memory.c 114.5 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
#include <asm/pgalloc.h>
71
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
72 73 74 75
#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_VMA(next-addr != HPAGE_PMD_SIZE, vma);
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
		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);
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
		if (pud_trans_huge(*src_pud) || pud_devmap(*src_pud)) {
			int err;

			VM_BUG_ON_VMA(next-addr != HPAGE_PUD_SIZE, vma);
			err = copy_huge_pud(dst_mm, src_mm,
					    dst_pud, src_pud, addr, vma);
			if (err == -ENOMEM)
				return -ENOMEM;
			if (!err)
				continue;
			/* fall through */
		}
L
Linus Torvalds 已提交
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
		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;
1063 1064 1065
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1066
	int ret;
L
Linus Torvalds 已提交
1067

1068 1069 1070 1071 1072 1073
	/*
	 * 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.
	 */
1074 1075 1076
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1077

L
Linus Torvalds 已提交
1078 1079 1080
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1081
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1082 1083 1084 1085
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1086
		ret = track_pfn_copy(vma);
1087 1088 1089 1090
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1091 1092 1093 1094 1095 1096
	/*
	 * 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.
	 */
1097 1098 1099 1100 1101 1102
	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
1103 1104

	ret = 0;
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110
	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
1111 1112 1113 1114 1115
		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
1116
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1117

1118 1119
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1120
	return ret;
L
Linus Torvalds 已提交
1121 1122
}

1123
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1124
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1125
				unsigned long addr, unsigned long end,
1126
				struct zap_details *details)
L
Linus Torvalds 已提交
1127
{
N
Nick Piggin 已提交
1128
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1129
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1130
	int rss[NR_MM_COUNTERS];
1131
	spinlock_t *ptl;
1132
	pte_t *start_pte;
1133
	pte_t *pte;
1134
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1135

1136
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1137
again:
1138
	init_rss_vec(rss);
1139 1140
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1141
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1142 1143
	do {
		pte_t ptent = *pte;
1144
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1145
			continue;
1146
		}
1147

L
Linus Torvalds 已提交
1148
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1149
			struct page *page;
1150

1151
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1152 1153 1154 1155 1156 1157 1158
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1159
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1160 1161
					continue;
			}
N
Nick Piggin 已提交
1162
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1163
							tlb->fullmm);
L
Linus Torvalds 已提交
1164 1165 1166
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1167 1168

			if (!PageAnon(page)) {
1169 1170
				if (pte_dirty(ptent)) {
					force_flush = 1;
1171
					set_page_dirty(page);
1172
				}
1173
				if (pte_young(ptent) &&
1174
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1175
					mark_page_accessed(page);
1176
			}
1177
			rss[mm_counter(page)]--;
1178
			page_remove_rmap(page, false);
1179 1180
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1181
			if (unlikely(__tlb_remove_page(tlb, page))) {
1182
				force_flush = 1;
1183
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1184
				break;
1185
			}
L
Linus Torvalds 已提交
1186 1187
			continue;
		}
1188 1189
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1190
			continue;
K
KAMEZAWA Hiroyuki 已提交
1191

1192 1193 1194 1195 1196
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1197

1198
			page = migration_entry_to_page(entry);
1199
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1200
		}
1201 1202
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1203
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1204
	} while (pte++, addr += PAGE_SIZE, addr != end);
1205

K
KAMEZAWA Hiroyuki 已提交
1206
	add_mm_rss_vec(mm, rss);
1207
	arch_leave_lazy_mmu_mode();
1208

1209
	/* Do the actual TLB flush before dropping ptl */
1210
	if (force_flush)
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
		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);
1223
		if (addr != end)
P
Peter Zijlstra 已提交
1224 1225 1226
			goto again;
	}

1227
	return addr;
L
Linus Torvalds 已提交
1228 1229
}

1230
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1231
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1232
				unsigned long addr, unsigned long end,
1233
				struct zap_details *details)
L
Linus Torvalds 已提交
1234 1235 1236 1237 1238 1239 1240
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1241
		if (pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1242
			if (next - addr != HPAGE_PMD_SIZE) {
1243 1244
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1245
				__split_huge_pmd(vma, pmd, addr, false, NULL);
S
Shaohua Li 已提交
1246
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1247
				goto next;
1248 1249
			/* fall through */
		}
1250 1251 1252 1253 1254 1255 1256 1257 1258
		/*
		 * 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;
1259
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1260
next:
1261 1262
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1263 1264

	return addr;
L
Linus Torvalds 已提交
1265 1266
}

1267
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1268
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1269
				unsigned long addr, unsigned long end,
1270
				struct zap_details *details)
L
Linus Torvalds 已提交
1271 1272 1273 1274 1275 1276 1277
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1278 1279 1280 1281 1282 1283 1284 1285
		if (pud_trans_huge(*pud) || pud_devmap(*pud)) {
			if (next - addr != HPAGE_PUD_SIZE) {
				VM_BUG_ON_VMA(!rwsem_is_locked(&tlb->mm->mmap_sem), vma);
				split_huge_pud(vma, pud, addr);
			} else if (zap_huge_pud(tlb, vma, pud, addr))
				goto next;
			/* fall through */
		}
1286
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1287
			continue;
1288
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1289 1290
next:
		cond_resched();
1291
	} while (pud++, addr = next, addr != end);
1292 1293

	return addr;
L
Linus Torvalds 已提交
1294 1295
}

M
Michal Hocko 已提交
1296
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1297 1298 1299
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1300 1301 1302 1303 1304 1305 1306 1307 1308
{
	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);
1309
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1310
			continue;
1311 1312
		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1313 1314
	tlb_end_vma(tlb, vma);
}
1315

1316 1317 1318

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1319
		unsigned long end_addr,
1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
		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;

1331 1332 1333
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1334
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1335
		untrack_pfn(vma, 0, 0);
1336 1337 1338 1339 1340 1341 1342

	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
1343
			 * cleanup path of mmap_region. When
1344
			 * hugetlbfs ->mmap method fails,
1345
			 * mmap_region() nullifies vma->vm_file
1346 1347 1348 1349
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1350
			if (vma->vm_file) {
1351
				i_mmap_lock_write(vma->vm_file->f_mapping);
1352
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1353
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1354
			}
1355 1356 1357
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1358 1359 1360 1361
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1362
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1363 1364 1365 1366
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1367
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377
 *
 * 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
1378
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1379
		struct vm_area_struct *vma, unsigned long start_addr,
1380
		unsigned long end_addr)
L
Linus Torvalds 已提交
1381
{
A
Andrea Arcangeli 已提交
1382
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1383

A
Andrea Arcangeli 已提交
1384
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1385
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1386
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1387
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1388 1389 1390 1391 1392
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1393
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1394
 * @size: number of bytes to zap
1395 1396
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1397
 */
1398
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1399
		unsigned long size)
L
Linus Torvalds 已提交
1400 1401
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1402
	struct mmu_gather tlb;
1403
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1404 1405

	lru_add_drain();
1406
	tlb_gather_mmu(&tlb, mm, start, end);
1407
	update_hiwater_rss(mm);
1408 1409
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1410
		unmap_single_vma(&tlb, vma, start, end, NULL);
1411 1412
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1413 1414
}

1415 1416 1417 1418 1419
/**
 * 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
1420
 * @details: details of shared cache invalidation
1421 1422
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1423
 */
1424
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1425 1426 1427
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1428
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1429 1430 1431
	unsigned long end = address + size;

	lru_add_drain();
1432
	tlb_gather_mmu(&tlb, mm, address, end);
1433
	update_hiwater_rss(mm);
1434
	mmu_notifier_invalidate_range_start(mm, address, end);
1435
	unmap_single_vma(&tlb, vma, address, end, details);
1436
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1437
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1438 1439
}

1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
/**
 * 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;
1458
	zap_page_range_single(vma, address, size, NULL);
1459 1460 1461 1462
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1463
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1464
			spinlock_t **ptl)
1465 1466 1467 1468
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
1469
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
1470 1471
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
1472
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
1473
		}
1474 1475 1476 1477
	}
	return NULL;
}

1478 1479 1480 1481 1482 1483 1484
/*
 * 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
1485 1486
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1487
{
N
Nick Piggin 已提交
1488
	struct mm_struct *mm = vma->vm_mm;
1489
	int retval;
1490
	pte_t *pte;
1491 1492
	spinlock_t *ptl;

1493
	retval = -EINVAL;
1494
	if (PageAnon(page))
1495
		goto out;
1496 1497
	retval = -ENOMEM;
	flush_dcache_page(page);
1498
	pte = get_locked_pte(mm, addr, &ptl);
1499
	if (!pte)
1500
		goto out;
1501 1502 1503 1504 1505 1506
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1507
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1508
	page_add_file_rmap(page, false);
1509 1510 1511
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1512 1513
	pte_unmap_unlock(pte, ptl);
	return retval;
1514 1515 1516 1517 1518 1519
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1520 1521 1522 1523 1524 1525
/**
 * 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
 *
1526 1527 1528 1529 1530 1531
 * 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
1532
 * (see split_page()).
1533 1534 1535 1536 1537 1538 1539 1540
 *
 * 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.
1541 1542 1543 1544 1545
 *
 * 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.
1546
 */
N
Nick Piggin 已提交
1547 1548
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1549 1550 1551 1552 1553
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1554 1555 1556 1557 1558
	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
1559
	return insert_page(vma, addr, page, vma->vm_page_prot);
1560
}
1561
EXPORT_SYMBOL(vm_insert_page);
1562

N
Nick Piggin 已提交
1563
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1564
			pfn_t pfn, pgprot_t prot)
N
Nick Piggin 已提交
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
{
	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.. */
1580 1581 1582 1583
	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
1584
	set_pte_at(mm, addr, pte, entry);
1585
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1586 1587 1588 1589 1590 1591 1592 1593

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

N
Nick Piggin 已提交
1594 1595 1596 1597 1598 1599
/**
 * 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
 *
1600
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1601 1602 1603 1604
 * 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
1605 1606 1607 1608 1609
 *
 * 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
1610 1611
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1612
			unsigned long pfn)
A
Andy Lutomirski 已提交
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
{
	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
1635
{
1636
	int ret;
N
Nick Piggin 已提交
1637 1638 1639 1640 1641 1642
	/*
	 * 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
1643 1644 1645 1646 1647
	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
1648

N
Nick Piggin 已提交
1649 1650
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1651 1652

	track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
1653

1654
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot);
1655 1656

	return ret;
N
Nick Piggin 已提交
1657
}
A
Andy Lutomirski 已提交
1658
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1659

N
Nick Piggin 已提交
1660
int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1661
			pfn_t pfn)
N
Nick Piggin 已提交
1662
{
1663 1664
	pgprot_t pgprot = vma->vm_page_prot;

N
Nick Piggin 已提交
1665
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1666

N
Nick Piggin 已提交
1667 1668
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1669 1670

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1671

N
Nick Piggin 已提交
1672 1673 1674 1675
	/*
	 * 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
1676 1677
	 * 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
1678
	 */
1679
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1680 1681
		struct page *page;

1682 1683 1684 1685 1686 1687
		/*
		 * 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));
1688
		return insert_page(vma, addr, page, pgprot);
N
Nick Piggin 已提交
1689
	}
1690
	return insert_pfn(vma, addr, pfn, pgprot);
N
Nick Piggin 已提交
1691
}
N
Nick Piggin 已提交
1692
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1693

L
Linus Torvalds 已提交
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703
/*
 * 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
1704
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1705

H
Hugh Dickins 已提交
1706
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1707 1708
	if (!pte)
		return -ENOMEM;
1709
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1710 1711
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1712
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1713 1714
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1715
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1716
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
	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;
1731
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760
	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;
}

1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
/**
 * 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
1771 1772 1773 1774 1775
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;
1776
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1777
	struct mm_struct *mm = vma->vm_mm;
1778
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1779 1780 1781 1782 1783 1784 1785
	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).
1786 1787 1788
	 *   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.
1789 1790 1791 1792
	 *   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
1793 1794 1795 1796
	 *
	 * 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".
1797
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1798
	 */
1799 1800 1801
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1802
		vma->vm_pgoff = pfn;
1803 1804
	}

1805
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1806
	if (err)
1807
		return -EINVAL;
L
Linus Torvalds 已提交
1808

1809
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821

	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);
1822 1823

	if (err)
1824
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1825

L
Linus Torvalds 已提交
1826 1827 1828 1829
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
/**
 * 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);

1877 1878 1879 1880 1881 1882
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;
1883
	pgtable_t token;
1884
	spinlock_t *uninitialized_var(ptl);
1885 1886 1887 1888 1889 1890 1891 1892 1893

	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));

1894 1895
	arch_enter_lazy_mmu_mode();

1896
	token = pmd_pgtable(*pmd);
1897 1898

	do {
1899
		err = fn(pte++, token, addr, data);
1900 1901
		if (err)
			break;
1902
	} while (addr += PAGE_SIZE, addr != end);
1903

1904 1905
	arch_leave_lazy_mmu_mode();

1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
	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
1919 1920
	BUG_ON(pud_huge(*pud));

1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
	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;
1962
	unsigned long end = addr + size;
1963 1964
	int err;

1965 1966 1967
	if (WARN_ON(addr >= end))
		return -EINVAL;

1968 1969 1970 1971 1972 1973 1974
	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);
1975

1976 1977 1978 1979
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

1980
/*
1981 1982 1983 1984 1985
 * 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;
1986
 * and do_anonymous_page can safely check later on).
1987
 */
H
Hugh Dickins 已提交
1988
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
1989 1990 1991 1992 1993
				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
1994 1995
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
1996
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
1997
		spin_unlock(ptl);
1998 1999 2000 2001 2002 2003
	}
#endif
	pte_unmap(page_table);
	return same;
}

2004
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2005
{
2006 2007
	debug_dma_assert_idle(src);

2008 2009 2010 2011 2012 2013 2014
	/*
	 * 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)) {
2015
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2016 2017 2018 2019 2020 2021 2022 2023 2024
		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))
2025
			clear_page(kaddr);
2026
		kunmap_atomic(kaddr);
2027
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2028 2029
	} else
		copy_user_highpage(dst, src, va, vma);
2030 2031
}

2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
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;
}

2046 2047 2048 2049 2050 2051
/*
 * 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.
 */
2052
static int do_page_mkwrite(struct vm_fault *vmf)
2053 2054
{
	int ret;
2055 2056
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2057

2058
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2059

2060
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2061 2062
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076
	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;
}

2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
/*
 * Handle dirtying of a page in shared file mapping on a write fault.
 *
 * The function expects the page to be locked and unlocks it.
 */
static void fault_dirty_shared_page(struct vm_area_struct *vma,
				    struct page *page)
{
	struct address_space *mapping;
	bool dirtied;
	bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;

	dirtied = set_page_dirty(page);
	VM_BUG_ON_PAGE(PageAnon(page), page);
	/*
	 * 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.
	 */
	mapping = page_rmapping(page);
	unlock_page(page);

	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);
}

2112 2113 2114 2115 2116 2117 2118 2119
/*
 * 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.
 */
2120
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2121
	__releases(vmf->ptl)
2122
{
J
Jan Kara 已提交
2123
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2124
	struct page *page = vmf->page;
2125 2126 2127 2128 2129 2130 2131 2132 2133
	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
2134 2135
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2136
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2137 2138 2139
	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);
2140 2141
}

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
/*
 * 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
2158
static int wp_page_copy(struct vm_fault *vmf)
2159
{
J
Jan Kara 已提交
2160
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2161
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2162
	struct page *old_page = vmf->page;
2163 2164 2165
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
J
Jan Kara 已提交
2166
	const unsigned long mmun_start = vmf->address & PAGE_MASK;
K
Kirill A. Shutemov 已提交
2167
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2168 2169 2170 2171 2172
	struct mem_cgroup *memcg;

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

J
Jan Kara 已提交
2173
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2174 2175
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2176 2177 2178
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2179
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2180
				vmf->address);
2181 2182
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2183
		cow_user_page(new_page, old_page, vmf->address, vma);
2184 2185
	}

2186
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2187 2188
		goto oom_free_new;

2189 2190
	__SetPageUptodate(new_page);

2191 2192 2193 2194 2195
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2196
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2197
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2198 2199
		if (old_page) {
			if (!PageAnon(old_page)) {
2200 2201
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2202 2203 2204 2205 2206
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2207
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2208 2209 2210 2211 2212 2213 2214 2215
		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
2216 2217
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2218
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2219 2220 2221 2222 2223 2224
		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
2225 2226
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249
		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.
			 */
2250
			page_remove_rmap(old_page, false);
2251 2252 2253 2254 2255 2256
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2257
		mem_cgroup_cancel_charge(new_page, memcg, false);
2258 2259 2260
	}

	if (new_page)
2261
		put_page(new_page);
2262

J
Jan Kara 已提交
2263
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2264 2265 2266 2267 2268 2269 2270 2271
	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 */
2272 2273
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2274 2275
			unlock_page(old_page);
		}
2276
		put_page(old_page);
2277 2278 2279
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2280
	put_page(new_page);
2281 2282
oom:
	if (old_page)
2283
		put_page(old_page);
2284 2285 2286
	return VM_FAULT_OOM;
}

2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
/**
 * finish_mkwrite_fault - finish page fault for a shared mapping, making PTE
 *			  writeable once the page is prepared
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a write page fault in a
 * shared mapping due to PTE being read-only once the mapped page is prepared.
 * It handles locking of PTE and modifying it. The function returns
 * VM_FAULT_WRITE on success, 0 when PTE got changed before we acquired PTE
 * lock.
 *
 * The function expects the page to be locked or other protection against
 * concurrent faults / writeback (such as DAX radix tree locks).
 */
int finish_mkwrite_fault(struct vm_fault *vmf)
{
	WARN_ON_ONCE(!(vmf->vma->vm_flags & VM_SHARED));
	vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd, vmf->address,
				       &vmf->ptl);
	/*
	 * We might have raced with another page fault while we released the
	 * pte_offset_map_lock.
	 */
	if (!pte_same(*vmf->pte, vmf->orig_pte)) {
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2313
		return VM_FAULT_NOPAGE;
2314 2315
	}
	wp_page_reuse(vmf);
2316
	return 0;
2317 2318
}

2319 2320 2321 2322
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
J
Jan Kara 已提交
2323
static int wp_pfn_shared(struct vm_fault *vmf)
2324
{
J
Jan Kara 已提交
2325
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2326

2327 2328 2329
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
		int ret;

J
Jan Kara 已提交
2330
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2331
		vmf->flags |= FAULT_FLAG_MKWRITE;
2332
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2333
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2334
			return ret;
2335
		return finish_mkwrite_fault(vmf);
2336
	}
2337 2338
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2339 2340
}

J
Jan Kara 已提交
2341
static int wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2342
	__releases(vmf->ptl)
2343
{
J
Jan Kara 已提交
2344
	struct vm_area_struct *vma = vmf->vma;
2345

J
Jan Kara 已提交
2346
	get_page(vmf->page);
2347 2348 2349 2350

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

J
Jan Kara 已提交
2351
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2352
		tmp = do_page_mkwrite(vmf);
2353 2354
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2355
			put_page(vmf->page);
2356 2357
			return tmp;
		}
2358
		tmp = finish_mkwrite_fault(vmf);
2359
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2360 2361
			unlock_page(vmf->page);
			put_page(vmf->page);
2362
			return tmp;
2363
		}
2364 2365
	} else {
		wp_page_reuse(vmf);
2366
		lock_page(vmf->page);
2367
	}
2368 2369
	fault_dirty_shared_page(vma, vmf->page);
	put_page(vmf->page);
2370

2371
	return VM_FAULT_WRITE;
2372 2373
}

L
Linus Torvalds 已提交
2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
/*
 * 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.
 *
2388 2389 2390
 * 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
2391
 */
J
Jan Kara 已提交
2392
static int do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2393
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2394
{
J
Jan Kara 已提交
2395
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2396

J
Jan Kara 已提交
2397 2398
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2399
		/*
2400 2401
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2402 2403
		 *
		 * We should not cow pages in a shared writeable mapping.
2404
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2405 2406 2407
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2408
			return wp_pfn_shared(vmf);
2409

J
Jan Kara 已提交
2410
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2411
		return wp_page_copy(vmf);
2412
	}
L
Linus Torvalds 已提交
2413

2414
	/*
P
Peter Zijlstra 已提交
2415 2416
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2417
	 */
J
Jan Kara 已提交
2418
	if (PageAnon(vmf->page) && !PageKsm(vmf->page)) {
2419
		int total_mapcount;
J
Jan Kara 已提交
2420 2421
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2422
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2423
			lock_page(vmf->page);
J
Jan Kara 已提交
2424 2425
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2426
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2427
				unlock_page(vmf->page);
J
Jan Kara 已提交
2428
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2429
				put_page(vmf->page);
2430
				return 0;
2431
			}
J
Jan Kara 已提交
2432
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2433
		}
J
Jan Kara 已提交
2434
		if (reuse_swap_page(vmf->page, &total_mapcount)) {
2435 2436 2437 2438 2439 2440 2441 2442
			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.
				 */
J
Jan Kara 已提交
2443
				page_move_anon_rmap(vmf->page, vma);
2444
			}
J
Jan Kara 已提交
2445
			unlock_page(vmf->page);
2446 2447
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2448
		}
J
Jan Kara 已提交
2449
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2450
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2451
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2452
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2453 2454 2455 2456 2457
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2458
	get_page(vmf->page);
2459

J
Jan Kara 已提交
2460
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2461
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2462 2463
}

2464
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2465 2466 2467
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2468
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2469 2470
}

2471
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2472 2473 2474 2475 2476
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2477
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2478 2479 2480
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2481
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2482 2483 2484 2485 2486 2487 2488
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2489
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2490 2491
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2492
				details);
L
Linus Torvalds 已提交
2493 2494 2495 2496
	}
}

/**
2497 2498 2499 2500
 * 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
2501
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2502 2503
 * @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
2504
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
 * 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
2516
	struct zap_details details = { };
L
Linus Torvalds 已提交
2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533
	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;

2534
	i_mmap_lock_write(mapping);
2535
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2536
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2537
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2538 2539 2540 2541
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2542 2543
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2544 2545 2546 2547
 * 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
2548
 */
J
Jan Kara 已提交
2549
int do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2550
{
J
Jan Kara 已提交
2551
	struct vm_area_struct *vma = vmf->vma;
2552
	struct page *page, *swapcache;
2553
	struct mem_cgroup *memcg;
2554
	swp_entry_t entry;
L
Linus Torvalds 已提交
2555
	pte_t pte;
2556
	int locked;
2557
	int exclusive = 0;
N
Nick Piggin 已提交
2558
	int ret = 0;
L
Linus Torvalds 已提交
2559

J
Jan Kara 已提交
2560
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte))
2561
		goto out;
2562

J
Jan Kara 已提交
2563
	entry = pte_to_swp_entry(vmf->orig_pte);
2564 2565
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2566 2567
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2568 2569 2570
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2571
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2572
			ret = VM_FAULT_SIGBUS;
2573
		}
2574 2575
		goto out;
	}
2576
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
2577 2578
	page = lookup_swap_cache(entry);
	if (!page) {
J
Jan Kara 已提交
2579 2580
		page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE, vma,
					vmf->address);
L
Linus Torvalds 已提交
2581 2582
		if (!page) {
			/*
2583 2584
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2585
			 */
J
Jan Kara 已提交
2586 2587
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2588
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2589
				ret = VM_FAULT_OOM;
2590
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2591
			goto unlock;
L
Linus Torvalds 已提交
2592 2593 2594 2595
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2596
		count_vm_event(PGMAJFAULT);
K
Kirill A. Shutemov 已提交
2597
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
2598
	} else if (PageHWPoison(page)) {
2599 2600 2601 2602
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2603 2604
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2605
		swapcache = page;
2606
		goto out_release;
L
Linus Torvalds 已提交
2607 2608
	}

2609
	swapcache = page;
J
Jan Kara 已提交
2610
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2611

2612
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2613 2614 2615 2616
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2617

A
Andrea Arcangeli 已提交
2618
	/*
2619 2620 2621 2622
	 * 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
2623
	 */
2624
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2625 2626
		goto out_page;

J
Jan Kara 已提交
2627
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2628 2629 2630 2631
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2632 2633
	}

K
Kirill A. Shutemov 已提交
2634 2635
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2636
		ret = VM_FAULT_OOM;
2637
		goto out_page;
2638 2639
	}

L
Linus Torvalds 已提交
2640
	/*
2641
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2642
	 */
J
Jan Kara 已提交
2643 2644
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2645
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2646 2647 2648 2649 2650
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2651 2652
	}

2653 2654 2655 2656 2657 2658 2659 2660 2661
	/*
	 * 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
2662

K
Kirill A. Shutemov 已提交
2663 2664
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2665
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2666
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2667
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2668
		vmf->flags &= ~FAULT_FLAG_WRITE;
2669
		ret |= VM_FAULT_WRITE;
2670
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2671 2672
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
2673
	if (pte_swp_soft_dirty(vmf->orig_pte))
2674
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2675
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
2676
	vmf->orig_pte = pte;
2677
	if (page == swapcache) {
J
Jan Kara 已提交
2678
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
2679
		mem_cgroup_commit_charge(page, memcg, true, false);
2680
		activate_page(page);
2681
	} else { /* ksm created a completely new copy */
J
Jan Kara 已提交
2682
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2683
		mem_cgroup_commit_charge(page, memcg, false, false);
2684 2685
		lru_cache_add_active_or_unevictable(page, vma);
	}
L
Linus Torvalds 已提交
2686

2687
	swap_free(entry);
2688 2689
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2690
		try_to_free_swap(page);
2691
	unlock_page(page);
2692
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2693 2694 2695 2696 2697 2698 2699 2700 2701
		/*
		 * 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);
2702
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2703
	}
2704

J
Jan Kara 已提交
2705
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
2706
		ret |= do_wp_page(vmf);
2707 2708
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2709 2710 2711 2712
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2713
	update_mmu_cache(vma, vmf->address, vmf->pte);
2714
unlock:
J
Jan Kara 已提交
2715
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
2716 2717
out:
	return ret;
2718
out_nomap:
2719
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
2720
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2721
out_page:
2722
	unlock_page(page);
2723
out_release:
2724
	put_page(page);
2725
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2726
		unlock_page(swapcache);
2727
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2728
	}
2729
	return ret;
L
Linus Torvalds 已提交
2730 2731
}

2732
/*
2733 2734
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
2735 2736 2737 2738 2739 2740
 * 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) {
2741 2742 2743 2744 2745 2746 2747 2748 2749 2750
		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;
2751

2752
		return expand_downwards(vma, address - PAGE_SIZE);
2753
	}
2754 2755 2756 2757 2758 2759 2760
	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;

2761
		return expand_upwards(vma, address + PAGE_SIZE);
2762
	}
2763 2764 2765
	return 0;
}

L
Linus Torvalds 已提交
2766
/*
2767 2768 2769
 * 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
2770
 */
J
Jan Kara 已提交
2771
static int do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2772
{
J
Jan Kara 已提交
2773
	struct vm_area_struct *vma = vmf->vma;
2774
	struct mem_cgroup *memcg;
2775
	struct page *page;
L
Linus Torvalds 已提交
2776 2777
	pte_t entry;

2778 2779 2780 2781
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2782
	/* Check if we need to add a guard page to the stack */
J
Jan Kara 已提交
2783
	if (check_stack_guard_page(vma, vmf->address) < 0)
2784
		return VM_FAULT_SIGSEGV;
2785

2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
	/*
	 * 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
2796
	if (pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))
2797 2798 2799
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
2800
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
2801 2802
		return 0;

2803
	/* Use the zero-page for reads */
J
Jan Kara 已提交
2804
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
2805
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
2806
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
2807
						vma->vm_page_prot));
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))
H
Hugh Dickins 已提交
2811
			goto unlock;
2812 2813
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2814 2815
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
2816
		}
H
Hugh Dickins 已提交
2817 2818 2819
		goto setpte;
	}

N
Nick Piggin 已提交
2820 2821 2822
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
2823
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
2824 2825
	if (!page)
		goto oom;
2826

K
Kirill A. Shutemov 已提交
2827
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
2828 2829
		goto oom_free_page;

2830 2831 2832 2833 2834
	/*
	 * 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
2835
	__SetPageUptodate(page);
2836

N
Nick Piggin 已提交
2837
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2838 2839
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2840

J
Jan Kara 已提交
2841 2842 2843
	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
2844
		goto release;
H
Hugh Dickins 已提交
2845

2846 2847
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2848
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2849
		mem_cgroup_cancel_charge(page, memcg, false);
2850
		put_page(page);
J
Jan Kara 已提交
2851
		return handle_userfault(vmf, VM_UFFD_MISSING);
2852 2853
	}

K
Kirill A. Shutemov 已提交
2854
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
2855
	page_add_new_anon_rmap(page, vma, vmf->address, false);
2856
	mem_cgroup_commit_charge(page, memcg, false, false);
2857
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
2858
setpte:
J
Jan Kara 已提交
2859
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
2860 2861

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2862
	update_mmu_cache(vma, vmf->address, vmf->pte);
2863
unlock:
J
Jan Kara 已提交
2864
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
2865
	return 0;
2866
release:
2867
	mem_cgroup_cancel_charge(page, memcg, false);
2868
	put_page(page);
2869
	goto unlock;
2870
oom_free_page:
2871
	put_page(page);
2872
oom:
L
Linus Torvalds 已提交
2873 2874 2875
	return VM_FAULT_OOM;
}

2876 2877 2878 2879 2880
/*
 * 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
2881
static int __do_fault(struct vm_fault *vmf)
2882
{
J
Jan Kara 已提交
2883
	struct vm_area_struct *vma = vmf->vma;
2884 2885
	int ret;

2886
	ret = vma->vm_ops->fault(vmf);
2887
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
2888
			    VM_FAULT_DONE_COW)))
2889
		return ret;
2890

2891
	if (unlikely(PageHWPoison(vmf->page))) {
2892
		if (ret & VM_FAULT_LOCKED)
2893 2894
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
2895
		vmf->page = NULL;
2896 2897 2898 2899
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
2900
		lock_page(vmf->page);
2901
	else
2902
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
2903 2904 2905 2906

	return ret;
}

J
Jan Kara 已提交
2907
static int pte_alloc_one_map(struct vm_fault *vmf)
2908
{
J
Jan Kara 已提交
2909
	struct vm_area_struct *vma = vmf->vma;
2910

J
Jan Kara 已提交
2911
	if (!pmd_none(*vmf->pmd))
2912
		goto map_pte;
J
Jan Kara 已提交
2913 2914 2915 2916
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
2917 2918 2919 2920
			goto map_pte;
		}

		atomic_long_inc(&vma->vm_mm->nr_ptes);
J
Jan Kara 已提交
2921 2922 2923 2924
		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))) {
2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938
		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
2939
	if (pmd_trans_unstable(vmf->pmd) || pmd_devmap(*vmf->pmd))
2940 2941
		return VM_FAULT_NOPAGE;

J
Jan Kara 已提交
2942 2943
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
2944 2945 2946
	return 0;
}

2947
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960

#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
2961
static void deposit_prealloc_pte(struct vm_fault *vmf)
2962
{
J
Jan Kara 已提交
2963
	struct vm_area_struct *vma = vmf->vma;
2964

J
Jan Kara 已提交
2965
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
2966 2967 2968 2969 2970
	/*
	 * 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
2971
	vmf->prealloc_pte = 0;
2972 2973
}

J
Jan Kara 已提交
2974
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
2975
{
J
Jan Kara 已提交
2976 2977 2978
	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
2979 2980 2981 2982 2983 2984 2985 2986 2987
	pmd_t entry;
	int i, ret;

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

2988 2989 2990 2991
	/*
	 * 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
2992 2993 2994
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm, vmf->address);
		if (!vmf->prealloc_pte)
2995 2996 2997 2998
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
2999 3000
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011
		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);
3012 3013 3014 3015
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3016
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3017

J
Jan Kara 已提交
3018
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3019

J
Jan Kara 已提交
3020
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3021 3022 3023

	/* fault is handled */
	ret = 0;
3024
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3025
out:
J
Jan Kara 已提交
3026
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3027 3028 3029
	return ret;
}
#else
J
Jan Kara 已提交
3030
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3031 3032 3033 3034 3035 3036
{
	BUILD_BUG();
	return 0;
}
#endif

3037
/**
3038 3039
 * 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.
3040
 *
J
Jan Kara 已提交
3041
 * @vmf: fault environment
3042
 * @memcg: memcg to charge page (only for private mappings)
3043 3044
 * @page: page to map
 *
J
Jan Kara 已提交
3045 3046
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3047 3048 3049 3050
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
J
Jan Kara 已提交
3051
int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3052
		struct page *page)
3053
{
J
Jan Kara 已提交
3054 3055
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3056
	pte_t entry;
K
Kirill A. Shutemov 已提交
3057 3058
	int ret;

J
Jan Kara 已提交
3059
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3060
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3061 3062 3063
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3064
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3065
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3066
			return ret;
K
Kirill A. Shutemov 已提交
3067
	}
3068

J
Jan Kara 已提交
3069 3070
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3071
		if (ret)
H
Hugh Dickins 已提交
3072
			return ret;
3073 3074 3075
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3076 3077
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3078

3079 3080 3081 3082
	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
3083 3084
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3085
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3086
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3087 3088
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3089
	} else {
3090
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3091
		page_add_file_rmap(page, false);
3092
	}
J
Jan Kara 已提交
3093
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3094 3095

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3096
	update_mmu_cache(vma, vmf->address, vmf->pte);
3097

H
Hugh Dickins 已提交
3098
	return 0;
3099 3100
}

3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132

/**
 * finish_fault - finish page fault once we have prepared the page to fault
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a page fault once the
 * page to fault in is prepared. It handles locking of PTEs, inserts PTE for
 * given page, adds reverse page mapping, handles memcg charges and LRU
 * addition. The function returns 0 on success, VM_FAULT_ code in case of
 * error.
 *
 * The function expects the page to be locked and on success it consumes a
 * reference of a page being mapped (for the PTE which maps it).
 */
int finish_fault(struct vm_fault *vmf)
{
	struct page *page;
	int ret;

	/* Did we COW the page? */
	if ((vmf->flags & FAULT_FLAG_WRITE) &&
	    !(vmf->vma->vm_flags & VM_SHARED))
		page = vmf->cow_page;
	else
		page = vmf->page;
	ret = alloc_set_pte(vmf, vmf->memcg, page);
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3133 3134
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3135 3136 3137

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3138
{
3139
	*val = fault_around_bytes;
3140 3141 3142
	return 0;
}

3143 3144 3145 3146 3147
/*
 * 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.
 */
3148
static int fault_around_bytes_set(void *data, u64 val)
3149
{
3150
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3151
		return -EINVAL;
3152 3153 3154 3155
	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 */
3156 3157
	return 0;
}
3158 3159
DEFINE_SIMPLE_ATTRIBUTE(fault_around_bytes_fops,
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3160 3161 3162 3163 3164

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

3165 3166
	ret = debugfs_create_file("fault_around_bytes", 0644, NULL, NULL,
			&fault_around_bytes_fops);
3167
	if (!ret)
3168
		pr_warn("Failed to create fault_around_bytes in debugfs");
3169 3170 3171 3172
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3173

3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196
/*
 * 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.
 */
3197
static int do_fault_around(struct vm_fault *vmf)
3198
{
J
Jan Kara 已提交
3199
	unsigned long address = vmf->address, nr_pages, mask;
3200
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3201
	pgoff_t end_pgoff;
3202
	int off, ret = 0;
3203

3204
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3205 3206
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3207 3208
	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
3209
	start_pgoff -= off;
3210 3211

	/*
K
Kirill A. Shutemov 已提交
3212 3213
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3214
	 */
K
Kirill A. Shutemov 已提交
3215
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3216
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3217
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3218
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3219
			start_pgoff + nr_pages - 1);
3220

J
Jan Kara 已提交
3221 3222 3223 3224
	if (pmd_none(*vmf->pmd)) {
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm,
						  vmf->address);
		if (!vmf->prealloc_pte)
3225
			goto out;
3226
		smp_wmb(); /* See comment in __pte_alloc() */
3227 3228
	}

J
Jan Kara 已提交
3229
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3230 3231

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3232
	if (pmd_trans_huge(*vmf->pmd)) {
3233 3234 3235 3236 3237
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3238
	if (!vmf->pte)
3239 3240 3241
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3242 3243
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3244
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3245
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3246
out:
J
Jan Kara 已提交
3247 3248
	vmf->address = address;
	vmf->pte = NULL;
3249
	return ret;
3250 3251
}

3252
static int do_read_fault(struct vm_fault *vmf)
3253
{
J
Jan Kara 已提交
3254
	struct vm_area_struct *vma = vmf->vma;
3255 3256 3257 3258 3259 3260 3261
	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).
	 */
3262
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3263
		ret = do_fault_around(vmf);
3264 3265
		if (ret)
			return ret;
3266
	}
3267

J
Jan Kara 已提交
3268
	ret = __do_fault(vmf);
3269 3270 3271
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3272
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3273
	unlock_page(vmf->page);
3274
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3275
		put_page(vmf->page);
3276 3277 3278
	return ret;
}

3279
static int do_cow_fault(struct vm_fault *vmf)
3280
{
J
Jan Kara 已提交
3281
	struct vm_area_struct *vma = vmf->vma;
3282 3283 3284 3285 3286
	int ret;

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

J
Jan Kara 已提交
3287 3288
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3289 3290
		return VM_FAULT_OOM;

J
Jan Kara 已提交
3291
	if (mem_cgroup_try_charge(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3292
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3293
		put_page(vmf->cow_page);
3294 3295 3296
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3297
	ret = __do_fault(vmf);
3298 3299
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3300 3301
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3302

3303
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3304
	__SetPageUptodate(vmf->cow_page);
3305

3306
	ret |= finish_fault(vmf);
3307 3308
	unlock_page(vmf->page);
	put_page(vmf->page);
3309 3310
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3311 3312
	return ret;
uncharge_out:
3313
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3314
	put_page(vmf->cow_page);
3315 3316 3317
	return ret;
}

3318
static int do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3319
{
J
Jan Kara 已提交
3320
	struct vm_area_struct *vma = vmf->vma;
3321
	int ret, tmp;
3322

J
Jan Kara 已提交
3323
	ret = __do_fault(vmf);
3324
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3325
		return ret;
L
Linus Torvalds 已提交
3326 3327

	/*
3328 3329
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3330
	 */
3331
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3332
		unlock_page(vmf->page);
3333
		tmp = do_page_mkwrite(vmf);
3334 3335
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3336
			put_page(vmf->page);
3337
			return tmp;
3338
		}
3339 3340
	}

3341
	ret |= finish_fault(vmf);
3342 3343
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3344 3345
		unlock_page(vmf->page);
		put_page(vmf->page);
3346
		return ret;
L
Linus Torvalds 已提交
3347
	}
N
Nick Piggin 已提交
3348

3349
	fault_dirty_shared_page(vma, vmf->page);
3350
	return ret;
3351
}
3352

3353 3354 3355 3356 3357 3358
/*
 * 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
3359
static int do_fault(struct vm_fault *vmf)
3360
{
J
Jan Kara 已提交
3361
	struct vm_area_struct *vma = vmf->vma;
H
Hugh Dickins 已提交
3362
	int ret;
3363

3364 3365
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
H
Hugh Dickins 已提交
3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379
		ret = VM_FAULT_SIGBUS;
	else if (!(vmf->flags & FAULT_FLAG_WRITE))
		ret = do_read_fault(vmf);
	else if (!(vma->vm_flags & VM_SHARED))
		ret = do_cow_fault(vmf);
	else
		ret = do_shared_fault(vmf);

	/* preallocated pagetable is unused: free it */
	if (vmf->prealloc_pte) {
		pte_free(vma->vm_mm, vmf->prealloc_pte);
		vmf->prealloc_pte = 0;
	}
	return ret;
3380 3381
}

3382
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3383 3384
				unsigned long addr, int page_nid,
				int *flags)
3385 3386 3387 3388
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3389
	if (page_nid == numa_node_id()) {
3390
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3391 3392
		*flags |= TNF_FAULT_LOCAL;
	}
3393 3394 3395 3396

	return mpol_misplaced(page, vma, addr);
}

J
Jan Kara 已提交
3397
static int do_numa_page(struct vm_fault *vmf)
3398
{
J
Jan Kara 已提交
3399
	struct vm_area_struct *vma = vmf->vma;
3400
	struct page *page = NULL;
3401
	int page_nid = -1;
3402
	int last_cpupid;
3403
	int target_nid;
3404
	bool migrated = false;
J
Jan Kara 已提交
3405
	pte_t pte = vmf->orig_pte;
3406
	bool was_writable = pte_write(pte);
3407
	int flags = 0;
3408 3409 3410 3411 3412 3413

	/*
	* 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.
	*
3414 3415 3416
	* 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.
3417
	*/
J
Jan Kara 已提交
3418 3419 3420 3421
	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);
3422 3423 3424
		goto out;
	}

3425 3426 3427
	/* Make it present again */
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3428 3429
	if (was_writable)
		pte = pte_mkwrite(pte);
J
Jan Kara 已提交
3430 3431
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
	update_mmu_cache(vma, vmf->address, vmf->pte);
3432

J
Jan Kara 已提交
3433
	page = vm_normal_page(vma, vmf->address, pte);
3434
	if (!page) {
J
Jan Kara 已提交
3435
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3436 3437 3438
		return 0;
	}

3439 3440
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3441
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3442 3443 3444
		return 0;
	}

3445
	/*
3446 3447 3448 3449 3450 3451
	 * 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.
3452
	 */
3453
	if (!pte_write(pte))
3454 3455
		flags |= TNF_NO_GROUP;

3456 3457 3458 3459 3460 3461 3462
	/*
	 * 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;

3463
	last_cpupid = page_cpupid_last(page);
3464
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3465
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3466
			&flags);
J
Jan Kara 已提交
3467
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3468 3469 3470 3471 3472 3473
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3474
	migrated = migrate_misplaced_page(page, vma, target_nid);
3475
	if (migrated) {
3476
		page_nid = target_nid;
3477
		flags |= TNF_MIGRATED;
3478 3479
	} else
		flags |= TNF_MIGRATE_FAIL;
3480 3481

out:
3482
	if (page_nid != -1)
3483
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3484 3485 3486
	return 0;
}

J
Jan Kara 已提交
3487
static int create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3488
{
3489
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3490
		return do_huge_pmd_anonymous_page(vmf);
3491 3492
	if (vmf->vma->vm_ops->huge_fault)
		return vmf->vma->vm_ops->huge_fault(vmf);
M
Matthew Wilcox 已提交
3493 3494 3495
	return VM_FAULT_FALLBACK;
}

J
Jan Kara 已提交
3496
static int wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3497
{
J
Jan Kara 已提交
3498 3499
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3500 3501
	if (vmf->vma->vm_ops->huge_fault)
		return vmf->vma->vm_ops->huge_fault(vmf);
K
Kirill A. Shutemov 已提交
3502 3503

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3504 3505
	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
3506

M
Matthew Wilcox 已提交
3507 3508 3509
	return VM_FAULT_FALLBACK;
}

3510 3511 3512 3513 3514
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538
static int create_huge_pud(struct vm_fault *vmf)
{
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/* No support for anonymous transparent PUD pages yet */
	if (vma_is_anonymous(vmf->vma))
		return VM_FAULT_FALLBACK;
	if (vmf->vma->vm_ops->huge_fault)
		return vmf->vma->vm_ops->huge_fault(vmf);
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

static int wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
{
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/* No support for anonymous transparent PUD pages yet */
	if (vma_is_anonymous(vmf->vma))
		return VM_FAULT_FALLBACK;
	if (vmf->vma->vm_ops->huge_fault)
		return vmf->vma->vm_ops->huge_fault(vmf);
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3539 3540 3541 3542 3543 3544 3545 3546 3547
/*
 * 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).
 *
3548 3549
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3550
 *
3551 3552
 * 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
3553
 */
J
Jan Kara 已提交
3554
static int handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3555 3556 3557
{
	pte_t entry;

J
Jan Kara 已提交
3558
	if (unlikely(pmd_none(*vmf->pmd))) {
3559 3560 3561 3562 3563 3564
		/*
		 * 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
3565
		vmf->pte = NULL;
3566 3567
	} else {
		/* See comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3568
		if (pmd_trans_unstable(vmf->pmd) || pmd_devmap(*vmf->pmd))
3569 3570 3571 3572 3573 3574 3575
			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
3576
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3577
		vmf->orig_pte = *vmf->pte;
3578 3579 3580 3581 3582 3583 3584 3585 3586 3587

		/*
		 * 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();
J
Jan Kara 已提交
3588
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3589 3590
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3591
		}
L
Linus Torvalds 已提交
3592 3593
	}

J
Jan Kara 已提交
3594 3595 3596
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3597
		else
J
Jan Kara 已提交
3598
			return do_fault(vmf);
3599 3600
	}

J
Jan Kara 已提交
3601 3602
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3603

J
Jan Kara 已提交
3604 3605
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3606

J
Jan Kara 已提交
3607 3608
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3609
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3610
	if (unlikely(!pte_same(*vmf->pte, entry)))
3611
		goto unlock;
J
Jan Kara 已提交
3612
	if (vmf->flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3613
		if (!pte_write(entry))
J
Jan Kara 已提交
3614
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3615 3616 3617
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3618 3619 3620
	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);
3621 3622 3623 3624 3625 3626 3627
	} 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
3628 3629
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3630
	}
3631
unlock:
J
Jan Kara 已提交
3632
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3633
	return 0;
L
Linus Torvalds 已提交
3634 3635 3636 3637
}

/*
 * By the time we get here, we already hold the mm semaphore
3638 3639 3640
 *
 * 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
3641
 */
3642 3643
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3644
{
J
Jan Kara 已提交
3645
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3646
		.vma = vma,
3647
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3648
		.flags = flags,
3649
		.pgoff = linear_page_index(vma, address),
3650
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3651
	};
3652
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3653
	pgd_t *pgd;
3654
	int ret;
L
Linus Torvalds 已提交
3655 3656

	pgd = pgd_offset(mm, address);
3657 3658 3659

	vmf.pud = pud_alloc(mm, pgd, address);
	if (!vmf.pud)
H
Hugh Dickins 已提交
3660
		return VM_FAULT_OOM;
3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688
	if (pud_none(*vmf.pud) && transparent_hugepage_enabled(vma)) {
		vmf.flags |= FAULT_FLAG_SIZE_PUD;
		ret = create_huge_pud(&vmf);
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
	} else {
		pud_t orig_pud = *vmf.pud;

		barrier();
		if (pud_trans_huge(orig_pud) || pud_devmap(orig_pud)) {
			unsigned int dirty = flags & FAULT_FLAG_WRITE;

			vmf.flags |= FAULT_FLAG_SIZE_PUD;

			/* NUMA case for anonymous PUDs would go here */

			if (dirty && !pud_write(orig_pud)) {
				ret = wp_huge_pud(&vmf, orig_pud);
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
			} else {
				huge_pud_set_accessed(&vmf, orig_pud);
				return 0;
			}
		}
	}

	vmf.pmd = pmd_alloc(mm, vmf.pud, address);
J
Jan Kara 已提交
3689
	if (!vmf.pmd)
H
Hugh Dickins 已提交
3690
		return VM_FAULT_OOM;
J
Jan Kara 已提交
3691
	if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) {
3692 3693
		vmf.flags |= FAULT_FLAG_SIZE_PMD;
		ret = create_huge_pmd(&vmf);
3694 3695
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3696 3697
		/* fall through path, remove PMD flag */
		vmf.flags &= ~FAULT_FLAG_SIZE_PMD;
3698
	} else {
J
Jan Kara 已提交
3699
		pmd_t orig_pmd = *vmf.pmd;
3700

3701
		barrier();
3702
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
3703
			vmf.flags |= FAULT_FLAG_SIZE_PMD;
3704
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
3705
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
3706

J
Jan Kara 已提交
3707
			if ((vmf.flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3708
					!pmd_write(orig_pmd)) {
J
Jan Kara 已提交
3709
				ret = wp_huge_pmd(&vmf, orig_pmd);
3710 3711
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3712 3713
				/* fall through path, remove PUD flag */
				vmf.flags &= ~FAULT_FLAG_SIZE_PUD;
3714
			} else {
J
Jan Kara 已提交
3715
				huge_pmd_set_accessed(&vmf, orig_pmd);
3716
				return 0;
3717
			}
3718 3719 3720
		}
	}

J
Jan Kara 已提交
3721
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
3722 3723
}

3724 3725 3726 3727 3728 3729
/*
 * 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().
 */
3730 3731
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
3732 3733 3734 3735 3736 3737
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
3738
	mem_cgroup_count_vm_event(vma->vm_mm, PGFAULT);
3739 3740 3741 3742 3743 3744 3745 3746 3747

	/* 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)
3748
		mem_cgroup_oom_enable();
3749

K
Kirill A. Shutemov 已提交
3750 3751 3752 3753 3754 3755 3756 3757 3758
	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);
3759

3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770
	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);
	}
3771

3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784
	/*
	 * 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;

3785 3786
	return ret;
}
3787
EXPORT_SYMBOL_GPL(handle_mm_fault);
3788

L
Linus Torvalds 已提交
3789 3790 3791
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3792
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3793
 */
3794
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3795
{
H
Hugh Dickins 已提交
3796 3797
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3798
		return -ENOMEM;
L
Linus Torvalds 已提交
3799

3800 3801
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3802
	spin_lock(&mm->page_table_lock);
3803
	if (pgd_present(*pgd))		/* Another has populated it */
3804
		pud_free(mm, new);
3805 3806
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3807
	spin_unlock(&mm->page_table_lock);
3808
	return 0;
L
Linus Torvalds 已提交
3809 3810 3811 3812 3813 3814
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3815
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3816
 */
3817
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3818
{
3819
	spinlock_t *ptl;
H
Hugh Dickins 已提交
3820 3821
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3822
		return -ENOMEM;
L
Linus Torvalds 已提交
3823

3824 3825
	smp_wmb(); /* See comment in __pte_alloc */

3826
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
3827
#ifndef __ARCH_HAS_4LEVEL_HACK
3828 3829
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
3830
		pud_populate(mm, pud, new);
3831
	} else	/* Another has populated it */
3832
		pmd_free(mm, new);
3833 3834 3835
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
3836
		pgd_populate(mm, pud, new);
3837 3838
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
3839
#endif /* __ARCH_HAS_4LEVEL_HACK */
3840
	spin_unlock(ptl);
3841
	return 0;
3842
}
L
Linus Torvalds 已提交
3843 3844
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
3845 3846
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
		pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861
{
	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);
3862
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3863

R
Ross Zwisler 已提交
3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891
		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;
}

3892 3893
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
3894 3895 3896 3897 3898
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
R
Ross Zwisler 已提交
3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912
			   !(res = __follow_pte_pmd(mm, address, ptepp, NULL,
					   ptlp)));
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
			   !(res = __follow_pte_pmd(mm, address, ptepp, pmdpp,
					   ptlp)));
3913 3914
	return res;
}
R
Ross Zwisler 已提交
3915
EXPORT_SYMBOL(follow_pte_pmd);
3916

J
Johannes Weiner 已提交
3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945
/**
 * 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);

3946
#ifdef CONFIG_HAVE_IOREMAP_PROT
3947 3948 3949
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
3950
{
3951
	int ret = -EINVAL;
3952 3953 3954
	pte_t *ptep, pte;
	spinlock_t *ptl;

3955 3956
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
3957

3958
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
3959
		goto out;
3960
	pte = *ptep;
3961

3962 3963 3964 3965
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
3966
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
3967

3968
	ret = 0;
3969 3970 3971
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
3972
	return ret;
3973 3974 3975 3976 3977 3978 3979
}

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
3980
	void __iomem *maddr;
3981 3982
	int offset = addr & (PAGE_SIZE-1);

3983
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
3984 3985
		return -EINVAL;

3986
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
3987 3988 3989 3990 3991 3992 3993 3994
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
3995
EXPORT_SYMBOL_GPL(generic_access_phys);
3996 3997
#endif

3998
/*
3999 4000
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4001
 */
4002
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4003
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4004 4005 4006
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4007
	int write = gup_flags & FOLL_WRITE;
4008 4009

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4010
	/* ignore errors, just check how much was successfully transferred */
4011 4012 4013
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4014
		struct page *page = NULL;
4015

4016
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4017
				gup_flags, &page, &vma, NULL);
4018
		if (ret <= 0) {
4019 4020 4021
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4022 4023 4024 4025 4026
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4027
			if (!vma || vma->vm_start > addr)
4028 4029 4030 4031 4032 4033 4034
				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;
4035
#endif
4036
		} else {
4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051
			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);
4052
			put_page(page);
4053 4054 4055 4056 4057 4058 4059 4060 4061
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4062

S
Stephen Wilson 已提交
4063
/**
4064
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4065 4066 4067 4068
 * @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
4069
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4070 4071 4072 4073
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4074
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4075
{
4076
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4077 4078
}

4079 4080 4081 4082 4083 4084
/*
 * 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,
4085
		void *buf, int len, unsigned int gup_flags)
4086 4087 4088 4089 4090 4091 4092 4093
{
	struct mm_struct *mm;
	int ret;

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

4094
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4095

4096 4097 4098 4099
	mmput(mm);

	return ret;
}
4100
EXPORT_SYMBOL_GPL(access_process_vm);
4101

4102 4103 4104 4105 4106 4107 4108 4109
/*
 * 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;

4110 4111 4112 4113 4114 4115 4116
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

4117 4118 4119 4120 4121 4122
	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
4123
			char *p;
4124

M
Miklos Szeredi 已提交
4125
			p = file_path(f, buf, PAGE_SIZE);
4126 4127
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4128
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4129 4130 4131 4132 4133
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4134
	up_read(&mm->mmap_sem);
4135
}
4136

4137
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4138
void __might_fault(const char *file, int line)
4139
{
4140 4141 4142 4143 4144 4145 4146 4147
	/*
	 * 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;
4148
	if (pagefault_disabled())
4149
		return;
4150 4151
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4152
	if (current->mm)
4153
		might_lock_read(&current->mm->mmap_sem);
4154
#endif
4155
}
4156
EXPORT_SYMBOL(__might_fault);
4157
#endif
A
Andrea Arcangeli 已提交
4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227

#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);
	}
}
4228 4229 4230

long copy_huge_page_from_user(struct page *dst_page,
				const void __user *usr_src,
4231 4232
				unsigned int pages_per_huge_page,
				bool allow_pagefault)
4233 4234 4235 4236 4237 4238 4239
{
	void *src = (void *)usr_src;
	void *page_kaddr;
	unsigned long i, rc = 0;
	unsigned long ret_val = pages_per_huge_page * PAGE_SIZE;

	for (i = 0; i < pages_per_huge_page; i++) {
4240 4241 4242 4243
		if (allow_pagefault)
			page_kaddr = kmap(dst_page + i);
		else
			page_kaddr = kmap_atomic(dst_page + i);
4244 4245 4246
		rc = copy_from_user(page_kaddr,
				(const void __user *)(src + i * PAGE_SIZE),
				PAGE_SIZE);
4247 4248 4249 4250
		if (allow_pagefault)
			kunmap(dst_page + i);
		else
			kunmap_atomic(page_kaddr);
4251 4252 4253 4254 4255 4256 4257 4258 4259

		ret_val -= (PAGE_SIZE - rc);
		if (rc)
			break;

		cond_resched();
	}
	return ret_val;
}
A
Andrea Arcangeli 已提交
4260
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4261

4262
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4263 4264 4265 4266 4267 4268 4269 4270 4271

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);
}

4272
bool ptlock_alloc(struct page *page)
4273 4274 4275
{
	spinlock_t *ptl;

4276
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4277 4278
	if (!ptl)
		return false;
4279
	page->ptl = ptl;
4280 4281 4282
	return true;
}

4283
void ptlock_free(struct page *page)
4284
{
4285
	kmem_cache_free(page_ptl_cachep, page->ptl);
4286 4287
}
#endif