Re: [PATCH 1/1] powerpc/mm: Add memory barrier in __hugepte_alloc()

[Michal Hocko]

On Wed 06-04-16 15:39:17, Aneesh Kumar K.V wrote:
> Michal Hocko <mhocko@xxxxxxxxxx> writes:
> 
> > [ text/plain ]
> > On Tue 05-04-16 12:05:47, Sukadev Bhattiprolu wrote:
> > [...]
> >> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
> >> index d991b9e..081f679 100644
> >> --- a/arch/powerpc/mm/hugetlbpage.c
> >> +++ b/arch/powerpc/mm/hugetlbpage.c
> >> @@ -81,6 +81,13 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
> >>  	if (! new)
> >>  		return -ENOMEM;
> >>  
> >> +	/*
> >> +	 * Make sure other cpus find the hugepd set only after a
> >> +	 * properly initialized page table is visible to them.
> >> +	 * For more details look for comment in __pte_alloc().
> >> +	 */
> >> +	smp_wmb();
> >> +
> >
> > what is the pairing memory barrier?
> >
> >>  	spin_lock(&mm->page_table_lock);
> >>  #ifdef CONFIG_PPC_FSL_BOOK3E
> >>  	/*
> 
> This is documented in __pte_alloc(). I didn't want to repeat the same
> here.
> 
> 	/*
> 	 * Ensure all pte setup (eg. pte page lock and page clearing) are
> 	 * visible before the pte is made visible to other CPUs by being
> 	 * put into page tables.
> 	 *
> 	 * The other side of the story is the pointer chasing in the page
> 	 * table walking code (when walking the page table without locking;
> 	 * ie. most of the time). Fortunately, these data accesses consist
> 	 * of a chain of data-dependent loads, meaning most CPUs (alpha
> 	 * being the notable exception) will already guarantee loads are
> 	 * seen in-order. See the alpha page table accessors for the
> 	 * smp_read_barrier_depends() barriers in page table walking code.
> 	 */
> 	smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */

OK, I have missed the reference to __pte_alloc. My bad!

-- 
Michal Hocko
SUSE Labs