By checking huge_pte_none(), we incorrectly classify PTE markers as
"present". Instead, check huge_pte_none_mostly(), classifying PTE
markers the same as if the PTE were completely blank.
PTE markers, unlike other kinds of swap entries, don't reference any
physical page and don't indicate that a physical page was mapped
previously. As such, treat them as non-present for the sake of
mincore().
Fixes: 5c041f5d1f23 ("mm: teach core mm about pte markers")
Signed-off-by: James Houghton <[email protected]>
---
mm/mincore.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/mm/mincore.c b/mm/mincore.c
index cd69b9db0081..d359650b0f75 100644
--- a/mm/mincore.c
+++ b/mm/mincore.c
@@ -33,7 +33,7 @@ static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
* Hugepages under user process are always in RAM and never
* swapped out, but theoretically it needs to be checked.
*/
- present = pte && !huge_pte_none(huge_ptep_get(pte));
+ present = pte && !huge_pte_none_mostly(huge_ptep_get(pte));
for (; addr != end; vec++, addr += PAGE_SIZE)
*vec = present;
walk->private = vec;
--
2.40.0.rc0.216.gc4246ad0f0-goog
On Thu, Mar 02, 2023 at 10:24:04PM +0000, James Houghton wrote:
> By checking huge_pte_none(), we incorrectly classify PTE markers as
> "present". Instead, check huge_pte_none_mostly(), classifying PTE
> markers the same as if the PTE were completely blank.
>
> PTE markers, unlike other kinds of swap entries, don't reference any
> physical page and don't indicate that a physical page was mapped
> previously. As such, treat them as non-present for the sake of
> mincore().
>
> Fixes: 5c041f5d1f23 ("mm: teach core mm about pte markers")
> Signed-off-by: James Houghton <[email protected]>
Acked-by: Peter Xu <[email protected]>
Thanks, James.
--
Peter Xu
On Thu, 2 Mar 2023 17:32:48 -0500 Peter Xu <[email protected]> wrote:
> On Thu, Mar 02, 2023 at 10:24:04PM +0000, James Houghton wrote:
> > By checking huge_pte_none(), we incorrectly classify PTE markers as
> > "present". Instead, check huge_pte_none_mostly(), classifying PTE
> > markers the same as if the PTE were completely blank.
> >
> > PTE markers, unlike other kinds of swap entries, don't reference any
> > physical page and don't indicate that a physical page was mapped
> > previously. As such, treat them as non-present for the sake of
> > mincore().
> >
> > Fixes: 5c041f5d1f23 ("mm: teach core mm about pte markers")
> > Signed-off-by: James Houghton <[email protected]>
>
> Acked-by: Peter Xu <[email protected]>
>
I think I'll add cc:stable to this.
On 02.03.23 23:24, James Houghton wrote:
> By checking huge_pte_none(), we incorrectly classify PTE markers as
> "present". Instead, check huge_pte_none_mostly(), classifying PTE
> markers the same as if the PTE were completely blank.
>
> PTE markers, unlike other kinds of swap entries, don't reference any
> physical page and don't indicate that a physical page was mapped
> previously. As such, treat them as non-present for the sake of
> mincore().
>
> Fixes: 5c041f5d1f23 ("mm: teach core mm about pte markers")
> Signed-off-by: James Houghton <[email protected]>
> ---
> mm/mincore.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/mincore.c b/mm/mincore.c
> index cd69b9db0081..d359650b0f75 100644
> --- a/mm/mincore.c
> +++ b/mm/mincore.c
> @@ -33,7 +33,7 @@ static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
> * Hugepages under user process are always in RAM and never
> * swapped out, but theoretically it needs to be checked.
> */
> - present = pte && !huge_pte_none(huge_ptep_get(pte));
> + present = pte && !huge_pte_none_mostly(huge_ptep_get(pte));
> for (; addr != end; vec++, addr += PAGE_SIZE)
> *vec = present;
> walk->private = vec;
Acked-by: David Hildenbrand <[email protected]>
--
Thanks,
David / dhildenb