migrate.c 85.4 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
/*
3
 * Memory Migration functionality - linux/mm/migrate.c
4 5 6 7 8 9 10 11 12
 *
 * 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
13
 * Christoph Lameter
14 15 16
 */

#include <linux/migrate.h>
17
#include <linux/export.h>
18
#include <linux/swap.h>
19
#include <linux/swapops.h>
20
#include <linux/pagemap.h>
21
#include <linux/buffer_head.h>
22
#include <linux/mm_inline.h>
23
#include <linux/nsproxy.h>
24
#include <linux/pagevec.h>
25
#include <linux/ksm.h>
26 27 28 29
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
30
#include <linux/writeback.h>
31 32
#include <linux/mempolicy.h>
#include <linux/vmalloc.h>
33
#include <linux/security.h>
34
#include <linux/backing-dev.h>
35
#include <linux/compaction.h>
36
#include <linux/syscalls.h>
37
#include <linux/compat.h>
38
#include <linux/hugetlb.h>
39
#include <linux/hugetlb_cgroup.h>
40
#include <linux/gfp.h>
41
#include <linux/pagewalk.h>
42
#include <linux/pfn_t.h>
43
#include <linux/memremap.h>
44
#include <linux/userfaultfd_k.h>
45
#include <linux/balloon_compaction.h>
46
#include <linux/mmu_notifier.h>
47
#include <linux/page_idle.h>
48
#include <linux/page_owner.h>
49
#include <linux/sched/mm.h>
50
#include <linux/ptrace.h>
51
#include <linux/oom.h>
52
#include <linux/memory.h>
53

54 55
#include <asm/tlbflush.h>

56 57 58
#define CREATE_TRACE_POINTS
#include <trace/events/migrate.h>

59 60
#include "internal.h"

61
int isolate_movable_page(struct page *page, isolate_mode_t mode)
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
{
	struct address_space *mapping;

	/*
	 * Avoid burning cycles with pages that are yet under __free_pages(),
	 * or just got freed under us.
	 *
	 * In case we 'win' a race for a movable page being freed under us and
	 * raise its refcount preventing __free_pages() from doing its job
	 * the put_page() at the end of this block will take care of
	 * release this page, thus avoiding a nasty leakage.
	 */
	if (unlikely(!get_page_unless_zero(page)))
		goto out;

	/*
	 * Check PageMovable before holding a PG_lock because page's owner
	 * assumes anybody doesn't touch PG_lock of newly allocated page
80
	 * so unconditionally grabbing the lock ruins page's owner side.
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111
	 */
	if (unlikely(!__PageMovable(page)))
		goto out_putpage;
	/*
	 * As movable pages are not isolated from LRU lists, concurrent
	 * compaction threads can race against page migration functions
	 * as well as race against the releasing a page.
	 *
	 * In order to avoid having an already isolated movable page
	 * being (wrongly) re-isolated while it is under migration,
	 * or to avoid attempting to isolate pages being released,
	 * lets be sure we have the page lock
	 * before proceeding with the movable page isolation steps.
	 */
	if (unlikely(!trylock_page(page)))
		goto out_putpage;

	if (!PageMovable(page) || PageIsolated(page))
		goto out_no_isolated;

	mapping = page_mapping(page);
	VM_BUG_ON_PAGE(!mapping, page);

	if (!mapping->a_ops->isolate_page(page, mode))
		goto out_no_isolated;

	/* Driver shouldn't use PG_isolated bit of page->flags */
	WARN_ON_ONCE(PageIsolated(page));
	__SetPageIsolated(page);
	unlock_page(page);

112
	return 0;
113 114 115 116 117 118

out_no_isolated:
	unlock_page(page);
out_putpage:
	put_page(page);
out:
119
	return -EBUSY;
120 121
}

122
static void putback_movable_page(struct page *page)
123 124 125 126 127 128 129 130
{
	struct address_space *mapping;

	mapping = page_mapping(page);
	mapping->a_ops->putback_page(page);
	__ClearPageIsolated(page);
}

131 132 133 134
/*
 * Put previously isolated pages back onto the appropriate lists
 * from where they were once taken off for compaction/migration.
 *
135 136 137
 * 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().
138 139 140 141 142 143
 */
void putback_movable_pages(struct list_head *l)
{
	struct page *page;
	struct page *page2;

144
	list_for_each_entry_safe(page, page2, l, lru) {
145 146 147 148
		if (unlikely(PageHuge(page))) {
			putback_active_hugepage(page);
			continue;
		}
149
		list_del(&page->lru);
150 151 152 153 154
		/*
		 * We isolated non-lru movable page so here we can use
		 * __PageMovable because LRU page's mapping cannot have
		 * PAGE_MAPPING_MOVABLE.
		 */
155
		if (unlikely(__PageMovable(page))) {
156 157 158 159 160 161 162 163 164
			VM_BUG_ON_PAGE(!PageIsolated(page), page);
			lock_page(page);
			if (PageMovable(page))
				putback_movable_page(page);
			else
				__ClearPageIsolated(page);
			unlock_page(page);
			put_page(page);
		} else {
165
			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
166
					page_is_file_lru(page), -thp_nr_pages(page));
167
			putback_lru_page(page);
168
		}
169 170 171
	}
}

172 173 174
/*
 * Restore a potential migration pte to a working pte entry
 */
175
static bool remove_migration_pte(struct page *page, struct vm_area_struct *vma,
176
				 unsigned long addr, void *old)
177
{
178 179 180 181 182 183 184 185
	struct page_vma_mapped_walk pvmw = {
		.page = old,
		.vma = vma,
		.address = addr,
		.flags = PVMW_SYNC | PVMW_MIGRATION,
	};
	struct page *new;
	pte_t pte;
186 187
	swp_entry_t entry;

188 189
	VM_BUG_ON_PAGE(PageTail(page), page);
	while (page_vma_mapped_walk(&pvmw)) {
190 191 192 193 194
		if (PageKsm(page))
			new = page;
		else
			new = page - pvmw.page->index +
				linear_page_index(vma, pvmw.address);
195

196 197 198 199 200 201 202 203 204
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
		/* PMD-mapped THP migration entry */
		if (!pvmw.pte) {
			VM_BUG_ON_PAGE(PageHuge(page) || !PageTransCompound(page), page);
			remove_migration_pmd(&pvmw, new);
			continue;
		}
#endif

205 206 207 208
		get_page(new);
		pte = pte_mkold(mk_pte(new, READ_ONCE(vma->vm_page_prot)));
		if (pte_swp_soft_dirty(*pvmw.pte))
			pte = pte_mksoft_dirty(pte);
209

210 211 212 213
		/*
		 * Recheck VMA as permissions can change since migration started
		 */
		entry = pte_to_swp_entry(*pvmw.pte);
214
		if (is_writable_migration_entry(entry))
215
			pte = maybe_mkwrite(pte, vma);
216 217
		else if (pte_swp_uffd_wp(*pvmw.pte))
			pte = pte_mkuffd_wp(pte);
218

219
		if (unlikely(is_device_private_page(new))) {
220 221 222 223 224 225
			if (pte_write(pte))
				entry = make_writable_device_private_entry(
							page_to_pfn(new));
			else
				entry = make_readable_device_private_entry(
							page_to_pfn(new));
226
			pte = swp_entry_to_pte(entry);
227 228
			if (pte_swp_soft_dirty(*pvmw.pte))
				pte = pte_swp_mksoft_dirty(pte);
229 230
			if (pte_swp_uffd_wp(*pvmw.pte))
				pte = pte_swp_mkuffd_wp(pte);
231
		}
232

233
#ifdef CONFIG_HUGETLB_PAGE
234
		if (PageHuge(new)) {
235 236
			unsigned int shift = huge_page_shift(hstate_vma(vma));

237
			pte = pte_mkhuge(pte);
238
			pte = arch_make_huge_pte(pte, shift, vma->vm_flags);
239
			set_huge_pte_at(vma->vm_mm, pvmw.address, pvmw.pte, pte);
240 241 242 243
			if (PageAnon(new))
				hugepage_add_anon_rmap(new, vma, pvmw.address);
			else
				page_dup_rmap(new, true);
244 245 246 247
		} else
#endif
		{
			set_pte_at(vma->vm_mm, pvmw.address, pvmw.pte, pte);
248

249 250 251 252 253
			if (PageAnon(new))
				page_add_anon_rmap(new, vma, pvmw.address, false);
			else
				page_add_file_rmap(new, false);
		}
254 255 256
		if (vma->vm_flags & VM_LOCKED && !PageTransCompound(new))
			mlock_vma_page(new);

257 258 259
		if (PageTransHuge(page) && PageMlocked(page))
			clear_page_mlock(page);

260 261 262
		/* No need to invalidate - it was non-present before */
		update_mmu_cache(vma, pvmw.address, pvmw.pte);
	}
263

264
	return true;
265 266
}

267 268 269 270
/*
 * Get rid of all migration entries and replace them by
 * references to the indicated page.
 */
271
void remove_migration_ptes(struct page *old, struct page *new, bool locked)
272
{
273 274 275 276 277
	struct rmap_walk_control rwc = {
		.rmap_one = remove_migration_pte,
		.arg = old,
	};

278 279 280 281
	if (locked)
		rmap_walk_locked(new, &rwc);
	else
		rmap_walk(new, &rwc);
282 283
}

284 285 286 287 288
/*
 * 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.
 */
289
void __migration_entry_wait(struct mm_struct *mm, pte_t *ptep,
290
				spinlock_t *ptl)
291
{
292
	pte_t pte;
293 294 295
	swp_entry_t entry;
	struct page *page;

296
	spin_lock(ptl);
297 298 299 300 301 302 303 304
	pte = *ptep;
	if (!is_swap_pte(pte))
		goto out;

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

305
	page = pfn_swap_entry_to_page(entry);
306
	page = compound_head(page);
307

308
	/*
309
	 * Once page cache replacement of page migration started, page_count
310 311
	 * is zero; but we must not call put_and_wait_on_page_locked() without
	 * a ref. Use get_page_unless_zero(), and just fault again if it fails.
312 313 314
	 */
	if (!get_page_unless_zero(page))
		goto out;
315
	pte_unmap_unlock(ptep, ptl);
316
	put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
317 318 319 320 321
	return;
out:
	pte_unmap_unlock(ptep, ptl);
}

322 323 324 325 326 327 328 329
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);
}

330 331
void migration_entry_wait_huge(struct vm_area_struct *vma,
		struct mm_struct *mm, pte_t *pte)
332
{
333
	spinlock_t *ptl = huge_pte_lockptr(hstate_vma(vma), mm, pte);
334 335 336
	__migration_entry_wait(mm, pte, ptl);
}

337 338 339 340 341 342 343 344 345
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
void pmd_migration_entry_wait(struct mm_struct *mm, pmd_t *pmd)
{
	spinlock_t *ptl;
	struct page *page;

	ptl = pmd_lock(mm, pmd);
	if (!is_pmd_migration_entry(*pmd))
		goto unlock;
346
	page = pfn_swap_entry_to_page(pmd_to_swp_entry(*pmd));
347 348 349
	if (!get_page_unless_zero(page))
		goto unlock;
	spin_unlock(ptl);
350
	put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
351 352 353 354 355 356
	return;
unlock:
	spin_unlock(ptl);
}
#endif

357
static int expected_page_refs(struct address_space *mapping, struct page *page)
358 359 360 361
{
	int expected_count = 1;

	/*
362
	 * Device private pages have an extra refcount as they are
363 364 365
	 * ZONE_DEVICE pages.
	 */
	expected_count += is_device_private_page(page);
366
	if (mapping)
367
		expected_count += compound_nr(page) + page_has_private(page);
368 369 370 371

	return expected_count;
}

372
/*
373
 * Replace the page in the mapping.
374 375 376 377
 *
 * The number of remaining references must be:
 * 1 for anonymous pages without a mapping
 * 2 for pages with a mapping
378
 * 3 for pages with a mapping and PagePrivate/PagePrivate2 set.
379
 */
380 381
int folio_migrate_mapping(struct address_space *mapping,
		struct folio *newfolio, struct folio *folio, int extra_count)
382
{
383
	XA_STATE(xas, &mapping->i_pages, folio_index(folio));
384 385
	struct zone *oldzone, *newzone;
	int dirty;
386 387
	int expected_count = expected_page_refs(mapping, &folio->page) + extra_count;
	long nr = folio_nr_pages(folio);
388

389
	if (!mapping) {
390
		/* Anonymous page without mapping */
391
		if (folio_ref_count(folio) != expected_count)
392
			return -EAGAIN;
393 394

		/* No turning back from here */
395 396 397 398
		newfolio->index = folio->index;
		newfolio->mapping = folio->mapping;
		if (folio_test_swapbacked(folio))
			__folio_set_swapbacked(newfolio);
399

400
		return MIGRATEPAGE_SUCCESS;
401 402
	}

403 404
	oldzone = folio_zone(folio);
	newzone = folio_zone(newfolio);
405

406
	xas_lock_irq(&xas);
407
	if (!folio_ref_freeze(folio, expected_count)) {
408
		xas_unlock_irq(&xas);
409 410 411
		return -EAGAIN;
	}

412
	/*
413
	 * Now we know that no one else is looking at the folio:
414
	 * no turning back from here.
415
	 */
416 417 418 419 420 421 422 423
	newfolio->index = folio->index;
	newfolio->mapping = folio->mapping;
	folio_ref_add(newfolio, nr); /* add cache reference */
	if (folio_test_swapbacked(folio)) {
		__folio_set_swapbacked(newfolio);
		if (folio_test_swapcache(folio)) {
			folio_set_swapcache(newfolio);
			newfolio->private = folio_get_private(folio);
424 425
		}
	} else {
426
		VM_BUG_ON_FOLIO(folio_test_swapcache(folio), folio);
427 428
	}

429
	/* Move dirty while page refs frozen and newpage not yet exposed */
430
	dirty = folio_test_dirty(folio);
431
	if (dirty) {
432 433
		folio_clear_dirty(folio);
		folio_set_dirty(newfolio);
434 435
	}

436 437
	xas_store(&xas, newfolio);
	if (nr > 1) {
438 439
		int i;

440
		for (i = 1; i < nr; i++) {
441
			xas_next(&xas);
442
			xas_store(&xas, newfolio);
443 444
		}
	}
445 446

	/*
447 448
	 * Drop cache reference from old page by unfreezing
	 * to one less reference.
449 450
	 * We know this isn't the last reference.
	 */
451
	folio_ref_unfreeze(folio, expected_count - nr);
452

453
	xas_unlock(&xas);
454 455
	/* Leave irq disabled to prevent preemption while updating stats */

456 457 458 459 460 461 462
	/*
	 * 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
463
	 * via NR_FILE_PAGES and NR_ANON_MAPPED if they
464 465
	 * are mapped to swap space.
	 */
466
	if (newzone != oldzone) {
467 468 469
		struct lruvec *old_lruvec, *new_lruvec;
		struct mem_cgroup *memcg;

470
		memcg = folio_memcg(folio);
471 472 473
		old_lruvec = mem_cgroup_lruvec(memcg, oldzone->zone_pgdat);
		new_lruvec = mem_cgroup_lruvec(memcg, newzone->zone_pgdat);

474 475
		__mod_lruvec_state(old_lruvec, NR_FILE_PAGES, -nr);
		__mod_lruvec_state(new_lruvec, NR_FILE_PAGES, nr);
476
		if (folio_test_swapbacked(folio) && !folio_test_swapcache(folio)) {
477 478
			__mod_lruvec_state(old_lruvec, NR_SHMEM, -nr);
			__mod_lruvec_state(new_lruvec, NR_SHMEM, nr);
479
		}
480
#ifdef CONFIG_SWAP
481
		if (folio_test_swapcache(folio)) {
482 483 484 485
			__mod_lruvec_state(old_lruvec, NR_SWAPCACHE, -nr);
			__mod_lruvec_state(new_lruvec, NR_SWAPCACHE, nr);
		}
#endif
486
		if (dirty && mapping_can_writeback(mapping)) {
487 488 489 490
			__mod_lruvec_state(old_lruvec, NR_FILE_DIRTY, -nr);
			__mod_zone_page_state(oldzone, NR_ZONE_WRITE_PENDING, -nr);
			__mod_lruvec_state(new_lruvec, NR_FILE_DIRTY, nr);
			__mod_zone_page_state(newzone, NR_ZONE_WRITE_PENDING, nr);
491
		}
492
	}
493
	local_irq_enable();
494

495
	return MIGRATEPAGE_SUCCESS;
496
}
497
EXPORT_SYMBOL(folio_migrate_mapping);
498

499 500
/*
 * The expected number of remaining references is the same as that
501
 * of folio_migrate_mapping().
502 503 504 505
 */
int migrate_huge_page_move_mapping(struct address_space *mapping,
				   struct page *newpage, struct page *page)
{
506
	XA_STATE(xas, &mapping->i_pages, page_index(page));
507 508
	int expected_count;

509
	xas_lock_irq(&xas);
510
	expected_count = 2 + page_has_private(page);
511 512
	if (page_count(page) != expected_count || xas_load(&xas) != page) {
		xas_unlock_irq(&xas);
513 514 515
		return -EAGAIN;
	}

516
	if (!page_ref_freeze(page, expected_count)) {
517
		xas_unlock_irq(&xas);
518 519 520
		return -EAGAIN;
	}

521 522
	newpage->index = page->index;
	newpage->mapping = page->mapping;
523

524 525
	get_page(newpage);

526
	xas_store(&xas, newpage);
527

528
	page_ref_unfreeze(page, expected_count - 1);
529

530
	xas_unlock_irq(&xas);
531

532
	return MIGRATEPAGE_SUCCESS;
533 534
}

535
/*
536
 * Copy the flags and some other ancillary information
537
 */
538
void folio_migrate_flags(struct folio *newfolio, struct folio *folio)
539
{
540 541
	int cpupid;

542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558
	if (folio_test_error(folio))
		folio_set_error(newfolio);
	if (folio_test_referenced(folio))
		folio_set_referenced(newfolio);
	if (folio_test_uptodate(folio))
		folio_mark_uptodate(newfolio);
	if (folio_test_clear_active(folio)) {
		VM_BUG_ON_FOLIO(folio_test_unevictable(folio), folio);
		folio_set_active(newfolio);
	} else if (folio_test_clear_unevictable(folio))
		folio_set_unevictable(newfolio);
	if (folio_test_workingset(folio))
		folio_set_workingset(newfolio);
	if (folio_test_checked(folio))
		folio_set_checked(newfolio);
	if (folio_test_mappedtodisk(folio))
		folio_set_mappedtodisk(newfolio);
559

560
	/* Move dirty on pages not done by folio_migrate_mapping() */
561 562
	if (folio_test_dirty(folio))
		folio_set_dirty(newfolio);
563

564 565 566 567
	if (folio_test_young(folio))
		folio_set_young(newfolio);
	if (folio_test_idle(folio))
		folio_set_idle(newfolio);
568

569 570 571 572
	/*
	 * Copy NUMA information to the new page, to prevent over-eager
	 * future migrations of this same page.
	 */
573 574
	cpupid = page_cpupid_xchg_last(&folio->page, -1);
	page_cpupid_xchg_last(&newfolio->page, cpupid);
575

576
	folio_migrate_ksm(newfolio, folio);
577 578 579 580
	/*
	 * Please do not reorder this without considering how mm/ksm.c's
	 * get_ksm_page() depends upon ksm_migrate_page() and PageSwapCache().
	 */
581 582 583
	if (folio_test_swapcache(folio))
		folio_clear_swapcache(folio);
	folio_clear_private(folio);
584 585

	/* page->private contains hugetlb specific flags */
586 587
	if (!folio_test_hugetlb(folio))
		folio->private = NULL;
588 589 590 591 592

	/*
	 * If any waiters have accumulated on the new page then
	 * wake them up.
	 */
593 594
	if (folio_test_writeback(newfolio))
		folio_end_writeback(newfolio);
595

596 597 598 599 600
	/*
	 * PG_readahead shares the same bit with PG_reclaim.  The above
	 * end_page_writeback() may clear PG_readahead mistakenly, so set the
	 * bit after that.
	 */
601 602
	if (folio_test_readahead(folio))
		folio_set_readahead(newfolio);
603

604
	folio_copy_owner(newfolio, folio);
605

606
	if (!folio_test_hugetlb(folio))
607
		mem_cgroup_migrate(folio, newfolio);
608
}
609
EXPORT_SYMBOL(folio_migrate_flags);
610

611
void folio_migrate_copy(struct folio *newfolio, struct folio *folio)
612
{
613 614
	folio_copy(newfolio, folio);
	folio_migrate_flags(newfolio, folio);
615
}
616
EXPORT_SYMBOL(folio_migrate_copy);
617

618 619 620 621
/************************************************************
 *                    Migration functions
 ***********************************************************/

622
/*
623
 * Common logic to directly migrate a single LRU page suitable for
624
 * pages that do not use PagePrivate/PagePrivate2.
625 626 627
 *
 * Pages are locked upon entry and exit.
 */
628
int migrate_page(struct address_space *mapping,
629 630
		struct page *newpage, struct page *page,
		enum migrate_mode mode)
631
{
632 633
	struct folio *newfolio = page_folio(newpage);
	struct folio *folio = page_folio(page);
634 635
	int rc;

636
	BUG_ON(folio_test_writeback(folio));	/* Writeback must be complete */
637

638
	rc = folio_migrate_mapping(mapping, newfolio, folio, 0);
639

640
	if (rc != MIGRATEPAGE_SUCCESS)
641 642
		return rc;

643
	if (mode != MIGRATE_SYNC_NO_COPY)
644
		folio_migrate_copy(newfolio, folio);
645
	else
646
		folio_migrate_flags(newfolio, folio);
647
	return MIGRATEPAGE_SUCCESS;
648 649 650
}
EXPORT_SYMBOL(migrate_page);

651
#ifdef CONFIG_BLOCK
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689
/* Returns true if all buffers are successfully locked */
static bool buffer_migrate_lock_buffers(struct buffer_head *head,
							enum migrate_mode mode)
{
	struct buffer_head *bh = head;

	/* Simple case, sync compaction */
	if (mode != MIGRATE_ASYNC) {
		do {
			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 {
		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;
			bh = head;
			while (bh != failed_bh) {
				unlock_buffer(bh);
				bh = bh->b_this_page;
			}
			return false;
		}

		bh = bh->b_this_page;
	} while (bh != head);
	return true;
}

690 691 692
static int __buffer_migrate_page(struct address_space *mapping,
		struct page *newpage, struct page *page, enum migrate_mode mode,
		bool check_refs)
693 694 695
{
	struct buffer_head *bh, *head;
	int rc;
696
	int expected_count;
697 698

	if (!page_has_buffers(page))
699
		return migrate_page(mapping, newpage, page, mode);
700

701
	/* Check whether page does not have extra refs before we do more work */
702
	expected_count = expected_page_refs(mapping, page);
703 704
	if (page_count(page) != expected_count)
		return -EAGAIN;
705

706 707 708
	head = page_buffers(page);
	if (!buffer_migrate_lock_buffers(head, mode))
		return -EAGAIN;
709

710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729
	if (check_refs) {
		bool busy;
		bool invalidated = false;

recheck_buffers:
		busy = false;
		spin_lock(&mapping->private_lock);
		bh = head;
		do {
			if (atomic_read(&bh->b_count)) {
				busy = true;
				break;
			}
			bh = bh->b_this_page;
		} while (bh != head);
		if (busy) {
			if (invalidated) {
				rc = -EAGAIN;
				goto unlock_buffers;
			}
730
			spin_unlock(&mapping->private_lock);
731 732 733 734 735 736
			invalidate_bh_lrus();
			invalidated = true;
			goto recheck_buffers;
		}
	}

737
	rc = migrate_page_move_mapping(mapping, newpage, page, 0);
738
	if (rc != MIGRATEPAGE_SUCCESS)
739
		goto unlock_buffers;
740

741
	attach_page_private(newpage, detach_page_private(page));
742 743 744 745 746 747 748 749

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

	} while (bh != head);

750 751 752 753
	if (mode != MIGRATE_SYNC_NO_COPY)
		migrate_page_copy(newpage, page);
	else
		migrate_page_states(newpage, page);
754

755 756
	rc = MIGRATEPAGE_SUCCESS;
unlock_buffers:
757 758
	if (check_refs)
		spin_unlock(&mapping->private_lock);
759 760 761 762 763 764 765
	bh = head;
	do {
		unlock_buffer(bh);
		bh = bh->b_this_page;

	} while (bh != head);

766
	return rc;
767
}
768 769 770 771 772 773 774 775 776 777 778

/*
 * Migration function for pages with buffers. This function can only be used
 * if the underlying filesystem guarantees that no other references to "page"
 * exist. For example attached buffer heads are accessed only under page lock.
 */
int buffer_migrate_page(struct address_space *mapping,
		struct page *newpage, struct page *page, enum migrate_mode mode)
{
	return __buffer_migrate_page(mapping, newpage, page, mode, false);
}
779
EXPORT_SYMBOL(buffer_migrate_page);
780 781 782 783 784 785 786 787 788 789 790 791

/*
 * Same as above except that this variant is more careful and checks that there
 * are also no buffer head references. This function is the right one for
 * mappings where buffer heads are directly looked up and referenced (such as
 * block device mappings).
 */
int buffer_migrate_page_norefs(struct address_space *mapping,
		struct page *newpage, struct page *page, enum migrate_mode mode)
{
	return __buffer_migrate_page(mapping, newpage, page, mode, true);
}
792
#endif
793

794 795 796 797
/*
 * Writeback a page to clean the dirty state
 */
static int writeout(struct address_space *mapping, struct page *page)
798
{
799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
	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;

816
	/*
817 818 819 820 821 822
	 * 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.
823
	 */
824
	remove_migration_ptes(page, page, false);
825

826
	rc = mapping->a_ops->writepage(page, &wbc);
827

828 829 830 831
	if (rc != AOP_WRITEPAGE_ACTIVATE)
		/* unlocked. Relock */
		lock_page(page);

832
	return (rc < 0) ? -EIO : -EAGAIN;
833 834 835 836 837 838
}

/*
 * Default handling if a filesystem does not provide a migration function.
 */
static int fallback_migrate_page(struct address_space *mapping,
839
	struct page *newpage, struct page *page, enum migrate_mode mode)
840
{
841
	if (PageDirty(page)) {
842
		/* Only writeback pages in full synchronous migration */
843 844 845 846 847
		switch (mode) {
		case MIGRATE_SYNC:
		case MIGRATE_SYNC_NO_COPY:
			break;
		default:
848
			return -EBUSY;
849
		}
850
		return writeout(mapping, page);
851
	}
852 853 854 855 856

	/*
	 * Buffers may be managed in a filesystem specific way.
	 * We must have no buffers or drop them.
	 */
857
	if (page_has_private(page) &&
858
	    !try_to_release_page(page, GFP_KERNEL))
859
		return mode == MIGRATE_SYNC ? -EAGAIN : -EBUSY;
860

861
	return migrate_page(mapping, newpage, page, mode);
862 863
}

864 865 866 867 868 869
/*
 * 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.
870 871 872
 *
 * Return value:
 *   < 0 - error code
873
 *  MIGRATEPAGE_SUCCESS - success
874
 */
875
static int move_to_new_page(struct page *newpage, struct page *page,
876
				enum migrate_mode mode)
877 878
{
	struct address_space *mapping;
879 880
	int rc = -EAGAIN;
	bool is_lru = !__PageMovable(page);
881

882 883
	VM_BUG_ON_PAGE(!PageLocked(page), page);
	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
884 885

	mapping = page_mapping(page);
886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903

	if (likely(is_lru)) {
		if (!mapping)
			rc = migrate_page(mapping, newpage, page, mode);
		else if (mapping->a_ops->migratepage)
			/*
			 * 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.
			 */
			rc = mapping->a_ops->migratepage(mapping, newpage,
							page, mode);
		else
			rc = fallback_migrate_page(mapping, newpage,
							page, mode);
	} else {
904
		/*
905 906
		 * In case of non-lru page, it could be released after
		 * isolation step. In that case, we shouldn't try migration.
907
		 */
908 909 910 911 912 913 914 915 916 917 918 919
		VM_BUG_ON_PAGE(!PageIsolated(page), page);
		if (!PageMovable(page)) {
			rc = MIGRATEPAGE_SUCCESS;
			__ClearPageIsolated(page);
			goto out;
		}

		rc = mapping->a_ops->migratepage(mapping, newpage,
						page, mode);
		WARN_ON_ONCE(rc == MIGRATEPAGE_SUCCESS &&
			!PageIsolated(page));
	}
920

921 922 923 924 925
	/*
	 * When successful, old pagecache page->mapping must be cleared before
	 * page is freed; but stats require that PageAnon be left as PageAnon.
	 */
	if (rc == MIGRATEPAGE_SUCCESS) {
926 927 928 929 930 931 932 933 934 935 936
		if (__PageMovable(page)) {
			VM_BUG_ON_PAGE(!PageIsolated(page), page);

			/*
			 * We clear PG_movable under page_lock so any compactor
			 * cannot try to migrate this page.
			 */
			__ClearPageIsolated(page);
		}

		/*
937
		 * Anonymous and movable page->mapping will be cleared by
938 939 940 941
		 * free_pages_prepare so don't reset it here for keeping
		 * the type to work PageAnon, for example.
		 */
		if (!PageMappingFlags(page))
942
			page->mapping = NULL;
943

944
		if (likely(!is_zone_device_page(newpage)))
945 946
			flush_dcache_page(newpage);

947
	}
948
out:
949 950 951
	return rc;
}

952
static int __unmap_and_move(struct page *page, struct page *newpage,
953
				int force, enum migrate_mode mode)
954
{
955
	int rc = -EAGAIN;
956
	bool page_was_mapped = false;
957
	struct anon_vma *anon_vma = NULL;
958
	bool is_lru = !__PageMovable(page);
959

N
Nick Piggin 已提交
960
	if (!trylock_page(page)) {
961
		if (!force || mode == MIGRATE_ASYNC)
962
			goto out;
963 964 965 966 967 968 969

		/*
		 * 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.
970
		 * mpage_readahead). If an allocation happens for the
971 972 973 974 975 976 977
		 * 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)
978
			goto out;
979

980 981 982 983
		lock_page(page);
	}

	if (PageWriteback(page)) {
984
		/*
985
		 * Only in the case of a full synchronous migration is it
986 987 988
		 * 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
989
		 */
990 991 992 993 994
		switch (mode) {
		case MIGRATE_SYNC:
		case MIGRATE_SYNC_NO_COPY:
			break;
		default:
995
			rc = -EBUSY;
996
			goto out_unlock;
997 998
		}
		if (!force)
999
			goto out_unlock;
1000 1001
		wait_on_page_writeback(page);
	}
1002

1003
	/*
1004
	 * By try_to_migrate(), page->mapcount goes down to 0 here. In this case,
1005
	 * we cannot notice that anon_vma is freed while we migrates a page.
1006
	 * This get_anon_vma() delays freeing anon_vma pointer until the end
1007
	 * of migration. File cache pages are no problem because of page_lock()
1008 1009
	 * File Caches may use write_page() or lock_page() in migration, then,
	 * just care Anon page here.
1010 1011 1012 1013 1014 1015
	 *
	 * Only page_get_anon_vma() understands the subtleties of
	 * getting a hold on an anon_vma from outside one of its mms.
	 * But if we cannot get anon_vma, then we won't need it anyway,
	 * because that implies that the anon page is no longer mapped
	 * (and cannot be remapped so long as we hold the page lock).
1016
	 */
1017
	if (PageAnon(page) && !PageKsm(page))
1018
		anon_vma = page_get_anon_vma(page);
1019

1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
	/*
	 * Block others from accessing the new page when we get around to
	 * establishing additional references. We are usually the only one
	 * holding a reference to newpage at this point. We used to have a BUG
	 * here if trylock_page(newpage) fails, but would like to allow for
	 * cases where there might be a race with the previous use of newpage.
	 * This is much like races on refcount of oldpage: just don't BUG().
	 */
	if (unlikely(!trylock_page(newpage)))
		goto out_unlock;

1031 1032 1033 1034 1035
	if (unlikely(!is_lru)) {
		rc = move_to_new_page(newpage, page, mode);
		goto out_unlock_both;
	}

1036
	/*
1037 1038 1039 1040 1041
	 * 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.
1042
	 * 2. An orphaned page (see truncate_cleanup_page) might have
1043 1044 1045 1046
	 * 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.
1047
	 */
1048
	if (!page->mapping) {
1049
		VM_BUG_ON_PAGE(PageAnon(page), page);
1050
		if (page_has_private(page)) {
1051
			try_to_free_buffers(page);
1052
			goto out_unlock_both;
1053
		}
1054 1055
	} else if (page_mapped(page)) {
		/* Establish migration ptes */
1056 1057
		VM_BUG_ON_PAGE(PageAnon(page) && !PageKsm(page) && !anon_vma,
				page);
1058
		try_to_migrate(page, 0);
1059
		page_was_mapped = true;
1060
	}
1061

1062
	if (!page_mapped(page))
1063
		rc = move_to_new_page(newpage, page, mode);
1064

1065 1066
	if (page_was_mapped)
		remove_migration_ptes(page,
1067
			rc == MIGRATEPAGE_SUCCESS ? newpage : page, false);
1068

1069 1070 1071
out_unlock_both:
	unlock_page(newpage);
out_unlock:
1072
	/* Drop an anon_vma reference if we took one */
1073
	if (anon_vma)
1074
		put_anon_vma(anon_vma);
1075
	unlock_page(page);
1076
out:
1077 1078 1079 1080
	/*
	 * If migration is successful, decrease refcount of the newpage
	 * which will not free the page because new page owner increased
	 * refcounter. As well, if it is LRU page, add the page to LRU
1081 1082 1083 1084
	 * list in here. Use the old state of the isolated source page to
	 * determine if we migrated a LRU page. newpage was already unlocked
	 * and possibly modified by its owner - don't rely on the page
	 * state.
1085 1086
	 */
	if (rc == MIGRATEPAGE_SUCCESS) {
1087
		if (unlikely(!is_lru))
1088 1089 1090 1091 1092
			put_page(newpage);
		else
			putback_lru_page(newpage);
	}

1093 1094
	return rc;
}
1095

1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144

/*
 * node_demotion[] example:
 *
 * Consider a system with two sockets.  Each socket has
 * three classes of memory attached: fast, medium and slow.
 * Each memory class is placed in its own NUMA node.  The
 * CPUs are placed in the node with the "fast" memory.  The
 * 6 NUMA nodes (0-5) might be split among the sockets like
 * this:
 *
 *	Socket A: 0, 1, 2
 *	Socket B: 3, 4, 5
 *
 * When Node 0 fills up, its memory should be migrated to
 * Node 1.  When Node 1 fills up, it should be migrated to
 * Node 2.  The migration path start on the nodes with the
 * processors (since allocations default to this node) and
 * fast memory, progress through medium and end with the
 * slow memory:
 *
 *	0 -> 1 -> 2 -> stop
 *	3 -> 4 -> 5 -> stop
 *
 * This is represented in the node_demotion[] like this:
 *
 *	{  1, // Node 0 migrates to 1
 *	   2, // Node 1 migrates to 2
 *	  -1, // Node 2 does not migrate
 *	   4, // Node 3 migrates to 4
 *	   5, // Node 4 migrates to 5
 *	  -1} // Node 5 does not migrate
 */

/*
 * Writes to this array occur without locking.  Cycles are
 * not allowed: Node X demotes to Y which demotes to X...
 *
 * If multiple reads are performed, a single rcu_read_lock()
 * must be held over all reads to ensure that no cycles are
 * observed.
 */
static int node_demotion[MAX_NUMNODES] __read_mostly =
	{[0 ...  MAX_NUMNODES - 1] = NUMA_NO_NODE};

/**
 * next_demotion_node() - Get the next node in the demotion path
 * @node: The starting node to lookup the next node
 *
1145
 * Return: node id for next memory node in the demotion path hierarchy
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
 * from @node; NUMA_NO_NODE if @node is terminal.  This does not keep
 * @node online or guarantee that it *continues* to be the next demotion
 * target.
 */
int next_demotion_node(int node)
{
	int target;

	/*
	 * node_demotion[] is updated without excluding this
	 * function from running.  RCU doesn't provide any
	 * compiler barriers, so the READ_ONCE() is required
	 * to avoid compiler reordering or read merging.
	 *
	 * Make sure to use RCU over entire code blocks if
	 * node_demotion[] reads need to be consistent.
	 */
	rcu_read_lock();
	target = READ_ONCE(node_demotion[node]);
	rcu_read_unlock();

	return target;
}

1170 1171 1172 1173
/*
 * Obtain the lock on page, remove all ptes and migrate the page
 * to the newly allocated page in newpage.
 */
1174
static int unmap_and_move(new_page_t get_new_page,
1175 1176
				   free_page_t put_new_page,
				   unsigned long private, struct page *page,
1177
				   int force, enum migrate_mode mode,
1178 1179
				   enum migrate_reason reason,
				   struct list_head *ret)
1180
{
1181
	int rc = MIGRATEPAGE_SUCCESS;
1182
	struct page *newpage = NULL;
1183

1184
	if (!thp_migration_supported() && PageTransHuge(page))
1185
		return -ENOSYS;
1186

1187 1188
	if (page_count(page) == 1) {
		/* page was freed from under us. So we are done. */
1189 1190
		ClearPageActive(page);
		ClearPageUnevictable(page);
1191 1192 1193 1194 1195 1196
		if (unlikely(__PageMovable(page))) {
			lock_page(page);
			if (!PageMovable(page))
				__ClearPageIsolated(page);
			unlock_page(page);
		}
1197 1198 1199
		goto out;
	}

1200 1201 1202 1203
	newpage = get_new_page(page, private);
	if (!newpage)
		return -ENOMEM;

1204
	rc = __unmap_and_move(page, newpage, force, mode);
1205
	if (rc == MIGRATEPAGE_SUCCESS)
1206
		set_page_owner_migrate_reason(newpage, reason);
1207

1208
out:
1209
	if (rc != -EAGAIN) {
1210 1211 1212
		/*
		 * A page that has been migrated has all references
		 * removed and will be freed. A page that has not been
1213
		 * migrated will have kept its references and be restored.
1214 1215
		 */
		list_del(&page->lru);
1216
	}
1217

1218 1219 1220 1221 1222 1223
	/*
	 * If migration is successful, releases reference grabbed during
	 * isolation. Otherwise, restore the page to right list unless
	 * we want to retry.
	 */
	if (rc == MIGRATEPAGE_SUCCESS) {
1224 1225 1226 1227 1228 1229
		/*
		 * Compaction can migrate also non-LRU pages which are
		 * not accounted to NR_ISOLATED_*. They can be recognized
		 * as __PageMovable
		 */
		if (likely(!__PageMovable(page)))
1230
			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
1231
					page_is_file_lru(page), -thp_nr_pages(page));
1232

1233
		if (reason != MR_MEMORY_FAILURE)
1234
			/*
1235
			 * We release the page in page_handle_poison.
1236
			 */
1237
			put_page(page);
1238
	} else {
1239 1240
		if (rc != -EAGAIN)
			list_add_tail(&page->lru, ret);
1241

1242 1243 1244 1245
		if (put_new_page)
			put_new_page(newpage, private);
		else
			put_page(newpage);
1246
	}
1247

1248 1249 1250
	return rc;
}

1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
/*
 * 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,
1270 1271
				free_page_t put_new_page, unsigned long private,
				struct page *hpage, int force,
1272 1273
				enum migrate_mode mode, int reason,
				struct list_head *ret)
1274
{
1275
	int rc = -EAGAIN;
1276
	int page_was_mapped = 0;
1277
	struct page *new_hpage;
1278
	struct anon_vma *anon_vma = NULL;
1279
	struct address_space *mapping = NULL;
1280

1281
	/*
1282
	 * Migratability of hugepages depends on architectures and their size.
1283 1284 1285 1286 1287
	 * 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.
	 */
1288
	if (!hugepage_migration_supported(page_hstate(hpage))) {
1289
		list_move_tail(&hpage->lru, ret);
1290
		return -ENOSYS;
1291
	}
1292

1293 1294 1295 1296 1297 1298
	if (page_count(hpage) == 1) {
		/* page was freed from under us. So we are done. */
		putback_active_hugepage(hpage);
		return MIGRATEPAGE_SUCCESS;
	}

1299
	new_hpage = get_new_page(hpage, private);
1300 1301 1302 1303
	if (!new_hpage)
		return -ENOMEM;

	if (!trylock_page(hpage)) {
1304
		if (!force)
1305
			goto out;
1306 1307 1308 1309 1310 1311 1312
		switch (mode) {
		case MIGRATE_SYNC:
		case MIGRATE_SYNC_NO_COPY:
			break;
		default:
			goto out;
		}
1313 1314 1315
		lock_page(hpage);
	}

1316 1317 1318 1319 1320
	/*
	 * Check for pages which are in the process of being freed.  Without
	 * page_mapping() set, hugetlbfs specific move page routine will not
	 * be called and we could leak usage counts for subpools.
	 */
1321
	if (hugetlb_page_subpool(hpage) && !page_mapping(hpage)) {
1322 1323 1324 1325
		rc = -EBUSY;
		goto out_unlock;
	}

1326 1327
	if (PageAnon(hpage))
		anon_vma = page_get_anon_vma(hpage);
1328

1329 1330 1331
	if (unlikely(!trylock_page(new_hpage)))
		goto put_anon;

1332
	if (page_mapped(hpage)) {
1333
		bool mapping_locked = false;
1334
		enum ttu_flags ttu = 0;
1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349

		if (!PageAnon(hpage)) {
			/*
			 * In shared mappings, try_to_unmap could potentially
			 * call huge_pmd_unshare.  Because of this, take
			 * semaphore in write mode here and set TTU_RMAP_LOCKED
			 * to let lower levels know we have taken the lock.
			 */
			mapping = hugetlb_page_mapping_lock_write(hpage);
			if (unlikely(!mapping))
				goto unlock_put_anon;

			mapping_locked = true;
			ttu |= TTU_RMAP_LOCKED;
		}
1350

1351
		try_to_migrate(hpage, ttu);
1352
		page_was_mapped = 1;
1353 1354 1355

		if (mapping_locked)
			i_mmap_unlock_write(mapping);
1356
	}
1357 1358

	if (!page_mapped(hpage))
1359
		rc = move_to_new_page(new_hpage, hpage, mode);
1360

1361
	if (page_was_mapped)
1362
		remove_migration_ptes(hpage,
1363
			rc == MIGRATEPAGE_SUCCESS ? new_hpage : hpage, false);
1364

1365
unlock_put_anon:
1366 1367 1368
	unlock_page(new_hpage);

put_anon:
H
Hugh Dickins 已提交
1369
	if (anon_vma)
1370
		put_anon_vma(anon_vma);
1371

1372
	if (rc == MIGRATEPAGE_SUCCESS) {
1373
		move_hugetlb_state(hpage, new_hpage, reason);
1374 1375
		put_new_page = NULL;
	}
1376

1377
out_unlock:
1378
	unlock_page(hpage);
1379
out:
1380
	if (rc == MIGRATEPAGE_SUCCESS)
1381
		putback_active_hugepage(hpage);
1382
	else if (rc != -EAGAIN)
1383
		list_move_tail(&hpage->lru, ret);
1384 1385 1386 1387 1388 1389

	/*
	 * If migration was not successful and there's a freeing callback, use
	 * it.  Otherwise, put_page() will drop the reference grabbed during
	 * isolation.
	 */
1390
	if (put_new_page)
1391 1392
		put_new_page(new_hpage, private);
	else
1393
		putback_active_hugepage(new_hpage);
1394

1395 1396 1397
	return rc;
}

1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411
static inline int try_split_thp(struct page *page, struct page **page2,
				struct list_head *from)
{
	int rc = 0;

	lock_page(page);
	rc = split_huge_page_to_list(page, from);
	unlock_page(page);
	if (!rc)
		list_safe_reset_next(page, *page2, lru);

	return rc;
}

1412
/*
1413 1414
 * migrate_pages - migrate the pages specified in a list, to the free pages
 *		   supplied as the target for the page migration
1415
 *
1416 1417 1418
 * @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.
1419 1420
 * @put_new_page:	The function used to free target pages if migration
 *			fails, or NULL if no special handling is necessary.
1421 1422 1423 1424
 * @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.
1425 1426
 * @ret_succeeded:	Set to the number of pages migrated successfully if
 *			the caller passes a non-NULL pointer.
1427
 *
1428 1429
 * 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.
1430 1431
 * It is caller's responsibility to call putback_movable_pages() to return pages
 * to the LRU or free list only if ret != 0.
1432
 *
1433
 * Returns the number of pages that were not migrated, or an error code.
1434
 */
1435
int migrate_pages(struct list_head *from, new_page_t get_new_page,
1436
		free_page_t put_new_page, unsigned long private,
1437
		enum migrate_mode mode, int reason, unsigned int *ret_succeeded)
1438
{
1439
	int retry = 1;
1440
	int thp_retry = 1;
1441
	int nr_failed = 0;
1442
	int nr_succeeded = 0;
1443 1444 1445
	int nr_thp_succeeded = 0;
	int nr_thp_failed = 0;
	int nr_thp_split = 0;
1446
	int pass = 0;
1447
	bool is_thp = false;
1448 1449 1450
	struct page *page;
	struct page *page2;
	int swapwrite = current->flags & PF_SWAPWRITE;
1451
	int rc, nr_subpages;
1452
	LIST_HEAD(ret_pages);
1453
	bool nosplit = (reason == MR_NUMA_MISPLACED);
1454

1455 1456
	trace_mm_migrate_pages_start(mode, reason);

1457 1458 1459
	if (!swapwrite)
		current->flags |= PF_SWAPWRITE;

1460
	for (pass = 0; pass < 10 && (retry || thp_retry); pass++) {
1461
		retry = 0;
1462
		thp_retry = 0;
1463

1464
		list_for_each_entry_safe(page, page2, from, lru) {
1465
retry:
1466 1467 1468 1469 1470
			/*
			 * THP statistics is based on the source huge page.
			 * Capture required information that might get lost
			 * during migration.
			 */
1471
			is_thp = PageTransHuge(page) && !PageHuge(page);
1472
			nr_subpages = thp_nr_pages(page);
1473
			cond_resched();
1474

1475 1476
			if (PageHuge(page))
				rc = unmap_and_move_huge_page(get_new_page,
1477
						put_new_page, private, page,
1478 1479
						pass > 2, mode, reason,
						&ret_pages);
1480
			else
1481
				rc = unmap_and_move(get_new_page, put_new_page,
1482
						private, page, pass > 2, mode,
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
						reason, &ret_pages);
			/*
			 * The rules are:
			 *	Success: non hugetlb page will be freed, hugetlb
			 *		 page will be put back
			 *	-EAGAIN: stay on the from list
			 *	-ENOMEM: stay on the from list
			 *	Other errno: put on ret_pages list then splice to
			 *		     from list
			 */
1493
			switch(rc) {
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520
			/*
			 * THP migration might be unsupported or the
			 * allocation could've failed so we should
			 * retry on the same page with the THP split
			 * to base pages.
			 *
			 * Head page is retried immediately and tail
			 * pages are added to the tail of the list so
			 * we encounter them after the rest of the list
			 * is processed.
			 */
			case -ENOSYS:
				/* THP migration is unsupported */
				if (is_thp) {
					if (!try_split_thp(page, &page2, from)) {
						nr_thp_split++;
						goto retry;
					}

					nr_thp_failed++;
					nr_failed += nr_subpages;
					break;
				}

				/* Hugetlb migration is unsupported */
				nr_failed++;
				break;
1521
			case -ENOMEM:
1522
				/*
1523 1524
				 * When memory is low, don't bother to try to migrate
				 * other pages, just exit.
1525
				 * THP NUMA faulting doesn't split THP to retry.
1526
				 */
1527
				if (is_thp && !nosplit) {
1528
					if (!try_split_thp(page, &page2, from)) {
1529
						nr_thp_split++;
1530 1531
						goto retry;
					}
1532

1533 1534 1535 1536
					nr_thp_failed++;
					nr_failed += nr_subpages;
					goto out;
				}
1537
				nr_failed++;
1538
				goto out;
1539
			case -EAGAIN:
1540 1541 1542 1543
				if (is_thp) {
					thp_retry++;
					break;
				}
1544
				retry++;
1545
				break;
1546
			case MIGRATEPAGE_SUCCESS:
1547 1548 1549 1550 1551
				if (is_thp) {
					nr_thp_succeeded++;
					nr_succeeded += nr_subpages;
					break;
				}
1552
				nr_succeeded++;
1553 1554
				break;
			default:
1555
				/*
1556
				 * Permanent failure (-EBUSY, etc.):
1557 1558 1559 1560
				 * unlike -EAGAIN case, the failed page is
				 * removed from migration page list and not
				 * retried in the next outer loop.
				 */
1561 1562 1563 1564 1565
				if (is_thp) {
					nr_thp_failed++;
					nr_failed += nr_subpages;
					break;
				}
1566
				nr_failed++;
1567
				break;
1568
			}
1569 1570
		}
	}
1571 1572
	nr_failed += retry + thp_retry;
	nr_thp_failed += thp_retry;
1573
	rc = nr_failed;
1574
out:
1575 1576 1577 1578 1579 1580
	/*
	 * Put the permanent failure page back to migration list, they
	 * will be put back to the right list by the caller.
	 */
	list_splice(&ret_pages, from);

1581 1582 1583 1584 1585 1586 1587
	count_vm_events(PGMIGRATE_SUCCESS, nr_succeeded);
	count_vm_events(PGMIGRATE_FAIL, nr_failed);
	count_vm_events(THP_MIGRATION_SUCCESS, nr_thp_succeeded);
	count_vm_events(THP_MIGRATION_FAIL, nr_thp_failed);
	count_vm_events(THP_MIGRATION_SPLIT, nr_thp_split);
	trace_mm_migrate_pages(nr_succeeded, nr_failed, nr_thp_succeeded,
			       nr_thp_failed, nr_thp_split, mode, reason);
1588

1589 1590 1591
	if (!swapwrite)
		current->flags &= ~PF_SWAPWRITE;

1592 1593 1594
	if (ret_succeeded)
		*ret_succeeded = nr_succeeded;

1595
	return rc;
1596
}
1597

1598
struct page *alloc_migration_target(struct page *page, unsigned long private)
1599
{
1600 1601
	struct migration_target_control *mtc;
	gfp_t gfp_mask;
1602 1603
	unsigned int order = 0;
	struct page *new_page = NULL;
1604 1605 1606 1607 1608 1609 1610 1611
	int nid;
	int zidx;

	mtc = (struct migration_target_control *)private;
	gfp_mask = mtc->gfp_mask;
	nid = mtc->nid;
	if (nid == NUMA_NO_NODE)
		nid = page_to_nid(page);
1612

1613 1614 1615
	if (PageHuge(page)) {
		struct hstate *h = page_hstate(compound_head(page));

1616 1617
		gfp_mask = htlb_modify_alloc_mask(h, gfp_mask);
		return alloc_huge_page_nodemask(h, nid, mtc->nmask, gfp_mask);
1618
	}
1619 1620

	if (PageTransHuge(page)) {
1621 1622 1623 1624 1625
		/*
		 * clear __GFP_RECLAIM to make the migration callback
		 * consistent with regular THP allocations.
		 */
		gfp_mask &= ~__GFP_RECLAIM;
1626 1627 1628
		gfp_mask |= GFP_TRANSHUGE;
		order = HPAGE_PMD_ORDER;
	}
1629 1630
	zidx = zone_idx(page_zone(page));
	if (is_highmem_idx(zidx) || zidx == ZONE_MOVABLE)
1631 1632
		gfp_mask |= __GFP_HIGHMEM;

1633
	new_page = __alloc_pages(gfp_mask, order, nid, mtc->nmask);
1634 1635 1636 1637 1638 1639 1640

	if (new_page && PageTransHuge(new_page))
		prep_transhuge_page(new_page);

	return new_page;
}

1641 1642
#ifdef CONFIG_NUMA

1643
static int store_status(int __user *status, int start, int value, int nr)
1644
{
1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
	while (nr-- > 0) {
		if (put_user(value, status + start))
			return -EFAULT;
		start++;
	}

	return 0;
}

static int do_move_pages_to_node(struct mm_struct *mm,
		struct list_head *pagelist, int node)
{
	int err;
1658 1659 1660 1661
	struct migration_target_control mtc = {
		.nid = node,
		.gfp_mask = GFP_HIGHUSER_MOVABLE | __GFP_THISNODE,
	};
1662

1663
	err = migrate_pages(pagelist, alloc_migration_target, NULL,
1664
		(unsigned long)&mtc, MIGRATE_SYNC, MR_SYSCALL, NULL);
1665 1666 1667
	if (err)
		putback_movable_pages(pagelist);
	return err;
1668 1669 1670
}

/*
1671 1672
 * Resolves the given address to a struct page, isolates it from the LRU and
 * puts it to the given pagelist.
1673 1674 1675 1676 1677
 * Returns:
 *     errno - if the page cannot be found/isolated
 *     0 - when it doesn't have to be migrated because it is already on the
 *         target node
 *     1 - when it has been queued
1678
 */
1679 1680
static int add_page_for_migration(struct mm_struct *mm, unsigned long addr,
		int node, struct list_head *pagelist, bool migrate_all)
1681
{
1682 1683 1684
	struct vm_area_struct *vma;
	struct page *page;
	unsigned int follflags;
1685 1686
	int err;

1687
	mmap_read_lock(mm);
1688 1689 1690 1691
	err = -EFAULT;
	vma = find_vma(mm, addr);
	if (!vma || addr < vma->vm_start || !vma_migratable(vma))
		goto out;
1692

1693 1694 1695
	/* FOLL_DUMP to ignore special (like zero) pages */
	follflags = FOLL_GET | FOLL_DUMP;
	page = follow_page(vma, addr, follflags);
1696

1697 1698 1699
	err = PTR_ERR(page);
	if (IS_ERR(page))
		goto out;
1700

1701 1702 1703
	err = -ENOENT;
	if (!page)
		goto out;
1704

1705 1706 1707
	err = 0;
	if (page_to_nid(page) == node)
		goto out_putpage;
1708

1709 1710 1711
	err = -EACCES;
	if (page_mapcount(page) > 1 && !migrate_all)
		goto out_putpage;
1712

1713 1714 1715
	if (PageHuge(page)) {
		if (PageHead(page)) {
			isolate_huge_page(page, pagelist);
1716
			err = 1;
1717
		}
1718 1719
	} else {
		struct page *head;
1720

1721 1722
		head = compound_head(page);
		err = isolate_lru_page(head);
1723
		if (err)
1724
			goto out_putpage;
1725

1726
		err = 1;
1727 1728
		list_add_tail(&head->lru, pagelist);
		mod_node_page_state(page_pgdat(head),
1729
			NR_ISOLATED_ANON + page_is_file_lru(head),
1730
			thp_nr_pages(head));
1731 1732 1733 1734 1735 1736 1737 1738 1739
	}
out_putpage:
	/*
	 * Either remove the duplicate refcount from
	 * isolate_lru_page() or drop the page ref if it was
	 * not isolated.
	 */
	put_page(page);
out:
1740
	mmap_read_unlock(mm);
1741 1742 1743
	return err;
}

1744 1745 1746 1747 1748 1749
static int move_pages_and_store_status(struct mm_struct *mm, int node,
		struct list_head *pagelist, int __user *status,
		int start, int i, unsigned long nr_pages)
{
	int err;

1750 1751 1752
	if (list_empty(pagelist))
		return 0;

1753 1754 1755 1756 1757 1758
	err = do_move_pages_to_node(mm, pagelist, node);
	if (err) {
		/*
		 * Positive err means the number of failed
		 * pages to migrate.  Since we are going to
		 * abort and return the number of non-migrated
1759
		 * pages, so need to include the rest of the
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
		 * nr_pages that have not been attempted as
		 * well.
		 */
		if (err > 0)
			err += nr_pages - i - 1;
		return err;
	}
	return store_status(status, start, node, i - start);
}

1770 1771 1772 1773
/*
 * Migrate an array of page address onto an array of nodes and fill
 * the corresponding array of status.
 */
1774
static int do_pages_move(struct mm_struct *mm, nodemask_t task_nodes,
1775 1776 1777 1778 1779
			 unsigned long nr_pages,
			 const void __user * __user *pages,
			 const int __user *nodes,
			 int __user *status, int flags)
{
1780 1781 1782 1783
	int current_node = NUMA_NO_NODE;
	LIST_HEAD(pagelist);
	int start, i;
	int err = 0, err1;
1784

1785
	lru_cache_disable();
1786

1787 1788 1789 1790
	for (i = start = 0; i < nr_pages; i++) {
		const void __user *p;
		unsigned long addr;
		int node;
1791

1792 1793 1794 1795 1796
		err = -EFAULT;
		if (get_user(p, pages + i))
			goto out_flush;
		if (get_user(node, nodes + i))
			goto out_flush;
1797
		addr = (unsigned long)untagged_addr(p);
1798 1799 1800 1801 1802 1803

		err = -ENODEV;
		if (node < 0 || node >= MAX_NUMNODES)
			goto out_flush;
		if (!node_state(node, N_MEMORY))
			goto out_flush;
1804

1805 1806 1807 1808 1809 1810 1811 1812
		err = -EACCES;
		if (!node_isset(node, task_nodes))
			goto out_flush;

		if (current_node == NUMA_NO_NODE) {
			current_node = node;
			start = i;
		} else if (node != current_node) {
1813 1814
			err = move_pages_and_store_status(mm, current_node,
					&pagelist, status, start, i, nr_pages);
1815 1816 1817 1818
			if (err)
				goto out;
			start = i;
			current_node = node;
1819 1820
		}

1821 1822 1823 1824 1825 1826
		/*
		 * Errors in the page lookup or isolation are not fatal and we simply
		 * report them via status
		 */
		err = add_page_for_migration(mm, addr, current_node,
				&pagelist, flags & MPOL_MF_MOVE_ALL);
1827

1828
		if (err > 0) {
1829 1830 1831
			/* The page is successfully queued for migration */
			continue;
		}
1832

1833 1834 1835 1836 1837
		/*
		 * If the page is already on the target node (!err), store the
		 * node, otherwise, store the err.
		 */
		err = store_status(status, i, err ? : current_node, 1);
1838 1839
		if (err)
			goto out_flush;
1840

1841 1842
		err = move_pages_and_store_status(mm, current_node, &pagelist,
				status, start, i, nr_pages);
1843 1844
		if (err)
			goto out;
1845
		current_node = NUMA_NO_NODE;
1846
	}
1847 1848
out_flush:
	/* Make sure we do not overwrite the existing error */
1849 1850
	err1 = move_pages_and_store_status(mm, current_node, &pagelist,
				status, start, i, nr_pages);
1851
	if (err >= 0)
1852
		err = err1;
1853
out:
1854
	lru_cache_enable();
1855 1856 1857
	return err;
}

1858
/*
1859
 * Determine the nodes of an array of pages and store it in an array of status.
1860
 */
1861 1862
static void do_pages_stat_array(struct mm_struct *mm, unsigned long nr_pages,
				const void __user **pages, int *status)
1863
{
1864 1865
	unsigned long i;

1866
	mmap_read_lock(mm);
1867

1868
	for (i = 0; i < nr_pages; i++) {
1869
		unsigned long addr = (unsigned long)(*pages);
1870 1871
		struct vm_area_struct *vma;
		struct page *page;
1872
		int err = -EFAULT;
1873

1874 1875
		vma = vma_lookup(mm, addr);
		if (!vma)
1876 1877
			goto set_status;

1878 1879
		/* FOLL_DUMP to ignore special (like zero) pages */
		page = follow_page(vma, addr, FOLL_DUMP);
1880 1881 1882 1883 1884

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

1885
		err = page ? page_to_nid(page) : -ENOENT;
1886
set_status:
1887 1888 1889 1890 1891 1892
		*status = err;

		pages++;
		status++;
	}

1893
	mmap_read_unlock(mm);
1894 1895
}

1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
static int get_compat_pages_array(const void __user *chunk_pages[],
				  const void __user * __user *pages,
				  unsigned long chunk_nr)
{
	compat_uptr_t __user *pages32 = (compat_uptr_t __user *)pages;
	compat_uptr_t p;
	int i;

	for (i = 0; i < chunk_nr; i++) {
		if (get_user(p, pages32 + i))
			return -EFAULT;
		chunk_pages[i] = compat_ptr(p);
	}

	return 0;
}

1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924
/*
 * 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];

1925 1926
	while (nr_pages) {
		unsigned long chunk_nr;
1927

1928 1929 1930 1931
		chunk_nr = nr_pages;
		if (chunk_nr > DO_PAGES_STAT_CHUNK_NR)
			chunk_nr = DO_PAGES_STAT_CHUNK_NR;

1932 1933 1934 1935 1936 1937 1938 1939 1940
		if (in_compat_syscall()) {
			if (get_compat_pages_array(chunk_pages, pages,
						   chunk_nr))
				break;
		} else {
			if (copy_from_user(chunk_pages, pages,
				      chunk_nr * sizeof(*chunk_pages)))
				break;
		}
1941 1942 1943

		do_pages_stat_array(mm, chunk_nr, chunk_pages, chunk_status);

1944 1945
		if (copy_to_user(status, chunk_status, chunk_nr * sizeof(*status)))
			break;
1946

1947 1948 1949 1950 1951
		pages += chunk_nr;
		status += chunk_nr;
		nr_pages -= chunk_nr;
	}
	return nr_pages ? -EFAULT : 0;
1952 1953
}

1954
static struct mm_struct *find_mm_struct(pid_t pid, nodemask_t *mem_nodes)
1955 1956 1957 1958
{
	struct task_struct *task;
	struct mm_struct *mm;

1959 1960 1961 1962 1963 1964 1965 1966 1967
	/*
	 * There is no need to check if current process has the right to modify
	 * the specified process when they are same.
	 */
	if (!pid) {
		mmget(current->mm);
		*mem_nodes = cpuset_mems_allowed(current);
		return current->mm;
	}
1968 1969

	/* Find the mm_struct */
1970
	rcu_read_lock();
1971
	task = find_task_by_vpid(pid);
1972
	if (!task) {
1973
		rcu_read_unlock();
1974
		return ERR_PTR(-ESRCH);
1975
	}
1976
	get_task_struct(task);
1977 1978 1979

	/*
	 * Check if this process has the right to modify the specified
1980
	 * process. Use the regular "ptrace_may_access()" checks.
1981
	 */
1982
	if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS)) {
1983
		rcu_read_unlock();
1984
		mm = ERR_PTR(-EPERM);
1985
		goto out;
1986
	}
1987
	rcu_read_unlock();
1988

1989 1990
	mm = ERR_PTR(security_task_movememory(task));
	if (IS_ERR(mm))
1991
		goto out;
1992
	*mem_nodes = cpuset_mems_allowed(task);
1993
	mm = get_task_mm(task);
1994
out:
1995
	put_task_struct(task);
1996
	if (!mm)
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
		mm = ERR_PTR(-EINVAL);
	return mm;
}

/*
 * Move a list of pages in the address space of the currently executing
 * process.
 */
static int kernel_move_pages(pid_t pid, unsigned long nr_pages,
			     const void __user * __user *pages,
			     const int __user *nodes,
			     int __user *status, int flags)
{
	struct mm_struct *mm;
	int err;
	nodemask_t task_nodes;

	/* Check flags */
	if (flags & ~(MPOL_MF_MOVE|MPOL_MF_MOVE_ALL))
2016 2017
		return -EINVAL;

2018 2019 2020 2021 2022 2023 2024
	if ((flags & MPOL_MF_MOVE_ALL) && !capable(CAP_SYS_NICE))
		return -EPERM;

	mm = find_mm_struct(pid, &task_nodes);
	if (IS_ERR(mm))
		return PTR_ERR(mm);

2025 2026 2027 2028 2029
	if (nodes)
		err = do_pages_move(mm, task_nodes, nr_pages, pages,
				    nodes, status, flags);
	else
		err = do_pages_stat(mm, nr_pages, pages, status);
2030 2031 2032 2033 2034

	mmput(mm);
	return err;
}

2035 2036 2037 2038 2039 2040 2041 2042
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)
{
	return kernel_move_pages(pid, nr_pages, pages, nodes, status, flags);
}

2043 2044 2045 2046 2047 2048
#ifdef CONFIG_NUMA_BALANCING
/*
 * Returns true if this is a safe migration target node for misplaced NUMA
 * pages. Currently it only checks the watermarks which crude
 */
static bool migrate_balanced_pgdat(struct pglist_data *pgdat,
2049
				   unsigned long nr_migrate_pages)
2050 2051
{
	int z;
2052

2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
	for (z = pgdat->nr_zones - 1; z >= 0; z--) {
		struct zone *zone = pgdat->node_zones + z;

		if (!populated_zone(zone))
			continue;

		/* Avoid waking kswapd by allocating pages_to_migrate pages. */
		if (!zone_watermark_ok(zone, 0,
				       high_wmark_pages(zone) +
				       nr_migrate_pages,
2063
				       ZONE_MOVABLE, 0))
2064 2065 2066 2067 2068 2069 2070
			continue;
		return true;
	}
	return false;
}

static struct page *alloc_misplaced_dst_page(struct page *page,
2071
					   unsigned long data)
2072 2073 2074 2075
{
	int nid = (int) data;
	struct page *newpage;

2076
	newpage = __alloc_pages_node(nid,
2077 2078 2079
					 (GFP_HIGHUSER_MOVABLE |
					  __GFP_THISNODE | __GFP_NOMEMALLOC |
					  __GFP_NORETRY | __GFP_NOWARN) &
2080
					 ~__GFP_RECLAIM, 0);
2081

2082 2083 2084
	return newpage;
}

2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101
static struct page *alloc_misplaced_dst_page_thp(struct page *page,
						 unsigned long data)
{
	int nid = (int) data;
	struct page *newpage;

	newpage = alloc_pages_node(nid, (GFP_TRANSHUGE_LIGHT | __GFP_THISNODE),
				   HPAGE_PMD_ORDER);
	if (!newpage)
		goto out;

	prep_transhuge_page(newpage);

out:
	return newpage;
}

2102
static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
2103
{
2104
	int page_lru;
2105
	int nr_pages = thp_nr_pages(page);
2106

2107
	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
2108

2109 2110 2111 2112
	/* Do not migrate THP mapped by multiple processes */
	if (PageTransHuge(page) && total_mapcount(page) > 1)
		return 0;

2113
	/* Avoid migrating to a node that is nearly full */
2114
	if (!migrate_balanced_pgdat(pgdat, nr_pages))
2115
		return 0;
2116

2117 2118
	if (isolate_lru_page(page))
		return 0;
2119

2120
	page_lru = page_is_file_lru(page);
2121
	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
2122
			    nr_pages);
2123

2124
	/*
2125 2126 2127
	 * Isolating the page has taken another reference, so the
	 * caller's reference can be safely dropped without the page
	 * disappearing underneath us during migration.
2128 2129
	 */
	put_page(page);
2130
	return 1;
2131 2132 2133 2134 2135 2136 2137
}

/*
 * Attempt to migrate a misplaced page to the specified destination
 * node. Caller is expected to have an elevated reference count on
 * the page that will be dropped by this function before returning.
 */
2138 2139
int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
			   int node)
2140 2141
{
	pg_data_t *pgdat = NODE_DATA(node);
2142
	int isolated;
2143 2144
	int nr_remaining;
	LIST_HEAD(migratepages);
2145 2146
	new_page_t *new;
	bool compound;
2147
	int nr_pages = thp_nr_pages(page);
2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159

	/*
	 * PTE mapped THP or HugeTLB page can't reach here so the page could
	 * be either base page or THP.  And it must be head page if it is
	 * THP.
	 */
	compound = PageTransHuge(page);

	if (compound)
		new = alloc_misplaced_dst_page_thp;
	else
		new = alloc_misplaced_dst_page;
2160 2161

	/*
2162 2163
	 * Don't migrate file pages that are mapped in multiple processes
	 * with execute permissions as they are probably shared libraries.
2164
	 */
2165 2166
	if (page_mapcount(page) != 1 && page_is_file_lru(page) &&
	    (vma->vm_flags & VM_EXEC))
2167 2168
		goto out;

2169 2170 2171 2172
	/*
	 * Also do not migrate dirty pages as not all filesystems can move
	 * dirty pages in MIGRATE_ASYNC mode which is a waste of cycles.
	 */
2173
	if (page_is_file_lru(page) && PageDirty(page))
2174 2175
		goto out;

2176 2177 2178 2179 2180
	isolated = numamigrate_isolate_page(pgdat, page);
	if (!isolated)
		goto out;

	list_add(&page->lru, &migratepages);
2181
	nr_remaining = migrate_pages(&migratepages, *new, NULL, node,
2182
				     MIGRATE_ASYNC, MR_NUMA_MISPLACED, NULL);
2183
	if (nr_remaining) {
2184 2185
		if (!list_empty(&migratepages)) {
			list_del(&page->lru);
2186 2187
			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
					page_is_file_lru(page), -nr_pages);
2188 2189
			putback_lru_page(page);
		}
2190 2191
		isolated = 0;
	} else
2192
		count_vm_numa_events(NUMA_PAGE_MIGRATE, nr_pages);
2193 2194
	BUG_ON(!list_empty(&migratepages));
	return isolated;
2195 2196 2197 2198

out:
	put_page(page);
	return 0;
2199
}
2200
#endif /* CONFIG_NUMA_BALANCING */
2201
#endif /* CONFIG_NUMA */
2202

2203
#ifdef CONFIG_DEVICE_PRIVATE
2204
static int migrate_vma_collect_skip(unsigned long start,
2205 2206 2207 2208 2209 2210
				    unsigned long end,
				    struct mm_walk *walk)
{
	struct migrate_vma *migrate = walk->private;
	unsigned long addr;

2211
	for (addr = start; addr < end; addr += PAGE_SIZE) {
2212
		migrate->dst[migrate->npages] = 0;
2213
		migrate->src[migrate->npages++] = 0;
2214 2215 2216 2217 2218
	}

	return 0;
}

2219
static int migrate_vma_collect_hole(unsigned long start,
2220
				    unsigned long end,
2221
				    __always_unused int depth,
2222 2223 2224 2225 2226
				    struct mm_walk *walk)
{
	struct migrate_vma *migrate = walk->private;
	unsigned long addr;

2227 2228 2229 2230
	/* Only allow populating anonymous memory. */
	if (!vma_is_anonymous(walk->vma))
		return migrate_vma_collect_skip(start, end, walk);

2231
	for (addr = start; addr < end; addr += PAGE_SIZE) {
2232
		migrate->src[migrate->npages] = MIGRATE_PFN_MIGRATE;
2233
		migrate->dst[migrate->npages] = 0;
2234 2235
		migrate->npages++;
		migrate->cpages++;
2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
	}

	return 0;
}

static int migrate_vma_collect_pmd(pmd_t *pmdp,
				   unsigned long start,
				   unsigned long end,
				   struct mm_walk *walk)
{
	struct migrate_vma *migrate = walk->private;
	struct vm_area_struct *vma = walk->vma;
	struct mm_struct *mm = vma->vm_mm;
2249
	unsigned long addr = start, unmapped = 0;
2250 2251 2252 2253 2254
	spinlock_t *ptl;
	pte_t *ptep;

again:
	if (pmd_none(*pmdp))
2255
		return migrate_vma_collect_hole(start, end, -1, walk);
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270

	if (pmd_trans_huge(*pmdp)) {
		struct page *page;

		ptl = pmd_lock(mm, pmdp);
		if (unlikely(!pmd_trans_huge(*pmdp))) {
			spin_unlock(ptl);
			goto again;
		}

		page = pmd_page(*pmdp);
		if (is_huge_zero_page(page)) {
			spin_unlock(ptl);
			split_huge_pmd(vma, pmdp, addr);
			if (pmd_trans_unstable(pmdp))
2271
				return migrate_vma_collect_skip(start, end,
2272 2273 2274 2275 2276 2277 2278
								walk);
		} else {
			int ret;

			get_page(page);
			spin_unlock(ptl);
			if (unlikely(!trylock_page(page)))
2279
				return migrate_vma_collect_skip(start, end,
2280 2281 2282 2283
								walk);
			ret = split_huge_page(page);
			unlock_page(page);
			put_page(page);
2284 2285 2286 2287
			if (ret)
				return migrate_vma_collect_skip(start, end,
								walk);
			if (pmd_none(*pmdp))
2288
				return migrate_vma_collect_hole(start, end, -1,
2289 2290 2291 2292 2293
								walk);
		}
	}

	if (unlikely(pmd_bad(*pmdp)))
2294
		return migrate_vma_collect_skip(start, end, walk);
2295 2296

	ptep = pte_offset_map_lock(mm, pmdp, addr, &ptl);
2297 2298
	arch_enter_lazy_mmu_mode();

2299
	for (; addr < end; addr += PAGE_SIZE, ptep++) {
2300
		unsigned long mpfn = 0, pfn;
2301
		struct page *page;
2302
		swp_entry_t entry;
2303 2304 2305 2306
		pte_t pte;

		pte = *ptep;

2307
		if (pte_none(pte)) {
2308 2309 2310 2311
			if (vma_is_anonymous(vma)) {
				mpfn = MIGRATE_PFN_MIGRATE;
				migrate->cpages++;
			}
2312 2313 2314
			goto next;
		}

2315 2316 2317 2318 2319 2320 2321 2322 2323 2324
		if (!pte_present(pte)) {
			/*
			 * Only care about unaddressable device page special
			 * page table entry. Other special swap entries are not
			 * migratable, and we ignore regular swapped page.
			 */
			entry = pte_to_swp_entry(pte);
			if (!is_device_private_entry(entry))
				goto next;

2325
			page = pfn_swap_entry_to_page(entry);
2326 2327 2328
			if (!(migrate->flags &
				MIGRATE_VMA_SELECT_DEVICE_PRIVATE) ||
			    page->pgmap->owner != migrate->pgmap_owner)
2329 2330
				goto next;

2331 2332
			mpfn = migrate_pfn(page_to_pfn(page)) |
					MIGRATE_PFN_MIGRATE;
2333
			if (is_writable_device_private_entry(entry))
2334 2335
				mpfn |= MIGRATE_PFN_WRITE;
		} else {
2336
			if (!(migrate->flags & MIGRATE_VMA_SELECT_SYSTEM))
2337
				goto next;
2338
			pfn = pte_pfn(pte);
2339 2340 2341 2342 2343
			if (is_zero_pfn(pfn)) {
				mpfn = MIGRATE_PFN_MIGRATE;
				migrate->cpages++;
				goto next;
			}
2344
			page = vm_normal_page(migrate->vma, addr, pte);
2345 2346 2347 2348
			mpfn = migrate_pfn(pfn) | MIGRATE_PFN_MIGRATE;
			mpfn |= pte_write(pte) ? MIGRATE_PFN_WRITE : 0;
		}

2349 2350
		/* FIXME support THP */
		if (!page || !page->mapping || PageTransCompound(page)) {
2351
			mpfn = 0;
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365
			goto next;
		}

		/*
		 * By getting a reference on the page we pin it and that blocks
		 * any kind of migration. Side effect is that it "freezes" the
		 * pte.
		 *
		 * We drop this reference after isolating the page from the lru
		 * for non device page (device page are not on the lru and thus
		 * can't be dropped from it).
		 */
		get_page(page);

2366 2367 2368 2369 2370 2371 2372 2373
		/*
		 * Optimize for the common case where page is only mapped once
		 * in one process. If we can lock the page, then we can safely
		 * set up a special migration page table entry now.
		 */
		if (trylock_page(page)) {
			pte_t swp_pte;

2374
			migrate->cpages++;
2375 2376 2377
			ptep_get_and_clear(mm, addr, ptep);

			/* Setup special migration page table entry */
2378 2379 2380 2381 2382 2383
			if (mpfn & MIGRATE_PFN_WRITE)
				entry = make_writable_migration_entry(
							page_to_pfn(page));
			else
				entry = make_readable_migration_entry(
							page_to_pfn(page));
2384
			swp_pte = swp_entry_to_pte(entry);
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
			if (pte_present(pte)) {
				if (pte_soft_dirty(pte))
					swp_pte = pte_swp_mksoft_dirty(swp_pte);
				if (pte_uffd_wp(pte))
					swp_pte = pte_swp_mkuffd_wp(swp_pte);
			} else {
				if (pte_swp_soft_dirty(pte))
					swp_pte = pte_swp_mksoft_dirty(swp_pte);
				if (pte_swp_uffd_wp(pte))
					swp_pte = pte_swp_mkuffd_wp(swp_pte);
			}
2396 2397 2398 2399 2400 2401 2402 2403 2404
			set_pte_at(mm, addr, ptep, swp_pte);

			/*
			 * This is like regular unmap: we remove the rmap and
			 * drop page refcount. Page won't be freed, as we took
			 * a reference just above.
			 */
			page_remove_rmap(page, false);
			put_page(page);
2405 2406 2407

			if (pte_present(pte))
				unmapped++;
2408 2409 2410
		} else {
			put_page(page);
			mpfn = 0;
2411 2412
		}

2413
next:
2414
		migrate->dst[migrate->npages] = 0;
2415 2416
		migrate->src[migrate->npages++] = mpfn;
	}
2417
	arch_leave_lazy_mmu_mode();
2418 2419
	pte_unmap_unlock(ptep - 1, ptl);

2420 2421 2422 2423
	/* Only flush the TLB if we actually modified any entries */
	if (unmapped)
		flush_tlb_range(walk->vma, start, end);

2424 2425 2426
	return 0;
}

2427 2428 2429 2430 2431
static const struct mm_walk_ops migrate_vma_walk_ops = {
	.pmd_entry		= migrate_vma_collect_pmd,
	.pte_hole		= migrate_vma_collect_hole,
};

2432 2433 2434 2435 2436 2437 2438 2439 2440 2441
/*
 * migrate_vma_collect() - collect pages over a range of virtual addresses
 * @migrate: migrate struct containing all migration information
 *
 * This will walk the CPU page table. For each virtual address backed by a
 * valid page, it updates the src array and takes a reference on the page, in
 * order to pin the page until we lock it and unmap it.
 */
static void migrate_vma_collect(struct migrate_vma *migrate)
{
2442
	struct mmu_notifier_range range;
2443

2444 2445 2446 2447 2448
	/*
	 * Note that the pgmap_owner is passed to the mmu notifier callback so
	 * that the registered device driver can skip invalidating device
	 * private page mappings that won't be migrated.
	 */
2449 2450
	mmu_notifier_range_init_owner(&range, MMU_NOTIFY_MIGRATE, 0,
		migrate->vma, migrate->vma->vm_mm, migrate->start, migrate->end,
2451
		migrate->pgmap_owner);
2452
	mmu_notifier_invalidate_range_start(&range);
2453

2454 2455 2456 2457
	walk_page_range(migrate->vma->vm_mm, migrate->start, migrate->end,
			&migrate_vma_walk_ops, migrate);

	mmu_notifier_invalidate_range_end(&range);
2458 2459 2460 2461 2462 2463 2464 2465
	migrate->end = migrate->start + (migrate->npages << PAGE_SHIFT);
}

/*
 * migrate_vma_check_page() - check if page is pinned or not
 * @page: struct page to check
 *
 * Pinned pages cannot be migrated. This is the same test as in
2466
 * folio_migrate_mapping(), except that here we allow migration of a
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
 * ZONE_DEVICE page.
 */
static bool migrate_vma_check_page(struct page *page)
{
	/*
	 * One extra ref because caller holds an extra reference, either from
	 * isolate_lru_page() for a regular page, or migrate_vma_collect() for
	 * a device page.
	 */
	int extra = 1;

	/*
	 * FIXME support THP (transparent huge page), it is bit more complex to
	 * check them than regular pages, because they can be mapped with a pmd
	 * or with a pte (split pte mapping).
	 */
	if (PageCompound(page))
		return false;

2486 2487 2488 2489 2490 2491 2492 2493 2494
	/* Page from ZONE_DEVICE have one extra reference */
	if (is_zone_device_page(page)) {
		/*
		 * Private page can never be pin as they have no valid pte and
		 * GUP will fail for those. Yet if there is a pending migration
		 * a thread might try to wait on the pte migration entry and
		 * will bump the page reference count. Sadly there is no way to
		 * differentiate a regular pin from migration wait. Hence to
		 * avoid 2 racing thread trying to migrate back to CPU to enter
2495
		 * infinite loop (one stopping migration because the other is
2496 2497 2498 2499 2500
		 * waiting on pte migration entry). We always return true here.
		 *
		 * FIXME proper solution is to rework migration_entry_wait() so
		 * it does not need to take a reference on page.
		 */
2501
		return is_device_private_page(page);
2502 2503
	}

2504 2505 2506 2507
	/* For file back page */
	if (page_mapping(page))
		extra += 1 + page_has_private(page);

2508 2509 2510 2511 2512 2513 2514
	if ((page_count(page) - extra) > page_mapcount(page))
		return false;

	return true;
}

/*
2515
 * migrate_vma_unmap() - replace page mapping with special migration pte entry
2516 2517
 * @migrate: migrate struct containing all migration information
 *
2518 2519 2520 2521 2522 2523
 * Isolate pages from the LRU and replace mappings (CPU page table pte) with a
 * special migration pte entry and check if it has been pinned. Pinned pages are
 * restored because we cannot migrate them.
 *
 * This is the last step before we call the device driver callback to allocate
 * destination memory and copy contents of original page over to new page.
2524
 */
2525
static void migrate_vma_unmap(struct migrate_vma *migrate)
2526 2527
{
	const unsigned long npages = migrate->npages;
2528 2529
	const unsigned long start = migrate->start;
	unsigned long addr, i, restore = 0;
2530 2531 2532 2533
	bool allow_drain = true;

	lru_add_drain();

2534
	for (i = 0; i < npages; i++) {
2535 2536 2537 2538 2539
		struct page *page = migrate_pfn_to_page(migrate->src[i]);

		if (!page)
			continue;

2540 2541 2542 2543 2544 2545 2546
		/* ZONE_DEVICE pages are not on LRU */
		if (!is_zone_device_page(page)) {
			if (!PageLRU(page) && allow_drain) {
				/* Drain CPU's pagevec */
				lru_add_drain_all();
				allow_drain = false;
			}
2547

2548
			if (isolate_lru_page(page)) {
2549 2550 2551
				migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
				migrate->cpages--;
				restore++;
2552
				continue;
2553
			}
2554 2555 2556

			/* Drop the reference we took in collect */
			put_page(page);
2557 2558
		}

2559 2560
		if (page_mapped(page))
			try_to_migrate(page, 0);
2561

2562 2563 2564 2565
		if (page_mapped(page) || !migrate_vma_check_page(page)) {
			if (!is_zone_device_page(page)) {
				get_page(page);
				putback_lru_page(page);
2566 2567
			}

2568 2569 2570
			migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
			migrate->cpages--;
			restore++;
2571
			continue;
2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584
		}
	}

	for (addr = start, i = 0; i < npages && restore; addr += PAGE_SIZE, i++) {
		struct page *page = migrate_pfn_to_page(migrate->src[i]);

		if (!page || (migrate->src[i] & MIGRATE_PFN_MIGRATE))
			continue;

		remove_migration_ptes(page, page, false);

		migrate->src[i] = 0;
		unlock_page(page);
2585
		put_page(page);
2586 2587 2588 2589
		restore--;
	}
}

2590 2591
/**
 * migrate_vma_setup() - prepare to migrate a range of memory
2592
 * @args: contains the vma, start, and pfns arrays for the migration
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608
 *
 * Returns: negative errno on failures, 0 when 0 or more pages were migrated
 * without an error.
 *
 * Prepare to migrate a range of memory virtual address range by collecting all
 * the pages backing each virtual address in the range, saving them inside the
 * src array.  Then lock those pages and unmap them. Once the pages are locked
 * and unmapped, check whether each page is pinned or not.  Pages that aren't
 * pinned have the MIGRATE_PFN_MIGRATE flag set (by this function) in the
 * corresponding src array entry.  Then restores any pages that are pinned, by
 * remapping and unlocking those pages.
 *
 * The caller should then allocate destination memory and copy source memory to
 * it for all those entries (ie with MIGRATE_PFN_VALID and MIGRATE_PFN_MIGRATE
 * flag set).  Once these are allocated and copied, the caller must update each
 * corresponding entry in the dst array with the pfn value of the destination
2609 2610
 * page and with MIGRATE_PFN_VALID. Destination pages must be locked via
 * lock_page().
2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
 *
 * Note that the caller does not have to migrate all the pages that are marked
 * with MIGRATE_PFN_MIGRATE flag in src array unless this is a migration from
 * device memory to system memory.  If the caller cannot migrate a device page
 * back to system memory, then it must return VM_FAULT_SIGBUS, which has severe
 * consequences for the userspace process, so it must be avoided if at all
 * possible.
 *
 * For empty entries inside CPU page table (pte_none() or pmd_none() is true) we
 * do set MIGRATE_PFN_MIGRATE flag inside the corresponding source array thus
I
Ingo Molnar 已提交
2621 2622
 * allowing the caller to allocate device memory for those unbacked virtual
 * addresses.  For this the caller simply has to allocate device memory and
2623
 * properly set the destination entry like for regular migration.  Note that
I
Ingo Molnar 已提交
2624 2625
 * this can still fail, and thus inside the device driver you must check if the
 * migration was successful for those entries after calling migrate_vma_pages(),
2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644
 * just like for regular migration.
 *
 * After that, the callers must call migrate_vma_pages() to go over each entry
 * in the src array that has the MIGRATE_PFN_VALID and MIGRATE_PFN_MIGRATE flag
 * set. If the corresponding entry in dst array has MIGRATE_PFN_VALID flag set,
 * then migrate_vma_pages() to migrate struct page information from the source
 * struct page to the destination struct page.  If it fails to migrate the
 * struct page information, then it clears the MIGRATE_PFN_MIGRATE flag in the
 * src array.
 *
 * At this point all successfully migrated pages have an entry in the src
 * array with MIGRATE_PFN_VALID and MIGRATE_PFN_MIGRATE flag set and the dst
 * array entry with MIGRATE_PFN_VALID flag set.
 *
 * Once migrate_vma_pages() returns the caller may inspect which pages were
 * successfully migrated, and which were not.  Successfully migrated pages will
 * have the MIGRATE_PFN_MIGRATE flag set for their src array entry.
 *
 * It is safe to update device page table after migrate_vma_pages() because
2645
 * both destination and source page are still locked, and the mmap_lock is held
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691
 * in read mode (hence no one can unmap the range being migrated).
 *
 * Once the caller is done cleaning up things and updating its page table (if it
 * chose to do so, this is not an obligation) it finally calls
 * migrate_vma_finalize() to update the CPU page table to point to new pages
 * for successfully migrated pages or otherwise restore the CPU page table to
 * point to the original source pages.
 */
int migrate_vma_setup(struct migrate_vma *args)
{
	long nr_pages = (args->end - args->start) >> PAGE_SHIFT;

	args->start &= PAGE_MASK;
	args->end &= PAGE_MASK;
	if (!args->vma || is_vm_hugetlb_page(args->vma) ||
	    (args->vma->vm_flags & VM_SPECIAL) || vma_is_dax(args->vma))
		return -EINVAL;
	if (nr_pages <= 0)
		return -EINVAL;
	if (args->start < args->vma->vm_start ||
	    args->start >= args->vma->vm_end)
		return -EINVAL;
	if (args->end <= args->vma->vm_start || args->end > args->vma->vm_end)
		return -EINVAL;
	if (!args->src || !args->dst)
		return -EINVAL;

	memset(args->src, 0, sizeof(*args->src) * nr_pages);
	args->cpages = 0;
	args->npages = 0;

	migrate_vma_collect(args);

	if (args->cpages)
		migrate_vma_unmap(args);

	/*
	 * At this point pages are locked and unmapped, and thus they have
	 * stable content and can safely be copied to destination memory that
	 * is allocated by the drivers.
	 */
	return 0;

}
EXPORT_SYMBOL(migrate_vma_setup);

2692 2693 2694 2695 2696 2697 2698 2699
/*
 * This code closely matches the code in:
 *   __handle_mm_fault()
 *     handle_pte_fault()
 *       do_anonymous_page()
 * to map in an anonymous zero page but the struct page will be a ZONE_DEVICE
 * private page.
 */
2700 2701 2702
static void migrate_vma_insert_page(struct migrate_vma *migrate,
				    unsigned long addr,
				    struct page *page,
2703
				    unsigned long *src)
2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738
{
	struct vm_area_struct *vma = migrate->vma;
	struct mm_struct *mm = vma->vm_mm;
	bool flush = false;
	spinlock_t *ptl;
	pte_t entry;
	pgd_t *pgdp;
	p4d_t *p4dp;
	pud_t *pudp;
	pmd_t *pmdp;
	pte_t *ptep;

	/* Only allow populating anonymous memory */
	if (!vma_is_anonymous(vma))
		goto abort;

	pgdp = pgd_offset(mm, addr);
	p4dp = p4d_alloc(mm, pgdp, addr);
	if (!p4dp)
		goto abort;
	pudp = pud_alloc(mm, p4dp, addr);
	if (!pudp)
		goto abort;
	pmdp = pmd_alloc(mm, pudp, addr);
	if (!pmdp)
		goto abort;

	if (pmd_trans_huge(*pmdp) || pmd_devmap(*pmdp))
		goto abort;

	/*
	 * Use pte_alloc() instead of pte_alloc_map().  We can't run
	 * pte_offset_map() on pmds where a huge pmd might be created
	 * from a different thread.
	 *
2739
	 * pte_alloc_map() is safe to use under mmap_write_lock(mm) or when
2740 2741
	 * parallel threads are excluded by other means.
	 *
2742
	 * Here we only have mmap_read_lock(mm).
2743
	 */
2744
	if (pte_alloc(mm, pmdp))
2745 2746 2747 2748 2749 2750 2751 2752
		goto abort;

	/* See the comment in pte_alloc_one_map() */
	if (unlikely(pmd_trans_unstable(pmdp)))
		goto abort;

	if (unlikely(anon_vma_prepare(vma)))
		goto abort;
2753
	if (mem_cgroup_charge(page_folio(page), vma->vm_mm, GFP_KERNEL))
2754 2755 2756 2757 2758 2759 2760 2761 2762
		goto abort;

	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * preceding stores to the page contents become visible before
	 * the set_pte_at() write.
	 */
	__SetPageUptodate(page);

2763 2764 2765 2766
	if (is_zone_device_page(page)) {
		if (is_device_private_page(page)) {
			swp_entry_t swp_entry;

2767 2768 2769 2770 2771 2772
			if (vma->vm_flags & VM_WRITE)
				swp_entry = make_writable_device_private_entry(
							page_to_pfn(page));
			else
				swp_entry = make_readable_device_private_entry(
							page_to_pfn(page));
2773
			entry = swp_entry_to_pte(swp_entry);
2774 2775 2776 2777 2778 2779 2780
		} else {
			/*
			 * For now we only support migrating to un-addressable
			 * device memory.
			 */
			pr_warn_once("Unsupported ZONE_DEVICE page type.\n");
			goto abort;
2781
		}
2782 2783 2784 2785 2786 2787 2788 2789
	} else {
		entry = mk_pte(page, vma->vm_page_prot);
		if (vma->vm_flags & VM_WRITE)
			entry = pte_mkwrite(pte_mkdirty(entry));
	}

	ptep = pte_offset_map_lock(mm, pmdp, addr, &ptl);

2790 2791 2792
	if (check_stable_address_space(mm))
		goto unlock_abort;

2793 2794 2795
	if (pte_present(*ptep)) {
		unsigned long pfn = pte_pfn(*ptep);

2796 2797
		if (!is_zero_pfn(pfn))
			goto unlock_abort;
2798
		flush = true;
2799 2800
	} else if (!pte_none(*ptep))
		goto unlock_abort;
2801 2802

	/*
2803
	 * Check for userfaultfd but do not deliver the fault. Instead,
2804 2805
	 * just back off.
	 */
2806 2807
	if (userfaultfd_missing(vma))
		goto unlock_abort;
2808 2809

	inc_mm_counter(mm, MM_ANONPAGES);
2810
	page_add_new_anon_rmap(page, vma, addr, false);
2811
	if (!is_zone_device_page(page))
2812
		lru_cache_add_inactive_or_unevictable(page, vma);
2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829
	get_page(page);

	if (flush) {
		flush_cache_page(vma, addr, pte_pfn(*ptep));
		ptep_clear_flush_notify(vma, addr, ptep);
		set_pte_at_notify(mm, addr, ptep, entry);
		update_mmu_cache(vma, addr, ptep);
	} else {
		/* No need to invalidate - it was non-present before */
		set_pte_at(mm, addr, ptep, entry);
		update_mmu_cache(vma, addr, ptep);
	}

	pte_unmap_unlock(ptep, ptl);
	*src = MIGRATE_PFN_MIGRATE;
	return;

2830 2831
unlock_abort:
	pte_unmap_unlock(ptep, ptl);
2832 2833 2834 2835
abort:
	*src &= ~MIGRATE_PFN_MIGRATE;
}

2836
/**
2837 2838 2839 2840 2841 2842 2843
 * migrate_vma_pages() - migrate meta-data from src page to dst page
 * @migrate: migrate struct containing all migration information
 *
 * This migrates struct page meta-data from source struct page to destination
 * struct page. This effectively finishes the migration from source page to the
 * destination page.
 */
2844
void migrate_vma_pages(struct migrate_vma *migrate)
2845 2846 2847
{
	const unsigned long npages = migrate->npages;
	const unsigned long start = migrate->start;
2848 2849
	struct mmu_notifier_range range;
	unsigned long addr, i;
2850
	bool notified = false;
2851 2852 2853 2854 2855 2856 2857

	for (i = 0, addr = start; i < npages; addr += PAGE_SIZE, i++) {
		struct page *newpage = migrate_pfn_to_page(migrate->dst[i]);
		struct page *page = migrate_pfn_to_page(migrate->src[i]);
		struct address_space *mapping;
		int r;

2858 2859
		if (!newpage) {
			migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
2860
			continue;
2861 2862 2863
		}

		if (!page) {
2864
			if (!(migrate->src[i] & MIGRATE_PFN_MIGRATE))
2865 2866 2867
				continue;
			if (!notified) {
				notified = true;
2868

2869 2870 2871
				mmu_notifier_range_init_owner(&range,
					MMU_NOTIFY_MIGRATE, 0, migrate->vma,
					migrate->vma->vm_mm, addr, migrate->end,
2872
					migrate->pgmap_owner);
2873
				mmu_notifier_invalidate_range_start(&range);
2874 2875
			}
			migrate_vma_insert_page(migrate, addr, newpage,
2876
						&migrate->src[i]);
2877
			continue;
2878
		}
2879 2880 2881

		mapping = page_mapping(page);

2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
		if (is_zone_device_page(newpage)) {
			if (is_device_private_page(newpage)) {
				/*
				 * For now only support private anonymous when
				 * migrating to un-addressable device memory.
				 */
				if (mapping) {
					migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
					continue;
				}
2892
			} else {
2893 2894 2895 2896 2897 2898 2899 2900 2901
				/*
				 * Other types of ZONE_DEVICE page are not
				 * supported.
				 */
				migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
				continue;
			}
		}

2902 2903 2904 2905
		r = migrate_page(mapping, newpage, page, MIGRATE_SYNC_NO_COPY);
		if (r != MIGRATEPAGE_SUCCESS)
			migrate->src[i] &= ~MIGRATE_PFN_MIGRATE;
	}
2906

2907 2908 2909 2910 2911
	/*
	 * No need to double call mmu_notifier->invalidate_range() callback as
	 * the above ptep_clear_flush_notify() inside migrate_vma_insert_page()
	 * did already call it.
	 */
2912
	if (notified)
2913
		mmu_notifier_invalidate_range_only_end(&range);
2914
}
2915
EXPORT_SYMBOL(migrate_vma_pages);
2916

2917
/**
2918 2919 2920 2921 2922 2923 2924 2925 2926 2927
 * migrate_vma_finalize() - restore CPU page table entry
 * @migrate: migrate struct containing all migration information
 *
 * This replaces the special migration pte entry with either a mapping to the
 * new page if migration was successful for that page, or to the original page
 * otherwise.
 *
 * This also unlocks the pages and puts them back on the lru, or drops the extra
 * refcount, for device pages.
 */
2928
void migrate_vma_finalize(struct migrate_vma *migrate)
2929 2930 2931 2932 2933 2934 2935 2936
{
	const unsigned long npages = migrate->npages;
	unsigned long i;

	for (i = 0; i < npages; i++) {
		struct page *newpage = migrate_pfn_to_page(migrate->dst[i]);
		struct page *page = migrate_pfn_to_page(migrate->src[i]);

2937 2938 2939 2940 2941
		if (!page) {
			if (newpage) {
				unlock_page(newpage);
				put_page(newpage);
			}
2942
			continue;
2943 2944
		}

2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955
		if (!(migrate->src[i] & MIGRATE_PFN_MIGRATE) || !newpage) {
			if (newpage) {
				unlock_page(newpage);
				put_page(newpage);
			}
			newpage = page;
		}

		remove_migration_ptes(page, newpage, false);
		unlock_page(page);

2956 2957 2958 2959
		if (is_zone_device_page(page))
			put_page(page);
		else
			putback_lru_page(page);
2960 2961 2962

		if (newpage != page) {
			unlock_page(newpage);
2963 2964 2965 2966
			if (is_zone_device_page(newpage))
				put_page(newpage);
			else
				putback_lru_page(newpage);
2967 2968 2969
		}
	}
}
2970
EXPORT_SYMBOL(migrate_vma_finalize);
2971
#endif /* CONFIG_DEVICE_PRIVATE */
2972

2973
#if defined(CONFIG_HOTPLUG_CPU)
2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113
/* Disable reclaim-based migration. */
static void __disable_all_migrate_targets(void)
{
	int node;

	for_each_online_node(node)
		node_demotion[node] = NUMA_NO_NODE;
}

static void disable_all_migrate_targets(void)
{
	__disable_all_migrate_targets();

	/*
	 * Ensure that the "disable" is visible across the system.
	 * Readers will see either a combination of before+disable
	 * state or disable+after.  They will never see before and
	 * after state together.
	 *
	 * The before+after state together might have cycles and
	 * could cause readers to do things like loop until this
	 * function finishes.  This ensures they can only see a
	 * single "bad" read and would, for instance, only loop
	 * once.
	 */
	synchronize_rcu();
}

/*
 * Find an automatic demotion target for 'node'.
 * Failing here is OK.  It might just indicate
 * being at the end of a chain.
 */
static int establish_migrate_target(int node, nodemask_t *used)
{
	int migration_target;

	/*
	 * Can not set a migration target on a
	 * node with it already set.
	 *
	 * No need for READ_ONCE() here since this
	 * in the write path for node_demotion[].
	 * This should be the only thread writing.
	 */
	if (node_demotion[node] != NUMA_NO_NODE)
		return NUMA_NO_NODE;

	migration_target = find_next_best_node(node, used);
	if (migration_target == NUMA_NO_NODE)
		return NUMA_NO_NODE;

	node_demotion[node] = migration_target;

	return migration_target;
}

/*
 * When memory fills up on a node, memory contents can be
 * automatically migrated to another node instead of
 * discarded at reclaim.
 *
 * Establish a "migration path" which will start at nodes
 * with CPUs and will follow the priorities used to build the
 * page allocator zonelists.
 *
 * The difference here is that cycles must be avoided.  If
 * node0 migrates to node1, then neither node1, nor anything
 * node1 migrates to can migrate to node0.
 *
 * This function can run simultaneously with readers of
 * node_demotion[].  However, it can not run simultaneously
 * with itself.  Exclusion is provided by memory hotplug events
 * being single-threaded.
 */
static void __set_migration_target_nodes(void)
{
	nodemask_t next_pass	= NODE_MASK_NONE;
	nodemask_t this_pass	= NODE_MASK_NONE;
	nodemask_t used_targets = NODE_MASK_NONE;
	int node;

	/*
	 * Avoid any oddities like cycles that could occur
	 * from changes in the topology.  This will leave
	 * a momentary gap when migration is disabled.
	 */
	disable_all_migrate_targets();

	/*
	 * Allocations go close to CPUs, first.  Assume that
	 * the migration path starts at the nodes with CPUs.
	 */
	next_pass = node_states[N_CPU];
again:
	this_pass = next_pass;
	next_pass = NODE_MASK_NONE;
	/*
	 * To avoid cycles in the migration "graph", ensure
	 * that migration sources are not future targets by
	 * setting them in 'used_targets'.  Do this only
	 * once per pass so that multiple source nodes can
	 * share a target node.
	 *
	 * 'used_targets' will become unavailable in future
	 * passes.  This limits some opportunities for
	 * multiple source nodes to share a destination.
	 */
	nodes_or(used_targets, used_targets, this_pass);
	for_each_node_mask(node, this_pass) {
		int target_node = establish_migrate_target(node, &used_targets);

		if (target_node == NUMA_NO_NODE)
			continue;

		/*
		 * Visit targets from this pass in the next pass.
		 * Eventually, every node will have been part of
		 * a pass, and will become set in 'used_targets'.
		 */
		node_set(target_node, next_pass);
	}
	/*
	 * 'next_pass' contains nodes which became migration
	 * targets in this pass.  Make additional passes until
	 * no more migrations targets are available.
	 */
	if (!nodes_empty(next_pass))
		goto again;
}

/*
 * For callers that do not hold get_online_mems() already.
 */
static void set_migration_target_nodes(void)
{
	get_online_mems();
	__set_migration_target_nodes();
	put_online_mems();
}
3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126

/*
 * This leaves migrate-on-reclaim transiently disabled between
 * the MEM_GOING_OFFLINE and MEM_OFFLINE events.  This runs
 * whether reclaim-based migration is enabled or not, which
 * ensures that the user can turn reclaim-based migration at
 * any time without needing to recalculate migration targets.
 *
 * These callbacks already hold get_online_mems().  That is why
 * __set_migration_target_nodes() can be used as opposed to
 * set_migration_target_nodes().
 */
static int __meminit migrate_on_reclaim_callback(struct notifier_block *self,
3127
						 unsigned long action, void *_arg)
3128
{
3129 3130 3131 3132 3133 3134 3135 3136 3137 3138
	struct memory_notify *arg = _arg;

	/*
	 * Only update the node migration order when a node is
	 * changing status, like online->offline.  This avoids
	 * the overhead of synchronize_rcu() in most cases.
	 */
	if (arg->status_change_nid < 0)
		return notifier_from_errno(0);

3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171
	switch (action) {
	case MEM_GOING_OFFLINE:
		/*
		 * Make sure there are not transient states where
		 * an offline node is a migration target.  This
		 * will leave migration disabled until the offline
		 * completes and the MEM_OFFLINE case below runs.
		 */
		disable_all_migrate_targets();
		break;
	case MEM_OFFLINE:
	case MEM_ONLINE:
		/*
		 * Recalculate the target nodes once the node
		 * reaches its final state (online or offline).
		 */
		__set_migration_target_nodes();
		break;
	case MEM_CANCEL_OFFLINE:
		/*
		 * MEM_GOING_OFFLINE disabled all the migration
		 * targets.  Reenable them.
		 */
		__set_migration_target_nodes();
		break;
	case MEM_GOING_ONLINE:
	case MEM_CANCEL_ONLINE:
		break;
	}

	return notifier_from_errno(0);
}

3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190
/*
 * React to hotplug events that might affect the migration targets
 * like events that online or offline NUMA nodes.
 *
 * The ordering is also currently dependent on which nodes have
 * CPUs.  That means we need CPU on/offline notification too.
 */
static int migration_online_cpu(unsigned int cpu)
{
	set_migration_target_nodes();
	return 0;
}

static int migration_offline_cpu(unsigned int cpu)
{
	set_migration_target_nodes();
	return 0;
}

3191 3192 3193 3194
static int __init migrate_on_reclaim_init(void)
{
	int ret;

3195 3196
	ret = cpuhp_setup_state_nocalls(CPUHP_MM_DEMOTION_DEAD, "mm/demotion:offline",
					NULL, migration_offline_cpu);
3197 3198 3199 3200 3201 3202 3203
	/*
	 * In the unlikely case that this fails, the automatic
	 * migration targets may become suboptimal for nodes
	 * where N_CPU changes.  With such a small impact in a
	 * rare case, do not bother trying to do anything special.
	 */
	WARN_ON(ret < 0);
3204 3205 3206
	ret = cpuhp_setup_state(CPUHP_AP_MM_DEMOTION_ONLINE, "mm/demotion:online",
				migration_online_cpu, NULL);
	WARN_ON(ret < 0);
3207 3208 3209 3210 3211

	hotplug_memory_notifier(migrate_on_reclaim_callback, 100);
	return 0;
}
late_initcall(migrate_on_reclaim_init);
3212
#endif /* CONFIG_HOTPLUG_CPU */
3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273

bool numa_demotion_enabled = false;

#ifdef CONFIG_SYSFS
static ssize_t numa_demotion_enabled_show(struct kobject *kobj,
					  struct kobj_attribute *attr, char *buf)
{
	return sysfs_emit(buf, "%s\n",
			  numa_demotion_enabled ? "true" : "false");
}

static ssize_t numa_demotion_enabled_store(struct kobject *kobj,
					   struct kobj_attribute *attr,
					   const char *buf, size_t count)
{
	if (!strncmp(buf, "true", 4) || !strncmp(buf, "1", 1))
		numa_demotion_enabled = true;
	else if (!strncmp(buf, "false", 5) || !strncmp(buf, "0", 1))
		numa_demotion_enabled = false;
	else
		return -EINVAL;

	return count;
}

static struct kobj_attribute numa_demotion_enabled_attr =
	__ATTR(demotion_enabled, 0644, numa_demotion_enabled_show,
	       numa_demotion_enabled_store);

static struct attribute *numa_attrs[] = {
	&numa_demotion_enabled_attr.attr,
	NULL,
};

static const struct attribute_group numa_attr_group = {
	.attrs = numa_attrs,
};

static int __init numa_init_sysfs(void)
{
	int err;
	struct kobject *numa_kobj;

	numa_kobj = kobject_create_and_add("numa", mm_kobj);
	if (!numa_kobj) {
		pr_err("failed to create numa kobject\n");
		return -ENOMEM;
	}
	err = sysfs_create_group(numa_kobj, &numa_attr_group);
	if (err) {
		pr_err("failed to register numa group\n");
		goto delete_obj;
	}
	return 0;

delete_obj:
	kobject_put(numa_kobj);
	return err;
}
subsys_initcall(numa_init_sysfs);
#endif
反馈
建议
客服 返回
顶部