On Mon, 2012-06-11 at 05:17 -0400, kosaki.motohiro@xxxxxxxxx wrote:
> From: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
>
> commit cc9a6c8776 (cpuset: mm: reduce large amounts of memory barrier related
> damage v3) introduced a memory corruption.
>
> shmem_alloc_page() passes pseudo vma and it has one significant unique
> combination, vma->vm_ops=NULL and (vma->policy->flags & MPOL_F_SHARED).
>
> Now, get_vma_policy() does NOT increase a policy ref when vma->vm_ops=NULL
> and mpol_cond_put() DOES decrease a policy ref when a policy has MPOL_F_SHARED.
> Therefore, when alloc_pages_vma() goes 'goto retry_cpuset' path, a policy
> refcount will be decreased too much and therefore it will make a memory corruption.
[...]
> --- a/mm/mempolicy.c
> +++ b/mm/mempolicy.c
> @@ -1544,18 +1544,29 @@ struct mempolicy *get_vma_policy(struct task_struct *task,
> struct vm_area_struct *vma, unsigned long addr)
> {
> struct mempolicy *pol = task->mempolicy;
> + int got_ref;
= 0
And this should really be a bool.
> if (vma) {
> if (vma->vm_ops && vma->vm_ops->get_policy) {
> struct mempolicy *vpol = vma->vm_ops->get_policy(vma,
> addr);
> - if (vpol)
> + if (vpol) {
> pol = vpol;
> + got_ref = 1;
> + }
> } else if (vma->vm_policy)
> pol = vma->vm_policy;
> }
> if (!pol)
> pol = &default_policy;
> +
> + /*
> + * shmem_alloc_page() passes MPOL_F_SHARED policy with vma->vm_ops=NULL.
> + * Thus, we need to take additional ref for avoiding refcount imbalance.
> + */
> + if (!got_ref && mpol_needs_cond_ref(pol))
> + mpol_get(pol);
> +
> return pol;
> }
>
[...]
--
Ben Hutchings
Computers are not intelligent. They only think they are.
Attachment:
signature.asc
Description: This is a digitally signed message part