migrate.c 31.6 KB
Newer Older
C
Christoph Lameter 已提交
1 2 3 4 5 6 7 8 9 10 11
/*
 * Memory Migration functionality - linux/mm/migration.c
 *
 * Copyright (C) 2006 Silicon Graphics, Inc., Christoph Lameter
 *
 * Page migration was first developed in the context of the memory hotplug
 * project. The main authors of the migration code are:
 *
 * IWAMOTO Toshihiro <iwamoto@valinux.co.jp>
 * Hirokazu Takahashi <taka@valinux.co.jp>
 * Dave Hansen <haveblue@us.ibm.com>
C
Christoph Lameter 已提交
12
 * Christoph Lameter
C
Christoph Lameter 已提交
13 14 15 16 17
 */

#include <linux/migrate.h>
#include <linux/module.h>
#include <linux/swap.h>
18
#include <linux/swapops.h>
C
Christoph Lameter 已提交
19
#include <linux/pagemap.h>
20
#include <linux/buffer_head.h>
C
Christoph Lameter 已提交
21
#include <linux/mm_inline.h>
22
#include <linux/nsproxy.h>
C
Christoph Lameter 已提交
23
#include <linux/pagevec.h>
24
#include <linux/ksm.h>
C
Christoph Lameter 已提交
25 26 27 28
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
29
#include <linux/writeback.h>
30 31
#include <linux/mempolicy.h>
#include <linux/vmalloc.h>
32
#include <linux/security.h>
33
#include <linux/memcontrol.h>
34
#include <linux/syscalls.h>
N
Naoya Horiguchi 已提交
35
#include <linux/hugetlb.h>
36
#include <linux/gfp.h>
C
Christoph Lameter 已提交
37

38 39
#include <asm/tlbflush.h>

C
Christoph Lameter 已提交
40 41 42 43 44
#include "internal.h"

#define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))

/*
45
 * migrate_prep() needs to be called before we start compiling a list of pages
46 47
 * to be migrated using isolate_lru_page(). If scheduling work on other CPUs is
 * undesirable, use migrate_prep_local()
C
Christoph Lameter 已提交
48 49 50 51 52 53 54 55 56 57 58 59 60 61
 */
int migrate_prep(void)
{
	/*
	 * Clear the LRU lists so pages can be isolated.
	 * Note that pages may be moved off the LRU after we have
	 * drained them. Those pages will fail to migrate like other
	 * pages that may be busy.
	 */
	lru_add_drain_all();

	return 0;
}

62 63 64 65 66 67 68 69
/* Do the necessary work of migrate_prep but not if it involves other CPUs */
int migrate_prep_local(void)
{
	lru_add_drain();

	return 0;
}

C
Christoph Lameter 已提交
70
/*
L
Lee Schermerhorn 已提交
71 72
 * Add isolated pages on the list back to the LRU under page lock
 * to avoid leaking evictable pages back onto unevictable list.
C
Christoph Lameter 已提交
73
 */
74
void putback_lru_pages(struct list_head *l)
C
Christoph Lameter 已提交
75 76 77 78 79
{
	struct page *page;
	struct page *page2;

	list_for_each_entry_safe(page, page2, l, lru) {
80
		list_del(&page->lru);
K
KOSAKI Motohiro 已提交
81
		dec_zone_page_state(page, NR_ISOLATED_ANON +
82
				page_is_file_cache(page));
L
Lee Schermerhorn 已提交
83
		putback_lru_page(page);
C
Christoph Lameter 已提交
84 85 86
	}
}

87 88 89
/*
 * Restore a potential migration pte to a working pte entry
 */
90 91
static int remove_migration_pte(struct page *new, struct vm_area_struct *vma,
				 unsigned long addr, void *old)
92 93 94 95 96 97 98 99 100
{
	struct mm_struct *mm = vma->vm_mm;
	swp_entry_t entry;
 	pgd_t *pgd;
 	pud_t *pud;
 	pmd_t *pmd;
	pte_t *ptep, pte;
 	spinlock_t *ptl;

N
Naoya Horiguchi 已提交
101 102 103 104 105 106 107 108 109
	if (unlikely(PageHuge(new))) {
		ptep = huge_pte_offset(mm, addr);
		if (!ptep)
			goto out;
		ptl = &mm->page_table_lock;
	} else {
		pgd = pgd_offset(mm, addr);
		if (!pgd_present(*pgd))
			goto out;
110

N
Naoya Horiguchi 已提交
111 112 113
		pud = pud_offset(pgd, addr);
		if (!pud_present(*pud))
			goto out;
114

N
Naoya Horiguchi 已提交
115
		pmd = pmd_offset(pud, addr);
116 117
		if (pmd_trans_huge(*pmd))
			goto out;
N
Naoya Horiguchi 已提交
118 119
		if (!pmd_present(*pmd))
			goto out;
120

N
Naoya Horiguchi 已提交
121
		ptep = pte_offset_map(pmd, addr);
122

N
Naoya Horiguchi 已提交
123 124 125 126 127 128 129
		if (!is_swap_pte(*ptep)) {
			pte_unmap(ptep);
			goto out;
		}

		ptl = pte_lockptr(mm, pmd);
	}
130 131 132 133

 	spin_lock(ptl);
	pte = *ptep;
	if (!is_swap_pte(pte))
134
		goto unlock;
135 136 137

	entry = pte_to_swp_entry(pte);

138 139 140
	if (!is_migration_entry(entry) ||
	    migration_entry_to_page(entry) != old)
		goto unlock;
141 142 143 144 145

	get_page(new);
	pte = pte_mkold(mk_pte(new, vma->vm_page_prot));
	if (is_write_migration_entry(entry))
		pte = pte_mkwrite(pte);
A
Andi Kleen 已提交
146
#ifdef CONFIG_HUGETLB_PAGE
N
Naoya Horiguchi 已提交
147 148
	if (PageHuge(new))
		pte = pte_mkhuge(pte);
A
Andi Kleen 已提交
149
#endif
150
	flush_cache_page(vma, addr, pte_pfn(pte));
151
	set_pte_at(mm, addr, ptep, pte);
152

N
Naoya Horiguchi 已提交
153 154 155 156 157 158
	if (PageHuge(new)) {
		if (PageAnon(new))
			hugepage_add_anon_rmap(new, vma, addr);
		else
			page_dup_rmap(new);
	} else if (PageAnon(new))
159 160 161 162 163
		page_add_anon_rmap(new, vma, addr);
	else
		page_add_file_rmap(new);

	/* No need to invalidate - it was non-present before */
164
	update_mmu_cache(vma, addr, ptep);
165
unlock:
166
	pte_unmap_unlock(ptep, ptl);
167 168
out:
	return SWAP_AGAIN;
169 170
}

171 172 173 174 175 176
/*
 * Get rid of all migration entries and replace them by
 * references to the indicated page.
 */
static void remove_migration_ptes(struct page *old, struct page *new)
{
177
	rmap_walk(new, remove_migration_pte, old);
178 179
}

180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
/*
 * Something used the pte of a page under migration. We need to
 * get to the page and wait until migration is finished.
 * When we return from this function the fault will be retried.
 *
 * This function is called from do_swap_page().
 */
void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
				unsigned long address)
{
	pte_t *ptep, pte;
	spinlock_t *ptl;
	swp_entry_t entry;
	struct page *page;

	ptep = pte_offset_map_lock(mm, pmd, address, &ptl);
	pte = *ptep;
	if (!is_swap_pte(pte))
		goto out;

	entry = pte_to_swp_entry(pte);
	if (!is_migration_entry(entry))
		goto out;

	page = migration_entry_to_page(entry);

N
Nick Piggin 已提交
206 207 208 209 210 211 212 213 214
	/*
	 * Once radix-tree replacement of page migration started, page_count
	 * *must* be zero. And, we don't want to call wait_on_page_locked()
	 * against a page without get_page().
	 * So, we use get_page_unless_zero(), here. Even failed, page fault
	 * will occur again.
	 */
	if (!get_page_unless_zero(page))
		goto out;
215 216 217 218 219 220 221 222
	pte_unmap_unlock(ptep, ptl);
	wait_on_page_locked(page);
	put_page(page);
	return;
out:
	pte_unmap_unlock(ptep, ptl);
}

C
Christoph Lameter 已提交
223
/*
224
 * Replace the page in the mapping.
225 226 227 228
 *
 * The number of remaining references must be:
 * 1 for anonymous pages without a mapping
 * 2 for pages with a mapping
229
 * 3 for pages with a mapping and PagePrivate/PagePrivate2 set.
C
Christoph Lameter 已提交
230
 */
231 232
static int migrate_page_move_mapping(struct address_space *mapping,
		struct page *newpage, struct page *page)
C
Christoph Lameter 已提交
233
{
N
Nick Piggin 已提交
234
	int expected_count;
235
	void **pslot;
C
Christoph Lameter 已提交
236

237
	if (!mapping) {
238
		/* Anonymous page without mapping */
239 240 241 242 243
		if (page_count(page) != 1)
			return -EAGAIN;
		return 0;
	}

N
Nick Piggin 已提交
244
	spin_lock_irq(&mapping->tree_lock);
C
Christoph Lameter 已提交
245

246 247
	pslot = radix_tree_lookup_slot(&mapping->page_tree,
 					page_index(page));
C
Christoph Lameter 已提交
248

249
	expected_count = 2 + page_has_private(page);
N
Nick Piggin 已提交
250
	if (page_count(page) != expected_count ||
251
		radix_tree_deref_slot_protected(pslot, &mapping->tree_lock) != page) {
N
Nick Piggin 已提交
252
		spin_unlock_irq(&mapping->tree_lock);
253
		return -EAGAIN;
C
Christoph Lameter 已提交
254 255
	}

N
Nick Piggin 已提交
256
	if (!page_freeze_refs(page, expected_count)) {
N
Nick Piggin 已提交
257
		spin_unlock_irq(&mapping->tree_lock);
N
Nick Piggin 已提交
258 259 260
		return -EAGAIN;
	}

C
Christoph Lameter 已提交
261 262 263
	/*
	 * Now we know that no one else is looking at the page.
	 */
264
	get_page(newpage);	/* add cache reference */
C
Christoph Lameter 已提交
265 266 267 268 269
	if (PageSwapCache(page)) {
		SetPageSwapCache(newpage);
		set_page_private(newpage, page_private(page));
	}

270 271
	radix_tree_replace_slot(pslot, newpage);

N
Nick Piggin 已提交
272
	page_unfreeze_refs(page, expected_count);
273 274 275 276
	/*
	 * Drop cache reference from old page.
	 * We know this isn't the last reference.
	 */
C
Christoph Lameter 已提交
277
	__put_page(page);
278

279 280 281 282 283 284 285 286 287 288 289 290
	/*
	 * If moved to a different zone then also account
	 * the page for that zone. Other VM counters will be
	 * taken care of when we establish references to the
	 * new page and drop references to the old page.
	 *
	 * Note that anonymous pages are accounted for
	 * via NR_FILE_PAGES and NR_ANON_PAGES if they
	 * are mapped to swap space.
	 */
	__dec_zone_page_state(page, NR_FILE_PAGES);
	__inc_zone_page_state(newpage, NR_FILE_PAGES);
291 292 293 294
	if (PageSwapBacked(page)) {
		__dec_zone_page_state(page, NR_SHMEM);
		__inc_zone_page_state(newpage, NR_SHMEM);
	}
N
Nick Piggin 已提交
295
	spin_unlock_irq(&mapping->tree_lock);
C
Christoph Lameter 已提交
296 297 298 299

	return 0;
}

N
Naoya Horiguchi 已提交
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322
/*
 * The expected number of remaining references is the same as that
 * of migrate_page_move_mapping().
 */
int migrate_huge_page_move_mapping(struct address_space *mapping,
				   struct page *newpage, struct page *page)
{
	int expected_count;
	void **pslot;

	if (!mapping) {
		if (page_count(page) != 1)
			return -EAGAIN;
		return 0;
	}

	spin_lock_irq(&mapping->tree_lock);

	pslot = radix_tree_lookup_slot(&mapping->page_tree,
					page_index(page));

	expected_count = 2 + page_has_private(page);
	if (page_count(page) != expected_count ||
323
		radix_tree_deref_slot_protected(pslot, &mapping->tree_lock) != page) {
N
Naoya Horiguchi 已提交
324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
		spin_unlock_irq(&mapping->tree_lock);
		return -EAGAIN;
	}

	if (!page_freeze_refs(page, expected_count)) {
		spin_unlock_irq(&mapping->tree_lock);
		return -EAGAIN;
	}

	get_page(newpage);

	radix_tree_replace_slot(pslot, newpage);

	page_unfreeze_refs(page, expected_count);

	__put_page(page);

	spin_unlock_irq(&mapping->tree_lock);
	return 0;
}

C
Christoph Lameter 已提交
345 346 347
/*
 * Copy the page to its new location
 */
N
Naoya Horiguchi 已提交
348
void migrate_page_copy(struct page *newpage, struct page *page)
C
Christoph Lameter 已提交
349
{
N
Naoya Horiguchi 已提交
350 351 352 353
	if (PageHuge(page))
		copy_huge_page(newpage, page);
	else
		copy_highpage(newpage, page);
C
Christoph Lameter 已提交
354 355 356 357 358 359 360

	if (PageError(page))
		SetPageError(newpage);
	if (PageReferenced(page))
		SetPageReferenced(newpage);
	if (PageUptodate(page))
		SetPageUptodate(newpage);
L
Lee Schermerhorn 已提交
361 362
	if (TestClearPageActive(page)) {
		VM_BUG_ON(PageUnevictable(page));
C
Christoph Lameter 已提交
363
		SetPageActive(newpage);
364 365
	} else if (TestClearPageUnevictable(page))
		SetPageUnevictable(newpage);
C
Christoph Lameter 已提交
366 367 368 369 370 371 372
	if (PageChecked(page))
		SetPageChecked(newpage);
	if (PageMappedToDisk(page))
		SetPageMappedToDisk(newpage);

	if (PageDirty(page)) {
		clear_page_dirty_for_io(page);
N
Nick Piggin 已提交
373 374 375 376 377
		/*
		 * Want to mark the page and the radix tree as dirty, and
		 * redo the accounting that clear_page_dirty_for_io undid,
		 * but we can't use set_page_dirty because that function
		 * is actually a signal that all of the page has become dirty.
L
Lucas De Marchi 已提交
378
		 * Whereas only part of our page may be dirty.
N
Nick Piggin 已提交
379 380
		 */
		__set_page_dirty_nobuffers(newpage);
C
Christoph Lameter 已提交
381 382
 	}

N
Nick Piggin 已提交
383
	mlock_migrate_page(newpage, page);
384
	ksm_migrate_page(newpage, page);
N
Nick Piggin 已提交
385

C
Christoph Lameter 已提交
386 387 388 389 390 391 392 393 394 395 396 397 398
	ClearPageSwapCache(page);
	ClearPagePrivate(page);
	set_page_private(page, 0);
	page->mapping = NULL;

	/*
	 * If any waiters have accumulated on the new page then
	 * wake them up.
	 */
	if (PageWriteback(newpage))
		end_page_writeback(newpage);
}

399 400 401 402 403
/************************************************************
 *                    Migration functions
 ***********************************************************/

/* Always fail migration. Used for mappings that are not movable */
404 405
int fail_migrate_page(struct address_space *mapping,
			struct page *newpage, struct page *page)
406 407 408 409 410
{
	return -EIO;
}
EXPORT_SYMBOL(fail_migrate_page);

C
Christoph Lameter 已提交
411 412
/*
 * Common logic to directly migrate a single page suitable for
413
 * pages that do not use PagePrivate/PagePrivate2.
C
Christoph Lameter 已提交
414 415 416
 *
 * Pages are locked upon entry and exit.
 */
417 418
int migrate_page(struct address_space *mapping,
		struct page *newpage, struct page *page)
C
Christoph Lameter 已提交
419 420 421 422 423
{
	int rc;

	BUG_ON(PageWriteback(page));	/* Writeback must be complete */

424
	rc = migrate_page_move_mapping(mapping, newpage, page);
C
Christoph Lameter 已提交
425 426 427 428 429 430 431 432 433

	if (rc)
		return rc;

	migrate_page_copy(newpage, page);
	return 0;
}
EXPORT_SYMBOL(migrate_page);

434
#ifdef CONFIG_BLOCK
435 436 437 438 439
/*
 * Migration function for pages with buffers. This function can only be used
 * if the underlying filesystem guarantees that no other references to "page"
 * exist.
 */
440 441
int buffer_migrate_page(struct address_space *mapping,
		struct page *newpage, struct page *page)
442 443 444 445 446
{
	struct buffer_head *bh, *head;
	int rc;

	if (!page_has_buffers(page))
447
		return migrate_page(mapping, newpage, page);
448 449 450

	head = page_buffers(page);

451
	rc = migrate_page_move_mapping(mapping, newpage, page);
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491

	if (rc)
		return rc;

	bh = head;
	do {
		get_bh(bh);
		lock_buffer(bh);
		bh = bh->b_this_page;

	} while (bh != head);

	ClearPagePrivate(page);
	set_page_private(newpage, page_private(page));
	set_page_private(page, 0);
	put_page(page);
	get_page(newpage);

	bh = head;
	do {
		set_bh_page(bh, newpage, bh_offset(bh));
		bh = bh->b_this_page;

	} while (bh != head);

	SetPagePrivate(newpage);

	migrate_page_copy(newpage, page);

	bh = head;
	do {
		unlock_buffer(bh);
 		put_bh(bh);
		bh = bh->b_this_page;

	} while (bh != head);

	return 0;
}
EXPORT_SYMBOL(buffer_migrate_page);
492
#endif
493

494 495 496 497
/*
 * Writeback a page to clean the dirty state
 */
static int writeout(struct address_space *mapping, struct page *page)
498
{
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_NONE,
		.nr_to_write = 1,
		.range_start = 0,
		.range_end = LLONG_MAX,
		.for_reclaim = 1
	};
	int rc;

	if (!mapping->a_ops->writepage)
		/* No write method for the address space */
		return -EINVAL;

	if (!clear_page_dirty_for_io(page))
		/* Someone else already triggered a write */
		return -EAGAIN;

516
	/*
517 518 519 520 521 522
	 * A dirty page may imply that the underlying filesystem has
	 * the page on some queue. So the page must be clean for
	 * migration. Writeout may mean we loose the lock and the
	 * page state is no longer what we checked for earlier.
	 * At this point we know that the migration attempt cannot
	 * be successful.
523
	 */
524
	remove_migration_ptes(page, page);
525

526
	rc = mapping->a_ops->writepage(page, &wbc);
527

528 529 530 531
	if (rc != AOP_WRITEPAGE_ACTIVATE)
		/* unlocked. Relock */
		lock_page(page);

H
Hugh Dickins 已提交
532
	return (rc < 0) ? -EIO : -EAGAIN;
533 534 535 536 537 538 539 540 541 542
}

/*
 * Default handling if a filesystem does not provide a migration function.
 */
static int fallback_migrate_page(struct address_space *mapping,
	struct page *newpage, struct page *page)
{
	if (PageDirty(page))
		return writeout(mapping, page);
543 544 545 546 547

	/*
	 * Buffers may be managed in a filesystem specific way.
	 * We must have no buffers or drop them.
	 */
548
	if (page_has_private(page) &&
549 550 551 552 553 554
	    !try_to_release_page(page, GFP_KERNEL))
		return -EAGAIN;

	return migrate_page(mapping, newpage, page);
}

555 556 557 558 559 560
/*
 * Move a page to a newly allocated page
 * The page is locked and all ptes have been successfully removed.
 *
 * The new page will have replaced the old page if this function
 * is successful.
L
Lee Schermerhorn 已提交
561 562 563 564
 *
 * Return value:
 *   < 0 - error code
 *  == 0 - success
565
 */
566
static int move_to_new_page(struct page *newpage, struct page *page,
567
					int remap_swapcache, bool sync)
568 569 570 571 572 573 574 575 576
{
	struct address_space *mapping;
	int rc;

	/*
	 * Block others from accessing the page when we get around to
	 * establishing additional references. We are the only one
	 * holding a reference to the new page at this point.
	 */
N
Nick Piggin 已提交
577
	if (!trylock_page(newpage))
578 579 580 581 582
		BUG();

	/* Prepare mapping for the new page.*/
	newpage->index = page->index;
	newpage->mapping = page->mapping;
R
Rik van Riel 已提交
583 584
	if (PageSwapBacked(page))
		SetPageSwapBacked(newpage);
585 586 587 588

	mapping = page_mapping(page);
	if (!mapping)
		rc = migrate_page(mapping, newpage, page);
589
	else {
590
		/*
591 592 593
		 * Do not writeback pages if !sync and migratepage is
		 * not pointing to migrate_page() which is nonblocking
		 * (swapcache/tmpfs uses migratepage = migrate_page).
594
		 */
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610
		if (PageDirty(page) && !sync &&
		    mapping->a_ops->migratepage != migrate_page)
			rc = -EBUSY;
		else if (mapping->a_ops->migratepage)
			/*
			 * Most pages have a mapping and most filesystems
			 * should provide a migration function. Anonymous
			 * pages are part of swap space which also has its
			 * own migration function. This is the most common
			 * path for page migration.
			 */
			rc = mapping->a_ops->migratepage(mapping,
							newpage, page);
		else
			rc = fallback_migrate_page(mapping, newpage, page);
	}
611

612
	if (rc) {
613
		newpage->mapping = NULL;
614 615 616 617
	} else {
		if (remap_swapcache)
			remove_migration_ptes(page, newpage);
	}
618 619 620 621 622 623 624 625 626 627

	unlock_page(newpage);

	return rc;
}

/*
 * Obtain the lock on page, remove all ptes and migrate the page
 * to the newly allocated page in newpage.
 */
628
static int unmap_and_move(new_page_t get_new_page, unsigned long private,
629
			struct page *page, int force, bool offlining, bool sync)
630 631
{
	int rc = 0;
632 633
	int *result = NULL;
	struct page *newpage = get_new_page(page, private, &result);
634
	int remap_swapcache = 1;
635
	int charge = 0;
636
	struct mem_cgroup *mem;
637
	struct anon_vma *anon_vma = NULL;
638 639 640

	if (!newpage)
		return -ENOMEM;
641

L
Lee Schermerhorn 已提交
642
	if (page_count(page) == 1) {
643
		/* page was freed from under us. So we are done. */
644
		goto move_newpage;
L
Lee Schermerhorn 已提交
645
	}
646 647 648
	if (unlikely(PageTransHuge(page)))
		if (unlikely(split_huge_page(page)))
			goto move_newpage;
649

650
	/* prepare cgroup just returns 0 or -ENOMEM */
651
	rc = -EAGAIN;
652

N
Nick Piggin 已提交
653
	if (!trylock_page(page)) {
654
		if (!force || !sync)
655
			goto move_newpage;
656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672

		/*
		 * It's not safe for direct compaction to call lock_page.
		 * For example, during page readahead pages are added locked
		 * to the LRU. Later, when the IO completes the pages are
		 * marked uptodate and unlocked. However, the queueing
		 * could be merging multiple pages for one bio (e.g.
		 * mpage_readpages). If an allocation happens for the
		 * second or third page, the process can end up locking
		 * the same page twice and deadlocking. Rather than
		 * trying to be clever about what pages can be locked,
		 * avoid the use of lock_page for direct compaction
		 * altogether.
		 */
		if (current->flags & PF_MEMALLOC)
			goto move_newpage;

673 674 675
		lock_page(page);
	}

676 677 678 679 680 681 682 683 684 685 686 687 688 689
	/*
	 * Only memory hotplug's offline_pages() caller has locked out KSM,
	 * and can safely migrate a KSM page.  The other cases have skipped
	 * PageKsm along with PageReserved - but it is only now when we have
	 * the page lock that we can be certain it will not go KSM beneath us
	 * (KSM will not upgrade a page from PageAnon to PageKsm when it sees
	 * its pagecount raised, but only here do we take the page lock which
	 * serializes that).
	 */
	if (PageKsm(page) && !offlining) {
		rc = -EBUSY;
		goto unlock;
	}

690
	/* charge against new page */
691
	charge = mem_cgroup_prepare_migration(page, newpage, &mem, GFP_KERNEL);
692 693 694 695 696 697
	if (charge == -ENOMEM) {
		rc = -ENOMEM;
		goto unlock;
	}
	BUG_ON(charge);

698
	if (PageWriteback(page)) {
699 700 701 702 703 704 705 706 707
		/*
		 * For !sync, there is no point retrying as the retry loop
		 * is expected to be too short for PageWriteback to be cleared
		 */
		if (!sync) {
			rc = -EBUSY;
			goto uncharge;
		}
		if (!force)
708
			goto uncharge;
709 710 711
		wait_on_page_writeback(page);
	}
	/*
712 713
	 * By try_to_unmap(), page->mapcount goes down to 0 here. In this case,
	 * we cannot notice that anon_vma is freed while we migrates a page.
714
	 * This get_anon_vma() delays freeing anon_vma pointer until the end
715
	 * of migration. File cache pages are no problem because of page_lock()
716 717
	 * File Caches may use write_page() or lock_page() in migration, then,
	 * just care Anon page here.
718
	 */
719
	if (PageAnon(page)) {
720 721 722 723
		/*
		 * Only page_lock_anon_vma() understands the subtleties of
		 * getting a hold on an anon_vma from outside one of its mms.
		 */
724
		anon_vma = page_get_anon_vma(page);
725 726
		if (anon_vma) {
			/*
727
			 * Anon page
728 729
			 */
		} else if (PageSwapCache(page)) {
730 731 732 733 734 735 736 737 738 739 740 741 742 743
			/*
			 * We cannot be sure that the anon_vma of an unmapped
			 * swapcache page is safe to use because we don't
			 * know in advance if the VMA that this page belonged
			 * to still exists. If the VMA and others sharing the
			 * data have been freed, then the anon_vma could
			 * already be invalid.
			 *
			 * To avoid this possibility, swapcache pages get
			 * migrated but are not remapped when migration
			 * completes
			 */
			remap_swapcache = 0;
		} else {
744
			goto uncharge;
745
		}
746
	}
747

748
	/*
749 750 751 752 753 754 755 756 757 758
	 * Corner case handling:
	 * 1. When a new swap-cache page is read into, it is added to the LRU
	 * and treated as swapcache but it has no rmap yet.
	 * Calling try_to_unmap() against a page->mapping==NULL page will
	 * trigger a BUG.  So handle it here.
	 * 2. An orphaned page (see truncate_complete_page) might have
	 * fs-private metadata. The page can be picked up due to memory
	 * offlining.  Everywhere else except page reclaim, the page is
	 * invisible to the vm, so the page can not be migrated.  So try to
	 * free the metadata, so the page can be freed.
759
	 */
760
	if (!page->mapping) {
761 762
		VM_BUG_ON(PageAnon(page));
		if (page_has_private(page)) {
763
			try_to_free_buffers(page);
764
			goto uncharge;
765
		}
766
		goto skip_unmap;
767 768
	}

769
	/* Establish migration ptes or remove ptes */
770
	try_to_unmap(page, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);
771

772
skip_unmap:
773
	if (!page_mapped(page))
774
		rc = move_to_new_page(newpage, page, remap_swapcache, sync);
775

776
	if (rc && remap_swapcache)
777
		remove_migration_ptes(page, page);
778 779

	/* Drop an anon_vma reference if we took one */
780
	if (anon_vma)
781
		put_anon_vma(anon_vma);
782

783 784
uncharge:
	if (!charge)
785
		mem_cgroup_end_migration(mem, page, newpage, rc == 0);
786 787
unlock:
	unlock_page(page);
788

789
move_newpage:
790
	if (rc != -EAGAIN) {
791 792 793 794 795 796 797
 		/*
 		 * A page that has been migrated has all references
 		 * removed and will be freed. A page that has not been
 		 * migrated will have kepts its references and be
 		 * restored.
 		 */
 		list_del(&page->lru);
K
KOSAKI Motohiro 已提交
798
		dec_zone_page_state(page, NR_ISOLATED_ANON +
799
				page_is_file_cache(page));
L
Lee Schermerhorn 已提交
800
		putback_lru_page(page);
801
	}
802 803 804 805 806

	/*
	 * Move the new page to the LRU. If migration was not successful
	 * then this will free the page.
	 */
L
Lee Schermerhorn 已提交
807 808
	putback_lru_page(newpage);

809 810 811 812 813 814
	if (result) {
		if (rc)
			*result = rc;
		else
			*result = page_to_nid(newpage);
	}
815 816 817
	return rc;
}

N
Naoya Horiguchi 已提交
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837
/*
 * Counterpart of unmap_and_move_page() for hugepage migration.
 *
 * This function doesn't wait the completion of hugepage I/O
 * because there is no race between I/O and migration for hugepage.
 * Note that currently hugepage I/O occurs only in direct I/O
 * where no lock is held and PG_writeback is irrelevant,
 * and writeback status of all subpages are counted in the reference
 * count of the head page (i.e. if all subpages of a 2MB hugepage are
 * under direct I/O, the reference of the head page is 512 and a bit more.)
 * This means that when we try to migrate hugepage whose subpages are
 * doing direct I/O, some references remain after try_to_unmap() and
 * hugepage migration fails without data corruption.
 *
 * There is also no race when direct I/O is issued on the page under migration,
 * because then pte is replaced with migration swap entry and direct I/O code
 * will wait in the page fault for migration to complete.
 */
static int unmap_and_move_huge_page(new_page_t get_new_page,
				unsigned long private, struct page *hpage,
838
				int force, bool offlining, bool sync)
N
Naoya Horiguchi 已提交
839 840 841 842 843 844 845 846 847 848 849 850
{
	int rc = 0;
	int *result = NULL;
	struct page *new_hpage = get_new_page(hpage, private, &result);
	struct anon_vma *anon_vma = NULL;

	if (!new_hpage)
		return -ENOMEM;

	rc = -EAGAIN;

	if (!trylock_page(hpage)) {
851
		if (!force || !sync)
N
Naoya Horiguchi 已提交
852 853 854 855
			goto out;
		lock_page(hpage);
	}

856 857
	if (PageAnon(hpage))
		anon_vma = page_get_anon_vma(hpage);
N
Naoya Horiguchi 已提交
858 859 860 861

	try_to_unmap(hpage, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);

	if (!page_mapped(hpage))
862
		rc = move_to_new_page(new_hpage, hpage, 1, sync);
N
Naoya Horiguchi 已提交
863 864 865 866

	if (rc)
		remove_migration_ptes(hpage, hpage);

H
Hugh Dickins 已提交
867
	if (anon_vma)
868
		put_anon_vma(anon_vma);
N
Naoya Horiguchi 已提交
869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
out:
	unlock_page(hpage);

	if (rc != -EAGAIN) {
		list_del(&hpage->lru);
		put_page(hpage);
	}

	put_page(new_hpage);

	if (result) {
		if (rc)
			*result = rc;
		else
			*result = page_to_nid(new_hpage);
	}
	return rc;
}

C
Christoph Lameter 已提交
888 889 890
/*
 * migrate_pages
 *
891 892 893
 * The function takes one list of pages to migrate and a function
 * that determines from the page to be migrated and the private data
 * the target of the move and allocates the page.
C
Christoph Lameter 已提交
894 895 896
 *
 * The function returns after 10 attempts or if no pages
 * are movable anymore because to has become empty
897 898
 * or no retryable pages exist anymore.
 * Caller should call putback_lru_pages to return pages to the LRU
899
 * or free list only if ret != 0.
C
Christoph Lameter 已提交
900
 *
901
 * Return: Number of pages not migrated or error code.
C
Christoph Lameter 已提交
902
 */
903
int migrate_pages(struct list_head *from,
904
		new_page_t get_new_page, unsigned long private, bool offlining,
905
		bool sync)
C
Christoph Lameter 已提交
906
{
907
	int retry = 1;
C
Christoph Lameter 已提交
908 909 910 911 912 913 914 915 916 917
	int nr_failed = 0;
	int pass = 0;
	struct page *page;
	struct page *page2;
	int swapwrite = current->flags & PF_SWAPWRITE;
	int rc;

	if (!swapwrite)
		current->flags |= PF_SWAPWRITE;

918 919
	for(pass = 0; pass < 10 && retry; pass++) {
		retry = 0;
C
Christoph Lameter 已提交
920

921 922
		list_for_each_entry_safe(page, page2, from, lru) {
			cond_resched();
923

924
			rc = unmap_and_move(get_new_page, private,
925 926
						page, pass > 2, offlining,
						sync);
927

928
			switch(rc) {
929 930
			case -ENOMEM:
				goto out;
931
			case -EAGAIN:
932
				retry++;
933 934 935 936
				break;
			case 0:
				break;
			default:
937 938
				/* Permanent failure */
				nr_failed++;
939
				break;
940
			}
C
Christoph Lameter 已提交
941 942
		}
	}
943 944
	rc = 0;
out:
C
Christoph Lameter 已提交
945 946 947
	if (!swapwrite)
		current->flags &= ~PF_SWAPWRITE;

948 949
	if (rc)
		return rc;
C
Christoph Lameter 已提交
950

951
	return nr_failed + retry;
C
Christoph Lameter 已提交
952
}
953

N
Naoya Horiguchi 已提交
954
int migrate_huge_pages(struct list_head *from,
955
		new_page_t get_new_page, unsigned long private, bool offlining,
956
		bool sync)
N
Naoya Horiguchi 已提交
957 958 959 960 961 962 963 964 965 966 967 968 969 970 971
{
	int retry = 1;
	int nr_failed = 0;
	int pass = 0;
	struct page *page;
	struct page *page2;
	int rc;

	for (pass = 0; pass < 10 && retry; pass++) {
		retry = 0;

		list_for_each_entry_safe(page, page2, from, lru) {
			cond_resched();

			rc = unmap_and_move_huge_page(get_new_page,
972 973
					private, page, pass > 2, offlining,
					sync);
N
Naoya Horiguchi 已提交
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997

			switch(rc) {
			case -ENOMEM:
				goto out;
			case -EAGAIN:
				retry++;
				break;
			case 0:
				break;
			default:
				/* Permanent failure */
				nr_failed++;
				break;
			}
		}
	}
	rc = 0;
out:
	if (rc)
		return rc;

	return nr_failed + retry;
}

998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
#ifdef CONFIG_NUMA
/*
 * Move a list of individual pages
 */
struct page_to_node {
	unsigned long addr;
	struct page *page;
	int node;
	int status;
};

static struct page *new_page_node(struct page *p, unsigned long private,
		int **result)
{
	struct page_to_node *pm = (struct page_to_node *)private;

	while (pm->node != MAX_NUMNODES && pm->page != p)
		pm++;

	if (pm->node == MAX_NUMNODES)
		return NULL;

	*result = &pm->status;

1022
	return alloc_pages_exact_node(pm->node,
1023
				GFP_HIGHUSER_MOVABLE | GFP_THISNODE, 0);
1024 1025 1026 1027 1028 1029
}

/*
 * Move a set of pages as indicated in the pm array. The addr
 * field must be set to the virtual address of the page to be moved
 * and the node number must contain a valid target node.
1030
 * The pm array ends with node = MAX_NUMNODES.
1031
 */
1032 1033 1034
static int do_move_page_to_node_array(struct mm_struct *mm,
				      struct page_to_node *pm,
				      int migrate_all)
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
{
	int err;
	struct page_to_node *pp;
	LIST_HEAD(pagelist);

	down_read(&mm->mmap_sem);

	/*
	 * Build a list of pages to migrate
	 */
	for (pp = pm; pp->node != MAX_NUMNODES; pp++) {
		struct vm_area_struct *vma;
		struct page *page;

		err = -EFAULT;
		vma = find_vma(mm, pp->addr);
1051
		if (!vma || pp->addr < vma->vm_start || !vma_migratable(vma))
1052 1053
			goto set_status;

1054
		page = follow_page(vma, pp->addr, FOLL_GET|FOLL_SPLIT);
1055 1056 1057 1058 1059

		err = PTR_ERR(page);
		if (IS_ERR(page))
			goto set_status;

1060 1061 1062 1063
		err = -ENOENT;
		if (!page)
			goto set_status;

1064 1065
		/* Use PageReserved to check for zero page */
		if (PageReserved(page) || PageKsm(page))
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
			goto put_and_set;

		pp->page = page;
		err = page_to_nid(page);

		if (err == pp->node)
			/*
			 * Node already in the right place
			 */
			goto put_and_set;

		err = -EACCES;
		if (page_mapcount(page) > 1 &&
				!migrate_all)
			goto put_and_set;

1082
		err = isolate_lru_page(page);
1083
		if (!err) {
1084
			list_add_tail(&page->lru, &pagelist);
1085 1086 1087
			inc_zone_page_state(page, NR_ISOLATED_ANON +
					    page_is_file_cache(page));
		}
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
put_and_set:
		/*
		 * Either remove the duplicate refcount from
		 * isolate_lru_page() or drop the page ref if it was
		 * not isolated.
		 */
		put_page(page);
set_status:
		pp->status = err;
	}

1099
	err = 0;
1100
	if (!list_empty(&pagelist)) {
1101
		err = migrate_pages(&pagelist, new_page_node,
1102
				(unsigned long)pm, 0, true);
1103 1104 1105
		if (err)
			putback_lru_pages(&pagelist);
	}
1106 1107 1108 1109 1110

	up_read(&mm->mmap_sem);
	return err;
}

1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
/*
 * Migrate an array of page address onto an array of nodes and fill
 * the corresponding array of status.
 */
static int do_pages_move(struct mm_struct *mm, struct task_struct *task,
			 unsigned long nr_pages,
			 const void __user * __user *pages,
			 const int __user *nodes,
			 int __user *status, int flags)
{
1121
	struct page_to_node *pm;
1122
	nodemask_t task_nodes;
1123 1124 1125
	unsigned long chunk_nr_pages;
	unsigned long chunk_start;
	int err;
1126 1127 1128

	task_nodes = cpuset_mems_allowed(task);

1129 1130 1131
	err = -ENOMEM;
	pm = (struct page_to_node *)__get_free_page(GFP_KERNEL);
	if (!pm)
1132
		goto out;
1133 1134 1135

	migrate_prep();

1136
	/*
1137 1138
	 * Store a chunk of page_to_node array in a page,
	 * but keep the last one as a marker
1139
	 */
1140
	chunk_nr_pages = (PAGE_SIZE / sizeof(struct page_to_node)) - 1;
1141

1142 1143 1144 1145
	for (chunk_start = 0;
	     chunk_start < nr_pages;
	     chunk_start += chunk_nr_pages) {
		int j;
1146

1147 1148 1149 1150 1151 1152
		if (chunk_start + chunk_nr_pages > nr_pages)
			chunk_nr_pages = nr_pages - chunk_start;

		/* fill the chunk pm with addrs and nodes from user-space */
		for (j = 0; j < chunk_nr_pages; j++) {
			const void __user *p;
1153 1154
			int node;

1155 1156 1157 1158 1159 1160
			err = -EFAULT;
			if (get_user(p, pages + j + chunk_start))
				goto out_pm;
			pm[j].addr = (unsigned long) p;

			if (get_user(node, nodes + j + chunk_start))
1161 1162 1163
				goto out_pm;

			err = -ENODEV;
1164 1165 1166
			if (node < 0 || node >= MAX_NUMNODES)
				goto out_pm;

1167 1168 1169 1170 1171 1172 1173
			if (!node_state(node, N_HIGH_MEMORY))
				goto out_pm;

			err = -EACCES;
			if (!node_isset(node, task_nodes))
				goto out_pm;

1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
			pm[j].node = node;
		}

		/* End marker for this chunk */
		pm[chunk_nr_pages].node = MAX_NUMNODES;

		/* Migrate this chunk */
		err = do_move_page_to_node_array(mm, pm,
						 flags & MPOL_MF_MOVE_ALL);
		if (err < 0)
			goto out_pm;
1185 1186

		/* Return status information */
1187 1188
		for (j = 0; j < chunk_nr_pages; j++)
			if (put_user(pm[j].status, status + j + chunk_start)) {
1189
				err = -EFAULT;
1190 1191 1192 1193
				goto out_pm;
			}
	}
	err = 0;
1194 1195

out_pm:
1196
	free_page((unsigned long)pm);
1197 1198 1199 1200
out:
	return err;
}

1201
/*
1202
 * Determine the nodes of an array of pages and store it in an array of status.
1203
 */
1204 1205
static void do_pages_stat_array(struct mm_struct *mm, unsigned long nr_pages,
				const void __user **pages, int *status)
1206
{
1207 1208
	unsigned long i;

1209 1210
	down_read(&mm->mmap_sem);

1211
	for (i = 0; i < nr_pages; i++) {
1212
		unsigned long addr = (unsigned long)(*pages);
1213 1214
		struct vm_area_struct *vma;
		struct page *page;
1215
		int err = -EFAULT;
1216 1217

		vma = find_vma(mm, addr);
1218
		if (!vma || addr < vma->vm_start)
1219 1220
			goto set_status;

1221
		page = follow_page(vma, addr, 0);
1222 1223 1224 1225 1226

		err = PTR_ERR(page);
		if (IS_ERR(page))
			goto set_status;

1227 1228
		err = -ENOENT;
		/* Use PageReserved to check for zero page */
1229
		if (!page || PageReserved(page) || PageKsm(page))
1230 1231 1232 1233
			goto set_status;

		err = page_to_nid(page);
set_status:
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254
		*status = err;

		pages++;
		status++;
	}

	up_read(&mm->mmap_sem);
}

/*
 * Determine the nodes of a user array of pages and store it in
 * a user array of status.
 */
static int do_pages_stat(struct mm_struct *mm, unsigned long nr_pages,
			 const void __user * __user *pages,
			 int __user *status)
{
#define DO_PAGES_STAT_CHUNK_NR 16
	const void __user *chunk_pages[DO_PAGES_STAT_CHUNK_NR];
	int chunk_status[DO_PAGES_STAT_CHUNK_NR];

1255 1256
	while (nr_pages) {
		unsigned long chunk_nr;
1257

1258 1259 1260 1261 1262 1263
		chunk_nr = nr_pages;
		if (chunk_nr > DO_PAGES_STAT_CHUNK_NR)
			chunk_nr = DO_PAGES_STAT_CHUNK_NR;

		if (copy_from_user(chunk_pages, pages, chunk_nr * sizeof(*chunk_pages)))
			break;
1264 1265 1266

		do_pages_stat_array(mm, chunk_nr, chunk_pages, chunk_status);

1267 1268
		if (copy_to_user(status, chunk_status, chunk_nr * sizeof(*status)))
			break;
1269

1270 1271 1272 1273 1274
		pages += chunk_nr;
		status += chunk_nr;
		nr_pages -= chunk_nr;
	}
	return nr_pages ? -EFAULT : 0;
1275 1276 1277 1278 1279 1280
}

/*
 * Move a list of pages in the address space of the currently executing
 * process.
 */
1281 1282 1283 1284
SYSCALL_DEFINE6(move_pages, pid_t, pid, unsigned long, nr_pages,
		const void __user * __user *, pages,
		const int __user *, nodes,
		int __user *, status, int, flags)
1285
{
1286
	const struct cred *cred = current_cred(), *tcred;
1287 1288
	struct task_struct *task;
	struct mm_struct *mm;
1289
	int err;
1290 1291 1292 1293 1294 1295 1296 1297 1298

	/* Check flags */
	if (flags & ~(MPOL_MF_MOVE|MPOL_MF_MOVE_ALL))
		return -EINVAL;

	if ((flags & MPOL_MF_MOVE_ALL) && !capable(CAP_SYS_NICE))
		return -EPERM;

	/* Find the mm_struct */
1299
	rcu_read_lock();
1300
	task = pid ? find_task_by_vpid(pid) : current;
1301
	if (!task) {
1302
		rcu_read_unlock();
1303 1304 1305
		return -ESRCH;
	}
	mm = get_task_mm(task);
1306
	rcu_read_unlock();
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316

	if (!mm)
		return -EINVAL;

	/*
	 * Check if this process has the right to modify the specified
	 * process. The right exists if the process has administrative
	 * capabilities, superuser privileges or the same
	 * userid as the target process.
	 */
1317 1318
	rcu_read_lock();
	tcred = __task_cred(task);
1319 1320
	if (cred->euid != tcred->suid && cred->euid != tcred->uid &&
	    cred->uid  != tcred->suid && cred->uid  != tcred->uid &&
1321
	    !capable(CAP_SYS_NICE)) {
1322
		rcu_read_unlock();
1323
		err = -EPERM;
1324
		goto out;
1325
	}
1326
	rcu_read_unlock();
1327

1328 1329
 	err = security_task_movememory(task);
 	if (err)
1330
		goto out;
1331

1332 1333 1334 1335
	if (nodes) {
		err = do_pages_move(mm, task, nr_pages, pages, nodes, status,
				    flags);
	} else {
1336
		err = do_pages_stat(mm, nr_pages, pages, status);
1337 1338 1339 1340 1341 1342 1343
	}

out:
	mmput(mm);
	return err;
}

1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
/*
 * 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;

1355
	for (vma = mm->mmap; vma && !err; vma = vma->vm_next) {
1356 1357 1358 1359 1360 1361 1362 1363
 		if (vma->vm_ops && vma->vm_ops->migrate) {
 			err = vma->vm_ops->migrate(vma, to, from, flags);
 			if (err)
 				break;
 		}
 	}
 	return err;
}
1364
#endif