Re: [patch 9/9] mutex subsystem, XFS namespace collision fixes

[Ingo Molnar <mingo@elte.hu>]

* Christoph Hellwig <hch@infradead.org> wrote:

> > -#define mutex_init(lock, type, name)		sema_init(lock, 1)
> > -#define mutex_destroy(lock)			sema_init(lock, -99)
> > -#define mutex_lock(lock, num)			down(lock)
> > -#define mutex_trylock(lock)			(down_trylock(lock) ? 0 : 1)
> > -#define mutex_unlock(lock)			up(lock)
> > +#define xfs_mutex_init(lock, type, name)	sema_init(lock, 1)
> > +#define xfs_mutex_destroy(lock)			sema_init(lock, -99)
> > +#define xfs_mutex_lock(lock, num)		down(lock)
> > +#define xfs_mutex_trylock(lock)			(down_trylock(lock) ? 0 : 1)
> > +#define xfs_mutex_unlock(lock)			up(lock)
> 
> Again, this should really be using the mutex primitives (obviously 
> ;-)).

yeah - but i didnt want to impact something so large as XFS. Such a 
change has to be tested and validated - so i wanted to get the namespace 
collision out of the way first. But i'd be happy to add an XFS 
conversion patch ontop of these, provided someone tests it.

> While we're at it, maybe we should a mutex_destroy aswell?  it would 
> be non-mandatory and allow that a lock is gone for the debugging 
> variant.

right now the lock is gone from the debugging state once it's unlocked.  
I'll add mutex_destroy(), it should be rather easy (it can e.g. destroy 
mutex->magic).

	Ingo