migrate.c 48.9 KB
Newer Older
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>
Christoph Lameter's avatar
Christoph Lameter committed
12
 * Christoph Lameter
13 14 15
 */

#include <linux/migrate.h>
16
#include <linux/export.h>
17
#include <linux/swap.h>
18
#include <linux/swapops.h>
19
#include <linux/pagemap.h>
20
#include <linux/buffer_head.h>
21
#include <linux/mm_inline.h>
22
#include <linux/nsproxy.h>
23
#include <linux/pagevec.h>
24
#include <linux/ksm.h>
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>
35
#include <linux/hugetlb.h>
36
#include <linux/hugetlb_cgroup.h>
37
#include <linux/gfp.h>
38
#include <linux/balloon_compaction.h>
39
#include <linux/mmu_notifier.h>
40

41 42
#include <asm/tlbflush.h>

43 44 45
#define CREATE_TRACE_POINTS
#include <trace/events/migrate.h>

46 47 48
#include "internal.h"

/*
49
 * migrate_prep() needs to be called before we start compiling a list of pages
50 51
 * to be migrated using isolate_lru_page(). If scheduling work on other CPUs is
 * undesirable, use migrate_prep_local()
52 53 54 55 56 57 58 59 60 61 62 63 64 65
 */
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;
}

66 67 68 69 70 71 72 73
/* Do the necessary work of migrate_prep but not if it involves other CPUs */
int migrate_prep_local(void)
{
	lru_add_drain();

	return 0;
}

74 75 76 77
/*
 * Put previously isolated pages back onto the appropriate lists
 * from where they were once taken off for compaction/migration.
 *
78 79 80
 * This function shall be used whenever the isolated pageset has been
 * built from lru, balloon, hugetlbfs page. See isolate_migratepages_range()
 * and isolate_huge_page().
81 82 83 84 85 86
 */
void putback_movable_pages(struct list_head *l)
{
	struct page *page;
	struct page *page2;

87
	list_for_each_entry_safe(page, page2, l, lru) {
88 89 90 91
		if (unlikely(PageHuge(page))) {
			putback_active_hugepage(page);
			continue;
		}
92
		list_del(&page->lru);
93
		dec_zone_page_state(page, NR_ISOLATED_ANON +
94
				page_is_file_cache(page));
95
		if (unlikely(isolated_balloon_page(page)))
96 97 98
			balloon_page_putback(page);
		else
			putback_lru_page(page);
99 100 101
	}
}

102 103 104
/*
 * Restore a potential migration pte to a working pte entry
 */
105 106
static int remove_migration_pte(struct page *new, struct vm_area_struct *vma,
				 unsigned long addr, void *old)
107 108 109 110 111 112 113
{
	struct mm_struct *mm = vma->vm_mm;
	swp_entry_t entry;
 	pmd_t *pmd;
	pte_t *ptep, pte;
 	spinlock_t *ptl;

114 115 116 117
	if (unlikely(PageHuge(new))) {
		ptep = huge_pte_offset(mm, addr);
		if (!ptep)
			goto out;
118
		ptl = huge_pte_lockptr(hstate_vma(vma), mm, ptep);
119
	} else {
Bob Liu's avatar
Bob Liu committed
120 121
		pmd = mm_find_pmd(mm, addr);
		if (!pmd)
122
			goto out;
123

124
		ptep = pte_offset_map(pmd, addr);
125

126 127 128 129
		/*
		 * Peek to check is_swap_pte() before taking ptlock?  No, we
		 * can race mremap's move_ptes(), which skips anon_vma lock.
		 */
130 131 132

		ptl = pte_lockptr(mm, pmd);
	}
133 134 135 136

 	spin_lock(ptl);
	pte = *ptep;
	if (!is_swap_pte(pte))
137
		goto unlock;
138 139 140

	entry = pte_to_swp_entry(pte);

141 142 143
	if (!is_migration_entry(entry) ||
	    migration_entry_to_page(entry) != old)
		goto unlock;
144 145 146

	get_page(new);
	pte = pte_mkold(mk_pte(new, vma->vm_page_prot));
147 148
	if (pte_swp_soft_dirty(*ptep))
		pte = pte_mksoft_dirty(pte);
149 150

	/* Recheck VMA as permissions can change since migration started  */
151
	if (is_write_migration_entry(entry))
152 153
		pte = maybe_mkwrite(pte, vma);

154
#ifdef CONFIG_HUGETLB_PAGE
155
	if (PageHuge(new)) {
156
		pte = pte_mkhuge(pte);
157 158
		pte = arch_make_huge_pte(pte, vma, new, 0);
	}
159
#endif
160
	flush_dcache_page(new);
161
	set_pte_at(mm, addr, ptep, pte);
162

163 164 165 166 167 168
	if (PageHuge(new)) {
		if (PageAnon(new))
			hugepage_add_anon_rmap(new, vma, addr);
		else
			page_dup_rmap(new);
	} else if (PageAnon(new))
169 170 171 172 173
		page_add_anon_rmap(new, vma, addr);
	else
		page_add_file_rmap(new);

	/* No need to invalidate - it was non-present before */
174
	update_mmu_cache(vma, addr, ptep);
175
unlock:
176
	pte_unmap_unlock(ptep, ptl);
177 178
out:
	return SWAP_AGAIN;
179 180
}

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
/*
 * Congratulations to trinity for discovering this bug.
 * mm/fremap.c's remap_file_pages() accepts any range within a single vma to
 * convert that vma to VM_NONLINEAR; and generic_file_remap_pages() will then
 * replace the specified range by file ptes throughout (maybe populated after).
 * If page migration finds a page within that range, while it's still located
 * by vma_interval_tree rather than lost to i_mmap_nonlinear list, no problem:
 * zap_pte() clears the temporary migration entry before mmap_sem is dropped.
 * But if the migrating page is in a part of the vma outside the range to be
 * remapped, then it will not be cleared, and remove_migration_ptes() needs to
 * deal with it.  Fortunately, this part of the vma is of course still linear,
 * so we just need to use linear location on the nonlinear list.
 */
static int remove_linear_migration_ptes_from_nonlinear(struct page *page,
		struct address_space *mapping, void *arg)
{
	struct vm_area_struct *vma;
	/* hugetlbfs does not support remap_pages, so no huge pgoff worries */
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	unsigned long addr;

	list_for_each_entry(vma,
		&mapping->i_mmap_nonlinear, shared.nonlinear) {

		addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
		if (addr >= vma->vm_start && addr < vma->vm_end)
			remove_migration_pte(page, vma, addr, arg);
	}
	return SWAP_AGAIN;
}

212 213 214 215 216 217
/*
 * 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)
{
218 219 220
	struct rmap_walk_control rwc = {
		.rmap_one = remove_migration_pte,
		.arg = old,
221
		.file_nonlinear = remove_linear_migration_ptes_from_nonlinear,
222 223 224
	};

	rmap_walk(new, &rwc);
225 226
}

227 228 229 230 231
/*
 * 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.
 */
232 233
static void __migration_entry_wait(struct mm_struct *mm, pte_t *ptep,
				spinlock_t *ptl)
234
{
235
	pte_t pte;
236 237 238
	swp_entry_t entry;
	struct page *page;

239
	spin_lock(ptl);
240 241 242 243 244 245 246 247 248 249
	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);

250 251 252 253 254 255 256 257 258
	/*
	 * 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;
259 260 261 262 263 264 265 266
	pte_unmap_unlock(ptep, ptl);
	wait_on_page_locked(page);
	put_page(page);
	return;
out:
	pte_unmap_unlock(ptep, ptl);
}

267 268 269 270 271 272 273 274
void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
				unsigned long address)
{
	spinlock_t *ptl = pte_lockptr(mm, pmd);
	pte_t *ptep = pte_offset_map(pmd, address);
	__migration_entry_wait(mm, ptep, ptl);
}

275 276
void migration_entry_wait_huge(struct vm_area_struct *vma,
		struct mm_struct *mm, pte_t *pte)
277
{
278
	spinlock_t *ptl = huge_pte_lockptr(hstate_vma(vma), mm, pte);
279 280 281
	__migration_entry_wait(mm, pte, ptl);
}

282 283
#ifdef CONFIG_BLOCK
/* Returns true if all buffers are successfully locked */
284 285
static bool buffer_migrate_lock_buffers(struct buffer_head *head,
							enum migrate_mode mode)
286 287 288 289
{
	struct buffer_head *bh = head;

	/* Simple case, sync compaction */
290
	if (mode != MIGRATE_ASYNC) {
291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
		do {
			get_bh(bh);
			lock_buffer(bh);
			bh = bh->b_this_page;

		} while (bh != head);

		return true;
	}

	/* async case, we cannot block on lock_buffer so use trylock_buffer */
	do {
		get_bh(bh);
		if (!trylock_buffer(bh)) {
			/*
			 * We failed to lock the buffer and cannot stall in
			 * async migration. Release the taken locks
			 */
			struct buffer_head *failed_bh = bh;
			put_bh(failed_bh);
			bh = head;
			while (bh != failed_bh) {
				unlock_buffer(bh);
				put_bh(bh);
				bh = bh->b_this_page;
			}
			return false;
		}

		bh = bh->b_this_page;
	} while (bh != head);
	return true;
}
#else
static inline bool buffer_migrate_lock_buffers(struct buffer_head *head,
326
							enum migrate_mode mode)
327 328 329 330 331
{
	return true;
}
#endif /* CONFIG_BLOCK */

332
/*
333
 * Replace the page in the mapping.
334 335 336 337
 *
 * The number of remaining references must be:
 * 1 for anonymous pages without a mapping
 * 2 for pages with a mapping
338
 * 3 for pages with a mapping and PagePrivate/PagePrivate2 set.
339
 */
340
int migrate_page_move_mapping(struct address_space *mapping,
341
		struct page *newpage, struct page *page,
342 343
		struct buffer_head *head, enum migrate_mode mode,
		int extra_count)
344
{
345
	int expected_count = 1 + extra_count;
346
	void **pslot;
347

348
	if (!mapping) {
349
		/* Anonymous page without mapping */
350
		if (page_count(page) != expected_count)
351
			return -EAGAIN;
352
		return MIGRATEPAGE_SUCCESS;
353 354
	}

Nick Piggin's avatar
Nick Piggin committed
355
	spin_lock_irq(&mapping->tree_lock);
356

357 358
	pslot = radix_tree_lookup_slot(&mapping->page_tree,
 					page_index(page));
359

360
	expected_count += 1 + page_has_private(page);
361
	if (page_count(page) != expected_count ||
362
		radix_tree_deref_slot_protected(pslot, &mapping->tree_lock) != page) {
Nick Piggin's avatar
Nick Piggin committed
363
		spin_unlock_irq(&mapping->tree_lock);
364
		return -EAGAIN;
365 366
	}

367
	if (!page_freeze_refs(page, expected_count)) {
Nick Piggin's avatar
Nick Piggin committed
368
		spin_unlock_irq(&mapping->tree_lock);
369 370 371
		return -EAGAIN;
	}

372 373 374 375 376 377 378
	/*
	 * In the async migration case of moving a page with buffers, lock the
	 * buffers using trylock before the mapping is moved. If the mapping
	 * was moved, we later failed to lock the buffers and could not move
	 * the mapping back due to an elevated page count, we would have to
	 * block waiting on other references to be dropped.
	 */
379 380
	if (mode == MIGRATE_ASYNC && head &&
			!buffer_migrate_lock_buffers(head, mode)) {
381 382 383 384 385
		page_unfreeze_refs(page, expected_count);
		spin_unlock_irq(&mapping->tree_lock);
		return -EAGAIN;
	}

386 387 388
	/*
	 * Now we know that no one else is looking at the page.
	 */
389
	get_page(newpage);	/* add cache reference */
390 391 392 393 394
	if (PageSwapCache(page)) {
		SetPageSwapCache(newpage);
		set_page_private(newpage, page_private(page));
	}

395 396 397
	radix_tree_replace_slot(pslot, newpage);

	/*
398 399
	 * Drop cache reference from old page by unfreezing
	 * to one less reference.
400 401
	 * We know this isn't the last reference.
	 */
402
	page_unfreeze_refs(page, expected_count - 1);
403

404 405 406 407 408 409 410 411 412 413 414 415
	/*
	 * 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);
416
	if (!PageSwapCache(page) && PageSwapBacked(page)) {
417 418 419
		__dec_zone_page_state(page, NR_SHMEM);
		__inc_zone_page_state(newpage, NR_SHMEM);
	}
Nick Piggin's avatar
Nick Piggin committed
420
	spin_unlock_irq(&mapping->tree_lock);
421

422
	return MIGRATEPAGE_SUCCESS;
423 424
}

425 426 427 428 429 430 431 432 433 434 435 436 437
/*
 * 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;
438
		return MIGRATEPAGE_SUCCESS;
439 440 441 442 443 444 445 446 447
	}

	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 ||
448
		radix_tree_deref_slot_protected(pslot, &mapping->tree_lock) != page) {
449 450 451 452 453 454 455 456 457 458 459 460 461
		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);

462
	page_unfreeze_refs(page, expected_count - 1);
463 464

	spin_unlock_irq(&mapping->tree_lock);
465
	return MIGRATEPAGE_SUCCESS;
466 467
}

468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
/*
 * Gigantic pages are so large that we do not guarantee that page++ pointer
 * arithmetic will work across the entire page.  We need something more
 * specialized.
 */
static void __copy_gigantic_page(struct page *dst, struct page *src,
				int nr_pages)
{
	int i;
	struct page *dst_base = dst;
	struct page *src_base = src;

	for (i = 0; i < nr_pages; ) {
		cond_resched();
		copy_highpage(dst, src);

		i++;
		dst = mem_map_next(dst, dst_base, i);
		src = mem_map_next(src, src_base, i);
	}
}

static void copy_huge_page(struct page *dst, struct page *src)
{
	int i;
	int nr_pages;

	if (PageHuge(src)) {
		/* hugetlbfs page */
		struct hstate *h = page_hstate(src);
		nr_pages = pages_per_huge_page(h);

		if (unlikely(nr_pages > MAX_ORDER_NR_PAGES)) {
			__copy_gigantic_page(dst, src, nr_pages);
			return;
		}
	} else {
		/* thp page */
		BUG_ON(!PageTransHuge(src));
		nr_pages = hpage_nr_pages(src);
	}

	for (i = 0; i < nr_pages; i++) {
		cond_resched();
		copy_highpage(dst + i, src + i);
	}
}

516 517 518
/*
 * Copy the page to its new location
 */
519
void migrate_page_copy(struct page *newpage, struct page *page)
520
{
521 522
	int cpupid;

523
	if (PageHuge(page) || PageTransHuge(page))
524 525 526
		copy_huge_page(newpage, page);
	else
		copy_highpage(newpage, page);
527 528 529 530 531 532 533

	if (PageError(page))
		SetPageError(newpage);
	if (PageReferenced(page))
		SetPageReferenced(newpage);
	if (PageUptodate(page))
		SetPageUptodate(newpage);
534
	if (TestClearPageActive(page)) {
535
		VM_BUG_ON_PAGE(PageUnevictable(page), page);
536
		SetPageActive(newpage);
537 538
	} else if (TestClearPageUnevictable(page))
		SetPageUnevictable(newpage);
539 540 541 542 543 544 545
	if (PageChecked(page))
		SetPageChecked(newpage);
	if (PageMappedToDisk(page))
		SetPageMappedToDisk(newpage);

	if (PageDirty(page)) {
		clear_page_dirty_for_io(page);
546 547 548 549 550
		/*
		 * 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.
Lucas De Marchi's avatar
Lucas De Marchi committed
551
		 * Whereas only part of our page may be dirty.
552
		 */
553 554 555 556
		if (PageSwapBacked(page))
			SetPageDirty(newpage);
		else
			__set_page_dirty_nobuffers(newpage);
557 558
 	}

559 560 561 562 563 564 565
	/*
	 * Copy NUMA information to the new page, to prevent over-eager
	 * future migrations of this same page.
	 */
	cpupid = page_cpupid_xchg_last(page, -1);
	page_cpupid_xchg_last(newpage, cpupid);

566
	mlock_migrate_page(newpage, page);
567
	ksm_migrate_page(newpage, page);
568 569 570 571
	/*
	 * Please do not reorder this without considering how mm/ksm.c's
	 * get_ksm_page() depends upon ksm_migrate_page() and PageSwapCache().
	 */
572 573 574 575 576 577 578 579 580 581 582 583
	ClearPageSwapCache(page);
	ClearPagePrivate(page);
	set_page_private(page, 0);

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

584 585 586 587
/************************************************************
 *                    Migration functions
 ***********************************************************/

588 589
/*
 * Common logic to directly migrate a single page suitable for
590
 * pages that do not use PagePrivate/PagePrivate2.
591 592 593
 *
 * Pages are locked upon entry and exit.
 */
594
int migrate_page(struct address_space *mapping,
595 596
		struct page *newpage, struct page *page,
		enum migrate_mode mode)
597 598 599 600 601
{
	int rc;

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

602
	rc = migrate_page_move_mapping(mapping, newpage, page, NULL, mode, 0);
603

604
	if (rc != MIGRATEPAGE_SUCCESS)
605 606 607
		return rc;

	migrate_page_copy(newpage, page);
608
	return MIGRATEPAGE_SUCCESS;
609 610 611
}
EXPORT_SYMBOL(migrate_page);

612
#ifdef CONFIG_BLOCK
613 614 615 616 617
/*
 * Migration function for pages with buffers. This function can only be used
 * if the underlying filesystem guarantees that no other references to "page"
 * exist.
 */
618
int buffer_migrate_page(struct address_space *mapping,
619
		struct page *newpage, struct page *page, enum migrate_mode mode)
620 621 622 623 624
{
	struct buffer_head *bh, *head;
	int rc;

	if (!page_has_buffers(page))
625
		return migrate_page(mapping, newpage, page, mode);
626 627 628

	head = page_buffers(page);

629
	rc = migrate_page_move_mapping(mapping, newpage, page, head, mode, 0);
630

631
	if (rc != MIGRATEPAGE_SUCCESS)
632 633
		return rc;

634 635 636 637 638
	/*
	 * In the async case, migrate_page_move_mapping locked the buffers
	 * with an IRQ-safe spinlock held. In the sync case, the buffers
	 * need to be locked now
	 */
639 640
	if (mode != MIGRATE_ASYNC)
		BUG_ON(!buffer_migrate_lock_buffers(head, mode));
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666

	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);

667
	return MIGRATEPAGE_SUCCESS;
668 669
}
EXPORT_SYMBOL(buffer_migrate_page);
670
#endif
671

672 673 674 675
/*
 * Writeback a page to clean the dirty state
 */
static int writeout(struct address_space *mapping, struct page *page)
676
{
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
	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;

694
	/*
695 696 697 698 699 700
	 * 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.
701
	 */
702
	remove_migration_ptes(page, page);
703

704
	rc = mapping->a_ops->writepage(page, &wbc);
705

706 707 708 709
	if (rc != AOP_WRITEPAGE_ACTIVATE)
		/* unlocked. Relock */
		lock_page(page);

710
	return (rc < 0) ? -EIO : -EAGAIN;
711 712 713 714 715 716
}

/*
 * Default handling if a filesystem does not provide a migration function.
 */
static int fallback_migrate_page(struct address_space *mapping,
717
	struct page *newpage, struct page *page, enum migrate_mode mode)
718
{
719
	if (PageDirty(page)) {
720 721
		/* Only writeback pages in full synchronous migration */
		if (mode != MIGRATE_SYNC)
722
			return -EBUSY;
723
		return writeout(mapping, page);
724
	}
725 726 727 728 729

	/*
	 * Buffers may be managed in a filesystem specific way.
	 * We must have no buffers or drop them.
	 */
730
	if (page_has_private(page) &&
731 732 733
	    !try_to_release_page(page, GFP_KERNEL))
		return -EAGAIN;

734
	return migrate_page(mapping, newpage, page, mode);
735 736
}

737 738 739 740 741 742
/*
 * 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.
743 744 745
 *
 * Return value:
 *   < 0 - error code
746
 *  MIGRATEPAGE_SUCCESS - success
747
 */
748
static int move_to_new_page(struct page *newpage, struct page *page,
749
				int remap_swapcache, enum migrate_mode mode)
750 751 752 753 754 755 756 757 758
{
	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.
	 */
Nick Piggin's avatar
Nick Piggin committed
759
	if (!trylock_page(newpage))
760 761 762 763 764
		BUG();

	/* Prepare mapping for the new page.*/
	newpage->index = page->index;
	newpage->mapping = page->mapping;
765 766
	if (PageSwapBacked(page))
		SetPageSwapBacked(newpage);
767 768 769

	mapping = page_mapping(page);
	if (!mapping)
770
		rc = migrate_page(mapping, newpage, page, mode);
771
	else if (mapping->a_ops->migratepage)
772
		/*
773 774 775 776
		 * Most pages have a mapping and most filesystems provide a
		 * migratepage callback. Anonymous pages are part of swap
		 * space which also has its own migratepage callback. This
		 * is the most common path for page migration.
777
		 */
778
		rc = mapping->a_ops->migratepage(mapping,
779
						newpage, page, mode);
780
	else
781
		rc = fallback_migrate_page(mapping, newpage, page, mode);
782

783
	if (rc != MIGRATEPAGE_SUCCESS) {
784
		newpage->mapping = NULL;
785
	} else {
786
		mem_cgroup_migrate(page, newpage, false);
787 788
		if (remap_swapcache)
			remove_migration_ptes(page, newpage);
789
		page->mapping = NULL;
790
	}
791 792 793 794 795 796

	unlock_page(newpage);

	return rc;
}

797
static int __unmap_and_move(struct page *page, struct page *newpage,
798
				int force, enum migrate_mode mode)
799
{
800
	int rc = -EAGAIN;
801
	int remap_swapcache = 1;
802
	struct anon_vma *anon_vma = NULL;
803

Nick Piggin's avatar
Nick Piggin committed
804
	if (!trylock_page(page)) {
805
		if (!force || mode == MIGRATE_ASYNC)
806
			goto out;
807 808 809 810 811 812 813 814 815 816 817 818 819 820 821

		/*
		 * 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)
822
			goto out;
823

824 825 826 827
		lock_page(page);
	}

	if (PageWriteback(page)) {
828
		/*
829
		 * Only in the case of a full synchronous migration is it
830 831 832
		 * necessary to wait for PageWriteback. In the async case,
		 * the retry loop is too short and in the sync-light case,
		 * the overhead of stalling is too much
833
		 */
834
		if (mode != MIGRATE_SYNC) {
835
			rc = -EBUSY;
836
			goto out_unlock;
837 838
		}
		if (!force)
839
			goto out_unlock;
840 841 842
		wait_on_page_writeback(page);
	}
	/*
843 844
	 * 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.
845
	 * This get_anon_vma() delays freeing anon_vma pointer until the end
846
	 * of migration. File cache pages are no problem because of page_lock()
847 848
	 * File Caches may use write_page() or lock_page() in migration, then,
	 * just care Anon page here.
849
	 */
850
	if (PageAnon(page) && !PageKsm(page)) {
851
		/*
852
		 * Only page_lock_anon_vma_read() understands the subtleties of
853 854
		 * getting a hold on an anon_vma from outside one of its mms.
		 */
855
		anon_vma = page_get_anon_vma(page);
856 857
		if (anon_vma) {
			/*
858
			 * Anon page
859 860
			 */
		} else if (PageSwapCache(page)) {
861 862 863 864 865 866 867 868 869 870 871 872 873 874
			/*
			 * 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 {
875
			goto out_unlock;
876
		}
877
	}
878

879
	if (unlikely(isolated_balloon_page(page))) {
880 881 882 883 884 885 886 887
		/*
		 * A ballooned page does not need any special attention from
		 * physical to virtual reverse mapping procedures.
		 * Skip any attempt to unmap PTEs or to remap swap cache,
		 * in order to avoid burning cycles at rmap level, and perform
		 * the page migration right away (proteced by page lock).
		 */
		rc = balloon_page_migrate(newpage, page, mode);
888
		goto out_unlock;
889 890
	}

891
	/*
892 893 894 895 896 897 898 899 900 901
	 * 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.
902
	 */
903
	if (!page->mapping) {
904
		VM_BUG_ON_PAGE(PageAnon(page), page);
905
		if (page_has_private(page)) {
906
			try_to_free_buffers(page);
907
			goto out_unlock;
908
		}
909
		goto skip_unmap;
910 911
	}

912
	/* Establish migration ptes or remove ptes */
913
	try_to_unmap(page, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);
914

915
skip_unmap:
916
	if (!page_mapped(page))
917
		rc = move_to_new_page(newpage, page, remap_swapcache, mode);
918

919
	if (rc && remap_swapcache)
920
		remove_migration_ptes(page, page);
921 922

	/* Drop an anon_vma reference if we took one */
923
	if (anon_vma)
924
		put_anon_vma(anon_vma);
925

926
out_unlock:
927
	unlock_page(page);
928 929 930
out:
	return rc;
}
931

932 933 934 935
/*
 * Obtain the lock on page, remove all ptes and migrate the page
 * to the newly allocated page in newpage.
 */
936 937 938
static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
			unsigned long private, struct page *page, int force,
			enum migrate_mode mode)
939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
{
	int rc = 0;
	int *result = NULL;
	struct page *newpage = get_new_page(page, private, &result);

	if (!newpage)
		return -ENOMEM;

	if (page_count(page) == 1) {
		/* page was freed from under us. So we are done. */
		goto out;
	}

	if (unlikely(PageTransHuge(page)))
		if (unlikely(split_huge_page(page)))
			goto out;

956
	rc = __unmap_and_move(page, newpage, force, mode);
957

958
out:
959
	if (rc != -EAGAIN) {
960 961 962 963 964 965 966
		/*
		 * 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);
967
		dec_zone_page_state(page, NR_ISOLATED_ANON +
968
				page_is_file_cache(page));
969
		putback_lru_page(page);
970
	}
971

972
	/*
973 974 975
	 * If migration was not successful and there's a freeing callback, use
	 * it.  Otherwise, putback_lru_page() will drop the reference grabbed
	 * during isolation.
976
	 */
977 978
	if (rc != MIGRATEPAGE_SUCCESS && put_new_page) {
		ClearPageSwapBacked(newpage);
979
		put_new_page(newpage, private);
980 981 982
	} else if (unlikely(__is_movable_balloon_page(newpage))) {
		/* drop our reference, page already in the balloon */
		put_page(newpage);
983
	} else
984 985
		putback_lru_page(newpage);

986 987 988 989 990 991
	if (result) {
		if (rc)
			*result = rc;
		else
			*result = page_to_nid(newpage);
	}
992 993 994
	return rc;
}

995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
/*
 * 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,
1014 1015 1016
				free_page_t put_new_page, unsigned long private,
				struct page *hpage, int force,
				enum migrate_mode mode)
1017 1018 1019
{
	int rc = 0;
	int *result = NULL;
1020
	struct page *new_hpage;
1021 1022
	struct anon_vma *anon_vma = NULL;

1023 1024 1025 1026 1027 1028 1029
	/*
	 * Movability of hugepages depends on architectures and hugepage size.
	 * This check is necessary because some callers of hugepage migration
	 * like soft offline and memory hotremove don't walk through page
	 * tables or check whether the hugepage is pmd-based or not before
	 * kicking migration.
	 */
1030
	if (!hugepage_migration_supported(page_hstate(hpage))) {
1031
		putback_active_hugepage(hpage);
1032
		return -ENOSYS;
1033
	}
1034

1035
	new_hpage = get_new_page(hpage, private, &result);
1036 1037 1038 1039 1040 1041
	if (!new_hpage)
		return -ENOMEM;

	rc = -EAGAIN;

	if (!trylock_page(hpage)) {
1042
		if (!force || mode != MIGRATE_SYNC)
1043 1044 1045 1046
			goto out;
		lock_page(hpage);
	}

1047 1048
	if (PageAnon(hpage))
		anon_vma = page_get_anon_vma(hpage);
1049 1050 1051 1052

	try_to_unmap(hpage, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);

	if (!page_mapped(hpage))
1053
		rc = move_to_new_page(new_hpage, hpage, 1, mode);
1054

1055
	if (rc != MIGRATEPAGE_SUCCESS)
1056 1057
		remove_migration_ptes(hpage, hpage);

Hugh Dickins's avatar
Hugh Dickins committed
1058
	if (anon_vma)
1059
		put_anon_vma(anon_vma);
1060

1061
	if (rc == MIGRATEPAGE_SUCCESS)
1062 1063
		hugetlb_cgroup_migrate(hpage, new_hpage);

1064
	unlock_page(hpage);
1065
out:
1066 1067
	if (rc != -EAGAIN)
		putback_active_hugepage(hpage);
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078

	/*
	 * If migration was not successful and there's a freeing callback, use
	 * it.  Otherwise, put_page() will drop the reference grabbed during
	 * isolation.
	 */
	if (rc != MIGRATEPAGE_SUCCESS && put_new_page)
		put_new_page(new_hpage, private);
	else
		put_page(new_hpage);

1079 1080 1081 1082 1083 1084 1085 1086 1087
	if (result) {
		if (rc)
			*result = rc;
		else
			*result = page_to_nid(new_hpage);
	}
	return rc;
}

1088
/*
1089 1090
 * migrate_pages - migrate the pages specified in a list, to the free pages
 *		   supplied as the target for the page migration
1091
 *
1092 1093 1094
 * @from:		The list of pages to be migrated.
 * @get_new_page:	The function used to allocate free pages to be used
 *			as the target of the page migration.
1095 1096
 * @put_new_page:	The function used to free target pages if migration
 *			fails, or NULL if no special handling is necessary.
1097 1098 1099 1100
 * @private:		Private data to be passed on to get_new_page()
 * @mode:		The migration mode that specifies the constraints for
 *			page migration, if any.
 * @reason:		The reason for page migration.
1101
 *
1102 1103 1104
 * The function returns after 10 attempts or if no pages are movable any more
 * because the list has become empty or no retryable pages exist any more.
 * The caller should call putback_lru_pages() to return pages to the LRU
1105
 * or free list only if ret != 0.
1106
 *
1107
 * Returns the number of pages that were not migrated, or an error code.
1108
 */
1109
int migrate_pages(struct list_head *from, new_page_t get_new_page,
1110 1111
		free_page_t put_new_page, unsigned long private,
		enum migrate_mode mode, int reason)
1112
{
1113
	int retry = 1;
1114
	int nr_failed = 0;
1115
	int nr_succeeded = 0;
1116 1117 1118 1119 1120 1121 1122 1123 1124
	int pass = 0;
	struct page *page;
	struct page *page2;
	int swapwrite = current->flags & PF_SWAPWRITE;
	int rc;

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

1125 1126
	for(pass = 0; pass < 10 && retry; pass++) {
		retry = 0;
1127

1128 1129
		list_for_each_entry_safe(page, page2, from, lru) {
			cond_resched();
1130

1131 1132
			if (PageHuge(page))
				rc = unmap_and_move_huge_page(get_new_page,
1133 1134
						put_new_page, private, page,
						pass > 2, mode);
1135
			else
1136 1137
				rc = unmap_and_move(get_new_page, put_new_page,
						private, page, pass > 2, mode);
1138

1139
			switch(rc) {
1140 1141
			case -ENOMEM:
				goto out;
1142
			case -EAGAIN:
1143
				retry++;
1144
				break;
1145
			case MIGRATEPAGE_SUCCESS:
1146
				nr_succeeded++;
1147 1148
				break;
			default:
1149 1150 1151 1152 1153 1154
				/*
				 * Permanent failure (-EBUSY, -ENOSYS, etc.):
				 * unlike -EAGAIN case, the failed page is
				 * removed from migration page list and not
				 * retried in the next outer loop.
				 */
1155
				nr_failed++;
1156
				break;
1157
			}
1158 1159
		}
	}
1160
	rc = nr_failed + retry;
1161
out:
1162 1163 1164 1165
	if (nr_succeeded)
		count_vm_events(PGMIGRATE_SUCCESS, nr_succeeded);
	if (nr_failed)
		count_vm_events(PGMIGRATE_FAIL, nr_failed);
1166 1167
	trace_mm_migrate_pages(nr_succeeded, nr_failed, mode, reason);

1168 1169 1170
	if (!swapwrite)
		current->flags &= ~PF_SWAPWRITE;

1171
	return rc;
1172
}
1173

1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
#ifdef CONFIG_NUMA
/*
 * Move a list of individual pages
 */
struct page_to_node {
	unsigned long addr