Re: [patch -mm] slab: use CPU_LOCK_[ACQUIRE|RELEASE]

[Srivatsa Vaddagiri <vatsa@in.ibm.com>]

On Wed, Jan 10, 2007 at 10:20:28AM -0800, Christoph Lameter wrote:
> I have got a bad feeling about upcoming deadlock problems when looking at
> the mutex_lock / unlock code in cpuup_callback in slab.c. Branches
> that just obtain a lock or release a lock? I hope there is some
> control of  what happens between lock acquisition and release?

A cpu hotplug should happen between LOCK_ACQUIRE/RELEASE

> You are aware that this lock is taken for cache shrinking/destroy, tuning
> of cpu cache sizes, proc output and cache creation? Any of those run on
> the same processor should cause a deadlock.

Why? mutex_lock() taken in LOCK_ACQ will just block those functions
(cache create etc) from proceeding simultaneously as a hotplug event.
This per-subsystem mutex_lock() is supposed to be a replacement for the global
lock_cpu_hotplug() lock .. 

But the whole thing is changing again ..we will likely move towards a
process freezer based cpu hotplug locking ..all the lock_cpu_hotplugs()
and the existing LOCK_ACQ/RELS can go away when we do that ..

-- 
Regards,
vatsa
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/