lkml.org 
[lkml]   [2015]   [Feb]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRE: [RFC] mm: change mm_advise_free to clear page dirty
> -----Original Message-----
> From: Minchan Kim [mailto:minchan.kim@gmail.com] On Behalf Of Minchan Kim
> Sent: Friday, February 27, 2015 2:44 PM
> To: Wang, Yalin
> Cc: Michal Hocko; Andrew Morton; linux-kernel@vger.kernel.org; linux-
> mm@kvack.org; Rik van Riel; Johannes Weiner; Mel Gorman; Shaohua Li
> Subject: Re: [RFC] mm: change mm_advise_free to clear page dirty
>
> On Fri, Feb 27, 2015 at 01:48:48PM +0800, Wang, Yalin wrote:
> > > -----Original Message-----
> > > From: Minchan Kim [mailto:minchan.kim@gmail.com] On Behalf Of Minchan
> Kim
> > > Sent: Friday, February 27, 2015 1:28 PM
> > > To: Wang, Yalin
> > > Cc: Michal Hocko; Andrew Morton; linux-kernel@vger.kernel.org; linux-
> > > mm@kvack.org; Rik van Riel; Johannes Weiner; Mel Gorman; Shaohua Li
> > > Subject: Re: [RFC] mm: change mm_advise_free to clear page dirty
> > >
> > > Hello,
> > >
> > > On Fri, Feb 27, 2015 at 11:37:18AM +0800, Wang, Yalin wrote:
> > > > This patch add ClearPageDirty() to clear AnonPage dirty flag,
> > > > the Anonpage mapcount must be 1, so that this page is only used by
> > > > the current process, not shared by other process like fork().
> > > > if not clear page dirty for this anon page, the page will never be
> > > > treated as freeable.
> > >
> > > In case of anonymous page, it has PG_dirty when VM adds it to
> > > swap cache and clear it in clear_page_dirty_for_io. That's why
> > > I added ClearPageDirty if we found it in swapcache.
> > > What case am I missing? It would be better to understand if you
> > > describe specific scenario.
> > >
> > > Thanks.
> > >
> > > >
> > > > Signed-off-by: Yalin Wang <yalin.wang@sonymobile.com>
> > > > ---
> > > > mm/madvise.c | 15 +++++----------
> > > > 1 file changed, 5 insertions(+), 10 deletions(-)
> > > >
> > > > diff --git a/mm/madvise.c b/mm/madvise.c
> > > > index 6d0fcb8..257925a 100644
> > > > --- a/mm/madvise.c
> > > > +++ b/mm/madvise.c
> > > > @@ -297,22 +297,17 @@ static int madvise_free_pte_range(pmd_t *pmd,
> > > unsigned long addr,
> > > > continue;
> > > >
> > > > page = vm_normal_page(vma, addr, ptent);
> > > > - if (!page)
> > > > + if (!page || !PageAnon(page) || !trylock_page(page))
> > > > continue;
> > > >
> > > > if (PageSwapCache(page)) {
> > > > - if (!trylock_page(page))
> > > > + if (!try_to_free_swap(page))
> > > > continue;
> > > > -
> > > > - if (!try_to_free_swap(page)) {
> > > > - unlock_page(page);
> > > > - continue;
> > > > - }
> > > > -
> > > > - ClearPageDirty(page);
> > > > - unlock_page(page);
> > > > }
> > > >
> > > > + if (page_mapcount(page) == 1)
> > > > + ClearPageDirty(page);
> > > > + unlock_page(page);
> > > > /*
> > > > * Some of architecture(ex, PPC) don't update TLB
> > > > * with set_pte_at and tlb_remove_tlb_entry so for
> > > > --
> > Yes, for page which is in SwapCache, it is correct,
> > But for anon page which is not in SwapCache, it is always
> > PageDirty(), so we should also clear dirty bit to make it freeable,
>
> No. Every anon page starts from !PageDirty and it has PG_dirty
> only when it's addeded into swap cache. If vm_swap_full turns on,
> a page in swap cache could have PG_dirty via try_to_free_swap again.

mmm..
sometimes you can see an anon page PageDirty(), but it is not in swapcache,
for example, handle_pte_fault()-->do_swap_page()-->try_to_free_swap(),
at this time, the page is deleted from swapcache and is marked PageDirty(),


> So, Do you have concern about swapped-out pages when MADV_FREE is
> called? If so, please look at my patch.
>
> https://lkml.org/lkml/2015/2/25/43
>
> It will zap the swapped out page. So, this is not a issue any more?
>
> >
> > Another problem is that if an anon page is shared by more than one
> process,
> > This happened when fork(), the anon page will be copy on write,
> > In this case, we should not clear page dirty,
> > This is not correct for other process which don't call MADV_FREE syscall.
>
> You mean we shouldn't inherit MADV_FREE attribute?
> Why?

Is it correct behavior if code like this:

Parent:
ptr1 = malloc(len);
memset(ptr1, 'a', len);
fork();
if (I am parent)
madvise_free(ptr1, len);

child:
sleep(10);
parse_data(ptr1, len); // child may see zero, not 'a',
// is it the right behavior that the programer want?

Because child don't call madvise_free(), so it should see 'a', not zero page.
Isn't it ?
Thanks








\
 
 \ /
  Last update: 2015-02-27 09:01    [W:0.331 / U:0.160 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site