Re: [PATCHv6 36/36] thp: update documentation

[Vlastimil Babka]

On 06/22/2015 03:18 PM, Kirill A. Shutemov wrote:
> On Thu, Jun 11, 2015 at 02:30:15PM +0200, Vlastimil Babka wrote:
> > On 06/03/2015 07:06 PM, Kirill A. Shutemov wrote:
> > > The patch updates Documentation/vm/transhuge.txt to reflect changes in
> > > THP design.
> >
> > One thing I'm missing is info about the deferred splitting.
>
> Okay, I'll add this.

Thanks.

> > > Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> > > ---
> > >   Documentation/vm/transhuge.txt | 124 +++++++++++++++++++++++------------------
> > >   1 file changed, 69 insertions(+), 55 deletions(-)
> > >
> > > diff --git a/Documentation/vm/transhuge.txt b/Documentation/vm/transhuge.txt
> > > index 6b31cfbe2a9a..2352b12cae93 100644
> > > --- a/Documentation/vm/transhuge.txt
> > > +++ b/Documentation/vm/transhuge.txt
> > > @@ -35,10 +35,10 @@ miss is going to run faster.
> > >
> > >   == Design ==
> > >
> > > -- "graceful fallback": mm components which don't have transparent
> > > -  hugepage knowledge fall back to breaking a transparent hugepage and
> > > -  working on the regular pages and their respective regular pmd/pte
> > > -  mappings
> > > +- "graceful fallback": mm components which don't have transparent hugepage
> > > +  knowledge fall back to breaking huge pmd mapping into table of ptes and,
> > > +  if nesessary, split a transparent hugepage. Therefore these components
> >
> >          necessary
> > > +
> > > +split_huge_page uses migration entries to stabilize page->_count and
> > > +page->_mapcount.
> >
> > Hm, what if there's some physical memory scanner taking page->_count pins? I
> > think compaction shouldn't be an issue, but maybe some others?
>
> The only legitimate way scanner can get reference to a page is
> get_page_unless_zero(), right?

I think so.

> All tail pages has zero ->_count until atomic_add() in
> __split_huge_page_tail() -- get_page_unless_zero() will fail.
> After the atomic_add() we don't care about ->_count value.
> We already known how many references with should uncharge from
> head page.
>
> For head page get_page_unless_zero() will succeed and we don't
> mind. It's clear where reference should go after split: it will
> stay on head page.

I guess that works, but it's rather non-obvious thus worth documenting 
somewhere.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
Please read the FAQ at  http://www.tux.org/lkml/