Re: Strange behaviour with threads in zombie state

[Andrew Doran <ad%netbsd.org@localhost>]

On Wed, Sep 24, 2008 at 05:00:20AM +0200, Juan RP wrote:

> Before loading the module, the system process shows up as:
> 
> $ ps axl|grep system     
>    0   0    0   0   0  0    0  13060 -       DKl  ?     0:00.05 [system]
> 1000 329  468   0  95  0  116    944 piperd  R+   ttyp1 0:00.00 grep system 
> $
> 
> Once the module is loaded, all continues to be good but when it's unloaded
> the system process is in zombie state, as well as all kernel threads.
> 
> $ sudo modload test.kmod                                                      
>                                           
> $ modstat |grep test                                                          
>                                   
> test            misc    filesys 0       224     -
> $ ps axl|grep system            
>    0   0    0   0 125  0    0  12856 schedule DKl  ?     0:00.06 [system]
> $ sudo modunload test
> $ ps axl|grep system  
>    0   0    0   0 125  0    0  12860 schedule ZKl  ?     0:00.06 [system]
> 1000 310  468   0  95  0  116    948 piperd   S+   ttyp1 0:00.00 grep system 
> $
> 
> I noticed the same thing with gkrellm, which once started its state
> is zombie and it's only shown in top with the lwp mode.
> 
> What's the reason of this? I don't think it's always been that way...

An exiting thread must wait to be reaped by another. At most one zombie
thread per-process can persist until reclaimed. That shouldn't cause the
entire process to be marked as a zombie though. Either ps or the sysctl
reporting is obtuse.

Andrew