memory.c 109.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
/*
 *  linux/mm/memory.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

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

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

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

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

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

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

76 77
#include "internal.h"

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

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

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

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

EXPORT_SYMBOL(high_memory);

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

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

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

125 126
EXPORT_SYMBOL(zero_pfn);

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

K
KAMEZAWA Hiroyuki 已提交
137

138 139
#if defined(SPLIT_RSS_COUNTING)

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

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

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

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

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

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

static void check_sync_rss_stat(struct task_struct *task)
{
}

183 184 185 186
#endif /* SPLIT_RSS_COUNTING */

#ifdef HAVE_GENERIC_MMU_GATHER

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

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

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

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

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

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

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

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

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

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

	__tlb_reset_range(tlb);
239 240
}

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

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

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

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

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

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

	tlb_flush_mmu(tlb);

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

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

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

302
	VM_BUG_ON(!tlb->end);
303

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

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

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

#endif /* HAVE_GENERIC_MMU_GATHER */

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

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

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

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

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

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

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

	free_page((unsigned long)batch);
}

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

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

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

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

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

398
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
399

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

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

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

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

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

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

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

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

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

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

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

516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
	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;

532
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
533 534 535 536
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
537
		free_pud_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
538
	} while (pgd++, addr = next, addr != end);
539 540
}

541
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
542
		unsigned long floor, unsigned long ceiling)
543 544 545 546 547
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

548
		/*
N
npiggin@suse.de 已提交
549 550
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
551
		 */
552
		unlink_anon_vmas(vma);
553 554
		unlink_file_vma(vma);

555
		if (is_vm_hugetlb_page(vma)) {
556
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
557
				floor, next? next->vm_start: ceiling);
558 559 560 561 562
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
563
			       && !is_vm_hugetlb_page(next)) {
564 565
				vma = next;
				next = vma->vm_next;
566
				unlink_anon_vmas(vma);
567
				unlink_file_vma(vma);
568 569 570 571
			}
			free_pgd_range(tlb, addr, vma->vm_end,
				floor, next? next->vm_start: ceiling);
		}
572 573
		vma = next;
	}
L
Linus Torvalds 已提交
574 575
}

576
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
577
{
578
	spinlock_t *ptl;
579
	pgtable_t new = pte_alloc_one(mm, address);
580 581 582
	if (!new)
		return -ENOMEM;

583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
	/*
	 * 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 */

598
	ptl = pmd_lock(mm, pmd);
599
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
600
		atomic_long_inc(&mm->nr_ptes);
L
Linus Torvalds 已提交
601
		pmd_populate(mm, pmd, new);
602
		new = NULL;
603
	}
604
	spin_unlock(ptl);
605 606
	if (new)
		pte_free(mm, new);
607
	return 0;
L
Linus Torvalds 已提交
608 609
}

610
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
611
{
612 613 614 615
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

616 617
	smp_wmb(); /* See comment in __pte_alloc */

618
	spin_lock(&init_mm.page_table_lock);
619
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
620
		pmd_populate_kernel(&init_mm, pmd, new);
621
		new = NULL;
622
	}
623
	spin_unlock(&init_mm.page_table_lock);
624 625
	if (new)
		pte_free_kernel(&init_mm, new);
626
	return 0;
L
Linus Torvalds 已提交
627 628
}

K
KAMEZAWA Hiroyuki 已提交
629 630 631 632 633 634
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)
635
{
K
KAMEZAWA Hiroyuki 已提交
636 637
	int i;

638
	if (current->mm == mm)
639
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
640 641 642
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
643 644
}

N
Nick Piggin 已提交
645
/*
646 647 648
 * 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
649 650 651
 *
 * The calling function must still handle the error.
 */
652 653
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
654
{
655 656 657 658 659
	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;
660 661 662 663 664 665 666 667 668 669 670 671 672 673
	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) {
674 675
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
676 677 678 679 680 681
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
682 683 684 685

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

686 687 688
	pr_alert("BUG: Bad page map in process %s  pte:%08llx pmd:%08llx\n",
		 current->comm,
		 (long long)pte_val(pte), (long long)pmd_val(*pmd));
689
	if (page)
690
		dump_page(page, "bad pte");
691 692
	pr_alert("addr:%p vm_flags:%08lx anon_vma:%p mapping:%p index:%lx\n",
		 (void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
693 694 695
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
696 697 698 699 700
	pr_alert("file:%pD fault:%pf mmap:%pf readpage:%pf\n",
		 vma->vm_file,
		 vma->vm_ops ? vma->vm_ops->fault : NULL,
		 vma->vm_file ? vma->vm_file->f_op->mmap : NULL,
		 mapping ? mapping->a_ops->readpage : NULL);
N
Nick Piggin 已提交
701
	dump_stack();
702
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
703 704
}

H
Hugh Dickins 已提交
705
/*
N
Nick Piggin 已提交
706
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
707
 *
N
Nick Piggin 已提交
708 709 710
 * "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
711
 *
N
Nick Piggin 已提交
712 713 714 715 716 717 718 719
 * 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.
720
 *
J
Jared Hulbert 已提交
721 722
 * 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
723 724
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
725 726 727
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
728 729 730 731 732 733
 * 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
734 735
 *
 *
N
Nick Piggin 已提交
736
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
737 738 739 740 741 742 743 744 745
 *
 * 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
746
 */
N
Nick Piggin 已提交
747 748 749 750 751 752 753
#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
754
{
755
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
756 757

	if (HAVE_PTE_SPECIAL) {
758
		if (likely(!pte_special(pte)))
759
			goto check_pfn;
760 761
		if (vma->vm_ops && vma->vm_ops->find_special_page)
			return vma->vm_ops->find_special_page(vma, addr);
H
Hugh Dickins 已提交
762 763
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
H
Hugh Dickins 已提交
764
		if (!is_zero_pfn(pfn))
765
			print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
766 767 768 769 770
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
771 772 773 774 775 776
	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
777 778
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
779 780 781 782 783
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
784 785
	}

786 787
	if (is_zero_pfn(pfn))
		return NULL;
788 789 790 791 792
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
793 794

	/*
N
Nick Piggin 已提交
795 796
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
797
	 */
J
Jared Hulbert 已提交
798
out:
799
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
800 801
}

802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
				pmd_t pmd)
{
	unsigned long pfn = pmd_pfn(pmd);

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

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

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

L
Linus Torvalds 已提交
842 843 844 845 846 847
/*
 * 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
848
static inline unsigned long
L
Linus Torvalds 已提交
849
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
850
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
851
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
852
{
N
Nick Piggin 已提交
853
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
854 855 856 857 858
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876
		swp_entry_t entry = pte_to_swp_entry(pte);

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

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

877
			rss[mm_counter(page)]++;
878 879 880 881 882 883 884 885 886 887 888 889

			if (is_write_migration_entry(entry) &&
					is_cow_mapping(vm_flags)) {
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
				make_migration_entry_read(&entry);
				pte = swp_entry_to_pte(entry);
				if (pte_swp_soft_dirty(*src_pte))
					pte = pte_swp_mksoft_dirty(pte);
				set_pte_at(src_mm, addr, src_pte, pte);
890
			}
L
Linus Torvalds 已提交
891
		}
892
		goto out_set_pte;
L
Linus Torvalds 已提交
893 894 895 896 897 898
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
899
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
900
		ptep_set_wrprotect(src_mm, addr, src_pte);
901
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
902 903 904 905 906 907 908 909 910
	}

	/*
	 * 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);
911 912 913 914

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
915
		page_dup_rmap(page, false);
916
		rss[mm_counter(page)]++;
917
	}
918 919 920

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
921
	return 0;
L
Linus Torvalds 已提交
922 923
}

924
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
925 926
		   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
927
{
928
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
929
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
930
	spinlock_t *src_ptl, *dst_ptl;
931
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
932
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
933
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
934 935

again:
K
KAMEZAWA Hiroyuki 已提交
936 937
	init_rss_vec(rss);

H
Hugh Dickins 已提交
938
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
939 940
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
941
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
942
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
943
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
944 945
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
946
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
947 948 949 950 951 952

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
953 954 955
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
956
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
957 958
				break;
		}
L
Linus Torvalds 已提交
959 960 961 962
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
963 964 965 966
		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
967 968 969
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

970
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
971
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
972
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
973
	add_mm_rss_vec(dst_mm, rss);
974
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
975
	cond_resched();
H
Hugh Dickins 已提交
976 977 978 979 980 981

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
	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);
1000
		if (pmd_trans_huge(*src_pmd) || pmd_devmap(*src_pmd)) {
1001
			int err;
1002
			VM_BUG_ON(next-addr != HPAGE_PMD_SIZE);
1003 1004 1005 1006 1007 1008 1009 1010
			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
1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
		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;
1049 1050 1051
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1052
	int ret;
L
Linus Torvalds 已提交
1053

1054 1055 1056 1057 1058 1059
	/*
	 * 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.
	 */
1060 1061 1062
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1063

L
Linus Torvalds 已提交
1064 1065 1066
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1067
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1068 1069 1070 1071
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1072
		ret = track_pfn_copy(vma);
1073 1074 1075 1076
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1077 1078 1079 1080 1081 1082
	/*
	 * 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.
	 */
1083 1084 1085 1086 1087 1088
	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
1089 1090

	ret = 0;
L
Linus Torvalds 已提交
1091 1092 1093 1094 1095 1096
	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
1097 1098 1099 1100 1101
		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
1102
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1103

1104 1105
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1106
	return ret;
L
Linus Torvalds 已提交
1107 1108
}

1109
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1110
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1111
				unsigned long addr, unsigned long end,
1112
				struct zap_details *details)
L
Linus Torvalds 已提交
1113
{
N
Nick Piggin 已提交
1114
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1115
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1116
	int rss[NR_MM_COUNTERS];
1117
	spinlock_t *ptl;
1118
	pte_t *start_pte;
1119
	pte_t *pte;
1120
	swp_entry_t entry;
1121
	struct page *pending_page = NULL;
K
KAMEZAWA Hiroyuki 已提交
1122

P
Peter Zijlstra 已提交
1123
again:
1124
	init_rss_vec(rss);
1125 1126
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1127
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1128 1129
	do {
		pte_t ptent = *pte;
1130
		if (pte_none(ptent)) {
L
Linus Torvalds 已提交
1131
			continue;
1132
		}
1133

L
Linus Torvalds 已提交
1134
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1135
			struct page *page;
1136

1137
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
			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;
			}
N
Nick Piggin 已提交
1148
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1149
							tlb->fullmm);
L
Linus Torvalds 已提交
1150 1151 1152
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1153 1154

			if (!PageAnon(page)) {
1155
				if (pte_dirty(ptent)) {
M
Michal Hocko 已提交
1156 1157 1158 1159 1160 1161
					/*
					 * oom_reaper cannot tear down dirty
					 * pages
					 */
					if (unlikely(details && details->ignore_dirty))
						continue;
1162
					force_flush = 1;
1163
					set_page_dirty(page);
1164
				}
1165
				if (pte_young(ptent) &&
1166
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1167
					mark_page_accessed(page);
1168
			}
1169
			rss[mm_counter(page)]--;
1170
			page_remove_rmap(page, false);
1171 1172
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1173
			if (unlikely(__tlb_remove_page(tlb, page))) {
1174
				force_flush = 1;
1175
				pending_page = page;
1176
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1177
				break;
1178
			}
L
Linus Torvalds 已提交
1179 1180
			continue;
		}
M
Michal Hocko 已提交
1181 1182
		/* only check swap_entries if explicitly asked for in details */
		if (unlikely(details && !details->check_swap_entries))
L
Linus Torvalds 已提交
1183
			continue;
K
KAMEZAWA Hiroyuki 已提交
1184

1185 1186 1187 1188 1189
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1190

1191
			page = migration_entry_to_page(entry);
1192
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1193
		}
1194 1195
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1196
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1197
	} while (pte++, addr += PAGE_SIZE, addr != end);
1198

K
KAMEZAWA Hiroyuki 已提交
1199
	add_mm_rss_vec(mm, rss);
1200
	arch_leave_lazy_mmu_mode();
1201

1202
	/* Do the actual TLB flush before dropping ptl */
1203
	if (force_flush)
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
		tlb_flush_mmu_tlbonly(tlb);
	pte_unmap_unlock(start_pte, ptl);

	/*
	 * If we forced a TLB flush (either due to running out of
	 * batch buffers or because we needed to flush dirty TLB
	 * entries before releasing the ptl), free the batched
	 * memory too. Restart if we didn't do everything.
	 */
	if (force_flush) {
		force_flush = 0;
		tlb_flush_mmu_free(tlb);
1216 1217 1218 1219 1220
		if (pending_page) {
			/* remove the page with new size */
			__tlb_remove_pte_page(tlb, pending_page);
			pending_page = NULL;
		}
1221
		if (addr != end)
P
Peter Zijlstra 已提交
1222 1223 1224
			goto again;
	}

1225
	return addr;
L
Linus Torvalds 已提交
1226 1227
}

1228
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1229
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1230
				unsigned long addr, unsigned long end,
1231
				struct zap_details *details)
L
Linus Torvalds 已提交
1232 1233 1234 1235 1236 1237 1238
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1239
		if (pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1240
			if (next - addr != HPAGE_PMD_SIZE) {
1241 1242
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1243
				split_huge_pmd(vma, pmd, addr);
S
Shaohua Li 已提交
1244
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1245
				goto next;
1246 1247
			/* fall through */
		}
1248 1249 1250 1251 1252 1253 1254 1255 1256
		/*
		 * 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;
1257
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1258
next:
1259 1260
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1261 1262

	return addr;
L
Linus Torvalds 已提交
1263 1264
}

1265
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1266
				struct vm_area_struct *vma, pgd_t *pgd,
L
Linus Torvalds 已提交
1267
				unsigned long addr, unsigned long end,
1268
				struct zap_details *details)
L
Linus Torvalds 已提交
1269 1270 1271 1272 1273 1274 1275
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
1276
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1277
			continue;
1278 1279
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
	} while (pud++, addr = next, addr != end);
1280 1281

	return addr;
L
Linus Torvalds 已提交
1282 1283
}

M
Michal Hocko 已提交
1284
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1285 1286 1287
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293 1294 1295 1296
{
	pgd_t *pgd;
	unsigned long next;

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

1304 1305 1306

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1307
		unsigned long end_addr,
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
		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;

1319 1320 1321
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1322
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1323
		untrack_pfn(vma, 0, 0);
1324 1325 1326 1327 1328 1329 1330

	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
1331
			 * cleanup path of mmap_region. When
1332
			 * hugetlbfs ->mmap method fails,
1333
			 * mmap_region() nullifies vma->vm_file
1334 1335 1336 1337
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1338
			if (vma->vm_file) {
1339
				i_mmap_lock_write(vma->vm_file->f_mapping);
1340
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1341
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1342
			}
1343 1344 1345
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1346 1347 1348 1349
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1350
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1351 1352 1353 1354
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1355
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
 *
 * 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
1366
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1367
		struct vm_area_struct *vma, unsigned long start_addr,
1368
		unsigned long end_addr)
L
Linus Torvalds 已提交
1369
{
A
Andrea Arcangeli 已提交
1370
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1371

A
Andrea Arcangeli 已提交
1372
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1373
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1374
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1375
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1376 1377 1378 1379 1380
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1381
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1382
 * @size: number of bytes to zap
1383
 * @details: details of shared cache invalidation
1384 1385
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1386
 */
1387
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
L
Linus Torvalds 已提交
1388 1389 1390
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1391
	struct mmu_gather tlb;
1392
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1393 1394

	lru_add_drain();
1395
	tlb_gather_mmu(&tlb, mm, start, end);
1396
	update_hiwater_rss(mm);
1397 1398
	mmu_notifier_invalidate_range_start(mm, start, end);
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next)
1399
		unmap_single_vma(&tlb, vma, start, end, details);
1400 1401
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1402 1403
}

1404 1405 1406 1407 1408
/**
 * 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
1409
 * @details: details of shared cache invalidation
1410 1411
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1412
 */
1413
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1414 1415 1416
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1417
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1418 1419 1420
	unsigned long end = address + size;

	lru_add_drain();
1421
	tlb_gather_mmu(&tlb, mm, address, end);
1422
	update_hiwater_rss(mm);
1423
	mmu_notifier_invalidate_range_start(mm, address, end);
1424
	unmap_single_vma(&tlb, vma, address, end, details);
1425
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1426
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1427 1428
}

1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
/**
 * 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;
1447
	zap_page_range_single(vma, address, size, NULL);
1448 1449 1450 1451
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1452
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1453
			spinlock_t **ptl)
1454 1455 1456 1457
{
	pgd_t * pgd = pgd_offset(mm, addr);
	pud_t * pud = pud_alloc(mm, pgd, addr);
	if (pud) {
1458
		pmd_t * pmd = pmd_alloc(mm, pud, addr);
1459 1460
		if (pmd) {
			VM_BUG_ON(pmd_trans_huge(*pmd));
1461
			return pte_alloc_map_lock(mm, pmd, addr, ptl);
1462
		}
1463 1464 1465 1466
	}
	return NULL;
}

1467 1468 1469 1470 1471 1472 1473
/*
 * 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
1474 1475
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1476
{
N
Nick Piggin 已提交
1477
	struct mm_struct *mm = vma->vm_mm;
1478
	int retval;
1479
	pte_t *pte;
1480 1481
	spinlock_t *ptl;

1482
	retval = -EINVAL;
1483
	if (PageAnon(page))
1484
		goto out;
1485 1486
	retval = -ENOMEM;
	flush_dcache_page(page);
1487
	pte = get_locked_pte(mm, addr, &ptl);
1488
	if (!pte)
1489
		goto out;
1490 1491 1492 1493 1494 1495
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1496
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1497
	page_add_file_rmap(page, false);
1498 1499 1500
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1501 1502
	pte_unmap_unlock(pte, ptl);
	return retval;
1503 1504 1505 1506 1507 1508
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1509 1510 1511 1512 1513 1514
/**
 * 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
 *
1515 1516 1517 1518 1519 1520
 * 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
1521
 * (see split_page()).
1522 1523 1524 1525 1526 1527 1528 1529
 *
 * 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.
1530 1531 1532 1533 1534
 *
 * 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.
1535
 */
N
Nick Piggin 已提交
1536 1537
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1538 1539 1540 1541 1542
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1543 1544 1545 1546 1547
	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
1548
	return insert_page(vma, addr, page, vma->vm_page_prot);
1549
}
1550
EXPORT_SYMBOL(vm_insert_page);
1551

N
Nick Piggin 已提交
1552
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1553
			pfn_t pfn, pgprot_t prot)
N
Nick Piggin 已提交
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
{
	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.. */
1569 1570 1571 1572
	if (pfn_t_devmap(pfn))
		entry = pte_mkdevmap(pfn_t_pte(pfn, prot));
	else
		entry = pte_mkspecial(pfn_t_pte(pfn, prot));
N
Nick Piggin 已提交
1573
	set_pte_at(mm, addr, pte, entry);
1574
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1575 1576 1577 1578 1579 1580 1581 1582

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

N
Nick Piggin 已提交
1583 1584 1585 1586 1587 1588
/**
 * 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
 *
1589
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1590 1591 1592 1593
 * 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
1594 1595 1596 1597 1598
 *
 * 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
1599 1600
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1601
			unsigned long pfn)
A
Andy Lutomirski 已提交
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
{
	return vm_insert_pfn_prot(vma, addr, pfn, vma->vm_page_prot);
}
EXPORT_SYMBOL(vm_insert_pfn);

/**
 * vm_insert_pfn_prot - insert single pfn into user vma with specified pgprot
 * @vma: user vma to map to
 * @addr: target user address of this page
 * @pfn: source kernel pfn
 * @pgprot: pgprot flags for the inserted page
 *
 * This is exactly like vm_insert_pfn, except that it allows drivers to
 * to override pgprot on a per-page basis.
 *
 * This only makes sense for IO mappings, and it makes no sense for
 * cow mappings.  In general, using multiple vmas is preferable;
 * vm_insert_pfn_prot should only be used if using multiple VMAs is
 * impractical.
 */
int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn, pgprot_t pgprot)
N
Nick Piggin 已提交
1624
{
1625
	int ret;
N
Nick Piggin 已提交
1626 1627 1628 1629 1630 1631
	/*
	 * 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
1632 1633 1634 1635 1636
	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
1637

N
Nick Piggin 已提交
1638 1639
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1640
	if (track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV)))
1641 1642
		return -EINVAL;

1643
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot);
1644 1645

	return ret;
N
Nick Piggin 已提交
1646
}
A
Andy Lutomirski 已提交
1647
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1648

N
Nick Piggin 已提交
1649
int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1650
			pfn_t pfn)
N
Nick Piggin 已提交
1651 1652
{
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1653

N
Nick Piggin 已提交
1654 1655
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
N
Nick Piggin 已提交
1656

N
Nick Piggin 已提交
1657 1658 1659 1660
	/*
	 * If we don't have pte special, then we have to use the pfn_valid()
	 * based VM_MIXEDMAP scheme (see vm_normal_page), and thus we *must*
	 * refcount the page if pfn_valid is true (hence insert_page rather
H
Hugh Dickins 已提交
1661 1662
	 * than insert_pfn).  If a zero_pfn were inserted into a VM_MIXEDMAP
	 * without pte special, it would there be refcounted as a normal page.
N
Nick Piggin 已提交
1663
	 */
1664
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1665 1666
		struct page *page;

1667 1668 1669 1670 1671 1672
		/*
		 * At this point we are committed to insert_page()
		 * regardless of whether the caller specified flags that
		 * result in pfn_t_has_page() == false.
		 */
		page = pfn_to_page(pfn_t_to_pfn(pfn));
N
Nick Piggin 已提交
1673 1674 1675
		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
1676
}
N
Nick Piggin 已提交
1677
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1678

L
Linus Torvalds 已提交
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
/*
 * maps a range of physical memory into the requested pages. the old
 * mappings are removed. any references to nonexistent pages results
 * in null mappings (currently treated as "copy-on-access")
 */
static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pte_t *pte;
H
Hugh Dickins 已提交
1689
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1690

H
Hugh Dickins 已提交
1691
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1692 1693
	if (!pte)
		return -ENOMEM;
1694
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1695 1696
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1697
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1698 1699
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1700
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1701
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715
	return 0;
}

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

	pfn -= addr >> PAGE_SHIFT;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
1716
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745
	do {
		next = pmd_addr_end(addr, end);
		if (remap_pte_range(mm, pmd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot))
			return -ENOMEM;
	} while (pmd++, addr = next, addr != end);
	return 0;
}

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

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

1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
/**
 * remap_pfn_range - remap kernel memory to userspace
 * @vma: user vma to map to
 * @addr: target user address to start at
 * @pfn: physical address of kernel memory
 * @size: size of map area
 * @prot: page protection flags for this mapping
 *
 *  Note: this is only safe if the mm semaphore is held when called.
 */
L
Linus Torvalds 已提交
1756 1757 1758 1759 1760
int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
		    unsigned long pfn, unsigned long size, pgprot_t prot)
{
	pgd_t *pgd;
	unsigned long next;
1761
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1762
	struct mm_struct *mm = vma->vm_mm;
1763
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1764 1765 1766 1767 1768 1769 1770
	int err;

	/*
	 * Physically remapped pages are special. Tell the
	 * rest of the world about it:
	 *   VM_IO tells people not to look at these pages
	 *	(accesses can have side effects).
1771 1772 1773
	 *   VM_PFNMAP tells the core MM that the base pages are just
	 *	raw PFN mappings, and do not have a "struct page" associated
	 *	with them.
1774 1775 1776 1777
	 *   VM_DONTEXPAND
	 *      Disable vma merging and expanding with mremap().
	 *   VM_DONTDUMP
	 *      Omit vma from core dump, even when VM_IO turned off.
L
Linus Torvalds 已提交
1778 1779 1780 1781
	 *
	 * There's a horrible special case to handle copy-on-write
	 * behaviour that some programs depend on. We mark the "original"
	 * un-COW'ed pages by matching them up with "vma->vm_pgoff".
1782
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1783
	 */
1784 1785 1786
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1787
		vma->vm_pgoff = pfn;
1788 1789
	}

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

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

	BUG_ON(addr >= end);
	pfn -= addr >> PAGE_SHIFT;
	pgd = pgd_offset(mm, addr);
	flush_cache_range(vma, addr, end);
	do {
		next = pgd_addr_end(addr, end);
		err = remap_pud_range(mm, pgd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1807 1808

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

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

1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861
/**
 * vm_iomap_memory - remap memory to userspace
 * @vma: user vma to map to
 * @start: start of area
 * @len: size of area
 *
 * This is a simplified io_remap_pfn_range() for common driver use. The
 * driver just needs to give us the physical memory range to be mapped,
 * we'll figure out the rest from the vma information.
 *
 * NOTE! Some drivers might want to tweak vma->vm_page_prot first to get
 * whatever write-combining details or similar.
 */
int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
{
	unsigned long vm_len, pfn, pages;

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

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

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

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

1862 1863 1864 1865 1866 1867
static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pte_t *pte;
	int err;
1868
	pgtable_t token;
1869
	spinlock_t *uninitialized_var(ptl);
1870 1871 1872 1873 1874 1875 1876 1877 1878

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

	BUG_ON(pmd_huge(*pmd));

1879 1880
	arch_enter_lazy_mmu_mode();

1881
	token = pmd_pgtable(*pmd);
1882 1883

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

1889 1890
	arch_leave_lazy_mmu_mode();

1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903
	if (mm != &init_mm)
		pte_unmap_unlock(pte-1, ptl);
	return err;
}

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

A
Andi Kleen 已提交
1904 1905
	BUG_ON(pud_huge(*pud));

1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
	do {
		next = pmd_addr_end(addr, end);
		err = apply_to_pte_range(mm, pmd, addr, next, fn, data);
		if (err)
			break;
	} while (pmd++, addr = next, addr != end);
	return err;
}

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

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

/*
 * Scan a region of virtual memory, filling in page tables as necessary
 * and calling a provided function on each leaf page table.
 */
int apply_to_page_range(struct mm_struct *mm, unsigned long addr,
			unsigned long size, pte_fn_t fn, void *data)
{
	pgd_t *pgd;
	unsigned long next;
1947
	unsigned long end = addr + size;
1948 1949
	int err;

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

1953 1954 1955 1956 1957 1958 1959
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
		err = apply_to_pud_range(mm, pgd, addr, next, fn, data);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1960

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

1965
/*
1966 1967 1968 1969 1970
 * handle_pte_fault chooses page fault handler according to an entry which was
 * read non-atomically.  Before making any commitment, on those architectures
 * or configurations (e.g. i386 with PAE) which might give a mix of unmatched
 * parts, do_swap_page must check under lock before unmapping the pte and
 * proceeding (but do_wp_page is only called after already making such a check;
1971
 * and do_anonymous_page can safely check later on).
1972
 */
H
Hugh Dickins 已提交
1973
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
1974 1975 1976 1977 1978
				pte_t *page_table, pte_t orig_pte)
{
	int same = 1;
#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
	if (sizeof(pte_t) > sizeof(unsigned long)) {
H
Hugh Dickins 已提交
1979 1980
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
1981
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
1982
		spin_unlock(ptl);
1983 1984 1985 1986 1987 1988
	}
#endif
	pte_unmap(page_table);
	return same;
}

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

1993 1994 1995 1996 1997 1998 1999
	/*
	 * If the source page was a PFN mapping, we don't have
	 * a "struct page" for it. We do a best-effort copy by
	 * just copying from the original user address. If that
	 * fails, we just zero-fill it. Live with it.
	 */
	if (unlikely(!src)) {
2000
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2001 2002 2003 2004 2005 2006 2007 2008 2009
		void __user *uaddr = (void __user *)(va & PAGE_MASK);

		/*
		 * This really shouldn't fail, because the page is there
		 * in the page tables. But it might just be unreadable,
		 * in which case we just give up and fill the result with
		 * zeroes.
		 */
		if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE))
2010
			clear_page(kaddr);
2011
		kunmap_atomic(kaddr);
2012
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2013 2014
	} else
		copy_user_highpage(dst, src, va, vma);
2015 2016
}

2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
static gfp_t __get_fault_gfp_mask(struct vm_area_struct *vma)
{
	struct file *vm_file = vma->vm_file;

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

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

2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
/*
 * Notify the address space that the page is about to become writable so that
 * it can prohibit this or wait for the page to get into an appropriate state.
 *
 * We do this without the lock held, so that it can sleep if it needs to.
 */
static int do_page_mkwrite(struct vm_area_struct *vma, struct page *page,
	       unsigned long address)
{
	struct vm_fault vmf;
	int ret;

	vmf.virtual_address = (void __user *)(address & PAGE_MASK);
	vmf.pgoff = page->index;
	vmf.flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2046
	vmf.gfp_mask = __get_fault_gfp_mask(vma);
2047
	vmf.page = page;
2048
	vmf.cow_page = NULL;
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064

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

2065 2066 2067 2068 2069 2070 2071 2072
/*
 * Handle write page faults for pages that can be reused in the current vma
 *
 * This can happen either due to the mapping being with the VM_SHARED flag,
 * or due to us being the last reference standing to the page. In either
 * case, all we need to do here is to mark the page as writable and update
 * any related book-keeping.
 */
K
Kirill A. Shutemov 已提交
2073 2074 2075
static inline int wp_page_reuse(struct fault_env *fe, pte_t orig_pte,
			struct page *page, int page_mkwrite, int dirty_shared)
	__releases(fe->ptl)
2076
{
K
Kirill A. Shutemov 已提交
2077
	struct vm_area_struct *vma = fe->vma;
2078 2079 2080 2081 2082 2083 2084 2085 2086
	pte_t entry;
	/*
	 * Clear the pages cpupid information as the existing
	 * information potentially belongs to a now completely
	 * unrelated process.
	 */
	if (page)
		page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);

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

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

		if (!page_mkwrite)
			lock_page(page);

		dirtied = set_page_dirty(page);
		VM_BUG_ON_PAGE(PageAnon(page), page);
		mapping = page->mapping;
		unlock_page(page);
2105
		put_page(page);
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121

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

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

	return VM_FAULT_WRITE;
}

2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
/*
 * Handle the case of a page which we actually need to copy to a new page.
 *
 * Called with mmap_sem locked and the old page referenced, but
 * without the ptl held.
 *
 * High level logic flow:
 *
 * - Allocate a page, copy the content of the old page to the new one.
 * - Handle book keeping and accounting - cgroups, mmu-notifiers, etc.
 * - Take the PTL. If the pte changed, bail out and release the allocated page
 * - If the pte is still the way we remember it, update the page table and all
 *   relevant references. This includes dropping the reference the page-table
 *   held to the old page, as well as updating the rmap.
 * - In any case, unlock the PTL and drop the reference we took to the old page.
 */
K
Kirill A. Shutemov 已提交
2138 2139
static int wp_page_copy(struct fault_env *fe, pte_t orig_pte,
		struct page *old_page)
2140
{
K
Kirill A. Shutemov 已提交
2141 2142
	struct vm_area_struct *vma = fe->vma;
	struct mm_struct *mm = vma->vm_mm;
2143 2144 2145
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
K
Kirill A. Shutemov 已提交
2146 2147
	const unsigned long mmun_start = fe->address & PAGE_MASK;
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2148 2149 2150 2151 2152 2153
	struct mem_cgroup *memcg;

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

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

2165
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2166 2167
		goto oom_free_new;

2168 2169
	__SetPageUptodate(new_page);

2170 2171 2172 2173 2174
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
K
Kirill A. Shutemov 已提交
2175 2176
	fe->pte = pte_offset_map_lock(mm, fe->pmd, fe->address, &fe->ptl);
	if (likely(pte_same(*fe->pte, orig_pte))) {
2177 2178
		if (old_page) {
			if (!PageAnon(old_page)) {
2179 2180
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2181 2182 2183 2184 2185
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
K
Kirill A. Shutemov 已提交
2186
		flush_cache_page(vma, fe->address, pte_pfn(orig_pte));
2187 2188 2189 2190 2191 2192 2193 2194
		entry = mk_pte(new_page, vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
		/*
		 * Clear the pte entry and flush it first, before updating the
		 * pte with the new entry. This will avoid a race condition
		 * seen in the presence of one thread doing SMC and another
		 * thread doing COW.
		 */
K
Kirill A. Shutemov 已提交
2195 2196
		ptep_clear_flush_notify(vma, fe->address, fe->pte);
		page_add_new_anon_rmap(new_page, vma, fe->address, false);
2197
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2198 2199 2200 2201 2202 2203
		lru_cache_add_active_or_unevictable(new_page, vma);
		/*
		 * We call the notify macro here because, when using secondary
		 * mmu page tables (such as kvm shadow page tables), we want the
		 * new page to be mapped directly into the secondary page table.
		 */
K
Kirill A. Shutemov 已提交
2204 2205
		set_pte_at_notify(mm, fe->address, fe->pte, entry);
		update_mmu_cache(vma, fe->address, fe->pte);
2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
		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.
			 */
2229
			page_remove_rmap(old_page, false);
2230 2231 2232 2233 2234 2235
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2236
		mem_cgroup_cancel_charge(new_page, memcg, false);
2237 2238 2239
	}

	if (new_page)
2240
		put_page(new_page);
2241

K
Kirill A. Shutemov 已提交
2242
	pte_unmap_unlock(fe->pte, fe->ptl);
2243 2244 2245 2246 2247 2248 2249 2250
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
	if (old_page) {
		/*
		 * Don't let another task, with possibly unlocked vma,
		 * keep the mlocked page.
		 */
		if (page_copied && (vma->vm_flags & VM_LOCKED)) {
			lock_page(old_page);	/* LRU manipulation */
2251 2252
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2253 2254
			unlock_page(old_page);
		}
2255
		put_page(old_page);
2256 2257 2258
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2259
	put_page(new_page);
2260 2261
oom:
	if (old_page)
2262
		put_page(old_page);
2263 2264 2265
	return VM_FAULT_OOM;
}

2266 2267 2268 2269
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
K
Kirill A. Shutemov 已提交
2270
static int wp_pfn_shared(struct fault_env *fe,  pte_t orig_pte)
2271
{
K
Kirill A. Shutemov 已提交
2272 2273
	struct vm_area_struct *vma = fe->vma;

2274 2275 2276
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
		struct vm_fault vmf = {
			.page = NULL,
K
Kirill A. Shutemov 已提交
2277 2278 2279
			.pgoff = linear_page_index(vma, fe->address),
			.virtual_address =
				(void __user *)(fe->address & PAGE_MASK),
2280 2281 2282 2283
			.flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE,
		};
		int ret;

K
Kirill A. Shutemov 已提交
2284
		pte_unmap_unlock(fe->pte, fe->ptl);
2285 2286 2287
		ret = vma->vm_ops->pfn_mkwrite(vma, &vmf);
		if (ret & VM_FAULT_ERROR)
			return ret;
K
Kirill A. Shutemov 已提交
2288 2289
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
				&fe->ptl);
2290 2291 2292 2293
		/*
		 * We might have raced with another page fault while we
		 * released the pte_offset_map_lock.
		 */
K
Kirill A. Shutemov 已提交
2294 2295
		if (!pte_same(*fe->pte, orig_pte)) {
			pte_unmap_unlock(fe->pte, fe->ptl);
2296 2297 2298
			return 0;
		}
	}
K
Kirill A. Shutemov 已提交
2299
	return wp_page_reuse(fe, orig_pte, NULL, 0, 0);
2300 2301
}

K
Kirill A. Shutemov 已提交
2302 2303 2304
static int wp_page_shared(struct fault_env *fe, pte_t orig_pte,
		struct page *old_page)
	__releases(fe->ptl)
2305
{
K
Kirill A. Shutemov 已提交
2306
	struct vm_area_struct *vma = fe->vma;
2307 2308
	int page_mkwrite = 0;

2309
	get_page(old_page);
2310 2311 2312 2313

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

K
Kirill A. Shutemov 已提交
2314 2315
		pte_unmap_unlock(fe->pte, fe->ptl);
		tmp = do_page_mkwrite(vma, old_page, fe->address);
2316 2317
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
2318
			put_page(old_page);
2319 2320 2321 2322 2323 2324 2325 2326
			return tmp;
		}
		/*
		 * Since we dropped the lock we need to revalidate
		 * the PTE as someone else may have changed it.  If
		 * they did, we just return, as we can count on the
		 * MMU to tell us if they didn't also make it writable.
		 */
K
Kirill A. Shutemov 已提交
2327 2328 2329
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
						 &fe->ptl);
		if (!pte_same(*fe->pte, orig_pte)) {
2330
			unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2331
			pte_unmap_unlock(fe->pte, fe->ptl);
2332
			put_page(old_page);
2333 2334 2335 2336 2337
			return 0;
		}
		page_mkwrite = 1;
	}

K
Kirill A. Shutemov 已提交
2338
	return wp_page_reuse(fe, orig_pte, old_page, page_mkwrite, 1);
2339 2340
}

L
Linus Torvalds 已提交
2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354
/*
 * 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.
 *
2355 2356 2357
 * 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
2358
 */
K
Kirill A. Shutemov 已提交
2359 2360
static int do_wp_page(struct fault_env *fe, pte_t orig_pte)
	__releases(fe->ptl)
L
Linus Torvalds 已提交
2361
{
K
Kirill A. Shutemov 已提交
2362
	struct vm_area_struct *vma = fe->vma;
2363
	struct page *old_page;
L
Linus Torvalds 已提交
2364

K
Kirill A. Shutemov 已提交
2365
	old_page = vm_normal_page(vma, fe->address, orig_pte);
2366 2367
	if (!old_page) {
		/*
2368 2369
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2370 2371
		 *
		 * We should not cow pages in a shared writeable mapping.
2372
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2373 2374 2375
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
K
Kirill A. Shutemov 已提交
2376
			return wp_pfn_shared(fe, orig_pte);
2377

K
Kirill A. Shutemov 已提交
2378 2379
		pte_unmap_unlock(fe->pte, fe->ptl);
		return wp_page_copy(fe, orig_pte, old_page);
2380
	}
L
Linus Torvalds 已提交
2381

2382
	/*
P
Peter Zijlstra 已提交
2383 2384
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2385
	 */
H
Hugh Dickins 已提交
2386
	if (PageAnon(old_page) && !PageKsm(old_page)) {
2387
		int total_mapcount;
2388
		if (!trylock_page(old_page)) {
2389
			get_page(old_page);
K
Kirill A. Shutemov 已提交
2390
			pte_unmap_unlock(fe->pte, fe->ptl);
2391
			lock_page(old_page);
K
Kirill A. Shutemov 已提交
2392 2393 2394
			fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd,
					fe->address, &fe->ptl);
			if (!pte_same(*fe->pte, orig_pte)) {
2395
				unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2396
				pte_unmap_unlock(fe->pte, fe->ptl);
2397
				put_page(old_page);
2398
				return 0;
2399
			}
2400
			put_page(old_page);
P
Peter Zijlstra 已提交
2401
		}
2402 2403 2404 2405 2406 2407 2408 2409 2410
		if (reuse_swap_page(old_page, &total_mapcount)) {
			if (total_mapcount == 1) {
				/*
				 * The page is all ours. Move it to
				 * our anon_vma so the rmap code will
				 * not search our parent or siblings.
				 * Protected against the rmap code by
				 * the page lock.
				 */
2411
				page_move_anon_rmap(old_page, vma);
2412
			}
2413
			unlock_page(old_page);
K
Kirill A. Shutemov 已提交
2414
			return wp_page_reuse(fe, orig_pte, old_page, 0, 0);
2415
		}
2416
		unlock_page(old_page);
P
Peter Zijlstra 已提交
2417
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2418
					(VM_WRITE|VM_SHARED))) {
K
Kirill A. Shutemov 已提交
2419
		return wp_page_shared(fe, orig_pte, old_page);
L
Linus Torvalds 已提交
2420 2421 2422 2423 2424
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
2425
	get_page(old_page);
2426

K
Kirill A. Shutemov 已提交
2427 2428
	pte_unmap_unlock(fe->pte, fe->ptl);
	return wp_page_copy(fe, orig_pte, old_page);
L
Linus Torvalds 已提交
2429 2430
}

2431
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2432 2433 2434
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2435
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2436 2437
}

2438
static inline void unmap_mapping_range_tree(struct rb_root *root,
L
Linus Torvalds 已提交
2439 2440 2441 2442 2443
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2444
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2445 2446 2447
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2448
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2449 2450 2451 2452 2453 2454 2455
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2456
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2457 2458
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2459
				details);
L
Linus Torvalds 已提交
2460 2461 2462 2463
	}
}

/**
2464 2465 2466 2467
 * 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
2468
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2469 2470
 * @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
2471
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482
 * must keep the partial page.  In contrast, we must get rid of
 * partial pages.
 * @holelen: size of prospective hole in bytes.  This will be rounded
 * up to a PAGE_SIZE boundary.  A holelen of zero truncates to the
 * end of the file.
 * @even_cows: 1 when truncating a file, unmap even private COWed pages;
 * but 0 when invalidating pagecache, don't throw away private data.
 */
void unmap_mapping_range(struct address_space *mapping,
		loff_t const holebegin, loff_t const holelen, int even_cows)
{
M
Michal Hocko 已提交
2483
	struct zap_details details = { };
L
Linus Torvalds 已提交
2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
	pgoff_t hba = holebegin >> PAGE_SHIFT;
	pgoff_t hlen = (holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;

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

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

2501
	i_mmap_lock_write(mapping);
2502
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap)))
L
Linus Torvalds 已提交
2503
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2504
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2505 2506 2507 2508
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2509 2510
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2511 2512 2513 2514
 * We return with pte unmapped and unlocked.
 *
 * We return with the mmap_sem locked or unlocked in the same cases
 * as does filemap_fault().
L
Linus Torvalds 已提交
2515
 */
K
Kirill A. Shutemov 已提交
2516
int do_swap_page(struct fault_env *fe, pte_t orig_pte)
L
Linus Torvalds 已提交
2517
{
K
Kirill A. Shutemov 已提交
2518
	struct vm_area_struct *vma = fe->vma;
2519
	struct page *page, *swapcache;
2520
	struct mem_cgroup *memcg;
2521
	swp_entry_t entry;
L
Linus Torvalds 已提交
2522
	pte_t pte;
2523
	int locked;
2524
	int exclusive = 0;
N
Nick Piggin 已提交
2525
	int ret = 0;
L
Linus Torvalds 已提交
2526

K
Kirill A. Shutemov 已提交
2527
	if (!pte_unmap_same(vma->vm_mm, fe->pmd, fe->pte, orig_pte))
2528
		goto out;
2529 2530

	entry = pte_to_swp_entry(orig_pte);
2531 2532
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
K
Kirill A. Shutemov 已提交
2533
			migration_entry_wait(vma->vm_mm, fe->pmd, fe->address);
2534 2535 2536
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
K
Kirill A. Shutemov 已提交
2537
			print_bad_pte(vma, fe->address, orig_pte, NULL);
H
Hugh Dickins 已提交
2538
			ret = VM_FAULT_SIGBUS;
2539
		}
2540 2541
		goto out;
	}
2542
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
L
Linus Torvalds 已提交
2543 2544
	page = lookup_swap_cache(entry);
	if (!page) {
2545
		page = swapin_readahead(entry,
K
Kirill A. Shutemov 已提交
2546
					GFP_HIGHUSER_MOVABLE, vma, fe->address);
L
Linus Torvalds 已提交
2547 2548
		if (!page) {
			/*
2549 2550
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2551
			 */
K
Kirill A. Shutemov 已提交
2552 2553 2554
			fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd,
					fe->address, &fe->ptl);
			if (likely(pte_same(*fe->pte, orig_pte)))
L
Linus Torvalds 已提交
2555
				ret = VM_FAULT_OOM;
2556
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2557
			goto unlock;
L
Linus Torvalds 已提交
2558 2559 2560 2561
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2562
		count_vm_event(PGMAJFAULT);
K
Kirill A. Shutemov 已提交
2563
		mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
2564
	} else if (PageHWPoison(page)) {
2565 2566 2567 2568
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2569 2570
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2571
		swapcache = page;
2572
		goto out_release;
L
Linus Torvalds 已提交
2573 2574
	}

2575
	swapcache = page;
K
Kirill A. Shutemov 已提交
2576
	locked = lock_page_or_retry(page, vma->vm_mm, fe->flags);
R
Rik van Riel 已提交
2577

2578
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2579 2580 2581 2582
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2583

A
Andrea Arcangeli 已提交
2584
	/*
2585 2586 2587 2588
	 * 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
2589
	 */
2590
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2591 2592
		goto out_page;

K
Kirill A. Shutemov 已提交
2593
	page = ksm_might_need_to_copy(page, vma, fe->address);
2594 2595 2596 2597
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2598 2599
	}

K
Kirill A. Shutemov 已提交
2600 2601
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2602
		ret = VM_FAULT_OOM;
2603
		goto out_page;
2604 2605
	}

L
Linus Torvalds 已提交
2606
	/*
2607
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2608
	 */
K
Kirill A. Shutemov 已提交
2609 2610 2611
	fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
			&fe->ptl);
	if (unlikely(!pte_same(*fe->pte, orig_pte)))
2612 2613 2614 2615 2616
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2617 2618
	}

2619 2620 2621 2622 2623 2624 2625 2626 2627
	/*
	 * The page isn't present yet, go ahead with the fault.
	 *
	 * Be careful about the sequence of operations here.
	 * To get its accounting right, reuse_swap_page() must be called
	 * while the page is counted on swap but not yet in mapcount i.e.
	 * before page_add_anon_rmap() and swap_free(); try_to_free_swap()
	 * must be called after the swap_free(), or it will never succeed.
	 */
L
Linus Torvalds 已提交
2628

K
Kirill A. Shutemov 已提交
2629 2630
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2631
	pte = mk_pte(page, vma->vm_page_prot);
K
Kirill A. Shutemov 已提交
2632
	if ((fe->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2633
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
K
Kirill A. Shutemov 已提交
2634
		fe->flags &= ~FAULT_FLAG_WRITE;
2635
		ret |= VM_FAULT_WRITE;
2636
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2637 2638
	}
	flush_icache_page(vma, page);
2639 2640
	if (pte_swp_soft_dirty(orig_pte))
		pte = pte_mksoft_dirty(pte);
K
Kirill A. Shutemov 已提交
2641
	set_pte_at(vma->vm_mm, fe->address, fe->pte, pte);
2642
	if (page == swapcache) {
K
Kirill A. Shutemov 已提交
2643
		do_page_add_anon_rmap(page, vma, fe->address, exclusive);
2644
		mem_cgroup_commit_charge(page, memcg, true, false);
2645
	} else { /* ksm created a completely new copy */
K
Kirill A. Shutemov 已提交
2646
		page_add_new_anon_rmap(page, vma, fe->address, false);
2647
		mem_cgroup_commit_charge(page, memcg, false, false);
2648 2649
		lru_cache_add_active_or_unevictable(page, vma);
	}
L
Linus Torvalds 已提交
2650

2651
	swap_free(entry);
2652 2653
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2654
		try_to_free_swap(page);
2655
	unlock_page(page);
2656
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2657 2658 2659 2660 2661 2662 2663 2664 2665
		/*
		 * 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);
2666
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2667
	}
2668

K
Kirill A. Shutemov 已提交
2669 2670
	if (fe->flags & FAULT_FLAG_WRITE) {
		ret |= do_wp_page(fe, pte);
2671 2672
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2673 2674 2675 2676
		goto out;
	}

	/* No need to invalidate - it was non-present before */
K
Kirill A. Shutemov 已提交
2677
	update_mmu_cache(vma, fe->address, fe->pte);
2678
unlock:
K
Kirill A. Shutemov 已提交
2679
	pte_unmap_unlock(fe->pte, fe->ptl);
L
Linus Torvalds 已提交
2680 2681
out:
	return ret;
2682
out_nomap:
2683
	mem_cgroup_cancel_charge(page, memcg, false);
K
Kirill A. Shutemov 已提交
2684
	pte_unmap_unlock(fe->pte, fe->ptl);
2685
out_page:
2686
	unlock_page(page);
2687
out_release:
2688
	put_page(page);
2689
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
2690
		unlock_page(swapcache);
2691
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2692
	}
2693
	return ret;
L
Linus Torvalds 已提交
2694 2695
}

2696
/*
2697 2698
 * This is like a special single-page "expand_{down|up}wards()",
 * except we must first make sure that 'address{-|+}PAGE_SIZE'
2699 2700 2701 2702 2703 2704
 * 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) {
2705 2706 2707 2708 2709 2710 2711 2712 2713 2714
		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;
2715

2716
		return expand_downwards(vma, address - PAGE_SIZE);
2717
	}
2718 2719 2720 2721 2722 2723 2724
	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;

2725
		return expand_upwards(vma, address + PAGE_SIZE);
2726
	}
2727 2728 2729
	return 0;
}

L
Linus Torvalds 已提交
2730
/*
2731 2732 2733
 * 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
2734
 */
K
Kirill A. Shutemov 已提交
2735
static int do_anonymous_page(struct fault_env *fe)
L
Linus Torvalds 已提交
2736
{
K
Kirill A. Shutemov 已提交
2737
	struct vm_area_struct *vma = fe->vma;
2738
	struct mem_cgroup *memcg;
2739
	struct page *page;
L
Linus Torvalds 已提交
2740 2741
	pte_t entry;

2742 2743 2744 2745
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2746
	/* Check if we need to add a guard page to the stack */
K
Kirill A. Shutemov 已提交
2747
	if (check_stack_guard_page(vma, fe->address) < 0)
2748
		return VM_FAULT_SIGSEGV;
2749

2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
	/*
	 * Use pte_alloc() instead of pte_alloc_map().  We can't run
	 * pte_offset_map() on pmds where a huge pmd might be created
	 * from a different thread.
	 *
	 * pte_alloc_map() is safe to use under down_write(mmap_sem) or when
	 * parallel threads are excluded by other means.
	 *
	 * Here we only have down_read(mmap_sem).
	 */
	if (pte_alloc(vma->vm_mm, fe->pmd, fe->address))
		return VM_FAULT_OOM;

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

2767
	/* Use the zero-page for reads */
K
Kirill A. Shutemov 已提交
2768 2769 2770
	if (!(fe->flags & FAULT_FLAG_WRITE) &&
			!mm_forbids_zeropage(vma->vm_mm)) {
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(fe->address),
H
Hugh Dickins 已提交
2771
						vma->vm_page_prot));
K
Kirill A. Shutemov 已提交
2772 2773 2774
		fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
				&fe->ptl);
		if (!pte_none(*fe->pte))
H
Hugh Dickins 已提交
2775
			goto unlock;
2776 2777
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
K
Kirill A. Shutemov 已提交
2778 2779
			pte_unmap_unlock(fe->pte, fe->ptl);
			return handle_userfault(fe, VM_UFFD_MISSING);
2780
		}
H
Hugh Dickins 已提交
2781 2782 2783
		goto setpte;
	}

N
Nick Piggin 已提交
2784 2785 2786
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
K
Kirill A. Shutemov 已提交
2787
	page = alloc_zeroed_user_highpage_movable(vma, fe->address);
N
Nick Piggin 已提交
2788 2789
	if (!page)
		goto oom;
2790

K
Kirill A. Shutemov 已提交
2791
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
2792 2793
		goto oom_free_page;

2794 2795 2796 2797 2798
	/*
	 * 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
2799
	__SetPageUptodate(page);
2800

N
Nick Piggin 已提交
2801
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2802 2803
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2804

K
Kirill A. Shutemov 已提交
2805 2806 2807
	fe->pte = pte_offset_map_lock(vma->vm_mm, fe->pmd, fe->address,
			&fe->ptl);
	if (!pte_none(*fe->pte))
N
Nick Piggin 已提交
2808
		goto release;
H
Hugh Dickins 已提交
2809

2810 2811
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
K
Kirill A. Shutemov 已提交
2812
		pte_unmap_unlock(fe->pte, fe->ptl);
2813
		mem_cgroup_cancel_charge(page, memcg, false);
2814
		put_page(page);
K
Kirill A. Shutemov 已提交
2815
		return handle_userfault(fe, VM_UFFD_MISSING);
2816 2817
	}

K
Kirill A. Shutemov 已提交
2818 2819
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	page_add_new_anon_rmap(page, vma, fe->address, false);
2820
	mem_cgroup_commit_charge(page, memcg, false, false);
2821
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
2822
setpte:
K
Kirill A. Shutemov 已提交
2823
	set_pte_at(vma->vm_mm, fe->address, fe->pte, entry);
L
Linus Torvalds 已提交
2824 2825

	/* No need to invalidate - it was non-present before */
K
Kirill A. Shutemov 已提交
2826
	update_mmu_cache(vma, fe->address, fe->pte);
2827
unlock:
K
Kirill A. Shutemov 已提交
2828
	pte_unmap_unlock(fe->pte, fe->ptl);
N
Nick Piggin 已提交
2829
	return 0;
2830
release:
2831
	mem_cgroup_cancel_charge(page, memcg, false);
2832
	put_page(page);
2833
	goto unlock;
2834
oom_free_page:
2835
	put_page(page);
2836
oom:
L
Linus Torvalds 已提交
2837 2838 2839
	return VM_FAULT_OOM;
}

2840 2841 2842 2843 2844
/*
 * The mmap_sem must have been held on entry, and may have been
 * released depending on flags and vma->vm_ops->fault() return value.
 * See filemap_fault() and __lock_page_retry().
 */
K
Kirill A. Shutemov 已提交
2845 2846
static int __do_fault(struct fault_env *fe, pgoff_t pgoff,
		struct page *cow_page, struct page **page, void **entry)
2847
{
K
Kirill A. Shutemov 已提交
2848
	struct vm_area_struct *vma = fe->vma;
2849 2850 2851
	struct vm_fault vmf;
	int ret;

K
Kirill A. Shutemov 已提交
2852
	vmf.virtual_address = (void __user *)(fe->address & PAGE_MASK);
2853
	vmf.pgoff = pgoff;
K
Kirill A. Shutemov 已提交
2854
	vmf.flags = fe->flags;
2855
	vmf.page = NULL;
2856
	vmf.gfp_mask = __get_fault_gfp_mask(vma);
2857
	vmf.cow_page = cow_page;
2858 2859 2860 2861

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

	if (unlikely(PageHWPoison(vmf.page))) {
		if (ret & VM_FAULT_LOCKED)
			unlock_page(vmf.page);
2870
		put_page(vmf.page);
2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882
		return VM_FAULT_HWPOISON;
	}

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

	*page = vmf.page;
	return ret;
}

2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922
static int pte_alloc_one_map(struct fault_env *fe)
{
	struct vm_area_struct *vma = fe->vma;

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

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

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

K
Kirill A. Shutemov 已提交
2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

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

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

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

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

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

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

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

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

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

	/* fault is handled */
	ret = 0;
2971
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
out:
	spin_unlock(fe->ptl);
	return ret;
}
#else
static int do_set_pmd(struct fault_env *fe, struct page *page)
{
	BUILD_BUG();
	return 0;
}
#endif

2984
/**
2985 2986
 * alloc_set_pte - setup new PTE entry for given page and add reverse page
 * mapping. If needed, the fucntion allocates page table or use pre-allocated.
2987
 *
K
Kirill A. Shutemov 已提交
2988
 * @fe: fault environment
2989
 * @memcg: memcg to charge page (only for private mappings)
2990 2991
 * @page: page to map
 *
2992
 * Caller must take care of unlocking fe->ptl, if fe->pte is non-NULL on return.
2993 2994 2995 2996
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
2997 2998
int alloc_set_pte(struct fault_env *fe, struct mem_cgroup *memcg,
		struct page *page)
2999
{
K
Kirill A. Shutemov 已提交
3000 3001
	struct vm_area_struct *vma = fe->vma;
	bool write = fe->flags & FAULT_FLAG_WRITE;
3002
	pte_t entry;
K
Kirill A. Shutemov 已提交
3003 3004 3005 3006 3007 3008 3009 3010 3011 3012
	int ret;

	if (pmd_none(*fe->pmd) && PageTransCompound(page)) {
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

		ret = do_set_pmd(fe, page);
		if (ret != VM_FAULT_FALLBACK)
			return ret;
	}
3013

3014
	if (!fe->pte) {
K
Kirill A. Shutemov 已提交
3015
		ret = pte_alloc_one_map(fe);
3016 3017 3018 3019 3020 3021 3022 3023
		if (ret)
			return ret;
	}

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

3024 3025 3026 3027
	flush_icache_page(vma, page);
	entry = mk_pte(page, vma->vm_page_prot);
	if (write)
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3028 3029
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3030
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
K
Kirill A. Shutemov 已提交
3031
		page_add_new_anon_rmap(page, vma, fe->address, false);
3032 3033
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3034
	} else {
3035
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3036
		page_add_file_rmap(page, false);
3037
	}
K
Kirill A. Shutemov 已提交
3038
	set_pte_at(vma->vm_mm, fe->address, fe->pte, entry);
3039 3040

	/* no need to invalidate: a not-present page won't be cached */
K
Kirill A. Shutemov 已提交
3041
	update_mmu_cache(vma, fe->address, fe->pte);
3042 3043

	return 0;
3044 3045
}

3046 3047
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3048 3049 3050

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3051
{
3052
	*val = fault_around_bytes;
3053 3054 3055
	return 0;
}

3056 3057 3058 3059 3060
/*
 * fault_around_pages() and fault_around_mask() expects fault_around_bytes
 * rounded down to nearest page order. It's what do_fault_around() expects to
 * see.
 */
3061
static int fault_around_bytes_set(void *data, u64 val)
3062
{
3063
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3064
		return -EINVAL;
3065 3066 3067 3068
	if (val > PAGE_SIZE)
		fault_around_bytes = rounddown_pow_of_two(val);
	else
		fault_around_bytes = PAGE_SIZE; /* rounddown_pow_of_two(0) is undefined */
3069 3070
	return 0;
}
3071 3072
DEFINE_SIMPLE_ATTRIBUTE(fault_around_bytes_fops,
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3073 3074 3075 3076 3077

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

3078 3079
	ret = debugfs_create_file("fault_around_bytes", 0644, NULL, NULL,
			&fault_around_bytes_fops);
3080
	if (!ret)
3081
		pr_warn("Failed to create fault_around_bytes in debugfs");
3082 3083 3084 3085
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3086

3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
/*
 * do_fault_around() tries to map few pages around the fault address. The hope
 * is that the pages will be needed soon and this will lower the number of
 * faults to handle.
 *
 * It uses vm_ops->map_pages() to map the pages, which skips the page if it's
 * not ready to be mapped: not up-to-date, locked, etc.
 *
 * This function is called with the page table lock taken. In the split ptlock
 * case the page table lock only protects only those entries which belong to
 * the page table corresponding to the fault address.
 *
 * This function doesn't cross the VMA boundaries, in order to call map_pages()
 * only once.
 *
 * fault_around_pages() defines how many pages we'll try to map.
 * do_fault_around() expects it to return a power of two less than or equal to
 * PTRS_PER_PTE.
 *
 * The virtual address of the area that we map is naturally aligned to the
 * fault_around_pages() value (and therefore to page order).  This way it's
 * easier to guarantee that we don't cross page table boundaries.
 */
3110
static int do_fault_around(struct fault_env *fe, pgoff_t start_pgoff)
3111
{
3112
	unsigned long address = fe->address, nr_pages, mask;
K
Kirill A. Shutemov 已提交
3113
	pgoff_t end_pgoff;
3114
	int off, ret = 0;
3115

3116
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3117 3118
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

3119 3120
	fe->address = max(address & mask, fe->vma->vm_start);
	off = ((address - fe->address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
K
Kirill A. Shutemov 已提交
3121
	start_pgoff -= off;
3122 3123

	/*
K
Kirill A. Shutemov 已提交
3124 3125
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3126
	 */
K
Kirill A. Shutemov 已提交
3127
	end_pgoff = start_pgoff -
3128
		((fe->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3129
		PTRS_PER_PTE - 1;
K
Kirill A. Shutemov 已提交
3130 3131
	end_pgoff = min3(end_pgoff, vma_pages(fe->vma) + fe->vma->vm_pgoff - 1,
			start_pgoff + nr_pages - 1);
3132

3133 3134 3135
	if (pmd_none(*fe->pmd)) {
		fe->prealloc_pte = pte_alloc_one(fe->vma->vm_mm, fe->address);
		smp_wmb(); /* See comment in __pte_alloc() */
3136 3137
	}

K
Kirill A. Shutemov 已提交
3138
	fe->vma->vm_ops->map_pages(fe, start_pgoff, end_pgoff);
3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159

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

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

	/* check if the page fault is solved */
	fe->pte -= (fe->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*fe->pte))
		ret = VM_FAULT_NOPAGE;
	pte_unmap_unlock(fe->pte, fe->ptl);
K
Kirill A. Shutemov 已提交
3160 3161
out:
	fe->address = address;
3162 3163
	fe->pte = NULL;
	return ret;
3164 3165
}

3166
static int do_read_fault(struct fault_env *fe, pgoff_t pgoff)
3167
{
K
Kirill A. Shutemov 已提交
3168
	struct vm_area_struct *vma = fe->vma;
3169
	struct page *fault_page;
3170 3171 3172 3173 3174 3175 3176
	int ret = 0;

	/*
	 * Let's call ->map_pages() first and use ->fault() as fallback
	 * if page by the offset is not ready to be mapped (cold cache or
	 * something).
	 */
3177
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3178 3179 3180
		ret = do_fault_around(fe, pgoff);
		if (ret)
			return ret;
3181
	}
3182

K
Kirill A. Shutemov 已提交
3183
	ret = __do_fault(fe, pgoff, NULL, &fault_page, NULL);
3184 3185 3186
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3187 3188
	ret |= alloc_set_pte(fe, NULL, fault_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3189
		pte_unmap_unlock(fe->pte, fe->ptl);
3190
	unlock_page(fault_page);
3191 3192
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		put_page(fault_page);
3193 3194 3195
	return ret;
}

3196
static int do_cow_fault(struct fault_env *fe, pgoff_t pgoff)
3197
{
K
Kirill A. Shutemov 已提交
3198
	struct vm_area_struct *vma = fe->vma;
3199
	struct page *fault_page, *new_page;
3200
	void *fault_entry;
3201
	struct mem_cgroup *memcg;
3202 3203 3204 3205 3206
	int ret;

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

K
Kirill A. Shutemov 已提交
3207
	new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, fe->address);
3208 3209 3210
	if (!new_page)
		return VM_FAULT_OOM;

K
Kirill A. Shutemov 已提交
3211 3212
	if (mem_cgroup_try_charge(new_page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
3213
		put_page(new_page);
3214 3215 3216
		return VM_FAULT_OOM;
	}

K
Kirill A. Shutemov 已提交
3217
	ret = __do_fault(fe, pgoff, new_page, &fault_page, &fault_entry);
3218 3219 3220
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;

3221
	if (!(ret & VM_FAULT_DAX_LOCKED))
K
Kirill A. Shutemov 已提交
3222
		copy_user_highpage(new_page, fault_page, fe->address, vma);
3223 3224
	__SetPageUptodate(new_page);

3225 3226
	ret |= alloc_set_pte(fe, memcg, new_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3227
		pte_unmap_unlock(fe->pte, fe->ptl);
3228
	if (!(ret & VM_FAULT_DAX_LOCKED)) {
3229
		unlock_page(fault_page);
3230
		put_page(fault_page);
3231
	} else {
3232
		dax_unlock_mapping_entry(vma->vm_file->f_mapping, pgoff);
3233
	}
3234 3235
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3236 3237
	return ret;
uncharge_out:
3238
	mem_cgroup_cancel_charge(new_page, memcg, false);
3239
	put_page(new_page);
3240 3241 3242
	return ret;
}

3243
static int do_shared_fault(struct fault_env *fe, pgoff_t pgoff)
L
Linus Torvalds 已提交
3244
{
K
Kirill A. Shutemov 已提交
3245
	struct vm_area_struct *vma = fe->vma;
3246 3247 3248 3249
	struct page *fault_page;
	struct address_space *mapping;
	int dirtied = 0;
	int ret, tmp;
3250

K
Kirill A. Shutemov 已提交
3251
	ret = __do_fault(fe, pgoff, NULL, &fault_page, NULL);
3252
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3253
		return ret;
L
Linus Torvalds 已提交
3254 3255

	/*
3256 3257
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3258
	 */
3259 3260
	if (vma->vm_ops->page_mkwrite) {
		unlock_page(fault_page);
K
Kirill A. Shutemov 已提交
3261
		tmp = do_page_mkwrite(vma, fault_page, fe->address);
3262 3263
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
3264
			put_page(fault_page);
3265
			return tmp;
3266
		}
3267 3268
	}

3269 3270
	ret |= alloc_set_pte(fe, NULL, fault_page);
	if (fe->pte)
K
Kirill A. Shutemov 已提交
3271
		pte_unmap_unlock(fe->pte, fe->ptl);
3272 3273
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
3274
		unlock_page(fault_page);
3275
		put_page(fault_page);
3276
		return ret;
L
Linus Torvalds 已提交
3277
	}
N
Nick Piggin 已提交
3278

3279 3280
	if (set_page_dirty(fault_page))
		dirtied = 1;
3281 3282 3283 3284 3285 3286
	/*
	 * Take a local copy of the address_space - page.mapping may be zeroed
	 * by truncate after unlock_page().   The address_space itself remains
	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
	 * release semantics to prevent the compiler from undoing this copying.
	 */
3287
	mapping = page_rmapping(fault_page);
3288 3289 3290 3291 3292 3293 3294
	unlock_page(fault_page);
	if ((dirtied || vma->vm_ops->page_mkwrite) && mapping) {
		/*
		 * Some device drivers do not set page.mapping but still
		 * dirty their pages
		 */
		balance_dirty_pages_ratelimited(mapping);
3295
	}
3296

3297
	if (!vma->vm_ops->page_mkwrite)
3298
		file_update_time(vma->vm_file);
N
Nick Piggin 已提交
3299

3300
	return ret;
3301
}
3302

3303 3304 3305 3306 3307 3308
/*
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults).
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
 */
3309
static int do_fault(struct fault_env *fe)
3310
{
K
Kirill A. Shutemov 已提交
3311 3312
	struct vm_area_struct *vma = fe->vma;
	pgoff_t pgoff = linear_page_index(vma, fe->address);
3313

3314 3315 3316
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
		return VM_FAULT_SIGBUS;
K
Kirill A. Shutemov 已提交
3317
	if (!(fe->flags & FAULT_FLAG_WRITE))
3318
		return do_read_fault(fe, pgoff);
3319
	if (!(vma->vm_flags & VM_SHARED))
3320 3321
		return do_cow_fault(fe, pgoff);
	return do_shared_fault(fe, pgoff);
3322 3323
}

3324
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3325 3326
				unsigned long addr, int page_nid,
				int *flags)
3327 3328 3329 3330
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3331
	if (page_nid == numa_node_id()) {
3332
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3333 3334
		*flags |= TNF_FAULT_LOCAL;
	}
3335 3336 3337 3338

	return mpol_misplaced(page, vma, addr);
}

K
Kirill A. Shutemov 已提交
3339
static int do_numa_page(struct fault_env *fe, pte_t pte)
3340
{
K
Kirill A. Shutemov 已提交
3341
	struct vm_area_struct *vma = fe->vma;
3342
	struct page *page = NULL;
3343
	int page_nid = -1;
3344
	int last_cpupid;
3345
	int target_nid;
3346
	bool migrated = false;
3347
	bool was_writable = pte_write(pte);
3348
	int flags = 0;
3349

3350 3351 3352
	/* A PROT_NONE fault should not end up here */
	BUG_ON(!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)));

3353 3354 3355 3356 3357
	/*
	* 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.
	*
3358 3359 3360
	* We can safely just do a "set_pte_at()", because the old
	* page table entry is not accessible, so there would be no
	* concurrent hardware modifications to the PTE.
3361
	*/
K
Kirill A. Shutemov 已提交
3362 3363 3364 3365
	fe->ptl = pte_lockptr(vma->vm_mm, fe->pmd);
	spin_lock(fe->ptl);
	if (unlikely(!pte_same(*fe->pte, pte))) {
		pte_unmap_unlock(fe->pte, fe->ptl);
3366 3367 3368
		goto out;
	}

3369 3370 3371
	/* Make it present again */
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3372 3373
	if (was_writable)
		pte = pte_mkwrite(pte);
K
Kirill A. Shutemov 已提交
3374 3375
	set_pte_at(vma->vm_mm, fe->address, fe->pte, pte);
	update_mmu_cache(vma, fe->address, fe->pte);
3376

K
Kirill A. Shutemov 已提交
3377
	page = vm_normal_page(vma, fe->address, pte);
3378
	if (!page) {
K
Kirill A. Shutemov 已提交
3379
		pte_unmap_unlock(fe->pte, fe->ptl);
3380 3381 3382
		return 0;
	}

3383 3384
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
K
Kirill A. Shutemov 已提交
3385
		pte_unmap_unlock(fe->pte, fe->ptl);
3386 3387 3388
		return 0;
	}

3389
	/*
3390 3391 3392 3393 3394 3395
	 * Avoid grouping on RO pages in general. RO pages shouldn't hurt as
	 * much anyway since they can be in shared cache state. This misses
	 * the case where a mapping is writable but the process never writes
	 * to it but pte_write gets cleared during protection updates and
	 * pte_dirty has unpredictable behaviour between PTE scan updates,
	 * background writeback, dirty balancing and application behaviour.
3396
	 */
3397
	if (!(vma->vm_flags & VM_WRITE))
3398 3399
		flags |= TNF_NO_GROUP;

3400 3401 3402 3403 3404 3405 3406
	/*
	 * Flag if the page is shared between multiple address spaces. This
	 * is later used when determining whether to group tasks together
	 */
	if (page_mapcount(page) > 1 && (vma->vm_flags & VM_SHARED))
		flags |= TNF_SHARED;

3407
	last_cpupid = page_cpupid_last(page);
3408
	page_nid = page_to_nid(page);
K
Kirill A. Shutemov 已提交
3409 3410 3411
	target_nid = numa_migrate_prep(page, vma, fe->address, page_nid,
			&flags);
	pte_unmap_unlock(fe->pte, fe->ptl);
3412 3413 3414 3415 3416 3417
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3418
	migrated = migrate_misplaced_page(page, vma, target_nid);
3419
	if (migrated) {
3420
		page_nid = target_nid;
3421
		flags |= TNF_MIGRATED;
3422 3423
	} else
		flags |= TNF_MIGRATE_FAIL;
3424 3425

out:
3426
	if (page_nid != -1)
3427
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3428 3429 3430
	return 0;
}

K
Kirill A. Shutemov 已提交
3431
static int create_huge_pmd(struct fault_env *fe)
M
Matthew Wilcox 已提交
3432
{
K
Kirill A. Shutemov 已提交
3433
	struct vm_area_struct *vma = fe->vma;
3434
	if (vma_is_anonymous(vma))
K
Kirill A. Shutemov 已提交
3435
		return do_huge_pmd_anonymous_page(fe);
M
Matthew Wilcox 已提交
3436
	if (vma->vm_ops->pmd_fault)
K
Kirill A. Shutemov 已提交
3437 3438
		return vma->vm_ops->pmd_fault(vma, fe->address, fe->pmd,
				fe->flags);
M
Matthew Wilcox 已提交
3439 3440 3441
	return VM_FAULT_FALLBACK;
}

K
Kirill A. Shutemov 已提交
3442
static int wp_huge_pmd(struct fault_env *fe, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3443
{
K
Kirill A. Shutemov 已提交
3444 3445 3446 3447 3448
	if (vma_is_anonymous(fe->vma))
		return do_huge_pmd_wp_page(fe, orig_pmd);
	if (fe->vma->vm_ops->pmd_fault)
		return fe->vma->vm_ops->pmd_fault(fe->vma, fe->address, fe->pmd,
				fe->flags);
K
Kirill A. Shutemov 已提交
3449 3450 3451 3452 3453

	/* COW handled on pte level: split pmd */
	VM_BUG_ON_VMA(fe->vma->vm_flags & VM_SHARED, fe->vma);
	split_huge_pmd(fe->vma, fe->pmd, fe->address);

M
Matthew Wilcox 已提交
3454 3455 3456
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3457 3458 3459 3460 3461 3462 3463 3464 3465
/*
 * 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).
 *
3466 3467
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3468
 *
3469 3470
 * The mmap_sem may have been released depending on flags and our return value.
 * See filemap_fault() and __lock_page_or_retry().
L
Linus Torvalds 已提交
3471
 */
K
Kirill A. Shutemov 已提交
3472
static int handle_pte_fault(struct fault_env *fe)
L
Linus Torvalds 已提交
3473 3474 3475
{
	pte_t entry;

3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506
	if (unlikely(pmd_none(*fe->pmd))) {
		/*
		 * Leave __pte_alloc() until later: because vm_ops->fault may
		 * want to allocate huge page, and if we expose page table
		 * for an instant, it will be difficult to retract from
		 * concurrent faults and from rmap lookups.
		 */
		fe->pte = NULL;
	} else {
		/* See comment in pte_alloc_one_map() */
		if (pmd_trans_unstable(fe->pmd) || pmd_devmap(*fe->pmd))
			return 0;
		/*
		 * A regular pmd is established and it can't morph into a huge
		 * pmd from under us anymore at this point because we hold the
		 * mmap_sem read mode and khugepaged takes it in write mode.
		 * So now it's safe to run pte_offset_map().
		 */
		fe->pte = pte_offset_map(fe->pmd, fe->address);

		entry = *fe->pte;

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
		 * CONFIG_32BIT=y, so READ_ONCE or ACCESS_ONCE cannot guarantee
		 * atomic accesses.  The code below just needs a consistent
		 * view for the ifs and we later double check anyway with the
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
3507
		if (pte_none(entry)) {
3508 3509
			pte_unmap(fe->pte);
			fe->pte = NULL;
3510
		}
L
Linus Torvalds 已提交
3511 3512
	}

3513 3514 3515 3516 3517 3518 3519 3520 3521 3522
	if (!fe->pte) {
		if (vma_is_anonymous(fe->vma))
			return do_anonymous_page(fe);
		else
			return do_fault(fe);
	}

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

3523
	if (pte_protnone(entry))
K
Kirill A. Shutemov 已提交
3524
		return do_numa_page(fe, entry);
3525

K
Kirill A. Shutemov 已提交
3526 3527 3528
	fe->ptl = pte_lockptr(fe->vma->vm_mm, fe->pmd);
	spin_lock(fe->ptl);
	if (unlikely(!pte_same(*fe->pte, entry)))
3529
		goto unlock;
K
Kirill A. Shutemov 已提交
3530
	if (fe->flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3531
		if (!pte_write(entry))
K
Kirill A. Shutemov 已提交
3532
			return do_wp_page(fe, entry);
L
Linus Torvalds 已提交
3533 3534 3535
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
K
Kirill A. Shutemov 已提交
3536 3537 3538
	if (ptep_set_access_flags(fe->vma, fe->address, fe->pte, entry,
				fe->flags & FAULT_FLAG_WRITE)) {
		update_mmu_cache(fe->vma, fe->address, fe->pte);
3539 3540 3541 3542 3543 3544 3545
	} else {
		/*
		 * This is needed only for protection faults but the arch code
		 * is not yet telling us if this is a protection fault or not.
		 * This still avoids useless tlb flushes for .text page faults
		 * with threads.
		 */
K
Kirill A. Shutemov 已提交
3546 3547
		if (fe->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(fe->vma, fe->address);
3548
	}
3549
unlock:
K
Kirill A. Shutemov 已提交
3550
	pte_unmap_unlock(fe->pte, fe->ptl);
N
Nick Piggin 已提交
3551
	return 0;
L
Linus Torvalds 已提交
3552 3553 3554 3555
}

/*
 * By the time we get here, we already hold the mm semaphore
3556 3557 3558
 *
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
L
Linus Torvalds 已提交
3559
 */
3560 3561
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3562
{
K
Kirill A. Shutemov 已提交
3563 3564 3565 3566 3567
	struct fault_env fe = {
		.vma = vma,
		.address = address,
		.flags = flags,
	};
3568
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3569 3570 3571 3572 3573 3574
	pgd_t *pgd;
	pud_t *pud;

	pgd = pgd_offset(mm, address);
	pud = pud_alloc(mm, pgd, address);
	if (!pud)
H
Hugh Dickins 已提交
3575
		return VM_FAULT_OOM;
K
Kirill A. Shutemov 已提交
3576 3577
	fe.pmd = pmd_alloc(mm, pud, address);
	if (!fe.pmd)
H
Hugh Dickins 已提交
3578
		return VM_FAULT_OOM;
K
Kirill A. Shutemov 已提交
3579 3580
	if (pmd_none(*fe.pmd) && transparent_hugepage_enabled(vma)) {
		int ret = create_huge_pmd(&fe);
3581 3582
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3583
	} else {
K
Kirill A. Shutemov 已提交
3584
		pmd_t orig_pmd = *fe.pmd;
3585 3586
		int ret;

3587
		barrier();
3588
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
3589
			if (pmd_protnone(orig_pmd))
K
Kirill A. Shutemov 已提交
3590
				return do_huge_pmd_numa_page(&fe, orig_pmd);
3591

K
Kirill A. Shutemov 已提交
3592 3593 3594
			if ((fe.flags & FAULT_FLAG_WRITE) &&
					!pmd_write(orig_pmd)) {
				ret = wp_huge_pmd(&fe, orig_pmd);
3595 3596
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3597
			} else {
K
Kirill A. Shutemov 已提交
3598
				huge_pmd_set_accessed(&fe, orig_pmd);
3599
				return 0;
3600
			}
3601 3602 3603
		}
	}

K
Kirill A. Shutemov 已提交
3604
	return handle_pte_fault(&fe);
L
Linus Torvalds 已提交
3605 3606
}

3607 3608 3609 3610 3611 3612
/*
 * By the time we get here, we already hold the mm semaphore
 *
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
 */
3613 3614
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
3615 3616 3617 3618 3619 3620
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
3621
	mem_cgroup_count_vm_event(vma->vm_mm, PGFAULT);
3622 3623 3624 3625 3626 3627 3628 3629 3630

	/* 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)
3631
		mem_cgroup_oom_enable();
3632

K
Kirill A. Shutemov 已提交
3633 3634 3635 3636 3637 3638 3639 3640 3641
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
3642

3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
	if (flags & FAULT_FLAG_USER) {
		mem_cgroup_oom_disable();
                /*
                 * The task may have entered a memcg OOM situation but
                 * if the allocation error was handled gracefully (no
                 * VM_FAULT_OOM), there is no need to kill anything.
                 * Just clean up the OOM state peacefully.
                 */
                if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
                        mem_cgroup_oom_synchronize(false);
	}
3654

3655 3656
	return ret;
}
3657
EXPORT_SYMBOL_GPL(handle_mm_fault);
3658

L
Linus Torvalds 已提交
3659 3660 3661
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
3662
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3663
 */
3664
int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
L
Linus Torvalds 已提交
3665
{
H
Hugh Dickins 已提交
3666 3667
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
3668
		return -ENOMEM;
L
Linus Torvalds 已提交
3669

3670 3671
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3672
	spin_lock(&mm->page_table_lock);
3673
	if (pgd_present(*pgd))		/* Another has populated it */
3674
		pud_free(mm, new);
3675 3676
	else
		pgd_populate(mm, pgd, new);
H
Hugh Dickins 已提交
3677
	spin_unlock(&mm->page_table_lock);
3678
	return 0;
L
Linus Torvalds 已提交
3679 3680 3681 3682 3683 3684
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
3685
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
3686
 */
3687
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
3688
{
H
Hugh Dickins 已提交
3689 3690
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
3691
		return -ENOMEM;
L
Linus Torvalds 已提交
3692

3693 3694
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
3695
	spin_lock(&mm->page_table_lock);
L
Linus Torvalds 已提交
3696
#ifndef __ARCH_HAS_4LEVEL_HACK
3697 3698
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
3699
		pud_populate(mm, pud, new);
3700
	} else	/* Another has populated it */
3701
		pmd_free(mm, new);
3702 3703 3704
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
3705
		pgd_populate(mm, pud, new);
3706 3707
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
3708
#endif /* __ARCH_HAS_4LEVEL_HACK */
H
Hugh Dickins 已提交
3709
	spin_unlock(&mm->page_table_lock);
3710
	return 0;
3711
}
L
Linus Torvalds 已提交
3712 3713
#endif /* __PAGETABLE_PMD_FOLDED */

3714
static int __follow_pte(struct mm_struct *mm, unsigned long address,
J
Johannes Weiner 已提交
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730
		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);
3731
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751
	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;
}

3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762
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
3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
/**
 * 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);

3792
#ifdef CONFIG_HAVE_IOREMAP_PROT
3793 3794 3795
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
3796
{
3797
	int ret = -EINVAL;
3798 3799 3800
	pte_t *ptep, pte;
	spinlock_t *ptl;

3801 3802
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
3803

3804
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
3805
		goto out;
3806
	pte = *ptep;
3807

3808 3809 3810 3811
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
3812
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
3813

3814
	ret = 0;
3815 3816 3817
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
3818
	return ret;
3819 3820 3821 3822 3823 3824 3825
}

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
3826
	void __iomem *maddr;
3827 3828
	int offset = addr & (PAGE_SIZE-1);

3829
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
3830 3831
		return -EINVAL;

3832
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
3833 3834 3835 3836 3837 3838 3839 3840
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
3841
EXPORT_SYMBOL_GPL(generic_access_phys);
3842 3843
#endif

3844
/*
3845 3846
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
3847
 */
3848 3849
static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
		unsigned long addr, void *buf, int len, int write)
3850 3851 3852 3853 3854
{
	struct vm_area_struct *vma;
	void *old_buf = buf;

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
3855
	/* ignore errors, just check how much was successfully transferred */
3856 3857 3858
	while (len) {
		int bytes, ret, offset;
		void *maddr;
3859
		struct page *page = NULL;
3860

3861
		ret = get_user_pages_remote(tsk, mm, addr, 1,
3862
				write, 1, &page, &vma);
3863
		if (ret <= 0) {
3864 3865 3866
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
3867 3868 3869 3870 3871
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
3872
			if (!vma || vma->vm_start > addr)
3873 3874 3875 3876 3877 3878 3879
				break;
			if (vma->vm_ops && vma->vm_ops->access)
				ret = vma->vm_ops->access(vma, addr, buf,
							  len, write);
			if (ret <= 0)
				break;
			bytes = ret;
3880
#endif
3881
		} else {
3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896
			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);
3897
			put_page(page);
3898 3899 3900 3901 3902 3903 3904 3905 3906
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
3907

S
Stephen Wilson 已提交
3908
/**
3909
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
 * @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);
}

3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944
/*
 * 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;
}

3945 3946 3947 3948 3949 3950 3951 3952
/*
 * 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;

3953 3954 3955 3956 3957 3958 3959
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

3960 3961 3962 3963 3964 3965
	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
3966
			char *p;
3967

M
Miklos Szeredi 已提交
3968
			p = file_path(f, buf, PAGE_SIZE);
3969 3970
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
3971
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
3972 3973 3974 3975 3976
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
3977
	up_read(&mm->mmap_sem);
3978
}
3979

3980
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
3981
void __might_fault(const char *file, int line)
3982
{
3983 3984 3985 3986 3987 3988 3989 3990
	/*
	 * 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;
3991
	if (pagefault_disabled())
3992
		return;
3993 3994
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
3995
	if (current->mm)
3996
		might_lock_read(&current->mm->mmap_sem);
3997
#endif
3998
}
3999
EXPORT_SYMBOL(__might_fault);
4000
#endif
A
Andrea Arcangeli 已提交
4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 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 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071

#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 */
4072

4073
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4074 4075 4076 4077 4078 4079 4080 4081 4082

static struct kmem_cache *page_ptl_cachep;

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

4083
bool ptlock_alloc(struct page *page)
4084 4085 4086
{
	spinlock_t *ptl;

4087
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4088 4089
	if (!ptl)
		return false;
4090
	page->ptl = ptl;
4091 4092 4093
	return true;
}

4094
void ptlock_free(struct page *page)
4095
{
4096
	kmem_cache_free(page_ptl_cachep, page->ptl);
4097 4098
}
#endif