On Sat 19-03-22 18:42:33, Miaohe Lin wrote:
[...]
> This would be triggered easily with below code snippet in my virtual machine:
>
> shmid = shmget((key_t)5566, 1024 * PAGE_SIZE, 0666|IPC_CREAT);
> shm = shmat(shmid, 0, 0);
> loop {
> mbind(shm, 1024 * PAGE_SIZE, MPOL_LOCAL, mask, maxnode, 0);
> mbind(shm + 128 * PAGE_SIZE, 128 * PAGE_SIZE, MPOL_DEFAULT, mask, maxnode, 0);
> }
>
> If there're many process doing the above work, mpol_new will be leaked easily.
> So should I resend this patch with Cc stable? But it seems I'am not supposed
> to make this decision and the maintainer will take care of this?
I would just add
Fixes: 42288fe366c4 ("mm: mempolicy: Convert shared_policy mutex to spinlock")
Cc: stable # 3.8
And also add your above reproducer snippet added to the original changelog.
This would be more then enough to conclude the importance.
Thank you for working hard on this!
--
Michal Hocko
SUSE Labs
