memory.c 116.8 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>
L
Linus Torvalds 已提交
62

A
Alexey Dobriyan 已提交
63
#include <asm/io.h>
L
Linus Torvalds 已提交
64 65 66 67 68 69
#include <asm/pgalloc.h>
#include <asm/uaccess.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

70 71
#include "internal.h"

72 73 74 75
#ifdef LAST_NID_NOT_IN_PAGE_FLAGS
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_nid.
#endif

A
Andy Whitcroft 已提交
76
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
/* 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);

96 97 98 99 100 101 102 103 104 105 106 107
/*
 * 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
108 109 110 111

static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
112
	return 1;
113 114 115
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
116
unsigned long zero_pfn __read_mostly;
H
Hugh Dickins 已提交
117
unsigned long highest_memmap_pfn __read_mostly;
H
Hugh Dickins 已提交
118 119 120 121 122 123 124 125 126 127

/*
 * 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);
128

K
KAMEZAWA Hiroyuki 已提交
129

130 131
#if defined(SPLIT_RSS_COUNTING)

132
void sync_mm_rss(struct mm_struct *mm)
133 134 135 136
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
137 138 139
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
140 141
		}
	}
142
	current->rss_stat.events = 0;
143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
}

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))
164
		sync_mm_rss(task->mm);
165
}
166
#else /* SPLIT_RSS_COUNTING */
167 168 169 170 171 172 173 174

#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)
{
}

175 176 177 178 179 180 181 182 183 184 185 186 187 188
#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;
	}

189 190 191
	if (tlb->batch_count == MAX_GATHER_BATCH_COUNT)
		return 0;

192 193 194 195
	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
	if (!batch)
		return 0;

196
	tlb->batch_count++;
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
	batch->next = NULL;
	batch->nr   = 0;
	batch->max  = MAX_GATHER_BATCH;

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

	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).
 */
212
void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long start, unsigned long end)
213 214 215
{
	tlb->mm = mm;

216 217
	/* Is it from 0 to ~0? */
	tlb->fullmm     = !(start | (end+1));
218
	tlb->need_flush_all = 0;
219 220
	tlb->start	= start;
	tlb->end	= end;
221 222 223 224 225
	tlb->need_flush = 0;
	tlb->local.next = NULL;
	tlb->local.nr   = 0;
	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
	tlb->active     = &tlb->local;
226
	tlb->batch_count = 0;
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242

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

void tlb_flush_mmu(struct mmu_gather *tlb)
{
	struct mmu_gather_batch *batch;

	if (!tlb->need_flush)
		return;
	tlb->need_flush = 0;
	tlb_flush(tlb);
#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb_table_flush(tlb);
243 244
#endif

245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
	for (batch = &tlb->local; batch; batch = batch->next) {
		free_pages_and_swap_cache(batch->pages, batch->nr);
		batch->nr = 0;
	}
	tlb->active = &tlb->local;
}

/* 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
282
	VM_BUG_ON(!tlb->need_flush);
283 284 285 286 287 288

	batch = tlb->active;
	batch->pages[batch->nr++] = page;
	if (batch->nr == batch->max) {
		if (!tlb_next_batch(tlb))
			return 0;
289
		batch = tlb->active;
290 291 292 293 294 295 296 297
	}
	VM_BUG_ON(batch->nr > batch->max);

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

#endif /* HAVE_GENERIC_MMU_GATHER */

298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
#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);
}

373
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
374

L
Linus Torvalds 已提交
375 376 377 378
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
379 380
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
381
{
382
	pgtable_t token = pmd_pgtable(*pmd);
383
	pmd_clear(pmd);
384
	pte_free_tlb(tlb, token, addr);
385
	tlb->mm->nr_ptes--;
L
Linus Torvalds 已提交
386 387
}

388 389 390
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
391 392 393
{
	pmd_t *pmd;
	unsigned long next;
394
	unsigned long start;
L
Linus Torvalds 已提交
395

396
	start = addr;
L
Linus Torvalds 已提交
397 398 399 400 401
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
402
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
403 404
	} while (pmd++, addr = next, addr != end);

405 406 407 408 409 410 411
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
412
	}
413 414 415 416 417
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
418
	pmd_free_tlb(tlb, pmd, start);
L
Linus Torvalds 已提交
419 420
}

421 422 423
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
424 425 426
{
	pud_t *pud;
	unsigned long next;
427
	unsigned long start;
L
Linus Torvalds 已提交
428

429
	start = addr;
L
Linus Torvalds 已提交
430 431 432 433 434
	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
435
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
436 437
	} while (pud++, addr = next, addr != end);

438 439 440 441 442 443 444
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
445
	}
446 447 448 449 450
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(pgd, start);
	pgd_clear(pgd);
451
	pud_free_tlb(tlb, pud, start);
L
Linus Torvalds 已提交
452 453 454
}

/*
455 456
 * This function frees user-level page tables of a process.
 *
L
Linus Torvalds 已提交
457 458
 * Must be called with pagetable lock held.
 */
459
void free_pgd_range(struct mmu_gather *tlb,
460 461
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
462 463 464
{
	pgd_t *pgd;
	unsigned long next;
465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490

	/*
	 * 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
491

492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
	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;

508
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
509 510 511 512
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
513
		free_pud_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
514
	} while (pgd++, addr = next, addr != end);
515 516
}

517
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
518
		unsigned long floor, unsigned long ceiling)
519 520 521 522 523
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

524
		/*
N
npiggin@suse.de 已提交
525 526
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
527
		 */
528
		unlink_anon_vmas(vma);
529 530
		unlink_file_vma(vma);

531
		if (is_vm_hugetlb_page(vma)) {
532
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
533
				floor, next? next->vm_start: ceiling);
534 535 536 537 538
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
539
			       && !is_vm_hugetlb_page(next)) {
540 541
				vma = next;
				next = vma->vm_next;
542
				unlink_anon_vmas(vma);
543
				unlink_file_vma(vma);
544 545 546 547
			}
			free_pgd_range(tlb, addr, vma->vm_end,
				floor, next? next->vm_start: ceiling);
		}
548 549
		vma = next;
	}
L
Linus Torvalds 已提交
550 551
}

552 553
int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
		pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
554
{
555
	pgtable_t new = pte_alloc_one(mm, address);
556
	int wait_split_huge_page;
557 558 559
	if (!new)
		return -ENOMEM;

560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
	/*
	 * 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 */

H
Hugh Dickins 已提交
575
	spin_lock(&mm->page_table_lock);
576 577
	wait_split_huge_page = 0;
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
L
Linus Torvalds 已提交
578 579
		mm->nr_ptes++;
		pmd_populate(mm, pmd, new);
580
		new = NULL;
581 582
	} else if (unlikely(pmd_trans_splitting(*pmd)))
		wait_split_huge_page = 1;
H
Hugh Dickins 已提交
583
	spin_unlock(&mm->page_table_lock);
584 585
	if (new)
		pte_free(mm, new);
586 587
	if (wait_split_huge_page)
		wait_split_huge_page(vma->anon_vma, pmd);
588
	return 0;
L
Linus Torvalds 已提交
589 590
}

591
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
592
{
593 594 595 596
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

597 598
	smp_wmb(); /* See comment in __pte_alloc */

599
	spin_lock(&init_mm.page_table_lock);
600
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
601
		pmd_populate_kernel(&init_mm, pmd, new);
602
		new = NULL;
603 604
	} else
		VM_BUG_ON(pmd_trans_splitting(*pmd));
605
	spin_unlock(&init_mm.page_table_lock);
606 607
	if (new)
		pte_free_kernel(&init_mm, new);
608
	return 0;
L
Linus Torvalds 已提交
609 610
}

K
KAMEZAWA Hiroyuki 已提交
611 612 613 614 615 616
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)
617
{
K
KAMEZAWA Hiroyuki 已提交
618 619
	int i;

620
	if (current->mm == mm)
621
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
622 623 624
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
625 626
}

N
Nick Piggin 已提交
627
/*
628 629 630
 * 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
631 632 633
 *
 * The calling function must still handle the error.
 */
634 635
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
636
{
637 638 639 640 641
	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;
642 643 644 645 646 647 648 649 650 651 652 653 654 655
	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) {
656 657
			printk(KERN_ALERT
				"BUG: Bad page map: %lu messages suppressed\n",
658 659 660 661 662 663 664
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
665 666 667 668

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

669 670
	printk(KERN_ALERT
		"BUG: Bad page map in process %s  pte:%08llx pmd:%08llx\n",
671 672
		current->comm,
		(long long)pte_val(pte), (long long)pmd_val(*pmd));
673 674
	if (page)
		dump_page(page);
675
	printk(KERN_ALERT
676 677 678 679 680 681
		"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
682 683
		printk(KERN_ALERT "vma->vm_ops->fault: %pSR\n",
		       vma->vm_ops->fault);
684
	if (vma->vm_file && vma->vm_file->f_op)
J
Joe Perches 已提交
685 686
		printk(KERN_ALERT "vma->vm_file->f_op->mmap: %pSR\n",
		       vma->vm_file->f_op->mmap);
N
Nick Piggin 已提交
687
	dump_stack();
688
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
689 690
}

691
static inline bool is_cow_mapping(vm_flags_t flags)
692 693 694 695
{
	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
}

H
Hugh Dickins 已提交
696
/*
N
Nick Piggin 已提交
697
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
698
 *
N
Nick Piggin 已提交
699 700 701
 * "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
702
 *
N
Nick Piggin 已提交
703 704 705 706 707 708 709 710
 * 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.
711
 *
J
Jared Hulbert 已提交
712 713
 * 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
714 715
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
716 717 718
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
719 720 721 722 723 724
 * 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
725 726
 *
 *
N
Nick Piggin 已提交
727
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
728 729 730 731 732 733 734 735 736
 *
 * 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
737
 */
N
Nick Piggin 已提交
738 739 740 741 742 743 744
#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
745
{
746
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
747 748

	if (HAVE_PTE_SPECIAL) {
749 750
		if (likely(!pte_special(pte)))
			goto check_pfn;
H
Hugh Dickins 已提交
751 752
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
H
Hugh Dickins 已提交
753
		if (!is_zero_pfn(pfn))
754
			print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
755 756 757 758 759
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
760 761 762 763 764 765
	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
766 767
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
768 769 770 771 772
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
773 774
	}

H
Hugh Dickins 已提交
775 776
	if (is_zero_pfn(pfn))
		return NULL;
777 778 779 780 781
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
782 783

	/*
N
Nick Piggin 已提交
784 785
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
786
	 */
J
Jared Hulbert 已提交
787
out:
788
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
789 790
}

L
Linus Torvalds 已提交
791 792 793 794 795 796
/*
 * 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
797
static inline unsigned long
L
Linus Torvalds 已提交
798
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
799
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
800
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
801
{
N
Nick Piggin 已提交
802
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
803 804 805 806 807 808
	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)) {
809 810
			swp_entry_t entry = pte_to_swp_entry(pte);

H
Hugh Dickins 已提交
811 812 813
			if (swap_duplicate(entry) < 0)
				return entry.val;

L
Linus Torvalds 已提交
814 815 816
			/* make sure dst_mm is on swapoff's mmlist. */
			if (unlikely(list_empty(&dst_mm->mmlist))) {
				spin_lock(&mmlist_lock);
817 818 819
				if (list_empty(&dst_mm->mmlist))
					list_add(&dst_mm->mmlist,
						 &src_mm->mmlist);
L
Linus Torvalds 已提交
820 821
				spin_unlock(&mmlist_lock);
			}
K
KAMEZAWA Hiroyuki 已提交
822 823
			if (likely(!non_swap_entry(entry)))
				rss[MM_SWAPENTS]++;
824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839
			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);
840 841
					if (pte_swp_soft_dirty(*src_pte))
						pte = pte_swp_mksoft_dirty(pte);
842 843
					set_pte_at(src_mm, addr, src_pte, pte);
				}
844
			}
L
Linus Torvalds 已提交
845
		}
846
		goto out_set_pte;
L
Linus Torvalds 已提交
847 848 849 850 851 852
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
853
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
854
		ptep_set_wrprotect(src_mm, addr, src_pte);
855
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
856 857 858 859 860 861 862 863 864
	}

	/*
	 * 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);
865 866 867 868

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
H
Hugh Dickins 已提交
869
		page_dup_rmap(page);
K
KAMEZAWA Hiroyuki 已提交
870 871 872 873
		if (PageAnon(page))
			rss[MM_ANONPAGES]++;
		else
			rss[MM_FILEPAGES]++;
874
	}
875 876 877

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
878
	return 0;
L
Linus Torvalds 已提交
879 880
}

881 882 883
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
884
{
885
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
886
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
887
	spinlock_t *src_ptl, *dst_ptl;
888
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
889
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
890
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
891 892

again:
K
KAMEZAWA Hiroyuki 已提交
893 894
	init_rss_vec(rss);

H
Hugh Dickins 已提交
895
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
896 897
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
898
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
899
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
900
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
901 902
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
903
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
904 905 906 907 908 909

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
910 911 912
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
913
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
914 915
				break;
		}
L
Linus Torvalds 已提交
916 917 918 919
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
920 921 922 923
		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
924 925 926
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

927
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
928
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
929
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
930
	add_mm_rss_vec(dst_mm, rss);
931
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
932
	cond_resched();
H
Hugh Dickins 已提交
933 934 935 936 937 938

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956
	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);
957 958
		if (pmd_trans_huge(*src_pmd)) {
			int err;
959
			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
960 961 962 963 964 965 966 967
			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
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
		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;
1006 1007 1008
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1009
	int ret;
L
Linus Torvalds 已提交
1010

1011 1012 1013 1014 1015 1016
	/*
	 * 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.
	 */
1017 1018
	if (!(vma->vm_flags & (VM_HUGETLB | VM_NONLINEAR |
			       VM_PFNMAP | VM_MIXEDMAP))) {
1019 1020 1021 1022
		if (!vma->anon_vma)
			return 0;
	}

L
Linus Torvalds 已提交
1023 1024 1025
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1026
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1027 1028 1029 1030
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1031
		ret = track_pfn_copy(vma);
1032 1033 1034 1035
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1036 1037 1038 1039 1040 1041
	/*
	 * 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.
	 */
1042 1043 1044 1045 1046 1047
	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
1048 1049

	ret = 0;
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054 1055
	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
1056 1057 1058 1059 1060
		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
1061
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1062

1063 1064
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1065
	return ret;
L
Linus Torvalds 已提交
1066 1067
}

1068
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1069
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1070
				unsigned long addr, unsigned long end,
1071
				struct zap_details *details)
L
Linus Torvalds 已提交
1072
{
N
Nick Piggin 已提交
1073
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1074
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1075
	int rss[NR_MM_COUNTERS];
1076
	spinlock_t *ptl;
1077
	pte_t *start_pte;
1078
	pte_t *pte;
K
KAMEZAWA Hiroyuki 已提交
1079

P
Peter Zijlstra 已提交
1080
again:
1081
	init_rss_vec(rss);
1082 1083
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1084
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1085 1086
	do {
		pte_t ptent = *pte;
1087
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1088
			continue;
1089
		}
1090

L
Linus Torvalds 已提交
1091
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1092
			struct page *page;
1093

1094
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
			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
1113
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1114
							tlb->fullmm);
L
Linus Torvalds 已提交
1115 1116 1117 1118 1119
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
			if (unlikely(details) && details->nonlinear_vma
			    && linear_page_index(details->nonlinear_vma,
1120 1121 1122 1123 1124 1125
						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
1126
			if (PageAnon(page))
K
KAMEZAWA Hiroyuki 已提交
1127
				rss[MM_ANONPAGES]--;
1128 1129 1130
			else {
				if (pte_dirty(ptent))
					set_page_dirty(page);
1131
				if (pte_young(ptent) &&
1132
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1133
					mark_page_accessed(page);
K
KAMEZAWA Hiroyuki 已提交
1134
				rss[MM_FILEPAGES]--;
1135
			}
1136
			page_remove_rmap(page);
1137 1138
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
P
Peter Zijlstra 已提交
1139 1140 1141
			force_flush = !__tlb_remove_page(tlb, page);
			if (force_flush)
				break;
L
Linus Torvalds 已提交
1142 1143 1144 1145 1146 1147 1148 1149
			continue;
		}
		/*
		 * If details->check_mapping, we leave swap entries;
		 * if details->nonlinear_vma, we leave file entries.
		 */
		if (unlikely(details))
			continue;
1150 1151 1152
		if (pte_file(ptent)) {
			if (unlikely(!(vma->vm_flags & VM_NONLINEAR)))
				print_bad_pte(vma, addr, ptent, NULL);
K
KAMEZAWA Hiroyuki 已提交
1153 1154 1155 1156 1157
		} else {
			swp_entry_t entry = pte_to_swp_entry(ptent);

			if (!non_swap_entry(entry))
				rss[MM_SWAPENTS]--;
1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
			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
1168 1169 1170
			if (unlikely(!free_swap_and_cache(entry)))
				print_bad_pte(vma, addr, ptent, NULL);
		}
1171
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1172
	} while (pte++, addr += PAGE_SIZE, addr != end);
1173

K
KAMEZAWA Hiroyuki 已提交
1174
	add_mm_rss_vec(mm, rss);
1175
	arch_leave_lazy_mmu_mode();
1176
	pte_unmap_unlock(start_pte, ptl);
1177

P
Peter Zijlstra 已提交
1178 1179 1180 1181 1182 1183
	/*
	 * mmu_gather ran out of room to batch pages, we break out of
	 * the PTE lock to avoid doing the potential expensive TLB invalidate
	 * and page-free while holding it.
	 */
	if (force_flush) {
1184 1185
		unsigned long old_end;

P
Peter Zijlstra 已提交
1186
		force_flush = 0;
1187

1188 1189 1190 1191 1192 1193
		/*
		 * Flush the TLB just for the previous segment,
		 * then update the range to be the remaining
		 * TLB range.
		 */
		old_end = tlb->end;
1194
		tlb->end = addr;
1195

P
Peter Zijlstra 已提交
1196
		tlb_flush_mmu(tlb);
1197 1198 1199 1200 1201

		tlb->start = addr;
		tlb->end = old_end;

		if (addr != end)
P
Peter Zijlstra 已提交
1202 1203 1204
			goto again;
	}

1205
	return addr;
L
Linus Torvalds 已提交
1206 1207
}

1208
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1209
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1210
				unsigned long addr, unsigned long end,
1211
				struct zap_details *details)
L
Linus Torvalds 已提交
1212 1213 1214 1215 1216 1217 1218
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1219
		if (pmd_trans_huge(*pmd)) {
1220
			if (next - addr != HPAGE_PMD_SIZE) {
1221 1222 1223 1224 1225 1226 1227 1228 1229
#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
1230
				split_huge_page_pmd(vma, addr, pmd);
S
Shaohua Li 已提交
1231
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1232
				goto next;
1233 1234
			/* fall through */
		}
1235 1236 1237 1238 1239 1240 1241 1242 1243
		/*
		 * 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;
1244
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1245
next:
1246 1247
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1248 1249

	return addr;
L
Linus Torvalds 已提交
1250 1251
}

1252
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1253
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1254
				unsigned long addr, unsigned long end,
1255
				struct zap_details *details)
L
Linus Torvalds 已提交
1256 1257 1258 1259 1260 1261 1262
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1263
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1264
			continue;
1265 1266
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
	} while (pud++, addr = next, addr != end);
1267 1268

	return addr;
L
Linus Torvalds 已提交
1269 1270
}

A
Al Viro 已提交
1271 1272 1273 1274
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
1275 1276 1277 1278 1279 1280 1281 1282
{
	pgd_t *pgd;
	unsigned long next;

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

	BUG_ON(addr >= end);
1283
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
1284 1285 1286 1287
	tlb_start_vma(tlb, vma);
	pgd = pgd_offset(vma->vm_mm, addr);
	do {
		next = pgd_addr_end(addr, end);
1288
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1289
			continue;
1290 1291
		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1292
	tlb_end_vma(tlb, vma);
1293
	mem_cgroup_uncharge_end();
L
Linus Torvalds 已提交
1294
}
1295

1296 1297 1298

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1299
		unsigned long end_addr,
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310
		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;

1311 1312 1313
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1314
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1315
		untrack_pfn(vma, 0, 0);
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329

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

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1342
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1343 1344 1345 1346
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1347
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
 *
 * 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
1358
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1359
		struct vm_area_struct *vma, unsigned long start_addr,
1360
		unsigned long end_addr)
L
Linus Torvalds 已提交
1361
{
A
Andrea Arcangeli 已提交
1362
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1363

A
Andrea Arcangeli 已提交
1364
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1365
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1366
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1367
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1368 1369 1370 1371 1372
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1373
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1374 1375
 * @size: number of bytes to zap
 * @details: details of nonlinear truncation or shared cache invalidation
1376 1377
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1378
 */
1379
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
L
Linus Torvalds 已提交
1380 1381 1382
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1383
	struct mmu_gather tlb;
1384
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1385 1386

	lru_add_drain();
1387
	tlb_gather_mmu(&tlb, mm, start, end);
1388
	update_hiwater_rss(mm);
1389 1390
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1391
		unmap_single_vma(&tlb, vma, start, end, details);
1392 1393
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1394 1395
}

1396 1397 1398 1399 1400 1401 1402 1403
/**
 * 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
1404
 */
1405
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1406 1407 1408
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1409
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1410 1411 1412
	unsigned long end = address + size;

	lru_add_drain();
1413
	tlb_gather_mmu(&tlb, mm, address, end);
1414
	update_hiwater_rss(mm);
1415
	mmu_notifier_invalidate_range_start(mm, address, end);
1416
	unmap_single_vma(&tlb, vma, address, end, details);
1417
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1418
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1419 1420
}

1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
/**
 * 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;
1439
	zap_page_range_single(vma, address, size, NULL);
1440 1441 1442 1443
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

J
Johannes Weiner 已提交
1444
/**
1445
 * follow_page_mask - look up a page descriptor from a user-virtual address
J
Johannes Weiner 已提交
1446 1447 1448
 * @vma: vm_area_struct mapping @address
 * @address: virtual address to look up
 * @flags: flags modifying lookup behaviour
1449
 * @page_mask: on output, *page_mask is set according to the size of the page
J
Johannes Weiner 已提交
1450 1451 1452 1453 1454 1455
 *
 * @flags can have FOLL_ flags set, defined in <linux/mm.h>
 *
 * Returns the mapped (struct page *), %NULL if no mapping exists, or
 * an error pointer if there is a mapping to something not represented
 * by a page descriptor (see also vm_normal_page()).
L
Linus Torvalds 已提交
1456
 */
1457 1458 1459
struct page *follow_page_mask(struct vm_area_struct *vma,
			      unsigned long address, unsigned int flags,
			      unsigned int *page_mask)
L
Linus Torvalds 已提交
1460 1461 1462 1463 1464
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep, pte;
1465
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1466
	struct page *page;
1467
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1468

1469 1470
	*page_mask = 0;

1471 1472 1473 1474 1475
	page = follow_huge_addr(mm, address, flags & FOLL_WRITE);
	if (!IS_ERR(page)) {
		BUG_ON(flags & FOLL_GET);
		goto out;
	}
L
Linus Torvalds 已提交
1476

1477
	page = NULL;
L
Linus Torvalds 已提交
1478 1479
	pgd = pgd_offset(mm, address);
	if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
1480
		goto no_page_table;
L
Linus Torvalds 已提交
1481 1482

	pud = pud_offset(pgd, address);
A
Andi Kleen 已提交
1483
	if (pud_none(*pud))
1484
		goto no_page_table;
1485
	if (pud_huge(*pud) && vma->vm_flags & VM_HUGETLB) {
1486 1487
		if (flags & FOLL_GET)
			goto out;
A
Andi Kleen 已提交
1488 1489 1490 1491 1492 1493
		page = follow_huge_pud(mm, address, pud, flags & FOLL_WRITE);
		goto out;
	}
	if (unlikely(pud_bad(*pud)))
		goto no_page_table;

L
Linus Torvalds 已提交
1494
	pmd = pmd_offset(pud, address);
1495
	if (pmd_none(*pmd))
1496
		goto no_page_table;
1497
	if (pmd_huge(*pmd) && vma->vm_flags & VM_HUGETLB) {
1498
		page = follow_huge_pmd(mm, address, pmd, flags & FOLL_WRITE);
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
		if (flags & FOLL_GET) {
			/*
			 * Refcount on tail pages are not well-defined and
			 * shouldn't be taken. The caller should handle a NULL
			 * return when trying to follow tail pages.
			 */
			if (PageHead(page))
				get_page(page);
			else {
				page = NULL;
				goto out;
			}
		}
L
Linus Torvalds 已提交
1512
		goto out;
1513
	}
1514 1515
	if ((flags & FOLL_NUMA) && pmd_numa(*pmd))
		goto no_page_table;
1516
	if (pmd_trans_huge(*pmd)) {
1517
		if (flags & FOLL_SPLIT) {
1518
			split_huge_page_pmd(vma, address, pmd);
1519 1520
			goto split_fallthrough;
		}
1521 1522 1523 1524 1525 1526
		spin_lock(&mm->page_table_lock);
		if (likely(pmd_trans_huge(*pmd))) {
			if (unlikely(pmd_trans_splitting(*pmd))) {
				spin_unlock(&mm->page_table_lock);
				wait_split_huge_page(vma->anon_vma, pmd);
			} else {
1527
				page = follow_trans_huge_pmd(vma, address,
1528 1529
							     pmd, flags);
				spin_unlock(&mm->page_table_lock);
1530
				*page_mask = HPAGE_PMD_NR - 1;
1531 1532 1533 1534 1535 1536
				goto out;
			}
		} else
			spin_unlock(&mm->page_table_lock);
		/* fall through */
	}
1537
split_fallthrough:
1538 1539 1540
	if (unlikely(pmd_bad(*pmd)))
		goto no_page_table;

1541
	ptep = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1542 1543

	pte = *ptep;
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
	if (!pte_present(pte)) {
		swp_entry_t entry;
		/*
		 * KSM's break_ksm() relies upon recognizing a ksm page
		 * even while it is being migrated, so for that case we
		 * need migration_entry_wait().
		 */
		if (likely(!(flags & FOLL_MIGRATION)))
			goto no_page;
		if (pte_none(pte) || pte_file(pte))
			goto no_page;
		entry = pte_to_swp_entry(pte);
		if (!is_migration_entry(entry))
			goto no_page;
		pte_unmap_unlock(ptep, ptl);
		migration_entry_wait(mm, pmd, address);
		goto split_fallthrough;
	}
1562 1563
	if ((flags & FOLL_NUMA) && pte_numa(pte))
		goto no_page;
1564 1565
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;
H
Hugh Dickins 已提交
1566

1567
	page = vm_normal_page(vma, address, pte);
H
Hugh Dickins 已提交
1568 1569
	if (unlikely(!page)) {
		if ((flags & FOLL_DUMP) ||
H
Hugh Dickins 已提交
1570
		    !is_zero_pfn(pte_pfn(pte)))
H
Hugh Dickins 已提交
1571 1572 1573
			goto bad_page;
		page = pte_page(pte);
	}
L
Linus Torvalds 已提交
1574

1575
	if (flags & FOLL_GET)
1576
		get_page_foll(page);
1577 1578 1579 1580
	if (flags & FOLL_TOUCH) {
		if ((flags & FOLL_WRITE) &&
		    !pte_dirty(pte) && !PageDirty(page))
			set_page_dirty(page);
1581 1582 1583 1584 1585
		/*
		 * pte_mkyoung() would be more correct here, but atomic care
		 * is needed to avoid losing the dirty bit: it is easier to use
		 * mark_page_accessed().
		 */
1586 1587
		mark_page_accessed(page);
	}
1588
	if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) {
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
		/*
		 * The preliminary mapping check is mainly to avoid the
		 * pointless overhead of lock_page on the ZERO_PAGE
		 * which might bounce very badly if there is contention.
		 *
		 * If the page is already locked, we don't need to
		 * handle it now - vmscan will handle it later if and
		 * when it attempts to reclaim the page.
		 */
		if (page->mapping && trylock_page(page)) {
			lru_add_drain();  /* push cached pages to LRU */
			/*
1601 1602 1603 1604
			 * Because we lock page here, and migration is
			 * blocked by the pte's page reference, and we
			 * know the page is still mapped, we don't even
			 * need to check for file-cache page truncation.
1605
			 */
1606
			mlock_vma_page(page);
1607 1608 1609
			unlock_page(page);
		}
	}
1610 1611
unlock:
	pte_unmap_unlock(ptep, ptl);
L
Linus Torvalds 已提交
1612
out:
1613
	return page;
L
Linus Torvalds 已提交
1614

1615 1616 1617 1618 1619 1620 1621 1622
bad_page:
	pte_unmap_unlock(ptep, ptl);
	return ERR_PTR(-EFAULT);

no_page:
	pte_unmap_unlock(ptep, ptl);
	if (!pte_none(pte))
		return page;
H
Hugh Dickins 已提交
1623

1624 1625 1626
no_page_table:
	/*
	 * When core dumping an enormous anonymous area that nobody
H
Hugh Dickins 已提交
1627 1628 1629 1630 1631
	 * has touched so far, we don't want to allocate unnecessary pages or
	 * page tables.  Return error instead of NULL to skip handle_mm_fault,
	 * then get_dump_page() will return NULL to leave a hole in the dump.
	 * But we can only make this optimization where a hole would surely
	 * be zero-filled if handle_mm_fault() actually did handle it.
1632
	 */
H
Hugh Dickins 已提交
1633 1634 1635
	if ((flags & FOLL_DUMP) &&
	    (!vma->vm_ops || !vma->vm_ops->fault))
		return ERR_PTR(-EFAULT);
1636
	return page;
L
Linus Torvalds 已提交
1637 1638
}

1639 1640
static inline int stack_guard_page(struct vm_area_struct *vma, unsigned long addr)
{
1641 1642
	return stack_guard_page_start(vma, addr) ||
	       stack_guard_page_end(vma, addr+PAGE_SIZE);
1643 1644
}

H
Huang Ying 已提交
1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
/**
 * __get_user_pages() - pin user pages in memory
 * @tsk:	task_struct of target task
 * @mm:		mm_struct of target mm
 * @start:	starting user address
 * @nr_pages:	number of pages from start to pin
 * @gup_flags:	flags modifying pin behaviour
 * @pages:	array that receives pointers to the pages pinned.
 *		Should be at least nr_pages long. Or NULL, if caller
 *		only intends to ensure the pages are faulted in.
 * @vmas:	array of pointers to vmas corresponding to each page.
 *		Or NULL if the caller does not require them.
 * @nonblocking: whether waiting for disk IO or mmap_sem contention
 *
 * Returns number of pages pinned. This may be fewer than the number
 * requested. If nr_pages is 0 or negative, returns 0. If no pages
 * were pinned, returns -errno. Each page returned must be released
 * with a put_page() call when it is finished with. vmas will only
 * remain valid while mmap_sem is held.
 *
 * Must be called with mmap_sem held for read or write.
 *
 * __get_user_pages walks a process's page tables and takes a reference to
 * each struct page that each user address corresponds to at a given
 * instant. That is, it takes the page that would be accessed if a user
 * thread accesses the given user virtual address at that instant.
 *
 * This does not guarantee that the page exists in the user mappings when
 * __get_user_pages returns, and there may even be a completely different
 * page there in some cases (eg. if mmapped pagecache has been invalidated
 * and subsequently re faulted). However it does guarantee that the page
 * won't be freed completely. And mostly callers simply care that the page
 * contains data that was valid *at some point in time*. Typically, an IO
 * or similar operation cannot guarantee anything stronger anyway because
 * locks can't be held over the syscall boundary.
 *
 * If @gup_flags & FOLL_WRITE == 0, the page must not be written to. If
 * the page is written to, set_page_dirty (or set_page_dirty_lock, as
 * appropriate) must be called after the page is finished with, and
 * before put_page is called.
 *
 * If @nonblocking != NULL, __get_user_pages will not wait for disk IO
 * or mmap_sem contention, and if waiting is needed to pin all pages,
 * *@nonblocking will be set to 0.
 *
 * In most cases, get_user_pages or get_user_pages_fast should be used
 * instead of __get_user_pages. __get_user_pages should be used only if
 * you need some special @gup_flags.
 */
1694 1695 1696 1697
long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
		unsigned long start, unsigned long nr_pages,
		unsigned int gup_flags, struct page **pages,
		struct vm_area_struct **vmas, int *nonblocking)
L
Linus Torvalds 已提交
1698
{
1699
	long i;
H
Hugh Dickins 已提交
1700
	unsigned long vm_flags;
1701
	unsigned int page_mask;
L
Linus Torvalds 已提交
1702

1703
	if (!nr_pages)
1704
		return 0;
H
Hugh Dickins 已提交
1705 1706 1707

	VM_BUG_ON(!!pages != !!(gup_flags & FOLL_GET));

L
Linus Torvalds 已提交
1708 1709
	/* 
	 * Require read or write permissions.
H
Hugh Dickins 已提交
1710
	 * If FOLL_FORCE is set, we only require the "MAY" flags.
L
Linus Torvalds 已提交
1711
	 */
H
Hugh Dickins 已提交
1712 1713 1714 1715
	vm_flags  = (gup_flags & FOLL_WRITE) ?
			(VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
	vm_flags &= (gup_flags & FOLL_FORCE) ?
			(VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728

	/*
	 * If FOLL_FORCE and FOLL_NUMA are both set, handle_mm_fault
	 * would be called on PROT_NONE ranges. We must never invoke
	 * handle_mm_fault on PROT_NONE ranges or the NUMA hinting
	 * page faults would unprotect the PROT_NONE ranges if
	 * _PAGE_NUMA and _PAGE_PROTNONE are sharing the same pte/pmd
	 * bitflag. So to avoid that, don't set FOLL_NUMA if
	 * FOLL_FORCE is set.
	 */
	if (!(gup_flags & FOLL_FORCE))
		gup_flags |= FOLL_NUMA;

L
Linus Torvalds 已提交
1729 1730 1731
	i = 0;

	do {
1732
		struct vm_area_struct *vma;
L
Linus Torvalds 已提交
1733 1734

		vma = find_extend_vma(mm, start);
1735
		if (!vma && in_gate_area(mm, start)) {
L
Linus Torvalds 已提交
1736 1737 1738 1739 1740
			unsigned long pg = start & PAGE_MASK;
			pgd_t *pgd;
			pud_t *pud;
			pmd_t *pmd;
			pte_t *pte;
N
Nick Piggin 已提交
1741 1742

			/* user gate pages are read-only */
H
Hugh Dickins 已提交
1743
			if (gup_flags & FOLL_WRITE)
L
Linus Torvalds 已提交
1744 1745 1746 1747 1748 1749 1750 1751 1752
				return i ? : -EFAULT;
			if (pg > TASK_SIZE)
				pgd = pgd_offset_k(pg);
			else
				pgd = pgd_offset_gate(mm, pg);
			BUG_ON(pgd_none(*pgd));
			pud = pud_offset(pgd, pg);
			BUG_ON(pud_none(*pud));
			pmd = pmd_offset(pud, pg);
1753 1754
			if (pmd_none(*pmd))
				return i ? : -EFAULT;
1755
			VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1756
			pte = pte_offset_map(pmd, pg);
1757 1758 1759 1760
			if (pte_none(*pte)) {
				pte_unmap(pte);
				return i ? : -EFAULT;
			}
1761
			vma = get_gate_vma(mm);
L
Linus Torvalds 已提交
1762
			if (pages) {
1763 1764
				struct page *page;

1765
				page = vm_normal_page(vma, start, *pte);
1766 1767 1768 1769 1770 1771 1772 1773 1774
				if (!page) {
					if (!(gup_flags & FOLL_DUMP) &&
					     is_zero_pfn(pte_pfn(*pte)))
						page = pte_page(*pte);
					else {
						pte_unmap(pte);
						return i ? : -EFAULT;
					}
				}
1775
				pages[i] = page;
1776
				get_page(page);
L
Linus Torvalds 已提交
1777 1778
			}
			pte_unmap(pte);
1779
			page_mask = 0;
1780
			goto next_page;
L
Linus Torvalds 已提交
1781 1782
		}

N
Nick Piggin 已提交
1783 1784
		if (!vma ||
		    (vma->vm_flags & (VM_IO | VM_PFNMAP)) ||
H
Hugh Dickins 已提交
1785
		    !(vm_flags & vma->vm_flags))
L
Linus Torvalds 已提交
1786 1787
			return i ? : -EFAULT;

H
Hugh Dickins 已提交
1788 1789
		if (is_vm_hugetlb_page(vma)) {
			i = follow_hugetlb_page(mm, vma, pages, vmas,
H
Hugh Dickins 已提交
1790
					&start, &nr_pages, i, gup_flags);
H
Hugh Dickins 已提交
1791 1792
			continue;
		}
1793

L
Linus Torvalds 已提交
1794
		do {
1795
			struct page *page;
H
Hugh Dickins 已提交
1796
			unsigned int foll_flags = gup_flags;
1797
			unsigned int page_increm;
L
Linus Torvalds 已提交
1798

1799
			/*
1800
			 * If we have a pending SIGKILL, don't keep faulting
H
Hugh Dickins 已提交
1801
			 * pages and potentially allocating memory.
1802
			 */
H
Hugh Dickins 已提交
1803
			if (unlikely(fatal_signal_pending(current)))
1804
				return i ? i : -ERESTARTSYS;
1805

1806
			cond_resched();
1807 1808
			while (!(page = follow_page_mask(vma, start,
						foll_flags, &page_mask))) {
1809
				int ret;
1810 1811
				unsigned int fault_flags = 0;

1812 1813 1814 1815 1816
				/* For mlock, just skip the stack guard page. */
				if (foll_flags & FOLL_MLOCK) {
					if (stack_guard_page(vma, start))
						goto next_page;
				}
1817 1818 1819 1820
				if (foll_flags & FOLL_WRITE)
					fault_flags |= FAULT_FLAG_WRITE;
				if (nonblocking)
					fault_flags |= FAULT_FLAG_ALLOW_RETRY;
1821 1822
				if (foll_flags & FOLL_NOWAIT)
					fault_flags |= (FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_RETRY_NOWAIT);
1823

1824
				ret = handle_mm_fault(mm, vma, start,
1825
							fault_flags);
1826

N
Nick Piggin 已提交
1827 1828 1829
				if (ret & VM_FAULT_ERROR) {
					if (ret & VM_FAULT_OOM)
						return i ? i : -ENOMEM;
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
					if (ret & (VM_FAULT_HWPOISON |
						   VM_FAULT_HWPOISON_LARGE)) {
						if (i)
							return i;
						else if (gup_flags & FOLL_HWPOISON)
							return -EHWPOISON;
						else
							return -EFAULT;
					}
					if (ret & VM_FAULT_SIGBUS)
N
Nick Piggin 已提交
1840 1841 1842
						return i ? i : -EFAULT;
					BUG();
				}
1843 1844 1845 1846 1847 1848 1849

				if (tsk) {
					if (ret & VM_FAULT_MAJOR)
						tsk->maj_flt++;
					else
						tsk->min_flt++;
				}
N
Nick Piggin 已提交
1850

1851
				if (ret & VM_FAULT_RETRY) {
1852 1853
					if (nonblocking)
						*nonblocking = 0;
1854 1855 1856
					return i;
				}

1857
				/*
N
Nick Piggin 已提交
1858 1859 1860 1861
				 * The VM_FAULT_WRITE bit tells us that
				 * do_wp_page has broken COW when necessary,
				 * even if maybe_mkwrite decided not to set
				 * pte_write. We can thus safely do subsequent
1862 1863 1864 1865 1866 1867
				 * page lookups as if they were reads. But only
				 * do so when looping for pte_write is futile:
				 * in some cases userspace may also be wanting
				 * to write to the gotten user page, which a
				 * read fault here might prevent (a readonly
				 * page might get reCOWed by userspace write).
1868
				 */
1869 1870
				if ((ret & VM_FAULT_WRITE) &&
				    !(vma->vm_flags & VM_WRITE))
1871
					foll_flags &= ~FOLL_WRITE;
N
Nick Piggin 已提交
1872

1873
				cond_resched();
L
Linus Torvalds 已提交
1874
			}
1875 1876
			if (IS_ERR(page))
				return i ? i : PTR_ERR(page);
L
Linus Torvalds 已提交
1877
			if (pages) {
1878
				pages[i] = page;
1879

1880
				flush_anon_page(vma, page, start);
1881
				flush_dcache_page(page);
1882
				page_mask = 0;
L
Linus Torvalds 已提交
1883
			}
1884
next_page:
1885
			if (vmas) {
L
Linus Torvalds 已提交
1886
				vmas[i] = vma;
1887 1888 1889 1890 1891 1892 1893 1894
				page_mask = 0;
			}
			page_increm = 1 + (~(start >> PAGE_SHIFT) & page_mask);
			if (page_increm > nr_pages)
				page_increm = nr_pages;
			i += page_increm;
			start += page_increm * PAGE_SIZE;
			nr_pages -= page_increm;
1895 1896
		} while (nr_pages && start < vma->vm_end);
	} while (nr_pages);
L
Linus Torvalds 已提交
1897 1898
	return i;
}
H
Huang Ying 已提交
1899
EXPORT_SYMBOL(__get_user_pages);
N
Nick Piggin 已提交
1900

1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957
/*
 * fixup_user_fault() - manually resolve a user page fault
 * @tsk:	the task_struct to use for page fault accounting, or
 *		NULL if faults are not to be recorded.
 * @mm:		mm_struct of target mm
 * @address:	user address
 * @fault_flags:flags to pass down to handle_mm_fault()
 *
 * This is meant to be called in the specific scenario where for locking reasons
 * we try to access user memory in atomic context (within a pagefault_disable()
 * section), this returns -EFAULT, and we want to resolve the user fault before
 * trying again.
 *
 * Typically this is meant to be used by the futex code.
 *
 * The main difference with get_user_pages() is that this function will
 * unconditionally call handle_mm_fault() which will in turn perform all the
 * necessary SW fixup of the dirty and young bits in the PTE, while
 * handle_mm_fault() only guarantees to update these in the struct page.
 *
 * This is important for some architectures where those bits also gate the
 * access permission to the page because they are maintained in software.  On
 * such architectures, gup() will not be enough to make a subsequent access
 * succeed.
 *
 * This should be called with the mm_sem held for read.
 */
int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
		     unsigned long address, unsigned int fault_flags)
{
	struct vm_area_struct *vma;
	int ret;

	vma = find_extend_vma(mm, address);
	if (!vma || address < vma->vm_start)
		return -EFAULT;

	ret = handle_mm_fault(mm, vma, address, fault_flags);
	if (ret & VM_FAULT_ERROR) {
		if (ret & VM_FAULT_OOM)
			return -ENOMEM;
		if (ret & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
			return -EHWPOISON;
		if (ret & VM_FAULT_SIGBUS)
			return -EFAULT;
		BUG();
	}
	if (tsk) {
		if (ret & VM_FAULT_MAJOR)
			tsk->maj_flt++;
		else
			tsk->min_flt++;
	}
	return 0;
}

/*
1958
 * get_user_pages() - pin user pages in memory
1959 1960
 * @tsk:	the task_struct to use for page fault accounting, or
 *		NULL if faults are not to be recorded.
1961 1962
 * @mm:		mm_struct of target mm
 * @start:	starting user address
1963
 * @nr_pages:	number of pages from start to pin
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974
 * @write:	whether pages will be written to by the caller
 * @force:	whether to force write access even if user mapping is
 *		readonly. This will result in the page being COWed even
 *		in MAP_SHARED mappings. You do not want this.
 * @pages:	array that receives pointers to the pages pinned.
 *		Should be at least nr_pages long. Or NULL, if caller
 *		only intends to ensure the pages are faulted in.
 * @vmas:	array of pointers to vmas corresponding to each page.
 *		Or NULL if the caller does not require them.
 *
 * Returns number of pages pinned. This may be fewer than the number
1975
 * requested. If nr_pages is 0 or negative, returns 0. If no pages
1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
 * were pinned, returns -errno. Each page returned must be released
 * with a put_page() call when it is finished with. vmas will only
 * remain valid while mmap_sem is held.
 *
 * Must be called with mmap_sem held for read or write.
 *
 * get_user_pages walks a process's page tables and takes a reference to
 * each struct page that each user address corresponds to at a given
 * instant. That is, it takes the page that would be accessed if a user
 * thread accesses the given user virtual address at that instant.
 *
 * This does not guarantee that the page exists in the user mappings when
 * get_user_pages returns, and there may even be a completely different
 * page there in some cases (eg. if mmapped pagecache has been invalidated
 * and subsequently re faulted). However it does guarantee that the page
 * won't be freed completely. And mostly callers simply care that the page
 * contains data that was valid *at some point in time*. Typically, an IO
 * or similar operation cannot guarantee anything stronger anyway because
 * locks can't be held over the syscall boundary.
 *
 * If write=0, the page must not be written to. If the page is written to,
 * set_page_dirty (or set_page_dirty_lock, as appropriate) must be called
 * after the page is finished with, and before put_page is called.
 *
 * get_user_pages is typically used for fewer-copy IO operations, to get a
 * handle on the memory by some means other than accesses via the user virtual
 * addresses. The pages may be submitted for DMA to devices or accessed via
 * their kernel linear mapping (via the kmap APIs). Care should be taken to
 * use the correct cache flushing APIs.
 *
 * See also get_user_pages_fast, for performance critical applications.
 */
2008 2009 2010
long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
		unsigned long start, unsigned long nr_pages, int write,
		int force, struct page **pages, struct vm_area_struct **vmas)
N
Nick Piggin 已提交
2011
{
H
Hugh Dickins 已提交
2012
	int flags = FOLL_TOUCH;
N
Nick Piggin 已提交
2013

H
Hugh Dickins 已提交
2014 2015
	if (pages)
		flags |= FOLL_GET;
N
Nick Piggin 已提交
2016
	if (write)
H
Hugh Dickins 已提交
2017
		flags |= FOLL_WRITE;
N
Nick Piggin 已提交
2018
	if (force)
H
Hugh Dickins 已提交
2019
		flags |= FOLL_FORCE;
N
Nick Piggin 已提交
2020

2021 2022
	return __get_user_pages(tsk, mm, start, nr_pages, flags, pages, vmas,
				NULL);
N
Nick Piggin 已提交
2023
}
L
Linus Torvalds 已提交
2024 2025
EXPORT_SYMBOL(get_user_pages);

H
Hugh Dickins 已提交
2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046
/**
 * get_dump_page() - pin user page in memory while writing it to core dump
 * @addr: user address
 *
 * Returns struct page pointer of user page pinned for dump,
 * to be freed afterwards by page_cache_release() or put_page().
 *
 * Returns NULL on any kind of failure - a hole must then be inserted into
 * the corefile, to preserve alignment with its headers; and also returns
 * NULL wherever the ZERO_PAGE, or an anonymous pte_none, has been found -
 * allowing a hole to be left in the corefile to save diskspace.
 *
 * Called without mmap_sem, but after all other threads have been killed.
 */
#ifdef CONFIG_ELF_CORE
struct page *get_dump_page(unsigned long addr)
{
	struct vm_area_struct *vma;
	struct page *page;

	if (__get_user_pages(current, current->mm, addr, 1,
2047 2048
			     FOLL_FORCE | FOLL_DUMP | FOLL_GET, &page, &vma,
			     NULL) < 1)
H
Hugh Dickins 已提交
2049 2050 2051 2052 2053 2054
		return NULL;
	flush_cache_page(vma, addr, page_to_pfn(page));
	return page;
}
#endif /* CONFIG_ELF_CORE */

2055
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
2056
			spinlock_t **ptl)
2057 2058 2059 2060
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
2061
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
2062 2063
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
2064
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
2065
		}
2066 2067 2068 2069
	}
	return NULL;
}

2070 2071 2072 2073 2074 2075 2076
/*
 * 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
2077 2078
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
2079
{
N
Nick Piggin 已提交
2080
	struct mm_struct *mm = vma->vm_mm;
2081
	int retval;
2082
	pte_t *pte;
2083 2084
	spinlock_t *ptl;

2085
	retval = -EINVAL;
2086
	if (PageAnon(page))
2087
		goto out;
2088 2089
	retval = -ENOMEM;
	flush_dcache_page(page);
2090
	pte = get_locked_pte(mm, addr, &ptl);
2091
	if (!pte)
2092
		goto out;
2093 2094 2095 2096 2097 2098
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
2099
	inc_mm_counter_fast(mm, MM_FILEPAGES);
2100 2101 2102 2103
	page_add_file_rmap(page);
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
2104 2105
	pte_unmap_unlock(pte, ptl);
	return retval;
2106 2107 2108 2109 2110 2111
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

2112 2113 2114 2115 2116 2117
/**
 * 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
 *
2118 2119 2120 2121 2122 2123
 * 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
2124
 * (see split_page()).
2125 2126 2127 2128 2129 2130 2131 2132
 *
 * 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.
2133 2134 2135 2136 2137
 *
 * 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.
2138
 */
N
Nick Piggin 已提交
2139 2140
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
2141 2142 2143 2144 2145
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
2146 2147 2148 2149 2150
	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
2151
	return insert_page(vma, addr, page, vma->vm_page_prot);
2152
}
2153
EXPORT_SYMBOL(vm_insert_page);
2154

N
Nick Piggin 已提交
2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
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);
2174
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
2175 2176 2177 2178 2179 2180 2181 2182

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

N
Nick Piggin 已提交
2183 2184 2185 2186 2187 2188
/**
 * 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
 *
2189
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
2190 2191 2192 2193
 * 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
2194 2195 2196 2197 2198
 *
 * 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
2199 2200
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
2201
			unsigned long pfn)
N
Nick Piggin 已提交
2202
{
2203
	int ret;
2204
	pgprot_t pgprot = vma->vm_page_prot;
N
Nick Piggin 已提交
2205 2206 2207 2208 2209 2210
	/*
	 * 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
2211 2212 2213 2214 2215
	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
2216

N
Nick Piggin 已提交
2217 2218
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
2219
	if (track_pfn_insert(vma, &pgprot, pfn))
2220 2221
		return -EINVAL;

2222
	ret = insert_pfn(vma, addr, pfn, pgprot);
2223 2224

	return ret;
N
Nick Piggin 已提交
2225 2226
}
EXPORT_SYMBOL(vm_insert_pfn);
N
Nick Piggin 已提交
2227

N
Nick Piggin 已提交
2228 2229 2230 2231
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
2232

N
Nick Piggin 已提交
2233 2234
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
N
Nick Piggin 已提交
2235

N
Nick Piggin 已提交
2236 2237 2238 2239
	/*
	 * 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
2240 2241
	 * 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
2242 2243 2244 2245 2246 2247 2248 2249
	 */
	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
2250
}
N
Nick Piggin 已提交
2251
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
2252

L
Linus Torvalds 已提交
2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
/*
 * 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
2263
	spinlock_t *ptl;
L
Linus Torvalds 已提交
2264

H
Hugh Dickins 已提交
2265
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
2266 2267
	if (!pte)
		return -ENOMEM;
2268
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
2269 2270
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
2271
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
2272 2273
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
2274
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
2275
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289
	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;
2290
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
	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;
}

2320 2321 2322 2323 2324 2325 2326 2327 2328 2329
/**
 * 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
2330 2331 2332 2333 2334
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;
2335
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
2336 2337 2338 2339 2340 2341 2342 2343
	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).
2344 2345 2346
	 *   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.
2347 2348 2349 2350
	 *   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
2351 2352 2353 2354
	 *
	 * 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".
2355
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
2356
	 */
2357 2358 2359
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
2360
		vma->vm_pgoff = pfn;
2361 2362 2363 2364
	}

	err = track_pfn_remap(vma, &prot, pfn, addr, PAGE_ALIGN(size));
	if (err)
2365
		return -EINVAL;
L
Linus Torvalds 已提交
2366

2367
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379

	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);
2380 2381

	if (err)
2382
		untrack_pfn(vma, pfn, PAGE_ALIGN(size));
2383

L
Linus Torvalds 已提交
2384 2385 2386 2387
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434
/**
 * 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);

2435 2436 2437 2438 2439 2440
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;
2441
	pgtable_t token;
2442
	spinlock_t *uninitialized_var(ptl);
2443 2444 2445 2446 2447 2448 2449 2450 2451

	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));

2452 2453
	arch_enter_lazy_mmu_mode();

2454
	token = pmd_pgtable(*pmd);
2455 2456

	do {
2457
		err = fn(pte++, token, addr, data);
2458 2459
		if (err)
			break;
2460
	} while (addr += PAGE_SIZE, addr != end);
2461

2462 2463
	arch_leave_lazy_mmu_mode();

2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
	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
2477 2478
	BUG_ON(pud_huge(*pud));

2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
	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;
2520
	unsigned long end = addr + size;
2521 2522 2523 2524 2525 2526 2527 2528 2529 2530
	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);
2531

2532 2533 2534 2535
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2536 2537 2538 2539
/*
 * 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
2540
 * might give a mix of unmatched parts, do_swap_page and do_nonlinear_fault
2541 2542
 * must check under lock before unmapping the pte and proceeding
 * (but do_wp_page is only called after already making such a check;
2543
 * and do_anonymous_page can safely check later on).
2544
 */
H
Hugh Dickins 已提交
2545
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2546 2547 2548 2549 2550
				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
2551 2552
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2553
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2554
		spin_unlock(ptl);
2555 2556 2557 2558 2559 2560
	}
#endif
	pte_unmap(page_table);
	return same;
}

2561
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2562 2563 2564 2565 2566 2567 2568 2569
{
	/*
	 * 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)) {
2570
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2571 2572 2573 2574 2575 2576 2577 2578 2579
		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))
2580
			clear_page(kaddr);
2581
		kunmap_atomic(kaddr);
2582
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2583 2584
	} else
		copy_user_highpage(dst, src, va, vma);
2585 2586
}

L
Linus Torvalds 已提交
2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600
/*
 * 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.
 *
2601 2602 2603
 * 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
2604
 */
2605 2606
static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
2607
		spinlock_t *ptl, pte_t orig_pte)
2608
	__releases(ptl)
L
Linus Torvalds 已提交
2609
{
2610
	struct page *old_page, *new_page = NULL;
L
Linus Torvalds 已提交
2611
	pte_t entry;
2612
	int ret = 0;
2613
	int page_mkwrite = 0;
2614
	struct page *dirty_page = NULL;
2615 2616
	unsigned long mmun_start = 0;	/* For mmu_notifiers */
	unsigned long mmun_end = 0;	/* For mmu_notifiers */
L
Linus Torvalds 已提交
2617

2618
	old_page = vm_normal_page(vma, address, orig_pte);
2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629
	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;
2630
		goto gotten;
2631
	}
L
Linus Torvalds 已提交
2632

2633
	/*
P
Peter Zijlstra 已提交
2634 2635
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2636
	 */
H
Hugh Dickins 已提交
2637
	if (PageAnon(old_page) && !PageKsm(old_page)) {
2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
		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
2649
		}
2650
		if (reuse_swap_page(old_page)) {
2651 2652 2653 2654 2655 2656
			/*
			 * 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);
2657 2658 2659
			unlock_page(old_page);
			goto reuse;
		}
2660
		unlock_page(old_page);
P
Peter Zijlstra 已提交
2661
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2662
					(VM_WRITE|VM_SHARED))) {
P
Peter Zijlstra 已提交
2663 2664 2665 2666 2667
		/*
		 * Only catch write-faults on shared writable pages,
		 * read-only shared pages can get COWed by
		 * get_user_pages(.write=1, .force=1).
		 */
2668
		if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
2669 2670 2671 2672 2673 2674 2675 2676 2677
			struct vm_fault vmf;
			int tmp;

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

2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688
			/*
			 * 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.
			 */
			page_cache_get(old_page);
			pte_unmap_unlock(page_table, ptl);

2689 2690 2691 2692
			tmp = vma->vm_ops->page_mkwrite(vma, &vmf);
			if (unlikely(tmp &
					(VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
				ret = tmp;
2693
				goto unwritable_page;
2694
			}
N
Nick Piggin 已提交
2695 2696 2697 2698 2699 2700 2701 2702 2703
			if (unlikely(!(tmp & VM_FAULT_LOCKED))) {
				lock_page(old_page);
				if (!old_page->mapping) {
					ret = 0; /* retry the fault */
					unlock_page(old_page);
					goto unwritable_page;
				}
			} else
				VM_BUG_ON(!PageLocked(old_page));
2704 2705 2706 2707 2708 2709 2710 2711 2712

			/*
			 * 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
2713 2714
			if (!pte_same(*page_table, orig_pte)) {
				unlock_page(old_page);
2715
				goto unlock;
N
Nick Piggin 已提交
2716
			}
2717 2718

			page_mkwrite = 1;
L
Linus Torvalds 已提交
2719
		}
2720 2721
		dirty_page = old_page;
		get_page(dirty_page);
2722

2723
reuse:
2724 2725 2726
		flush_cache_page(vma, address, pte_pfn(orig_pte));
		entry = pte_mkyoung(orig_pte);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
2727
		if (ptep_set_access_flags(vma, address, page_table, entry,1))
2728
			update_mmu_cache(vma, address, page_table);
2729
		pte_unmap_unlock(page_table, ptl);
2730
		ret |= VM_FAULT_WRITE;
2731 2732 2733 2734 2735 2736 2737 2738 2739 2740

		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.
		 *
2741
		 * __do_fault is protected similarly.
2742 2743 2744 2745
		 */
		if (!page_mkwrite) {
			wait_on_page_locked(dirty_page);
			set_page_dirty_balance(dirty_page, page_mkwrite);
2746 2747 2748
			/* file_update_time outside page_lock */
			if (vma->vm_file)
				file_update_time(vma->vm_file);
2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
		}
		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
2767 2768 2769 2770 2771
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
N
Nick Piggin 已提交
2772
	page_cache_get(old_page);
H
Hugh Dickins 已提交
2773
gotten:
2774
	pte_unmap_unlock(page_table, ptl);
L
Linus Torvalds 已提交
2775 2776

	if (unlikely(anon_vma_prepare(vma)))
2777
		goto oom;
H
Hugh Dickins 已提交
2778

H
Hugh Dickins 已提交
2779
	if (is_zero_pfn(pte_pfn(orig_pte))) {
H
Hugh Dickins 已提交
2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790
		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);

K
KAMEZAWA Hiroyuki 已提交
2791
	if (mem_cgroup_newpage_charge(new_page, mm, GFP_KERNEL))
2792 2793
		goto oom_free_new;

2794
	mmun_start  = address & PAGE_MASK;
2795
	mmun_end    = mmun_start + PAGE_SIZE;
2796 2797
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

L
Linus Torvalds 已提交
2798 2799 2800
	/*
	 * Re-check the pte - we dropped the lock
	 */
2801
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
2802
	if (likely(pte_same(*page_table, orig_pte))) {
H
Hugh Dickins 已提交
2803 2804
		if (old_page) {
			if (!PageAnon(old_page)) {
2805 2806
				dec_mm_counter_fast(mm, MM_FILEPAGES);
				inc_mm_counter_fast(mm, MM_ANONPAGES);
H
Hugh Dickins 已提交
2807 2808
			}
		} else
2809
			inc_mm_counter_fast(mm, MM_ANONPAGES);
2810
		flush_cache_page(vma, address, pte_pfn(orig_pte));
2811 2812
		entry = mk_pte(new_page, vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
2813 2814 2815 2816 2817 2818
		/*
		 * 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.
		 */
2819
		ptep_clear_flush(vma, address, page_table);
N
Nick Piggin 已提交
2820
		page_add_new_anon_rmap(new_page, vma, address);
2821 2822 2823 2824 2825 2826
		/*
		 * 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);
2827
		update_mmu_cache(vma, address, page_table);
N
Nick Piggin 已提交
2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850
		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.
			 */
2851
			page_remove_rmap(old_page);
N
Nick Piggin 已提交
2852 2853
		}

L
Linus Torvalds 已提交
2854 2855
		/* Free the old page.. */
		new_page = old_page;
N
Nick Piggin 已提交
2856
		ret |= VM_FAULT_WRITE;
2857 2858 2859
	} else
		mem_cgroup_uncharge_page(new_page);

2860 2861
	if (new_page)
		page_cache_release(new_page);
2862
unlock:
2863
	pte_unmap_unlock(page_table, ptl);
2864
	if (mmun_end > mmun_start)
2865
		mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877
	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
2878
	return ret;
2879
oom_free_new:
2880
	page_cache_release(new_page);
2881
oom:
2882
	if (old_page)
H
Hugh Dickins 已提交
2883
		page_cache_release(old_page);
L
Linus Torvalds 已提交
2884
	return VM_FAULT_OOM;
2885 2886 2887

unwritable_page:
	page_cache_release(old_page);
2888
	return ret;
L
Linus Torvalds 已提交
2889 2890
}

2891
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2892 2893 2894
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2895
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2896 2897
}

2898
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2899 2900 2901 2902 2903
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2904
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2905 2906 2907
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2908
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2909 2910 2911 2912 2913 2914 2915 2916
		/* 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;

2917
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2918 2919
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2920
				details);
L
Linus Torvalds 已提交
2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
	}
}

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.
	 */
2935
	list_for_each_entry(vma, head, shared.nonlinear) {
L
Linus Torvalds 已提交
2936
		details->nonlinear_vma = vma;
2937
		unmap_mapping_range_vma(vma, vma->vm_start, vma->vm_end, details);
L
Linus Torvalds 已提交
2938 2939 2940 2941
	}
}

/**
2942
 * 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
2943
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2944 2945
 * @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
2946
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977
 * 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;


2978
	mutex_lock(&mapping->i_mmap_mutex);
2979
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2980 2981 2982
		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);
2983
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
2984 2985 2986 2987
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2988 2989 2990
 * 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
2991
 */
2992 2993
static int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
2994
		unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
2995
{
2996
	spinlock_t *ptl;
2997
	struct page *page, *swapcache;
2998
	swp_entry_t entry;
L
Linus Torvalds 已提交
2999
	pte_t pte;
3000
	int locked;
3001
	struct mem_cgroup *ptr;
3002
	int exclusive = 0;
N
Nick Piggin 已提交
3003
	int ret = 0;
L
Linus Torvalds 已提交
3004

H
Hugh Dickins 已提交
3005
	if (!pte_unmap_same(mm, pmd, page_table, orig_pte))
3006
		goto out;
3007 3008

	entry = pte_to_swp_entry(orig_pte);
3009 3010 3011 3012 3013 3014 3015
	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
3016
			ret = VM_FAULT_SIGBUS;
3017
		}
3018 3019
		goto out;
	}
3020
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
3021 3022
	page = lookup_swap_cache(entry);
	if (!page) {
3023 3024
		page = swapin_readahead(entry,
					GFP_HIGHUSER_MOVABLE, vma, address);
L
Linus Torvalds 已提交
3025 3026
		if (!page) {
			/*
3027 3028
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
3029
			 */
3030
			page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
3031 3032
			if (likely(pte_same(*page_table, orig_pte)))
				ret = VM_FAULT_OOM;
3033
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
3034
			goto unlock;
L
Linus Torvalds 已提交
3035 3036 3037 3038
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
3039
		count_vm_event(PGMAJFAULT);
3040
		mem_cgroup_count_vm_event(mm, PGMAJFAULT);
3041
	} else if (PageHWPoison(page)) {
3042 3043 3044 3045
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
3046 3047
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
3048
		swapcache = page;
3049
		goto out_release;
L
Linus Torvalds 已提交
3050 3051
	}

3052
	swapcache = page;
3053
	locked = lock_page_or_retry(page, mm, flags);
R
Rik van Riel 已提交
3054

3055
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
3056 3057 3058 3059
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
3060

A
Andrea Arcangeli 已提交
3061
	/*
3062 3063 3064 3065
	 * 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
3066
	 */
3067
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
3068 3069
		goto out_page;

3070 3071 3072 3073 3074
	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
3075 3076
	}

K
KAMEZAWA Hiroyuki 已提交
3077
	if (mem_cgroup_try_charge_swapin(mm, page, GFP_KERNEL, &ptr)) {
3078
		ret = VM_FAULT_OOM;
3079
		goto out_page;
3080 3081
	}

L
Linus Torvalds 已提交
3082
	/*
3083
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
3084
	 */
3085
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
H
Hugh Dickins 已提交
3086
	if (unlikely(!pte_same(*page_table, orig_pte)))
3087 3088 3089 3090 3091
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
3092 3093
	}

3094 3095 3096 3097 3098 3099 3100 3101
	/*
	 * 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.
3102 3103 3104 3105
	 * 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().
3106
	 */
L
Linus Torvalds 已提交
3107

3108
	inc_mm_counter_fast(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
3109
	dec_mm_counter_fast(mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
3110
	pte = mk_pte(page, vma->vm_page_prot);
3111
	if ((flags & FAULT_FLAG_WRITE) && reuse_swap_page(page)) {
L
Linus Torvalds 已提交
3112
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
3113
		flags &= ~FAULT_FLAG_WRITE;
3114
		ret |= VM_FAULT_WRITE;
3115
		exclusive = 1;
L
Linus Torvalds 已提交
3116 3117
	}
	flush_icache_page(vma, page);
3118 3119
	if (pte_swp_soft_dirty(orig_pte))
		pte = pte_mksoft_dirty(pte);
L
Linus Torvalds 已提交
3120
	set_pte_at(mm, address, page_table, pte);
3121
	if (page == swapcache)
3122
		do_page_add_anon_rmap(page, vma, address, exclusive);
3123 3124
	else /* ksm created a completely new copy */
		page_add_new_anon_rmap(page, vma, address);
3125 3126
	/* It's better to call commit-charge after rmap is established */
	mem_cgroup_commit_charge_swapin(page, ptr);
L
Linus Torvalds 已提交
3127

3128
	swap_free(entry);
N
Nick Piggin 已提交
3129
	if (vm_swap_full() || (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
3130
		try_to_free_swap(page);
3131
	unlock_page(page);
3132
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
		/*
		 * 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);
	}
3144

3145
	if (flags & FAULT_FLAG_WRITE) {
3146 3147 3148
		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
3149 3150 3151 3152
		goto out;
	}

	/* No need to invalidate - it was non-present before */
3153
	update_mmu_cache(vma, address, page_table);
3154
unlock:
3155
	pte_unmap_unlock(page_table, ptl);
L
Linus Torvalds 已提交
3156 3157
out:
	return ret;
3158
out_nomap:
3159
	mem_cgroup_cancel_charge_swapin(ptr);
3160
	pte_unmap_unlock(page_table, ptl);
3161
out_page:
3162
	unlock_page(page);
3163
out_release:
3164
	page_cache_release(page);
3165
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
3166 3167 3168
		unlock_page(swapcache);
		page_cache_release(swapcache);
	}
3169
	return ret;
L
Linus Torvalds 已提交
3170 3171
}

3172
/*
3173 3174
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
3175 3176 3177 3178 3179 3180
 * 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) {
3181 3182 3183 3184 3185 3186 3187 3188 3189 3190
		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;
3191

3192
		expand_downwards(vma, address - PAGE_SIZE);
3193
	}
3194 3195 3196 3197 3198 3199 3200 3201 3202
	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);
	}
3203 3204 3205
	return 0;
}

L
Linus Torvalds 已提交
3206
/*
3207 3208 3209
 * 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
3210
 */
3211 3212
static int do_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
3213
		unsigned int flags)
L
Linus Torvalds 已提交
3214
{
3215 3216
	struct page *page;
	spinlock_t *ptl;
L
Linus Torvalds 已提交
3217 3218
	pte_t entry;

3219 3220 3221 3222
	pte_unmap(page_table);

	/* Check if we need to add a guard page to the stack */
	if (check_stack_guard_page(vma, address) < 0)
3223 3224
		return VM_FAULT_SIGBUS;

3225
	/* Use the zero-page for reads */
H
Hugh Dickins 已提交
3226 3227 3228
	if (!(flags & FAULT_FLAG_WRITE)) {
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(address),
						vma->vm_page_prot));
3229
		page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
H
Hugh Dickins 已提交
3230 3231 3232 3233 3234
		if (!pte_none(*page_table))
			goto unlock;
		goto setpte;
	}

N
Nick Piggin 已提交
3235 3236 3237 3238 3239 3240
	/* 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;
3241 3242 3243 3244 3245
	/*
	 * 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
3246
	__SetPageUptodate(page);
3247

K
KAMEZAWA Hiroyuki 已提交
3248
	if (mem_cgroup_newpage_charge(page, mm, GFP_KERNEL))
3249 3250
		goto oom_free_page;

N
Nick Piggin 已提交
3251
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
3252 3253
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
3254

N
Nick Piggin 已提交
3255
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
3256
	if (!pte_none(*page_table))
N
Nick Piggin 已提交
3257
		goto release;
H
Hugh Dickins 已提交
3258

3259
	inc_mm_counter_fast(mm, MM_ANONPAGES);
N
Nick Piggin 已提交
3260
	page_add_new_anon_rmap(page, vma, address);
H
Hugh Dickins 已提交
3261
setpte:
3262
	set_pte_at(mm, address, page_table, entry);
L
Linus Torvalds 已提交
3263 3264

	/* No need to invalidate - it was non-present before */
3265
	update_mmu_cache(vma, address, page_table);
3266
unlock:
3267
	pte_unmap_unlock(page_table, ptl);
N
Nick Piggin 已提交
3268
	return 0;
3269
release:
3270
	mem_cgroup_uncharge_page(page);
3271 3272
	page_cache_release(page);
	goto unlock;
3273
oom_free_page:
3274
	page_cache_release(page);
3275
oom:
L
Linus Torvalds 已提交
3276 3277 3278 3279
	return VM_FAULT_OOM;
}

/*
3280
 * __do_fault() tries to create a new page mapping. It aggressively
L
Linus Torvalds 已提交
3281
 * tries to share with existing pages, but makes a separate copy if
3282 3283
 * the FAULT_FLAG_WRITE is set in the flags parameter in order to avoid
 * the next page fault.
L
Linus Torvalds 已提交
3284 3285 3286 3287
 *
 * As this is called only for pages that do not currently exist, we
 * do not need to flush old virtual caches or the TLB.
 *
3288
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
3289
 * but allow concurrent faults), and pte neither mapped nor locked.
3290
 * We return with mmap_sem still held, but pte unmapped and unlocked.
L
Linus Torvalds 已提交
3291
 */
3292
static int __do_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3293
		unsigned long address, pmd_t *pmd,
3294
		pgoff_t pgoff, unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
3295
{
3296
	pte_t *page_table;
3297
	spinlock_t *ptl;
N
Nick Piggin 已提交
3298
	struct page *page;
3299
	struct page *cow_page;
L
Linus Torvalds 已提交
3300 3301
	pte_t entry;
	int anon = 0;
3302
	struct page *dirty_page = NULL;
N
Nick Piggin 已提交
3303 3304
	struct vm_fault vmf;
	int ret;
3305
	int page_mkwrite = 0;
3306

3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326
	/*
	 * If we do COW later, allocate page befor taking lock_page()
	 * on the file cache page. This will reduce lock holding time.
	 */
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {

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

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

		if (mem_cgroup_newpage_charge(cow_page, mm, GFP_KERNEL)) {
			page_cache_release(cow_page);
			return VM_FAULT_OOM;
		}
	} else
		cow_page = NULL;

N
Nick Piggin 已提交
3327 3328 3329 3330
	vmf.virtual_address = (void __user *)(address & PAGE_MASK);
	vmf.pgoff = pgoff;
	vmf.flags = flags;
	vmf.page = NULL;
L
Linus Torvalds 已提交
3331

N
Nick Piggin 已提交
3332
	ret = vma->vm_ops->fault(vma, &vmf);
3333 3334
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
			    VM_FAULT_RETRY)))
3335
		goto uncharge_out;
L
Linus Torvalds 已提交
3336

3337 3338 3339
	if (unlikely(PageHWPoison(vmf.page))) {
		if (ret & VM_FAULT_LOCKED)
			unlock_page(vmf.page);
3340 3341
		ret = VM_FAULT_HWPOISON;
		goto uncharge_out;
3342 3343
	}

3344
	/*
N
Nick Piggin 已提交
3345
	 * For consistency in subsequent calls, make the faulted page always
3346 3347
	 * locked.
	 */
N
Nick Piggin 已提交
3348
	if (unlikely(!(ret & VM_FAULT_LOCKED)))
N
Nick Piggin 已提交
3349
		lock_page(vmf.page);
3350
	else
N
Nick Piggin 已提交
3351
		VM_BUG_ON(!PageLocked(vmf.page));
3352

L
Linus Torvalds 已提交
3353 3354 3355
	/*
	 * Should we do an early C-O-W break?
	 */
N
Nick Piggin 已提交
3356
	page = vmf.page;
3357
	if (flags & FAULT_FLAG_WRITE) {
3358
		if (!(vma->vm_flags & VM_SHARED)) {
3359
			page = cow_page;
3360
			anon = 1;
N
Nick Piggin 已提交
3361
			copy_user_highpage(page, vmf.page, address, vma);
N
Nick Piggin 已提交
3362
			__SetPageUptodate(page);
3363
		} else {
3364 3365
			/*
			 * If the page will be shareable, see if the backing
3366
			 * address space wants to know that the page is about
3367 3368
			 * to become writable
			 */
3369
			if (vma->vm_ops->page_mkwrite) {
3370 3371
				int tmp;

3372
				unlock_page(page);
N
Nick Piggin 已提交
3373
				vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
3374 3375 3376 3377
				tmp = vma->vm_ops->page_mkwrite(vma, &vmf);
				if (unlikely(tmp &
					  (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
					ret = tmp;
N
Nick Piggin 已提交
3378
					goto unwritable_page;
N
Nick Piggin 已提交
3379
				}
N
Nick Piggin 已提交
3380 3381 3382 3383 3384 3385 3386 3387 3388
				if (unlikely(!(tmp & VM_FAULT_LOCKED))) {
					lock_page(page);
					if (!page->mapping) {
						ret = 0; /* retry the fault */
						unlock_page(page);
						goto unwritable_page;
					}
				} else
					VM_BUG_ON(!PageLocked(page));
3389
				page_mkwrite = 1;
3390 3391
			}
		}
3392

L
Linus Torvalds 已提交
3393 3394
	}

3395
	page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
3396 3397 3398 3399 3400 3401

	/*
	 * This silly early PAGE_DIRTY setting removes a race
	 * due to the bad i386 page protection. But it's valid
	 * for other architectures too.
	 *
3402
	 * Note that if FAULT_FLAG_WRITE is set, we either now have
L
Linus Torvalds 已提交
3403 3404 3405 3406 3407
	 * an exclusive copy of the page, or this is a shared mapping,
	 * so we can make it writable and dirty to avoid having to
	 * handle that later.
	 */
	/* Only go through if we didn't race with anybody else... */
3408
	if (likely(pte_same(*page_table, orig_pte))) {
3409 3410
		flush_icache_page(vma, page);
		entry = mk_pte(page, vma->vm_page_prot);
3411
		if (flags & FAULT_FLAG_WRITE)
L
Linus Torvalds 已提交
3412
			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
3413 3414
		else if (pte_file(orig_pte) && pte_file_soft_dirty(orig_pte))
			pte_mksoft_dirty(entry);
L
Linus Torvalds 已提交
3415
		if (anon) {
3416
			inc_mm_counter_fast(mm, MM_ANONPAGES);
3417
			page_add_new_anon_rmap(page, vma, address);
3418
		} else {
3419
			inc_mm_counter_fast(mm, MM_FILEPAGES);
3420
			page_add_file_rmap(page);
3421
			if (flags & FAULT_FLAG_WRITE) {
3422
				dirty_page = page;
3423 3424
				get_page(dirty_page);
			}
3425
		}
3426
		set_pte_at(mm, address, page_table, entry);
3427 3428

		/* no need to invalidate: a not-present page won't be cached */
3429
		update_mmu_cache(vma, address, page_table);
L
Linus Torvalds 已提交
3430
	} else {
3431 3432
		if (cow_page)
			mem_cgroup_uncharge_page(cow_page);
3433 3434 3435
		if (anon)
			page_cache_release(page);
		else
3436
			anon = 1; /* no anon but release faulted_page */
L
Linus Torvalds 已提交
3437 3438
	}

3439
	pte_unmap_unlock(page_table, ptl);
3440

N
Nick Piggin 已提交
3441 3442
	if (dirty_page) {
		struct address_space *mapping = page->mapping;
3443
		int dirtied = 0;
3444

N
Nick Piggin 已提交
3445
		if (set_page_dirty(dirty_page))
3446
			dirtied = 1;
N
Nick Piggin 已提交
3447
		unlock_page(dirty_page);
3448
		put_page(dirty_page);
3449
		if ((dirtied || page_mkwrite) && mapping) {
N
Nick Piggin 已提交
3450 3451 3452 3453 3454 3455 3456 3457
			/*
			 * Some device drivers do not set page.mapping but still
			 * dirty their pages
			 */
			balance_dirty_pages_ratelimited(mapping);
		}

		/* file_update_time outside page_lock */
3458
		if (vma->vm_file && !page_mkwrite)
N
Nick Piggin 已提交
3459 3460 3461 3462 3463
			file_update_time(vma->vm_file);
	} else {
		unlock_page(vmf.page);
		if (anon)
			page_cache_release(vmf.page);
3464
	}
3465

N
Nick Piggin 已提交
3466
	return ret;
N
Nick Piggin 已提交
3467 3468 3469 3470

unwritable_page:
	page_cache_release(page);
	return ret;
3471 3472 3473 3474 3475 3476 3477
uncharge_out:
	/* fs's fault handler get error */
	if (cow_page) {
		mem_cgroup_uncharge_page(cow_page);
		page_cache_release(cow_page);
	}
	return ret;
3478
}
3479

3480 3481
static int do_linear_fault(struct mm_struct *mm, struct vm_area_struct *vma,
		unsigned long address, pte_t *page_table, pmd_t *pmd,
3482
		unsigned int flags, pte_t orig_pte)
3483 3484
{
	pgoff_t pgoff = (((address & PAGE_MASK)
3485
			- vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
3486

3487 3488
	pte_unmap(page_table);
	return __do_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
3489 3490
}

L
Linus Torvalds 已提交
3491 3492 3493 3494
/*
 * Fault of a previously existing named mapping. Repopulate the pte
 * from the encoded file_pte if possible. This enables swappable
 * nonlinear vmas.
3495 3496 3497 3498
 *
 * 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
3499
 */
N
Nick Piggin 已提交
3500
static int do_nonlinear_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3501
		unsigned long address, pte_t *page_table, pmd_t *pmd,
3502
		unsigned int flags, pte_t orig_pte)
L
Linus Torvalds 已提交
3503
{
3504
	pgoff_t pgoff;
L
Linus Torvalds 已提交
3505

3506 3507
	flags |= FAULT_FLAG_NONLINEAR;

H
Hugh Dickins 已提交
3508
	if (!pte_unmap_same(mm, pmd, page_table, orig_pte))
N
Nick Piggin 已提交
3509
		return 0;
L
Linus Torvalds 已提交
3510

3511
	if (unlikely(!(vma->vm_flags & VM_NONLINEAR))) {
3512 3513 3514
		/*
		 * Page table corrupted: show pte and kill process.
		 */
3515
		print_bad_pte(vma, address, orig_pte, NULL);
H
Hugh Dickins 已提交
3516
		return VM_FAULT_SIGBUS;
3517 3518 3519
	}

	pgoff = pte_to_pgoff(orig_pte);
3520
	return __do_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
L
Linus Torvalds 已提交
3521 3522
}

3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534
int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
				unsigned long addr, int current_nid)
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
	if (current_nid == numa_node_id())
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);

	return mpol_misplaced(page, vma, addr);
}

3535 3536 3537
int do_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
		   unsigned long addr, pte_t pte, pte_t *ptep, pmd_t *pmd)
{
3538
	struct page *page = NULL;
3539
	spinlock_t *ptl;
3540 3541
	int current_nid = -1;
	int target_nid;
3542
	bool migrated = false;
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554

	/*
	* 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);
3555 3556 3557 3558 3559
	if (unlikely(!pte_same(*ptep, pte))) {
		pte_unmap_unlock(ptep, ptl);
		goto out;
	}

3560 3561 3562 3563 3564 3565 3566 3567 3568 3569
	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;
	}

3570
	current_nid = page_to_nid(page);
3571
	target_nid = numa_migrate_prep(page, vma, addr, current_nid);
3572
	pte_unmap_unlock(ptep, ptl);
3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583
	if (target_nid == -1) {
		/*
		 * Account for the fault against the current node if it not
		 * being replaced regardless of where the page is located.
		 */
		current_nid = numa_node_id();
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3584 3585
	migrated = migrate_misplaced_page(page, target_nid);
	if (migrated)
3586 3587 3588
		current_nid = target_nid;

out:
3589
	if (current_nid != -1)
3590
		task_numa_fault(current_nid, 1, migrated);
3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
	return 0;
}

/* NUMA hinting page fault entry point for regular pmds */
#ifdef CONFIG_NUMA_BALANCING
static int do_pmd_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
		     unsigned long addr, pmd_t *pmdp)
{
	pmd_t pmd;
	pte_t *pte, *orig_pte;
	unsigned long _addr = addr & PMD_MASK;
	unsigned long offset;
	spinlock_t *ptl;
	bool numa = false;
3605
	int local_nid = numa_node_id();
3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627

	spin_lock(&mm->page_table_lock);
	pmd = *pmdp;
	if (pmd_numa(pmd)) {
		set_pmd_at(mm, _addr, pmdp, pmd_mknonnuma(pmd));
		numa = true;
	}
	spin_unlock(&mm->page_table_lock);

	if (!numa)
		return 0;

	/* we're in a page fault so some vma must be in the range */
	BUG_ON(!vma);
	BUG_ON(vma->vm_start >= _addr + PMD_SIZE);
	offset = max(_addr, vma->vm_start) & ~PMD_MASK;
	VM_BUG_ON(offset >= PMD_SIZE);
	orig_pte = pte = pte_offset_map_lock(mm, pmdp, _addr, &ptl);
	pte += offset >> PAGE_SHIFT;
	for (addr = _addr + offset; addr < _addr + PMD_SIZE; pte++, addr += PAGE_SIZE) {
		pte_t pteval = *pte;
		struct page *page;
3628 3629
		int curr_nid = local_nid;
		int target_nid;
3630
		bool migrated;
3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647
		if (!pte_present(pteval))
			continue;
		if (!pte_numa(pteval))
			continue;
		if (addr >= vma->vm_end) {
			vma = find_vma(mm, addr);
			/* there's a pte present so there must be a vma */
			BUG_ON(!vma);
			BUG_ON(addr < vma->vm_start);
		}
		if (pte_numa(pteval)) {
			pteval = pte_mknonnuma(pteval);
			set_pte_at(mm, addr, pte, pteval);
		}
		page = vm_normal_page(vma, addr, pteval);
		if (unlikely(!page))
			continue;
3648 3649 3650 3651
		/* only check non-shared pages */
		if (unlikely(page_mapcount(page) != 1))
			continue;

3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663
		/*
		 * Note that the NUMA fault is later accounted to either
		 * the node that is currently running or where the page is
		 * migrated to.
		 */
		curr_nid = local_nid;
		target_nid = numa_migrate_prep(page, vma, addr,
					       page_to_nid(page));
		if (target_nid == -1) {
			put_page(page);
			continue;
		}
3664

3665 3666
		/* Migrate to the requested node */
		pte_unmap_unlock(pte, ptl);
3667 3668
		migrated = migrate_misplaced_page(page, target_nid);
		if (migrated)
3669
			curr_nid = target_nid;
3670
		task_numa_fault(curr_nid, 1, migrated);
3671

3672
		pte = pte_offset_map_lock(mm, pmdp, addr, &ptl);
3673 3674 3675 3676 3677 3678 3679 3680 3681 3682
	}
	pte_unmap_unlock(orig_pte, ptl);

	return 0;
}
#else
static int do_pmd_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
		     unsigned long addr, pmd_t *pmdp)
{
	BUG();
A
Andrew Morton 已提交
3683
	return 0;
3684 3685 3686
}
#endif /* CONFIG_NUMA_BALANCING */

L
Linus Torvalds 已提交
3687 3688 3689 3690 3691 3692 3693 3694 3695
/*
 * 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
3696 3697 3698
 * 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
3699
 */
3700
static int handle_pte_fault(struct mm_struct *mm,
3701 3702
		     struct vm_area_struct *vma, unsigned long address,
		     pte_t *pte, pmd_t *pmd, unsigned int flags)
L
Linus Torvalds 已提交
3703 3704
{
	pte_t entry;
3705
	spinlock_t *ptl;
L
Linus Torvalds 已提交
3706

3707
	entry = *pte;
L
Linus Torvalds 已提交
3708
	if (!pte_present(entry)) {
3709
		if (pte_none(entry)) {
J
Jes Sorensen 已提交
3710
			if (vma->vm_ops) {
N
Nick Piggin 已提交
3711
				if (likely(vma->vm_ops->fault))
3712
					return do_linear_fault(mm, vma, address,
3713
						pte, pmd, flags, entry);
J
Jes Sorensen 已提交
3714 3715
			}
			return do_anonymous_page(mm, vma, address,
3716
						 pte, pmd, flags);
3717
		}
L
Linus Torvalds 已提交
3718
		if (pte_file(entry))
N
Nick Piggin 已提交
3719
			return do_nonlinear_fault(mm, vma, address,
3720
					pte, pmd, flags, entry);
3721
		return do_swap_page(mm, vma, address,
3722
					pte, pmd, flags, entry);
L
Linus Torvalds 已提交
3723 3724
	}

3725 3726 3727
	if (pte_numa(entry))
		return do_numa_page(mm, vma, address, entry, pte, pmd);

H
Hugh Dickins 已提交
3728
	ptl = pte_lockptr(mm, pmd);
3729 3730 3731
	spin_lock(ptl);
	if (unlikely(!pte_same(*pte, entry)))
		goto unlock;
3732
	if (flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3733
		if (!pte_write(entry))
3734 3735
			return do_wp_page(mm, vma, address,
					pte, pmd, ptl, entry);
L
Linus Torvalds 已提交
3736 3737 3738
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
3739
	if (ptep_set_access_flags(vma, address, pte, entry, flags & FAULT_FLAG_WRITE)) {
3740
		update_mmu_cache(vma, address, pte);
3741 3742 3743 3744 3745 3746 3747
	} 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.
		 */
3748
		if (flags & FAULT_FLAG_WRITE)
3749
			flush_tlb_fix_spurious_fault(vma, address);
3750
	}
3751 3752
unlock:
	pte_unmap_unlock(pte, ptl);
N
Nick Piggin 已提交
3753
	return 0;
L
Linus Torvalds 已提交
3754 3755 3756 3757 3758
}

/*
 * By the time we get here, we already hold the mm semaphore
 */
3759 3760
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
3761 3762 3763 3764 3765 3766
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

3767
	if (unlikely(is_vm_hugetlb_page(vma)))
3768
		return hugetlb_fault(mm, vma, address, flags);
L
Linus Torvalds 已提交
3769

3770
retry:
L
Linus Torvalds 已提交
3771 3772 3773
	pgd = pgd_offset(mm, address);
	pud = pud_alloc(mm, pgd, address);
	if (!pud)
H
Hugh Dickins 已提交
3774
		return VM_FAULT_OOM;
L
Linus Torvalds 已提交
3775 3776
	pmd = pmd_alloc(mm, pud, address);
	if (!pmd)
H
Hugh Dickins 已提交
3777
		return VM_FAULT_OOM;
3778
	if (pmd_none(*pmd) && transparent_hugepage_enabled(vma)) {
3779
		int ret = VM_FAULT_FALLBACK;
3780
		if (!vma->vm_ops)
3781 3782 3783 3784
			ret = do_huge_pmd_anonymous_page(mm, vma, address,
					pmd, flags);
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3785 3786
	} else {
		pmd_t orig_pmd = *pmd;
3787 3788
		int ret;

3789 3790
		barrier();
		if (pmd_trans_huge(orig_pmd)) {
3791 3792
			unsigned int dirty = flags & FAULT_FLAG_WRITE;

3793 3794 3795 3796 3797 3798 3799 3800
			/*
			 * 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;

3801
			if (pmd_numa(orig_pmd))
3802
				return do_huge_pmd_numa_page(mm, vma, address,
3803 3804
							     orig_pmd, pmd);

3805
			if (dirty && !pmd_write(orig_pmd)) {
3806 3807 3808 3809 3810 3811 3812 3813 3814 3815
				ret = do_huge_pmd_wp_page(mm, vma, address, pmd,
							  orig_pmd);
				/*
				 * If COW results in an oom, the huge pmd will
				 * have been split, so retry the fault on the
				 * pte for a smaller charge.
				 */
				if (unlikely(ret & VM_FAULT_OOM))
					goto retry;
				return ret;
3816 3817 3818
			} else {
				huge_pmd_set_accessed(mm, vma, address, pmd,
						      orig_pmd, dirty);
3819
			}
3820

3821 3822 3823 3824
			return 0;
		}
	}

3825 3826 3827
	if (pmd_numa(*pmd))
		return do_pmd_numa_page(mm, vma, address, pmd);

3828 3829 3830 3831 3832
	/*
	 * 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.
	 */
3833 3834
	if (unlikely(pmd_none(*pmd)) &&
	    unlikely(__pte_alloc(mm, vma, pmd, address)))
H
Hugh Dickins 已提交
3835
		return VM_FAULT_OOM;
3836 3837 3838 3839 3840 3841 3842 3843 3844 3845
	/* 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
3846

3847
	return handle_pte_fault(mm, vma, address, pte, pmd, flags);
L
Linus Torvalds 已提交
3848 3849
}

3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874
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)
		mem_cgroup_enable_oom();

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

	if (flags & FAULT_FLAG_USER)
		mem_cgroup_disable_oom();

3875 3876 3877
	if (WARN_ON(task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM)))
		mem_cgroup_oom_synchronize();

3878 3879 3880
	return ret;
}

L
Linus Torvalds 已提交
3881 3882 3883
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3884
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3885
 */
3886
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3887
{
H
Hugh Dickins 已提交
3888 3889
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3890
		return -ENOMEM;
L
Linus Torvalds 已提交
3891

3892 3893
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3894
	spin_lock(&mm->page_table_lock);
3895
	if (pgd_present(*pgd))		/* Another has populated it */
3896
		pud_free(mm, new);
3897 3898
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3899
	spin_unlock(&mm->page_table_lock);
3900
	return 0;
L
Linus Torvalds 已提交
3901 3902 3903 3904 3905 3906
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3907
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3908
 */
3909
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3910
{
H
Hugh Dickins 已提交
3911 3912
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3913
		return -ENOMEM;
L
Linus Torvalds 已提交
3914

3915 3916
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3917
	spin_lock(&mm->page_table_lock);
L
Linus Torvalds 已提交
3918
#ifndef __ARCH_HAS_4LEVEL_HACK
3919
	if (pud_present(*pud))		/* Another has populated it */
3920
		pmd_free(mm, new);
3921 3922
	else
		pud_populate(mm, pud, new);
L
Linus Torvalds 已提交
3923
#else
3924
	if (pgd_present(*pud))		/* Another has populated it */
3925
		pmd_free(mm, new);
3926 3927
	else
		pgd_populate(mm, pud, new);
L
Linus Torvalds 已提交
3928
#endif /* __ARCH_HAS_4LEVEL_HACK */
H
Hugh Dickins 已提交
3929
	spin_unlock(&mm->page_table_lock);
3930
	return 0;
3931
}
L
Linus Torvalds 已提交
3932 3933 3934 3935 3936
#endif /* __PAGETABLE_PMD_FOLDED */

#if !defined(__HAVE_ARCH_GATE_AREA)

#if defined(AT_SYSINFO_EHDR)
3937
static struct vm_area_struct gate_vma;
L
Linus Torvalds 已提交
3938 3939 3940 3941 3942 3943

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
3944 3945
	gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
	gate_vma.vm_page_prot = __P101;
3946

L
Linus Torvalds 已提交
3947 3948 3949 3950 3951
	return 0;
}
__initcall(gate_vma_init);
#endif

3952
struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
L
Linus Torvalds 已提交
3953 3954 3955 3956 3957 3958 3959 3960
{
#ifdef AT_SYSINFO_EHDR
	return &gate_vma;
#else
	return NULL;
#endif
}

3961
int in_gate_area_no_mm(unsigned long addr)
L
Linus Torvalds 已提交
3962 3963 3964 3965 3966 3967 3968 3969 3970
{
#ifdef AT_SYSINFO_EHDR
	if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
		return 1;
#endif
	return 0;
}

#endif	/* __HAVE_ARCH_GATE_AREA */
3971

3972
static int __follow_pte(struct mm_struct *mm, unsigned long address,
J
Johannes Weiner 已提交
3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988
		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);
3989
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009
	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;
}

4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020
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
4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049
/**
 * 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);

4050
#ifdef CONFIG_HAVE_IOREMAP_PROT
4051 4052 4053
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4054
{
4055
	int ret = -EINVAL;
4056 4057 4058
	pte_t *ptep, pte;
	spinlock_t *ptl;

4059 4060
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4061

4062
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4063
		goto out;
4064
	pte = *ptep;
4065

4066 4067 4068 4069
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4070
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4071

4072
	ret = 0;
4073 4074 4075
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4076
	return ret;
4077 4078 4079 4080 4081 4082 4083
}

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
4084
	void __iomem *maddr;
4085 4086
	int offset = addr & (PAGE_SIZE-1);

4087
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098
		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;
}
4099
EXPORT_SYMBOL_GPL(generic_access_phys);
4100 4101
#endif

4102
/*
4103 4104
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4105
 */
4106 4107
static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
		unsigned long addr, void *buf, int len, int write)
4108 4109 4110 4111 4112
{
	struct vm_area_struct *vma;
	void *old_buf = buf;

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4113
	/* ignore errors, just check how much was successfully transferred */
4114 4115 4116
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4117
		struct page *page = NULL;
4118 4119 4120

		ret = get_user_pages(tsk, mm, addr, 1,
				write, 1, &page, &vma);
4121 4122 4123 4124 4125 4126 4127
		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);
4128
			if (!vma || vma->vm_start > addr)
4129 4130 4131 4132 4133 4134 4135 4136
				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;
4137
		} else {
4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153
			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);
4154 4155 4156 4157 4158 4159 4160 4161 4162
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4163

S
Stephen Wilson 已提交
4164
/**
4165
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179
 * @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);
}

4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200
/*
 * 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;
}

4201 4202 4203 4204 4205 4206 4207 4208
/*
 * 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;

4209 4210 4211 4212 4213 4214 4215
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

4216 4217 4218 4219 4220 4221
	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
4222
			char *p;
4223

4224
			p = d_path(&f->f_path, buf, PAGE_SIZE);
4225 4226
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4227
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4228 4229 4230 4231 4232
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4233
	up_read(&mm->mmap_sem);
4234
}
4235

4236
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4237 4238
void might_fault(void)
{
4239 4240 4241 4242 4243 4244 4245 4246 4247
	/*
	 * 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;

4248 4249 4250 4251 4252
	/*
	 * 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.
	 */
4253 4254 4255 4256 4257 4258
	if (in_atomic())
		return;

	__might_sleep(__FILE__, __LINE__, 0);

	if (current->mm)
4259 4260 4261 4262
		might_lock_read(&current->mm->mmap_sem);
}
EXPORT_SYMBOL(might_fault);
#endif
A
Andrea Arcangeli 已提交
4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333

#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 */