]> err.no Git - linux-2.6/commitdiff
[PATCH] page migration cleanup: pass "mapping" to migration functions
authorChristoph Lameter <clameter@sgi.com>
Fri, 23 Jun 2006 09:03:33 +0000 (02:03 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Fri, 23 Jun 2006 14:42:50 +0000 (07:42 -0700)
Change handling of address spaces.

Pass a pointer to the address space in which the page is migrated to all
migration function.  This avoids repeatedly having to retrieve the address
space pointer from the page and checking it for validity.  The old page
mapping will change once migration has gone to a certain step, so it is less
confusing to have the pointer always available.

Move the setting of the mapping and index for the new page into
migrate_pages().

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
include/linux/fs.h
include/linux/migrate.h
mm/migrate.c

index c823a3815e24f71ce76a490e08c31b18397db3ca..e917403f4d58fdaa64282db6204e49db0e2808db 100644 (file)
@@ -376,7 +376,8 @@ struct address_space_operations {
        struct page* (*get_xip_page)(struct address_space *, sector_t,
                        int);
        /* migrate the contents of a page to the specified target */
-       int (*migratepage) (struct page *, struct page *);
+       int (*migratepage) (struct address_space *,
+                       struct page *, struct page *);
 };
 
 struct backing_dev_info;
@@ -1772,7 +1773,8 @@ extern void simple_release_fs(struct vfsmount **mount, int *count);
 extern ssize_t simple_read_from_buffer(void __user *, size_t, loff_t *, const void *, size_t);
 
 #ifdef CONFIG_MIGRATION
-extern int buffer_migrate_page(struct page *, struct page *);
+extern int buffer_migrate_page(struct address_space *,
+                               struct page *, struct page *);
 #else
 #define buffer_migrate_page NULL
 #endif
index e8d3b08cc3542310df75cbe0d9bb7f45446d476c..287c47b5e5dfb64505debd78289f8f08be4466ad 100644 (file)
@@ -6,12 +6,14 @@
 #ifdef CONFIG_MIGRATION
 extern int isolate_lru_page(struct page *p, struct list_head *pagelist);
 extern int putback_lru_pages(struct list_head *l);
-extern int migrate_page(struct page *, struct page *);
+extern int migrate_page(struct address_space *,
+                       struct page *, struct page *);
 extern int migrate_pages(struct list_head *l, struct list_head *t,
                struct list_head *moved, struct list_head *failed);
 extern int migrate_pages_to(struct list_head *pagelist,
                        struct vm_area_struct *vma, int dest);
-extern int fail_migrate_page(struct page *, struct page *);
+extern int fail_migrate_page(struct address_space *,
+                       struct page *, struct page *);
 
 extern int migrate_prep(void);
 
index 8095c607a494d11582500fed427bea090710443b..f65e69d9452711a12e00897dd4e00733fbb863d0 100644 (file)
@@ -173,15 +173,11 @@ retry:
  * 2 for pages with a mapping
  * 3 for pages with a mapping and PagePrivate set.
  */
-static int migrate_page_move_mapping(struct page *newpage,
-                               struct page *page)
+static int migrate_page_move_mapping(struct address_space *mapping,
+               struct page *newpage, struct page *page)
 {
-       struct address_space *mapping = page_mapping(page);
        struct page **radix_pointer;
 
-       if (!mapping)
-               return -EAGAIN;
-
        write_lock_irq(&mapping->tree_lock);
 
        radix_pointer = (struct page **)radix_tree_lookup_slot(
@@ -197,15 +193,8 @@ static int migrate_page_move_mapping(struct page *newpage,
 
        /*
         * Now we know that no one else is looking at the page.
-        *
-        * Certain minimal information about a page must be available
-        * in order for other subsystems to properly handle the page if they
-        * find it through the radix tree update before we are finished
-        * copying the page.
         */
        get_page(newpage);
-       newpage->index = page->index;
-       newpage->mapping = page->mapping;
        if (PageSwapCache(page)) {
                SetPageSwapCache(newpage);
                set_page_private(newpage, page_private(page));
@@ -262,7 +251,8 @@ static void migrate_page_copy(struct page *newpage, struct page *page)
  ***********************************************************/
 
 /* Always fail migration. Used for mappings that are not movable */
-int fail_migrate_page(struct page *newpage, struct page *page)
+int fail_migrate_page(struct address_space *mapping,
+                       struct page *newpage, struct page *page)
 {
        return -EIO;
 }
@@ -274,13 +264,14 @@ EXPORT_SYMBOL(fail_migrate_page);
  *
  * Pages are locked upon entry and exit.
  */
-int migrate_page(struct page *newpage, struct page *page)
+int migrate_page(struct address_space *mapping,
+               struct page *newpage, struct page *page)
 {
        int rc;
 
        BUG_ON(PageWriteback(page));    /* Writeback must be complete */
 
-       rc = migrate_page_move_mapping(newpage, page);
+       rc = migrate_page_move_mapping(mapping, newpage, page);
 
        if (rc)
                return rc;
@@ -305,21 +296,18 @@ EXPORT_SYMBOL(migrate_page);
  * if the underlying filesystem guarantees that no other references to "page"
  * exist.
  */
-int buffer_migrate_page(struct page *newpage, struct page *page)
+int buffer_migrate_page(struct address_space *mapping,
+               struct page *newpage, struct page *page)
 {
-       struct address_space *mapping = page->mapping;
        struct buffer_head *bh, *head;
        int rc;
 
-       if (!mapping)
-               return -EAGAIN;
-
        if (!page_has_buffers(page))
-               return migrate_page(newpage, page);
+               return migrate_page(mapping, newpage, page);
 
        head = page_buffers(page);
 
-       rc = migrate_page_move_mapping(newpage, page);
+       rc = migrate_page_move_mapping(mapping, newpage, page);
 
        if (rc)
                return rc;
@@ -448,9 +436,6 @@ redo:
                        goto next;
                }
 
-               newpage = lru_to_page(to);
-               lock_page(newpage);
-
                /*
                 * Establish swap ptes for anonymous pages or destroy pte
                 * maps for files.
@@ -473,11 +458,18 @@ redo:
                rc = -EPERM;
                if (try_to_unmap(page, 1) == SWAP_FAIL)
                        /* A vma has VM_LOCKED set -> permanent failure */
-                       goto unlock_both;
+                       goto unlock_page;
 
                rc = -EAGAIN;
                if (page_mapped(page))
-                       goto unlock_both;
+                       goto unlock_page;
+
+               newpage = lru_to_page(to);
+               lock_page(newpage);
+               /* Prepare mapping for the new page.*/
+               newpage->index = page->index;
+               newpage->mapping = page->mapping;
+
                /*
                 * Pages are properly locked and writeback is complete.
                 * Try to migrate the page.
@@ -494,7 +486,8 @@ redo:
                         * own migration function. This is the most common
                         * path for page migration.
                         */
-                       rc = mapping->a_ops->migratepage(newpage, page);
+                       rc = mapping->a_ops->migratepage(mapping,
+                                                       newpage, page);
                        goto unlock_both;
                 }
 
@@ -524,7 +517,7 @@ redo:
                 */
                if (!page_has_buffers(page) ||
                    try_to_release_page(page, GFP_KERNEL)) {
-                       rc = migrate_page(newpage, page);
+                       rc = migrate_page(mapping, newpage, page);
                        goto unlock_both;
                }
 
@@ -553,12 +546,17 @@ unlock_page:
                unlock_page(page);
 
 next:
-               if (rc == -EAGAIN) {
-                       retry++;
-               } else if (rc) {
-                       /* Permanent failure */
-                       list_move(&page->lru, failed);
-                       nr_failed++;
+               if (rc) {
+                       if (newpage)
+                               newpage->mapping = NULL;
+
+                       if (rc == -EAGAIN)
+                               retry++;
+                       else {
+                               /* Permanent failure */
+                               list_move(&page->lru, failed);
+                               nr_failed++;
+                       }
                } else {
                        if (newpage) {
                                /* Successful migration. Return page to LRU */