPID   PID  PGID   SID TTY      TPGID STAT   UID   TIME COMMAND
>>>         0     1     1     1 ?           -1 Ss       0   0:00 /sbin/init
>>>         1    13    10    10 ?           -1 Sl     101   0:00 rsyslogd
>>> -c4
>>>         1    54    54    54 ?           -1 Ss       0   0:00
>>> /usr/sbin/sshd
>>>         1    68    68    68 tty1        81 Ss       0   0:00 /bin/login
>>>     --           68    74    74    68 tty1        81 S        0  
>>> 0:00  \_ -bash
>>>        74    81    81    68 tty1        81 R+       0   0:00      \_
>>> ps axjf
>>>   
>>>       
>> Very weird ...
>>
>> Another one :)
>>
>> strace -f -eclone,setpgid bash
>> and then /bin/true (or whatever).
>>
>>
>>
>>     
> At same time, I was upgrading my kernel from 2.6.31 to 2.6.33.
> And it works now.
> I have done a rollback to reproduce. Clearly, my old kernel is the issue.
>   

Yep, I am suspecting a bug in the 2.6.31 kernel.

Do you mind to give the ouput of the strace for each kernel ?
Do you have a patched kernel ? I used the 2.6.31 but I don't recall 
falling onto this problem ...
> After some searches, it seems that my 2.6.31 kernel loosed 2 config
> items from my previous config :
> CONFIG_CGROUP_CPUACCT
> CONFIG_CGROUP_SCHED
>
> do you think this is the problem ?
>   
At the fist glance, I would say no,  I don't think this is in the 
"clone" / "copy_proces" kernel code path.