Re: [PATCH] mm: move huge_pmd_set_accessed out of huge_memory.c

[Shi, Yang]

On 4/22/2016 2:48 AM, Kirill A. Shutemov wrote:
> On Thu, Apr 21, 2016 at 03:56:07PM -0700, Shi, Yang wrote:
> > On 4/21/2016 12:30 AM, Kirill A. Shutemov wrote:
> > > On Wed, Apr 20, 2016 at 02:00:11PM -0700, Shi, Yang wrote:
> > > > Hi folks,
> > > >
> > > > I didn't realize pmd_* functions are protected by
> > > > CONFIG_TRANSPARENT_HUGEPAGE on the most architectures before I made this
> > > > change.
> > > >
> > > > Before I fix all the affected architectures code, I want to check if you
> > > > guys think this change is worth or not?
> > > >
> > > > Thanks,
> > > > Yang
> > > >
> > > > On 4/20/2016 11:24 AM, Yang Shi wrote:
> > > > > huge_pmd_set_accessed is only called by __handle_mm_fault from memory.c,
> > > > > move the definition to memory.c and make it static like create_huge_pmd and
> > > > > wp_huge_pmd.
> > > > >
> > > > > Signed-off-by: Yang Shi <yang.shi@xxxxxxxxxx>
> > >
> > > On pte side we have the same functionality open-coded. Should we do the
> > > same for pmd? Or change pte side the same way?
> >
> > Sorry, I don't quite understand you. Do you mean pte_* functions?
>
> See handle_pte_fault(), we do the same for pte there what
> huge_pmd_set_accessed() does for pmd.

Thanks for directing to this code.

> I think we should be consistent here: either both are abstructed into
> functions or both open-coded.

I'm supposed functions sound better. However, do_wp_page has to be 
called with pte lock acquired. So, the abstracted function has to call it.

Thanks,
Yang


>