On Mon 13-05-19 10:37:01, Weikang shi wrote:
> From: swkhack <swkhack@xxxxxxxxx>
>
> In 64-bit machine,the value of "vma->vm_end - vma->vm_start"
> maybe negative in 32bit int and the "count >> PAGE_SHIFT"'s result
> will be wrong.So change the local variable and return
> value to unsigned long will fix the problem.
>
> Signed-off-by: swkhack <swkhack@xxxxxxxxx>
Fixes: 0cf2f6f6dc60 ("mm: mlock: check against vma for actual mlock() size")
Acked-by: Michal Hocko <mhocko@xxxxxxxx>
Most users probably never noticed because large mlocked areas are not
allowed by default. So I am not really sure this is worth backporting to
stable trees.
> ---
> mm/mlock.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/mm/mlock.c b/mm/mlock.c
> index 080f3b364..d614163f5 100644
> --- a/mm/mlock.c
> +++ b/mm/mlock.c
> @@ -636,11 +636,11 @@ static int apply_vma_lock_flags(unsigned long start, size_t len,
> * is also counted.
> * Return value: previously mlocked page counts
> */
> -static int count_mm_mlocked_page_nr(struct mm_struct *mm,
> +static unsigned long count_mm_mlocked_page_nr(struct mm_struct *mm,
> unsigned long start, size_t len)
> {
> struct vm_area_struct *vma;
> - int count = 0;
> + unsigned long count = 0;
>
> if (mm == NULL)
> mm = current->mm;
> --
> 2.17.1
--
Michal Hocko
SUSE Labs
