Hi Jiri,
On Fri, Jun 20, 2014 at 09:13:32AM +0200, Jiri Slaby wrote:
> From: Naoya Horiguchi <[email protected]>
>
> This patch does NOT apply to the 3.12 stable tree. If you still want
> it applied, please provide a backport.
At kernel 3.12 queue_pages_hugetlb_pmd_range() used vma->vm_mm->page_table_lock
for page table lock, which was introduced at commit e2d8cf40552 "migrate: add
hugepage migration code to migrate_pages()".
After 3.12 in mainline kernel we had the lock conversion into split pmd lock
at commit cb900f4121544 "mm, hugetlb: convert hugetlbfs to use split pmd lock".
Stable-3.12 doesn't have split pmd lock, so just adding pte_present check
with keeping vma->vm_mm->page_table_lock should be correct.
Here is the diff:
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 390bdab01c3c..ad4df869c907 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -1353,7 +1353,7 @@ static int gather_hugetbl_stats(pte_t *pte, unsigned long
hmask,
struct numa_maps *md;
struct page *page;
- if (pte_none(*pte))
+ if (!pte_present(*pte))
return 0;
page = pte_page(*pte);
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 927a69cf354a..a005cc9f6f18 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -525,9 +525,13 @@ static void queue_pages_hugetlb_pmd_range(struct
vm_area_struct *vma,
#ifdef CONFIG_HUGETLB_PAGE
int nid;
struct page *page;
+ pte_t entry;
spin_lock(&vma->vm_mm->page_table_lock);
- page = pte_page(huge_ptep_get((pte_t *)pmd));
+ entry = huge_ptep_get((pte_t *)pmd);
+ if (!pte_present(entry))
+ goto unlock;
+ page = pte_page(entry);
nid = page_to_nid(page);
if (node_isset(nid, *nodes) == !!(flags & MPOL_MF_INVERT))
goto unlock;
--
Thanks,
Naoya Horiguchi
> ===============
>
> commit d4c54919ed86302094c0ca7d48a8cbd4ee753e92 upstream.
>
> The age table walker doesn't check non-present hugetlb entry in common
> path, so hugetlb_entry() callbacks must check it. The reason for this
> behavior is that some callers want to handle it in its own way.
>
> [ I think that reason is bogus, btw - it should just do what the regular
> code does, which is to call the "pte_hole()" function for such hugetlb
> entries - Linus]
>
> However, some callers don't check it now, which causes unpredictable
> result, for example when we have a race between migrating hugepage and
> reading /proc/pid/numa_maps. This patch fixes it by adding !pte_present
> checks on buggy callbacks.
>
> This bug exists for years and got visible by introducing hugepage
> migration.
>
> ChangeLog v2:
> - fix if condition (check !pte_present() instead of pte_present())
>
> Reported-by: Sasha Levin <[email protected]>
> Signed-off-by: Naoya Horiguchi <[email protected]>
> Cc: Rik van Riel <[email protected]>
> Cc: <[email protected]> [3.12+]
> Signed-off-by: Andrew Morton <[email protected]>
> [ Backported to 3.15. Signed-off-by: Josh Boyer <[email protected]> ]
> Signed-off-by: Linus Torvalds <[email protected]>
> ---
> fs/proc/task_mmu.c | 2 +-
> mm/mempolicy.c | 6 +++++-
> 2 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
> index 442177b1119a..c4b2646b6d7c 100644
> --- a/fs/proc/task_mmu.c
> +++ b/fs/proc/task_mmu.c
> @@ -1351,7 +1351,7 @@ static int gather_hugetbl_stats(pte_t *pte, unsigned
> long hmask,
> struct numa_maps *md;
> struct page *page;
>
> - if (pte_none(*pte))
> + if (!pte_present(*pte))
> return 0;
>
> page = pte_page(*pte);
> diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> index 78e1472933ea..30cc47f8ffa0 100644
> --- a/mm/mempolicy.c
> +++ b/mm/mempolicy.c
> @@ -526,9 +526,13 @@ static void queue_pages_hugetlb_pmd_range(struct
> vm_area_struct *vma,
> int nid;
> struct page *page;
> spinlock_t *ptl;
> + pte_t entry;
>
> ptl = huge_pte_lock(hstate_vma(vma), vma->vm_mm, (pte_t *)pmd);
> - page = pte_page(huge_ptep_get((pte_t *)pmd));
> + entry = huge_ptep_get((pte_t *)pmd);
> + if (!pte_present(entry))
> + goto unlock;
> + page = pte_page(entry);
> nid = page_to_nid(page);
> if (node_isset(nid, *nodes) == !!(flags & MPOL_MF_INVERT))
> goto unlock;
> --
> 2.0.0
>
>
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
