Re: [PATCH] mm: hugetlb: flush dcache before returning zeroed huge page to userspace

[Will Deacon <will.deacon@arm.com>]

On Thu, Jul 12, 2012 at 12:16:59PM +0100, Michal Hocko wrote:
> On Wed 11-07-12 18:48:02, Will Deacon wrote:
> > Just to confirm, the following quick hack at least results in the correct
> > flushing for me (on ARM):
> > 
> > 
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index e198831..7a7c9d3 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -1141,6 +1141,7 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
> >         }
> >  
> >         set_page_private(page, (unsigned long)spool);
> > +       clear_bit(PG_arch_1, &page->flags);
> >  
> >         vma_commit_reservation(h, vma, addr);
> >  
> > 
> > 
> > The question is whether we should tidy that up for the core code or get
> > architectures to clear the bit in arch_make_huge_pte (which also seems to
> > work).
> 
> This should go into arch specific code IMO. Even the page flag name
> suggests this shouldn't be in the base code.

Well, the comment in linux/page-flags.h does state that:

 * PG_arch_1 is an architecture specific page state bit.  The generic code
 * guarantees that this bit is cleared for a page when it first is entered into
 * the page cache.

so it's not completely clear cut that the architecture should be responsible
for clearing this bit when allocating pages from the hugepage pool.

Will