Messages in this thread | | | Date | Wed, 18 Aug 2010 12:02:00 +0900 | From | Naoya Horiguchi <> | Subject | Re: [PATCH 2/9] hugetlb: add allocate function for hugepage migration |
| |
Hi,
On Mon, Aug 16, 2010 at 11:51:30PM -0700, David Rientjes wrote: > On Tue, 10 Aug 2010, Naoya Horiguchi wrote: ... > > +/* > > + * This allocation function is useful in the context where vma is irrelevant. > > + * E.g. soft-offlining uses this function because it only cares physical > > + * address of error page. > > + */ > > +struct page *alloc_huge_page_no_vma_node(struct hstate *h, int nid) > > +{ > > + struct page *page; > > + > > + spin_lock(&hugetlb_lock); > > + get_mems_allowed(); > > Why is this calling get_mems_allowed()? dequeue_huge_page_node() isn't > concerned if nid can be allocated by current in this context.
OK, I'll remove this.
> > + page = dequeue_huge_page_node(h, nid); > > + put_mems_allowed(); > > + spin_unlock(&hugetlb_lock); > > + > > + if (!page) { > > + page = alloc_buddy_huge_page_node(h, nid); > > + if (!page) { > > + __count_vm_event(HTLB_BUDDY_PGALLOC_FAIL); > > return NULL; > > - } > > - prep_new_huge_page(h, page, nid); > > + } else > > + __count_vm_event(HTLB_BUDDY_PGALLOC); > > } > > > > + set_page_refcounted(page); > > Possibility of NULL pointer dereference?
I think this allocate function returns without calling set_page_refcounted() if page == NULL. Or do you mean another point?
Thanks, Naoya Horiguchi
| |