[PATCH] page migration: Support a vma migration function

Hooks for calling vma specific migration functions With this patch a vma may define a vma->vm_ops->migrate function. That function may perform page migration on its own (some vmas may not contain page structs and therefore cannot be handled by regular page migration. Pages in a vma may require special preparatory treatment before migration is possible etc) . Only mmap_sem is held when the migration function is called. The migrate() function gets passed two sets of nodemasks describing the source and the target of the migration. The flags parameter either contains MPOL_MF_MOVE which means that only pages used exclusively by the specified mm should be moved or MPOL_MF_MOVE_ALL which means that pages shared with other processes should also be moved. The migration function returns 0 on success or an error condition. An error condition will prevent regular page migration from occurring. On its own this patch cannot be included since there are no users for this functionality. But it seems that the uncached allocator will need this functionality at some point. Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] page migration: Support a vma migration function
Hooks for calling vma specific migration functions With this patch a vma may define a vma->vm_ops->migrate function. That function may perform page migration on its own (some vmas may not contain page structs and therefore cannot be handled by regular page migration. Pages in a vma may require special preparatory treatment before migration is possible etc) . Only mmap_sem is held when the migration function is called. The migrate() function gets passed two sets of nodemasks describing the source and the target of the migration. The flags parameter either contains MPOL_MF_MOVE which means that only pages used exclusively by the specified mm should be moved or MPOL_MF_MOVE_ALL which means that pages shared with other processes should also be moved. The migration function returns 0 on success or an error condition. An error condition will prevent regular page migration from occurring. On its own this patch cannot be included since there are no users for this functionality. But it seems that the uncached allocator will need this functionality at some point. Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Hugh Dickins <hugh@veritas.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Christoph Lameter · Linus Torvalds
1 parent 68402ddc67
Showing 4 changed files with 37 additions and 2 deletions Side-by-side Diff
include/linux/migrate.h
include/linux/mm.h
mm/mempolicy.c
mm/migrate.c
@@ -16,7 +16,9 @@
 			struct page *, struct page *);
  
 extern int migrate_prep(void);
-
+extern int migrate_vmas(struct mm_struct *mm,
+		const nodemask_t *from, const nodemask_t *to,
+		unsigned long flags);
 #else
  
 static inline int isolate_lru_page(struct page *p, struct list_head *list)
@@ -29,6 +31,13 @@
 			struct vm_area_struct *vma, int dest) { return 0; }
  
 static inline int migrate_prep(void) { return -ENOSYS; }
+
+static inline int migrate_vmas(struct mm_struct *mm,
+		const nodemask_t *from, const nodemask_t *to,
+		unsigned long flags)
+{
+	return -ENOSYS;
+}
  
 /* Possible settings for the migrate_page() method in address_operations */
 #define migrate_page NULL
@@ -206,6 +206,8 @@
 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 					unsigned long addr);
+	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
+		const nodemask_t *to, unsigned long flags);
 #endif
 };
  
@@ -632,6 +632,10 @@
  
   	down_read(&mm->mmap_sem);
  
+	err = migrate_vmas(mm, from_nodes, to_nodes, flags);
+	if (err)
+		goto out;
+
 /*
  * Find a 'source' bit set in 'tmp' whose corresponding 'dest'
  * bit in 'to' is not also set in 'tmp'.  Clear the found 'source'
@@ -691,7 +695,7 @@
 		if (err < 0)
 			break;
 	}
-
+out:
 	up_read(&mm->mmap_sem);
 	if (err < 0)
 		return err;
@@ -975,4 +975,25 @@
 	return err;
 }
 #endif
+
+/*
+ * Call migration functions in the vma_ops that may prepare
+ * memory in a vm for migration. migration functions may perform
+ * the migration for vmas that do not have an underlying page struct.
+ */
+int migrate_vmas(struct mm_struct *mm, const nodemask_t *to,
+	const nodemask_t *from, unsigned long flags)
+{
+ 	struct vm_area_struct *vma;
+ 	int err = 0;
+
+ 	for(vma = mm->mmap; vma->vm_next && !err; vma = vma->vm_next) {
+ 		if (vma->vm_ops && vma->vm_ops->migrate) {
+ 			err = vma->vm_ops->migrate(vma, to, from, flags);
+ 			if (err)
+ 				break;
+ 		}
+ 	}
+ 	return err;
+}
...	...	@@ -16,7 +16,9 @@
16	16	struct page , struct page );
17	17
18	18	extern int migrate_prep(void);
19		-
	19	+extern int migrate_vmas(struct mm_struct *mm,
	20	+ const nodemask_t from, const nodemask_t to,
	21	+ unsigned long flags);
20	22	#else
21	23
22	24	static inline int isolate_lru_page(struct page p, struct list_head list)
...	...	@@ -29,6 +31,13 @@
29	31	struct vm_area_struct *vma, int dest) { return 0; }
30	32
31	33	static inline int migrate_prep(void) { return -ENOSYS; }
	34	+
	35	+static inline int migrate_vmas(struct mm_struct *mm,
	36	+ const nodemask_t from, const nodemask_t to,
	37	+ unsigned long flags)
	38	+{
	39	+ return -ENOSYS;
	40	+}
32	41
33	42	/* Possible settings for the migrate_page() method in address_operations */
34	43	#define migrate_page NULL
...	...	@@ -206,6 +206,8 @@
206	206	int (set_policy)(struct vm_area_struct vma, struct mempolicy *new);
207	207	struct mempolicy (get_policy)(struct vm_area_struct *vma,
208	208	unsigned long addr);
	209	+ int (migrate)(struct vm_area_struct vma, const nodemask_t *from,
	210	+ const nodemask_t *to, unsigned long flags);
209	211	#endif
210	212	};
211	213
...	...	@@ -632,6 +632,10 @@
632	632
633	633	down_read(&mm->mmap_sem);
634	634
	635	+ err = migrate_vmas(mm, from_nodes, to_nodes, flags);
	636	+ if (err)
	637	+ goto out;
	638	+
635	639	/*
636	640	* Find a 'source' bit set in 'tmp' whose corresponding 'dest'
637	641	* bit in 'to' is not also set in 'tmp'. Clear the found 'source'
...	...	@@ -691,7 +695,7 @@
691	695	if (err < 0)
692	696	break;
693	697	}
694		-
	698	+out:
695	699	up_read(&mm->mmap_sem);
696	700	if (err < 0)
697	701	return err;
...	...	@@ -975,4 +975,25 @@
975	975	return err;
976	976	}
977	977	#endif
	978	+
	979	+/*
	980	+ * Call migration functions in the vma_ops that may prepare
	981	+ * memory in a vm for migration. migration functions may perform
	982	+ * the migration for vmas that do not have an underlying page struct.
	983	+ */
	984	+int migrate_vmas(struct mm_struct mm, const nodemask_t to,
	985	+ const nodemask_t *from, unsigned long flags)
	986	+{
	987	+ struct vm_area_struct *vma;
	988	+ int err = 0;
	989	+
	990	+ for(vma = mm->mmap; vma->vm_next && !err; vma = vma->vm_next) {
	991	+ if (vma->vm_ops && vma->vm_ops->migrate) {
	992	+ err = vma->vm_ops->migrate(vma, to, from, flags);
	993	+ if (err)
	994	+ break;
	995	+ }
	996	+ }
	997	+ return err;
	998	+}