Re: [PATCH v4 3/9] liblockdep: Add public headers for pthread_mutex_t implementation

[Peter Zijlstra <peterz@infradead.org>]

On Tue, May 28, 2013 at 03:33:51PM -0400, Sasha Levin wrote:
> On 05/22/2013 05:22 AM, Peter Zijlstra wrote:
> > They will however then also want all the 'normal' lockdep annotations to
> > deal with that like:
> > 
> > liblockdep_pthread_mutex_lock_nested()
> > liblockdep_pthread_mutex_lock_nest_lock()
> > 
> > *phew* and here I always though pthread_mutex_* was a long prefix...
> > 
> > Also, the above doesn't have the full lockstat contention hooks -- not
> > sure that's on purpose or not.
> 
> I was quietly hoping on leaving this out in the initial version of liblockdep
> and start adding this and the rest of the toys that come with lockdep once we
> figure out whether this code will go into the kernel tree or not.
> 
> Should I be adding them now?

I think you'll need them very quicky one you actually start using this
stuff, but sure, add them when you need them.

> >> +
> >> +static inline int liblockdep_pthread_mutex_unlock(liblockdep_pthread_mutex_t *lock)
> >> +{
> >> +	lock_release(&lock->dep_map, 0, (unsigned long)_RET_IP_);
> >> +	return pthread_mutex_unlock(&lock->mutex);
> >> +}
> >> +
> >> +static inline int liblockdep_pthread_mutex_trylock(liblockdep_pthread_mutex_t *lock)
> >> +{
> >> +	lock_acquire(&lock->dep_map, 0, 1, 0, 2, NULL, (unsigned long)_RET_IP_);
> >> +	return pthread_mutex_trylock(&lock->mutex) == 0 ? 1 : 0;
> >> +}
> >> +
> >> +static inline int liblockdep_pthread_mutex_destroy(liblockdep_pthread_mutex_t *lock)
> >> +{
> >> +	return pthread_mutex_destroy(&lock->mutex);
> >> +}
> >> +
> >> +#ifdef __USE_LIBLOCKDEP
> >> +
> >> +#define pthread_mutex_t         liblockdep_pthread_mutex_t
> >> +#define pthread_mutex_init      liblockdep_pthread_mutex_init
> >> +#define pthread_mutex_lock      liblockdep_pthread_mutex_lock
> >> +#define pthread_mutex_unlock    liblockdep_pthread_mutex_unlock
> >> +#define pthread_mutex_trylock   liblockdep_pthread_mutex_trylock
> >> +#define pthread_mutex_destroy   liblockdep_pthread_mutex_destroy
> > 
> > Given the liblockdep_* things use 'proper' classes do you want this
> > mapping? If you do, should we use the same alias nonsense glibc uses or
> > are CPP macros good enough for us?
> > 
> 
> I think that this will be good enough for our purpose, why wouldn't these
> simple macros be enough?

Suppose you have a funny someone who added a function called:
pthread_mutex_lock_obj() or somesuch, imagine what the CPP thing will
make of that :-)

Then again, people doing that might deserve what they get ;-)