Re: [PATCH 02/10] mm,madvise: call soft_offline_page() without MF_COUNT_INCREASED

[Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>]

Hi David,

On Wed, Sep 11, 2019 at 12:23:24PM +0200, David Hildenbrand wrote:
> On 10.09.19 12:30, Oscar Salvador wrote:
> > From: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> > 
> > Currently madvise_inject_error() pins the target via get_user_pages_fast.
> > The call to get_user_pages_fast is only to get the respective page
> > of a given address, but it is the job of the memory-poisoning handler
> > to deal with races, so drop the refcount grabbed by get_user_pages_fast.
> > 
> > Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> > Signed-off-by: Oscar Salvador <osalvador@suse.de>
> > ---
> >  mm/madvise.c | 25 +++++++++++--------------
> >  1 file changed, 11 insertions(+), 14 deletions(-)
> > 
> > diff --git a/mm/madvise.c b/mm/madvise.c
> > index 6e023414f5c1..fbe6d402232c 100644
> > --- a/mm/madvise.c
> > +++ b/mm/madvise.c
> > @@ -883,6 +883,16 @@ static int madvise_inject_error(int behavior,
> >  		ret = get_user_pages_fast(start, 1, 0, &page);
> >  		if (ret != 1)
> >  			return ret;
> > +		/*
> > +		 * The get_user_pages_fast() is just to get the pfn of the
> > +		 * given address, and the refcount has nothing to do with
> > +		 * what we try to test, so it should be released immediately.
> > +		 * This is racy but it's intended because the real hardware
> > +		 * errors could happen at any moment and memory error handlers
> > +		 * must properly handle the race.
> > +		 */
> > +		put_page(page);
> > +
> 
> I wonder if it would be clearer to do that after the page has been fully
> used  - e.g. after getting the pfn and the order (and then e.g.,
> symbolically setting the page pointer to 0).

Yes, this could be called just after page_to_pfn() below.

> I guess the important part of this patch is to not have an elevated
> refcount while calling soft_offline_page().
> 

That's right.

> >  		pfn = page_to_pfn(page);
> >  
> >  		/*
> > @@ -892,16 +902,11 @@ static int madvise_inject_error(int behavior,
> >  		 */
> >  		order = compound_order(compound_head(page));
> >  
> > -		if (PageHWPoison(page)) {
> > -			put_page(page);
> > -			continue;
> > -		}
> 
> This change is not reflected in the changelog. I would have expected
> that only the put_page() would go. If this should go completely, I
> suggest a separate patch.
> 

I forget why I tried to remove the if block, and now I think only the
put_page() should go as you point out.

Thanks for the comment.

- Naoya