[PATCH -mm v6 12/13] mm: /proc/pid/clear_refs: avoid split_huge_page()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



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>
Reviewed-by: Cyrill Gorcunov <gorcunov@xxxxxxxxxx>
---
 fs/proc/task_mmu.c | 47 ++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 44 insertions(+), 3 deletions(-)

diff --git mmotm-2014-07-30-15-57.orig/fs/proc/task_mmu.c mmotm-2014-07-30-15-57/fs/proc/task_mmu.c
index 084d750f6177..5b71471b9647 100644
--- mmotm-2014-07-30-15-57.orig/fs/proc/task_mmu.c
+++ mmotm-2014-07-30-15-57/fs/proc/task_mmu.c
@@ -712,10 +712,10 @@ 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
@@ -734,9 +734,35 @@ 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)
 {
@@ -746,7 +772,22 @@ 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, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]