[PATCH 3/4] mm: numa: Mark huge PTEs young when clearing NUMA hinting faults

[Linus Torvalds]

On Sat, Mar 7, 2015 at 7:20 AM, Mel Gorman <mgorman at suse.de> wrote:
>         pmd = pmd_modify(pmd, vma->vm_page_prot);
> +       pmd = pmd_mkyoung(pmd);

Hmm. I *thought* this should be unnecessary. vm_page_prot alreadty has
the accessed bit set, and we kind of depend on the initial page table
setup and mk_pte() and friends (ie all new pages are installed
"young").

But it looks like I am wrong - the way we use _[H]PAGE_CHG_MASK means
that we always take the accessed and dirty bits from the old entry,
ignoring the bit in vm_page_prot.

I wonder if we should just make pte/pmd_modify() work the way I
*thought* they worked (remove the masking of vm_page_prot bits).

So the patch isn't wrong. It's just that we *migth* instead just do
something like this:

    arch/x86/include/asm/pgtable.h | 4 ++--
    1 file changed, 2 insertions(+), 2 deletions(-)

   diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
   index a0c35bf6cb92..79b898bb9e18 100644
   --- a/arch/x86/include/asm/pgtable.h
   +++ b/arch/x86/include/asm/pgtable.h
   @@ -355,7 +355,7 @@ static inline pte_t pte_modify(pte_t pte,
pgprot_t newprot)
            * the newprot (if present):
            */
           val &= _PAGE_CHG_MASK;
   -       val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
   +       val |= massage_pgprot(newprot);

           return __pte(val);
    }
   @@ -365,7 +365,7 @@ static inline pmd_t pmd_modify(pmd_t pmd,
pgprot_t newprot)
           pmdval_t val = pmd_val(pmd);

           val &= _HPAGE_CHG_MASK;
   -       val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
   +       val |= massage_pgprot(newprot);

           return __pmd(val);
    }

instead, and remove the mkyoung. Completely untested, but that "just
or in the new protection bits" is what pnf_pte() does just a few lines
above this.

                        Linus