Re: [RFC PATCH v5 1/2] Use lock_device_hotplug() in cpu_probe_store() and cpu_release_store()

[Li Zhong <zhong@linux.vnet.ibm.com>]

On Fri, 2014-04-25 at 08:28 -0400, Tejun Heo wrote:
> On Fri, Apr 25, 2014 at 09:56:10AM +0800, Li Zhong wrote:
> > > A nests cpu subsys mutex under s_active of the online node.  B nests
> > > s_active of the online node under the cpu subsys mutex.  What am I
> > > missing?
> > 
> > From the above two chain, I think the problem is cpu_subsys_mutex and
> > s_active(online), which is the deadlock we are trying to solve in patch
> 
> Yeap, that one was what I was thinking about.
> 
> > #2. I seems to me s_active(release) here doesn't have lock issues? 
> 
> And your patch doesn't change the situation for online.  

I'm a little confused here...

To clarify for this patch(#1): 

Currently(before patch#2), the s_active(online) and cpu subsys mutex
deadlock is solved by lock_device_hotplug_sysfs(). (lockdep warnings are
still there, so we want to change the implementation of
lock_device_hotplug_sysfs() in patch #2 to take out of s_active_online,
instead of device_hotplug_lock, which is grabbed before cpu subsys
mutex). 

But here, for the probe/release, I think we really don't need to use
this _sysfs() version. 

So it is replaced with lock_device_hotplug() directly to avoid two
unnecessary conversions of lock_device_hotplug_sysfs(), that otherwise
need to be done in patch #2.

Thanks, Zhong

> Alright,
> thanks a lot for the explanation.
>