memory.c 101.8 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>
P
Peter Zijlstra 已提交
53
#include <linux/writeback.h>
54
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
55
#include <linux/mmu_notifier.h>
56 57 58
#include <linux/kallsyms.h>
#include <linux/swapops.h>
#include <linux/elf.h>
59
#include <linux/gfp.h>
60
#include <linux/migrate.h>
A
Andy Shevchenko 已提交
61
#include <linux/string.h>
62
#include <linux/dma-debug.h>
63
#include <linux/debugfs.h>
L
Linus Torvalds 已提交
64

A
Alexey Dobriyan 已提交
65
#include <asm/io.h>
L
Linus Torvalds 已提交
66 67 68 69 70 71
#include <asm/pgalloc.h>
#include <asm/uaccess.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

72 73
#include "internal.h"

74 75
#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
76 77
#endif

A
Andy Whitcroft 已提交
78
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
/* 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);

98 99 100 101 102 103 104 105 106 107 108 109
/*
 * 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
110 111 112 113

static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
114
	return 1;
115 116 117
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
118
unsigned long zero_pfn __read_mostly;
H
Hugh Dickins 已提交
119
unsigned long highest_memmap_pfn __read_mostly;
H
Hugh Dickins 已提交
120 121 122 123 124 125 126 127 128 129

/*
 * 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);
130

K
KAMEZAWA Hiroyuki 已提交
131

132 133
#if defined(SPLIT_RSS_COUNTING)

134
void sync_mm_rss(struct mm_struct *mm)
135 136 137 138
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
139 140 141
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
142 143
		}
	}
144
	current->rss_stat.events = 0;
145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165
}

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))
166
		sync_mm_rss(task->mm);
167
}
168
#else /* SPLIT_RSS_COUNTING */
169 170 171 172 173 174 175 176

#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)
{
}

177 178 179 180 181 182 183 184 185 186 187 188 189 190
#endif /* SPLIT_RSS_COUNTING */

#ifdef HAVE_GENERIC_MMU_GATHER

static int tlb_next_batch(struct mmu_gather *tlb)
{
	struct mmu_gather_batch *batch;

	batch = tlb->active;
	if (batch->next) {
		tlb->active = batch->next;
		return 1;
	}

191 192 193
	if (tlb->batch_count == MAX_GATHER_BATCH_COUNT)
		return 0;

194 195 196 197
	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
	if (!batch)
		return 0;

198
	tlb->batch_count++;
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
	batch->next = NULL;
	batch->nr   = 0;
	batch->max  = MAX_GATHER_BATCH;

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

	return 1;
}

/* 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).
 */
214
void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end)
215 216 217
{
	tlb->mm = mm;

218 219
	/* Is it from 0 to ~0? */
	tlb->fullmm     = !(start | (end+1));
220
	tlb->need_flush_all = 0;
221 222
	tlb->start	= start;
	tlb->end	= end;
223 224 225 226 227
	tlb->need_flush = 0;
	tlb->local.next = NULL;
	tlb->local.nr   = 0;
	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
	tlb->active     = &tlb->local;
228
	tlb->batch_count = 0;
229 230 231 232 233 234

#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb->batch = NULL;
#endif
}

235
static void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb)
236 237 238 239 240
{
	tlb->need_flush = 0;
	tlb_flush(tlb);
#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb_table_flush(tlb);
241
#endif
242 243 244 245 246
}

static void tlb_flush_mmu_free(struct mmu_gather *tlb)
{
	struct mmu_gather_batch *batch;
247

248 249 250 251 252 253 254
	for (batch = &tlb->local; batch; batch = batch->next) {
		free_pages_and_swap_cache(batch->pages, batch->nr);
		batch->nr = 0;
	}
	tlb->active = &tlb->local;
}

255 256 257 258 259 260 261 262
void tlb_flush_mmu(struct mmu_gather *tlb)
{
	if (!tlb->need_flush)
		return;
	tlb_flush_mmu_tlbonly(tlb);
	tlb_flush_mmu_free(tlb);
}

263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
/* 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().
 */
int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
{
	struct mmu_gather_batch *batch;

S
Shaohua Li 已提交
293
	VM_BUG_ON(!tlb->need_flush);
294 295 296 297 298 299

	batch = tlb->active;
	batch->pages[batch->nr++] = page;
	if (batch->nr == batch->max) {
		if (!tlb_next_batch(tlb))
			return 0;
300
		batch = tlb->active;
301
	}
302
	VM_BUG_ON_PAGE(batch->nr > batch->max, page);
303 304 305 306 307 308

	return batch->max - batch->nr;
}

#endif /* HAVE_GENERIC_MMU_GATHER */

309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
#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;

	tlb->need_flush = 1;

	/*
	 * 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);
}

384
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
385

L
Linus Torvalds 已提交
386 387 388 389
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
390 391
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
392
{
393
	pgtable_t token = pmd_pgtable(*pmd);
394
	pmd_clear(pmd);
395
	pte_free_tlb(tlb, token, addr);
396
	atomic_long_dec(&tlb->mm->nr_ptes);
L
Linus Torvalds 已提交
397 398
}

399 400 401
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
402 403 404
{
	pmd_t *pmd;
	unsigned long next;
405
	unsigned long start;
L
Linus Torvalds 已提交
406

407
	start = addr;
L
Linus Torvalds 已提交
408 409 410 411 412
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
413
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
414 415
	} while (pmd++, addr = next, addr != end);

416 417 418 419 420 421 422
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
423
	}
424 425 426 427 428
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
429
	pmd_free_tlb(tlb, pmd, start);
L
Linus Torvalds 已提交
430 431
}

432 433 434
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
435 436 437
{
	pud_t *pud;
	unsigned long next;
438
	unsigned long start;
L
Linus Torvalds 已提交
439

440
	start = addr;
L
Linus Torvalds 已提交
441 442 443 444 445
	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
446
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
447 448
	} while (pud++, addr = next, addr != end);

449 450 451 452 453 454 455
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
456
	}
457 458 459 460 461
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(pgd, start);
	pgd_clear(pgd);
462
	pud_free_tlb(tlb, pud, start);
L
Linus Torvalds 已提交
463 464 465
}

/*
466
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
467
 */
468
void free_pgd_range(struct mmu_gather *tlb,
469 470
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
471 472 473
{
	pgd_t *pgd;
	unsigned long next;
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499

	/*
	 * 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
500

501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
	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;

517
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
518 519 520 521
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
522
		free_pud_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
523
	} while (pgd++, addr = next, addr != end);
524 525
}

526
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
527
		unsigned long floor, unsigned long ceiling)
528 529 530 531 532
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

533
		/*
N
npiggin@suse.de 已提交
534 535
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
536
		 */
537
		unlink_anon_vmas(vma);
538 539
		unlink_file_vma(vma);

540
		if (is_vm_hugetlb_page(vma)) {
541
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
542
				floor, next? next->vm_start: ceiling);
543 544 545 546 547
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
548
			       && !is_vm_hugetlb_page(next)) {
549 550
				vma = next;
				next = vma->vm_next;
551
				unlink_anon_vmas(vma);
552
				unlink_file_vma(vma);
553 554 555 556
			}
			free_pgd_range(tlb, addr, vma->vm_end,
				floor, next? next->vm_start: ceiling);
		}
557 558
		vma = next;
	}
L
Linus Torvalds 已提交
559 560
}

561 562
int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
		pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
563
{
564
	spinlock_t *ptl;
565
	pgtable_t new = pte_alloc_one(mm, address);
566
	int wait_split_huge_page;
567 568 569
	if (!new)
		return -ENOMEM;

570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
	/*
	 * 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 */

585
	ptl = pmd_lock(mm, pmd);
586 587
	wait_split_huge_page = 0;
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
588
		atomic_long_inc(&mm->nr_ptes);
L
Linus Torvalds 已提交
589
		pmd_populate(mm, pmd, new);
590
		new = NULL;
591 592
	} else if (unlikely(pmd_trans_splitting(*pmd)))
		wait_split_huge_page = 1;
593
	spin_unlock(ptl);
594 595
	if (new)
		pte_free(mm, new);
596 597
	if (wait_split_huge_page)
		wait_split_huge_page(vma->anon_vma, pmd);
598
	return 0;
L
Linus Torvalds 已提交
599 600
}

601
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
602
{
603 604 605 606
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

607 608
	smp_wmb(); /* See comment in __pte_alloc */

609
	spin_lock(&init_mm.page_table_lock);
610
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
611
		pmd_populate_kernel(&init_mm, pmd, new);
612
		new = NULL;
613 614
	} else
		VM_BUG_ON(pmd_trans_splitting(*pmd));
615
	spin_unlock(&init_mm.page_table_lock);
616 617
	if (new)
		pte_free_kernel(&init_mm, new);
618
	return 0;
L
Linus Torvalds 已提交
619 620
}

K
KAMEZAWA Hiroyuki 已提交
621 622 623 624 625 626
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)
627
{
K
KAMEZAWA Hiroyuki 已提交
628 629
	int i;

630
	if (current->mm == mm)
631
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
632 633 634
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
635 636
}

N
Nick Piggin 已提交
637
/*
638 639 640
 * 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
641 642 643
 *
 * The calling function must still handle the error.
 */
644 645
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
646
{
647 648 649 650 651
	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;
652 653 654 655 656 657 658 659 660 661 662 663 664 665
	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) {
666 667
			printk(KERN_ALERT
				"BUG: Bad page map: %lu messages suppressed\n",
668 669 670 671 672 673 674
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
675 676 677 678

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

679 680
	printk(KERN_ALERT
		"BUG: Bad page map in process %s  pte:%08llx pmd:%08llx\n",
681 682
		current->comm,
		(long long)pte_val(pte), (long long)pmd_val(*pmd));
683
	if (page)
684
		dump_page(page, "bad pte");
685
	printk(KERN_ALERT
686 687 688 689 690 691
		"addr:%p vm_flags:%08lx anon_vma:%p mapping:%p index:%lx\n",
		(void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
	if (vma->vm_ops)
J
Joe Perches 已提交
692 693
		printk(KERN_ALERT "vma->vm_ops->fault: %pSR\n",
		       vma->vm_ops->fault);
A
Al Viro 已提交
694
	if (vma->vm_file)
J
Joe Perches 已提交
695 696
		printk(KERN_ALERT "vma->vm_file->f_op->mmap: %pSR\n",
		       vma->vm_file->f_op->mmap);
N
Nick Piggin 已提交
697
	dump_stack();
698
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
699 700
}

H
Hugh Dickins 已提交
701
/*
N
Nick Piggin 已提交
702
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
703
 *
N
Nick Piggin 已提交
704 705 706
 * "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
707
 *
N
Nick Piggin 已提交
708 709 710 711 712 713 714 715
 * 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.
716
 *
J
Jared Hulbert 已提交
717 718
 * 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
719 720
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
721 722 723
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
724 725 726 727 728 729
 * 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
730 731
 *
 *
N
Nick Piggin 已提交
732
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
733 734 735 736 737 738 739 740 741
 *
 * 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
742
 */
N
Nick Piggin 已提交
743 744 745 746 747 748 749
#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
750
{
751
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
752 753

	if (HAVE_PTE_SPECIAL) {
754
		if (likely(!pte_special(pte) || pte_numa(pte)))
755
			goto check_pfn;
H
Hugh Dickins 已提交
756 757
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
H
Hugh Dickins 已提交
758
		if (!is_zero_pfn(pfn))
759
			print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
760 761 762 763 764
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
765 766 767 768 769 770
	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
771 772
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
773 774 775 776 777
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
778 779
	}

780 781 782 783 784
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
785

786 787 788
	if (is_zero_pfn(pfn))
		return NULL;

789
	/*
N
Nick Piggin 已提交
790 791
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
792
	 */
J
Jared Hulbert 已提交
793
out:
794
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
795 796
}

L
Linus Torvalds 已提交
797 798 799 800 801 802
/*
 * 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
803
static inline unsigned long
L
Linus Torvalds 已提交
804
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
805
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
806
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
807
{
N
Nick Piggin 已提交
808
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
809 810 811 812 813 814
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
		if (!pte_file(pte)) {
815 816
			swp_entry_t entry = pte_to_swp_entry(pte);

H
Hugh Dickins 已提交
817 818 819
			if (swap_duplicate(entry) < 0)
				return entry.val;

L
Linus Torvalds 已提交
820 821 822
			/* make sure dst_mm is on swapoff's mmlist. */
			if (unlikely(list_empty(&dst_mm->mmlist))) {
				spin_lock(&mmlist_lock);
823 824 825
				if (list_empty(&dst_mm->mmlist))
					list_add(&dst_mm->mmlist,
						 &src_mm->mmlist);
L
Linus Torvalds 已提交
826 827
				spin_unlock(&mmlist_lock);
			}
K
KAMEZAWA Hiroyuki 已提交
828 829
			if (likely(!non_swap_entry(entry)))
				rss[MM_SWAPENTS]++;
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
			else if (is_migration_entry(entry)) {
				page = migration_entry_to_page(entry);

				if (PageAnon(page))
					rss[MM_ANONPAGES]++;
				else
					rss[MM_FILEPAGES]++;

				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);
846 847
					if (pte_swp_soft_dirty(*src_pte))
						pte = pte_swp_mksoft_dirty(pte);
848 849
					set_pte_at(src_mm, addr, src_pte, pte);
				}
850
			}
L
Linus Torvalds 已提交
851
		}
852
		goto out_set_pte;
L
Linus Torvalds 已提交
853 854 855 856 857 858
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
859
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
860
		ptep_set_wrprotect(src_mm, addr, src_pte);
861
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
862 863 864 865 866 867 868 869 870
	}

	/*
	 * 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);
871 872 873 874

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
H
Hugh Dickins 已提交
875
		page_dup_rmap(page);
K
KAMEZAWA Hiroyuki 已提交
876 877 878 879
		if (PageAnon(page))
			rss[MM_ANONPAGES]++;
		else
			rss[MM_FILEPAGES]++;
880
	}
881 882 883

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
884
	return 0;
L
Linus Torvalds 已提交
885 886
}

887 888 889
int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		   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
890
{
891
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
892
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
893
	spinlock_t *src_ptl, *dst_ptl;
894
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
895
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
896
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
897 898

again:
K
KAMEZAWA Hiroyuki 已提交
899 900
	init_rss_vec(rss);

H
Hugh Dickins 已提交
901
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
902 903
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
904
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
905
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
906
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
907 908
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
909
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
910 911 912 913 914 915

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
916 917 918
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
919
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
920 921
				break;
		}
L
Linus Torvalds 已提交
922 923 924 925
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
926 927 928 929
		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
930 931 932
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

933
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
934
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
935
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
936
	add_mm_rss_vec(dst_mm, rss);
937
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
938
	cond_resched();
H
Hugh Dickins 已提交
939 940 941 942 943 944

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962
	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);
963 964
		if (pmd_trans_huge(*src_pmd)) {
			int err;
965
			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
966 967 968 969 970 971 972 973
			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
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
		if (pmd_none_or_clear_bad(src_pmd))
			continue;
		if (copy_pte_range(dst_mm, src_mm, dst_pmd, src_pmd,
						vma, addr, next))
			return -ENOMEM;
	} while (dst_pmd++, src_pmd++, addr = next, addr != end);
	return 0;
}

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

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

int copy_page_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		struct vm_area_struct *vma)
{
	pgd_t *src_pgd, *dst_pgd;
	unsigned long next;
	unsigned long addr = vma->vm_start;
	unsigned long end = vma->vm_end;
1012 1013 1014
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1015
	int ret;
L
Linus Torvalds 已提交
1016

1017 1018 1019 1020 1021 1022
	/*
	 * 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.
	 */
1023 1024
	if (!(vma->vm_flags & (VM_HUGETLB | VM_NONLINEAR |
			       VM_PFNMAP | VM_MIXEDMAP))) {
1025 1026 1027 1028
		if (!vma->anon_vma)
			return 0;
	}

L
Linus Torvalds 已提交
1029 1030 1031
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1032
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1033 1034 1035 1036
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1037
		ret = track_pfn_copy(vma);
1038 1039 1040 1041
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1042 1043 1044 1045 1046 1047
	/*
	 * 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.
	 */
1048 1049 1050 1051 1052 1053
	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
1054 1055

	ret = 0;
L
Linus Torvalds 已提交
1056 1057 1058 1059 1060 1061
	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
1062 1063 1064 1065 1066
		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
1067
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1068

1069 1070
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1071
	return ret;
L
Linus Torvalds 已提交
1072 1073
}

1074
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1075
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1076
				unsigned long addr, unsigned long end,
1077
				struct zap_details *details)
L
Linus Torvalds 已提交
1078
{
N
Nick Piggin 已提交
1079
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1080
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1081
	int rss[NR_MM_COUNTERS];
1082
	spinlock_t *ptl;
1083
	pte_t *start_pte;
1084
	pte_t *pte;
K
KAMEZAWA Hiroyuki 已提交
1085

P
Peter Zijlstra 已提交
1086
again:
1087
	init_rss_vec(rss);
1088 1089
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1090
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1091 1092
	do {
		pte_t ptent = *pte;
1093
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1094
			continue;
1095
		}
1096

L
Linus Torvalds 已提交
1097
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1098
			struct page *page;
1099

1100
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
				    details->check_mapping != page->mapping)
					continue;
				/*
				 * Each page->index must be checked when
				 * invalidating or truncating nonlinear.
				 */
				if (details->nonlinear_vma &&
				    (page->index < details->first_index ||
				     page->index > details->last_index))
					continue;
			}
N
Nick Piggin 已提交
1119
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1120
							tlb->fullmm);
L
Linus Torvalds 已提交
1121 1122 1123 1124 1125
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
			if (unlikely(details) && details->nonlinear_vma
			    && linear_page_index(details->nonlinear_vma,
1126 1127 1128 1129 1130 1131
						addr) != page->index) {
				pte_t ptfile = pgoff_to_pte(page->index);
				if (pte_soft_dirty(ptent))
					pte_file_mksoft_dirty(ptfile);
				set_pte_at(mm, addr, pte, ptfile);
			}
L
Linus Torvalds 已提交
1132
			if (PageAnon(page))
K
KAMEZAWA Hiroyuki 已提交
1133
				rss[MM_ANONPAGES]--;
1134
			else {
1135 1136
				if (pte_dirty(ptent)) {
					force_flush = 1;
1137
					set_page_dirty(page);
1138
				}
1139
				if (pte_young(ptent) &&
1140
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1141
					mark_page_accessed(page);
K
KAMEZAWA Hiroyuki 已提交
1142
				rss[MM_FILEPAGES]--;
1143
			}
1144
			page_remove_rmap(page);
1145 1146
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1147 1148
			if (unlikely(!__tlb_remove_page(tlb, page))) {
				force_flush = 1;
P
Peter Zijlstra 已提交
1149
				break;
1150
			}
L
Linus Torvalds 已提交
1151 1152 1153 1154 1155 1156 1157 1158
			continue;
		}
		/*
		 * If details->check_mapping, we leave swap entries;
		 * if details->nonlinear_vma, we leave file entries.
		 */
		if (unlikely(details))
			continue;
1159 1160 1161
		if (pte_file(ptent)) {
			if (unlikely(!(vma->vm_flags & VM_NONLINEAR)))
				print_bad_pte(vma, addr, ptent, NULL);
K
KAMEZAWA Hiroyuki 已提交
1162 1163 1164 1165 1166
		} else {
			swp_entry_t entry = pte_to_swp_entry(ptent);

			if (!non_swap_entry(entry))
				rss[MM_SWAPENTS]--;
1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
			else if (is_migration_entry(entry)) {
				struct page *page;

				page = migration_entry_to_page(entry);

				if (PageAnon(page))
					rss[MM_ANONPAGES]--;
				else
					rss[MM_FILEPAGES]--;
			}
K
KAMEZAWA Hiroyuki 已提交
1177 1178 1179
			if (unlikely(!free_swap_and_cache(entry)))
				print_bad_pte(vma, addr, ptent, NULL);
		}
1180
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1181
	} while (pte++, addr += PAGE_SIZE, addr != end);
1182

K
KAMEZAWA Hiroyuki 已提交
1183
	add_mm_rss_vec(mm, rss);
1184
	arch_leave_lazy_mmu_mode();
1185

1186
	/* Do the actual TLB flush before dropping ptl */
P
Peter Zijlstra 已提交
1187
	if (force_flush) {
1188 1189 1190 1191 1192 1193 1194 1195
		unsigned long old_end;

		/*
		 * Flush the TLB just for the previous segment,
		 * then update the range to be the remaining
		 * TLB range.
		 */
		old_end = tlb->end;
1196
		tlb->end = addr;
1197
		tlb_flush_mmu_tlbonly(tlb);
1198 1199
		tlb->start = addr;
		tlb->end = old_end;
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
	}
	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);
1212 1213

		if (addr != end)
P
Peter Zijlstra 已提交
1214 1215 1216
			goto again;
	}

1217
	return addr;
L
Linus Torvalds 已提交
1218 1219
}

1220
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1221
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1222
				unsigned long addr, unsigned long end,
1223
				struct zap_details *details)
L
Linus Torvalds 已提交
1224 1225 1226 1227 1228 1229 1230
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1231
		if (pmd_trans_huge(*pmd)) {
1232
			if (next - addr != HPAGE_PMD_SIZE) {
1233 1234 1235 1236 1237 1238 1239 1240 1241
#ifdef CONFIG_DEBUG_VM
				if (!rwsem_is_locked(&tlb->mm->mmap_sem)) {
					pr_err("%s: mmap_sem is unlocked! addr=0x%lx end=0x%lx vma->vm_start=0x%lx vma->vm_end=0x%lx\n",
						__func__, addr, end,
						vma->vm_start,
						vma->vm_end);
					BUG();
				}
#endif
1242
				split_huge_page_pmd(vma, addr, pmd);
S
Shaohua Li 已提交
1243
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1244
				goto next;
1245 1246
			/* fall through */
		}
1247 1248 1249 1250 1251 1252 1253 1254 1255
		/*
		 * 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;
1256
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1257
next:
1258 1259
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1260 1261

	return addr;
L
Linus Torvalds 已提交
1262 1263
}

1264
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1265
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1266
				unsigned long addr, unsigned long end,
1267
				struct zap_details *details)
L
Linus Torvalds 已提交
1268 1269 1270 1271 1272 1273 1274
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1275
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1276
			continue;
1277 1278
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
	} while (pud++, addr = next, addr != end);
1279 1280

	return addr;
L
Linus Torvalds 已提交
1281 1282
}

A
Al Viro 已提交
1283 1284 1285 1286
static void unmap_page_range(struct mmu_gather *tlb,
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1287 1288 1289 1290 1291 1292 1293 1294
{
	pgd_t *pgd;
	unsigned long next;

	if (details && !details->check_mapping && !details->nonlinear_vma)
		details = NULL;

	BUG_ON(addr >= end);
1295
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
1296 1297 1298 1299
	tlb_start_vma(tlb, vma);
	pgd = pgd_offset(vma->vm_mm, addr);
	do {
		next = pgd_addr_end(addr, end);
1300
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1301
			continue;
1302 1303
		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1304
	tlb_end_vma(tlb, vma);
1305
	mem_cgroup_uncharge_end();
L
Linus Torvalds 已提交
1306
}
1307

1308 1309 1310

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1311
		unsigned long end_addr,
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
		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;

1323 1324 1325
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1326
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1327
		untrack_pfn(vma, 0, 0);
1328 1329 1330 1331 1332 1333 1334

	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
1335
			 * cleanup path of mmap_region. When
1336
			 * hugetlbfs ->mmap method fails,
1337
			 * mmap_region() nullifies vma->vm_file
1338 1339 1340 1341
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1342 1343
			if (vma->vm_file) {
				mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
1344
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1345 1346
				mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
			}
1347 1348 1349
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1350 1351 1352 1353
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1354
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1355 1356 1357 1358
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1359
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
 *
 * 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
1370
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1371
		struct vm_area_struct *vma, unsigned long start_addr,
1372
		unsigned long end_addr)
L
Linus Torvalds 已提交
1373
{
A
Andrea Arcangeli 已提交
1374
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1375

A
Andrea Arcangeli 已提交
1376
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1377
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1378
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1379
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1380 1381 1382 1383 1384
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1385
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1386 1387
 * @size: number of bytes to zap
 * @details: details of nonlinear truncation or shared cache invalidation
1388 1389
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1390
 */
1391
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
L
Linus Torvalds 已提交
1392 1393 1394
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1395
	struct mmu_gather tlb;
1396
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1397 1398

	lru_add_drain();
1399
	tlb_gather_mmu(&tlb, mm, start, end);
1400
	update_hiwater_rss(mm);
1401 1402
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1403
		unmap_single_vma(&tlb, vma, start, end, details);
1404 1405
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1406 1407
}

1408 1409 1410 1411 1412 1413 1414 1415
/**
 * 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
 * @details: details of nonlinear truncation or shared cache invalidation
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1416
 */
1417
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1418 1419 1420
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1421
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1422 1423 1424
	unsigned long end = address + size;

	lru_add_drain();
1425
	tlb_gather_mmu(&tlb, mm, address, end);
1426
	update_hiwater_rss(mm);
1427
	mmu_notifier_invalidate_range_start(mm, address, end);
1428
	unmap_single_vma(&tlb, vma, address, end, details);
1429
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1430
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1431 1432
}

1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
/**
 * 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;
1451
	zap_page_range_single(vma, address, size, NULL);
1452 1453 1454 1455
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1456
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1457
			spinlock_t **ptl)
1458 1459 1460 1461
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
1462
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
1463 1464
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
1465
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
1466
		}
1467 1468 1469 1470
	}
	return NULL;
}

1471 1472 1473 1474 1475 1476 1477
/*
 * 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
1478 1479
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1480
{
N
Nick Piggin 已提交
1481
	struct mm_struct *mm = vma->vm_mm;
1482
	int retval;
1483
	pte_t *pte;
1484 1485
	spinlock_t *ptl;

1486
	retval = -EINVAL;
1487
	if (PageAnon(page))
1488
		goto out;
1489 1490
	retval = -ENOMEM;
	flush_dcache_page(page);
1491
	pte = get_locked_pte(mm, addr, &ptl);
1492
	if (!pte)
1493
		goto out;
1494 1495 1496 1497 1498 1499
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1500
	inc_mm_counter_fast(mm, MM_FILEPAGES);
1501 1502 1503 1504
	page_add_file_rmap(page);
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1505 1506
	pte_unmap_unlock(pte, ptl);
	return retval;
1507 1508 1509 1510 1511 1512
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1513 1514 1515 1516 1517 1518
/**
 * 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
 *
1519 1520 1521 1522 1523 1524
 * 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
1525
 * (see split_page()).
1526 1527 1528 1529 1530 1531 1532 1533
 *
 * 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.
1534 1535 1536 1537 1538
 *
 * 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.
1539
 */
N
Nick Piggin 已提交
1540 1541
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1542 1543 1544 1545 1546
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1547 1548 1549 1550 1551
	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
1552
	return insert_page(vma, addr, page, vma->vm_page_prot);
1553
}
1554
EXPORT_SYMBOL(vm_insert_page);
1555

N
Nick Piggin 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn, pgprot_t prot)
{
	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.. */
	entry = pte_mkspecial(pfn_pte(pfn, prot));
	set_pte_at(mm, addr, pte, entry);
1575
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1576 1577 1578 1579 1580 1581 1582 1583

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

N
Nick Piggin 已提交
1584 1585 1586 1587 1588 1589
/**
 * 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
 *
1590
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1591 1592 1593 1594
 * 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
1595 1596 1597 1598 1599
 *
 * 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
1600 1601
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1602
			unsigned long pfn)
N
Nick Piggin 已提交
1603
{
1604
	int ret;
1605
	pgprot_t pgprot = vma->vm_page_prot;
N
Nick Piggin 已提交
1606 1607 1608 1609 1610 1611
	/*
	 * 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
1612 1613 1614 1615 1616
	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
1617

N
Nick Piggin 已提交
1618 1619
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1620
	if (track_pfn_insert(vma, &pgprot, pfn))
1621 1622
		return -EINVAL;

1623
	ret = insert_pfn(vma, addr, pfn, pgprot);
1624 1625

	return ret;
N
Nick Piggin 已提交
1626 1627
}
EXPORT_SYMBOL(vm_insert_pfn);
N
Nick Piggin 已提交
1628

N
Nick Piggin 已提交
1629 1630 1631 1632
int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn)
{
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1633

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

N
Nick Piggin 已提交
1637 1638 1639 1640
	/*
	 * 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
1641 1642
	 * 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
1643 1644 1645 1646 1647 1648 1649 1650
	 */
	if (!HAVE_PTE_SPECIAL && pfn_valid(pfn)) {
		struct page *page;

		page = pfn_to_page(pfn);
		return insert_page(vma, addr, page, vma->vm_page_prot);
	}
	return insert_pfn(vma, addr, pfn, vma->vm_page_prot);
N
Nick Piggin 已提交
1651
}
N
Nick Piggin 已提交
1652
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1653

L
Linus Torvalds 已提交
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
/*
 * 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
1664
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1665

H
Hugh Dickins 已提交
1666
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1667 1668
	if (!pte)
		return -ENOMEM;
1669
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1670 1671
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1672
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1673 1674
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1675
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1676
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690
	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;
1691
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
	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;
}

1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
/**
 * 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
1731 1732 1733 1734 1735
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;
1736
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1737 1738 1739 1740 1741 1742 1743 1744
	struct mm_struct *mm = vma->vm_mm;
	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).
1745 1746 1747
	 *   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.
1748 1749 1750 1751
	 *   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
1752 1753 1754 1755
	 *
	 * 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".
1756
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1757
	 */
1758 1759 1760
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1761
		vma->vm_pgoff = pfn;
1762 1763 1764 1765
	}

	err = track_pfn_remap(vma, &prot, pfn, addr, PAGE_ALIGN(size));
	if (err)
1766
		return -EINVAL;
L
Linus Torvalds 已提交
1767

1768
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780

	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);
1781 1782

	if (err)
1783
		untrack_pfn(vma, pfn, PAGE_ALIGN(size));
1784

L
Linus Torvalds 已提交
1785 1786 1787 1788
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
/**
 * 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);

1836 1837 1838 1839 1840 1841
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;
1842
	pgtable_t token;
1843
	spinlock_t *uninitialized_var(ptl);
1844 1845 1846 1847 1848 1849 1850 1851 1852

	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));

1853 1854
	arch_enter_lazy_mmu_mode();

1855
	token = pmd_pgtable(*pmd);
1856 1857

	do {
1858
		err = fn(pte++, token, addr, data);
1859 1860
		if (err)
			break;
1861
	} while (addr += PAGE_SIZE, addr != end);
1862

1863 1864
	arch_leave_lazy_mmu_mode();

1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877
	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
1878 1879
	BUG_ON(pud_huge(*pud));

1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
	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;
1921
	unsigned long end = addr + size;
1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	int err;

	BUG_ON(addr >= end);
	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);
1932

1933 1934 1935 1936
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

1937 1938 1939 1940
/*
 * 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
1941
 * might give a mix of unmatched parts, do_swap_page and do_nonlinear_fault
1942 1943
 * must check under lock before unmapping the pte and proceeding
 * (but do_wp_page is only called after already making such a check;
1944
 * and do_anonymous_page can safely check later on).
1945
 */
H
Hugh Dickins 已提交
1946
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
1947 1948 1949 1950 1951
				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
1952 1953
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
1954
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
1955
		spin_unlock(ptl);
1956 1957 1958 1959 1960 1961
	}
#endif
	pte_unmap(page_table);
	return same;
}

1962
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
1963
{
1964 1965
	debug_dma_assert_idle(src);

1966 1967 1968 1969 1970 1971 1972
	/*
	 * 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)) {
1973
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
1974 1975 1976 1977 1978 1979 1980 1981 1982
		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))
1983
			clear_page(kaddr);
1984
		kunmap_atomic(kaddr);
1985
		flush_dcache_page(dst);
N
Nick Piggin 已提交
1986 1987
	} else
		copy_user_highpage(dst, src, va, vma);
1988 1989
}

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
/*
 * Notify the address space that the page is about to become writable so that
 * it can prohibit this or wait for the page to get into an appropriate state.
 *
 * We do this without the lock held, so that it can sleep if it needs to.
 */
static int do_page_mkwrite(struct vm_area_struct *vma, struct page *page,
	       unsigned long address)
{
	struct vm_fault vmf;
	int ret;

	vmf.virtual_address = (void __user *)(address & PAGE_MASK);
	vmf.pgoff = page->index;
	vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
	vmf.page = page;

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

L
Linus Torvalds 已提交
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
/*
 * 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.
 *
2036 2037 2038
 * 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
2039
 */
2040 2041
static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
2042
		spinlock_t *ptl, pte_t orig_pte)
2043
	__releases(ptl)
L
Linus Torvalds 已提交
2044
{
2045
	struct page *old_page, *new_page = NULL;
L
Linus Torvalds 已提交
2046
	pte_t entry;
2047
	int ret = 0;
2048
	int page_mkwrite = 0;
2049
	struct page *dirty_page = NULL;
2050 2051
	unsigned long mmun_start = 0;	/* For mmu_notifiers */
	unsigned long mmun_end = 0;	/* For mmu_notifiers */
L
Linus Torvalds 已提交
2052

2053
	old_page = vm_normal_page(vma, address, orig_pte);
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
	if (!old_page) {
		/*
		 * VM_MIXEDMAP !pfn_valid() case
		 *
		 * We should not cow pages in a shared writeable mapping.
		 * Just mark the pages writable as we can't do any dirty
		 * accounting on raw pfn maps.
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
			goto reuse;
2065
		goto gotten;
2066
	}
L
Linus Torvalds 已提交
2067

2068
	/*
P
Peter Zijlstra 已提交
2069 2070
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2071
	 */
H
Hugh Dickins 已提交
2072
	if (PageAnon(old_page) && !PageKsm(old_page)) {
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
		if (!trylock_page(old_page)) {
			page_cache_get(old_page);
			pte_unmap_unlock(page_table, ptl);
			lock_page(old_page);
			page_table = pte_offset_map_lock(mm, pmd, address,
							 &ptl);
			if (!pte_same(*page_table, orig_pte)) {
				unlock_page(old_page);
				goto unlock;
			}
			page_cache_release(old_page);
P
Peter Zijlstra 已提交
2084
		}
2085
		if (reuse_swap_page(old_page)) {
2086 2087 2088 2089 2090 2091
			/*
			 * 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.
			 */
			page_move_anon_rmap(old_page, vma, address);
2092 2093 2094
			unlock_page(old_page);
			goto reuse;
		}
2095
		unlock_page(old_page);
P
Peter Zijlstra 已提交
2096
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2097
					(VM_WRITE|VM_SHARED))) {
P
Peter Zijlstra 已提交
2098 2099 2100 2101 2102
		/*
		 * Only catch write-faults on shared writable pages,
		 * read-only shared pages can get COWed by
		 * get_user_pages(.write=1, .force=1).
		 */
2103
		if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
2104
			int tmp;
2105 2106
			page_cache_get(old_page);
			pte_unmap_unlock(page_table, ptl);
2107 2108 2109 2110 2111
			tmp = do_page_mkwrite(vma, old_page, address);
			if (unlikely(!tmp || (tmp &
					(VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
				page_cache_release(old_page);
				return tmp;
2112
			}
2113 2114 2115 2116 2117 2118 2119 2120
			/*
			 * Since we dropped the lock we need to revalidate
			 * the PTE as someone else may have changed it.  If
			 * they did, we just return, as we can count on the
			 * MMU to tell us if they didn't also make it writable.
			 */
			page_table = pte_offset_map_lock(mm, pmd, address,
							 &ptl);
N
Nick Piggin 已提交
2121 2122
			if (!pte_same(*page_table, orig_pte)) {
				unlock_page(old_page);
2123
				goto unlock;
N
Nick Piggin 已提交
2124
			}
2125 2126

			page_mkwrite = 1;
L
Linus Torvalds 已提交
2127
		}
2128 2129
		dirty_page = old_page;
		get_page(dirty_page);
2130

2131
reuse:
2132 2133 2134 2135 2136 2137 2138 2139
		/*
		 * Clear the pages cpupid information as the existing
		 * information potentially belongs to a now completely
		 * unrelated process.
		 */
		if (old_page)
			page_cpupid_xchg_last(old_page, (1 << LAST_CPUPID_SHIFT) - 1);

2140 2141 2142
		flush_cache_page(vma, address, pte_pfn(orig_pte));
		entry = pte_mkyoung(orig_pte);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
2143
		if (ptep_set_access_flags(vma, address, page_table, entry,1))
2144
			update_mmu_cache(vma, address, page_table);
2145
		pte_unmap_unlock(page_table, ptl);
2146
		ret |= VM_FAULT_WRITE;
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156

		if (!dirty_page)
			return ret;

		/*
		 * Yes, Virginia, this is actually required to prevent a race
		 * with clear_page_dirty_for_io() from clearing the page dirty
		 * bit after it clear all dirty ptes, but before a racing
		 * do_wp_page installs a dirty pte.
		 *
2157
		 * do_shared_fault is protected similarly.
2158 2159 2160
		 */
		if (!page_mkwrite) {
			wait_on_page_locked(dirty_page);
2161
			set_page_dirty_balance(dirty_page);
2162 2163 2164
			/* file_update_time outside page_lock */
			if (vma->vm_file)
				file_update_time(vma->vm_file);
2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182
		}
		put_page(dirty_page);
		if (page_mkwrite) {
			struct address_space *mapping = dirty_page->mapping;

			set_page_dirty(dirty_page);
			unlock_page(dirty_page);
			page_cache_release(dirty_page);
			if (mapping)	{
				/*
				 * Some device drivers do not set page.mapping
				 * but still dirty their pages
				 */
				balance_dirty_pages_ratelimited(mapping);
			}
		}

		return ret;
L
Linus Torvalds 已提交
2183 2184 2185 2186 2187
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
N
Nick Piggin 已提交
2188
	page_cache_get(old_page);
H
Hugh Dickins 已提交
2189
gotten:
2190
	pte_unmap_unlock(page_table, ptl);
L
Linus Torvalds 已提交
2191 2192

	if (unlikely(anon_vma_prepare(vma)))
2193
		goto oom;
H
Hugh Dickins 已提交
2194

H
Hugh Dickins 已提交
2195
	if (is_zero_pfn(pte_pfn(orig_pte))) {
H
Hugh Dickins 已提交
2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
		new_page = alloc_zeroed_user_highpage_movable(vma, address);
		if (!new_page)
			goto oom;
	} else {
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
		if (!new_page)
			goto oom;
		cow_user_page(new_page, old_page, address, vma);
	}
	__SetPageUptodate(new_page);

2207
	if (mem_cgroup_charge_anon(new_page, mm, GFP_KERNEL))
2208 2209
		goto oom_free_new;

2210
	mmun_start  = address & PAGE_MASK;
2211
	mmun_end    = mmun_start + PAGE_SIZE;
2212 2213
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

L
Linus Torvalds 已提交
2214 2215 2216
	/*
	 * Re-check the pte - we dropped the lock
	 */
2217
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
2218
	if (likely(pte_same(*page_table, orig_pte))) {
H
Hugh Dickins 已提交
2219 2220
		if (old_page) {
			if (!PageAnon(old_page)) {
2221 2222
				dec_mm_counter_fast(mm, MM_FILEPAGES);
				inc_mm_counter_fast(mm, MM_ANONPAGES);
H
Hugh Dickins 已提交
2223 2224
			}
		} else
2225
			inc_mm_counter_fast(mm, MM_ANONPAGES);
2226
		flush_cache_page(vma, address, pte_pfn(orig_pte));
2227 2228
		entry = mk_pte(new_page, vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
2229 2230 2231 2232 2233 2234
		/*
		 * 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.
		 */
2235
		ptep_clear_flush(vma, address, page_table);
N
Nick Piggin 已提交
2236
		page_add_new_anon_rmap(new_page, vma, address);
2237 2238 2239 2240 2241 2242
		/*
		 * 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.
		 */
		set_pte_at_notify(mm, address, page_table, entry);
2243
		update_mmu_cache(vma, address, page_table);
N
Nick Piggin 已提交
2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
		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.
			 */
2267
			page_remove_rmap(old_page);
N
Nick Piggin 已提交
2268 2269
		}

L
Linus Torvalds 已提交
2270 2271
		/* Free the old page.. */
		new_page = old_page;
N
Nick Piggin 已提交
2272
		ret |= VM_FAULT_WRITE;
2273 2274 2275
	} else
		mem_cgroup_uncharge_page(new_page);

2276 2277
	if (new_page)
		page_cache_release(new_page);
2278
unlock:
2279
	pte_unmap_unlock(page_table, ptl);
2280
	if (mmun_end > mmun_start)
2281
		mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293
	if (old_page) {
		/*
		 * Don't let another task, with possibly unlocked vma,
		 * keep the mlocked page.
		 */
		if ((ret & VM_FAULT_WRITE) && (vma->vm_flags & VM_LOCKED)) {
			lock_page(old_page);	/* LRU manipulation */
			munlock_vma_page(old_page);
			unlock_page(old_page);
		}
		page_cache_release(old_page);
	}
N
Nick Piggin 已提交
2294
	return ret;
2295
oom_free_new:
2296
	page_cache_release(new_page);
2297
oom:
2298
	if (old_page)
H
Hugh Dickins 已提交
2299
		page_cache_release(old_page);
L
Linus Torvalds 已提交
2300 2301 2302
	return VM_FAULT_OOM;
}

2303
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2304 2305 2306
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2307
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2308 2309
}

2310
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2311 2312 2313 2314 2315
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2316
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2317 2318 2319
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2320
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2321 2322 2323 2324 2325 2326 2327 2328
		/* Assume for now that PAGE_CACHE_SHIFT == PAGE_SHIFT */
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2329
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2330 2331
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2332
				details);
L
Linus Torvalds 已提交
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
	}
}

static inline void unmap_mapping_range_list(struct list_head *head,
					    struct zap_details *details)
{
	struct vm_area_struct *vma;

	/*
	 * In nonlinear VMAs there is no correspondence between virtual address
	 * offset and file offset.  So we must perform an exhaustive search
	 * across *all* the pages in each nonlinear VMA, not just the pages
	 * whose virtual address lies outside the file truncation point.
	 */
2347
	list_for_each_entry(vma, head, shared.nonlinear) {
L
Linus Torvalds 已提交
2348
		details->nonlinear_vma = vma;
2349
		unmap_mapping_range_vma(vma, vma->vm_start, vma->vm_end, details);
L
Linus Torvalds 已提交
2350 2351 2352 2353
	}
}

/**
2354
 * 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
2355
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2356 2357
 * @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
2358
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389
 * 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)
{
	struct zap_details details;
	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.nonlinear_vma = NULL;
	details.first_index = hba;
	details.last_index = hba + hlen - 1;
	if (details.last_index < details.first_index)
		details.last_index = ULONG_MAX;


2390
	mutex_lock(&mapping->i_mmap_mutex);
2391
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2392 2393 2394
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
2395
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
2396 2397 2398 2399
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2400 2401 2402
 * 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
2403
 */
2404 2405
static int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
2406
		unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
2407
{
2408
	spinlock_t *ptl;
2409
	struct page *page, *swapcache;
2410
	swp_entry_t entry;
L
Linus Torvalds 已提交
2411
	pte_t pte;
2412
	int locked;
2413
	struct mem_cgroup *ptr;
2414
	int exclusive = 0;
N
Nick Piggin 已提交
2415
	int ret = 0;
L
Linus Torvalds 已提交
2416

H
Hugh Dickins 已提交
2417
	if (!pte_unmap_same(mm, pmd, page_table, orig_pte))
2418
		goto out;
2419 2420

	entry = pte_to_swp_entry(orig_pte);
2421 2422 2423 2424 2425 2426 2427
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
			migration_entry_wait(mm, pmd, address);
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
			print_bad_pte(vma, address, orig_pte, NULL);
H
Hugh Dickins 已提交
2428
			ret = VM_FAULT_SIGBUS;
2429
		}
2430 2431
		goto out;
	}
2432
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
2433 2434
	page = lookup_swap_cache(entry);
	if (!page) {
2435 2436
		page = swapin_readahead(entry,
					GFP_HIGHUSER_MOVABLE, vma, address);
L
Linus Torvalds 已提交
2437 2438
		if (!page) {
			/*
2439 2440
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2441
			 */
2442
			page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
2443 2444
			if (likely(pte_same(*page_table, orig_pte)))
				ret = VM_FAULT_OOM;
2445
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2446
			goto unlock;
L
Linus Torvalds 已提交
2447 2448 2449 2450
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2451
		count_vm_event(PGMAJFAULT);
2452
		mem_cgroup_count_vm_event(mm, PGMAJFAULT);
2453
	} else if (PageHWPoison(page)) {
2454 2455 2456 2457
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2458 2459
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2460
		swapcache = page;
2461
		goto out_release;
L
Linus Torvalds 已提交
2462 2463
	}

2464
	swapcache = page;
2465
	locked = lock_page_or_retry(page, mm, flags);
R
Rik van Riel 已提交
2466

2467
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2468 2469 2470 2471
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2472

A
Andrea Arcangeli 已提交
2473
	/*
2474 2475 2476 2477
	 * 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
2478
	 */
2479
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2480 2481
		goto out_page;

2482 2483 2484 2485 2486
	page = ksm_might_need_to_copy(page, vma, address);
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2487 2488
	}

K
KAMEZAWA Hiroyuki 已提交
2489
	if (mem_cgroup_try_charge_swapin(mm, page, GFP_KERNEL, &ptr)) {
2490
		ret = VM_FAULT_OOM;
2491
		goto out_page;
2492 2493
	}

L
Linus Torvalds 已提交
2494
	/*
2495
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2496
	 */
2497
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
H
Hugh Dickins 已提交
2498
	if (unlikely(!pte_same(*page_table, orig_pte)))
2499 2500 2501 2502 2503
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2504 2505
	}

2506 2507 2508 2509 2510 2511 2512 2513
	/*
	 * 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.
2514 2515 2516 2517
	 * Because delete_from_swap_page() may be called by reuse_swap_page(),
	 * mem_cgroup_commit_charge_swapin() may not be able to find swp_entry
	 * in page->private. In this case, a record in swap_cgroup  is silently
	 * discarded at swap_free().
2518
	 */
L
Linus Torvalds 已提交
2519

2520
	inc_mm_counter_fast(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
2521
	dec_mm_counter_fast(mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2522
	pte = mk_pte(page, vma->vm_page_prot);
2523
	if ((flags & FAULT_FLAG_WRITE) && reuse_swap_page(page)) {
L
Linus Torvalds 已提交
2524
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
2525
		flags &= ~FAULT_FLAG_WRITE;
2526
		ret |= VM_FAULT_WRITE;
2527
		exclusive = 1;
L
Linus Torvalds 已提交
2528 2529
	}
	flush_icache_page(vma, page);
2530 2531
	if (pte_swp_soft_dirty(orig_pte))
		pte = pte_mksoft_dirty(pte);
L
Linus Torvalds 已提交
2532
	set_pte_at(mm, address, page_table, pte);
2533
	if (page == swapcache)
2534
		do_page_add_anon_rmap(page, vma, address, exclusive);
2535 2536
	else /* ksm created a completely new copy */
		page_add_new_anon_rmap(page, vma, address);
2537 2538
	/* It's better to call commit-charge after rmap is established */
	mem_cgroup_commit_charge_swapin(page, ptr);
L
Linus Torvalds 已提交
2539

2540
	swap_free(entry);
N
Nick Piggin 已提交
2541
	if (vm_swap_full() || (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2542
		try_to_free_swap(page);
2543
	unlock_page(page);
2544
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555
		/*
		 * 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);
		page_cache_release(swapcache);
	}
2556

2557
	if (flags & FAULT_FLAG_WRITE) {
2558 2559 2560
		ret |= do_wp_page(mm, vma, address, page_table, pmd, ptl, pte);
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2561 2562 2563 2564
		goto out;
	}

	/* No need to invalidate - it was non-present before */
2565
	update_mmu_cache(vma, address, page_table);
2566
unlock:
2567
	pte_unmap_unlock(page_table, ptl);
L
Linus Torvalds 已提交
2568 2569
out:
	return ret;
2570
out_nomap:
2571
	mem_cgroup_cancel_charge_swapin(ptr);
2572
	pte_unmap_unlock(page_table, ptl);
2573
out_page:
2574
	unlock_page(page);
2575
out_release:
2576
	page_cache_release(page);
2577
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2578 2579 2580
		unlock_page(swapcache);
		page_cache_release(swapcache);
	}
2581
	return ret;
L
Linus Torvalds 已提交
2582 2583
}

2584
/*
2585 2586
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
2587 2588 2589 2590 2591 2592
 * 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) {
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602
		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;
2603

2604
		expand_downwards(vma, address - PAGE_SIZE);
2605
	}
2606 2607 2608 2609 2610 2611 2612 2613 2614
	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;

		expand_upwards(vma, address + PAGE_SIZE);
	}
2615 2616 2617
	return 0;
}

L
Linus Torvalds 已提交
2618
/*
2619 2620 2621
 * 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
2622
 */
2623 2624
static int do_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
2625
		unsigned int flags)
L
Linus Torvalds 已提交
2626
{
2627 2628
	struct page *page;
	spinlock_t *ptl;
L
Linus Torvalds 已提交
2629 2630
	pte_t entry;

2631 2632 2633 2634
	pte_unmap(page_table);

	/* Check if we need to add a guard page to the stack */
	if (check_stack_guard_page(vma, address) < 0)
2635 2636
		return VM_FAULT_SIGBUS;

2637
	/* Use the zero-page for reads */
H
Hugh Dickins 已提交
2638 2639 2640
	if (!(flags & FAULT_FLAG_WRITE)) {
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(address),
						vma->vm_page_prot));
2641
		page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
H
Hugh Dickins 已提交
2642 2643 2644 2645 2646
		if (!pte_none(*page_table))
			goto unlock;
		goto setpte;
	}

N
Nick Piggin 已提交
2647 2648 2649 2650 2651 2652
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
	page = alloc_zeroed_user_highpage_movable(vma, address);
	if (!page)
		goto oom;
2653 2654 2655 2656 2657
	/*
	 * 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
2658
	__SetPageUptodate(page);
2659

2660
	if (mem_cgroup_charge_anon(page, mm, GFP_KERNEL))
2661 2662
		goto oom_free_page;

N
Nick Piggin 已提交
2663
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2664 2665
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2666

N
Nick Piggin 已提交
2667
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
2668
	if (!pte_none(*page_table))
N
Nick Piggin 已提交
2669
		goto release;
H
Hugh Dickins 已提交
2670

2671
	inc_mm_counter_fast(mm, MM_ANONPAGES);
N
Nick Piggin 已提交
2672
	page_add_new_anon_rmap(page, vma, address);
H
Hugh Dickins 已提交
2673
setpte:
2674
	set_pte_at(mm, address, page_table, entry);
L
Linus Torvalds 已提交
2675 2676

	/* No need to invalidate - it was non-present before */
2677
	update_mmu_cache(vma, address, page_table);
2678
unlock:
2679
	pte_unmap_unlock(page_table, ptl);
N
Nick Piggin 已提交
2680
	return 0;
2681
release:
2682
	mem_cgroup_uncharge_page(page);
2683 2684
	page_cache_release(page);
	goto unlock;
2685
oom_free_page:
2686
	page_cache_release(page);
2687
oom:
L
Linus Torvalds 已提交
2688 2689 2690
	return VM_FAULT_OOM;
}

2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721
static int __do_fault(struct vm_area_struct *vma, unsigned long address,
		pgoff_t pgoff, unsigned int flags, struct page **page)
{
	struct vm_fault vmf;
	int ret;

	vmf.virtual_address = (void __user *)(address & PAGE_MASK);
	vmf.pgoff = pgoff;
	vmf.flags = flags;
	vmf.page = NULL;

	ret = vma->vm_ops->fault(vma, &vmf);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

	if (unlikely(PageHWPoison(vmf.page))) {
		if (ret & VM_FAULT_LOCKED)
			unlock_page(vmf.page);
		page_cache_release(vmf.page);
		return VM_FAULT_HWPOISON;
	}

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

	*page = vmf.page;
	return ret;
}

2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737
/**
 * do_set_pte - setup new PTE entry for given page and add reverse page mapping.
 *
 * @vma: virtual memory area
 * @address: user virtual address
 * @page: page to map
 * @pte: pointer to target page table entry
 * @write: true, if new entry is writable
 * @anon: true, if it's anonymous page
 *
 * Caller must hold page table lock relevant for @pte.
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
void do_set_pte(struct vm_area_struct *vma, unsigned long address,
2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760
		struct page *page, pte_t *pte, bool write, bool anon)
{
	pte_t entry;

	flush_icache_page(vma, page);
	entry = mk_pte(page, vma->vm_page_prot);
	if (write)
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
	else if (pte_file(*pte) && pte_file_soft_dirty(*pte))
		pte_mksoft_dirty(entry);
	if (anon) {
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
		page_add_new_anon_rmap(page, vma, address);
	} else {
		inc_mm_counter_fast(vma->vm_mm, MM_FILEPAGES);
		page_add_file_rmap(page);
	}
	set_pte_at(vma->vm_mm, address, pte, entry);

	/* no need to invalidate: a not-present page won't be cached */
	update_mmu_cache(vma, address, pte);
}

2761
static unsigned long fault_around_bytes = rounddown_pow_of_two(65536);
2762 2763 2764

static inline unsigned long fault_around_pages(void)
{
2765
	return fault_around_bytes >> PAGE_SHIFT;
2766 2767 2768 2769
}

static inline unsigned long fault_around_mask(void)
{
2770
	return ~(fault_around_bytes - 1) & PAGE_MASK;
2771
}
2772

2773 2774
#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
2775
{
2776
	*val = fault_around_bytes;
2777 2778 2779
	return 0;
}

2780 2781 2782 2783 2784
/*
 * 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.
 */
2785
static int fault_around_bytes_set(void *data, u64 val)
2786
{
2787
	if (val / PAGE_SIZE > PTRS_PER_PTE)
2788
		return -EINVAL;
2789 2790 2791 2792
	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 */
2793 2794
	return 0;
}
2795 2796
DEFINE_SIMPLE_ATTRIBUTE(fault_around_bytes_fops,
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
2797 2798 2799 2800 2801

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

2802 2803
	ret = debugfs_create_file("fault_around_bytes", 0644, NULL, NULL,
			&fault_around_bytes_fops);
2804
	if (!ret)
2805
		pr_warn("Failed to create fault_around_bytes in debugfs");
2806 2807 2808 2809
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
2810

2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833
/*
 * 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.
 */
2834 2835 2836 2837 2838 2839 2840 2841
static void do_fault_around(struct vm_area_struct *vma, unsigned long address,
		pte_t *pte, pgoff_t pgoff, unsigned int flags)
{
	unsigned long start_addr;
	pgoff_t max_pgoff;
	struct vm_fault vmf;
	int off;

2842
	start_addr = max(address & fault_around_mask(), vma->vm_start);
2843 2844 2845 2846 2847 2848
	off = ((address - start_addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
	pte -= off;
	pgoff -= off;

	/*
	 *  max_pgoff is either end of page table or end of vma
2849
	 *  or fault_around_pages() from pgoff, depending what is nearest.
2850 2851 2852 2853
	 */
	max_pgoff = pgoff - ((start_addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
		PTRS_PER_PTE - 1;
	max_pgoff = min3(max_pgoff, vma_pages(vma) + vma->vm_pgoff - 1,
2854
			pgoff + fault_around_pages() - 1);
2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873

	/* Check if it makes any sense to call ->map_pages */
	while (!pte_none(*pte)) {
		if (++pgoff > max_pgoff)
			return;
		start_addr += PAGE_SIZE;
		if (start_addr >= vma->vm_end)
			return;
		pte++;
	}

	vmf.virtual_address = (void __user *) start_addr;
	vmf.pte = pte;
	vmf.pgoff = pgoff;
	vmf.max_pgoff = max_pgoff;
	vmf.flags = flags;
	vma->vm_ops->map_pages(vma, &vmf);
}

2874 2875 2876 2877 2878 2879
static int do_read_fault(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pmd_t *pmd,
		pgoff_t pgoff, unsigned int flags, pte_t orig_pte)
{
	struct page *fault_page;
	spinlock_t *ptl;
2880
	pte_t *pte;
2881 2882 2883 2884 2885 2886 2887
	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).
	 */
2888 2889
	if (vma->vm_ops->map_pages && !(flags & FAULT_FLAG_NONLINEAR) &&
	    fault_around_pages() > 1) {
2890 2891 2892 2893 2894 2895
		pte = pte_offset_map_lock(mm, pmd, address, &ptl);
		do_fault_around(vma, address, pte, pgoff, flags);
		if (!pte_same(*pte, orig_pte))
			goto unlock_out;
		pte_unmap_unlock(pte, ptl);
	}
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907

	ret = __do_fault(vma, address, pgoff, flags, &fault_page);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
	if (unlikely(!pte_same(*pte, orig_pte))) {
		pte_unmap_unlock(pte, ptl);
		unlock_page(fault_page);
		page_cache_release(fault_page);
		return ret;
	}
2908
	do_set_pte(vma, address, fault_page, pte, false, false);
2909
	unlock_page(fault_page);
2910 2911
unlock_out:
	pte_unmap_unlock(pte, ptl);
2912 2913 2914
	return ret;
}

2915 2916 2917 2918 2919 2920
static int do_cow_fault(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pmd_t *pmd,
		pgoff_t pgoff, unsigned int flags, pte_t orig_pte)
{
	struct page *fault_page, *new_page;
	spinlock_t *ptl;
2921
	pte_t *pte;
2922 2923 2924 2925 2926 2927 2928 2929 2930
	int ret;

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

	new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
	if (!new_page)
		return VM_FAULT_OOM;

2931
	if (mem_cgroup_charge_anon(new_page, mm, GFP_KERNEL)) {
2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949
		page_cache_release(new_page);
		return VM_FAULT_OOM;
	}

	ret = __do_fault(vma, address, pgoff, flags, &fault_page);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;

	copy_user_highpage(new_page, fault_page, address, vma);
	__SetPageUptodate(new_page);

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
	if (unlikely(!pte_same(*pte, orig_pte))) {
		pte_unmap_unlock(pte, ptl);
		unlock_page(fault_page);
		page_cache_release(fault_page);
		goto uncharge_out;
	}
2950
	do_set_pte(vma, address, new_page, pte, true, true);
2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
	pte_unmap_unlock(pte, ptl);
	unlock_page(fault_page);
	page_cache_release(fault_page);
	return ret;
uncharge_out:
	mem_cgroup_uncharge_page(new_page);
	page_cache_release(new_page);
	return ret;
}

2961
static int do_shared_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2962
		unsigned long address, pmd_t *pmd,
2963
		pgoff_t pgoff, unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
2964
{
2965 2966
	struct page *fault_page;
	struct address_space *mapping;
2967
	spinlock_t *ptl;
2968
	pte_t *pte;
2969 2970
	int dirtied = 0;
	int ret, tmp;
2971

2972 2973
	ret = __do_fault(vma, address, pgoff, flags, &fault_page);
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
2974
		return ret;
L
Linus Torvalds 已提交
2975 2976

	/*
2977 2978
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
2979
	 */
2980 2981 2982 2983 2984
	if (vma->vm_ops->page_mkwrite) {
		unlock_page(fault_page);
		tmp = do_page_mkwrite(vma, fault_page, address);
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
2985
			page_cache_release(fault_page);
2986
			return tmp;
2987
		}
2988 2989
	}

2990 2991 2992 2993 2994 2995
	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
	if (unlikely(!pte_same(*pte, orig_pte))) {
		pte_unmap_unlock(pte, ptl);
		unlock_page(fault_page);
		page_cache_release(fault_page);
		return ret;
L
Linus Torvalds 已提交
2996
	}
2997
	do_set_pte(vma, address, fault_page, pte, true, false);
2998
	pte_unmap_unlock(pte, ptl);
N
Nick Piggin 已提交
2999

3000 3001 3002 3003 3004 3005 3006 3007 3008 3009
	if (set_page_dirty(fault_page))
		dirtied = 1;
	mapping = fault_page->mapping;
	unlock_page(fault_page);
	if ((dirtied || vma->vm_ops->page_mkwrite) && mapping) {
		/*
		 * Some device drivers do not set page.mapping but still
		 * dirty their pages
		 */
		balance_dirty_pages_ratelimited(mapping);
3010
	}
3011

3012 3013 3014
	/* file_update_time outside page_lock */
	if (vma->vm_file && !vma->vm_ops->page_mkwrite)
		file_update_time(vma->vm_file);
N
Nick Piggin 已提交
3015

3016
	return ret;
3017
}
3018

3019 3020
static int do_linear_fault(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
3021
		unsigned int flags, pte_t orig_pte)
3022 3023
{
	pgoff_t pgoff = (((address & PAGE_MASK)
3024
			- vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
3025

3026
	pte_unmap(page_table);
3027 3028 3029
	if (!(flags & FAULT_FLAG_WRITE))
		return do_read_fault(mm, vma, address, pmd, pgoff, flags,
				orig_pte);
3030 3031 3032
	if (!(vma->vm_flags & VM_SHARED))
		return do_cow_fault(mm, vma, address, pmd, pgoff, flags,
				orig_pte);
3033
	return do_shared_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
3034 3035
}

L
Linus Torvalds 已提交
3036 3037 3038 3039
/*
 * Fault of a previously existing named mapping. Repopulate the pte
 * from the encoded file_pte if possible. This enables swappable
 * nonlinear vmas.
3040 3041 3042 3043
 *
 * 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
3044
 */
N
Nick Piggin 已提交
3045
static int do_nonlinear_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3046
		unsigned long address, pte_t *page_table, pmd_t *pmd,
3047
		unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
3048
{
3049
	pgoff_t pgoff;
L
Linus Torvalds 已提交
3050

3051 3052
	flags |= FAULT_FLAG_NONLINEAR;

H
Hugh Dickins 已提交
3053
	if (!pte_unmap_same(mm, pmd, page_table, orig_pte))
N
Nick Piggin 已提交
3054
		return 0;
L
Linus Torvalds 已提交
3055

3056
	if (unlikely(!(vma->vm_flags & VM_NONLINEAR))) {
3057 3058 3059
		/*
		 * Page table corrupted: show pte and kill process.
		 */
3060
		print_bad_pte(vma, address, orig_pte, NULL);
H
Hugh Dickins 已提交
3061
		return VM_FAULT_SIGBUS;
3062 3063 3064
	}

	pgoff = pte_to_pgoff(orig_pte);
3065 3066 3067
	if (!(flags & FAULT_FLAG_WRITE))
		return do_read_fault(mm, vma, address, pmd, pgoff, flags,
				orig_pte);
3068 3069 3070
	if (!(vma->vm_flags & VM_SHARED))
		return do_cow_fault(mm, vma, address, pmd, pgoff, flags,
				orig_pte);
3071
	return do_shared_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
L
Linus Torvalds 已提交
3072 3073
}

3074
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3075 3076
				unsigned long addr, int page_nid,
				int *flags)
3077 3078 3079 3080
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3081
	if (page_nid == numa_node_id()) {
3082
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3083 3084
		*flags |= TNF_FAULT_LOCAL;
	}
3085 3086 3087 3088

	return mpol_misplaced(page, vma, addr);
}

3089
static int do_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
3090 3091
		   unsigned long addr, pte_t pte, pte_t *ptep, pmd_t *pmd)
{
3092
	struct page *page = NULL;
3093
	spinlock_t *ptl;
3094
	int page_nid = -1;
3095
	int last_cpupid;
3096
	int target_nid;
3097
	bool migrated = false;
3098
	int flags = 0;
3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110

	/*
	* 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.
	*
	* ptep_modify_prot_start is not called as this is clearing
	* the _PAGE_NUMA bit and it is not really expected that there
	* would be concurrent hardware modifications to the PTE.
	*/
	ptl = pte_lockptr(mm, pmd);
	spin_lock(ptl);
3111 3112 3113 3114 3115
	if (unlikely(!pte_same(*ptep, pte))) {
		pte_unmap_unlock(ptep, ptl);
		goto out;
	}

3116 3117 3118 3119 3120 3121 3122 3123 3124
	pte = pte_mknonnuma(pte);
	set_pte_at(mm, addr, ptep, pte);
	update_mmu_cache(vma, addr, ptep);

	page = vm_normal_page(vma, addr, pte);
	if (!page) {
		pte_unmap_unlock(ptep, ptl);
		return 0;
	}
3125
	BUG_ON(is_zero_pfn(page_to_pfn(page)));
3126

3127 3128 3129 3130 3131 3132 3133 3134
	/*
	 * Avoid grouping on DSO/COW pages in specific and RO pages
	 * in general, RO pages shouldn't hurt as much anyway since
	 * they can be in shared cache state.
	 */
	if (!pte_write(pte))
		flags |= TNF_NO_GROUP;

3135 3136 3137 3138 3139 3140 3141
	/*
	 * 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;

3142
	last_cpupid = page_cpupid_last(page);
3143
	page_nid = page_to_nid(page);
3144
	target_nid = numa_migrate_prep(page, vma, addr, page_nid, &flags);
3145
	pte_unmap_unlock(ptep, ptl);
3146 3147 3148 3149 3150 3151
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3152
	migrated = migrate_misplaced_page(page, vma, target_nid);
3153
	if (migrated) {
3154
		page_nid = target_nid;
3155 3156
		flags |= TNF_MIGRATED;
	}
3157 3158

out:
3159
	if (page_nid != -1)
3160
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3161 3162 3163
	return 0;
}

L
Linus Torvalds 已提交
3164 3165 3166 3167 3168 3169 3170 3171 3172
/*
 * 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).
 *
H
Hugh Dickins 已提交
3173 3174 3175
 * 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
3176
 */
3177
static int handle_pte_fault(struct mm_struct *mm,
3178 3179
		     struct vm_area_struct *vma, unsigned long address,
		     pte_t *pte, pmd_t *pmd, unsigned int flags)
L
Linus Torvalds 已提交
3180 3181
{
	pte_t entry;
3182
	spinlock_t *ptl;
L
Linus Torvalds 已提交
3183

3184
	entry = ACCESS_ONCE(*pte);
L
Linus Torvalds 已提交
3185
	if (!pte_present(entry)) {
3186
		if (pte_none(entry)) {
J
Jes Sorensen 已提交
3187
			if (vma->vm_ops) {
N
Nick Piggin 已提交
3188
				if (likely(vma->vm_ops->fault))
3189
					return do_linear_fault(mm, vma, address,
3190
						pte, pmd, flags, entry);
J
Jes Sorensen 已提交
3191 3192
			}
			return do_anonymous_page(mm, vma, address,
3193
						 pte, pmd, flags);
3194
		}
L
Linus Torvalds 已提交
3195
		if (pte_file(entry))
N
Nick Piggin 已提交
3196
			return do_nonlinear_fault(mm, vma, address,
3197
					pte, pmd, flags, entry);
3198
		return do_swap_page(mm, vma, address,
3199
					pte, pmd, flags, entry);
L
Linus Torvalds 已提交
3200 3201
	}

3202 3203 3204
	if (pte_numa(entry))
		return do_numa_page(mm, vma, address, entry, pte, pmd);

H
Hugh Dickins 已提交
3205
	ptl = pte_lockptr(mm, pmd);
3206 3207 3208
	spin_lock(ptl);
	if (unlikely(!pte_same(*pte, entry)))
		goto unlock;
3209
	if (flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3210
		if (!pte_write(entry))
3211 3212
			return do_wp_page(mm, vma, address,
					pte, pmd, ptl, entry);
L
Linus Torvalds 已提交
3213 3214 3215
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
3216
	if (ptep_set_access_flags(vma, address, pte, entry, flags & FAULT_FLAG_WRITE)) {
3217
		update_mmu_cache(vma, address, pte);
3218 3219 3220 3221 3222 3223 3224
	} 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.
		 */
3225
		if (flags & FAULT_FLAG_WRITE)
3226
			flush_tlb_fix_spurious_fault(vma, address);
3227
	}
3228 3229
unlock:
	pte_unmap_unlock(pte, ptl);
N
Nick Piggin 已提交
3230
	return 0;
L
Linus Torvalds 已提交
3231 3232 3233 3234 3235
}

/*
 * By the time we get here, we already hold the mm semaphore
 */
3236 3237
static int __handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
			     unsigned long address, unsigned int flags)
L
Linus Torvalds 已提交
3238 3239 3240 3241 3242 3243
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

3244
	if (unlikely(is_vm_hugetlb_page(vma)))
3245
		return hugetlb_fault(mm, vma, address, flags);
L
Linus Torvalds 已提交
3246 3247 3248 3249

	pgd = pgd_offset(mm, address);
	pud = pud_alloc(mm, pgd, address);
	if (!pud)
H
Hugh Dickins 已提交
3250
		return VM_FAULT_OOM;
L
Linus Torvalds 已提交
3251 3252
	pmd = pmd_alloc(mm, pud, address);
	if (!pmd)
H
Hugh Dickins 已提交
3253
		return VM_FAULT_OOM;
3254
	if (pmd_none(*pmd) && transparent_hugepage_enabled(vma)) {
3255
		int ret = VM_FAULT_FALLBACK;
3256
		if (!vma->vm_ops)
3257 3258 3259 3260
			ret = do_huge_pmd_anonymous_page(mm, vma, address,
					pmd, flags);
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3261 3262
	} else {
		pmd_t orig_pmd = *pmd;
3263 3264
		int ret;

3265 3266
		barrier();
		if (pmd_trans_huge(orig_pmd)) {
3267 3268
			unsigned int dirty = flags & FAULT_FLAG_WRITE;

3269 3270 3271 3272 3273 3274 3275 3276
			/*
			 * If the pmd is splitting, return and retry the
			 * the fault.  Alternative: wait until the split
			 * is done, and goto retry.
			 */
			if (pmd_trans_splitting(orig_pmd))
				return 0;

3277
			if (pmd_numa(orig_pmd))
3278
				return do_huge_pmd_numa_page(mm, vma, address,
3279 3280
							     orig_pmd, pmd);

3281
			if (dirty && !pmd_write(orig_pmd)) {
3282 3283
				ret = do_huge_pmd_wp_page(mm, vma, address, pmd,
							  orig_pmd);
3284 3285
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3286 3287 3288
			} else {
				huge_pmd_set_accessed(mm, vma, address, pmd,
						      orig_pmd, dirty);
3289
				return 0;
3290
			}
3291 3292 3293 3294 3295 3296 3297 3298
		}
	}

	/*
	 * Use __pte_alloc instead of pte_alloc_map, because we can't
	 * run pte_offset_map on the pmd, if an huge pmd could
	 * materialize from under us from a different thread.
	 */
3299 3300
	if (unlikely(pmd_none(*pmd)) &&
	    unlikely(__pte_alloc(mm, vma, pmd, address)))
H
Hugh Dickins 已提交
3301
		return VM_FAULT_OOM;
3302 3303 3304 3305 3306 3307 3308 3309 3310 3311
	/* if an huge pmd materialized from under us just retry later */
	if (unlikely(pmd_trans_huge(*pmd)))
		return 0;
	/*
	 * A regular pmd is established and it can't morph into a huge pmd
	 * from under us anymore at this point because we hold the mmap_sem
	 * read mode and khugepaged takes it in write mode. So now it's
	 * safe to run pte_offset_map().
	 */
	pte = pte_offset_map(pmd, address);
L
Linus Torvalds 已提交
3312

3313
	return handle_pte_fault(mm, vma, address, pte, pmd, flags);
L
Linus Torvalds 已提交
3314 3315
}

3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333
int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
		    unsigned long address, unsigned int flags)
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
	mem_cgroup_count_vm_event(mm, PGFAULT);

	/* 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)
3334
		mem_cgroup_oom_enable();
3335 3336 3337

	ret = __handle_mm_fault(mm, vma, address, flags);

3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
	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);
	}
3349

3350 3351 3352
	return ret;
}

L
Linus Torvalds 已提交
3353 3354 3355
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3356
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3357
 */
3358
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3359
{
H
Hugh Dickins 已提交
3360 3361
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3362
		return -ENOMEM;
L
Linus Torvalds 已提交
3363

3364 3365
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3366
	spin_lock(&mm->page_table_lock);
3367
	if (pgd_present(*pgd))		/* Another has populated it */
3368
		pud_free(mm, new);
3369 3370
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3371
	spin_unlock(&mm->page_table_lock);
3372
	return 0;
L
Linus Torvalds 已提交
3373 3374 3375 3376 3377 3378
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3379
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3380
 */
3381
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3382
{
H
Hugh Dickins 已提交
3383 3384
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3385
		return -ENOMEM;
L
Linus Torvalds 已提交
3386

3387 3388
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3389
	spin_lock(&mm->page_table_lock);
L
Linus Torvalds 已提交
3390
#ifndef __ARCH_HAS_4LEVEL_HACK
3391
	if (pud_present(*pud))		/* Another has populated it */
3392
		pmd_free(mm, new);
3393 3394
	else
		pud_populate(mm, pud, new);
L
Linus Torvalds 已提交
3395
#else
3396
	if (pgd_present(*pud))		/* Another has populated it */
3397
		pmd_free(mm, new);
3398 3399
	else
		pgd_populate(mm, pud, new);
L
Linus Torvalds 已提交
3400
#endif /* __ARCH_HAS_4LEVEL_HACK */
H
Hugh Dickins 已提交
3401
	spin_unlock(&mm->page_table_lock);
3402
	return 0;
3403
}
L
Linus Torvalds 已提交
3404 3405 3406 3407 3408
#endif /* __PAGETABLE_PMD_FOLDED */

#if !defined(__HAVE_ARCH_GATE_AREA)

#if defined(AT_SYSINFO_EHDR)
3409
static struct vm_area_struct gate_vma;
L
Linus Torvalds 已提交
3410 3411 3412 3413 3414 3415

static int __init gate_vma_init(void)
{
	gate_vma.vm_mm = NULL;
	gate_vma.vm_start = FIXADDR_USER_START;
	gate_vma.vm_end = FIXADDR_USER_END;
R
Roland McGrath 已提交
3416 3417
	gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
	gate_vma.vm_page_prot = __P101;
3418

L
Linus Torvalds 已提交
3419 3420 3421 3422 3423
	return 0;
}
__initcall(gate_vma_init);
#endif

3424
struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
L
Linus Torvalds 已提交
3425 3426 3427 3428 3429 3430 3431 3432
{
#ifdef AT_SYSINFO_EHDR
	return &gate_vma;
#else
	return NULL;
#endif
}

3433
int in_gate_area_no_mm(unsigned long addr)
L
Linus Torvalds 已提交
3434 3435 3436 3437 3438 3439 3440 3441 3442
{
#ifdef AT_SYSINFO_EHDR
	if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
		return 1;
#endif
	return 0;
}

#endif	/* __HAVE_ARCH_GATE_AREA */
3443

3444
static int __follow_pte(struct mm_struct *mm, unsigned long address,
J
Johannes Weiner 已提交
3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460
		pte_t **ptepp, spinlock_t **ptlp)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

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

	pmd = pmd_offset(pud, address);
3461
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481
	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
		goto out;

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

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

3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
{
	int res;

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

J
Johannes Weiner 已提交
3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521
/**
 * 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);

3522
#ifdef CONFIG_HAVE_IOREMAP_PROT
3523 3524 3525
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
3526
{
3527
	int ret = -EINVAL;
3528 3529 3530
	pte_t *ptep, pte;
	spinlock_t *ptl;

3531 3532
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
3533

3534
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
3535
		goto out;
3536
	pte = *ptep;
3537

3538 3539 3540 3541
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
3542
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
3543

3544
	ret = 0;
3545 3546 3547
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
3548
	return ret;
3549 3550 3551 3552 3553 3554 3555
}

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
3556
	void __iomem *maddr;
3557 3558
	int offset = addr & (PAGE_SIZE-1);

3559
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
		return -EINVAL;

	maddr = ioremap_prot(phys_addr, PAGE_SIZE, prot);
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
3571
EXPORT_SYMBOL_GPL(generic_access_phys);
3572 3573
#endif

3574
/*
3575 3576
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
3577
 */
3578 3579
static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
		unsigned long addr, void *buf, int len, int write)
3580 3581 3582 3583 3584
{
	struct vm_area_struct *vma;
	void *old_buf = buf;

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
3585
	/* ignore errors, just check how much was successfully transferred */
3586 3587 3588
	while (len) {
		int bytes, ret, offset;
		void *maddr;
3589
		struct page *page = NULL;
3590 3591 3592

		ret = get_user_pages(tsk, mm, addr, 1,
				write, 1, &page, &vma);
3593 3594 3595 3596 3597 3598 3599
		if (ret <= 0) {
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
#ifdef CONFIG_HAVE_IOREMAP_PROT
			vma = find_vma(mm, addr);
3600
			if (!vma || vma->vm_start > addr)
3601 3602 3603 3604 3605 3606 3607 3608
				break;
			if (vma->vm_ops && vma->vm_ops->access)
				ret = vma->vm_ops->access(vma, addr, buf,
							  len, write);
			if (ret <= 0)
#endif
				break;
			bytes = ret;
3609
		} else {
3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625
			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);
			page_cache_release(page);
3626 3627 3628 3629 3630 3631 3632 3633 3634
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
3635

S
Stephen Wilson 已提交
3636
/**
3637
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651
 * @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
 * @write:	whether the access is a write
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
		void *buf, int len, int write)
{
	return __access_remote_vm(NULL, mm, addr, buf, len, write);
}

3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672
/*
 * 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,
		void *buf, int len, int write)
{
	struct mm_struct *mm;
	int ret;

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

	ret = __access_remote_vm(tsk, mm, addr, buf, len, write);
	mmput(mm);

	return ret;
}

3673 3674 3675 3676 3677 3678 3679 3680
/*
 * 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;

3681 3682 3683 3684 3685 3686 3687
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

3688 3689 3690 3691 3692 3693
	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
3694
			char *p;
3695

3696
			p = d_path(&f->f_path, buf, PAGE_SIZE);
3697 3698
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
3699
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
3700 3701 3702 3703 3704
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
3705
	up_read(&mm->mmap_sem);
3706
}
3707

3708
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
3709 3710
void might_fault(void)
{
3711 3712 3713 3714 3715 3716 3717 3718 3719
	/*
	 * 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;

3720 3721 3722 3723 3724
	/*
	 * it would be nicer only to annotate paths which are not under
	 * pagefault_disable, however that requires a larger audit and
	 * providing helpers like get_user_atomic.
	 */
3725 3726 3727 3728 3729 3730
	if (in_atomic())
		return;

	__might_sleep(__FILE__, __LINE__, 0);

	if (current->mm)
3731 3732 3733 3734
		might_lock_read(&current->mm->mmap_sem);
}
EXPORT_SYMBOL(might_fault);
#endif
A
Andrea Arcangeli 已提交
3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805

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

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

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

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

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

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

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

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

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

	might_sleep();
	for (i = 0; i < pages_per_huge_page; i++) {
		cond_resched();
		copy_user_highpage(dst + i, src + i, addr + i*PAGE_SIZE, vma);
	}
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3806

3807
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
3808 3809 3810 3811 3812 3813 3814 3815 3816

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);
}

3817
bool ptlock_alloc(struct page *page)
3818 3819 3820
{
	spinlock_t *ptl;

3821
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
3822 3823
	if (!ptl)
		return false;
3824
	page->ptl = ptl;
3825 3826 3827
	return true;
}

3828
void ptlock_free(struct page *page)
3829
{
3830
	kmem_cache_free(page_ptl_cachep, page->ptl);
3831 3832
}
#endif