Re: [patch 1/2] read_barrier_depends fixlets

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

> Nick Piggin <npiggin@suse.de> wrote:
> 
> > While considering the impact of read_barrier_depends, it occurred to
> > me that it should really be really a noop for the compiler.
> 
> If you're defining it so, then you need to adjust memory-barriers.txt too.
> 
> 	========================
> 	EXPLICIT KERNEL BARRIERS
> 	========================
> 	...
> 	CPU MEMORY BARRIERS
> 	-------------------
> 
> 	The Linux kernel has eight basic CPU memory barriers:
> 
> 		TYPE		MANDATORY		SMP CONDITIONAL
> 		===============	=======================	===========================
> 		GENERAL		mb()			smp_mb()
> 		WRITE		wmb()			smp_wmb()
> 		READ		rmb()			smp_rmb()
> 		DATA DEPENDENCY	read_barrier_depends()	smp_read_barrier_depends()
> 
> 
> 	All CPU memory barriers unconditionally imply compiler barriers.
> 
> That last line needs modification, perhaps to say:
> 
> 	General, read and write memory barriers unconditionally imply general
> 	compiler barriers; data dependency barriers, however, imply a barrier
> 	only for the specific access being performed due to the fact that the
> 	instructions must be performed in a specific order.

And to make sure the compiler preserves the ordering, you also need
the ACCESS_ONCE() in the general case.

							Thanx, Paul