On 10/4/19 9:37 AM, Miles Chen wrote:
> In __set_page_owner_handle(), we should loop over page
> [0...(1 << order) - 1] and setup their page_owner structures.
>
> Currently, __set_page_owner_handle() update page_ext at the end of
> the loop, sets the page_owner of (page + 0) twice and
> misses the page_owner of (page + (1 << order) - 1).
>
> Fix it by updating the page_ext at the start of the loop.
>
> In i == 0 case:
> for (i = 0; i < (1 << order); i++) {
> page_owner = get_page_owner(page_ext); <- page_ext belongs to page + 0
> ...
> page_ext = lookup_page_ext(page + i); <- lookup_page_ext(page + 0)
> }
>
> Cc: Andrew Morton <[email protected]>
> Cc: Vlastimil Babka <[email protected]>
> Cc: Michal Hocko <[email protected]>
> Signed-off-by: Miles Chen <[email protected]>
> Fixes: 7e2f2a0cd17c ("mm, page_owner: record page owner for each subpage")
Thanks. Kirill spotted it earlier and there's a fix pending:
https://lore.kernel.org/linux-mm/[email protected]/
> ---
> mm/page_owner.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/mm/page_owner.c b/mm/page_owner.c
> index dee931184788..110c3e1987f2 100644
> --- a/mm/page_owner.c
> +++ b/mm/page_owner.c
> @@ -178,6 +178,7 @@ static inline void __set_page_owner_handle(struct page
> *page,
> int i;
>
> for (i = 0; i < (1 << order); i++) {
> + page_ext = lookup_page_ext(page + i);
> page_owner = get_page_owner(page_ext);
> page_owner->handle = handle;
> page_owner->order = order;
> @@ -185,8 +186,6 @@ static inline void __set_page_owner_handle(struct page
> *page,
> page_owner->last_migrate_reason = -1;
> __set_bit(PAGE_EXT_OWNER, &page_ext->flags);
> __set_bit(PAGE_EXT_OWNER_ACTIVE, &page_ext->flags);
> -
> - page_ext = lookup_page_ext(page + i);
> }
> }
>
>
