[PATCH v3 12/13] mm: /proc/pid/clear_refs: avoid split_huge_page()

[Naoya Horiguchi]

From: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>

Currently pagewalker splits all THP pages on any clear_refs request.  It's
not necessary.  We can handle this on PMD level.

One side effect is that soft dirty will potentially see more dirty memory,
since we will mark whole THP page dirty at once.

Sanity checked with CRIU test suite. More testing is required.

ChangeLog:
- move code for thp to clear_refs_pte_range()

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Pavel Emelyanov <xemul@xxxxxxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxx>
Signed-off-by: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
Cc: Cyrill Gorcunov <gorcunov@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
 fs/proc/task_mmu.c | 47 ++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 44 insertions(+), 3 deletions(-)

diff --git v3.16-rc1.orig/fs/proc/task_mmu.c v3.16-rc1/fs/proc/task_mmu.c
index ac50a829320a..2acbe144152c 100644
--- v3.16-rc1.orig/fs/proc/task_mmu.c
+++ v3.16-rc1/fs/proc/task_mmu.c
@@ -719,10 +719,10 @@ struct clear_refs_private {
 	enum clear_refs_types type;
 };
 
+#ifdef CONFIG_MEM_SOFT_DIRTY
 static inline void clear_soft_dirty(struct vm_area_struct *vma,
 		unsigned long addr, pte_t *pte)
 {
-#ifdef CONFIG_MEM_SOFT_DIRTY
 	/*
 	 * The soft-dirty tracker uses #PF-s to catch writes
 	 * to pages, so write-protect the pte as well. See the
@@ -741,9 +741,35 @@ static inline void clear_soft_dirty(struct vm_area_struct *vma,
 	}
 
 	set_pte_at(vma->vm_mm, addr, pte, ptent);
-#endif
 }
 
+static inline void clear_soft_dirty_pmd(struct vm_area_struct *vma,
+		unsigned long addr, pmd_t *pmdp)
+{
+	pmd_t pmd = *pmdp;
+
+	pmd = pmd_wrprotect(pmd);
+	pmd = pmd_clear_flags(pmd, _PAGE_SOFT_DIRTY);
+
+	if (vma->vm_flags & VM_SOFTDIRTY)
+		vma->vm_flags &= ~VM_SOFTDIRTY;
+
+	set_pmd_at(vma->vm_mm, addr, pmdp, pmd);
+}
+
+#else
+
+static inline void clear_soft_dirty(struct vm_area_struct *vma,
+		unsigned long addr, pte_t *pte)
+{
+}
+
+static inline void clear_soft_dirty_pmd(struct vm_area_struct *vma,
+		unsigned long addr, pmd_t *pmdp)
+{
+}
+#endif
+
 static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr,
 				unsigned long end, struct mm_walk *walk)
 {
@@ -753,7 +779,22 @@ static int clear_refs_pte_range(pmd_t *pmd, unsigned long addr,
 	spinlock_t *ptl;
 	struct page *page;
 
-	split_huge_page_pmd(vma, addr, pmd);
+	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
+		if (cp->type == CLEAR_REFS_SOFT_DIRTY) {
+			clear_soft_dirty_pmd(vma, addr, pmd);
+			goto out;
+		}
+
+		page = pmd_page(*pmd);
+
+		/* Clear accessed and referenced bits. */
+		pmdp_test_and_clear_young(vma, addr, pmd);
+		ClearPageReferenced(page);
+out:
+		spin_unlock(ptl);
+		return 0;
+	}
+
 	if (pmd_trans_unstable(pmd))
 		return 0;
 
-- 
1.9.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/