ksm: rmap_walk to remove_migation_ptes

A side-effect of making ksm pages swappable is that they have to be placed
on the LRUs: which then exposes them to isolate_lru_page() and hence to
page migration.

Add rmap_walk() for remove_migration_ptes() to use: rmap_walk_anon() and
rmap_walk_file() in rmap.c, but rmap_walk_ksm() in ksm.c.  Perhaps some
consolidation with existing code is possible, but don't attempt that yet
(try_to_unmap needs to handle nonlinears, but migration pte removal does
not).

rmap_walk() is sadly less general than it appears: rmap_walk_anon(), like
remove_anon_migration_ptes() which it replaces, avoids calling
page_lock_anon_vma(), because that includes a page_mapped() test which
fails when all migration ptes are in place.  That was valid when NUMA page
migration was introduced (holding mmap_sem provided the missing guarantee
that anon_vma's slab had not already been destroyed), but I believe not
valid in the memory hotremove case added since.

For now do the same as before, and consider the best way to fix that
unlikely race later on.  When fixed, we can probably use rmap_walk() on
hwpoisoned ksm pages too: for now, they remain among hwpoison's various
exceptions (its PageKsm test comes before the page is locked, but its
page_lock_anon_vma fails safely if an anon gets upgraded).

Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: Izik Eidus <ieidus@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Chris Wright <chrisw@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

5 files changed, 181 insertions, 67 deletions

diff --git a/include/linux/ksm.h b/include/linux/ksm.h
index 157d83d..bed5f16 100644
--- a/include/linux/ksm.h
+++ b/include/linux/ksm.h
@@ -88,6 +88,9 @@ static inline struct page *ksm_might_need_to_copy(struct page *page,
 int page_referenced_ksm(struct page *page,
 			struct mem_cgroup *memcg, unsigned long *vm_flags);
 int try_to_unmap_ksm(struct page *page, enum ttu_flags flags);
+int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page *,
+		  struct vm_area_struct *, unsigned long, void *), void *arg);
+void ksm_migrate_page(struct page *newpage, struct page *oldpage);
 
 #else  /* !CONFIG_KSM */
 
@@ -127,6 +130,16 @@ static inline int try_to_unmap_ksm(struct page *page, enum ttu_flags flags)
 {
 	return 0;
 }
+
+static inline int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page*,
+		struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+	return 0;
+}
+
+static inline void ksm_migrate_page(struct page *newpage, struct page *oldpage)
+{
+}
 #endif /* !CONFIG_KSM */
 
 #endif /* __LINUX_KSM_H */
diff --git a/include/linux/rmap.h b/include/linux/rmap.h
index 980094a..b019ae6 100644
--- a/include/linux/rmap.h
+++ b/include/linux/rmap.h
@@ -164,6 +164,12 @@ struct anon_vma *page_lock_anon_vma(struct page *page);
 void page_unlock_anon_vma(struct anon_vma *anon_vma);
 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 
+/*
+ * Called by migrate.c to remove migration ptes, but might be used more later.
+ */
+int rmap_walk(struct page *page, int (*rmap_one)(struct page *,
+		struct vm_area_struct *, unsigned long, void *), void *arg);
+
 #else	/* !CONFIG_MMU */
 
 #define anon_vma_init()		do {} while (0)
diff --git a/mm/ksm.c b/mm/ksm.c
index 20f46a7..dfdc292 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -1656,6 +1656,71 @@ out:
 	return ret;
 }
 
+#ifdef CONFIG_MIGRATION
+int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page *,
+		  struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+	struct stable_node *stable_node;
+	struct hlist_node *hlist;
+	struct rmap_item *rmap_item;
+	int ret = SWAP_AGAIN;
+	int search_new_forks = 0;
+
+	VM_BUG_ON(!PageKsm(page));
+	VM_BUG_ON(!PageLocked(page));
+
+	stable_node = page_stable_node(page);
+	if (!stable_node)
+		return ret;
+again:
+	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
+		struct anon_vma *anon_vma = rmap_item->anon_vma;
+		struct vm_area_struct *vma;
+
+		spin_lock(&anon_vma->lock);
+		list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
+			if (rmap_item->address < vma->vm_start ||
+			    rmap_item->address >= vma->vm_end)
+				continue;
+			/*
+			 * Initially we examine only the vma which covers this
+			 * rmap_item; but later, if there is still work to do,
+			 * we examine covering vmas in other mms: in case they
+			 * were forked from the original since ksmd passed.
+			 */
+			if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
+				continue;
+
+			ret = rmap_one(page, vma, rmap_item->address, arg);
+			if (ret != SWAP_AGAIN) {
+				spin_unlock(&anon_vma->lock);
+				goto out;
+			}
+		}
+		spin_unlock(&anon_vma->lock);
+	}
+	if (!search_new_forks++)
+		goto again;
+out:
+	return ret;
+}
+
+void ksm_migrate_page(struct page *newpage, struct page *oldpage)
+{
+	struct stable_node *stable_node;
+
+	VM_BUG_ON(!PageLocked(oldpage));
+	VM_BUG_ON(!PageLocked(newpage));
+	VM_BUG_ON(newpage->mapping != oldpage->mapping);
+
+	stable_node = page_stable_node(newpage);
+	if (stable_node) {
+		VM_BUG_ON(stable_node->page != oldpage);
+		stable_node->page = newpage;
+	}
+}
+#endif /* CONFIG_MIGRATION */
+
 #ifdef CONFIG_SYSFS
 /*
  * This all compiles without CONFIG_SYSFS, but is a waste of space.
diff --git a/mm/migrate.c b/mm/migrate.c
index 367272d..0b71474 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -21,6 +21,7 @@
 #include <linux/mm_inline.h>
 #include <linux/nsproxy.h>
 #include <linux/pagevec.h>
+#include <linux/ksm.h>
 #include <linux/rmap.h>
 #include <linux/topology.h>
 #include <linux/cpu.h>
@@ -78,8 +79,8 @@ int putback_lru_pages(struct list_head *l)
 /*
  * Restore a potential migration pte to a working pte entry
  */
-static void remove_migration_pte(struct vm_area_struct *vma,
-		struct page *old, struct page *new)
+static int remove_migration_pte(struct page *new, struct vm_area_struct *vma,
+				 unsigned long addr, void *old)
 {
 	struct mm_struct *mm = vma->vm_mm;
 	swp_entry_t entry;
@@ -88,40 +89,37 @@ static void remove_migration_pte(struct vm_area_struct *vma,
  	pmd_t *pmd;
 	pte_t *ptep, pte;
  	spinlock_t *ptl;
-	unsigned long addr = page_address_in_vma(new, vma);
-
-	if (addr == -EFAULT)
-		return;
 
  	pgd = pgd_offset(mm, addr);
 	if (!pgd_present(*pgd))
-                return;
+		goto out;
 
 	pud = pud_offset(pgd, addr);
 	if (!pud_present(*pud))
-                return;
+		goto out;
 
 	pmd = pmd_offset(pud, addr);
 	if (!pmd_present(*pmd))
-		return;
+		goto out;
 
 	ptep = pte_offset_map(pmd, addr);
 
 	if (!is_swap_pte(*ptep)) {
 		pte_unmap(ptep);
- 		return;
+		goto out;
  	}
 
  	ptl = pte_lockptr(mm, pmd);
  	spin_lock(ptl);
 	pte = *ptep;
 	if (!is_swap_pte(pte))
-		goto out;
+		goto unlock;
 
 	entry = pte_to_swp_entry(pte);
 
-	if (!is_migration_entry(entry) || migration_entry_to_page(entry) != old)
-		goto out;
+	if (!is_migration_entry(entry) ||
+	    migration_entry_to_page(entry) != old)
+		goto unlock;
 
 	get_page(new);
 	pte = pte_mkold(mk_pte(new, vma->vm_page_prot));
@@ -137,55 +135,10 @@ static void remove_migration_pte(struct vm_area_struct *vma,
 
 	/* No need to invalidate - it was non-present before */
 	update_mmu_cache(vma, addr, pte);
-
-out:
+unlock:
 	pte_unmap_unlock(ptep, ptl);
-}
-
-/*
- * Note that remove_file_migration_ptes will only work on regular mappings,
- * Nonlinear mappings do not use migration entries.
- */
-static void remove_file_migration_ptes(struct page *old, struct page *new)
-{
-	struct vm_area_struct *vma;
-	struct address_space *mapping = new->mapping;
-	struct prio_tree_iter iter;
-	pgoff_t pgoff = new->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
-
-	if (!mapping)
-		return;
-
-	spin_lock(&mapping->i_mmap_lock);
-
-	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff)
-		remove_migration_pte(vma, old, new);
-
-	spin_unlock(&mapping->i_mmap_lock);
-}
-
-/*
- * Must hold mmap_sem lock on at least one of the vmas containing
- * the page so that the anon_vma cannot vanish.
- */
-static void remove_anon_migration_ptes(struct page *old, struct page *new)
-{
-	struct anon_vma *anon_vma;
-	struct vm_area_struct *vma;
-
-	/*
-	 * We hold the mmap_sem lock. So no need to call page_lock_anon_vma.
-	 */
-	anon_vma = page_anon_vma(new);
-	if (!anon_vma)
-		return;
-
-	spin_lock(&anon_vma->lock);
-
-	list_for_each_entry(vma, &anon_vma->head, anon_vma_node)
-		remove_migration_pte(vma, old, new);
-
-	spin_unlock(&anon_vma->lock);
+out:
+	return SWAP_AGAIN;
 }
 
 /*
@@ -194,10 +147,7 @@ static void remove_anon_migration_ptes(struct page *old, struct page *new)
  */
 static void remove_migration_ptes(struct page *old, struct page *new)
 {
-	if (PageAnon(new))
-		remove_anon_migration_ptes(old, new);
-	else
-		remove_file_migration_ptes(old, new);
+	rmap_walk(new, remove_migration_pte, old);
 }
 
 /*
@@ -358,6 +308,7 @@ static void migrate_page_copy(struct page *newpage, struct page *page)
  	}
 
 	mlock_migrate_page(newpage, page);
+	ksm_migrate_page(newpage, page);
 
 	ClearPageSwapCache(page);
 	ClearPagePrivate(page);
@@ -577,9 +528,9 @@ static int move_to_new_page(struct page *newpage, struct page *page)
 	else
 		rc = fallback_migrate_page(mapping, newpage, page);
 
-	if (!rc) {
+	if (!rc)
 		remove_migration_ptes(page, newpage);
-	} else
+	else
 		newpage->mapping = NULL;
 
 	unlock_page(newpage);
diff --git a/mm/rmap.c b/mm/rmap.c
index 2e38e90..c81bedd 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1203,3 +1203,82 @@ int try_to_munlock(struct page *page)
 	else
 		return try_to_unmap_file(page, TTU_MUNLOCK);
 }
+
+#ifdef CONFIG_MIGRATION
+/*
+ * rmap_walk() and its helpers rmap_walk_anon() and rmap_walk_file():
+ * Called by migrate.c to remove migration ptes, but might be used more later.
+ */
+static int rmap_walk_anon(struct page *page, int (*rmap_one)(struct page *,
+		struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+	struct anon_vma *anon_vma;
+	struct vm_area_struct *vma;
+	int ret = SWAP_AGAIN;
+
+	/*
+	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma()
+	 * because that depends on page_mapped(); but not all its usages
+	 * are holding mmap_sem, which also gave the necessary guarantee
+	 * (that this anon_vma's slab has not already been destroyed).
+	 * This needs to be reviewed later: avoiding page_lock_anon_vma()
+	 * is risky, and currently limits the usefulness of rmap_walk().
+	 */
+	anon_vma = page_anon_vma(page);
+	if (!anon_vma)
+		return ret;
+	spin_lock(&anon_vma->lock);
+	list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
+		unsigned long address = vma_address(page, vma);
+		if (address == -EFAULT)
+			continue;
+		ret = rmap_one(page, vma, address, arg);
+		if (ret != SWAP_AGAIN)
+			break;
+	}
+	spin_unlock(&anon_vma->lock);
+	return ret;
+}
+
+static int rmap_walk_file(struct page *page, int (*rmap_one)(struct page *,
+		struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+	struct address_space *mapping = page->mapping;
+	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
+	struct vm_area_struct *vma;
+	struct prio_tree_iter iter;
+	int ret = SWAP_AGAIN;
+
+	if (!mapping)
+		return ret;
+	spin_lock(&mapping->i_mmap_lock);
+	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
+		unsigned long address = vma_address(page, vma);
+		if (address == -EFAULT)
+			continue;
+		ret = rmap_one(page, vma, address, arg);
+		if (ret != SWAP_AGAIN)
+			break;
+	}
+	/*
+	 * No nonlinear handling: being always shared, nonlinear vmas
+	 * never contain migration ptes.  Decide what to do about this
+	 * limitation to linear when we need rmap_walk() on nonlinear.
+	 */
+	spin_unlock(&mapping->i_mmap_lock);
+	return ret;
+}
+
+int rmap_walk(struct page *page, int (*rmap_one)(struct page *,
+		struct vm_area_struct *, unsigned long, void *), void *arg)
+{
+	VM_BUG_ON(!PageLocked(page));
+
+	if (unlikely(PageKsm(page)))
+		return rmap_walk_ksm(page, rmap_one, arg);
+	else if (PageAnon(page))
+		return rmap_walk_anon(page, rmap_one, arg);
+	else
+		return rmap_walk_file(page, rmap_one, arg);
+}
+#endif /* CONFIG_MIGRATION */