[Ocfs2-tools-devel] [PATCH 2/2] o2hbmonitor: Limit number of active instances

[Tristan Ye]

Sunil Mushran wrote:
> On 12/08/2010 06:23 PM, Tristan Ye wrote:
>>    I'm just wondering how you can go with POSIX semaphores on known 
>> 'SIGKILL'
>> problem, it was just incapable of undoing the locks after SIGKILL 
>> being delivered.
>> Also, my testing wasn't supportive to your patches as well.
>>
>>    Like what Joel said, System V semaphores have no troubling deal 
>> with the issue
>> by 'SEM_UNDO'.
>
> So you see /dev/shm/sem.o2hbmonitor after kill -9?
>
> If that is the case, then sem_undo patch will be the best approach, imho.

Yep,

    I did see the named semaphores right there after kill -9, just 
curious why you succeeded,
Or you were just issuing 'kill' instead of 'kill -9'?