On 22.11.19 23:41, Navid Emamdoost wrote:
> In the implementation of __gup_benchmark_ioctl() memory is leaked if the
> passed cmd is invalid. Release pages before returning -1.
>
> Fixes: 714a3a1ebafe ("mm/gup_benchmark.c: add additional pinning methods")
> Signed-off-by: Navid Emamdoost <navid.emamdoost@xxxxxxxxx>
> ---
> mm/gup_benchmark.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/mm/gup_benchmark.c b/mm/gup_benchmark.c
> index 7dd602d7f8db..33ede5727523 100644
> --- a/mm/gup_benchmark.c
> +++ b/mm/gup_benchmark.c
> @@ -23,7 +23,7 @@ static int __gup_benchmark_ioctl(unsigned int cmd,
> struct gup_benchmark *gup)
> {
> ktime_t start_time, end_time;
> - unsigned long i, nr_pages, addr, next;
> + unsigned long i, j, nr_pages, addr, next;
> int nr;
> struct page **pages;
>
> @@ -63,6 +63,12 @@ static int __gup_benchmark_ioctl(unsigned int cmd,
> NULL);
> break;
> default:
> + for (j = 0; j < i; j++) {
> + if (!pages[j])
> + break;
> + put_page(pages[j]);
> + }
We didn't pin any pages, why should we release them?
IMHO, all that's needed is the ...
> + kvfree(pages);
.. here
> return -1;
> }
>
Not sure how often CONFIG_GUP_BENCHMARK is really enabled. Maybe we should
Cc: stable@xxxxxxxxxxxxxxx # v4.20+
With only the kvfree(pages)
Reviewed-by: David Hildenbrand <david@xxxxxxxxxx>
--
Thanks,
David / dhildenb
