memory.c 111.9 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
	VM_WARN_ON(tlb->page_size != page_size);
304

305
	batch = tlb->active;
306 307 308 309 310
	/*
	 * Add the page and check if we are full. If so
	 * force a flush.
	 */
	batch->pages[batch->nr++] = page;
311 312
	if (batch->nr == batch->max) {
		if (!tlb_next_batch(tlb))
313
			return true;
314
		batch = tlb->active;
315
	}
316
	VM_BUG_ON_PAGE(batch->nr > batch->max, page);
317

318
	return false;
319 320 321 322
}

#endif /* HAVE_GENERIC_MMU_GATHER */

323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395
#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);
}

396
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
397

L
Linus Torvalds 已提交
398 399 400 401
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
402 403
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
404
{
405
	pgtable_t token = pmd_pgtable(*pmd);
406
	pmd_clear(pmd);
407
	pte_free_tlb(tlb, token, addr);
408
	atomic_long_dec(&tlb->mm->nr_ptes);
L
Linus Torvalds 已提交
409 410
}

411 412 413
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
414 415 416
{
	pmd_t *pmd;
	unsigned long next;
417
	unsigned long start;
L
Linus Torvalds 已提交
418

419
	start = addr;
L
Linus Torvalds 已提交
420 421 422 423 424
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
425
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
426 427
	} while (pmd++, addr = next, addr != end);

428 429 430 431 432 433 434
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
435
	}
436 437 438 439 440
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
441
	pmd_free_tlb(tlb, pmd, start);
442
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
443 444
}

445 446 447
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
448 449 450
{
	pud_t *pud;
	unsigned long next;
451
	unsigned long start;
L
Linus Torvalds 已提交
452

453
	start = addr;
L
Linus Torvalds 已提交
454 455 456 457 458
	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
459
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
460 461
	} while (pud++, addr = next, addr != end);

462 463 464 465 466 467 468
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
469
	}
470 471 472 473 474
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(pgd, start);
	pgd_clear(pgd);
475
	pud_free_tlb(tlb, pud, start);
L
Linus Torvalds 已提交
476 477 478
}

/*
479
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
480
 */
481
void free_pgd_range(struct mmu_gather *tlb,
482 483
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
484 485 486
{
	pgd_t *pgd;
	unsigned long next;
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512

	/*
	 * 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
513

514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
	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;
529 530 531 532 533
	/*
	 * 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);
534
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
535 536 537 538
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
539
		free_pud_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
540
	} while (pgd++, addr = next, addr != end);
541 542
}

543
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
544
		unsigned long floor, unsigned long ceiling)
545 546 547 548 549
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

550
		/*
N
npiggin@suse.de 已提交
551 552
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
553
		 */
554
		unlink_anon_vmas(vma);
555 556
		unlink_file_vma(vma);

557
		if (is_vm_hugetlb_page(vma)) {
558
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
559
				floor, next? next->vm_start: ceiling);
560 561 562 563 564
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
565
			       && !is_vm_hugetlb_page(next)) {
566 567
				vma = next;
				next = vma->vm_next;
568
				unlink_anon_vmas(vma);
569
				unlink_file_vma(vma);
570 571 572 573
			}
			free_pgd_range(tlb, addr, vma->vm_end,
				floor, next? next->vm_start: ceiling);
		}
574 575
		vma = next;
	}
L
Linus Torvalds 已提交
576 577
}

578
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
579
{
580
	spinlock_t *ptl;
581
	pgtable_t new = pte_alloc_one(mm, address);
582 583 584
	if (!new)
		return -ENOMEM;

585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
	/*
	 * 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 */

600
	ptl = pmd_lock(mm, pmd);
601
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
602
		atomic_long_inc(&mm->nr_ptes);
L
Linus Torvalds 已提交
603
		pmd_populate(mm, pmd, new);
604
		new = NULL;
605
	}
606
	spin_unlock(ptl);
607 608
	if (new)
		pte_free(mm, new);
609
	return 0;
L
Linus Torvalds 已提交
610 611
}

612
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
613
{
614 615 616 617
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

618 619
	smp_wmb(); /* See comment in __pte_alloc */

620
	spin_lock(&init_mm.page_table_lock);
621
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
622
		pmd_populate_kernel(&init_mm, pmd, new);
623
		new = NULL;
624
	}
625
	spin_unlock(&init_mm.page_table_lock);
626 627
	if (new)
		pte_free_kernel(&init_mm, new);
628
	return 0;
L
Linus Torvalds 已提交
629 630
}

K
KAMEZAWA Hiroyuki 已提交
631 632 633 634 635 636
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)
637
{
K
KAMEZAWA Hiroyuki 已提交
638 639
	int i;

640
	if (current->mm == mm)
641
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
642 643 644
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
645 646
}

N
Nick Piggin 已提交
647
/*
648 649 650
 * 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
651 652 653
 *
 * The calling function must still handle the error.
 */
654 655
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
656
{
657 658 659 660 661
	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;
662 663 664 665 666 667 668 669 670 671 672 673 674 675
	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) {
676 677
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
678 679 680 681 682 683
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
684 685 686 687

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

688 689 690
	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));
691
	if (page)
692
		dump_page(page, "bad pte");
693 694
	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);
695 696 697
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
698 699 700 701 702
	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
703
	dump_stack();
704
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
705 706
}

H
Hugh Dickins 已提交
707
/*
N
Nick Piggin 已提交
708
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
709
 *
N
Nick Piggin 已提交
710 711 712
 * "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
713
 *
N
Nick Piggin 已提交
714 715 716 717 718 719 720 721
 * 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.
722
 *
J
Jared Hulbert 已提交
723 724
 * 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
725 726
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
727 728 729
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
730 731 732 733 734 735
 * 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
736 737
 *
 *
N
Nick Piggin 已提交
738
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
739 740 741 742 743 744 745 746 747
 *
 * 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
748
 */
N
Nick Piggin 已提交
749 750 751 752 753 754 755
#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
756
{
757
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
758 759

	if (HAVE_PTE_SPECIAL) {
760
		if (likely(!pte_special(pte)))
761
			goto check_pfn;
762 763
		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
764 765
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
H
Hugh Dickins 已提交
766
		if (!is_zero_pfn(pfn))
767
			print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
768 769 770 771 772
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
773 774 775 776 777 778
	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
779 780
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
781 782 783 784 785
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
786 787
	}

788 789
	if (is_zero_pfn(pfn))
		return NULL;
790 791 792 793 794
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
795 796

	/*
N
Nick Piggin 已提交
797 798
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
799
	 */
J
Jared Hulbert 已提交
800
out:
801
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
802 803
}

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
#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
844 845 846 847 848 849
/*
 * 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
850
static inline unsigned long
L
Linus Torvalds 已提交
851
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
852
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
853
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
854
{
N
Nick Piggin 已提交
855
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
856 857 858 859 860
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
		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);

879
			rss[mm_counter(page)]++;
880 881 882 883 884 885 886 887 888 889 890 891

			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);
892
			}
L
Linus Torvalds 已提交
893
		}
894
		goto out_set_pte;
L
Linus Torvalds 已提交
895 896 897 898 899 900
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
901
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
902
		ptep_set_wrprotect(src_mm, addr, src_pte);
903
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
904 905 906 907 908 909 910 911 912
	}

	/*
	 * 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);
913 914 915 916

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
917
		page_dup_rmap(page, false);
918
		rss[mm_counter(page)]++;
919
	}
920 921 922

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
923
	return 0;
L
Linus Torvalds 已提交
924 925
}

926
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
927 928
		   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
929
{
930
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
931
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
932
	spinlock_t *src_ptl, *dst_ptl;
933
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
934
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
935
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
936 937

again:
K
KAMEZAWA Hiroyuki 已提交
938 939
	init_rss_vec(rss);

H
Hugh Dickins 已提交
940
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
941 942
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
943
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
944
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
945
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
946 947
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
948
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
949 950 951 952 953 954

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
955 956 957
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
958
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
959 960
				break;
		}
L
Linus Torvalds 已提交
961 962 963 964
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
965 966 967 968
		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
969 970 971
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

972
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
973
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
974
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
975
	add_mm_rss_vec(dst_mm, rss);
976
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
977
	cond_resched();
H
Hugh Dickins 已提交
978 979 980 981 982 983

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
	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);
1002
		if (pmd_trans_huge(*src_pmd) || pmd_devmap(*src_pmd)) {
1003
			int err;
1004
			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
1005 1006 1007 1008 1009 1010 1011 1012
			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
1013 1014 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
		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;
1051 1052 1053
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1054
	int ret;
L
Linus Torvalds 已提交
1055

1056 1057 1058 1059 1060 1061
	/*
	 * 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.
	 */
1062 1063 1064
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1065

L
Linus Torvalds 已提交
1066 1067 1068
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1069
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1070 1071 1072 1073
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1074
		ret = track_pfn_copy(vma);
1075 1076 1077 1078
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1079 1080 1081 1082 1083 1084
	/*
	 * 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.
	 */
1085 1086 1087 1088 1089 1090
	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
1091 1092

	ret = 0;
L
Linus Torvalds 已提交
1093 1094 1095 1096 1097 1098
	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
1099 1100 1101 1102 1103
		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
1104
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1105

1106 1107
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1108
	return ret;
L
Linus Torvalds 已提交
1109 1110
}

1111
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1112
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1113
				unsigned long addr, unsigned long end,
1114
				struct zap_details *details)
L
Linus Torvalds 已提交
1115
{
N
Nick Piggin 已提交
1116
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1117
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1118
	int rss[NR_MM_COUNTERS];
1119
	spinlock_t *ptl;
1120
	pte_t *start_pte;
1121
	pte_t *pte;
1122
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1123

1124
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1125
again:
1126
	init_rss_vec(rss);
1127 1128
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1129
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1130 1131
	do {
		pte_t ptent = *pte;
1132
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1133
			continue;
1134
		}
1135

L
Linus Torvalds 已提交
1136
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1137
			struct page *page;
1138

1139
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1140 1141 1142 1143 1144 1145 1146
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1147
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1148 1149
					continue;
			}
N
Nick Piggin 已提交
1150
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1151
							tlb->fullmm);
L
Linus Torvalds 已提交
1152 1153 1154
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1155 1156

			if (!PageAnon(page)) {
1157
				if (pte_dirty(ptent)) {
M
Michal Hocko 已提交
1158 1159 1160 1161 1162 1163
					/*
					 * oom_reaper cannot tear down dirty
					 * pages
					 */
					if (unlikely(details && details->ignore_dirty))
						continue;
1164
					force_flush = 1;
1165
					set_page_dirty(page);
1166
				}
1167
				if (pte_young(ptent) &&
1168
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1169
					mark_page_accessed(page);
1170
			}
1171
			rss[mm_counter(page)]--;
1172
			page_remove_rmap(page, false);
1173 1174
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1175
			if (unlikely(__tlb_remove_page(tlb, page))) {
1176
				force_flush = 1;
1177
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1178
				break;
1179
			}
L
Linus Torvalds 已提交
1180 1181
			continue;
		}
M
Michal Hocko 已提交
1182 1183
		/* only check swap_entries if explicitly asked for in details */
		if (unlikely(details && !details->check_swap_entries))
L
Linus Torvalds 已提交
1184
			continue;
K
KAMEZAWA Hiroyuki 已提交
1185

1186 1187 1188 1189 1190
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1191

1192
			page = migration_entry_to_page(entry);
1193
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1194
		}
1195 1196
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1197
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1198
	} while (pte++, addr += PAGE_SIZE, addr != end);
1199

K
KAMEZAWA Hiroyuki 已提交
1200
	add_mm_rss_vec(mm, rss);
1201
	arch_leave_lazy_mmu_mode();
1202

1203
	/* Do the actual TLB flush before dropping ptl */
1204
	if (force_flush)
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
		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);
1217
		if (addr != end)
P
Peter Zijlstra 已提交
1218 1219 1220
			goto again;
	}

1221
	return addr;
L
Linus Torvalds 已提交
1222 1223
}

1224
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1225
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1226
				unsigned long addr, unsigned long end,
1227
				struct zap_details *details)
L
Linus Torvalds 已提交
1228 1229 1230 1231 1232 1233 1234
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1235
		if (pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1236
			if (next - addr != HPAGE_PMD_SIZE) {
1237 1238
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1239
				__split_huge_pmd(vma, pmd, addr, false, NULL);
S
Shaohua Li 已提交
1240
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1241
				goto next;
1242 1243
			/* fall through */
		}
1244 1245 1246 1247 1248 1249 1250 1251 1252
		/*
		 * 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;
1253
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1254
next:
1255 1256
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1257 1258

	return addr;
L
Linus Torvalds 已提交
1259 1260
}

1261
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1262
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1263
				unsigned long addr, unsigned long end,
1264
				struct zap_details *details)
L
Linus Torvalds 已提交
1265 1266 1267 1268 1269 1270 1271
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1272
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1273
			continue;
1274 1275
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
	} while (pud++, addr = next, addr != end);
1276 1277

	return addr;
L
Linus Torvalds 已提交
1278 1279
}

M
Michal Hocko 已提交
1280
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1281 1282 1283
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1284 1285 1286 1287 1288 1289 1290 1291 1292
{
	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);
1293
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1294
			continue;
1295 1296
		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1297 1298
	tlb_end_vma(tlb, vma);
}
1299

1300 1301 1302

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1303
		unsigned long end_addr,
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
		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;

1315 1316 1317
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1318
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1319
		untrack_pfn(vma, 0, 0);
1320 1321 1322 1323 1324 1325 1326

	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
1327
			 * cleanup path of mmap_region. When
1328
			 * hugetlbfs ->mmap method fails,
1329
			 * mmap_region() nullifies vma->vm_file
1330 1331 1332 1333
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1334
			if (vma->vm_file) {
1335
				i_mmap_lock_write(vma->vm_file->f_mapping);
1336
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1337
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1338
			}
1339 1340 1341
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1342 1343 1344 1345
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1346
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1347 1348 1349 1350
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1351
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
 *
 * 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
1362
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1363
		struct vm_area_struct *vma, unsigned long start_addr,
1364
		unsigned long end_addr)
L
Linus Torvalds 已提交
1365
{
A
Andrea Arcangeli 已提交
1366
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1367

A
Andrea Arcangeli 已提交
1368
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1369
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1370
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1371
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1372 1373 1374 1375 1376
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1377
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1378
 * @size: number of bytes to zap
1379
 * @details: details of shared cache invalidation
1380 1381
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1382
 */
1383
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
L
Linus Torvalds 已提交
1384 1385 1386
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1387
	struct mmu_gather tlb;
1388
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1389 1390

	lru_add_drain();
1391
	tlb_gather_mmu(&tlb, mm, start, end);
1392
	update_hiwater_rss(mm);
1393 1394
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1395
		unmap_single_vma(&tlb, vma, start, end, details);
1396 1397
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1398 1399
}

1400 1401 1402 1403 1404
/**
 * 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
1405
 * @details: details of shared cache invalidation
1406 1407
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1408
 */
1409
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1410 1411 1412
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1413
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1414 1415 1416
	unsigned long end = address + size;

	lru_add_drain();
1417
	tlb_gather_mmu(&tlb, mm, address, end);
1418
	update_hiwater_rss(mm);
1419
	mmu_notifier_invalidate_range_start(mm, address, end);
1420
	unmap_single_vma(&tlb, vma, address, end, details);
1421
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1422
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1423 1424
}

1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
/**
 * 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;
1443
	zap_page_range_single(vma, address, size, NULL);
1444 1445 1446 1447
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1448
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1449
			spinlock_t **ptl)
1450 1451 1452 1453
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
1454
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
1455 1456
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
1457
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
1458
		}
1459 1460 1461 1462
	}
	return NULL;
}

1463 1464 1465 1466 1467 1468 1469
/*
 * 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
1470 1471
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1472
{
N
Nick Piggin 已提交
1473
	struct mm_struct *mm = vma->vm_mm;
1474
	int retval;
1475
	pte_t *pte;
1476 1477
	spinlock_t *ptl;

1478
	retval = -EINVAL;
1479
	if (PageAnon(page))
1480
		goto out;
1481 1482
	retval = -ENOMEM;
	flush_dcache_page(page);
1483
	pte = get_locked_pte(mm, addr, &ptl);
1484
	if (!pte)
1485
		goto out;
1486 1487 1488 1489 1490 1491
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1492
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1493
	page_add_file_rmap(page, false);
1494 1495 1496
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1497 1498
	pte_unmap_unlock(pte, ptl);
	return retval;
1499 1500 1501 1502 1503 1504
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1505 1506 1507 1508 1509 1510
/**
 * 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
 *
1511 1512 1513 1514 1515 1516
 * 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
1517
 * (see split_page()).
1518 1519 1520 1521 1522 1523 1524 1525
 *
 * 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.
1526 1527 1528 1529 1530
 *
 * 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.
1531
 */
N
Nick Piggin 已提交
1532 1533
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1534 1535 1536 1537 1538
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1539 1540 1541 1542 1543
	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
1544
	return insert_page(vma, addr, page, vma->vm_page_prot);
1545
}
1546
EXPORT_SYMBOL(vm_insert_page);
1547

N
Nick Piggin 已提交
1548
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1549
			pfn_t pfn, pgprot_t prot)
N
Nick Piggin 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
{
	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.. */
1565 1566 1567 1568
	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
1569
	set_pte_at(mm, addr, pte, entry);
1570
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1571 1572 1573 1574 1575 1576 1577 1578

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

N
Nick Piggin 已提交
1579 1580 1581 1582 1583 1584
/**
 * 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
 *
1585
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1586 1587 1588 1589
 * 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
1590 1591 1592 1593 1594
 *
 * 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
1595 1596
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1597
			unsigned long pfn)
A
Andy Lutomirski 已提交
1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
{
	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
1620
{
1621
	int ret;
N
Nick Piggin 已提交
1622 1623 1624 1625 1626 1627
	/*
	 * 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
1628 1629 1630 1631 1632
	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
1633

N
Nick Piggin 已提交
1634 1635
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1636 1637

	track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
1638

1639
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot);
1640 1641

	return ret;
N
Nick Piggin 已提交
1642
}
A
Andy Lutomirski 已提交
1643
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1644

N
Nick Piggin 已提交
1645
int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1646
			pfn_t pfn)
N
Nick Piggin 已提交
1647
{
1648 1649
	pgprot_t pgprot = vma->vm_page_prot;

N
Nick Piggin 已提交
1650
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1651

N
Nick Piggin 已提交
1652 1653
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1654 1655

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1656

N
Nick Piggin 已提交
1657 1658 1659 1660
	/*
	 * 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
1661 1662
	 * 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
1663
	 */
1664
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1665 1666
		struct page *page;

1667 1668 1669 1670 1671 1672
		/*
		 * 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));
1673
		return insert_page(vma, addr, page, pgprot);
N
Nick Piggin 已提交
1674
	}
1675
	return insert_pfn(vma, addr, pfn, pgprot);
N
Nick Piggin 已提交
1676
}
N
Nick Piggin 已提交
1677
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1678

L
Linus Torvalds 已提交
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
/*
 * 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
1689
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1690

H
Hugh Dickins 已提交
1691
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1692 1693
	if (!pte)
		return -ENOMEM;
1694
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1695 1696
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1697
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1698 1699
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1700
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1701
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715
	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;
1716
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745
	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;
}

1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
/**
 * 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
1756 1757 1758 1759 1760
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;
1761
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1762
	struct mm_struct *mm = vma->vm_mm;
1763
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1764 1765 1766 1767 1768 1769 1770
	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).
1771 1772 1773
	 *   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.
1774 1775 1776 1777
	 *   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
1778 1779 1780 1781
	 *
	 * 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".
1782
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1783
	 */
1784 1785 1786
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1787
		vma->vm_pgoff = pfn;
1788 1789
	}

1790
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1791
	if (err)
1792
		return -EINVAL;
L
Linus Torvalds 已提交
1793

1794
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806

	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);
1807 1808

	if (err)
1809
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1810

L
Linus Torvalds 已提交
1811 1812 1813 1814
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1815 1816 1817 1818 1819 1820 1821 1822 1823 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
/**
 * 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);

1862 1863 1864 1865 1866 1867
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;
1868
	pgtable_t token;
1869
	spinlock_t *uninitialized_var(ptl);
1870 1871 1872 1873 1874 1875 1876 1877 1878

	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));

1879 1880
	arch_enter_lazy_mmu_mode();

1881
	token = pmd_pgtable(*pmd);
1882 1883

	do {
1884
		err = fn(pte++, token, addr, data);
1885 1886
		if (err)
			break;
1887
	} while (addr += PAGE_SIZE, addr != end);
1888

1889 1890
	arch_leave_lazy_mmu_mode();

1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903
	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
1904 1905
	BUG_ON(pud_huge(*pud));

1906 1907 1908 1909 1910 1911 1912 1913 1914 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
	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;
1947
	unsigned long end = addr + size;
1948 1949
	int err;

1950 1951 1952
	if (WARN_ON(addr >= end))
		return -EINVAL;

1953 1954 1955 1956 1957 1958 1959
	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);
1960

1961 1962 1963 1964
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

1965
/*
1966 1967 1968 1969 1970
 * 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;
1971
 * and do_anonymous_page can safely check later on).
1972
 */
H
Hugh Dickins 已提交
1973
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
1974 1975 1976 1977 1978
				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
1979 1980
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
1981
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
1982
		spin_unlock(ptl);
1983 1984 1985 1986 1987 1988
	}
#endif
	pte_unmap(page_table);
	return same;
}

1989
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
1990
{
1991 1992
	debug_dma_assert_idle(src);

1993 1994 1995 1996 1997 1998 1999
	/*
	 * 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)) {
2000
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2001 2002 2003 2004 2005 2006 2007 2008 2009
		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))
2010
			clear_page(kaddr);
2011
		kunmap_atomic(kaddr);
2012
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2013 2014
	} else
		copy_user_highpage(dst, src, va, vma);
2015 2016
}

2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
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;
}

2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
/*
 * 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;
2046
	vmf.gfp_mask = __get_fault_gfp_mask(vma);
2047
	vmf.page = page;
2048
	vmf.cow_page = NULL;
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064

	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;
}

2065 2066 2067 2068 2069 2070 2071 2072
/*
 * 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.
 */
J
Jan Kara 已提交
2073
static inline int wp_page_reuse(struct vm_fault *vmf, pte_t orig_pte,
K
Kirill A. Shutemov 已提交
2074
			struct page *page, int page_mkwrite, int dirty_shared)
J
Jan Kara 已提交
2075
	__releases(vmf->ptl)
2076
{
J
Jan Kara 已提交
2077
	struct vm_area_struct *vma = vmf->vma;
2078 2079 2080 2081 2082 2083 2084 2085 2086
	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);

J
Jan Kara 已提交
2087
	flush_cache_page(vma, vmf->address, pte_pfn(orig_pte));
2088 2089
	entry = pte_mkyoung(orig_pte);
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2090 2091 2092
	if (ptep_set_access_flags(vma, vmf->address, vmf->pte, entry, 1))
		update_mmu_cache(vma, vmf->address, vmf->pte);
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104

	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);
2105
		put_page(page);
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121

		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;
}

2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
/*
 * 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.
 */
J
Jan Kara 已提交
2138
static int wp_page_copy(struct vm_fault *vmf, pte_t orig_pte,
K
Kirill A. Shutemov 已提交
2139
		struct page *old_page)
2140
{
J
Jan Kara 已提交
2141
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2142
	struct mm_struct *mm = vma->vm_mm;
2143 2144 2145
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
J
Jan Kara 已提交
2146
	const unsigned long mmun_start = vmf->address & PAGE_MASK;
K
Kirill A. Shutemov 已提交
2147
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2148 2149 2150 2151 2152 2153
	struct mem_cgroup *memcg;

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

	if (is_zero_pfn(pte_pfn(orig_pte))) {
J
Jan Kara 已提交
2154 2155
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2156 2157 2158
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2159
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2160
				vmf->address);
2161 2162
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2163
		cow_user_page(new_page, old_page, vmf->address, vma);
2164 2165
	}

2166
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2167 2168
		goto oom_free_new;

2169 2170
	__SetPageUptodate(new_page);

2171 2172 2173 2174 2175
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2176 2177
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
	if (likely(pte_same(*vmf->pte, orig_pte))) {
2178 2179
		if (old_page) {
			if (!PageAnon(old_page)) {
2180 2181
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2182 2183 2184 2185 2186
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2187
		flush_cache_page(vma, vmf->address, pte_pfn(orig_pte));
2188 2189 2190 2191 2192 2193 2194 2195
		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.
		 */
J
Jan Kara 已提交
2196 2197
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2198
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2199 2200 2201 2202 2203 2204
		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.
		 */
J
Jan Kara 已提交
2205 2206
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
		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.
			 */
2230
			page_remove_rmap(old_page, false);
2231 2232 2233 2234 2235 2236
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2237
		mem_cgroup_cancel_charge(new_page, memcg, false);
2238 2239 2240
	}

	if (new_page)
2241
		put_page(new_page);
2242

J
Jan Kara 已提交
2243
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2244 2245 2246 2247 2248 2249 2250 2251
	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 */
2252 2253
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2254 2255
			unlock_page(old_page);
		}
2256
		put_page(old_page);
2257 2258 2259
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2260
	put_page(new_page);
2261 2262
oom:
	if (old_page)
2263
		put_page(old_page);
2264 2265 2266
	return VM_FAULT_OOM;
}

2267 2268 2269 2270
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
J
Jan Kara 已提交
2271
static int wp_pfn_shared(struct vm_fault *vmf, pte_t orig_pte)
2272
{
J
Jan Kara 已提交
2273
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2274

2275
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
J
Jan Kara 已提交
2276
		struct vm_fault vmf2 = {
2277
			.page = NULL,
J
Jan Kara 已提交
2278
			.pgoff = linear_page_index(vma, vmf->address),
K
Kirill A. Shutemov 已提交
2279
			.virtual_address =
J
Jan Kara 已提交
2280
				(void __user *)(vmf->address & PAGE_MASK),
2281 2282 2283 2284
			.flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE,
		};
		int ret;

J
Jan Kara 已提交
2285 2286
		pte_unmap_unlock(vmf->pte, vmf->ptl);
		ret = vma->vm_ops->pfn_mkwrite(vma, &vmf2);
2287 2288
		if (ret & VM_FAULT_ERROR)
			return ret;
J
Jan Kara 已提交
2289 2290
		vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
				vmf->address, &vmf->ptl);
2291 2292 2293 2294
		/*
		 * We might have raced with another page fault while we
		 * released the pte_offset_map_lock.
		 */
J
Jan Kara 已提交
2295 2296
		if (!pte_same(*vmf->pte, orig_pte)) {
			pte_unmap_unlock(vmf->pte, vmf->ptl);
2297 2298 2299
			return 0;
		}
	}
J
Jan Kara 已提交
2300
	return wp_page_reuse(vmf, orig_pte, NULL, 0, 0);
2301 2302
}

J
Jan Kara 已提交
2303
static int wp_page_shared(struct vm_fault *vmf, pte_t orig_pte,
K
Kirill A. Shutemov 已提交
2304
		struct page *old_page)
J
Jan Kara 已提交
2305
	__releases(vmf->ptl)
2306
{
J
Jan Kara 已提交
2307
	struct vm_area_struct *vma = vmf->vma;
2308 2309
	int page_mkwrite = 0;

2310
	get_page(old_page);
2311 2312 2313 2314

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

J
Jan Kara 已提交
2315 2316
		pte_unmap_unlock(vmf->pte, vmf->ptl);
		tmp = do_page_mkwrite(vma, old_page, vmf->address);
2317 2318
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
2319
			put_page(old_page);
2320 2321 2322 2323 2324 2325 2326 2327
			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.
		 */
J
Jan Kara 已提交
2328 2329 2330
		vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
						vmf->address, &vmf->ptl);
		if (!pte_same(*vmf->pte, orig_pte)) {
2331
			unlock_page(old_page);
J
Jan Kara 已提交
2332
			pte_unmap_unlock(vmf->pte, vmf->ptl);
2333
			put_page(old_page);
2334 2335 2336 2337 2338
			return 0;
		}
		page_mkwrite = 1;
	}

J
Jan Kara 已提交
2339
	return wp_page_reuse(vmf, orig_pte, old_page, page_mkwrite, 1);
2340 2341
}

L
Linus Torvalds 已提交
2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355
/*
 * 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.
 *
2356 2357 2358
 * 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
2359
 */
J
Jan Kara 已提交
2360 2361
static int do_wp_page(struct vm_fault *vmf, pte_t orig_pte)
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2362
{
J
Jan Kara 已提交
2363
	struct vm_area_struct *vma = vmf->vma;
2364
	struct page *old_page;
L
Linus Torvalds 已提交
2365

J
Jan Kara 已提交
2366
	old_page = vm_normal_page(vma, vmf->address, orig_pte);
2367 2368
	if (!old_page) {
		/*
2369 2370
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2371 2372
		 *
		 * We should not cow pages in a shared writeable mapping.
2373
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2374 2375 2376
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2377
			return wp_pfn_shared(vmf, orig_pte);
2378

J
Jan Kara 已提交
2379 2380
		pte_unmap_unlock(vmf->pte, vmf->ptl);
		return wp_page_copy(vmf, orig_pte, old_page);
2381
	}
L
Linus Torvalds 已提交
2382

2383
	/*
P
Peter Zijlstra 已提交
2384 2385
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2386
	 */
H
Hugh Dickins 已提交
2387
	if (PageAnon(old_page) && !PageKsm(old_page)) {
2388
		int total_mapcount;
2389
		if (!trylock_page(old_page)) {
2390
			get_page(old_page);
J
Jan Kara 已提交
2391
			pte_unmap_unlock(vmf->pte, vmf->ptl);
2392
			lock_page(old_page);
J
Jan Kara 已提交
2393 2394 2395
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
			if (!pte_same(*vmf->pte, orig_pte)) {
2396
				unlock_page(old_page);
J
Jan Kara 已提交
2397
				pte_unmap_unlock(vmf->pte, vmf->ptl);
2398
				put_page(old_page);
2399
				return 0;
2400
			}
2401
			put_page(old_page);
P
Peter Zijlstra 已提交
2402
		}
2403 2404 2405 2406 2407 2408 2409 2410 2411
		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.
				 */
2412
				page_move_anon_rmap(old_page, vma);
2413
			}
2414
			unlock_page(old_page);
J
Jan Kara 已提交
2415
			return wp_page_reuse(vmf, orig_pte, old_page, 0, 0);
2416
		}
2417
		unlock_page(old_page);
P
Peter Zijlstra 已提交
2418
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2419
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2420
		return wp_page_shared(vmf, orig_pte, old_page);
L
Linus Torvalds 已提交
2421 2422 2423 2424 2425
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
2426
	get_page(old_page);
2427

J
Jan Kara 已提交
2428 2429
	pte_unmap_unlock(vmf->pte, vmf->ptl);
	return wp_page_copy(vmf, orig_pte, old_page);
L
Linus Torvalds 已提交
2430 2431
}

2432
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2433 2434 2435
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2436
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2437 2438
}

2439
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2440 2441 2442 2443 2444
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2445
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2446 2447 2448
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2449
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2450 2451 2452 2453 2454 2455 2456
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2457
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2458 2459
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2460
				details);
L
Linus Torvalds 已提交
2461 2462 2463 2464
	}
}

/**
2465 2466 2467 2468
 * 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
2469
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2470 2471
 * @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
2472
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
 * 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
2484
	struct zap_details details = { };
L
Linus Torvalds 已提交
2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501
	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;

2502
	i_mmap_lock_write(mapping);
2503
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2504
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2505
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2506 2507 2508 2509
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2510 2511
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2512 2513 2514 2515
 * 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
2516
 */
J
Jan Kara 已提交
2517
int do_swap_page(struct vm_fault *vmf, pte_t orig_pte)
L
Linus Torvalds 已提交
2518
{
J
Jan Kara 已提交
2519
	struct vm_area_struct *vma = vmf->vma;
2520
	struct page *page, *swapcache;
2521
	struct mem_cgroup *memcg;
2522
	swp_entry_t entry;
L
Linus Torvalds 已提交
2523
	pte_t pte;
2524
	int locked;
2525
	int exclusive = 0;
N
Nick Piggin 已提交
2526
	int ret = 0;
L
Linus Torvalds 已提交
2527

J
Jan Kara 已提交
2528
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, orig_pte))
2529
		goto out;
2530 2531

	entry = pte_to_swp_entry(orig_pte);
2532 2533
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2534 2535
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2536 2537 2538
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2539
			print_bad_pte(vma, vmf->address, orig_pte, NULL);
H
Hugh Dickins 已提交
2540
			ret = VM_FAULT_SIGBUS;
2541
		}
2542 2543
		goto out;
	}
2544
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
2545 2546
	page = lookup_swap_cache(entry);
	if (!page) {
J
Jan Kara 已提交
2547 2548
		page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE, vma,
					vmf->address);
L
Linus Torvalds 已提交
2549 2550
		if (!page) {
			/*
2551 2552
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2553
			 */
J
Jan Kara 已提交
2554 2555 2556
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
			if (likely(pte_same(*vmf->pte, orig_pte)))
L
Linus Torvalds 已提交
2557
				ret = VM_FAULT_OOM;
2558
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2559
			goto unlock;
L
Linus Torvalds 已提交
2560 2561 2562 2563
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2564
		count_vm_event(PGMAJFAULT);
K
Kirill A. Shutemov 已提交
2565
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
2566
	} else if (PageHWPoison(page)) {
2567 2568 2569 2570
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2571 2572
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2573
		swapcache = page;
2574
		goto out_release;
L
Linus Torvalds 已提交
2575 2576
	}

2577
	swapcache = page;
J
Jan Kara 已提交
2578
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2579

2580
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2581 2582 2583 2584
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2585

A
Andrea Arcangeli 已提交
2586
	/*
2587 2588 2589 2590
	 * 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
2591
	 */
2592
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2593 2594
		goto out_page;

J
Jan Kara 已提交
2595
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2596 2597 2598 2599
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2600 2601
	}

K
Kirill A. Shutemov 已提交
2602 2603
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2604
		ret = VM_FAULT_OOM;
2605
		goto out_page;
2606 2607
	}

L
Linus Torvalds 已提交
2608
	/*
2609
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2610
	 */
J
Jan Kara 已提交
2611 2612 2613
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
	if (unlikely(!pte_same(*vmf->pte, orig_pte)))
2614 2615 2616 2617 2618
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2619 2620
	}

2621 2622 2623 2624 2625 2626 2627 2628 2629
	/*
	 * 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
2630

K
Kirill A. Shutemov 已提交
2631 2632
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2633
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2634
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2635
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2636
		vmf->flags &= ~FAULT_FLAG_WRITE;
2637
		ret |= VM_FAULT_WRITE;
2638
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2639 2640
	}
	flush_icache_page(vma, page);
2641 2642
	if (pte_swp_soft_dirty(orig_pte))
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2643
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
2644
	if (page == swapcache) {
J
Jan Kara 已提交
2645
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
2646
		mem_cgroup_commit_charge(page, memcg, true, false);
2647
		activate_page(page);
2648
	} else { /* ksm created a completely new copy */
J
Jan Kara 已提交
2649
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2650
		mem_cgroup_commit_charge(page, memcg, false, false);
2651 2652
		lru_cache_add_active_or_unevictable(page, vma);
	}
L
Linus Torvalds 已提交
2653

2654
	swap_free(entry);
2655 2656
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2657
		try_to_free_swap(page);
2658
	unlock_page(page);
2659
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2660 2661 2662 2663 2664 2665 2666 2667 2668
		/*
		 * 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);
2669
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2670
	}
2671

J
Jan Kara 已提交
2672 2673
	if (vmf->flags & FAULT_FLAG_WRITE) {
		ret |= do_wp_page(vmf, pte);
2674 2675
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2676 2677 2678 2679
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2680
	update_mmu_cache(vma, vmf->address, vmf->pte);
2681
unlock:
J
Jan Kara 已提交
2682
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
2683 2684
out:
	return ret;
2685
out_nomap:
2686
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
2687
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2688
out_page:
2689
	unlock_page(page);
2690
out_release:
2691
	put_page(page);
2692
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2693
		unlock_page(swapcache);
2694
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2695
	}
2696
	return ret;
L
Linus Torvalds 已提交
2697 2698
}

2699
/*
2700 2701
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
2702 2703 2704 2705 2706 2707
 * 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) {
2708 2709 2710 2711 2712 2713 2714 2715 2716 2717
		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;
2718

2719
		return expand_downwards(vma, address - PAGE_SIZE);
2720
	}
2721 2722 2723 2724 2725 2726 2727
	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;

2728
		return expand_upwards(vma, address + PAGE_SIZE);
2729
	}
2730 2731 2732
	return 0;
}

L
Linus Torvalds 已提交
2733
/*
2734 2735 2736
 * 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
2737
 */
J
Jan Kara 已提交
2738
static int do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2739
{
J
Jan Kara 已提交
2740
	struct vm_area_struct *vma = vmf->vma;
2741
	struct mem_cgroup *memcg;
2742
	struct page *page;
L
Linus Torvalds 已提交
2743 2744
	pte_t entry;

2745 2746 2747 2748
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2749
	/* Check if we need to add a guard page to the stack */
J
Jan Kara 已提交
2750
	if (check_stack_guard_page(vma, vmf->address) < 0)
2751
		return VM_FAULT_SIGSEGV;
2752

2753 2754 2755 2756 2757 2758 2759 2760 2761 2762
	/*
	 * 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).
	 */
J
Jan Kara 已提交
2763
	if (pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))
2764 2765 2766
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
2767
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
2768 2769
		return 0;

2770
	/* Use the zero-page for reads */
J
Jan Kara 已提交
2771
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
2772
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
2773
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
2774
						vma->vm_page_prot));
J
Jan Kara 已提交
2775 2776 2777
		vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
				vmf->address, &vmf->ptl);
		if (!pte_none(*vmf->pte))
H
Hugh Dickins 已提交
2778
			goto unlock;
2779 2780
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2781 2782
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
2783
		}
H
Hugh Dickins 已提交
2784 2785 2786
		goto setpte;
	}

N
Nick Piggin 已提交
2787 2788 2789
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
2790
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
2791 2792
	if (!page)
		goto oom;
2793

K
Kirill A. Shutemov 已提交
2794
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
2795 2796
		goto oom_free_page;

2797 2798 2799 2800 2801
	/*
	 * 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
2802
	__SetPageUptodate(page);
2803

N
Nick Piggin 已提交
2804
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2805 2806
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2807

J
Jan Kara 已提交
2808 2809 2810
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
	if (!pte_none(*vmf->pte))
N
Nick Piggin 已提交
2811
		goto release;
H
Hugh Dickins 已提交
2812

2813 2814
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2815
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2816
		mem_cgroup_cancel_charge(page, memcg, false);
2817
		put_page(page);
J
Jan Kara 已提交
2818
		return handle_userfault(vmf, VM_UFFD_MISSING);
2819 2820
	}

K
Kirill A. Shutemov 已提交
2821
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
2822
	page_add_new_anon_rmap(page, vma, vmf->address, false);
2823
	mem_cgroup_commit_charge(page, memcg, false, false);
2824
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
2825
setpte:
J
Jan Kara 已提交
2826
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
2827 2828

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2829
	update_mmu_cache(vma, vmf->address, vmf->pte);
2830
unlock:
J
Jan Kara 已提交
2831
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
2832
	return 0;
2833
release:
2834
	mem_cgroup_cancel_charge(page, memcg, false);
2835
	put_page(page);
2836
	goto unlock;
2837
oom_free_page:
2838
	put_page(page);
2839
oom:
L
Linus Torvalds 已提交
2840 2841 2842
	return VM_FAULT_OOM;
}

2843 2844 2845 2846 2847
/*
 * 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().
 */
J
Jan Kara 已提交
2848
static int __do_fault(struct vm_fault *vmf, pgoff_t pgoff,
K
Kirill A. Shutemov 已提交
2849
		struct page *cow_page, struct page **page, void **entry)
2850
{
J
Jan Kara 已提交
2851 2852
	struct vm_area_struct *vma = vmf->vma;
	struct vm_fault vmf2;
2853 2854
	int ret;

J
Jan Kara 已提交
2855 2856 2857 2858 2859 2860
	vmf2.virtual_address = (void __user *)(vmf->address & PAGE_MASK);
	vmf2.pgoff = pgoff;
	vmf2.flags = vmf->flags;
	vmf2.page = NULL;
	vmf2.gfp_mask = __get_fault_gfp_mask(vma);
	vmf2.cow_page = cow_page;
2861

J
Jan Kara 已提交
2862
	ret = vma->vm_ops->fault(vma, &vmf2);
2863 2864
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;
2865
	if (ret & VM_FAULT_DAX_LOCKED) {
J
Jan Kara 已提交
2866
		*entry = vmf2.entry;
2867 2868
		return ret;
	}
2869

J
Jan Kara 已提交
2870
	if (unlikely(PageHWPoison(vmf2.page))) {
2871
		if (ret & VM_FAULT_LOCKED)
J
Jan Kara 已提交
2872 2873
			unlock_page(vmf2.page);
		put_page(vmf2.page);
2874 2875 2876 2877
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
J
Jan Kara 已提交
2878
		lock_page(vmf2.page);
2879
	else
J
Jan Kara 已提交
2880
		VM_BUG_ON_PAGE(!PageLocked(vmf2.page), vmf2.page);
2881

J
Jan Kara 已提交
2882
	*page = vmf2.page;
2883 2884 2885
	return ret;
}

J
Jan Kara 已提交
2886
static int pte_alloc_one_map(struct vm_fault *vmf)
2887
{
J
Jan Kara 已提交
2888
	struct vm_area_struct *vma = vmf->vma;
2889

J
Jan Kara 已提交
2890
	if (!pmd_none(*vmf->pmd))
2891
		goto map_pte;
J
Jan Kara 已提交
2892 2893 2894 2895
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
2896 2897 2898 2899
			goto map_pte;
		}

		atomic_long_inc(&vma->vm_mm->nr_ptes);
J
Jan Kara 已提交
2900 2901 2902 2903
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
		vmf->prealloc_pte = 0;
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))) {
2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917
		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.
	 */
J
Jan Kara 已提交
2918
	if (pmd_trans_unstable(vmf->pmd) || pmd_devmap(*vmf->pmd))
2919 2920
		return VM_FAULT_NOPAGE;

J
Jan Kara 已提交
2921 2922
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
2923 2924 2925
	return 0;
}

2926
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939

#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;
}

J
Jan Kara 已提交
2940
static void deposit_prealloc_pte(struct vm_fault *vmf)
2941
{
J
Jan Kara 已提交
2942
	struct vm_area_struct *vma = vmf->vma;
2943

J
Jan Kara 已提交
2944
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
2945 2946 2947 2948 2949
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
	atomic_long_inc(&vma->vm_mm->nr_ptes);
J
Jan Kara 已提交
2950
	vmf->prealloc_pte = 0;
2951 2952
}

J
Jan Kara 已提交
2953
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
2954
{
J
Jan Kara 已提交
2955 2956 2957
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
	unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
K
Kirill A. Shutemov 已提交
2958 2959 2960 2961 2962 2963 2964 2965 2966
	pmd_t entry;
	int i, ret;

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

2967 2968 2969 2970
	/*
	 * Archs like ppc64 need additonal space to store information
	 * related to pte entry. Use the preallocated table for that.
	 */
J
Jan Kara 已提交
2971 2972 2973
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm, vmf->address);
		if (!vmf->prealloc_pte)
2974 2975 2976 2977
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
2978 2979
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990
		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);
2991 2992 2993 2994
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
2995
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
2996

J
Jan Kara 已提交
2997
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
2998

J
Jan Kara 已提交
2999
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3000 3001 3002

	/* fault is handled */
	ret = 0;
3003
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3004
out:
3005 3006 3007 3008 3009
	/*
	 * If we are going to fallback to pte mapping, do a
	 * withdraw with pmd lock held.
	 */
	if (arch_needs_pgtable_deposit() && ret == VM_FAULT_FALLBACK)
J
Jan Kara 已提交
3010 3011 3012
		vmf->prealloc_pte = pgtable_trans_huge_withdraw(vma->vm_mm,
								vmf->pmd);
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3013 3014 3015
	return ret;
}
#else
J
Jan Kara 已提交
3016
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3017 3018 3019 3020 3021 3022
{
	BUILD_BUG();
	return 0;
}
#endif

3023
/**
3024 3025
 * 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.
3026
 *
J
Jan Kara 已提交
3027
 * @vmf: fault environment
3028
 * @memcg: memcg to charge page (only for private mappings)
3029 3030
 * @page: page to map
 *
J
Jan Kara 已提交
3031 3032
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3033 3034 3035 3036
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
J
Jan Kara 已提交
3037
int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3038
		struct page *page)
3039
{
J
Jan Kara 已提交
3040 3041
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3042
	pte_t entry;
K
Kirill A. Shutemov 已提交
3043 3044
	int ret;

J
Jan Kara 已提交
3045
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3046
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3047 3048 3049
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3050
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3051
		if (ret != VM_FAULT_FALLBACK)
3052
			goto fault_handled;
K
Kirill A. Shutemov 已提交
3053
	}
3054

J
Jan Kara 已提交
3055 3056
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3057
		if (ret)
3058
			goto fault_handled;
3059 3060 3061
	}

	/* Re-check under ptl */
J
Jan Kara 已提交
3062
	if (unlikely(!pte_none(*vmf->pte))) {
3063 3064 3065
		ret = VM_FAULT_NOPAGE;
		goto fault_handled;
	}
3066

3067 3068 3069 3070
	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
3071 3072
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3073
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3074
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3075 3076
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3077
	} else {
3078
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3079
		page_add_file_rmap(page, false);
3080
	}
J
Jan Kara 已提交
3081
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3082 3083

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3084
	update_mmu_cache(vma, vmf->address, vmf->pte);
3085
	ret = 0;
3086

3087 3088
fault_handled:
	/* preallocated pagetable is unused: free it */
J
Jan Kara 已提交
3089 3090 3091
	if (vmf->prealloc_pte) {
		pte_free(vmf->vma->vm_mm, vmf->prealloc_pte);
		vmf->prealloc_pte = 0;
3092 3093
	}
	return ret;
3094 3095
}

3096 3097
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3098 3099 3100

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3101
{
3102
	*val = fault_around_bytes;
3103 3104 3105
	return 0;
}

3106 3107 3108 3109 3110
/*
 * 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.
 */
3111
static int fault_around_bytes_set(void *data, u64 val)
3112
{
3113
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3114
		return -EINVAL;
3115 3116 3117 3118
	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 */
3119 3120
	return 0;
}
3121 3122
DEFINE_SIMPLE_ATTRIBUTE(fault_around_bytes_fops,
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3123 3124 3125 3126 3127

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

3128 3129
	ret = debugfs_create_file("fault_around_bytes", 0644, NULL, NULL,
			&fault_around_bytes_fops);
3130
	if (!ret)
3131
		pr_warn("Failed to create fault_around_bytes in debugfs");
3132 3133 3134 3135
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3136

3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159
/*
 * 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.
 */
J
Jan Kara 已提交
3160
static int do_fault_around(struct vm_fault *vmf, pgoff_t start_pgoff)
3161
{
J
Jan Kara 已提交
3162
	unsigned long address = vmf->address, nr_pages, mask;
K
Kirill A. Shutemov 已提交
3163
	pgoff_t end_pgoff;
3164
	int off, ret = 0;
3165

3166
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3167 3168
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3169 3170
	vmf->address = max(address & mask, vmf->vma->vm_start);
	off = ((address - vmf->address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
K
Kirill A. Shutemov 已提交
3171
	start_pgoff -= off;
3172 3173

	/*
K
Kirill A. Shutemov 已提交
3174 3175
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3176
	 */
K
Kirill A. Shutemov 已提交
3177
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3178
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3179
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3180
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3181
			start_pgoff + nr_pages - 1);
3182

J
Jan Kara 已提交
3183 3184 3185 3186
	if (pmd_none(*vmf->pmd)) {
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm,
						  vmf->address);
		if (!vmf->prealloc_pte)
3187
			goto out;
3188
		smp_wmb(); /* See comment in __pte_alloc() */
3189 3190
	}

J
Jan Kara 已提交
3191
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3192 3193

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3194
	if (pmd_trans_huge(*vmf->pmd)) {
3195 3196 3197 3198 3199
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3200
	if (!vmf->pte)
3201 3202 3203
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3204 3205
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3206
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3207
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3208
out:
J
Jan Kara 已提交
3209 3210
	vmf->address = address;
	vmf->pte = NULL;
3211
	return ret;
3212 3213
}

J
Jan Kara 已提交
3214
static int do_read_fault(struct vm_fault *vmf, pgoff_t pgoff)
3215
{
J
Jan Kara 已提交
3216
	struct vm_area_struct *vma = vmf->vma;
3217
	struct page *fault_page;
3218 3219 3220 3221 3222 3223 3224
	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).
	 */
3225
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
J
Jan Kara 已提交
3226
		ret = do_fault_around(vmf, pgoff);
3227 3228
		if (ret)
			return ret;
3229
	}
3230

J
Jan Kara 已提交
3231
	ret = __do_fault(vmf, pgoff, NULL, &fault_page, NULL);
3232 3233 3234
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

J
Jan Kara 已提交
3235 3236 3237
	ret |= alloc_set_pte(vmf, NULL, fault_page);
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3238
	unlock_page(fault_page);
3239 3240
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		put_page(fault_page);
3241 3242 3243
	return ret;
}

J
Jan Kara 已提交
3244
static int do_cow_fault(struct vm_fault *vmf, pgoff_t pgoff)
3245
{
J
Jan Kara 已提交
3246
	struct vm_area_struct *vma = vmf->vma;
3247
	struct page *fault_page, *new_page;
3248
	void *fault_entry;
3249
	struct mem_cgroup *memcg;
3250 3251 3252 3253 3254
	int ret;

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

J
Jan Kara 已提交
3255
	new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
3256 3257 3258
	if (!new_page)
		return VM_FAULT_OOM;

K
Kirill A. Shutemov 已提交
3259 3260
	if (mem_cgroup_try_charge(new_page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
3261
		put_page(new_page);
3262 3263 3264
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3265
	ret = __do_fault(vmf, pgoff, new_page, &fault_page, &fault_entry);
3266 3267 3268
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;

3269
	if (!(ret & VM_FAULT_DAX_LOCKED))
J
Jan Kara 已提交
3270
		copy_user_highpage(new_page, fault_page, vmf->address, vma);
3271 3272
	__SetPageUptodate(new_page);

J
Jan Kara 已提交
3273 3274 3275
	ret |= alloc_set_pte(vmf, memcg, new_page);
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3276
	if (!(ret & VM_FAULT_DAX_LOCKED)) {
3277
		unlock_page(fault_page);
3278
		put_page(fault_page);
3279
	} else {
3280
		dax_unlock_mapping_entry(vma->vm_file->f_mapping, pgoff);
3281
	}
3282 3283
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3284 3285
	return ret;
uncharge_out:
3286
	mem_cgroup_cancel_charge(new_page, memcg, false);
3287
	put_page(new_page);
3288 3289 3290
	return ret;
}

J
Jan Kara 已提交
3291
static int do_shared_fault(struct vm_fault *vmf, pgoff_t pgoff)
L
Linus Torvalds 已提交
3292
{
J
Jan Kara 已提交
3293
	struct vm_area_struct *vma = vmf->vma;
3294 3295 3296 3297
	struct page *fault_page;
	struct address_space *mapping;
	int dirtied = 0;
	int ret, tmp;
3298

J
Jan Kara 已提交
3299
	ret = __do_fault(vmf, pgoff, NULL, &fault_page, NULL);
3300
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3301
		return ret;
L
Linus Torvalds 已提交
3302 3303

	/*
3304 3305
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3306
	 */
3307 3308
	if (vma->vm_ops->page_mkwrite) {
		unlock_page(fault_page);
J
Jan Kara 已提交
3309
		tmp = do_page_mkwrite(vma, fault_page, vmf->address);
3310 3311
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
3312
			put_page(fault_page);
3313
			return tmp;
3314
		}
3315 3316
	}

J
Jan Kara 已提交
3317 3318 3319
	ret |= alloc_set_pte(vmf, NULL, fault_page);
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3320 3321
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
3322
		unlock_page(fault_page);
3323
		put_page(fault_page);
3324
		return ret;
L
Linus Torvalds 已提交
3325
	}
N
Nick Piggin 已提交
3326

3327 3328
	if (set_page_dirty(fault_page))
		dirtied = 1;
3329 3330 3331 3332 3333 3334
	/*
	 * 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.
	 */
3335
	mapping = page_rmapping(fault_page);
3336 3337 3338 3339 3340 3341 3342
	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);
3343
	}
3344

3345
	if (!vma->vm_ops->page_mkwrite)
3346
		file_update_time(vma->vm_file);
N
Nick Piggin 已提交
3347

3348
	return ret;
3349
}
3350

3351 3352 3353 3354 3355 3356
/*
 * 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().
 */
J
Jan Kara 已提交
3357
static int do_fault(struct vm_fault *vmf)
3358
{
J
Jan Kara 已提交
3359 3360
	struct vm_area_struct *vma = vmf->vma;
	pgoff_t pgoff = linear_page_index(vma, vmf->address);
3361

3362 3363 3364
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
		return VM_FAULT_SIGBUS;
J
Jan Kara 已提交
3365 3366
	if (!(vmf->flags & FAULT_FLAG_WRITE))
		return do_read_fault(vmf, pgoff);
3367
	if (!(vma->vm_flags & VM_SHARED))
J
Jan Kara 已提交
3368 3369
		return do_cow_fault(vmf, pgoff);
	return do_shared_fault(vmf, pgoff);
3370 3371
}

3372
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3373 3374
				unsigned long addr, int page_nid,
				int *flags)
3375 3376 3377 3378
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3379
	if (page_nid == numa_node_id()) {
3380
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3381 3382
		*flags |= TNF_FAULT_LOCAL;
	}
3383 3384 3385 3386

	return mpol_misplaced(page, vma, addr);
}

J
Jan Kara 已提交
3387
static int do_numa_page(struct vm_fault *vmf, pte_t pte)
3388
{
J
Jan Kara 已提交
3389
	struct vm_area_struct *vma = vmf->vma;
3390
	struct page *page = NULL;
3391
	int page_nid = -1;
3392
	int last_cpupid;
3393
	int target_nid;
3394
	bool migrated = false;
3395
	bool was_writable = pte_write(pte);
3396
	int flags = 0;
3397 3398 3399 3400 3401 3402

	/*
	* 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.
	*
3403 3404 3405
	* 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.
3406
	*/
J
Jan Kara 已提交
3407 3408 3409 3410
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
	if (unlikely(!pte_same(*vmf->pte, pte))) {
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3411 3412 3413
		goto out;
	}

3414 3415 3416
	/* Make it present again */
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3417 3418
	if (was_writable)
		pte = pte_mkwrite(pte);
J
Jan Kara 已提交
3419 3420
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
	update_mmu_cache(vma, vmf->address, vmf->pte);
3421

J
Jan Kara 已提交
3422
	page = vm_normal_page(vma, vmf->address, pte);
3423
	if (!page) {
J
Jan Kara 已提交
3424
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3425 3426 3427
		return 0;
	}

3428 3429
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3430
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3431 3432 3433
		return 0;
	}

3434
	/*
3435 3436 3437 3438 3439 3440
	 * 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.
3441
	 */
3442
	if (!pte_write(pte))
3443 3444
		flags |= TNF_NO_GROUP;

3445 3446 3447 3448 3449 3450 3451
	/*
	 * 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;

3452
	last_cpupid = page_cpupid_last(page);
3453
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3454
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3455
			&flags);
J
Jan Kara 已提交
3456
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3457 3458 3459 3460 3461 3462
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3463
	migrated = migrate_misplaced_page(page, vma, target_nid);
3464
	if (migrated) {
3465
		page_nid = target_nid;
3466
		flags |= TNF_MIGRATED;
3467 3468
	} else
		flags |= TNF_MIGRATE_FAIL;
3469 3470

out:
3471
	if (page_nid != -1)
3472
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3473 3474 3475
	return 0;
}

J
Jan Kara 已提交
3476
static int create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3477
{
J
Jan Kara 已提交
3478
	struct vm_area_struct *vma = vmf->vma;
3479
	if (vma_is_anonymous(vma))
J
Jan Kara 已提交
3480
		return do_huge_pmd_anonymous_page(vmf);
M
Matthew Wilcox 已提交
3481
	if (vma->vm_ops->pmd_fault)
J
Jan Kara 已提交
3482 3483
		return vma->vm_ops->pmd_fault(vma, vmf->address, vmf->pmd,
				vmf->flags);
M
Matthew Wilcox 已提交
3484 3485 3486
	return VM_FAULT_FALLBACK;
}

J
Jan Kara 已提交
3487
static int wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3488
{
J
Jan Kara 已提交
3489 3490 3491 3492 3493
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
	if (vmf->vma->vm_ops->pmd_fault)
		return vmf->vma->vm_ops->pmd_fault(vmf->vma, vmf->address,
						   vmf->pmd, vmf->flags);
K
Kirill A. Shutemov 已提交
3494 3495

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3496 3497
	VM_BUG_ON_VMA(vmf->vma->vm_flags & VM_SHARED, vmf->vma);
	__split_huge_pmd(vmf->vma, vmf->pmd, vmf->address, false, NULL);
K
Kirill A. Shutemov 已提交
3498

M
Matthew Wilcox 已提交
3499 3500 3501
	return VM_FAULT_FALLBACK;
}

3502 3503 3504 3505 3506
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
3507 3508 3509 3510 3511 3512 3513 3514 3515
/*
 * 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).
 *
3516 3517
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3518
 *
3519 3520
 * 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
3521
 */
J
Jan Kara 已提交
3522
static int handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3523 3524 3525
{
	pte_t entry;

J
Jan Kara 已提交
3526
	if (unlikely(pmd_none(*vmf->pmd))) {
3527 3528 3529 3530 3531 3532
		/*
		 * 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.
		 */
J
Jan Kara 已提交
3533
		vmf->pte = NULL;
3534 3535
	} else {
		/* See comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3536
		if (pmd_trans_unstable(vmf->pmd) || pmd_devmap(*vmf->pmd))
3537 3538 3539 3540 3541 3542 3543
			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().
		 */
J
Jan Kara 已提交
3544
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
3545

J
Jan Kara 已提交
3546
		entry = *vmf->pte;
3547 3548 3549 3550 3551 3552 3553 3554 3555 3556

		/*
		 * 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();
3557
		if (pte_none(entry)) {
J
Jan Kara 已提交
3558 3559
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3560
		}
L
Linus Torvalds 已提交
3561 3562
	}

J
Jan Kara 已提交
3563 3564 3565
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3566
		else
J
Jan Kara 已提交
3567
			return do_fault(vmf);
3568 3569 3570
	}

	if (!pte_present(entry))
J
Jan Kara 已提交
3571
		return do_swap_page(vmf, entry);
3572

J
Jan Kara 已提交
3573 3574
	if (pte_protnone(entry) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf, entry);
3575

J
Jan Kara 已提交
3576 3577 3578
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
	if (unlikely(!pte_same(*vmf->pte, entry)))
3579
		goto unlock;
J
Jan Kara 已提交
3580
	if (vmf->flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3581
		if (!pte_write(entry))
J
Jan Kara 已提交
3582
			return do_wp_page(vmf, entry);
L
Linus Torvalds 已提交
3583 3584 3585
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3586 3587 3588
	if (ptep_set_access_flags(vmf->vma, vmf->address, vmf->pte, entry,
				vmf->flags & FAULT_FLAG_WRITE)) {
		update_mmu_cache(vmf->vma, vmf->address, vmf->pte);
3589 3590 3591 3592 3593 3594 3595
	} 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.
		 */
J
Jan Kara 已提交
3596 3597
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3598
	}
3599
unlock:
J
Jan Kara 已提交
3600
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3601
	return 0;
L
Linus Torvalds 已提交
3602 3603 3604 3605
}

/*
 * By the time we get here, we already hold the mm semaphore
3606 3607 3608
 *
 * 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
3609
 */
3610 3611
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3612
{
J
Jan Kara 已提交
3613
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3614 3615 3616 3617
		.vma = vma,
		.address = address,
		.flags = flags,
	};
3618
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3619 3620 3621 3622 3623 3624
	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
3625
		return VM_FAULT_OOM;
J
Jan Kara 已提交
3626 3627
	vmf.pmd = pmd_alloc(mm, pud, address);
	if (!vmf.pmd)
H
Hugh Dickins 已提交
3628
		return VM_FAULT_OOM;
J
Jan Kara 已提交
3629 3630
	if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) {
		int ret = create_huge_pmd(&vmf);
3631 3632
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3633
	} else {
J
Jan Kara 已提交
3634
		pmd_t orig_pmd = *vmf.pmd;
3635 3636
		int ret;

3637
		barrier();
3638
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
3639
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
3640
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
3641

J
Jan Kara 已提交
3642
			if ((vmf.flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3643
					!pmd_write(orig_pmd)) {
J
Jan Kara 已提交
3644
				ret = wp_huge_pmd(&vmf, orig_pmd);
3645 3646
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3647
			} else {
J
Jan Kara 已提交
3648
				huge_pmd_set_accessed(&vmf, orig_pmd);
3649
				return 0;
3650
			}
3651 3652 3653
		}
	}

J
Jan Kara 已提交
3654
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
3655 3656
}

3657 3658 3659 3660 3661 3662
/*
 * 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().
 */
3663 3664
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
3665 3666 3667 3668 3669 3670
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
3671
	mem_cgroup_count_vm_event(vma->vm_mm, PGFAULT);
3672 3673 3674 3675 3676 3677 3678 3679 3680

	/* 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)
3681
		mem_cgroup_oom_enable();
3682

K
Kirill A. Shutemov 已提交
3683 3684 3685 3686 3687 3688 3689 3690 3691
	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);
3692

3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703
	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);
	}
3704

3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717
	/*
	 * 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;

3718 3719
	return ret;
}
3720
EXPORT_SYMBOL_GPL(handle_mm_fault);
3721

L
Linus Torvalds 已提交
3722 3723 3724
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3725
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3726
 */
3727
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3728
{
H
Hugh Dickins 已提交
3729 3730
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3731
		return -ENOMEM;
L
Linus Torvalds 已提交
3732

3733 3734
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3735
	spin_lock(&mm->page_table_lock);
3736
	if (pgd_present(*pgd))		/* Another has populated it */
3737
		pud_free(mm, new);
3738 3739
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3740
	spin_unlock(&mm->page_table_lock);
3741
	return 0;
L
Linus Torvalds 已提交
3742 3743 3744 3745 3746 3747
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3748
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3749
 */
3750
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3751
{
H
Hugh Dickins 已提交
3752 3753
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3754
		return -ENOMEM;
L
Linus Torvalds 已提交
3755

3756 3757
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3758
	spin_lock(&mm->page_table_lock);
L
Linus Torvalds 已提交
3759
#ifndef __ARCH_HAS_4LEVEL_HACK
3760 3761
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
3762
		pud_populate(mm, pud, new);
3763
	} else	/* Another has populated it */
3764
		pmd_free(mm, new);
3765 3766 3767
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
3768
		pgd_populate(mm, pud, new);
3769 3770
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
3771
#endif /* __ARCH_HAS_4LEVEL_HACK */
H
Hugh Dickins 已提交
3772
	spin_unlock(&mm->page_table_lock);
3773
	return 0;
3774
}
L
Linus Torvalds 已提交
3775 3776
#endif /* __PAGETABLE_PMD_FOLDED */

3777
static int __follow_pte(struct mm_struct *mm, unsigned long address,
J
Johannes Weiner 已提交
3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793
		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);
3794
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
	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;
}

3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825
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
3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854
/**
 * 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);

3855
#ifdef CONFIG_HAVE_IOREMAP_PROT
3856 3857 3858
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
3859
{
3860
	int ret = -EINVAL;
3861 3862 3863
	pte_t *ptep, pte;
	spinlock_t *ptl;

3864 3865
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
3866

3867
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
3868
		goto out;
3869
	pte = *ptep;
3870

3871 3872 3873 3874
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
3875
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
3876

3877
	ret = 0;
3878 3879 3880
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
3881
	return ret;
3882 3883 3884 3885 3886 3887 3888
}

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
3889
	void __iomem *maddr;
3890 3891
	int offset = addr & (PAGE_SIZE-1);

3892
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
3893 3894
		return -EINVAL;

3895
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
3896 3897 3898 3899 3900 3901 3902 3903
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
3904
EXPORT_SYMBOL_GPL(generic_access_phys);
3905 3906
#endif

3907
/*
3908 3909
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
3910
 */
3911
static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
3912
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
3913 3914 3915
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
3916
	int write = gup_flags & FOLL_WRITE;
3917 3918

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
3919
	/* ignore errors, just check how much was successfully transferred */
3920 3921 3922
	while (len) {
		int bytes, ret, offset;
		void *maddr;
3923
		struct page *page = NULL;
3924

3925
		ret = get_user_pages_remote(tsk, mm, addr, 1,
3926
				gup_flags, &page, &vma, NULL);
3927
		if (ret <= 0) {
3928 3929 3930
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
3931 3932 3933 3934 3935
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
3936
			if (!vma || vma->vm_start > addr)
3937 3938 3939 3940 3941 3942 3943
				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;
3944
#endif
3945
		} else {
3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960
			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);
3961
			put_page(page);
3962 3963 3964 3965 3966 3967 3968 3969 3970
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
3971

S
Stephen Wilson 已提交
3972
/**
3973
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
3974 3975 3976 3977
 * @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
3978
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
3979 3980 3981 3982
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
3983
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
3984
{
3985
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
3986 3987
}

3988 3989 3990 3991 3992 3993
/*
 * 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,
3994
		void *buf, int len, unsigned int gup_flags)
3995 3996 3997 3998 3999 4000 4001 4002
{
	struct mm_struct *mm;
	int ret;

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

4003
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4004

4005 4006 4007 4008
	mmput(mm);

	return ret;
}
4009
EXPORT_SYMBOL_GPL(access_process_vm);
4010

4011 4012 4013 4014 4015 4016 4017 4018
/*
 * 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;

4019 4020 4021 4022 4023 4024 4025
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

4026 4027 4028 4029 4030 4031
	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
4032
			char *p;
4033

M
Miklos Szeredi 已提交
4034
			p = file_path(f, buf, PAGE_SIZE);
4035 4036
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4037
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4038 4039 4040 4041 4042
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4043
	up_read(&mm->mmap_sem);
4044
}
4045

4046
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4047
void __might_fault(const char *file, int line)
4048
{
4049 4050 4051 4052 4053 4054 4055 4056
	/*
	 * 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;
4057
	if (pagefault_disabled())
4058
		return;
4059 4060
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4061
	if (current->mm)
4062
		might_lock_read(&current->mm->mmap_sem);
4063
#endif
4064
}
4065
EXPORT_SYMBOL(__might_fault);
4066
#endif
A
Andrea Arcangeli 已提交
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 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137

#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 */
4138

4139
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4140 4141 4142 4143 4144 4145 4146 4147 4148

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);
}

4149
bool ptlock_alloc(struct page *page)
4150 4151 4152
{
	spinlock_t *ptl;

4153
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4154 4155
	if (!ptl)
		return false;
4156
	page->ptl = ptl;
4157 4158 4159
	return true;
}

4160
void ptlock_free(struct page *page)
4161
{
4162
	kmem_cache_free(page_ptl_cachep, page->ptl);
4163 4164
}
#endif