memory.c 110.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
/*
 *  linux/mm/memory.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

/*
 * demand-loading started 01.12.91 - seems it is high on the list of
 * things wanted, and it should be easy to implement. - Linus
 */

/*
 * Ok, demand-loading was easy, shared pages a little bit tricker. Shared
 * pages started 02.12.91, seems to work. - Linus.
 *
 * Tested sharing by executing about 30 /bin/sh: under the old kernel it
 * would have taken more than the 6M I have free, but it worked well as
 * far as I could see.
 *
 * Also corrected some "invalidate()"s - I wasn't doing enough of them.
 */

/*
 * Real VM (paging to/from disk) started 18.12.91. Much more work and
 * thought has to go into this. Oh, well..
 * 19.12.91  -  works, somewhat. Sometimes I get faults, don't know why.
 *		Found it. Everything seems to work now.
 * 20.12.91  -  Ok, making the swap-device changeable like the root.
 */

/*
 * 05.04.94  -  Multi-page memory management added for v1.1.
 * 		Idea by Alex Bligh (alex@cconcepts.co.uk)
 *
 * 16.07.99  -  Support of BIGMEM added by Gerhard Wichert, Siemens AG
 *		(Gerhard.Wichert@pdb.siemens.de)
 *
 * Aug/Sep 2004 Changed to four level page tables (Andi Kleen)
 */

#include <linux/kernel_stat.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
H
Hugh Dickins 已提交
48
#include <linux/ksm.h>
L
Linus Torvalds 已提交
49
#include <linux/rmap.h>
50
#include <linux/export.h>
51
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
52
#include <linux/init.h>
53
#include <linux/pfn_t.h>
P
Peter Zijlstra 已提交
54
#include <linux/writeback.h>
55
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
56
#include <linux/mmu_notifier.h>
57 58 59
#include <linux/kallsyms.h>
#include <linux/swapops.h>
#include <linux/elf.h>
60
#include <linux/gfp.h>
61
#include <linux/migrate.h>
A
Andy Shevchenko 已提交
62
#include <linux/string.h>
63
#include <linux/dma-debug.h>
64
#include <linux/debugfs.h>
65
#include <linux/userfaultfd_k.h>
66
#include <linux/dax.h>
L
Linus Torvalds 已提交
67

A
Alexey Dobriyan 已提交
68
#include <asm/io.h>
69
#include <asm/mmu_context.h>
L
Linus Torvalds 已提交
70 71 72 73 74 75
#include <asm/pgalloc.h>
#include <asm/uaccess.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

76 77
#include "internal.h"

78 79
#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
80 81
#endif

A
Andy Whitcroft 已提交
82
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
/* use the per-pgdat data instead for discontigmem - mbligh */
unsigned long max_mapnr;
struct page *mem_map;

EXPORT_SYMBOL(max_mapnr);
EXPORT_SYMBOL(mem_map);
#endif

/*
 * A number of key systems in x86 including ioremap() rely on the assumption
 * that high_memory defines the upper bound on direct map memory, then end
 * of ZONE_NORMAL.  Under CONFIG_DISCONTIG this means that max_low_pfn and
 * highstart_pfn must be the same; there must be no gap between ZONE_NORMAL
 * and ZONE_HIGHMEM.
 */
void * high_memory;

EXPORT_SYMBOL(high_memory);

102 103 104 105 106 107 108 109 110 111 112 113
/*
 * Randomize the address space (stacks, mmaps, brk, etc.).
 *
 * ( When CONFIG_COMPAT_BRK=y we exclude brk from randomization,
 *   as ancient (libc5 based) binaries can segfault. )
 */
int randomize_va_space __read_mostly =
#ifdef CONFIG_COMPAT_BRK
					1;
#else
					2;
#endif
114 115 116 117

static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
118
	return 1;
119 120 121
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
122
unsigned long zero_pfn __read_mostly;
H
Hugh Dickins 已提交
123
unsigned long highest_memmap_pfn __read_mostly;
H
Hugh Dickins 已提交
124

125 126
EXPORT_SYMBOL(zero_pfn);

H
Hugh Dickins 已提交
127 128 129 130 131 132 133 134 135
/*
 * CONFIG_MMU architectures set up ZERO_PAGE in their paging_init()
 */
static int __init init_zero_pfn(void)
{
	zero_pfn = page_to_pfn(ZERO_PAGE(0));
	return 0;
}
core_initcall(init_zero_pfn);
136

K
KAMEZAWA Hiroyuki 已提交
137

138 139
#if defined(SPLIT_RSS_COUNTING)

140
void sync_mm_rss(struct mm_struct *mm)
141 142 143 144
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
145 146 147
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
148 149
		}
	}
150
	current->rss_stat.events = 0;
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
}

static void add_mm_counter_fast(struct mm_struct *mm, int member, int val)
{
	struct task_struct *task = current;

	if (likely(task->mm == mm))
		task->rss_stat.count[member] += val;
	else
		add_mm_counter(mm, member, val);
}
#define inc_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member, 1)
#define dec_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member, -1)

/* sync counter once per 64 page faults */
#define TASK_RSS_EVENTS_THRESH	(64)
static void check_sync_rss_stat(struct task_struct *task)
{
	if (unlikely(task != current))
		return;
	if (unlikely(task->rss_stat.events++ > TASK_RSS_EVENTS_THRESH))
172
		sync_mm_rss(task->mm);
173
}
174
#else /* SPLIT_RSS_COUNTING */
175 176 177 178 179 180 181 182

#define inc_mm_counter_fast(mm, member) inc_mm_counter(mm, member)
#define dec_mm_counter_fast(mm, member) dec_mm_counter(mm, member)

static void check_sync_rss_stat(struct task_struct *task)
{
}

183 184 185 186
#endif /* SPLIT_RSS_COUNTING */

#ifdef HAVE_GENERIC_MMU_GATHER

187
static bool tlb_next_batch(struct mmu_gather *tlb)
188 189 190 191 192 193
{
	struct mmu_gather_batch *batch;

	batch = tlb->active;
	if (batch->next) {
		tlb->active = batch->next;
194
		return true;
195 196
	}

197
	if (tlb->batch_count == MAX_GATHER_BATCH_COUNT)
198
		return false;
199

200 201
	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
	if (!batch)
202
		return false;
203

204
	tlb->batch_count++;
205 206 207 208 209 210 211
	batch->next = NULL;
	batch->nr   = 0;
	batch->max  = MAX_GATHER_BATCH;

	tlb->active->next = batch;
	tlb->active = batch;

212
	return true;
213 214 215 216 217 218 219
}

/* tlb_gather_mmu
 *	Called to initialize an (on-stack) mmu_gather structure for page-table
 *	tear-down from @mm. The @fullmm argument is used when @mm is without
 *	users and we're going to destroy the full address space (exit/execve).
 */
220
void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end)
221 222 223
{
	tlb->mm = mm;

224 225
	/* Is it from 0 to ~0? */
	tlb->fullmm     = !(start | (end+1));
226
	tlb->need_flush_all = 0;
227 228 229 230
	tlb->local.next = NULL;
	tlb->local.nr   = 0;
	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
	tlb->active     = &tlb->local;
231
	tlb->batch_count = 0;
232 233 234 235

#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb->batch = NULL;
#endif
236
	tlb->page_size = 0;
237 238

	__tlb_reset_range(tlb);
239 240
}

241
static void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb)
242
{
243 244 245
	if (!tlb->end)
		return;

246
	tlb_flush(tlb);
247
	mmu_notifier_invalidate_range(tlb->mm, tlb->start, tlb->end);
248 249
#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb_table_flush(tlb);
250
#endif
251
	__tlb_reset_range(tlb);
252 253 254 255 256
}

static void tlb_flush_mmu_free(struct mmu_gather *tlb)
{
	struct mmu_gather_batch *batch;
257

258
	for (batch = &tlb->local; batch && batch->nr; batch = batch->next) {
259 260 261 262 263 264
		free_pages_and_swap_cache(batch->pages, batch->nr);
		batch->nr = 0;
	}
	tlb->active = &tlb->local;
}

265 266 267 268 269 270
void tlb_flush_mmu(struct mmu_gather *tlb)
{
	tlb_flush_mmu_tlbonly(tlb);
	tlb_flush_mmu_free(tlb);
}

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
/* tlb_finish_mmu
 *	Called at the end of the shootdown operation to free up any resources
 *	that were required.
 */
void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
{
	struct mmu_gather_batch *batch, *next;

	tlb_flush_mmu(tlb);

	/* keep the page table cache within bounds */
	check_pgt_cache();

	for (batch = tlb->local.next; batch; batch = next) {
		next = batch->next;
		free_pages((unsigned long)batch, 0);
	}
	tlb->local.next = NULL;
}

/* __tlb_remove_page
 *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
 *	handling the additional races in SMP caused by other CPUs caching valid
 *	mappings in their TLBs. Returns the number of free page slots left.
 *	When out of page slots we must call tlb_flush_mmu().
296
 *returns true if the caller should flush.
297
 */
298
bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size)
299 300 301
{
	struct mmu_gather_batch *batch;

302
	VM_BUG_ON(!tlb->end);
303

304 305 306 307 308 309 310
	if (!tlb->page_size)
		tlb->page_size = page_size;
	else {
		if (page_size != tlb->page_size)
			return true;
	}

311 312 313
	batch = tlb->active;
	if (batch->nr == batch->max) {
		if (!tlb_next_batch(tlb))
314
			return true;
315
		batch = tlb->active;
316
	}
317
	VM_BUG_ON_PAGE(batch->nr > batch->max, page);
318

319 320
	batch->pages[batch->nr++] = page;
	return false;
321 322 323 324
}

#endif /* HAVE_GENERIC_MMU_GATHER */

325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
#ifdef CONFIG_HAVE_RCU_TABLE_FREE

/*
 * See the comment near struct mmu_table_batch.
 */

static void tlb_remove_table_smp_sync(void *arg)
{
	/* Simply deliver the interrupt */
}

static void tlb_remove_table_one(void *table)
{
	/*
	 * This isn't an RCU grace period and hence the page-tables cannot be
	 * assumed to be actually RCU-freed.
	 *
	 * It is however sufficient for software page-table walkers that rely on
	 * IRQ disabling. See the comment near struct mmu_table_batch.
	 */
	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
	__tlb_remove_table(table);
}

static void tlb_remove_table_rcu(struct rcu_head *head)
{
	struct mmu_table_batch *batch;
	int i;

	batch = container_of(head, struct mmu_table_batch, rcu);

	for (i = 0; i < batch->nr; i++)
		__tlb_remove_table(batch->tables[i]);

	free_page((unsigned long)batch);
}

void tlb_table_flush(struct mmu_gather *tlb)
{
	struct mmu_table_batch **batch = &tlb->batch;

	if (*batch) {
		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
		*batch = NULL;
	}
}

void tlb_remove_table(struct mmu_gather *tlb, void *table)
{
	struct mmu_table_batch **batch = &tlb->batch;

	/*
	 * When there's less then two users of this mm there cannot be a
	 * concurrent page-table walk.
	 */
	if (atomic_read(&tlb->mm->mm_users) < 2) {
		__tlb_remove_table(table);
		return;
	}

	if (*batch == NULL) {
		*batch = (struct mmu_table_batch *)__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
		if (*batch == NULL) {
			tlb_remove_table_one(table);
			return;
		}
		(*batch)->nr = 0;
	}
	(*batch)->tables[(*batch)->nr++] = table;
	if ((*batch)->nr == MAX_TABLE_BATCH)
		tlb_table_flush(tlb);
}

398
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
399

L
Linus Torvalds 已提交
400 401 402 403
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
404 405
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
406
{
407
	pgtable_t token = pmd_pgtable(*pmd);
408
	pmd_clear(pmd);
409
	pte_free_tlb(tlb, token, addr);
410
	atomic_long_dec(&tlb->mm->nr_ptes);
L
Linus Torvalds 已提交
411 412
}

413 414 415
static inline void free_pmd_range(struct mmu_gather *tlb, pud_t *pud,
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
416 417 418
{
	pmd_t *pmd;
	unsigned long next;
419
	unsigned long start;
L
Linus Torvalds 已提交
420

421
	start = addr;
L
Linus Torvalds 已提交
422 423 424 425 426
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
427
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
428 429
	} while (pmd++, addr = next, addr != end);

430 431 432 433 434 435 436
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
437
	}
438 439 440 441 442
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
443
	pmd_free_tlb(tlb, pmd, start);
444
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
445 446
}

447 448 449
static inline void free_pud_range(struct mmu_gather *tlb, pgd_t *pgd,
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
450 451 452
{
	pud_t *pud;
	unsigned long next;
453
	unsigned long start;
L
Linus Torvalds 已提交
454

455
	start = addr;
L
Linus Torvalds 已提交
456 457 458 459 460
	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
461
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
462 463
	} while (pud++, addr = next, addr != end);

464 465 466 467 468 469 470
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
471
	}
472 473 474 475 476
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(pgd, start);
	pgd_clear(pgd);
477
	pud_free_tlb(tlb, pud, start);
L
Linus Torvalds 已提交
478 479 480
}

/*
481
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
482
 */
483
void free_pgd_range(struct mmu_gather *tlb,
484 485
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
486 487 488
{
	pgd_t *pgd;
	unsigned long next;
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514

	/*
	 * The next few lines have given us lots of grief...
	 *
	 * Why are we testing PMD* at this top level?  Because often
	 * there will be no work to do at all, and we'd prefer not to
	 * go all the way down to the bottom just to discover that.
	 *
	 * Why all these "- 1"s?  Because 0 represents both the bottom
	 * of the address space and the top of it (using -1 for the
	 * top wouldn't help much: the masks would do the wrong thing).
	 * The rule is that addr 0 and floor 0 refer to the bottom of
	 * the address space, but end 0 and ceiling 0 refer to the top
	 * Comparisons need to use "end - 1" and "ceiling - 1" (though
	 * that end 0 case should be mythical).
	 *
	 * Wherever addr is brought up or ceiling brought down, we must
	 * be careful to reject "the opposite 0" before it confuses the
	 * subsequent tests.  But what about where end is brought down
	 * by PMD_SIZE below? no, end can't go down to 0 there.
	 *
	 * Whereas we round start (addr) and ceiling down, by different
	 * masks at different levels, in order to test whether a table
	 * now has no other vmas using it, so can be freed, we don't
	 * bother to round floor or end up - the tests don't need that.
	 */
L
Linus Torvalds 已提交
515

516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
	addr &= PMD_MASK;
	if (addr < floor) {
		addr += PMD_SIZE;
		if (!addr)
			return;
	}
	if (ceiling) {
		ceiling &= PMD_MASK;
		if (!ceiling)
			return;
	}
	if (end - 1 > ceiling - 1)
		end -= PMD_SIZE;
	if (addr > end - 1)
		return;
531 532 533 534 535
	/*
	 * We add page table cache pages with PAGE_SIZE,
	 * (see pte_free_tlb()), flush the tlb if we need
	 */
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
536
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
537 538 539 540
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
541
		free_pud_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
542
	} while (pgd++, addr = next, addr != end);
543 544
}

545
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
546
		unsigned long floor, unsigned long ceiling)
547 548 549 550 551
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

552
		/*
N
npiggin@suse.de 已提交
553 554
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
555
		 */
556
		unlink_anon_vmas(vma);
557 558
		unlink_file_vma(vma);

559
		if (is_vm_hugetlb_page(vma)) {
560
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
561
				floor, next? next->vm_start: ceiling);
562 563 564 565 566
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
567
			       && !is_vm_hugetlb_page(next)) {
568 569
				vma = next;
				next = vma->vm_next;
570
				unlink_anon_vmas(vma);
571
				unlink_file_vma(vma);
572 573 574 575
			}
			free_pgd_range(tlb, addr, vma->vm_end,
				floor, next? next->vm_start: ceiling);
		}
576 577
		vma = next;
	}
L
Linus Torvalds 已提交
578 579
}

580
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
581
{
582
	spinlock_t *ptl;
583
	pgtable_t new = pte_alloc_one(mm, address);
584 585 586
	if (!new)
		return -ENOMEM;

587 588 589 590 591 592 593 594 595 596 597 598 599 600 601
	/*
	 * Ensure all pte setup (eg. pte page lock and page clearing) are
	 * visible before the pte is made visible to other CPUs by being
	 * put into page tables.
	 *
	 * The other side of the story is the pointer chasing in the page
	 * table walking code (when walking the page table without locking;
	 * ie. most of the time). Fortunately, these data accesses consist
	 * of a chain of data-dependent loads, meaning most CPUs (alpha
	 * being the notable exception) will already guarantee loads are
	 * seen in-order. See the alpha page table accessors for the
	 * smp_read_barrier_depends() barriers in page table walking code.
	 */
	smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */

602
	ptl = pmd_lock(mm, pmd);
603
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
604
		atomic_long_inc(&mm->nr_ptes);
L
Linus Torvalds 已提交
605
		pmd_populate(mm, pmd, new);
606
		new = NULL;
607
	}
608
	spin_unlock(ptl);
609 610
	if (new)
		pte_free(mm, new);
611
	return 0;
L
Linus Torvalds 已提交
612 613
}

614
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
615
{
616 617 618 619
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

620 621
	smp_wmb(); /* See comment in __pte_alloc */

622
	spin_lock(&init_mm.page_table_lock);
623
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
624
		pmd_populate_kernel(&init_mm, pmd, new);
625
		new = NULL;
626
	}
627
	spin_unlock(&init_mm.page_table_lock);
628 629
	if (new)
		pte_free_kernel(&init_mm, new);
630
	return 0;
L
Linus Torvalds 已提交
631 632
}

K
KAMEZAWA Hiroyuki 已提交
633 634 635 636 637 638
static inline void init_rss_vec(int *rss)
{
	memset(rss, 0, sizeof(int) * NR_MM_COUNTERS);
}

static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
639
{
K
KAMEZAWA Hiroyuki 已提交
640 641
	int i;

642
	if (current->mm == mm)
643
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
644 645 646
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
647 648
}

N
Nick Piggin 已提交
649
/*
650 651 652
 * This function is called to print an error when a bad pte
 * is found. For example, we might have a PFN-mapped pte in
 * a region that doesn't allow it.
N
Nick Piggin 已提交
653 654 655
 *
 * The calling function must still handle the error.
 */
656 657
static void print_bad_pte(struct vm_area_struct *vma, unsigned long addr,
			  pte_t pte, struct page *page)
N
Nick Piggin 已提交
658
{
659 660 661 662 663
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
	pud_t *pud = pud_offset(pgd, addr);
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
664 665 666 667 668 669 670 671 672 673 674 675 676 677
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

	/*
	 * Allow a burst of 60 reports, then keep quiet for that minute;
	 * or allow a steady drip of one report per second.
	 */
	if (nr_shown == 60) {
		if (time_before(jiffies, resume)) {
			nr_unshown++;
			return;
		}
		if (nr_unshown) {
678 679
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
680 681 682 683 684 685
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
686 687 688 689

	mapping = vma->vm_file ? vma->vm_file->f_mapping : NULL;
	index = linear_page_index(vma, addr);

690 691 692
	pr_alert("BUG: Bad page map in process %s  pte:%08llx pmd:%08llx\n",
		 current->comm,
		 (long long)pte_val(pte), (long long)pmd_val(*pmd));
693
	if (page)
694
		dump_page(page, "bad pte");
695 696
	pr_alert("addr:%p vm_flags:%08lx anon_vma:%p mapping:%p index:%lx\n",
		 (void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
697 698 699
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
700 701 702 703 704
	pr_alert("file:%pD fault:%pf mmap:%pf readpage:%pf\n",
		 vma->vm_file,
		 vma->vm_ops ? vma->vm_ops->fault : NULL,
		 vma->vm_file ? vma->vm_file->f_op->mmap : NULL,
		 mapping ? mapping->a_ops->readpage : NULL);
N
Nick Piggin 已提交
705
	dump_stack();
706
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
707 708
}

H
Hugh Dickins 已提交
709
/*
N
Nick Piggin 已提交
710
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
711
 *
N
Nick Piggin 已提交
712 713 714
 * "Special" mappings do not wish to be associated with a "struct page" (either
 * it doesn't exist, or it exists but they don't want to touch it). In this
 * case, NULL is returned here. "Normal" mappings do have a struct page.
J
Jared Hulbert 已提交
715
 *
N
Nick Piggin 已提交
716 717 718 719 720 721 722 723
 * There are 2 broad cases. Firstly, an architecture may define a pte_special()
 * pte bit, in which case this function is trivial. Secondly, an architecture
 * may not have a spare pte bit, which requires a more complicated scheme,
 * described below.
 *
 * A raw VM_PFNMAP mapping (ie. one that is not COWed) is always considered a
 * special mapping (even if there are underlying and valid "struct pages").
 * COWed pages of a VM_PFNMAP are always normal.
724
 *
J
Jared Hulbert 已提交
725 726
 * The way we recognize COWed pages within VM_PFNMAP mappings is through the
 * rules set up by "remap_pfn_range()": the vma will have the VM_PFNMAP bit
N
Nick Piggin 已提交
727 728
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
729 730 731
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
732 733 734 735 736 737
 * And for normal mappings this is false.
 *
 * This restricts such mappings to be a linear translation from virtual address
 * to pfn. To get around this restriction, we allow arbitrary mappings so long
 * as the vma is not a COW mapping; in that case, we know that all ptes are
 * special (because none can have been COWed).
J
Jared Hulbert 已提交
738 739
 *
 *
N
Nick Piggin 已提交
740
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
741 742 743 744 745 746 747 748 749
 *
 * VM_MIXEDMAP mappings can likewise contain memory with or without "struct
 * page" backing, however the difference is that _all_ pages with a struct
 * page (that is, those where pfn_valid is true) are refcounted and considered
 * normal pages by the VM. The disadvantage is that pages are refcounted
 * (which can be slower and simply not an option for some PFNMAP users). The
 * advantage is that we don't have to follow the strict linearity rule of
 * PFNMAP mappings in order to support COWable mappings.
 *
H
Hugh Dickins 已提交
750
 */
N
Nick Piggin 已提交
751 752 753 754 755 756 757
#ifdef __HAVE_ARCH_PTE_SPECIAL
# define HAVE_PTE_SPECIAL 1
#else
# define HAVE_PTE_SPECIAL 0
#endif
struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
				pte_t pte)
H
Hugh Dickins 已提交
758
{
759
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
760 761

	if (HAVE_PTE_SPECIAL) {
762
		if (likely(!pte_special(pte)))
763
			goto check_pfn;
764 765
		if (vma->vm_ops && vma->vm_ops->find_special_page)
			return vma->vm_ops->find_special_page(vma, addr);
H
Hugh Dickins 已提交
766 767
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
H
Hugh Dickins 已提交
768
		if (!is_zero_pfn(pfn))
769
			print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
770 771 772 773 774
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
775 776 777 778 779 780
	if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
		if (vma->vm_flags & VM_MIXEDMAP) {
			if (!pfn_valid(pfn))
				return NULL;
			goto out;
		} else {
N
Nick Piggin 已提交
781 782
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
783 784 785 786 787
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
788 789
	}

790 791
	if (is_zero_pfn(pfn))
		return NULL;
792 793 794 795 796
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
797 798

	/*
N
Nick Piggin 已提交
799 800
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
801
	 */
J
Jared Hulbert 已提交
802
out:
803
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
804 805
}

806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
				pmd_t pmd)
{
	unsigned long pfn = pmd_pfn(pmd);

	/*
	 * There is no pmd_special() but there may be special pmds, e.g.
	 * in a direct-access (dax) mapping, so let's just replicate the
	 * !HAVE_PTE_SPECIAL case from vm_normal_page() here.
	 */
	if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
		if (vma->vm_flags & VM_MIXEDMAP) {
			if (!pfn_valid(pfn))
				return NULL;
			goto out;
		} else {
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
	}

	if (is_zero_pfn(pfn))
		return NULL;
	if (unlikely(pfn > highest_memmap_pfn))
		return NULL;

	/*
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
	 */
out:
	return pfn_to_page(pfn);
}
#endif

L
Linus Torvalds 已提交
846 847 848 849 850 851
/*
 * copy one vm_area from one task to the other. Assumes the page tables
 * already present in the new task to be cleared in the whole range
 * covered by this vma.
 */

H
Hugh Dickins 已提交
852
static inline unsigned long
L
Linus Torvalds 已提交
853
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
854
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
855
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
856
{
N
Nick Piggin 已提交
857
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
858 859 860 861 862
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880
		swp_entry_t entry = pte_to_swp_entry(pte);

		if (likely(!non_swap_entry(entry))) {
			if (swap_duplicate(entry) < 0)
				return entry.val;

			/* make sure dst_mm is on swapoff's mmlist. */
			if (unlikely(list_empty(&dst_mm->mmlist))) {
				spin_lock(&mmlist_lock);
				if (list_empty(&dst_mm->mmlist))
					list_add(&dst_mm->mmlist,
							&src_mm->mmlist);
				spin_unlock(&mmlist_lock);
			}
			rss[MM_SWAPENTS]++;
		} else if (is_migration_entry(entry)) {
			page = migration_entry_to_page(entry);

881
			rss[mm_counter(page)]++;
882 883 884 885 886 887 888 889 890 891 892 893

			if (is_write_migration_entry(entry) &&
					is_cow_mapping(vm_flags)) {
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
				make_migration_entry_read(&entry);
				pte = swp_entry_to_pte(entry);
				if (pte_swp_soft_dirty(*src_pte))
					pte = pte_swp_mksoft_dirty(pte);
				set_pte_at(src_mm, addr, src_pte, pte);
894
			}
L
Linus Torvalds 已提交
895
		}
896
		goto out_set_pte;
L
Linus Torvalds 已提交
897 898 899 900 901 902
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
903
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
904
		ptep_set_wrprotect(src_mm, addr, src_pte);
905
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
906 907 908 909 910 911 912 913 914
	}

	/*
	 * If it's a shared mapping, mark it clean in
	 * the child
	 */
	if (vm_flags & VM_SHARED)
		pte = pte_mkclean(pte);
	pte = pte_mkold(pte);
915 916 917 918

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
919
		page_dup_rmap(page, false);
920
		rss[mm_counter(page)]++;
921
	}
922 923 924

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
925
	return 0;
L
Linus Torvalds 已提交
926 927
}

928
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
929 930
		   pmd_t *dst_pmd, pmd_t *src_pmd, struct vm_area_struct *vma,
		   unsigned long addr, unsigned long end)
L
Linus Torvalds 已提交
931
{
932
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
933
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
934
	spinlock_t *src_ptl, *dst_ptl;
935
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
936
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
937
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
938 939

again:
K
KAMEZAWA Hiroyuki 已提交
940 941
	init_rss_vec(rss);

H
Hugh Dickins 已提交
942
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
943 944
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
945
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
946
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
947
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
948 949
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
950
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
951 952 953 954 955 956

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
957 958 959
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
960
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
961 962
				break;
		}
L
Linus Torvalds 已提交
963 964 965 966
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
967 968 969 970
		entry.val = copy_one_pte(dst_mm, src_mm, dst_pte, src_pte,
							vma, addr, rss);
		if (entry.val)
			break;
L
Linus Torvalds 已提交
971 972 973
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

974
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
975
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
976
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
977
	add_mm_rss_vec(dst_mm, rss);
978
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
979
	cond_resched();
H
Hugh Dickins 已提交
980 981 982 983 984 985

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
	if (addr != end)
		goto again;
	return 0;
}

static inline int copy_pmd_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		pud_t *dst_pud, pud_t *src_pud, struct vm_area_struct *vma,
		unsigned long addr, unsigned long end)
{
	pmd_t *src_pmd, *dst_pmd;
	unsigned long next;

	dst_pmd = pmd_alloc(dst_mm, dst_pud, addr);
	if (!dst_pmd)
		return -ENOMEM;
	src_pmd = pmd_offset(src_pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1004
		if (pmd_trans_huge(*src_pmd) || pmd_devmap(*src_pmd)) {
1005
			int err;
1006
			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
1007 1008 1009 1010 1011 1012 1013 1014
			err = copy_huge_pmd(dst_mm, src_mm,
					    dst_pmd, src_pmd, addr, vma);
			if (err == -ENOMEM)
				return -ENOMEM;
			if (!err)
				continue;
			/* fall through */
		}
L
Linus Torvalds 已提交
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
		if (pmd_none_or_clear_bad(src_pmd))
			continue;
		if (copy_pte_range(dst_mm, src_mm, dst_pmd, src_pmd,
						vma, addr, next))
			return -ENOMEM;
	} while (dst_pmd++, src_pmd++, addr = next, addr != end);
	return 0;
}

static inline int copy_pud_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		pgd_t *dst_pgd, pgd_t *src_pgd, struct vm_area_struct *vma,
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

	dst_pud = pud_alloc(dst_mm, dst_pgd, addr);
	if (!dst_pud)
		return -ENOMEM;
	src_pud = pud_offset(src_pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(src_pud))
			continue;
		if (copy_pmd_range(dst_mm, src_mm, dst_pud, src_pud,
						vma, addr, next))
			return -ENOMEM;
	} while (dst_pud++, src_pud++, addr = next, addr != end);
	return 0;
}

int copy_page_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		struct vm_area_struct *vma)
{
	pgd_t *src_pgd, *dst_pgd;
	unsigned long next;
	unsigned long addr = vma->vm_start;
	unsigned long end = vma->vm_end;
1053 1054 1055
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1056
	int ret;
L
Linus Torvalds 已提交
1057

1058 1059 1060 1061 1062 1063
	/*
	 * Don't copy ptes where a page fault will fill them correctly.
	 * Fork becomes much lighter when there are big shared or private
	 * readonly mappings. The tradeoff is that copy_page_range is more
	 * efficient than faulting.
	 */
1064 1065 1066
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1067

L
Linus Torvalds 已提交
1068 1069 1070
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1071
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1072 1073 1074 1075
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1076
		ret = track_pfn_copy(vma);
1077 1078 1079 1080
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1081 1082 1083 1084 1085 1086
	/*
	 * We need to invalidate the secondary MMU mappings only when
	 * there could be a permission downgrade on the ptes of the
	 * parent mm. And a permission downgrade will only happen if
	 * is_cow_mapping() returns true.
	 */
1087 1088 1089 1090 1091 1092
	is_cow = is_cow_mapping(vma->vm_flags);
	mmun_start = addr;
	mmun_end   = end;
	if (is_cow)
		mmu_notifier_invalidate_range_start(src_mm, mmun_start,
						    mmun_end);
A
Andrea Arcangeli 已提交
1093 1094

	ret = 0;
L
Linus Torvalds 已提交
1095 1096 1097 1098 1099 1100
	dst_pgd = pgd_offset(dst_mm, addr);
	src_pgd = pgd_offset(src_mm, addr);
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(src_pgd))
			continue;
A
Andrea Arcangeli 已提交
1101 1102 1103 1104 1105
		if (unlikely(copy_pud_range(dst_mm, src_mm, dst_pgd, src_pgd,
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1106
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1107

1108 1109
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1110
	return ret;
L
Linus Torvalds 已提交
1111 1112
}

1113
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1114
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1115
				unsigned long addr, unsigned long end,
1116
				struct zap_details *details)
L
Linus Torvalds 已提交
1117
{
N
Nick Piggin 已提交
1118
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1119
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1120
	int rss[NR_MM_COUNTERS];
1121
	spinlock_t *ptl;
1122
	pte_t *start_pte;
1123
	pte_t *pte;
1124
	swp_entry_t entry;
1125
	struct page *pending_page = NULL;
K
KAMEZAWA Hiroyuki 已提交
1126

1127
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1128
again:
1129
	init_rss_vec(rss);
1130 1131
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1132
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1133 1134
	do {
		pte_t ptent = *pte;
1135
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1136
			continue;
1137
		}
1138

L
Linus Torvalds 已提交
1139
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1140
			struct page *page;
1141

1142
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1143 1144 1145 1146 1147 1148 1149
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1150
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1151 1152
					continue;
			}
N
Nick Piggin 已提交
1153
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1154
							tlb->fullmm);
L
Linus Torvalds 已提交
1155 1156 1157
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1158 1159

			if (!PageAnon(page)) {
1160
				if (pte_dirty(ptent)) {
M
Michal Hocko 已提交
1161 1162 1163 1164 1165 1166
					/*
					 * oom_reaper cannot tear down dirty
					 * pages
					 */
					if (unlikely(details && details->ignore_dirty))
						continue;
1167
					force_flush = 1;
1168
					set_page_dirty(page);
1169
				}
1170
				if (pte_young(ptent) &&
1171
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1172
					mark_page_accessed(page);
1173
			}
1174
			rss[mm_counter(page)]--;
1175
			page_remove_rmap(page, false);
1176 1177
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1178
			if (unlikely(__tlb_remove_page(tlb, page))) {
1179
				force_flush = 1;
1180
				pending_page = page;
1181
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1182
				break;
1183
			}
L
Linus Torvalds 已提交
1184 1185
			continue;
		}
M
Michal Hocko 已提交
1186 1187
		/* only check swap_entries if explicitly asked for in details */
		if (unlikely(details && !details->check_swap_entries))
L
Linus Torvalds 已提交
1188
			continue;
K
KAMEZAWA Hiroyuki 已提交
1189

1190 1191 1192 1193 1194
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1195

1196
			page = migration_entry_to_page(entry);
1197
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1198
		}
1199 1200
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1201
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1202
	} while (pte++, addr += PAGE_SIZE, addr != end);
1203

K
KAMEZAWA Hiroyuki 已提交
1204
	add_mm_rss_vec(mm, rss);
1205
	arch_leave_lazy_mmu_mode();
1206

1207
	/* Do the actual TLB flush before dropping ptl */
1208
	if (force_flush)
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
		tlb_flush_mmu_tlbonly(tlb);
	pte_unmap_unlock(start_pte, ptl);

	/*
	 * If we forced a TLB flush (either due to running out of
	 * batch buffers or because we needed to flush dirty TLB
	 * entries before releasing the ptl), free the batched
	 * memory too. Restart if we didn't do everything.
	 */
	if (force_flush) {
		force_flush = 0;
		tlb_flush_mmu_free(tlb);
1221 1222 1223 1224 1225
		if (pending_page) {
			/* remove the page with new size */
			__tlb_remove_pte_page(tlb, pending_page);
			pending_page = NULL;
		}
1226
		if (addr != end)
P
Peter Zijlstra 已提交
1227 1228 1229
			goto again;
	}

1230
	return addr;
L
Linus Torvalds 已提交
1231 1232
}

1233
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1234
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1235
				unsigned long addr, unsigned long end,
1236
				struct zap_details *details)
L
Linus Torvalds 已提交
1237 1238 1239 1240 1241 1242 1243
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1244
		if (pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1245
			if (next - addr != HPAGE_PMD_SIZE) {
1246 1247
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1248
				__split_huge_pmd(vma, pmd, addr, false, NULL);
S
Shaohua Li 已提交
1249
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1250
				goto next;
1251 1252
			/* fall through */
		}
1253 1254 1255 1256 1257 1258 1259 1260 1261
		/*
		 * Here there can be other concurrent MADV_DONTNEED or
		 * trans huge page faults running, and if the pmd is
		 * none or trans huge it can change under us. This is
		 * because MADV_DONTNEED holds the mmap_sem in read
		 * mode.
		 */
		if (pmd_none_or_trans_huge_or_clear_bad(pmd))
			goto next;
1262
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1263
next:
1264 1265
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1266 1267

	return addr;
L
Linus Torvalds 已提交
1268 1269
}

1270
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1271
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1272
				unsigned long addr, unsigned long end,
1273
				struct zap_details *details)
L
Linus Torvalds 已提交
1274 1275 1276 1277 1278 1279 1280
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1281
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1282
			continue;
1283 1284
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
	} while (pud++, addr = next, addr != end);
1285 1286

	return addr;
L
Linus Torvalds 已提交
1287 1288
}

M
Michal Hocko 已提交
1289
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1290 1291 1292
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1293 1294 1295 1296 1297 1298 1299 1300 1301
{
	pgd_t *pgd;
	unsigned long next;

	BUG_ON(addr >= end);
	tlb_start_vma(tlb, vma);
	pgd = pgd_offset(vma->vm_mm, addr);
	do {
		next = pgd_addr_end(addr, end);
1302
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1303
			continue;
1304 1305
		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1306 1307
	tlb_end_vma(tlb, vma);
}
1308

1309 1310 1311

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1312
		unsigned long end_addr,
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323
		struct zap_details *details)
{
	unsigned long start = max(vma->vm_start, start_addr);
	unsigned long end;

	if (start >= vma->vm_end)
		return;
	end = min(vma->vm_end, end_addr);
	if (end <= vma->vm_start)
		return;

1324 1325 1326
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1327
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1328
		untrack_pfn(vma, 0, 0);
1329 1330 1331 1332 1333 1334 1335

	if (start != end) {
		if (unlikely(is_vm_hugetlb_page(vma))) {
			/*
			 * It is undesirable to test vma->vm_file as it
			 * should be non-null for valid hugetlb area.
			 * However, vm_file will be NULL in the error
1336
			 * cleanup path of mmap_region. When
1337
			 * hugetlbfs ->mmap method fails,
1338
			 * mmap_region() nullifies vma->vm_file
1339 1340 1341 1342
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1343
			if (vma->vm_file) {
1344
				i_mmap_lock_write(vma->vm_file->f_mapping);
1345
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1346
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1347
			}
1348 1349 1350
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1351 1352 1353 1354
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1355
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1356 1357 1358 1359
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1360
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
 *
 * Only addresses between `start' and `end' will be unmapped.
 *
 * The VMA list must be sorted in ascending virtual address order.
 *
 * unmap_vmas() assumes that the caller will flush the whole unmapped address
 * range after unmap_vmas() returns.  So the only responsibility here is to
 * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
 * drops the lock and schedules.
 */
A
Al Viro 已提交
1371
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1372
		struct vm_area_struct *vma, unsigned long start_addr,
1373
		unsigned long end_addr)
L
Linus Torvalds 已提交
1374
{
A
Andrea Arcangeli 已提交
1375
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1376

A
Andrea Arcangeli 已提交
1377
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1378
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1379
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1380
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1386
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1387
 * @size: number of bytes to zap
1388
 * @details: details of shared cache invalidation
1389 1390
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1391
 */
1392
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
L
Linus Torvalds 已提交
1393 1394 1395
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1396
	struct mmu_gather tlb;
1397
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1398 1399

	lru_add_drain();
1400
	tlb_gather_mmu(&tlb, mm, start, end);
1401
	update_hiwater_rss(mm);
1402 1403
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1404
		unmap_single_vma(&tlb, vma, start, end, details);
1405 1406
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1407 1408
}

1409 1410 1411 1412 1413
/**
 * zap_page_range_single - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
 * @address: starting address of pages to zap
 * @size: number of bytes to zap
1414
 * @details: details of shared cache invalidation
1415 1416
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1417
 */
1418
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1419 1420 1421
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1422
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1423 1424 1425
	unsigned long end = address + size;

	lru_add_drain();
1426
	tlb_gather_mmu(&tlb, mm, address, end);
1427
	update_hiwater_rss(mm);
1428
	mmu_notifier_invalidate_range_start(mm, address, end);
1429
	unmap_single_vma(&tlb, vma, address, end, details);
1430
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1431
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1432 1433
}

1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
/**
 * zap_vma_ptes - remove ptes mapping the vma
 * @vma: vm_area_struct holding ptes to be zapped
 * @address: starting address of pages to zap
 * @size: number of bytes to zap
 *
 * This function only unmaps ptes assigned to VM_PFNMAP vmas.
 *
 * The entire address range must be fully contained within the vma.
 *
 * Returns 0 if successful.
 */
int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
		unsigned long size)
{
	if (address < vma->vm_start || address + size > vma->vm_end ||
	    		!(vma->vm_flags & VM_PFNMAP))
		return -1;
1452
	zap_page_range_single(vma, address, size, NULL);
1453 1454 1455 1456
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1457
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1458
			spinlock_t **ptl)
1459 1460 1461 1462
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
1463
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
1464 1465
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
1466
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
1467
		}
1468 1469 1470 1471
	}
	return NULL;
}

1472 1473 1474 1475 1476 1477 1478
/*
 * This is the old fallback for page remapping.
 *
 * For historical reasons, it only allows reserved pages. Only
 * old drivers should use this, and they needed to mark their
 * pages reserved for the old functions anyway.
 */
N
Nick Piggin 已提交
1479 1480
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1481
{
N
Nick Piggin 已提交
1482
	struct mm_struct *mm = vma->vm_mm;
1483
	int retval;
1484
	pte_t *pte;
1485 1486
	spinlock_t *ptl;

1487
	retval = -EINVAL;
1488
	if (PageAnon(page))
1489
		goto out;
1490 1491
	retval = -ENOMEM;
	flush_dcache_page(page);
1492
	pte = get_locked_pte(mm, addr, &ptl);
1493
	if (!pte)
1494
		goto out;
1495 1496 1497 1498 1499 1500
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1501
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1502
	page_add_file_rmap(page, false);
1503 1504 1505
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1506 1507
	pte_unmap_unlock(pte, ptl);
	return retval;
1508 1509 1510 1511 1512 1513
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1514 1515 1516 1517 1518 1519
/**
 * vm_insert_page - insert single page into user vma
 * @vma: user vma to map to
 * @addr: target user address of this page
 * @page: source kernel page
 *
1520 1521 1522 1523 1524 1525
 * This allows drivers to insert individual pages they've allocated
 * into a user vma.
 *
 * The page has to be a nice clean _individual_ kernel allocation.
 * If you allocate a compound page, you need to have marked it as
 * such (__GFP_COMP), or manually just split the page up yourself
N
Nick Piggin 已提交
1526
 * (see split_page()).
1527 1528 1529 1530 1531 1532 1533 1534
 *
 * NOTE! Traditionally this was done with "remap_pfn_range()" which
 * took an arbitrary page protection parameter. This doesn't allow
 * that. Your vma protection will have to be set up correctly, which
 * means that if you want a shared writable mapping, you'd better
 * ask for a shared writable mapping!
 *
 * The page does not need to be reserved.
1535 1536 1537 1538 1539
 *
 * Usually this function is called from f_op->mmap() handler
 * under mm->mmap_sem write-lock, so it can change vma->vm_flags.
 * Caller must set VM_MIXEDMAP on vma if it wants to call this
 * function from other places, for example from page-fault handler.
1540
 */
N
Nick Piggin 已提交
1541 1542
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1543 1544 1545 1546 1547
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1548 1549 1550 1551 1552
	if (!(vma->vm_flags & VM_MIXEDMAP)) {
		BUG_ON(down_read_trylock(&vma->vm_mm->mmap_sem));
		BUG_ON(vma->vm_flags & VM_PFNMAP);
		vma->vm_flags |= VM_MIXEDMAP;
	}
N
Nick Piggin 已提交
1553
	return insert_page(vma, addr, page, vma->vm_page_prot);
1554
}
1555
EXPORT_SYMBOL(vm_insert_page);
1556

N
Nick Piggin 已提交
1557
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1558
			pfn_t pfn, pgprot_t prot)
N
Nick Piggin 已提交
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
{
	struct mm_struct *mm = vma->vm_mm;
	int retval;
	pte_t *pte, entry;
	spinlock_t *ptl;

	retval = -ENOMEM;
	pte = get_locked_pte(mm, addr, &ptl);
	if (!pte)
		goto out;
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
1574 1575 1576 1577
	if (pfn_t_devmap(pfn))
		entry = pte_mkdevmap(pfn_t_pte(pfn, prot));
	else
		entry = pte_mkspecial(pfn_t_pte(pfn, prot));
N
Nick Piggin 已提交
1578
	set_pte_at(mm, addr, pte, entry);
1579
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1580 1581 1582 1583 1584 1585 1586 1587

	retval = 0;
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

N
Nick Piggin 已提交
1588 1589 1590 1591 1592 1593
/**
 * vm_insert_pfn - insert single pfn into user vma
 * @vma: user vma to map to
 * @addr: target user address of this page
 * @pfn: source kernel pfn
 *
1594
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1595 1596 1597 1598
 * they've allocated into a user vma. Same comments apply.
 *
 * This function should only be called from a vm_ops->fault handler, and
 * in that case the handler should return NULL.
N
Nick Piggin 已提交
1599 1600 1601 1602 1603
 *
 * vma cannot be a COW mapping.
 *
 * As this is called only for pages that do not currently exist, we
 * do not need to flush old virtual caches or the TLB.
N
Nick Piggin 已提交
1604 1605
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1606
			unsigned long pfn)
A
Andy Lutomirski 已提交
1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
{
	return vm_insert_pfn_prot(vma, addr, pfn, vma->vm_page_prot);
}
EXPORT_SYMBOL(vm_insert_pfn);

/**
 * vm_insert_pfn_prot - insert single pfn into user vma with specified pgprot
 * @vma: user vma to map to
 * @addr: target user address of this page
 * @pfn: source kernel pfn
 * @pgprot: pgprot flags for the inserted page
 *
 * This is exactly like vm_insert_pfn, except that it allows drivers to
 * to override pgprot on a per-page basis.
 *
 * This only makes sense for IO mappings, and it makes no sense for
 * cow mappings.  In general, using multiple vmas is preferable;
 * vm_insert_pfn_prot should only be used if using multiple VMAs is
 * impractical.
 */
int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn, pgprot_t pgprot)
N
Nick Piggin 已提交
1629
{
1630
	int ret;
N
Nick Piggin 已提交
1631 1632 1633 1634 1635 1636
	/*
	 * Technically, architectures with pte_special can avoid all these
	 * restrictions (same for remap_pfn_range).  However we would like
	 * consistency in testing and feature parity among all, so we should
	 * try to keep these invariants in place for everybody.
	 */
J
Jared Hulbert 已提交
1637 1638 1639 1640 1641
	BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
	BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) ==
						(VM_PFNMAP|VM_MIXEDMAP));
	BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
	BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn));
N
Nick Piggin 已提交
1642

N
Nick Piggin 已提交
1643 1644
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1645 1646

	track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
1647

1648
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot);
1649 1650

	return ret;
N
Nick Piggin 已提交
1651
}
A
Andy Lutomirski 已提交
1652
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1653

N
Nick Piggin 已提交
1654
int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1655
			pfn_t pfn)
N
Nick Piggin 已提交
1656
{
1657 1658
	pgprot_t pgprot = vma->vm_page_prot;

N
Nick Piggin 已提交
1659
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1660

N
Nick Piggin 已提交
1661 1662
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1663 1664

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1665

N
Nick Piggin 已提交
1666 1667 1668 1669
	/*
	 * If we don't have pte special, then we have to use the pfn_valid()
	 * based VM_MIXEDMAP scheme (see vm_normal_page), and thus we *must*
	 * refcount the page if pfn_valid is true (hence insert_page rather
H
Hugh Dickins 已提交
1670 1671
	 * than insert_pfn).  If a zero_pfn were inserted into a VM_MIXEDMAP
	 * without pte special, it would there be refcounted as a normal page.
N
Nick Piggin 已提交
1672
	 */
1673
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1674 1675
		struct page *page;

1676 1677 1678 1679 1680 1681
		/*
		 * At this point we are committed to insert_page()
		 * regardless of whether the caller specified flags that
		 * result in pfn_t_has_page() == false.
		 */
		page = pfn_to_page(pfn_t_to_pfn(pfn));
1682
		return insert_page(vma, addr, page, pgprot);
N
Nick Piggin 已提交
1683
	}
1684
	return insert_pfn(vma, addr, pfn, pgprot);
N
Nick Piggin 已提交
1685
}
N
Nick Piggin 已提交
1686
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1687

L
Linus Torvalds 已提交
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
/*
 * maps a range of physical memory into the requested pages. the old
 * mappings are removed. any references to nonexistent pages results
 * in null mappings (currently treated as "copy-on-access")
 */
static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pte_t *pte;
H
Hugh Dickins 已提交
1698
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1699

H
Hugh Dickins 已提交
1700
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1701 1702
	if (!pte)
		return -ENOMEM;
1703
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1704 1705
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1706
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1707 1708
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1709
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1710
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
	return 0;
}

static inline int remap_pmd_range(struct mm_struct *mm, pud_t *pud,
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pmd_t *pmd;
	unsigned long next;

	pfn -= addr >> PAGE_SHIFT;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
1725
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754
	do {
		next = pmd_addr_end(addr, end);
		if (remap_pte_range(mm, pmd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot))
			return -ENOMEM;
	} while (pmd++, addr = next, addr != end);
	return 0;
}

static inline int remap_pud_range(struct mm_struct *mm, pgd_t *pgd,
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;

	pfn -= addr >> PAGE_SHIFT;
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
		if (remap_pmd_range(mm, pud, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot))
			return -ENOMEM;
	} while (pud++, addr = next, addr != end);
	return 0;
}

1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
/**
 * remap_pfn_range - remap kernel memory to userspace
 * @vma: user vma to map to
 * @addr: target user address to start at
 * @pfn: physical address of kernel memory
 * @size: size of map area
 * @prot: page protection flags for this mapping
 *
 *  Note: this is only safe if the mm semaphore is held when called.
 */
L
Linus Torvalds 已提交
1765 1766 1767 1768 1769
int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
		    unsigned long pfn, unsigned long size, pgprot_t prot)
{
	pgd_t *pgd;
	unsigned long next;
1770
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1771
	struct mm_struct *mm = vma->vm_mm;
1772
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1773 1774 1775 1776 1777 1778 1779
	int err;

	/*
	 * Physically remapped pages are special. Tell the
	 * rest of the world about it:
	 *   VM_IO tells people not to look at these pages
	 *	(accesses can have side effects).
1780 1781 1782
	 *   VM_PFNMAP tells the core MM that the base pages are just
	 *	raw PFN mappings, and do not have a "struct page" associated
	 *	with them.
1783 1784 1785 1786
	 *   VM_DONTEXPAND
	 *      Disable vma merging and expanding with mremap().
	 *   VM_DONTDUMP
	 *      Omit vma from core dump, even when VM_IO turned off.
L
Linus Torvalds 已提交
1787 1788 1789 1790
	 *
	 * There's a horrible special case to handle copy-on-write
	 * behaviour that some programs depend on. We mark the "original"
	 * un-COW'ed pages by matching them up with "vma->vm_pgoff".
1791
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1792
	 */
1793 1794 1795
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1796
		vma->vm_pgoff = pfn;
1797 1798
	}

1799
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1800
	if (err)
1801
		return -EINVAL;
L
Linus Torvalds 已提交
1802

1803
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815

	BUG_ON(addr >= end);
	pfn -= addr >> PAGE_SHIFT;
	pgd = pgd_offset(mm, addr);
	flush_cache_range(vma, addr, end);
	do {
		next = pgd_addr_end(addr, end);
		err = remap_pud_range(mm, pgd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1816 1817

	if (err)
1818
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1819

L
Linus Torvalds 已提交
1820 1821 1822 1823
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
/**
 * vm_iomap_memory - remap memory to userspace
 * @vma: user vma to map to
 * @start: start of area
 * @len: size of area
 *
 * This is a simplified io_remap_pfn_range() for common driver use. The
 * driver just needs to give us the physical memory range to be mapped,
 * we'll figure out the rest from the vma information.
 *
 * NOTE! Some drivers might want to tweak vma->vm_page_prot first to get
 * whatever write-combining details or similar.
 */
int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
{
	unsigned long vm_len, pfn, pages;

	/* Check that the physical memory area passed in looks valid */
	if (start + len < start)
		return -EINVAL;
	/*
	 * You *really* shouldn't map things that aren't page-aligned,
	 * but we've historically allowed it because IO memory might
	 * just have smaller alignment.
	 */
	len += start & ~PAGE_MASK;
	pfn = start >> PAGE_SHIFT;
	pages = (len + ~PAGE_MASK) >> PAGE_SHIFT;
	if (pfn + pages < pfn)
		return -EINVAL;

	/* We start the mapping 'vm_pgoff' pages into the area */
	if (vma->vm_pgoff > pages)
		return -EINVAL;
	pfn += vma->vm_pgoff;
	pages -= vma->vm_pgoff;

	/* Can we fit all of the mapping? */
	vm_len = vma->vm_end - vma->vm_start;
	if (vm_len >> PAGE_SHIFT > pages)
		return -EINVAL;

	/* Ok, let it rip */
	return io_remap_pfn_range(vma, vma->vm_start, pfn, vm_len, vma->vm_page_prot);
}
EXPORT_SYMBOL(vm_iomap_memory);

1871 1872 1873 1874 1875 1876
static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pte_t *pte;
	int err;
1877
	pgtable_t token;
1878
	spinlock_t *uninitialized_var(ptl);
1879 1880 1881 1882 1883 1884 1885 1886 1887

	pte = (mm == &init_mm) ?
		pte_alloc_kernel(pmd, addr) :
		pte_alloc_map_lock(mm, pmd, addr, &ptl);
	if (!pte)
		return -ENOMEM;

	BUG_ON(pmd_huge(*pmd));

1888 1889
	arch_enter_lazy_mmu_mode();

1890
	token = pmd_pgtable(*pmd);
1891 1892

	do {
1893
		err = fn(pte++, token, addr, data);
1894 1895
		if (err)
			break;
1896
	} while (addr += PAGE_SIZE, addr != end);
1897

1898 1899
	arch_leave_lazy_mmu_mode();

1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
	if (mm != &init_mm)
		pte_unmap_unlock(pte-1, ptl);
	return err;
}

static int apply_to_pmd_range(struct mm_struct *mm, pud_t *pud,
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pmd_t *pmd;
	unsigned long next;
	int err;

A
Andi Kleen 已提交
1913 1914
	BUG_ON(pud_huge(*pud));

1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
	do {
		next = pmd_addr_end(addr, end);
		err = apply_to_pte_range(mm, pmd, addr, next, fn, data);
		if (err)
			break;
	} while (pmd++, addr = next, addr != end);
	return err;
}

static int apply_to_pud_range(struct mm_struct *mm, pgd_t *pgd,
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
		err = apply_to_pmd_range(mm, pud, addr, next, fn, data);
		if (err)
			break;
	} while (pud++, addr = next, addr != end);
	return err;
}

/*
 * Scan a region of virtual memory, filling in page tables as necessary
 * and calling a provided function on each leaf page table.
 */
int apply_to_page_range(struct mm_struct *mm, unsigned long addr,
			unsigned long size, pte_fn_t fn, void *data)
{
	pgd_t *pgd;
	unsigned long next;
1956
	unsigned long end = addr + size;
1957 1958
	int err;

1959 1960 1961
	if (WARN_ON(addr >= end))
		return -EINVAL;

1962 1963 1964 1965 1966 1967 1968
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
		err = apply_to_pud_range(mm, pgd, addr, next, fn, data);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1969

1970 1971 1972 1973
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

1974
/*
1975 1976 1977 1978 1979
 * handle_pte_fault chooses page fault handler according to an entry which was
 * read non-atomically.  Before making any commitment, on those architectures
 * or configurations (e.g. i386 with PAE) which might give a mix of unmatched
 * parts, do_swap_page must check under lock before unmapping the pte and
 * proceeding (but do_wp_page is only called after already making such a check;
1980
 * and do_anonymous_page can safely check later on).
1981
 */
H
Hugh Dickins 已提交
1982
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
1983 1984 1985 1986 1987
				pte_t *page_table, pte_t orig_pte)
{
	int same = 1;
#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
	if (sizeof(pte_t) > sizeof(unsigned long)) {
H
Hugh Dickins 已提交
1988 1989
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
1990
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
1991
		spin_unlock(ptl);
1992 1993 1994 1995 1996 1997
	}
#endif
	pte_unmap(page_table);
	return same;
}

1998
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
1999
{
2000 2001
	debug_dma_assert_idle(src);

2002 2003 2004 2005 2006 2007 2008
	/*
	 * If the source page was a PFN mapping, we don't have
	 * a "struct page" for it. We do a best-effort copy by
	 * just copying from the original user address. If that
	 * fails, we just zero-fill it. Live with it.
	 */
	if (unlikely(!src)) {
2009
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2010 2011 2012 2013 2014 2015 2016 2017 2018
		void __user *uaddr = (void __user *)(va & PAGE_MASK);

		/*
		 * This really shouldn't fail, because the page is there
		 * in the page tables. But it might just be unreadable,
		 * in which case we just give up and fill the result with
		 * zeroes.
		 */
		if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE))
2019
			clear_page(kaddr);
2020
		kunmap_atomic(kaddr);
2021
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2022 2023
	} else
		copy_user_highpage(dst, src, va, vma);
2024 2025
}

2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039
static gfp_t __get_fault_gfp_mask(struct vm_area_struct *vma)
{
	struct file *vm_file = vma->vm_file;

	if (vm_file)
		return mapping_gfp_mask(vm_file->f_mapping) | __GFP_FS | __GFP_IO;

	/*
	 * Special mappings (e.g. VDSO) do not have any file so fake
	 * a default GFP_KERNEL for them.
	 */
	return GFP_KERNEL;
}

2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054
/*
 * Notify the address space that the page is about to become writable so that
 * it can prohibit this or wait for the page to get into an appropriate state.
 *
 * We do this without the lock held, so that it can sleep if it needs to.
 */
static int do_page_mkwrite(struct vm_area_struct *vma, struct page *page,
	       unsigned long address)
{
	struct vm_fault vmf;
	int ret;

	vmf.virtual_address = (void __user *)(address & PAGE_MASK);
	vmf.pgoff = page->index;
	vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2055
	vmf.gfp_mask = __get_fault_gfp_mask(vma);
2056
	vmf.page = page;
2057
	vmf.cow_page = NULL;
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073

	ret = vma->vm_ops->page_mkwrite(vma, &vmf);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))
		return ret;
	if (unlikely(!(ret & VM_FAULT_LOCKED))) {
		lock_page(page);
		if (!page->mapping) {
			unlock_page(page);
			return 0; /* retry */
		}
		ret |= VM_FAULT_LOCKED;
	} else
		VM_BUG_ON_PAGE(!PageLocked(page), page);
	return ret;
}

2074 2075 2076 2077 2078 2079 2080 2081
/*
 * Handle write page faults for pages that can be reused in the current vma
 *
 * This can happen either due to the mapping being with the VM_SHARED flag,
 * or due to us being the last reference standing to the page. In either
 * case, all we need to do here is to mark the page as writable and update
 * any related book-keeping.
 */
K
Kirill A. Shutemov 已提交
2082 2083 2084
static inline int wp_page_reuse(struct fault_env *fe, pte_t orig_pte,
			struct page *page, int page_mkwrite, int dirty_shared)
	__releases(fe->ptl)
2085
{
K
Kirill A. Shutemov 已提交
2086
	struct vm_area_struct *vma = fe->vma;
2087 2088 2089 2090 2091 2092 2093 2094 2095
	pte_t entry;
	/*
	 * Clear the pages cpupid information as the existing
	 * information potentially belongs to a now completely
	 * unrelated process.
	 */
	if (page)
		page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);

K
Kirill A. Shutemov 已提交
2096
	flush_cache_page(vma, fe->address, pte_pfn(orig_pte));
2097 2098
	entry = pte_mkyoung(orig_pte);
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
2099 2100 2101
	if (ptep_set_access_flags(vma, fe->address, fe->pte, entry, 1))
		update_mmu_cache(vma, fe->address, fe->pte);
	pte_unmap_unlock(fe->pte, fe->ptl);
2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113

	if (dirty_shared) {
		struct address_space *mapping;
		int dirtied;

		if (!page_mkwrite)
			lock_page(page);

		dirtied = set_page_dirty(page);
		VM_BUG_ON_PAGE(PageAnon(page), page);
		mapping = page->mapping;
		unlock_page(page);
2114
		put_page(page);
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130

		if ((dirtied || page_mkwrite) && mapping) {
			/*
			 * Some device drivers do not set page.mapping
			 * but still dirty their pages
			 */
			balance_dirty_pages_ratelimited(mapping);
		}

		if (!page_mkwrite)
			file_update_time(vma->vm_file);
	}

	return VM_FAULT_WRITE;
}

2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146
/*
 * Handle the case of a page which we actually need to copy to a new page.
 *
 * Called with mmap_sem locked and the old page referenced, but
 * without the ptl held.
 *
 * High level logic flow:
 *
 * - Allocate a page, copy the content of the old page to the new one.
 * - Handle book keeping and accounting - cgroups, mmu-notifiers, etc.
 * - Take the PTL. If the pte changed, bail out and release the allocated page
 * - If the pte is still the way we remember it, update the page table and all
 *   relevant references. This includes dropping the reference the page-table
 *   held to the old page, as well as updating the rmap.
 * - In any case, unlock the PTL and drop the reference we took to the old page.
 */
K
Kirill A. Shutemov 已提交
2147 2148
static int wp_page_copy(struct fault_env *fe, pte_t orig_pte,
		struct page *old_page)
2149
{
K
Kirill A. Shutemov 已提交
2150 2151
	struct vm_area_struct *vma = fe->vma;
	struct mm_struct *mm = vma->vm_mm;
2152 2153 2154
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
K
Kirill A. Shutemov 已提交
2155 2156
	const unsigned long mmun_start = fe->address & PAGE_MASK;
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2157 2158 2159 2160 2161 2162
	struct mem_cgroup *memcg;

	if (unlikely(anon_vma_prepare(vma)))
		goto oom;

	if (is_zero_pfn(pte_pfn(orig_pte))) {
K
Kirill A. Shutemov 已提交
2163
		new_page = alloc_zeroed_user_highpage_movable(vma, fe->address);
2164 2165 2166
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2167 2168
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
				fe->address);
2169 2170
		if (!new_page)
			goto oom;
K
Kirill A. Shutemov 已提交
2171
		cow_user_page(new_page, old_page, fe->address, vma);
2172 2173
	}

2174
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2175 2176
		goto oom_free_new;

2177 2178
	__SetPageUptodate(new_page);

2179 2180 2181 2182 2183
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
K
Kirill A. Shutemov 已提交
2184 2185
	fe->pte = pte_offset_map_lock(mm, fe->pmd, fe->address, &fe->ptl);
	if (likely(pte_same(*fe->pte, orig_pte))) {
2186 2187
		if (old_page) {
			if (!PageAnon(old_page)) {
2188 2189
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2190 2191 2192 2193 2194
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
K
Kirill A. Shutemov 已提交
2195
		flush_cache_page(vma, fe->address, pte_pfn(orig_pte));
2196 2197 2198 2199 2200 2201 2202 2203
		entry = mk_pte(new_page, vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
		/*
		 * Clear the pte entry and flush it first, before updating the
		 * pte with the new entry. This will avoid a race condition
		 * seen in the presence of one thread doing SMC and another
		 * thread doing COW.
		 */
K
Kirill A. Shutemov 已提交
2204 2205
		ptep_clear_flush_notify(vma, fe->address, fe->pte);
		page_add_new_anon_rmap(new_page, vma, fe->address, false);
2206
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2207 2208 2209 2210 2211 2212
		lru_cache_add_active_or_unevictable(new_page, vma);
		/*
		 * We call the notify macro here because, when using secondary
		 * mmu page tables (such as kvm shadow page tables), we want the
		 * new page to be mapped directly into the secondary page table.
		 */
K
Kirill A. Shutemov 已提交
2213 2214
		set_pte_at_notify(mm, fe->address, fe->pte, entry);
		update_mmu_cache(vma, fe->address, fe->pte);
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237
		if (old_page) {
			/*
			 * Only after switching the pte to the new page may
			 * we remove the mapcount here. Otherwise another
			 * process may come and find the rmap count decremented
			 * before the pte is switched to the new page, and
			 * "reuse" the old page writing into it while our pte
			 * here still points into it and can be read by other
			 * threads.
			 *
			 * The critical issue is to order this
			 * page_remove_rmap with the ptp_clear_flush above.
			 * Those stores are ordered by (if nothing else,)
			 * the barrier present in the atomic_add_negative
			 * in page_remove_rmap.
			 *
			 * Then the TLB flush in ptep_clear_flush ensures that
			 * no process can access the old page before the
			 * decremented mapcount is visible. And the old page
			 * cannot be reused until after the decremented
			 * mapcount is visible. So transitively, TLBs to
			 * old page will be flushed before it can be reused.
			 */
2238
			page_remove_rmap(old_page, false);
2239 2240 2241 2242 2243 2244
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2245
		mem_cgroup_cancel_charge(new_page, memcg, false);
2246 2247 2248
	}

	if (new_page)
2249
		put_page(new_page);
2250

K
Kirill A. Shutemov 已提交
2251
	pte_unmap_unlock(fe->pte, fe->ptl);
2252 2253 2254 2255 2256 2257 2258 2259
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
	if (old_page) {
		/*
		 * Don't let another task, with possibly unlocked vma,
		 * keep the mlocked page.
		 */
		if (page_copied && (vma->vm_flags & VM_LOCKED)) {
			lock_page(old_page);	/* LRU manipulation */
2260 2261
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2262 2263
			unlock_page(old_page);
		}
2264
		put_page(old_page);
2265 2266 2267
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2268
	put_page(new_page);
2269 2270
oom:
	if (old_page)
2271
		put_page(old_page);
2272 2273 2274
	return VM_FAULT_OOM;
}

2275 2276 2277 2278
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
K
Kirill A. Shutemov 已提交
2279
static int wp_pfn_shared(struct fault_env *fe,  pte_t orig_pte)
2280
{
K
Kirill A. Shutemov 已提交
2281 2282
	struct vm_area_struct *vma = fe->vma;

2283 2284 2285
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
		struct vm_fault vmf = {
			.page = NULL,
K
Kirill A. Shutemov 已提交
2286 2287 2288
			.pgoff = linear_page_index(vma, fe->address),
			.virtual_address =
				(void __user *)(fe->address & PAGE_MASK),
2289 2290 2291 2292
			.flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE,
		};
		int ret;

K
Kirill A. Shutemov 已提交
2293
		pte_unmap_unlock(fe->pte, fe->ptl);
2294 2295 2296
		ret = vma->vm_ops->pfn_mkwrite(vma, &vmf);
		if (ret & VM_FAULT_ERROR)
			return ret;
K
Kirill A. Shutemov 已提交
2297 2298
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
				&fe->ptl);
2299 2300 2301 2302
		/*
		 * We might have raced with another page fault while we
		 * released the pte_offset_map_lock.
		 */
K
Kirill A. Shutemov 已提交
2303 2304
		if (!pte_same(*fe->pte, orig_pte)) {
			pte_unmap_unlock(fe->pte, fe->ptl);
2305 2306 2307
			return 0;
		}
	}
K
Kirill A. Shutemov 已提交
2308
	return wp_page_reuse(fe, orig_pte, NULL, 0, 0);
2309 2310
}

K
Kirill A. Shutemov 已提交
2311 2312 2313
static int wp_page_shared(struct fault_env *fe, pte_t orig_pte,
		struct page *old_page)
	__releases(fe->ptl)
2314
{
K
Kirill A. Shutemov 已提交
2315
	struct vm_area_struct *vma = fe->vma;
2316 2317
	int page_mkwrite = 0;

2318
	get_page(old_page);
2319 2320 2321 2322

	if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
		int tmp;

K
Kirill A. Shutemov 已提交
2323 2324
		pte_unmap_unlock(fe->pte, fe->ptl);
		tmp = do_page_mkwrite(vma, old_page, fe->address);
2325 2326
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
2327
			put_page(old_page);
2328 2329 2330 2331 2332 2333 2334 2335
			return tmp;
		}
		/*
		 * Since we dropped the lock we need to revalidate
		 * the PTE as someone else may have changed it.  If
		 * they did, we just return, as we can count on the
		 * MMU to tell us if they didn't also make it writable.
		 */
K
Kirill A. Shutemov 已提交
2336 2337 2338
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
						 &fe->ptl);
		if (!pte_same(*fe->pte, orig_pte)) {
2339
			unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2340
			pte_unmap_unlock(fe->pte, fe->ptl);
2341
			put_page(old_page);
2342 2343 2344 2345 2346
			return 0;
		}
		page_mkwrite = 1;
	}

K
Kirill A. Shutemov 已提交
2347
	return wp_page_reuse(fe, orig_pte, old_page, page_mkwrite, 1);
2348 2349
}

L
Linus Torvalds 已提交
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
/*
 * This routine handles present pages, when users try to write
 * to a shared page. It is done by copying the page to a new address
 * and decrementing the shared-page counter for the old page.
 *
 * Note that this routine assumes that the protection checks have been
 * done by the caller (the low-level page fault routine in most cases).
 * Thus we can safely just mark it writable once we've done any necessary
 * COW.
 *
 * We also mark the page dirty at this point even though the page will
 * change only once the write actually happens. This avoids a few races,
 * and potentially makes it more efficient.
 *
2364 2365 2366
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), with pte both mapped and locked.
 * We return with mmap_sem still held, but pte unmapped and unlocked.
L
Linus Torvalds 已提交
2367
 */
K
Kirill A. Shutemov 已提交
2368 2369
static int do_wp_page(struct fault_env *fe, pte_t orig_pte)
	__releases(fe->ptl)
L
Linus Torvalds 已提交
2370
{
K
Kirill A. Shutemov 已提交
2371
	struct vm_area_struct *vma = fe->vma;
2372
	struct page *old_page;
L
Linus Torvalds 已提交
2373

K
Kirill A. Shutemov 已提交
2374
	old_page = vm_normal_page(vma, fe->address, orig_pte);
2375 2376
	if (!old_page) {
		/*
2377 2378
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2379 2380
		 *
		 * We should not cow pages in a shared writeable mapping.
2381
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2382 2383 2384
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
K
Kirill A. Shutemov 已提交
2385
			return wp_pfn_shared(fe, orig_pte);
2386

K
Kirill A. Shutemov 已提交
2387 2388
		pte_unmap_unlock(fe->pte, fe->ptl);
		return wp_page_copy(fe, orig_pte, old_page);
2389
	}
L
Linus Torvalds 已提交
2390

2391
	/*
P
Peter Zijlstra 已提交
2392 2393
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2394
	 */
H
Hugh Dickins 已提交
2395
	if (PageAnon(old_page) && !PageKsm(old_page)) {
2396
		int total_mapcount;
2397
		if (!trylock_page(old_page)) {
2398
			get_page(old_page);
K
Kirill A. Shutemov 已提交
2399
			pte_unmap_unlock(fe->pte, fe->ptl);
2400
			lock_page(old_page);
K
Kirill A. Shutemov 已提交
2401 2402 2403
			fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd,
					fe->address, &fe->ptl);
			if (!pte_same(*fe->pte, orig_pte)) {
2404
				unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2405
				pte_unmap_unlock(fe->pte, fe->ptl);
2406
				put_page(old_page);
2407
				return 0;
2408
			}
2409
			put_page(old_page);
P
Peter Zijlstra 已提交
2410
		}
2411 2412 2413 2414 2415 2416 2417 2418 2419
		if (reuse_swap_page(old_page, &total_mapcount)) {
			if (total_mapcount == 1) {
				/*
				 * The page is all ours. Move it to
				 * our anon_vma so the rmap code will
				 * not search our parent or siblings.
				 * Protected against the rmap code by
				 * the page lock.
				 */
2420
				page_move_anon_rmap(old_page, vma);
2421
			}
2422
			unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2423
			return wp_page_reuse(fe, orig_pte, old_page, 0, 0);
2424
		}
2425
		unlock_page(old_page);
P
Peter Zijlstra 已提交
2426
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2427
					(VM_WRITE|VM_SHARED))) {
K
Kirill A. Shutemov 已提交
2428
		return wp_page_shared(fe, orig_pte, old_page);
L
Linus Torvalds 已提交
2429 2430 2431 2432 2433
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
2434
	get_page(old_page);
2435

K
Kirill A. Shutemov 已提交
2436 2437
	pte_unmap_unlock(fe->pte, fe->ptl);
	return wp_page_copy(fe, orig_pte, old_page);
L
Linus Torvalds 已提交
2438 2439
}

2440
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2441 2442 2443
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2444
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2445 2446
}

2447
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2448 2449 2450 2451 2452
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2453
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2454 2455 2456
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2457
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2458 2459 2460 2461 2462 2463 2464
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2465
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2466 2467
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2468
				details);
L
Linus Torvalds 已提交
2469 2470 2471 2472
	}
}

/**
2473 2474 2475 2476
 * unmap_mapping_range - unmap the portion of all mmaps in the specified
 * address_space corresponding to the specified page range in the underlying
 * file.
 *
M
Martin Waitz 已提交
2477
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2478 2479
 * @holebegin: byte in first page to unmap, relative to the start of
 * the underlying file.  This will be rounded down to a PAGE_SIZE
N
npiggin@suse.de 已提交
2480
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
 * must keep the partial page.  In contrast, we must get rid of
 * partial pages.
 * @holelen: size of prospective hole in bytes.  This will be rounded
 * up to a PAGE_SIZE boundary.  A holelen of zero truncates to the
 * end of the file.
 * @even_cows: 1 when truncating a file, unmap even private COWed pages;
 * but 0 when invalidating pagecache, don't throw away private data.
 */
void unmap_mapping_range(struct address_space *mapping,
		loff_t const holebegin, loff_t const holelen, int even_cows)
{
M
Michal Hocko 已提交
2492
	struct zap_details details = { };
L
Linus Torvalds 已提交
2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509
	pgoff_t hba = holebegin >> PAGE_SHIFT;
	pgoff_t hlen = (holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;

	/* Check for overflow. */
	if (sizeof(holelen) > sizeof(hlen)) {
		long long holeend =
			(holebegin + holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;
		if (holeend & ~(long long)ULONG_MAX)
			hlen = ULONG_MAX - hba + 1;
	}

	details.check_mapping = even_cows? NULL: mapping;
	details.first_index = hba;
	details.last_index = hba + hlen - 1;
	if (details.last_index < details.first_index)
		details.last_index = ULONG_MAX;

2510
	i_mmap_lock_write(mapping);
2511
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2512
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2513
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2514 2515 2516 2517
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2518 2519
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2520 2521 2522 2523
 * We return with pte unmapped and unlocked.
 *
 * We return with the mmap_sem locked or unlocked in the same cases
 * as does filemap_fault().
L
Linus Torvalds 已提交
2524
 */
K
Kirill A. Shutemov 已提交
2525
int do_swap_page(struct fault_env *fe, pte_t orig_pte)
L
Linus Torvalds 已提交
2526
{
K
Kirill A. Shutemov 已提交
2527
	struct vm_area_struct *vma = fe->vma;
2528
	struct page *page, *swapcache;
2529
	struct mem_cgroup *memcg;
2530
	swp_entry_t entry;
L
Linus Torvalds 已提交
2531
	pte_t pte;
2532
	int locked;
2533
	int exclusive = 0;
N
Nick Piggin 已提交
2534
	int ret = 0;
L
Linus Torvalds 已提交
2535

K
Kirill A. Shutemov 已提交
2536
	if (!pte_unmap_same(vma->vm_mm, fe->pmd, fe->pte, orig_pte))
2537
		goto out;
2538 2539

	entry = pte_to_swp_entry(orig_pte);
2540 2541
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
K
Kirill A. Shutemov 已提交
2542
			migration_entry_wait(vma->vm_mm, fe->pmd, fe->address);
2543 2544 2545
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
K
Kirill A. Shutemov 已提交
2546
			print_bad_pte(vma, fe->address, orig_pte, NULL);
H
Hugh Dickins 已提交
2547
			ret = VM_FAULT_SIGBUS;
2548
		}
2549 2550
		goto out;
	}
2551
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
2552 2553
	page = lookup_swap_cache(entry);
	if (!page) {
2554
		page = swapin_readahead(entry,
K
Kirill A. Shutemov 已提交
2555
					GFP_HIGHUSER_MOVABLE, vma, fe->address);
L
Linus Torvalds 已提交
2556 2557
		if (!page) {
			/*
2558 2559
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2560
			 */
K
Kirill A. Shutemov 已提交
2561 2562 2563
			fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd,
					fe->address, &fe->ptl);
			if (likely(pte_same(*fe->pte, orig_pte)))
L
Linus Torvalds 已提交
2564
				ret = VM_FAULT_OOM;
2565
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2566
			goto unlock;
L
Linus Torvalds 已提交
2567 2568 2569 2570
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2571
		count_vm_event(PGMAJFAULT);
K
Kirill A. Shutemov 已提交
2572
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
2573
	} else if (PageHWPoison(page)) {
2574 2575 2576 2577
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2578 2579
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2580
		swapcache = page;
2581
		goto out_release;
L
Linus Torvalds 已提交
2582 2583
	}

2584
	swapcache = page;
K
Kirill A. Shutemov 已提交
2585
	locked = lock_page_or_retry(page, vma->vm_mm, fe->flags);
R
Rik van Riel 已提交
2586

2587
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2588 2589 2590 2591
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2592

A
Andrea Arcangeli 已提交
2593
	/*
2594 2595 2596 2597
	 * Make sure try_to_free_swap or reuse_swap_page or swapoff did not
	 * release the swapcache from under us.  The page pin, and pte_same
	 * test below, are not enough to exclude that.  Even if it is still
	 * swapcache, we need to check that the page's swap has not changed.
A
Andrea Arcangeli 已提交
2598
	 */
2599
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2600 2601
		goto out_page;

K
Kirill A. Shutemov 已提交
2602
	page = ksm_might_need_to_copy(page, vma, fe->address);
2603 2604 2605 2606
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2607 2608
	}

K
Kirill A. Shutemov 已提交
2609 2610
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2611
		ret = VM_FAULT_OOM;
2612
		goto out_page;
2613 2614
	}

L
Linus Torvalds 已提交
2615
	/*
2616
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2617
	 */
K
Kirill A. Shutemov 已提交
2618 2619 2620
	fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
			&fe->ptl);
	if (unlikely(!pte_same(*fe->pte, orig_pte)))
2621 2622 2623 2624 2625
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2626 2627
	}

2628 2629 2630 2631 2632 2633 2634 2635 2636
	/*
	 * The page isn't present yet, go ahead with the fault.
	 *
	 * Be careful about the sequence of operations here.
	 * To get its accounting right, reuse_swap_page() must be called
	 * while the page is counted on swap but not yet in mapcount i.e.
	 * before page_add_anon_rmap() and swap_free(); try_to_free_swap()
	 * must be called after the swap_free(), or it will never succeed.
	 */
L
Linus Torvalds 已提交
2637

K
Kirill A. Shutemov 已提交
2638 2639
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2640
	pte = mk_pte(page, vma->vm_page_prot);
K
Kirill A. Shutemov 已提交
2641
	if ((fe->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2642
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
K
Kirill A. Shutemov 已提交
2643
		fe->flags &= ~FAULT_FLAG_WRITE;
2644
		ret |= VM_FAULT_WRITE;
2645
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2646 2647
	}
	flush_icache_page(vma, page);
2648 2649
	if (pte_swp_soft_dirty(orig_pte))
		pte = pte_mksoft_dirty(pte);
K
Kirill A. Shutemov 已提交
2650
	set_pte_at(vma->vm_mm, fe->address, fe->pte, pte);
2651
	if (page == swapcache) {
K
Kirill A. Shutemov 已提交
2652
		do_page_add_anon_rmap(page, vma, fe->address, exclusive);
2653
		mem_cgroup_commit_charge(page, memcg, true, false);
2654
		activate_page(page);
2655
	} else { /* ksm created a completely new copy */
K
Kirill A. Shutemov 已提交
2656
		page_add_new_anon_rmap(page, vma, fe->address, false);
2657
		mem_cgroup_commit_charge(page, memcg, false, false);
2658 2659
		lru_cache_add_active_or_unevictable(page, vma);
	}
L
Linus Torvalds 已提交
2660

2661
	swap_free(entry);
2662 2663
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2664
		try_to_free_swap(page);
2665
	unlock_page(page);
2666
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2667 2668 2669 2670 2671 2672 2673 2674 2675
		/*
		 * Hold the lock to avoid the swap entry to be reused
		 * until we take the PT lock for the pte_same() check
		 * (to avoid false positives from pte_same). For
		 * further safety release the lock after the swap_free
		 * so that the swap count won't change under a
		 * parallel locked swapcache.
		 */
		unlock_page(swapcache);
2676
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2677
	}
2678

K
Kirill A. Shutemov 已提交
2679 2680
	if (fe->flags & FAULT_FLAG_WRITE) {
		ret |= do_wp_page(fe, pte);
2681 2682
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2683 2684 2685 2686
		goto out;
	}

	/* No need to invalidate - it was non-present before */
K
Kirill A. Shutemov 已提交
2687
	update_mmu_cache(vma, fe->address, fe->pte);
2688
unlock:
K
Kirill A. Shutemov 已提交
2689
	pte_unmap_unlock(fe->pte, fe->ptl);
L
Linus Torvalds 已提交
2690 2691
out:
	return ret;
2692
out_nomap:
2693
	mem_cgroup_cancel_charge(page, memcg, false);
K
Kirill A. Shutemov 已提交
2694
	pte_unmap_unlock(fe->pte, fe->ptl);
2695
out_page:
2696
	unlock_page(page);
2697
out_release:
2698
	put_page(page);
2699
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2700
		unlock_page(swapcache);
2701
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2702
	}
2703
	return ret;
L
Linus Torvalds 已提交
2704 2705
}

2706
/*
2707 2708
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
2709 2710 2711 2712 2713 2714
 * doesn't hit another vma.
 */
static inline int check_stack_guard_page(struct vm_area_struct *vma, unsigned long address)
{
	address &= PAGE_MASK;
	if ((vma->vm_flags & VM_GROWSDOWN) && address == vma->vm_start) {
2715 2716 2717 2718 2719 2720 2721 2722 2723 2724
		struct vm_area_struct *prev = vma->vm_prev;

		/*
		 * Is there a mapping abutting this one below?
		 *
		 * That's only ok if it's the same stack mapping
		 * that has gotten split..
		 */
		if (prev && prev->vm_end == address)
			return prev->vm_flags & VM_GROWSDOWN ? 0 : -ENOMEM;
2725

2726
		return expand_downwards(vma, address - PAGE_SIZE);
2727
	}
2728 2729 2730 2731 2732 2733 2734
	if ((vma->vm_flags & VM_GROWSUP) && address + PAGE_SIZE == vma->vm_end) {
		struct vm_area_struct *next = vma->vm_next;

		/* As VM_GROWSDOWN but s/below/above/ */
		if (next && next->vm_start == address + PAGE_SIZE)
			return next->vm_flags & VM_GROWSUP ? 0 : -ENOMEM;

2735
		return expand_upwards(vma, address + PAGE_SIZE);
2736
	}
2737 2738 2739
	return 0;
}

L
Linus Torvalds 已提交
2740
/*
2741 2742 2743
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
 * We return with mmap_sem still held, but pte unmapped and unlocked.
L
Linus Torvalds 已提交
2744
 */
K
Kirill A. Shutemov 已提交
2745
static int do_anonymous_page(struct fault_env *fe)
L
Linus Torvalds 已提交
2746
{
K
Kirill A. Shutemov 已提交
2747
	struct vm_area_struct *vma = fe->vma;
2748
	struct mem_cgroup *memcg;
2749
	struct page *page;
L
Linus Torvalds 已提交
2750 2751
	pte_t entry;

2752 2753 2754 2755
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2756
	/* Check if we need to add a guard page to the stack */
K
Kirill A. Shutemov 已提交
2757
	if (check_stack_guard_page(vma, fe->address) < 0)
2758
		return VM_FAULT_SIGSEGV;
2759

2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
	/*
	 * 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.
	 *
	 * pte_alloc_map() is safe to use under down_write(mmap_sem) or when
	 * parallel threads are excluded by other means.
	 *
	 * Here we only have down_read(mmap_sem).
	 */
	if (pte_alloc(vma->vm_mm, fe->pmd, fe->address))
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
	if (unlikely(pmd_trans_unstable(fe->pmd)))
		return 0;

2777
	/* Use the zero-page for reads */
K
Kirill A. Shutemov 已提交
2778 2779 2780
	if (!(fe->flags & FAULT_FLAG_WRITE) &&
			!mm_forbids_zeropage(vma->vm_mm)) {
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(fe->address),
H
Hugh Dickins 已提交
2781
						vma->vm_page_prot));
K
Kirill A. Shutemov 已提交
2782 2783 2784
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
				&fe->ptl);
		if (!pte_none(*fe->pte))
H
Hugh Dickins 已提交
2785
			goto unlock;
2786 2787
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
K
Kirill A. Shutemov 已提交
2788 2789
			pte_unmap_unlock(fe->pte, fe->ptl);
			return handle_userfault(fe, VM_UFFD_MISSING);
2790
		}
H
Hugh Dickins 已提交
2791 2792 2793
		goto setpte;
	}

N
Nick Piggin 已提交
2794 2795 2796
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
K
Kirill A. Shutemov 已提交
2797
	page = alloc_zeroed_user_highpage_movable(vma, fe->address);
N
Nick Piggin 已提交
2798 2799
	if (!page)
		goto oom;
2800

K
Kirill A. Shutemov 已提交
2801
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
2802 2803
		goto oom_free_page;

2804 2805 2806 2807 2808
	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * preceeding stores to the page contents become visible before
	 * the set_pte_at() write.
	 */
N
Nick Piggin 已提交
2809
	__SetPageUptodate(page);
2810

N
Nick Piggin 已提交
2811
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2812 2813
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2814

K
Kirill A. Shutemov 已提交
2815 2816 2817
	fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
			&fe->ptl);
	if (!pte_none(*fe->pte))
N
Nick Piggin 已提交
2818
		goto release;
H
Hugh Dickins 已提交
2819

2820 2821
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
K
Kirill A. Shutemov 已提交
2822
		pte_unmap_unlock(fe->pte, fe->ptl);
2823
		mem_cgroup_cancel_charge(page, memcg, false);
2824
		put_page(page);
K
Kirill A. Shutemov 已提交
2825
		return handle_userfault(fe, VM_UFFD_MISSING);
2826 2827
	}

K
Kirill A. Shutemov 已提交
2828 2829
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	page_add_new_anon_rmap(page, vma, fe->address, false);
2830
	mem_cgroup_commit_charge(page, memcg, false, false);
2831
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
2832
setpte:
K
Kirill A. Shutemov 已提交
2833
	set_pte_at(vma->vm_mm, fe->address, fe->pte, entry);
L
Linus Torvalds 已提交
2834 2835

	/* No need to invalidate - it was non-present before */
K
Kirill A. Shutemov 已提交
2836
	update_mmu_cache(vma, fe->address, fe->pte);
2837
unlock:
K
Kirill A. Shutemov 已提交
2838
	pte_unmap_unlock(fe->pte, fe->ptl);
N
Nick Piggin 已提交
2839
	return 0;
2840
release:
2841
	mem_cgroup_cancel_charge(page, memcg, false);
2842
	put_page(page);
2843
	goto unlock;
2844
oom_free_page:
2845
	put_page(page);
2846
oom:
L
Linus Torvalds 已提交
2847 2848 2849
	return VM_FAULT_OOM;
}

2850 2851 2852 2853 2854
/*
 * The mmap_sem must have been held on entry, and may have been
 * released depending on flags and vma->vm_ops->fault() return value.
 * See filemap_fault() and __lock_page_retry().
 */
K
Kirill A. Shutemov 已提交
2855 2856
static int __do_fault(struct fault_env *fe, pgoff_t pgoff,
		struct page *cow_page, struct page **page, void **entry)
2857
{
K
Kirill A. Shutemov 已提交
2858
	struct vm_area_struct *vma = fe->vma;
2859 2860 2861
	struct vm_fault vmf;
	int ret;

K
Kirill A. Shutemov 已提交
2862
	vmf.virtual_address = (void __user *)(fe->address & PAGE_MASK);
2863
	vmf.pgoff = pgoff;
K
Kirill A. Shutemov 已提交
2864
	vmf.flags = fe->flags;
2865
	vmf.page = NULL;
2866
	vmf.gfp_mask = __get_fault_gfp_mask(vma);
2867
	vmf.cow_page = cow_page;
2868 2869 2870 2871

	ret = vma->vm_ops->fault(vma, &vmf);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;
2872 2873 2874 2875
	if (ret & VM_FAULT_DAX_LOCKED) {
		*entry = vmf.entry;
		return ret;
	}
2876 2877 2878 2879

	if (unlikely(PageHWPoison(vmf.page))) {
		if (ret & VM_FAULT_LOCKED)
			unlock_page(vmf.page);
2880
		put_page(vmf.page);
2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
		lock_page(vmf.page);
	else
		VM_BUG_ON_PAGE(!PageLocked(vmf.page), vmf.page);

	*page = vmf.page;
	return ret;
}

2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932
static int pte_alloc_one_map(struct fault_env *fe)
{
	struct vm_area_struct *vma = fe->vma;

	if (!pmd_none(*fe->pmd))
		goto map_pte;
	if (fe->prealloc_pte) {
		fe->ptl = pmd_lock(vma->vm_mm, fe->pmd);
		if (unlikely(!pmd_none(*fe->pmd))) {
			spin_unlock(fe->ptl);
			goto map_pte;
		}

		atomic_long_inc(&vma->vm_mm->nr_ptes);
		pmd_populate(vma->vm_mm, fe->pmd, fe->prealloc_pte);
		spin_unlock(fe->ptl);
		fe->prealloc_pte = 0;
	} else if (unlikely(pte_alloc(vma->vm_mm, fe->pmd, fe->address))) {
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
	 * pmd_trans_unstable() instead of pmd_trans_huge() to ensure the pmd
	 * didn't become pmd_trans_huge under us and then back to pmd_none, as
	 * a result of MADV_DONTNEED running immediately after a huge pmd fault
	 * in a different thread of this mm, in turn leading to a misleading
	 * pmd_trans_huge() retval.  All we have to ensure is that it is a
	 * regular pmd that we can walk with pte_offset_map() and we can do that
	 * through an atomic read in C, which is what pmd_trans_unstable()
	 * provides.
	 */
	if (pmd_trans_unstable(fe->pmd) || pmd_devmap(*fe->pmd))
		return VM_FAULT_NOPAGE;

	fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
			&fe->ptl);
	return 0;
}

2933
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980

#define HPAGE_CACHE_INDEX_MASK (HPAGE_PMD_NR - 1)
static inline bool transhuge_vma_suitable(struct vm_area_struct *vma,
		unsigned long haddr)
{
	if (((vma->vm_start >> PAGE_SHIFT) & HPAGE_CACHE_INDEX_MASK) !=
			(vma->vm_pgoff & HPAGE_CACHE_INDEX_MASK))
		return false;
	if (haddr < vma->vm_start || haddr + HPAGE_PMD_SIZE > vma->vm_end)
		return false;
	return true;
}

static int do_set_pmd(struct fault_env *fe, struct page *page)
{
	struct vm_area_struct *vma = fe->vma;
	bool write = fe->flags & FAULT_FLAG_WRITE;
	unsigned long haddr = fe->address & HPAGE_PMD_MASK;
	pmd_t entry;
	int i, ret;

	if (!transhuge_vma_suitable(vma, haddr))
		return VM_FAULT_FALLBACK;

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

	fe->ptl = pmd_lock(vma->vm_mm, fe->pmd);
	if (unlikely(!pmd_none(*fe->pmd)))
		goto out;

	for (i = 0; i < HPAGE_PMD_NR; i++)
		flush_icache_page(vma, page + i);

	entry = mk_huge_pmd(page, vma->vm_page_prot);
	if (write)
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);

	add_mm_counter(vma->vm_mm, MM_FILEPAGES, HPAGE_PMD_NR);
	page_add_file_rmap(page, true);

	set_pmd_at(vma->vm_mm, haddr, fe->pmd, entry);

	update_mmu_cache_pmd(vma, haddr, fe->pmd);

	/* fault is handled */
	ret = 0;
2981
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
out:
	spin_unlock(fe->ptl);
	return ret;
}
#else
static int do_set_pmd(struct fault_env *fe, struct page *page)
{
	BUILD_BUG();
	return 0;
}
#endif

2994
/**
2995 2996
 * alloc_set_pte - setup new PTE entry for given page and add reverse page
 * mapping. If needed, the fucntion allocates page table or use pre-allocated.
2997
 *
K
Kirill A. Shutemov 已提交
2998
 * @fe: fault environment
2999
 * @memcg: memcg to charge page (only for private mappings)
3000 3001
 * @page: page to map
 *
3002
 * Caller must take care of unlocking fe->ptl, if fe->pte is non-NULL on return.
3003 3004 3005 3006
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
3007 3008
int alloc_set_pte(struct fault_env *fe, struct mem_cgroup *memcg,
		struct page *page)
3009
{
K
Kirill A. Shutemov 已提交
3010 3011
	struct vm_area_struct *vma = fe->vma;
	bool write = fe->flags & FAULT_FLAG_WRITE;
3012
	pte_t entry;
K
Kirill A. Shutemov 已提交
3013 3014
	int ret;

3015 3016
	if (pmd_none(*fe->pmd) && PageTransCompound(page) &&
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3017 3018 3019 3020 3021 3022 3023
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

		ret = do_set_pmd(fe, page);
		if (ret != VM_FAULT_FALLBACK)
			return ret;
	}
3024

3025
	if (!fe->pte) {
K
Kirill A. Shutemov 已提交
3026
		ret = pte_alloc_one_map(fe);
3027 3028 3029 3030 3031 3032 3033 3034
		if (ret)
			return ret;
	}

	/* Re-check under ptl */
	if (unlikely(!pte_none(*fe->pte)))
		return VM_FAULT_NOPAGE;

3035 3036 3037 3038
	flush_icache_page(vma, page);
	entry = mk_pte(page, vma->vm_page_prot);
	if (write)
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3039 3040
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3041
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
K
Kirill A. Shutemov 已提交
3042
		page_add_new_anon_rmap(page, vma, fe->address, false);
3043 3044
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3045
	} else {
3046
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3047
		page_add_file_rmap(page, false);
3048
	}
K
Kirill A. Shutemov 已提交
3049
	set_pte_at(vma->vm_mm, fe->address, fe->pte, entry);
3050 3051

	/* no need to invalidate: a not-present page won't be cached */
K
Kirill A. Shutemov 已提交
3052
	update_mmu_cache(vma, fe->address, fe->pte);
3053 3054

	return 0;
3055 3056
}

3057 3058
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3059 3060 3061

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3062
{
3063
	*val = fault_around_bytes;
3064 3065 3066
	return 0;
}

3067 3068 3069 3070 3071
/*
 * fault_around_pages() and fault_around_mask() expects fault_around_bytes
 * rounded down to nearest page order. It's what do_fault_around() expects to
 * see.
 */
3072
static int fault_around_bytes_set(void *data, u64 val)
3073
{
3074
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3075
		return -EINVAL;
3076 3077 3078 3079
	if (val > PAGE_SIZE)
		fault_around_bytes = rounddown_pow_of_two(val);
	else
		fault_around_bytes = PAGE_SIZE; /* rounddown_pow_of_two(0) is undefined */
3080 3081
	return 0;
}
3082 3083
DEFINE_SIMPLE_ATTRIBUTE(fault_around_bytes_fops,
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3084 3085 3086 3087 3088

static int __init fault_around_debugfs(void)
{
	void *ret;

3089 3090
	ret = debugfs_create_file("fault_around_bytes", 0644, NULL, NULL,
			&fault_around_bytes_fops);
3091
	if (!ret)
3092
		pr_warn("Failed to create fault_around_bytes in debugfs");
3093 3094 3095 3096
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3097

3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
/*
 * do_fault_around() tries to map few pages around the fault address. The hope
 * is that the pages will be needed soon and this will lower the number of
 * faults to handle.
 *
 * It uses vm_ops->map_pages() to map the pages, which skips the page if it's
 * not ready to be mapped: not up-to-date, locked, etc.
 *
 * This function is called with the page table lock taken. In the split ptlock
 * case the page table lock only protects only those entries which belong to
 * the page table corresponding to the fault address.
 *
 * This function doesn't cross the VMA boundaries, in order to call map_pages()
 * only once.
 *
 * fault_around_pages() defines how many pages we'll try to map.
 * do_fault_around() expects it to return a power of two less than or equal to
 * PTRS_PER_PTE.
 *
 * The virtual address of the area that we map is naturally aligned to the
 * fault_around_pages() value (and therefore to page order).  This way it's
 * easier to guarantee that we don't cross page table boundaries.
 */
3121
static int do_fault_around(struct fault_env *fe, pgoff_t start_pgoff)
3122
{
3123
	unsigned long address = fe->address, nr_pages, mask;
K
Kirill A. Shutemov 已提交
3124
	pgoff_t end_pgoff;
3125
	int off, ret = 0;
3126

3127
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3128 3129
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

3130 3131
	fe->address = max(address & mask, fe->vma->vm_start);
	off = ((address - fe->address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
K
Kirill A. Shutemov 已提交
3132
	start_pgoff -= off;
3133 3134

	/*
K
Kirill A. Shutemov 已提交
3135 3136
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3137
	 */
K
Kirill A. Shutemov 已提交
3138
	end_pgoff = start_pgoff -
3139
		((fe->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3140
		PTRS_PER_PTE - 1;
K
Kirill A. Shutemov 已提交
3141 3142
	end_pgoff = min3(end_pgoff, vma_pages(fe->vma) + fe->vma->vm_pgoff - 1,
			start_pgoff + nr_pages - 1);
3143

3144 3145
	if (pmd_none(*fe->pmd)) {
		fe->prealloc_pte = pte_alloc_one(fe->vma->vm_mm, fe->address);
3146 3147
		if (!fe->prealloc_pte)
			goto out;
3148
		smp_wmb(); /* See comment in __pte_alloc() */
3149 3150
	}

K
Kirill A. Shutemov 已提交
3151
	fe->vma->vm_ops->map_pages(fe, start_pgoff, end_pgoff);
3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172

	/* preallocated pagetable is unused: free it */
	if (fe->prealloc_pte) {
		pte_free(fe->vma->vm_mm, fe->prealloc_pte);
		fe->prealloc_pte = 0;
	}
	/* Huge page is mapped? Page fault is solved */
	if (pmd_trans_huge(*fe->pmd)) {
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
	if (!fe->pte)
		goto out;

	/* check if the page fault is solved */
	fe->pte -= (fe->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*fe->pte))
		ret = VM_FAULT_NOPAGE;
	pte_unmap_unlock(fe->pte, fe->ptl);
K
Kirill A. Shutemov 已提交
3173 3174
out:
	fe->address = address;
3175 3176
	fe->pte = NULL;
	return ret;
3177 3178
}

3179
static int do_read_fault(struct fault_env *fe, pgoff_t pgoff)
3180
{
K
Kirill A. Shutemov 已提交
3181
	struct vm_area_struct *vma = fe->vma;
3182
	struct page *fault_page;
3183 3184 3185 3186 3187 3188 3189
	int ret = 0;

	/*
	 * Let's call ->map_pages() first and use ->fault() as fallback
	 * if page by the offset is not ready to be mapped (cold cache or
	 * something).
	 */
3190
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3191 3192 3193
		ret = do_fault_around(fe, pgoff);
		if (ret)
			return ret;
3194
	}
3195

K
Kirill A. Shutemov 已提交
3196
	ret = __do_fault(fe, pgoff, NULL, &fault_page, NULL);
3197 3198 3199
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3200 3201
	ret |= alloc_set_pte(fe, NULL, fault_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3202
		pte_unmap_unlock(fe->pte, fe->ptl);
3203
	unlock_page(fault_page);
3204 3205
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		put_page(fault_page);
3206 3207 3208
	return ret;
}

3209
static int do_cow_fault(struct fault_env *fe, pgoff_t pgoff)
3210
{
K
Kirill A. Shutemov 已提交
3211
	struct vm_area_struct *vma = fe->vma;
3212
	struct page *fault_page, *new_page;
3213
	void *fault_entry;
3214
	struct mem_cgroup *memcg;
3215 3216 3217 3218 3219
	int ret;

	if (unlikely(anon_vma_prepare(vma)))
		return VM_FAULT_OOM;

K
Kirill A. Shutemov 已提交
3220
	new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, fe->address);
3221 3222 3223
	if (!new_page)
		return VM_FAULT_OOM;

K
Kirill A. Shutemov 已提交
3224 3225
	if (mem_cgroup_try_charge(new_page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
3226
		put_page(new_page);
3227 3228 3229
		return VM_FAULT_OOM;
	}

K
Kirill A. Shutemov 已提交
3230
	ret = __do_fault(fe, pgoff, new_page, &fault_page, &fault_entry);
3231 3232 3233
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;

3234
	if (!(ret & VM_FAULT_DAX_LOCKED))
K
Kirill A. Shutemov 已提交
3235
		copy_user_highpage(new_page, fault_page, fe->address, vma);
3236 3237
	__SetPageUptodate(new_page);

3238 3239
	ret |= alloc_set_pte(fe, memcg, new_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3240
		pte_unmap_unlock(fe->pte, fe->ptl);
3241
	if (!(ret & VM_FAULT_DAX_LOCKED)) {
3242
		unlock_page(fault_page);
3243
		put_page(fault_page);
3244
	} else {
3245
		dax_unlock_mapping_entry(vma->vm_file->f_mapping, pgoff);
3246
	}
3247 3248
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3249 3250
	return ret;
uncharge_out:
3251
	mem_cgroup_cancel_charge(new_page, memcg, false);
3252
	put_page(new_page);
3253 3254 3255
	return ret;
}

3256
static int do_shared_fault(struct fault_env *fe, pgoff_t pgoff)
L
Linus Torvalds 已提交
3257
{
K
Kirill A. Shutemov 已提交
3258
	struct vm_area_struct *vma = fe->vma;
3259 3260 3261 3262
	struct page *fault_page;
	struct address_space *mapping;
	int dirtied = 0;
	int ret, tmp;
3263

K
Kirill A. Shutemov 已提交
3264
	ret = __do_fault(fe, pgoff, NULL, &fault_page, NULL);
3265
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3266
		return ret;
L
Linus Torvalds 已提交
3267 3268

	/*
3269 3270
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3271
	 */
3272 3273
	if (vma->vm_ops->page_mkwrite) {
		unlock_page(fault_page);
K
Kirill A. Shutemov 已提交
3274
		tmp = do_page_mkwrite(vma, fault_page, fe->address);
3275 3276
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
3277
			put_page(fault_page);
3278
			return tmp;
3279
		}
3280 3281
	}

3282 3283
	ret |= alloc_set_pte(fe, NULL, fault_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3284
		pte_unmap_unlock(fe->pte, fe->ptl);
3285 3286
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
3287
		unlock_page(fault_page);
3288
		put_page(fault_page);
3289
		return ret;
L
Linus Torvalds 已提交
3290
	}
N
Nick Piggin 已提交
3291

3292 3293
	if (set_page_dirty(fault_page))
		dirtied = 1;
3294 3295 3296 3297 3298 3299
	/*
	 * Take a local copy of the address_space - page.mapping may be zeroed
	 * by truncate after unlock_page().   The address_space itself remains
	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
	 * release semantics to prevent the compiler from undoing this copying.
	 */
3300
	mapping = page_rmapping(fault_page);
3301 3302 3303 3304 3305 3306 3307
	unlock_page(fault_page);
	if ((dirtied || vma->vm_ops->page_mkwrite) && mapping) {
		/*
		 * Some device drivers do not set page.mapping but still
		 * dirty their pages
		 */
		balance_dirty_pages_ratelimited(mapping);
3308
	}
3309

3310
	if (!vma->vm_ops->page_mkwrite)
3311
		file_update_time(vma->vm_file);
N
Nick Piggin 已提交
3312

3313
	return ret;
3314
}
3315

3316 3317 3318 3319 3320 3321
/*
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults).
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
 */
3322
static int do_fault(struct fault_env *fe)
3323
{
K
Kirill A. Shutemov 已提交
3324 3325
	struct vm_area_struct *vma = fe->vma;
	pgoff_t pgoff = linear_page_index(vma, fe->address);
3326

3327 3328 3329
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
		return VM_FAULT_SIGBUS;
K
Kirill A. Shutemov 已提交
3330
	if (!(fe->flags & FAULT_FLAG_WRITE))
3331
		return do_read_fault(fe, pgoff);
3332
	if (!(vma->vm_flags & VM_SHARED))
3333 3334
		return do_cow_fault(fe, pgoff);
	return do_shared_fault(fe, pgoff);
3335 3336
}

3337
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3338 3339
				unsigned long addr, int page_nid,
				int *flags)
3340 3341 3342 3343
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3344
	if (page_nid == numa_node_id()) {
3345
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3346 3347
		*flags |= TNF_FAULT_LOCAL;
	}
3348 3349 3350 3351

	return mpol_misplaced(page, vma, addr);
}

K
Kirill A. Shutemov 已提交
3352
static int do_numa_page(struct fault_env *fe, pte_t pte)
3353
{
K
Kirill A. Shutemov 已提交
3354
	struct vm_area_struct *vma = fe->vma;
3355
	struct page *page = NULL;
3356
	int page_nid = -1;
3357
	int last_cpupid;
3358
	int target_nid;
3359
	bool migrated = false;
3360
	bool was_writable = pte_write(pte);
3361
	int flags = 0;
3362 3363 3364 3365 3366 3367

	/*
	* The "pte" at this point cannot be used safely without
	* validation through pte_unmap_same(). It's of NUMA type but
	* the pfn may be screwed if the read is non atomic.
	*
3368 3369 3370
	* We can safely just do a "set_pte_at()", because the old
	* page table entry is not accessible, so there would be no
	* concurrent hardware modifications to the PTE.
3371
	*/
K
Kirill A. Shutemov 已提交
3372 3373 3374 3375
	fe->ptl = pte_lockptr(vma->vm_mm, fe->pmd);
	spin_lock(fe->ptl);
	if (unlikely(!pte_same(*fe->pte, pte))) {
		pte_unmap_unlock(fe->pte, fe->ptl);
3376 3377 3378
		goto out;
	}

3379 3380 3381
	/* Make it present again */
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3382 3383
	if (was_writable)
		pte = pte_mkwrite(pte);
K
Kirill A. Shutemov 已提交
3384 3385
	set_pte_at(vma->vm_mm, fe->address, fe->pte, pte);
	update_mmu_cache(vma, fe->address, fe->pte);
3386

K
Kirill A. Shutemov 已提交
3387
	page = vm_normal_page(vma, fe->address, pte);
3388
	if (!page) {
K
Kirill A. Shutemov 已提交
3389
		pte_unmap_unlock(fe->pte, fe->ptl);
3390 3391 3392
		return 0;
	}

3393 3394
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
K
Kirill A. Shutemov 已提交
3395
		pte_unmap_unlock(fe->pte, fe->ptl);
3396 3397 3398
		return 0;
	}

3399
	/*
3400 3401 3402 3403 3404 3405
	 * Avoid grouping on RO pages in general. RO pages shouldn't hurt as
	 * much anyway since they can be in shared cache state. This misses
	 * the case where a mapping is writable but the process never writes
	 * to it but pte_write gets cleared during protection updates and
	 * pte_dirty has unpredictable behaviour between PTE scan updates,
	 * background writeback, dirty balancing and application behaviour.
3406
	 */
3407
	if (!pte_write(pte))
3408 3409
		flags |= TNF_NO_GROUP;

3410 3411 3412 3413 3414 3415 3416
	/*
	 * Flag if the page is shared between multiple address spaces. This
	 * is later used when determining whether to group tasks together
	 */
	if (page_mapcount(page) > 1 && (vma->vm_flags & VM_SHARED))
		flags |= TNF_SHARED;

3417
	last_cpupid = page_cpupid_last(page);
3418
	page_nid = page_to_nid(page);
K
Kirill A. Shutemov 已提交
3419 3420 3421
	target_nid = numa_migrate_prep(page, vma, fe->address, page_nid,
			&flags);
	pte_unmap_unlock(fe->pte, fe->ptl);
3422 3423 3424 3425 3426 3427
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3428
	migrated = migrate_misplaced_page(page, vma, target_nid);
3429
	if (migrated) {
3430
		page_nid = target_nid;
3431
		flags |= TNF_MIGRATED;
3432 3433
	} else
		flags |= TNF_MIGRATE_FAIL;
3434 3435

out:
3436
	if (page_nid != -1)
3437
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3438 3439 3440
	return 0;
}

K
Kirill A. Shutemov 已提交
3441
static int create_huge_pmd(struct fault_env *fe)
M
Matthew Wilcox 已提交
3442
{
K
Kirill A. Shutemov 已提交
3443
	struct vm_area_struct *vma = fe->vma;
3444
	if (vma_is_anonymous(vma))
K
Kirill A. Shutemov 已提交
3445
		return do_huge_pmd_anonymous_page(fe);
M
Matthew Wilcox 已提交
3446
	if (vma->vm_ops->pmd_fault)
K
Kirill A. Shutemov 已提交
3447 3448
		return vma->vm_ops->pmd_fault(vma, fe->address, fe->pmd,
				fe->flags);
M
Matthew Wilcox 已提交
3449 3450 3451
	return VM_FAULT_FALLBACK;
}

K
Kirill A. Shutemov 已提交
3452
static int wp_huge_pmd(struct fault_env *fe, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3453
{
K
Kirill A. Shutemov 已提交
3454 3455 3456 3457 3458
	if (vma_is_anonymous(fe->vma))
		return do_huge_pmd_wp_page(fe, orig_pmd);
	if (fe->vma->vm_ops->pmd_fault)
		return fe->vma->vm_ops->pmd_fault(fe->vma, fe->address, fe->pmd,
				fe->flags);
K
Kirill A. Shutemov 已提交
3459 3460 3461

	/* COW handled on pte level: split pmd */
	VM_BUG_ON_VMA(fe->vma->vm_flags & VM_SHARED, fe->vma);
3462
	__split_huge_pmd(fe->vma, fe->pmd, fe->address, false, NULL);
K
Kirill A. Shutemov 已提交
3463

M
Matthew Wilcox 已提交
3464 3465 3466
	return VM_FAULT_FALLBACK;
}

3467 3468 3469 3470 3471
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

L
Linus Torvalds 已提交
3472 3473 3474 3475 3476 3477 3478 3479 3480
/*
 * These routines also need to handle stuff like marking pages dirty
 * and/or accessed for architectures that don't do it in hardware (most
 * RISC architectures).  The early dirtying is also good on the i386.
 *
 * There is also a hook called "update_mmu_cache()" that architectures
 * with external mmu caches can use to update those (ie the Sparc or
 * PowerPC hashed page tables that act as extended TLBs).
 *
3481 3482
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3483
 *
3484 3485
 * The mmap_sem may have been released depending on flags and our return value.
 * See filemap_fault() and __lock_page_or_retry().
L
Linus Torvalds 已提交
3486
 */
K
Kirill A. Shutemov 已提交
3487
static int handle_pte_fault(struct fault_env *fe)
L
Linus Torvalds 已提交
3488 3489 3490
{
	pte_t entry;

3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521
	if (unlikely(pmd_none(*fe->pmd))) {
		/*
		 * Leave __pte_alloc() until later: because vm_ops->fault may
		 * want to allocate huge page, and if we expose page table
		 * for an instant, it will be difficult to retract from
		 * concurrent faults and from rmap lookups.
		 */
		fe->pte = NULL;
	} else {
		/* See comment in pte_alloc_one_map() */
		if (pmd_trans_unstable(fe->pmd) || pmd_devmap(*fe->pmd))
			return 0;
		/*
		 * A regular pmd is established and it can't morph into a huge
		 * pmd from under us anymore at this point because we hold the
		 * mmap_sem read mode and khugepaged takes it in write mode.
		 * So now it's safe to run pte_offset_map().
		 */
		fe->pte = pte_offset_map(fe->pmd, fe->address);

		entry = *fe->pte;

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
		 * CONFIG_32BIT=y, so READ_ONCE or ACCESS_ONCE cannot guarantee
		 * atomic accesses.  The code below just needs a consistent
		 * view for the ifs and we later double check anyway with the
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
3522
		if (pte_none(entry)) {
3523 3524
			pte_unmap(fe->pte);
			fe->pte = NULL;
3525
		}
L
Linus Torvalds 已提交
3526 3527
	}

3528 3529 3530 3531 3532 3533 3534 3535 3536 3537
	if (!fe->pte) {
		if (vma_is_anonymous(fe->vma))
			return do_anonymous_page(fe);
		else
			return do_fault(fe);
	}

	if (!pte_present(entry))
		return do_swap_page(fe, entry);

3538
	if (pte_protnone(entry) && vma_is_accessible(fe->vma))
K
Kirill A. Shutemov 已提交
3539
		return do_numa_page(fe, entry);
3540

K
Kirill A. Shutemov 已提交
3541 3542 3543
	fe->ptl = pte_lockptr(fe->vma->vm_mm, fe->pmd);
	spin_lock(fe->ptl);
	if (unlikely(!pte_same(*fe->pte, entry)))
3544
		goto unlock;
K
Kirill A. Shutemov 已提交
3545
	if (fe->flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3546
		if (!pte_write(entry))
K
Kirill A. Shutemov 已提交
3547
			return do_wp_page(fe, entry);
L
Linus Torvalds 已提交
3548 3549 3550
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
K
Kirill A. Shutemov 已提交
3551 3552 3553
	if (ptep_set_access_flags(fe->vma, fe->address, fe->pte, entry,
				fe->flags & FAULT_FLAG_WRITE)) {
		update_mmu_cache(fe->vma, fe->address, fe->pte);
3554 3555 3556 3557 3558 3559 3560
	} else {
		/*
		 * This is needed only for protection faults but the arch code
		 * is not yet telling us if this is a protection fault or not.
		 * This still avoids useless tlb flushes for .text page faults
		 * with threads.
		 */
K
Kirill A. Shutemov 已提交
3561 3562
		if (fe->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(fe->vma, fe->address);
3563
	}
3564
unlock:
K
Kirill A. Shutemov 已提交
3565
	pte_unmap_unlock(fe->pte, fe->ptl);
N
Nick Piggin 已提交
3566
	return 0;
L
Linus Torvalds 已提交
3567 3568 3569 3570
}

/*
 * By the time we get here, we already hold the mm semaphore
3571 3572 3573
 *
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
L
Linus Torvalds 已提交
3574
 */
3575 3576
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3577
{
K
Kirill A. Shutemov 已提交
3578 3579 3580 3581 3582
	struct fault_env fe = {
		.vma = vma,
		.address = address,
		.flags = flags,
	};
3583
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3584 3585 3586 3587 3588 3589
	pgd_t *pgd;
	pud_t *pud;

	pgd = pgd_offset(mm, address);
	pud = pud_alloc(mm, pgd, address);
	if (!pud)
H
Hugh Dickins 已提交
3590
		return VM_FAULT_OOM;
K
Kirill A. Shutemov 已提交
3591 3592
	fe.pmd = pmd_alloc(mm, pud, address);
	if (!fe.pmd)
H
Hugh Dickins 已提交
3593
		return VM_FAULT_OOM;
K
Kirill A. Shutemov 已提交
3594 3595
	if (pmd_none(*fe.pmd) && transparent_hugepage_enabled(vma)) {
		int ret = create_huge_pmd(&fe);
3596 3597
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3598
	} else {
K
Kirill A. Shutemov 已提交
3599
		pmd_t orig_pmd = *fe.pmd;
3600 3601
		int ret;

3602
		barrier();
3603
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
3604
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
K
Kirill A. Shutemov 已提交
3605
				return do_huge_pmd_numa_page(&fe, orig_pmd);
3606

K
Kirill A. Shutemov 已提交
3607 3608 3609
			if ((fe.flags & FAULT_FLAG_WRITE) &&
					!pmd_write(orig_pmd)) {
				ret = wp_huge_pmd(&fe, orig_pmd);
3610 3611
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3612
			} else {
K
Kirill A. Shutemov 已提交
3613
				huge_pmd_set_accessed(&fe, orig_pmd);
3614
				return 0;
3615
			}
3616 3617 3618
		}
	}

K
Kirill A. Shutemov 已提交
3619
	return handle_pte_fault(&fe);
L
Linus Torvalds 已提交
3620 3621
}

3622 3623 3624 3625 3626 3627
/*
 * By the time we get here, we already hold the mm semaphore
 *
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
 */
3628 3629
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
3630 3631 3632 3633 3634 3635
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
3636
	mem_cgroup_count_vm_event(vma->vm_mm, PGFAULT);
3637 3638 3639 3640 3641 3642 3643 3644 3645

	/* do counter updates before entering really critical section. */
	check_sync_rss_stat(current);

	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
3646
		mem_cgroup_oom_enable();
3647

K
Kirill A. Shutemov 已提交
3648 3649 3650 3651 3652 3653 3654 3655 3656
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
3657

3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668
	if (flags & FAULT_FLAG_USER) {
		mem_cgroup_oom_disable();
                /*
                 * The task may have entered a memcg OOM situation but
                 * if the allocation error was handled gracefully (no
                 * VM_FAULT_OOM), there is no need to kill anything.
                 * Just clean up the OOM state peacefully.
                 */
                if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
                        mem_cgroup_oom_synchronize(false);
	}
3669

3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682
	/*
	 * This mm has been already reaped by the oom reaper and so the
	 * refault cannot be trusted in general. Anonymous refaults would
	 * lose data and give a zero page instead e.g. This is especially
	 * problem for use_mm() because regular tasks will just die and
	 * the corrupted data will not be visible anywhere while kthread
	 * will outlive the oom victim and potentially propagate the date
	 * further.
	 */
	if (unlikely((current->flags & PF_KTHREAD) && !(ret & VM_FAULT_ERROR)
				&& test_bit(MMF_UNSTABLE, &vma->vm_mm->flags)))
		ret = VM_FAULT_SIGBUS;

3683 3684
	return ret;
}
3685
EXPORT_SYMBOL_GPL(handle_mm_fault);
3686

L
Linus Torvalds 已提交
3687 3688 3689
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3690
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3691
 */
3692
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3693
{
H
Hugh Dickins 已提交
3694 3695
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3696
		return -ENOMEM;
L
Linus Torvalds 已提交
3697

3698 3699
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3700
	spin_lock(&mm->page_table_lock);
3701
	if (pgd_present(*pgd))		/* Another has populated it */
3702
		pud_free(mm, new);
3703 3704
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3705
	spin_unlock(&mm->page_table_lock);
3706
	return 0;
L
Linus Torvalds 已提交
3707 3708 3709 3710 3711 3712
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3713
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3714
 */
3715
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3716
{
H
Hugh Dickins 已提交
3717 3718
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3719
		return -ENOMEM;
L
Linus Torvalds 已提交
3720

3721 3722
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3723
	spin_lock(&mm->page_table_lock);
L
Linus Torvalds 已提交
3724
#ifndef __ARCH_HAS_4LEVEL_HACK
3725 3726
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
3727
		pud_populate(mm, pud, new);
3728
	} else	/* Another has populated it */
3729
		pmd_free(mm, new);
3730 3731 3732
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
3733
		pgd_populate(mm, pud, new);
3734 3735
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
3736
#endif /* __ARCH_HAS_4LEVEL_HACK */
H
Hugh Dickins 已提交
3737
	spin_unlock(&mm->page_table_lock);
3738
	return 0;
3739
}
L
Linus Torvalds 已提交
3740 3741
#endif /* __PAGETABLE_PMD_FOLDED */

3742
static int __follow_pte(struct mm_struct *mm, unsigned long address,
J
Johannes Weiner 已提交
3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758
		pte_t **ptepp, spinlock_t **ptlp)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

	pgd = pgd_offset(mm, address);
	if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
		goto out;

	pud = pud_offset(pgd, address);
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
3759
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779
	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
		goto out;

	/* We cannot handle huge page PFN maps. Luckily they don't exist. */
	if (pmd_huge(*pmd))
		goto out;

	ptep = pte_offset_map_lock(mm, pmd, address, ptlp);
	if (!ptep)
		goto out;
	if (!pte_present(*ptep))
		goto unlock;
	*ptepp = ptep;
	return 0;
unlock:
	pte_unmap_unlock(ptep, *ptlp);
out:
	return -EINVAL;
}

3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
			   !(res = __follow_pte(mm, address, ptepp, ptlp)));
	return res;
}

J
Johannes Weiner 已提交
3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819
/**
 * follow_pfn - look up PFN at a user virtual address
 * @vma: memory mapping
 * @address: user virtual address
 * @pfn: location to store found PFN
 *
 * Only IO mappings and raw PFN mappings are allowed.
 *
 * Returns zero and the pfn at @pfn on success, -ve otherwise.
 */
int follow_pfn(struct vm_area_struct *vma, unsigned long address,
	unsigned long *pfn)
{
	int ret = -EINVAL;
	spinlock_t *ptl;
	pte_t *ptep;

	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		return ret;

	ret = follow_pte(vma->vm_mm, address, &ptep, &ptl);
	if (ret)
		return ret;
	*pfn = pte_pfn(*ptep);
	pte_unmap_unlock(ptep, ptl);
	return 0;
}
EXPORT_SYMBOL(follow_pfn);

3820
#ifdef CONFIG_HAVE_IOREMAP_PROT
3821 3822 3823
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
3824
{
3825
	int ret = -EINVAL;
3826 3827 3828
	pte_t *ptep, pte;
	spinlock_t *ptl;

3829 3830
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
3831

3832
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
3833
		goto out;
3834
	pte = *ptep;
3835

3836 3837 3838 3839
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
3840
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
3841

3842
	ret = 0;
3843 3844 3845
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
3846
	return ret;
3847 3848 3849 3850 3851 3852 3853
}

int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
			void *buf, int len, int write)
{
	resource_size_t phys_addr;
	unsigned long prot = 0;
K
KOSAKI Motohiro 已提交
3854
	void __iomem *maddr;
3855 3856
	int offset = addr & (PAGE_SIZE-1);

3857
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
3858 3859
		return -EINVAL;

3860
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
3861 3862 3863 3864 3865 3866 3867 3868
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
3869
EXPORT_SYMBOL_GPL(generic_access_phys);
3870 3871
#endif

3872
/*
3873 3874
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
3875
 */
3876
static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
3877
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
3878 3879 3880
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
3881
	int write = gup_flags & FOLL_WRITE;
3882 3883

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
3884
	/* ignore errors, just check how much was successfully transferred */
3885 3886 3887
	while (len) {
		int bytes, ret, offset;
		void *maddr;
3888
		struct page *page = NULL;
3889

3890
		ret = get_user_pages_remote(tsk, mm, addr, 1,
3891
				gup_flags, &page, &vma);
3892
		if (ret <= 0) {
3893 3894 3895
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
3896 3897 3898 3899 3900
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
3901
			if (!vma || vma->vm_start > addr)
3902 3903 3904 3905 3906 3907 3908
				break;
			if (vma->vm_ops && vma->vm_ops->access)
				ret = vma->vm_ops->access(vma, addr, buf,
							  len, write);
			if (ret <= 0)
				break;
			bytes = ret;
3909
#endif
3910
		} else {
3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925
			bytes = len;
			offset = addr & (PAGE_SIZE-1);
			if (bytes > PAGE_SIZE-offset)
				bytes = PAGE_SIZE-offset;

			maddr = kmap(page);
			if (write) {
				copy_to_user_page(vma, page, addr,
						  maddr + offset, buf, bytes);
				set_page_dirty_lock(page);
			} else {
				copy_from_user_page(vma, page, addr,
						    buf, maddr + offset, bytes);
			}
			kunmap(page);
3926
			put_page(page);
3927 3928 3929 3930 3931 3932 3933 3934 3935
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
3936

S
Stephen Wilson 已提交
3937
/**
3938
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
3939 3940 3941 3942
 * @mm:		the mm_struct of the target address space
 * @addr:	start address to access
 * @buf:	source or destination buffer
 * @len:	number of bytes to transfer
3943
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
3944 3945 3946 3947
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
3948
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
3949
{
3950
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
3951 3952
}

3953 3954 3955 3956 3957 3958
/*
 * Access another process' address space.
 * Source/target buffer must be kernel space,
 * Do not walk the page table directly, use get_user_pages
 */
int access_process_vm(struct task_struct *tsk, unsigned long addr,
3959
		void *buf, int len, unsigned int gup_flags)
3960 3961 3962 3963 3964 3965 3966 3967
{
	struct mm_struct *mm;
	int ret;

	mm = get_task_mm(tsk);
	if (!mm)
		return 0;

3968
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
3969

3970 3971 3972 3973 3974
	mmput(mm);

	return ret;
}

3975 3976 3977 3978 3979 3980 3981 3982
/*
 * Print the name of a VMA.
 */
void print_vma_addr(char *prefix, unsigned long ip)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;

3983 3984 3985 3986 3987 3988 3989
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

3990 3991 3992 3993 3994 3995
	down_read(&mm->mmap_sem);
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
		char *buf = (char *)__get_free_page(GFP_KERNEL);
		if (buf) {
A
Andy Shevchenko 已提交
3996
			char *p;
3997

M
Miklos Szeredi 已提交
3998
			p = file_path(f, buf, PAGE_SIZE);
3999 4000
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4001
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4002 4003 4004 4005 4006
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4007
	up_read(&mm->mmap_sem);
4008
}
4009

4010
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4011
void __might_fault(const char *file, int line)
4012
{
4013 4014 4015 4016 4017 4018 4019 4020
	/*
	 * Some code (nfs/sunrpc) uses socket ops on kernel memory while
	 * holding the mmap_sem, this is safe because kernel memory doesn't
	 * get paged out, therefore we'll never actually fault, and the
	 * below annotations will generate false positives.
	 */
	if (segment_eq(get_fs(), KERNEL_DS))
		return;
4021
	if (pagefault_disabled())
4022
		return;
4023 4024
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4025
	if (current->mm)
4026
		might_lock_read(&current->mm->mmap_sem);
4027
#endif
4028
}
4029
EXPORT_SYMBOL(__might_fault);
4030
#endif
A
Andrea Arcangeli 已提交
4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101

#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
static void clear_gigantic_page(struct page *page,
				unsigned long addr,
				unsigned int pages_per_huge_page)
{
	int i;
	struct page *p = page;

	might_sleep();
	for (i = 0; i < pages_per_huge_page;
	     i++, p = mem_map_next(p, page, i)) {
		cond_resched();
		clear_user_highpage(p, addr + i * PAGE_SIZE);
	}
}
void clear_huge_page(struct page *page,
		     unsigned long addr, unsigned int pages_per_huge_page)
{
	int i;

	if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES)) {
		clear_gigantic_page(page, addr, pages_per_huge_page);
		return;
	}

	might_sleep();
	for (i = 0; i < pages_per_huge_page; i++) {
		cond_resched();
		clear_user_highpage(page + i, addr + i * PAGE_SIZE);
	}
}

static void copy_user_gigantic_page(struct page *dst, struct page *src,
				    unsigned long addr,
				    struct vm_area_struct *vma,
				    unsigned int pages_per_huge_page)
{
	int i;
	struct page *dst_base = dst;
	struct page *src_base = src;

	for (i = 0; i < pages_per_huge_page; ) {
		cond_resched();
		copy_user_highpage(dst, src, addr + i*PAGE_SIZE, vma);

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

void copy_user_huge_page(struct page *dst, struct page *src,
			 unsigned long addr, struct vm_area_struct *vma,
			 unsigned int pages_per_huge_page)
{
	int i;

	if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES)) {
		copy_user_gigantic_page(dst, src, addr, vma,
					pages_per_huge_page);
		return;
	}

	might_sleep();
	for (i = 0; i < pages_per_huge_page; i++) {
		cond_resched();
		copy_user_highpage(dst + i, src + i, addr + i*PAGE_SIZE, vma);
	}
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4102

4103
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4104 4105 4106 4107 4108 4109 4110 4111 4112

static struct kmem_cache *page_ptl_cachep;

void __init ptlock_cache_init(void)
{
	page_ptl_cachep = kmem_cache_create("page->ptl", sizeof(spinlock_t), 0,
			SLAB_PANIC, NULL);
}

4113
bool ptlock_alloc(struct page *page)
4114 4115 4116
{
	spinlock_t *ptl;

4117
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4118 4119
	if (!ptl)
		return false;
4120
	page->ptl = ptl;
4121 4122 4123
	return true;
}

4124
void ptlock_free(struct page *page)
4125
{
4126
	kmem_cache_free(page_ptl_cachep, page->ptl);
4127 4128
}
#endif