mm: /proc/pid/clear_refs: avoid split_huge_page()

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.

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index f5ca96524f5f9f48c577cd05c15ea2fac37d4304..0e36c1e49fe30be1a7932bb2fb4d00e9ec9c84a3 100644 (file)
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -739,10 +739,10 @@ 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
@@ -759,9 +759,35 @@ 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)
 {
@@ -771,7 +797,22 @@ 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;