Skip to contentSkip to conversation

3/4 Migration Cache - move migration cache page to swap cache

RFC New Lee Schermerhorn 1 participant 1 message Feb 17, 2006

Subsystems (5)
Memory management MGLRU MM reclaim MM swap +1 more subsystems
lore
Share & tools
lore

Lee Schermerhorn

author
#0 lore raw

RFC 1/1 3/4 Migration Cache - move migration cache page to swap cache

Migration Cache "V8" 3/4

This patch modifies the swapfile.c "unuse_*" stack to support moving pages from migration cache to swap cache in case we have to "fall back to swap". This also allows vmscan.c:shrink_list() to move migration cache pages to swap cache when/if it wants to swap them out. shrink_list() should only find anon pages in the migration cache when/if we implement lazy page migration.

Because of the new usage, the patch renames the static "unuse_*" functions in swapfile.c to "update_*". In "update_pte_range", if the entry arg matches the page's private data, we perform the usual "unuse_pte()"; otherwise, this is an "update/move" operation and we "update_pte()".

Then, this patch implements the __migration_move_to_swap() function on top of the modified "update_*" stack.

Assumption: because this facility is used only for removing swap devices [sys_swapoff()] and direct page migration, it is not in a critical/fast path. Even so, this patch doesn't add significant overhead to the stack [he says, glibly].

Signed-off-by Lee Schermerhorn <[email protected]>
include/linux/swap.h +3 -1
Index: linux-2.6.16-rc3-mm1/include/linux/swap.h
===================================================================
@@ -197,6 +197,7 @@ extern int migrate_page_remove_reference
 extern unsigned long migrate_pages(struct list_head *l, struct list_head *t,
 		struct list_head *moved, struct list_head *failed);
 extern int fail_migrate_page(struct page *, struct page *);
+extern int remove_vma_swap(struct vm_area_struct *, struct page *);
 
 extern int add_to_migration_cache(struct page *, int);
 extern void migration_remove_entry(swp_entry_t, int);

@@ -206,6 +207,8 @@ extern void migration_remove_reference(s
 extern void __migration_remove_reference(struct page *, swp_entry_t,
 		 int);
 extern struct page *lookup_migration_cache(swp_entry_t);
+extern int __migration_move_to_swap(struct vm_area_struct *, struct page *,
+		swp_entry_t);
 #else
 static inline int isolate_lru_page(struct page *p) { return -ENOSYS; }
 static inline int putback_lru_pages(struct list_head *l) { return 0; }

@@ -279,7 +282,6 @@ extern int remove_exclusive_swap_page(st
 struct backing_dev_info;
 
 extern spinlock_t swap_lock;
-extern int remove_vma_swap(struct vm_area_struct *vma, struct page *page);
 
 /* linux/mm/thrash.c */
 extern struct mm_struct * swap_token_mm;
mm/swapfile.c +59 -10

137 lines not on this page · raw message

include/linux/rmap.h +1 -0
Index: linux-2.6.16-rc3-mm1/include/linux/rmap.h
===================================================================
@@ -93,6 +93,7 @@ static inline void page_dup_rmap(struct 
 int page_referenced(struct page *, int is_locked);
 int try_to_unmap(struct page *, int ignore_refs);
 void remove_from_swap(struct page *page);
+int migration_move_to_swap(struct page *);
 
 /*
  * Called from mm/filemap_xip.c to unmap empty zero page
mm/rmap.c +66 -0

72 lines not on this page · raw message

mm/vmscan.c +2 -5
Index: linux-2.6.16-rc3-mm1/mm/vmscan.c
===================================================================
@@ -464,11 +464,8 @@ static unsigned long shrink_page_list(st
 				if (!add_to_swap(page, GFP_ATOMIC))
 					goto activate_locked;
 			} else if (page_is_migration(page)) {
-				/*
-				 * For now, skip migration cache pages.
-				 * TODO:  move to swap cache [difficult?]
-				 */
-				goto keep_locked;
+				if (!migration_move_to_swap(page))
+					goto keep_locked;
 			}
 		}
 #endif /* CONFIG_SWAP */

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>