memory.c 125.5 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
/*
 *  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.
T
Tobin C Harding 已提交
33
 *              Idea by Alex Bligh (alex@cconcepts.co.uk)
L
Linus Torvalds 已提交
34 35 36 37 38 39 40 41 42
 *
 * 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>
43
#include <linux/sched/mm.h>
44
#include <linux/sched/coredump.h>
45
#include <linux/sched/numa_balancing.h>
46
#include <linux/sched/task.h>
L
Linus Torvalds 已提交
47 48 49 50 51
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
52
#include <linux/memremap.h>
H
Hugh Dickins 已提交
53
#include <linux/ksm.h>
L
Linus Torvalds 已提交
54
#include <linux/rmap.h>
55
#include <linux/export.h>
56
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
57
#include <linux/init.h>
58
#include <linux/pfn_t.h>
P
Peter Zijlstra 已提交
59
#include <linux/writeback.h>
60
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
61
#include <linux/mmu_notifier.h>
62 63 64
#include <linux/kallsyms.h>
#include <linux/swapops.h>
#include <linux/elf.h>
65
#include <linux/gfp.h>
66
#include <linux/migrate.h>
A
Andy Shevchenko 已提交
67
#include <linux/string.h>
68
#include <linux/dma-debug.h>
69
#include <linux/debugfs.h>
70
#include <linux/userfaultfd_k.h>
71
#include <linux/dax.h>
72
#include <linux/oom.h>
L
Linus Torvalds 已提交
73

A
Alexey Dobriyan 已提交
74
#include <asm/io.h>
75
#include <asm/mmu_context.h>
L
Linus Torvalds 已提交
76
#include <asm/pgalloc.h>
77
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
78 79 80 81
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

82 83
#include "internal.h"

84 85
#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
86 87
#endif

A
Andy Whitcroft 已提交
88
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
89 90 91
/* use the per-pgdat data instead for discontigmem - mbligh */
unsigned long max_mapnr;
EXPORT_SYMBOL(max_mapnr);
T
Tobin C Harding 已提交
92 93

struct page *mem_map;
L
Linus Torvalds 已提交
94 95 96 97 98 99 100 101 102 103
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.
 */
T
Tobin C Harding 已提交
104
void *high_memory;
L
Linus Torvalds 已提交
105 106
EXPORT_SYMBOL(high_memory);

107 108 109 110 111 112 113 114 115 116 117 118
/*
 * 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
119 120 121 122

static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
123
	return 1;
124 125 126
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
127
unsigned long zero_pfn __read_mostly;
128 129
EXPORT_SYMBOL(zero_pfn);

T
Tobin C Harding 已提交
130 131
unsigned long highest_memmap_pfn __read_mostly;

H
Hugh Dickins 已提交
132 133 134 135 136 137 138 139 140
/*
 * 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);
141

K
KAMEZAWA Hiroyuki 已提交
142

143 144
#if defined(SPLIT_RSS_COUNTING)

145
void sync_mm_rss(struct mm_struct *mm)
146 147 148 149
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
150 151 152
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
153 154
		}
	}
155
	current->rss_stat.events = 0;
156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
}

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))
177
		sync_mm_rss(task->mm);
178
}
179
#else /* SPLIT_RSS_COUNTING */
180 181 182 183 184 185 186 187

#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)
{
}

188 189 190 191
#endif /* SPLIT_RSS_COUNTING */

#ifdef HAVE_GENERIC_MMU_GATHER

192
static bool tlb_next_batch(struct mmu_gather *tlb)
193 194 195 196 197 198
{
	struct mmu_gather_batch *batch;

	batch = tlb->active;
	if (batch->next) {
		tlb->active = batch->next;
199
		return true;
200 201
	}

202
	if (tlb->batch_count == MAX_GATHER_BATCH_COUNT)
203
		return false;
204

205 206
	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
	if (!batch)
207
		return false;
208

209
	tlb->batch_count++;
210 211 212 213 214 215 216
	batch->next = NULL;
	batch->nr   = 0;
	batch->max  = MAX_GATHER_BATCH;

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

217
	return true;
218 219
}

M
Minchan Kim 已提交
220 221
void arch_tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm,
				unsigned long start, unsigned long end)
222 223 224
{
	tlb->mm = mm;

225 226
	/* Is it from 0 to ~0? */
	tlb->fullmm     = !(start | (end+1));
227
	tlb->need_flush_all = 0;
228 229 230 231
	tlb->local.next = NULL;
	tlb->local.nr   = 0;
	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
	tlb->active     = &tlb->local;
232
	tlb->batch_count = 0;
233 234 235 236

#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb->batch = NULL;
#endif
237
	tlb->page_size = 0;
238 239

	__tlb_reset_range(tlb);
240 241
}

242
static void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb)
243
{
244 245 246
	if (!tlb->end)
		return;

247
	tlb_flush(tlb);
248
	mmu_notifier_invalidate_range(tlb->mm, tlb->start, tlb->end);
249 250
#ifdef CONFIG_HAVE_RCU_TABLE_FREE
	tlb_table_flush(tlb);
251
#endif
252
	__tlb_reset_range(tlb);
253 254 255 256 257
}

static void tlb_flush_mmu_free(struct mmu_gather *tlb)
{
	struct mmu_gather_batch *batch;
258

259
	for (batch = &tlb->local; batch && batch->nr; batch = batch->next) {
260 261 262 263 264 265
		free_pages_and_swap_cache(batch->pages, batch->nr);
		batch->nr = 0;
	}
	tlb->active = &tlb->local;
}

266 267 268 269 270 271
void tlb_flush_mmu(struct mmu_gather *tlb)
{
	tlb_flush_mmu_tlbonly(tlb);
	tlb_flush_mmu_free(tlb);
}

272 273 274 275
/* tlb_finish_mmu
 *	Called at the end of the shootdown operation to free up any resources
 *	that were required.
 */
M
Minchan Kim 已提交
276
void arch_tlb_finish_mmu(struct mmu_gather *tlb,
277
		unsigned long start, unsigned long end, bool force)
278 279 280
{
	struct mmu_gather_batch *batch, *next;

281 282 283
	if (force)
		__tlb_adjust_range(tlb, start, end - start);

284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
	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().
301
 *returns true if the caller should flush.
302
 */
303
bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size)
304 305 306
{
	struct mmu_gather_batch *batch;

307
	VM_BUG_ON(!tlb->end);
308
	VM_WARN_ON(tlb->page_size != page_size);
309

310
	batch = tlb->active;
311 312 313 314 315
	/*
	 * Add the page and check if we are full. If so
	 * force a flush.
	 */
	batch->pages[batch->nr++] = page;
316 317
	if (batch->nr == batch->max) {
		if (!tlb_next_batch(tlb))
318
			return true;
319
		batch = tlb->active;
320
	}
321
	VM_BUG_ON_PAGE(batch->nr > batch->max, page);
322

323
	return false;
324 325 326 327
}

#endif /* HAVE_GENERIC_MMU_GATHER */

328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
#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);
}

401
#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
402

M
Minchan Kim 已提交
403 404 405 406 407 408 409 410 411
/* 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).
 */
void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm,
			unsigned long start, unsigned long end)
{
	arch_tlb_gather_mmu(tlb, mm, start, end);
412
	inc_tlb_flush_pending(tlb->mm);
M
Minchan Kim 已提交
413 414 415 416 417
}

void tlb_finish_mmu(struct mmu_gather *tlb,
		unsigned long start, unsigned long end)
{
418 419 420 421 422 423 424 425 426 427 428
	/*
	 * If there are parallel threads are doing PTE changes on same range
	 * under non-exclusive lock(e.g., mmap_sem read-side) but defer TLB
	 * flush by batching, a thread has stable TLB entry can fail to flush
	 * the TLB by observing pte_none|!pte_dirty, for example so flush TLB
	 * forcefully if we detect parallel PTE batching threads.
	 */
	bool force = mm_tlb_flush_nested(tlb->mm);

	arch_tlb_finish_mmu(tlb, start, end, force);
	dec_tlb_flush_pending(tlb->mm);
M
Minchan Kim 已提交
429 430
}

L
Linus Torvalds 已提交
431 432 433 434
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
435 436
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
437
{
438
	pgtable_t token = pmd_pgtable(*pmd);
439
	pmd_clear(pmd);
440
	pte_free_tlb(tlb, token, addr);
441
	mm_dec_nr_ptes(tlb->mm);
L
Linus Torvalds 已提交
442 443
}

444 445 446
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
447 448 449
{
	pmd_t *pmd;
	unsigned long next;
450
	unsigned long start;
L
Linus Torvalds 已提交
451

452
	start = addr;
L
Linus Torvalds 已提交
453 454 455 456 457
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
458
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
459 460
	} while (pmd++, addr = next, addr != end);

461 462 463 464 465 466 467
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
468
	}
469 470 471 472 473
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
474
	pmd_free_tlb(tlb, pmd, start);
475
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
476 477
}

478
static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
479 480
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
481 482 483
{
	pud_t *pud;
	unsigned long next;
484
	unsigned long start;
L
Linus Torvalds 已提交
485

486
	start = addr;
487
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
488 489 490 491
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
492
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
493 494
	} while (pud++, addr = next, addr != end);

495 496 497 498 499 500 501 502 503 504 505 506 507 508
	start &= P4D_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= P4D_MASK;
		if (!ceiling)
			return;
	}
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(p4d, start);
	p4d_clear(p4d);
	pud_free_tlb(tlb, pud, start);
K
Kirill A. Shutemov 已提交
509
	mm_dec_nr_puds(tlb->mm);
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
}

static inline void free_p4d_range(struct mmu_gather *tlb, pgd_t *pgd,
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
{
	p4d_t *p4d;
	unsigned long next;
	unsigned long start;

	start = addr;
	p4d = p4d_offset(pgd, addr);
	do {
		next = p4d_addr_end(addr, end);
		if (p4d_none_or_clear_bad(p4d))
			continue;
		free_pud_range(tlb, p4d, addr, next, floor, ceiling);
	} while (p4d++, addr = next, addr != end);

529 530 531 532 533 534 535
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
536
	}
537 538 539
	if (end - 1 > ceiling - 1)
		return;

540
	p4d = p4d_offset(pgd, start);
541
	pgd_clear(pgd);
542
	p4d_free_tlb(tlb, p4d, start);
L
Linus Torvalds 已提交
543 544 545
}

/*
546
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
547
 */
548
void free_pgd_range(struct mmu_gather *tlb,
549 550
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
551 552 553
{
	pgd_t *pgd;
	unsigned long next;
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579

	/*
	 * 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
580

581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
	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;
596 597 598 599 600
	/*
	 * We add page table cache pages with PAGE_SIZE,
	 * (see pte_free_tlb()), flush the tlb if we need
	 */
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
601
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
602 603 604 605
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
606
		free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
607
	} while (pgd++, addr = next, addr != end);
608 609
}

610
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
611
		unsigned long floor, unsigned long ceiling)
612 613 614 615 616
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

617
		/*
N
npiggin@suse.de 已提交
618 619
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
620
		 */
621
		unlink_anon_vmas(vma);
622 623
		unlink_file_vma(vma);

624
		if (is_vm_hugetlb_page(vma)) {
625
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
626
				floor, next ? next->vm_start : ceiling);
627 628 629 630 631
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
632
			       && !is_vm_hugetlb_page(next)) {
633 634
				vma = next;
				next = vma->vm_next;
635
				unlink_anon_vmas(vma);
636
				unlink_file_vma(vma);
637 638
			}
			free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
639
				floor, next ? next->vm_start : ceiling);
640
		}
641 642
		vma = next;
	}
L
Linus Torvalds 已提交
643 644
}

645
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
646
{
647
	spinlock_t *ptl;
648
	pgtable_t new = pte_alloc_one(mm, address);
649 650 651
	if (!new)
		return -ENOMEM;

652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
	/*
	 * 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 */

667
	ptl = pmd_lock(mm, pmd);
668
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
669
		mm_inc_nr_ptes(mm);
L
Linus Torvalds 已提交
670
		pmd_populate(mm, pmd, new);
671
		new = NULL;
672
	}
673
	spin_unlock(ptl);
674 675
	if (new)
		pte_free(mm, new);
676
	return 0;
L
Linus Torvalds 已提交
677 678
}

679
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
680
{
681 682 683 684
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

685 686
	smp_wmb(); /* See comment in __pte_alloc */

687
	spin_lock(&init_mm.page_table_lock);
688
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
689
		pmd_populate_kernel(&init_mm, pmd, new);
690
		new = NULL;
691
	}
692
	spin_unlock(&init_mm.page_table_lock);
693 694
	if (new)
		pte_free_kernel(&init_mm, new);
695
	return 0;
L
Linus Torvalds 已提交
696 697
}

K
KAMEZAWA Hiroyuki 已提交
698 699 700 701 702 703
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)
704
{
K
KAMEZAWA Hiroyuki 已提交
705 706
	int i;

707
	if (current->mm == mm)
708
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
709 710 711
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
712 713
}

N
Nick Piggin 已提交
714
/*
715 716 717
 * 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
718 719 720
 *
 * The calling function must still handle the error.
 */
721 722
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
723
{
724
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
725 726
	p4d_t *p4d = p4d_offset(pgd, addr);
	pud_t *pud = pud_offset(p4d, addr);
727 728 729
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
730 731 732 733 734 735 736 737 738 739 740 741 742 743
	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) {
744 745
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
746 747 748 749 750 751
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
752 753 754 755

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

756 757 758
	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));
759
	if (page)
760
		dump_page(page, "bad pte");
761 762
	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);
763 764 765
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
766 767 768 769 770
	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
771
	dump_stack();
772
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
773 774
}

H
Hugh Dickins 已提交
775
/*
N
Nick Piggin 已提交
776
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
777
 *
N
Nick Piggin 已提交
778 779 780
 * "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
781
 *
N
Nick Piggin 已提交
782 783 784 785 786 787 788 789
 * 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.
790
 *
J
Jared Hulbert 已提交
791 792
 * 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
793 794
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
795 796 797
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
798 799 800 801 802 803
 * 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
804 805
 *
 *
N
Nick Piggin 已提交
806
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
807 808 809 810 811 812 813 814 815
 *
 * 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
816
 */
N
Nick Piggin 已提交
817 818 819 820 821
#ifdef __HAVE_ARCH_PTE_SPECIAL
# define HAVE_PTE_SPECIAL 1
#else
# define HAVE_PTE_SPECIAL 0
#endif
822 823
struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
			     pte_t pte, bool with_public_device)
H
Hugh Dickins 已提交
824
{
825
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
826 827

	if (HAVE_PTE_SPECIAL) {
828
		if (likely(!pte_special(pte)))
829
			goto check_pfn;
830 831
		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
832 833
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
		if (is_zero_pfn(pfn))
			return NULL;

		/*
		 * Device public pages are special pages (they are ZONE_DEVICE
		 * pages but different from persistent memory). They behave
		 * allmost like normal pages. The difference is that they are
		 * not on the lru and thus should never be involve with any-
		 * thing that involve lru manipulation (mlock, numa balancing,
		 * ...).
		 *
		 * This is why we still want to return NULL for such page from
		 * vm_normal_page() so that we do not have to special case all
		 * call site of vm_normal_page().
		 */
849
		if (likely(pfn <= highest_memmap_pfn)) {
850 851 852 853 854 855 856 857 858
			struct page *page = pfn_to_page(pfn);

			if (is_device_public_page(page)) {
				if (with_public_device)
					return page;
				return NULL;
			}
		}
		print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
859 860 861 862 863
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
864 865 866 867 868 869
	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
870 871
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
872 873 874 875 876
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
877 878
	}

879 880
	if (is_zero_pfn(pfn))
		return NULL;
881 882 883 884 885
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
886 887

	/*
N
Nick Piggin 已提交
888 889
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
890
	 */
J
Jared Hulbert 已提交
891
out:
892
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
893 894
}

895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934
#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
935 936 937 938 939 940
/*
 * 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
941
static inline unsigned long
L
Linus Torvalds 已提交
942
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
943
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
944
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
945
{
N
Nick Piggin 已提交
946
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
947 948 949 950 951
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
		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);

970
			rss[mm_counter(page)]++;
971 972 973 974 975 976 977 978 979 980 981 982

			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);
983
			}
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
		} else if (is_device_private_entry(entry)) {
			page = device_private_entry_to_page(entry);

			/*
			 * Update rss count even for unaddressable pages, as
			 * they should treated just like normal pages in this
			 * respect.
			 *
			 * We will likely want to have some new rss counters
			 * for unaddressable pages, at some point. But for now
			 * keep things as they are.
			 */
			get_page(page);
			rss[mm_counter(page)]++;
			page_dup_rmap(page, false);

			/*
			 * We do not preserve soft-dirty information, because so
			 * far, checkpoint/restore is the only feature that
			 * requires that. And checkpoint/restore does not work
			 * when a device driver is involved (you cannot easily
			 * save and restore device driver state).
			 */
			if (is_write_device_private_entry(entry) &&
			    is_cow_mapping(vm_flags)) {
				make_device_private_entry_read(&entry);
				pte = swp_entry_to_pte(entry);
				set_pte_at(src_mm, addr, src_pte, pte);
			}
L
Linus Torvalds 已提交
1013
		}
1014
		goto out_set_pte;
L
Linus Torvalds 已提交
1015 1016 1017 1018 1019 1020
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
1021
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
1022
		ptep_set_wrprotect(src_mm, addr, src_pte);
1023
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
1024 1025 1026 1027 1028 1029 1030 1031 1032
	}

	/*
	 * 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);
1033 1034 1035 1036

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
1037
		page_dup_rmap(page, false);
1038
		rss[mm_counter(page)]++;
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
	} else if (pte_devmap(pte)) {
		page = pte_page(pte);

		/*
		 * Cache coherent device memory behave like regular page and
		 * not like persistent memory page. For more informations see
		 * MEMORY_DEVICE_CACHE_COHERENT in memory_hotplug.h
		 */
		if (is_device_public_page(page)) {
			get_page(page);
			page_dup_rmap(page, false);
			rss[mm_counter(page)]++;
		}
1052
	}
1053 1054 1055

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
1056
	return 0;
L
Linus Torvalds 已提交
1057 1058
}

1059
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
1060 1061
		   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
1062
{
1063
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
1064
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
1065
	spinlock_t *src_ptl, *dst_ptl;
1066
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
1067
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
1068
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
1069 1070

again:
K
KAMEZAWA Hiroyuki 已提交
1071 1072
	init_rss_vec(rss);

H
Hugh Dickins 已提交
1073
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
1074 1075
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
1076
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
1077
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
1078
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
1079 1080
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
1081
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1082 1083 1084 1085 1086 1087

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
1088 1089 1090
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
1091
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
1092 1093
				break;
		}
L
Linus Torvalds 已提交
1094 1095 1096 1097
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
1098 1099 1100 1101
		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
1102 1103 1104
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

1105
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1106
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
1107
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
1108
	add_mm_rss_vec(dst_mm, rss);
1109
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
1110
	cond_resched();
H
Hugh Dickins 已提交
1111 1112 1113 1114 1115 1116

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
	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);
1135 1136
		if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
			|| pmd_devmap(*src_pmd)) {
1137
			int err;
1138
			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
1139 1140 1141 1142 1143 1144 1145 1146
			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
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
		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,
1157
		p4d_t *dst_p4d, p4d_t *src_p4d, struct vm_area_struct *vma,
L
Linus Torvalds 已提交
1158 1159 1160 1161 1162
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

1163
	dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
L
Linus Torvalds 已提交
1164 1165
	if (!dst_pud)
		return -ENOMEM;
1166
	src_pud = pud_offset(src_p4d, addr);
L
Linus Torvalds 已提交
1167 1168
	do {
		next = pud_addr_end(addr, end);
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
		if (pud_trans_huge(*src_pud) || pud_devmap(*src_pud)) {
			int err;

			VM_BUG_ON_VMA(next-addr != HPAGE_PUD_SIZE, vma);
			err = copy_huge_pud(dst_mm, src_mm,
					    dst_pud, src_pud, addr, vma);
			if (err == -ENOMEM)
				return -ENOMEM;
			if (!err)
				continue;
			/* fall through */
		}
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186 1187 1188 1189
		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;
}

1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
static inline int copy_p4d_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)
{
	p4d_t *src_p4d, *dst_p4d;
	unsigned long next;

	dst_p4d = p4d_alloc(dst_mm, dst_pgd, addr);
	if (!dst_p4d)
		return -ENOMEM;
	src_p4d = p4d_offset(src_pgd, addr);
	do {
		next = p4d_addr_end(addr, end);
		if (p4d_none_or_clear_bad(src_p4d))
			continue;
		if (copy_pud_range(dst_mm, src_mm, dst_p4d, src_p4d,
						vma, addr, next))
			return -ENOMEM;
	} while (dst_p4d++, src_p4d++, addr = next, addr != end);
	return 0;
}

L
Linus Torvalds 已提交
1212 1213 1214 1215 1216 1217 1218
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;
1219 1220 1221
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1222
	int ret;
L
Linus Torvalds 已提交
1223

1224 1225 1226 1227 1228 1229
	/*
	 * 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.
	 */
1230 1231 1232
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1233

L
Linus Torvalds 已提交
1234 1235 1236
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1237
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1238 1239 1240 1241
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1242
		ret = track_pfn_copy(vma);
1243 1244 1245 1246
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1247 1248 1249 1250 1251 1252
	/*
	 * 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.
	 */
1253 1254 1255 1256 1257 1258
	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
1259 1260

	ret = 0;
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266
	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;
1267
		if (unlikely(copy_p4d_range(dst_mm, src_mm, dst_pgd, src_pgd,
A
Andrea Arcangeli 已提交
1268 1269 1270 1271
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1272
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1273

1274 1275
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1276
	return ret;
L
Linus Torvalds 已提交
1277 1278
}

1279
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1280
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1281
				unsigned long addr, unsigned long end,
1282
				struct zap_details *details)
L
Linus Torvalds 已提交
1283
{
N
Nick Piggin 已提交
1284
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1285
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1286
	int rss[NR_MM_COUNTERS];
1287
	spinlock_t *ptl;
1288
	pte_t *start_pte;
1289
	pte_t *pte;
1290
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1291

1292
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1293
again:
1294
	init_rss_vec(rss);
1295 1296
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1297
	flush_tlb_batched_pending(mm);
1298
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1299 1300
	do {
		pte_t ptent = *pte;
T
Tobin C Harding 已提交
1301
		if (pte_none(ptent))
L
Linus Torvalds 已提交
1302
			continue;
1303

L
Linus Torvalds 已提交
1304
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1305
			struct page *page;
1306

1307
			page = _vm_normal_page(vma, addr, ptent, true);
L
Linus Torvalds 已提交
1308 1309 1310 1311 1312 1313 1314
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1315
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1316 1317
					continue;
			}
N
Nick Piggin 已提交
1318
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1319
							tlb->fullmm);
L
Linus Torvalds 已提交
1320 1321 1322
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1323 1324

			if (!PageAnon(page)) {
1325 1326
				if (pte_dirty(ptent)) {
					force_flush = 1;
1327
					set_page_dirty(page);
1328
				}
1329
				if (pte_young(ptent) &&
1330
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1331
					mark_page_accessed(page);
1332
			}
1333
			rss[mm_counter(page)]--;
1334
			page_remove_rmap(page, false);
1335 1336
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1337
			if (unlikely(__tlb_remove_page(tlb, page))) {
1338
				force_flush = 1;
1339
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1340
				break;
1341
			}
L
Linus Torvalds 已提交
1342 1343
			continue;
		}
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366

		entry = pte_to_swp_entry(ptent);
		if (non_swap_entry(entry) && is_device_private_entry(entry)) {
			struct page *page = device_private_entry_to_page(entry);

			if (unlikely(details && details->check_mapping)) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping !=
				    page_rmapping(page))
					continue;
			}

			pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
			rss[mm_counter(page)]--;
			page_remove_rmap(page, false);
			put_page(page);
			continue;
		}

1367 1368
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1369
			continue;
K
KAMEZAWA Hiroyuki 已提交
1370

1371 1372 1373 1374 1375
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1376

1377
			page = migration_entry_to_page(entry);
1378
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1379
		}
1380 1381
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1382
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1383
	} while (pte++, addr += PAGE_SIZE, addr != end);
1384

K
KAMEZAWA Hiroyuki 已提交
1385
	add_mm_rss_vec(mm, rss);
1386
	arch_leave_lazy_mmu_mode();
1387

1388
	/* Do the actual TLB flush before dropping ptl */
1389
	if (force_flush)
1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
		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);
1402
		if (addr != end)
P
Peter Zijlstra 已提交
1403 1404 1405
			goto again;
	}

1406
	return addr;
L
Linus Torvalds 已提交
1407 1408
}

1409
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1410
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1411
				unsigned long addr, unsigned long end,
1412
				struct zap_details *details)
L
Linus Torvalds 已提交
1413 1414 1415 1416 1417 1418 1419
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1420
		if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1421
			if (next - addr != HPAGE_PMD_SIZE) {
1422 1423
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1424
				__split_huge_pmd(vma, pmd, addr, false, NULL);
S
Shaohua Li 已提交
1425
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1426
				goto next;
1427 1428
			/* fall through */
		}
1429 1430 1431 1432 1433 1434 1435 1436 1437
		/*
		 * 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;
1438
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1439
next:
1440 1441
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1442 1443

	return addr;
L
Linus Torvalds 已提交
1444 1445
}

1446
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
1447
				struct vm_area_struct *vma, p4d_t *p4d,
L
Linus Torvalds 已提交
1448
				unsigned long addr, unsigned long end,
1449
				struct zap_details *details)
L
Linus Torvalds 已提交
1450 1451 1452 1453
{
	pud_t *pud;
	unsigned long next;

1454
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
1455 1456
	do {
		next = pud_addr_end(addr, end);
1457 1458 1459 1460 1461 1462 1463 1464
		if (pud_trans_huge(*pud) || pud_devmap(*pud)) {
			if (next - addr != HPAGE_PUD_SIZE) {
				VM_BUG_ON_VMA(!rwsem_is_locked(&tlb->mm->mmap_sem), vma);
				split_huge_pud(vma, pud, addr);
			} else if (zap_huge_pud(tlb, vma, pud, addr))
				goto next;
			/* fall through */
		}
1465
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1466
			continue;
1467
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1468 1469
next:
		cond_resched();
1470
	} while (pud++, addr = next, addr != end);
1471 1472

	return addr;
L
Linus Torvalds 已提交
1473 1474
}

1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
static inline unsigned long zap_p4d_range(struct mmu_gather *tlb,
				struct vm_area_struct *vma, pgd_t *pgd,
				unsigned long addr, unsigned long end,
				struct zap_details *details)
{
	p4d_t *p4d;
	unsigned long next;

	p4d = p4d_offset(pgd, addr);
	do {
		next = p4d_addr_end(addr, end);
		if (p4d_none_or_clear_bad(p4d))
			continue;
		next = zap_pud_range(tlb, vma, p4d, addr, next, details);
	} while (p4d++, addr = next, addr != end);

	return addr;
}

M
Michal Hocko 已提交
1494
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1495 1496 1497
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1498 1499 1500 1501 1502 1503 1504 1505 1506
{
	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);
1507
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1508
			continue;
1509
		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
1510
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1511 1512
	tlb_end_vma(tlb, vma);
}
1513

1514 1515 1516

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1517
		unsigned long end_addr,
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
		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;

1529 1530 1531
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1532
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1533
		untrack_pfn(vma, 0, 0);
1534 1535 1536 1537 1538 1539 1540

	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
1541
			 * cleanup path of mmap_region. When
1542
			 * hugetlbfs ->mmap method fails,
1543
			 * mmap_region() nullifies vma->vm_file
1544 1545 1546 1547
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1548
			if (vma->vm_file) {
1549
				i_mmap_lock_write(vma->vm_file->f_mapping);
1550
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1551
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1552
			}
1553 1554 1555
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1556 1557 1558 1559
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1560
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1561 1562 1563 1564
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1565
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
 *
 * 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
1576
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1577
		struct vm_area_struct *vma, unsigned long start_addr,
1578
		unsigned long end_addr)
L
Linus Torvalds 已提交
1579
{
A
Andrea Arcangeli 已提交
1580
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1581

A
Andrea Arcangeli 已提交
1582
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1583
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1584
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1585
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1586 1587 1588 1589 1590
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1591
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1592
 * @size: number of bytes to zap
1593 1594
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1595
 */
1596
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1597
		unsigned long size)
L
Linus Torvalds 已提交
1598 1599
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1600
	struct mmu_gather tlb;
1601
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1602 1603

	lru_add_drain();
1604
	tlb_gather_mmu(&tlb, mm, start, end);
1605
	update_hiwater_rss(mm);
1606
	mmu_notifier_invalidate_range_start(mm, start, end);
1607
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next) {
1608
		unmap_single_vma(&tlb, vma, start, end, NULL);
1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620

		/*
		 * zap_page_range does not specify whether mmap_sem should be
		 * held for read or write. That allows parallel zap_page_range
		 * operations to unmap a PTE and defer a flush meaning that
		 * this call observes pte_none and fails to flush the TLB.
		 * Rather than adding a complex API, ensure that no stale
		 * TLB entries exist when this call returns.
		 */
		flush_tlb_range(vma, start, end);
	}

1621 1622
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1623 1624
}

1625 1626 1627 1628 1629
/**
 * 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
1630
 * @details: details of shared cache invalidation
1631 1632
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1633
 */
1634
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1635 1636 1637
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1638
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1639 1640 1641
	unsigned long end = address + size;

	lru_add_drain();
1642
	tlb_gather_mmu(&tlb, mm, address, end);
1643
	update_hiwater_rss(mm);
1644
	mmu_notifier_invalidate_range_start(mm, address, end);
1645
	unmap_single_vma(&tlb, vma, address, end, details);
1646
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1647
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1648 1649
}

1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
/**
 * 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;
1668
	zap_page_range_single(vma, address, size, NULL);
1669 1670 1671 1672
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1673
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1674
			spinlock_t **ptl)
1675
{
1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
	pgd_t *pgd;
	p4d_t *p4d;
	pud_t *pud;
	pmd_t *pmd;

	pgd = pgd_offset(mm, addr);
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
	pud = pud_alloc(mm, p4d, addr);
	if (!pud)
		return NULL;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return NULL;

	VM_BUG_ON(pmd_trans_huge(*pmd));
	return pte_alloc_map_lock(mm, pmd, addr, ptl);
1694 1695
}

1696 1697 1698 1699 1700 1701 1702
/*
 * 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
1703 1704
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1705
{
N
Nick Piggin 已提交
1706
	struct mm_struct *mm = vma->vm_mm;
1707
	int retval;
1708
	pte_t *pte;
1709 1710
	spinlock_t *ptl;

1711
	retval = -EINVAL;
1712
	if (PageAnon(page))
1713
		goto out;
1714 1715
	retval = -ENOMEM;
	flush_dcache_page(page);
1716
	pte = get_locked_pte(mm, addr, &ptl);
1717
	if (!pte)
1718
		goto out;
1719 1720 1721 1722 1723 1724
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1725
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1726
	page_add_file_rmap(page, false);
1727 1728 1729
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1730 1731
	pte_unmap_unlock(pte, ptl);
	return retval;
1732 1733 1734 1735 1736 1737
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1738 1739 1740 1741 1742 1743
/**
 * 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
 *
1744 1745 1746 1747 1748 1749
 * 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
1750
 * (see split_page()).
1751 1752 1753 1754 1755 1756 1757 1758
 *
 * 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.
1759 1760 1761 1762 1763
 *
 * 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.
1764
 */
N
Nick Piggin 已提交
1765 1766
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1767 1768 1769 1770 1771
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1772 1773 1774 1775 1776
	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
1777
	return insert_page(vma, addr, page, vma->vm_page_prot);
1778
}
1779
EXPORT_SYMBOL(vm_insert_page);
1780

N
Nick Piggin 已提交
1781
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
R
Ross Zwisler 已提交
1782
			pfn_t pfn, pgprot_t prot, bool mkwrite)
N
Nick Piggin 已提交
1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793
{
	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;
R
Ross Zwisler 已提交
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
	if (!pte_none(*pte)) {
		if (mkwrite) {
			/*
			 * For read faults on private mappings the PFN passed
			 * in may not match the PFN we have mapped if the
			 * mapped PFN is a writeable COW page.  In the mkwrite
			 * case we are creating a writable PTE for a shared
			 * mapping and we expect the PFNs to match.
			 */
			if (WARN_ON_ONCE(pte_pfn(*pte) != pfn_t_to_pfn(pfn)))
				goto out_unlock;
			entry = *pte;
			goto out_mkwrite;
		} else
			goto out_unlock;
	}
N
Nick Piggin 已提交
1810 1811

	/* Ok, finally just insert the thing.. */
1812 1813 1814 1815
	if (pfn_t_devmap(pfn))
		entry = pte_mkdevmap(pfn_t_pte(pfn, prot));
	else
		entry = pte_mkspecial(pfn_t_pte(pfn, prot));
R
Ross Zwisler 已提交
1816 1817 1818 1819 1820 1821 1822

out_mkwrite:
	if (mkwrite) {
		entry = pte_mkyoung(entry);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
	}

N
Nick Piggin 已提交
1823
	set_pte_at(mm, addr, pte, entry);
1824
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1825 1826 1827 1828 1829 1830 1831 1832

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

N
Nick Piggin 已提交
1833 1834 1835 1836 1837 1838
/**
 * 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
 *
1839
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1840 1841 1842 1843
 * 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
1844 1845 1846 1847 1848
 *
 * 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
1849 1850
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1851
			unsigned long pfn)
A
Andy Lutomirski 已提交
1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
{
	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
1874
{
1875
	int ret;
N
Nick Piggin 已提交
1876 1877 1878 1879 1880 1881
	/*
	 * 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
1882 1883 1884 1885 1886
	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
1887

N
Nick Piggin 已提交
1888 1889
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1890 1891

	track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
1892

R
Ross Zwisler 已提交
1893 1894
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
			false);
1895 1896

	return ret;
N
Nick Piggin 已提交
1897
}
A
Andy Lutomirski 已提交
1898
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1899

R
Ross Zwisler 已提交
1900 1901
static int __vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
			pfn_t pfn, bool mkwrite)
N
Nick Piggin 已提交
1902
{
1903 1904
	pgprot_t pgprot = vma->vm_page_prot;

N
Nick Piggin 已提交
1905
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1906

N
Nick Piggin 已提交
1907 1908
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1909 1910

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1911

N
Nick Piggin 已提交
1912 1913 1914 1915
	/*
	 * 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
1916 1917
	 * 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
1918
	 */
1919
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1920 1921
		struct page *page;

1922 1923 1924 1925 1926 1927
		/*
		 * 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));
1928
		return insert_page(vma, addr, page, pgprot);
N
Nick Piggin 已提交
1929
	}
R
Ross Zwisler 已提交
1930 1931 1932 1933 1934 1935 1936 1937
	return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
}

int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
			pfn_t pfn)
{
	return __vm_insert_mixed(vma, addr, pfn, false);

N
Nick Piggin 已提交
1938
}
N
Nick Piggin 已提交
1939
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1940

R
Ross Zwisler 已提交
1941 1942 1943 1944 1945 1946 1947
int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
			pfn_t pfn)
{
	return __vm_insert_mixed(vma, addr, pfn, true);
}
EXPORT_SYMBOL(vm_insert_mixed_mkwrite);

L
Linus Torvalds 已提交
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957
/*
 * 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
1958
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1959

H
Hugh Dickins 已提交
1960
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1961 1962
	if (!pte)
		return -ENOMEM;
1963
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1964 1965
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1966
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1967 1968
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1969
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1970
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
	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;
1985
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1986 1987 1988 1989 1990 1991 1992 1993 1994
	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;
}

1995
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
L
Linus Torvalds 已提交
1996 1997 1998 1999 2000 2001 2002
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;

	pfn -= addr >> PAGE_SHIFT;
2003
	pud = pud_alloc(mm, p4d, addr);
L
Linus Torvalds 已提交
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
	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;
}

2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034
static inline int remap_p4d_range(struct mm_struct *mm, pgd_t *pgd,
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	p4d_t *p4d;
	unsigned long next;

	pfn -= addr >> PAGE_SHIFT;
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
		if (remap_pud_range(mm, p4d, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot))
			return -ENOMEM;
	} while (p4d++, addr = next, addr != end);
	return 0;
}

2035 2036 2037 2038 2039 2040 2041 2042 2043 2044
/**
 * 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
2045 2046 2047 2048 2049
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;
2050
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
2051
	struct mm_struct *mm = vma->vm_mm;
2052
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
2053 2054 2055 2056 2057 2058 2059
	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).
2060 2061 2062
	 *   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.
2063 2064 2065 2066
	 *   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
2067 2068 2069 2070
	 *
	 * 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".
2071
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
2072
	 */
2073 2074 2075
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
2076
		vma->vm_pgoff = pfn;
2077 2078
	}

2079
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
2080
	if (err)
2081
		return -EINVAL;
L
Linus Torvalds 已提交
2082

2083
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
2084 2085 2086 2087 2088 2089 2090

	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);
2091
		err = remap_p4d_range(mm, pgd, addr, next,
L
Linus Torvalds 已提交
2092 2093 2094 2095
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2096 2097

	if (err)
2098
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
2099

L
Linus Torvalds 已提交
2100 2101 2102 2103
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150
/**
 * 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);

2151 2152 2153 2154 2155 2156
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;
2157
	pgtable_t token;
2158
	spinlock_t *uninitialized_var(ptl);
2159 2160 2161 2162 2163 2164 2165 2166 2167

	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));

2168 2169
	arch_enter_lazy_mmu_mode();

2170
	token = pmd_pgtable(*pmd);
2171 2172

	do {
2173
		err = fn(pte++, token, addr, data);
2174 2175
		if (err)
			break;
2176
	} while (addr += PAGE_SIZE, addr != end);
2177

2178 2179
	arch_leave_lazy_mmu_mode();

2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
	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
2193 2194
	BUG_ON(pud_huge(*pud));

2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
	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;
}

2207
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
2208 2209 2210 2211 2212 2213 2214
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

2215
	pud = pud_alloc(mm, p4d, addr);
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226
	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;
}

2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246
static int apply_to_p4d_range(struct mm_struct *mm, pgd_t *pgd,
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	p4d_t *p4d;
	unsigned long next;
	int err;

	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
		err = apply_to_pud_range(mm, p4d, addr, next, fn, data);
		if (err)
			break;
	} while (p4d++, addr = next, addr != end);
	return err;
}

2247 2248 2249 2250 2251 2252 2253 2254 2255
/*
 * 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;
2256
	unsigned long end = addr + size;
2257 2258
	int err;

2259 2260 2261
	if (WARN_ON(addr >= end))
		return -EINVAL;

2262 2263 2264
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
2265
		err = apply_to_p4d_range(mm, pgd, addr, next, fn, data);
2266 2267 2268
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2269

2270 2271 2272 2273
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2274
/*
2275 2276 2277 2278 2279
 * 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;
2280
 * and do_anonymous_page can safely check later on).
2281
 */
H
Hugh Dickins 已提交
2282
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2283 2284 2285 2286 2287
				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
2288 2289
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2290
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2291
		spin_unlock(ptl);
2292 2293 2294 2295 2296 2297
	}
#endif
	pte_unmap(page_table);
	return same;
}

2298
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2299
{
2300 2301
	debug_dma_assert_idle(src);

2302 2303 2304 2305 2306 2307 2308
	/*
	 * 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)) {
2309
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2310 2311 2312 2313 2314 2315 2316 2317 2318
		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))
2319
			clear_page(kaddr);
2320
		kunmap_atomic(kaddr);
2321
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2322 2323
	} else
		copy_user_highpage(dst, src, va, vma);
2324 2325
}

2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339
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;
}

2340 2341 2342 2343 2344 2345
/*
 * 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.
 */
2346
static int do_page_mkwrite(struct vm_fault *vmf)
2347 2348
{
	int ret;
2349 2350
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2351

2352
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2353

2354
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2355 2356
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370
	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;
}

2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
/*
 * Handle dirtying of a page in shared file mapping on a write fault.
 *
 * The function expects the page to be locked and unlocks it.
 */
static void fault_dirty_shared_page(struct vm_area_struct *vma,
				    struct page *page)
{
	struct address_space *mapping;
	bool dirtied;
	bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;

	dirtied = set_page_dirty(page);
	VM_BUG_ON_PAGE(PageAnon(page), page);
	/*
	 * Take a local copy of the address_space - page.mapping may be zeroed
	 * by truncate after unlock_page().   The address_space itself remains
	 * pinned by vma->vm_file's reference.  We rely on unlock_page()'s
	 * release semantics to prevent the compiler from undoing this copying.
	 */
	mapping = page_rmapping(page);
	unlock_page(page);

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

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

2406 2407 2408 2409 2410 2411 2412 2413
/*
 * 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.
 */
2414
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2415
	__releases(vmf->ptl)
2416
{
J
Jan Kara 已提交
2417
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2418
	struct page *page = vmf->page;
2419 2420 2421 2422 2423 2424 2425 2426 2427
	pte_t entry;
	/*
	 * Clear the pages cpupid information as the existing
	 * information potentially belongs to a now completely
	 * unrelated process.
	 */
	if (page)
		page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);

J
Jan Kara 已提交
2428 2429
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2430
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2431 2432 2433
	if (ptep_set_access_flags(vma, vmf->address, vmf->pte, entry, 1))
		update_mmu_cache(vma, vmf->address, vmf->pte);
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2434 2435
}

2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451
/*
 * Handle the case of a page which we actually need to copy to a new page.
 *
 * Called with mmap_sem locked and the old page referenced, but
 * without the ptl held.
 *
 * High level logic flow:
 *
 * - Allocate a page, copy the content of the old page to the new one.
 * - Handle book keeping and accounting - cgroups, mmu-notifiers, etc.
 * - Take the PTL. If the pte changed, bail out and release the allocated page
 * - If the pte is still the way we remember it, update the page table and all
 *   relevant references. This includes dropping the reference the page-table
 *   held to the old page, as well as updating the rmap.
 * - In any case, unlock the PTL and drop the reference we took to the old page.
 */
J
Jan Kara 已提交
2452
static int wp_page_copy(struct vm_fault *vmf)
2453
{
J
Jan Kara 已提交
2454
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2455
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2456
	struct page *old_page = vmf->page;
2457 2458 2459
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
J
Jan Kara 已提交
2460
	const unsigned long mmun_start = vmf->address & PAGE_MASK;
K
Kirill A. Shutemov 已提交
2461
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2462 2463 2464 2465 2466
	struct mem_cgroup *memcg;

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

J
Jan Kara 已提交
2467
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2468 2469
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2470 2471 2472
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2473
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2474
				vmf->address);
2475 2476
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2477
		cow_user_page(new_page, old_page, vmf->address, vma);
2478 2479
	}

2480
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2481 2482
		goto oom_free_new;

2483 2484
	__SetPageUptodate(new_page);

2485 2486 2487 2488 2489
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2490
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2491
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2492 2493
		if (old_page) {
			if (!PageAnon(old_page)) {
2494 2495
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2496 2497 2498 2499 2500
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2501
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2502 2503 2504 2505 2506 2507 2508 2509
		entry = mk_pte(new_page, vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
		/*
		 * Clear the pte entry and flush it first, before updating the
		 * pte with the new entry. This will avoid a race condition
		 * seen in the presence of one thread doing SMC and another
		 * thread doing COW.
		 */
J
Jan Kara 已提交
2510 2511
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2512
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2513 2514 2515 2516 2517 2518
		lru_cache_add_active_or_unevictable(new_page, vma);
		/*
		 * We call the notify macro here because, when using secondary
		 * mmu page tables (such as kvm shadow page tables), we want the
		 * new page to be mapped directly into the secondary page table.
		 */
J
Jan Kara 已提交
2519 2520
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
		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.
			 */
2544
			page_remove_rmap(old_page, false);
2545 2546 2547 2548 2549 2550
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2551
		mem_cgroup_cancel_charge(new_page, memcg, false);
2552 2553 2554
	}

	if (new_page)
2555
		put_page(new_page);
2556

J
Jan Kara 已提交
2557
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2558 2559 2560 2561 2562
	/*
	 * No need to double call mmu_notifier->invalidate_range() callback as
	 * the above ptep_clear_flush_notify() did already call it.
	 */
	mmu_notifier_invalidate_range_only_end(mm, mmun_start, mmun_end);
2563 2564 2565 2566 2567 2568 2569
	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 */
2570 2571
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2572 2573
			unlock_page(old_page);
		}
2574
		put_page(old_page);
2575 2576 2577
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2578
	put_page(new_page);
2579 2580
oom:
	if (old_page)
2581
		put_page(old_page);
2582 2583 2584
	return VM_FAULT_OOM;
}

2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610
/**
 * finish_mkwrite_fault - finish page fault for a shared mapping, making PTE
 *			  writeable once the page is prepared
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a write page fault in a
 * shared mapping due to PTE being read-only once the mapped page is prepared.
 * It handles locking of PTE and modifying it. The function returns
 * VM_FAULT_WRITE on success, 0 when PTE got changed before we acquired PTE
 * lock.
 *
 * The function expects the page to be locked or other protection against
 * concurrent faults / writeback (such as DAX radix tree locks).
 */
int finish_mkwrite_fault(struct vm_fault *vmf)
{
	WARN_ON_ONCE(!(vmf->vma->vm_flags & VM_SHARED));
	vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd, vmf->address,
				       &vmf->ptl);
	/*
	 * We might have raced with another page fault while we released the
	 * pte_offset_map_lock.
	 */
	if (!pte_same(*vmf->pte, vmf->orig_pte)) {
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2611
		return VM_FAULT_NOPAGE;
2612 2613
	}
	wp_page_reuse(vmf);
2614
	return 0;
2615 2616
}

2617 2618 2619 2620
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
J
Jan Kara 已提交
2621
static int wp_pfn_shared(struct vm_fault *vmf)
2622
{
J
Jan Kara 已提交
2623
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2624

2625 2626 2627
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
		int ret;

J
Jan Kara 已提交
2628
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2629
		vmf->flags |= FAULT_FLAG_MKWRITE;
2630
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2631
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2632
			return ret;
2633
		return finish_mkwrite_fault(vmf);
2634
	}
2635 2636
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2637 2638
}

J
Jan Kara 已提交
2639
static int wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2640
	__releases(vmf->ptl)
2641
{
J
Jan Kara 已提交
2642
	struct vm_area_struct *vma = vmf->vma;
2643

J
Jan Kara 已提交
2644
	get_page(vmf->page);
2645 2646 2647 2648

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

J
Jan Kara 已提交
2649
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2650
		tmp = do_page_mkwrite(vmf);
2651 2652
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2653
			put_page(vmf->page);
2654 2655
			return tmp;
		}
2656
		tmp = finish_mkwrite_fault(vmf);
2657
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2658 2659
			unlock_page(vmf->page);
			put_page(vmf->page);
2660
			return tmp;
2661
		}
2662 2663
	} else {
		wp_page_reuse(vmf);
2664
		lock_page(vmf->page);
2665
	}
2666 2667
	fault_dirty_shared_page(vma, vmf->page);
	put_page(vmf->page);
2668

2669
	return VM_FAULT_WRITE;
2670 2671
}

L
Linus Torvalds 已提交
2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
/*
 * 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.
 *
2686 2687 2688
 * 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
2689
 */
J
Jan Kara 已提交
2690
static int do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2691
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2692
{
J
Jan Kara 已提交
2693
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2694

J
Jan Kara 已提交
2695 2696
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2697
		/*
2698 2699
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2700 2701
		 *
		 * We should not cow pages in a shared writeable mapping.
2702
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2703 2704 2705
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2706
			return wp_pfn_shared(vmf);
2707

J
Jan Kara 已提交
2708
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2709
		return wp_page_copy(vmf);
2710
	}
L
Linus Torvalds 已提交
2711

2712
	/*
P
Peter Zijlstra 已提交
2713 2714
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2715
	 */
J
Jan Kara 已提交
2716
	if (PageAnon(vmf->page) && !PageKsm(vmf->page)) {
2717
		int total_map_swapcount;
J
Jan Kara 已提交
2718 2719
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2720
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2721
			lock_page(vmf->page);
J
Jan Kara 已提交
2722 2723
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2724
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2725
				unlock_page(vmf->page);
J
Jan Kara 已提交
2726
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2727
				put_page(vmf->page);
2728
				return 0;
2729
			}
J
Jan Kara 已提交
2730
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2731
		}
2732 2733
		if (reuse_swap_page(vmf->page, &total_map_swapcount)) {
			if (total_map_swapcount == 1) {
2734 2735 2736 2737 2738 2739 2740
				/*
				 * The page is all ours. Move it to
				 * our anon_vma so the rmap code will
				 * not search our parent or siblings.
				 * Protected against the rmap code by
				 * the page lock.
				 */
J
Jan Kara 已提交
2741
				page_move_anon_rmap(vmf->page, vma);
2742
			}
J
Jan Kara 已提交
2743
			unlock_page(vmf->page);
2744 2745
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2746
		}
J
Jan Kara 已提交
2747
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2748
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2749
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2750
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2751 2752 2753 2754 2755
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2756
	get_page(vmf->page);
2757

J
Jan Kara 已提交
2758
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2759
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2760 2761
}

2762
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2763 2764 2765
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2766
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2767 2768
}

2769
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
L
Linus Torvalds 已提交
2770 2771 2772 2773 2774
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2775
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2776 2777 2778
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2779
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2780 2781 2782 2783 2784 2785 2786
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2787
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2788 2789
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2790
				details);
L
Linus Torvalds 已提交
2791 2792 2793 2794
	}
}

/**
2795 2796 2797 2798
 * 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
2799
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2800 2801
 * @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
2802
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813
 * 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
2814
	struct zap_details details = { };
L
Linus Torvalds 已提交
2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
	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;
	}

T
Tobin C Harding 已提交
2826
	details.check_mapping = even_cows ? NULL : mapping;
L
Linus Torvalds 已提交
2827 2828 2829 2830 2831
	details.first_index = hba;
	details.last_index = hba + hlen - 1;
	if (details.last_index < details.first_index)
		details.last_index = ULONG_MAX;

2832
	i_mmap_lock_write(mapping);
2833
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
L
Linus Torvalds 已提交
2834
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2835
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2836 2837 2838 2839
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2840 2841
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2842 2843 2844 2845
 * 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
2846
 */
J
Jan Kara 已提交
2847
int do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2848
{
J
Jan Kara 已提交
2849
	struct vm_area_struct *vma = vmf->vma;
2850
	struct page *page = NULL, *swapcache = NULL;
2851
	struct mem_cgroup *memcg;
H
Huang Ying 已提交
2852
	struct vma_swap_readahead swap_ra;
2853
	swp_entry_t entry;
L
Linus Torvalds 已提交
2854
	pte_t pte;
2855
	int locked;
2856
	int exclusive = 0;
N
Nick Piggin 已提交
2857
	int ret = 0;
H
Huang Ying 已提交
2858
	bool vma_readahead = swap_use_vma_readahead();
L
Linus Torvalds 已提交
2859

H
Huang Ying 已提交
2860 2861 2862 2863 2864
	if (vma_readahead)
		page = swap_readahead_detect(vmf, &swap_ra);
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte)) {
		if (page)
			put_page(page);
2865
		goto out;
H
Huang Ying 已提交
2866
	}
2867

J
Jan Kara 已提交
2868
	entry = pte_to_swp_entry(vmf->orig_pte);
2869 2870
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2871 2872
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2873 2874 2875 2876 2877 2878 2879 2880
		} else if (is_device_private_entry(entry)) {
			/*
			 * For un-addressable device memory we call the pgmap
			 * fault handler callback. The callback must migrate
			 * the page back to some CPU accessible page.
			 */
			ret = device_private_entry_fault(vma, vmf->address, entry,
						 vmf->flags, vmf->pmd);
2881 2882 2883
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2884
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2885
			ret = VM_FAULT_SIGBUS;
2886
		}
2887 2888
		goto out;
	}
2889 2890


2891
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
H
Huang Ying 已提交
2892 2893 2894
	if (!page)
		page = lookup_swap_cache(entry, vma_readahead ? vma : NULL,
					 vmf->address);
L
Linus Torvalds 已提交
2895
	if (!page) {
2896 2897
		struct swap_info_struct *si = swp_swap_info(entry);

2898 2899
		if (si->flags & SWP_SYNCHRONOUS_IO &&
				__swap_count(si, entry) == 1) {
2900 2901 2902 2903 2904 2905 2906 2907 2908
			/* skip swapcache */
			page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
			if (page) {
				__SetPageLocked(page);
				__SetPageSwapBacked(page);
				set_page_private(page, entry.val);
				lru_cache_add_anon(page);
				swap_readpage(page, true);
			}
2909 2910 2911 2912 2913 2914 2915 2916
		} else {
			if (vma_readahead)
				page = do_swap_page_readahead(entry,
					GFP_HIGHUSER_MOVABLE, vmf, &swap_ra);
			else
				page = swapin_readahead(entry,
				       GFP_HIGHUSER_MOVABLE, vma, vmf->address);
			swapcache = page;
2917 2918
		}

L
Linus Torvalds 已提交
2919 2920
		if (!page) {
			/*
2921 2922
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2923
			 */
J
Jan Kara 已提交
2924 2925
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2926
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2927
				ret = VM_FAULT_OOM;
2928
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2929
			goto unlock;
L
Linus Torvalds 已提交
2930 2931 2932 2933
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2934
		count_vm_event(PGMAJFAULT);
2935
		count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
2936
	} else if (PageHWPoison(page)) {
2937 2938 2939 2940
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2941 2942
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2943
		swapcache = page;
2944
		goto out_release;
L
Linus Torvalds 已提交
2945 2946
	}

J
Jan Kara 已提交
2947
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2948

2949
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2950 2951 2952 2953
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2954

A
Andrea Arcangeli 已提交
2955
	/*
2956 2957 2958 2959
	 * 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
2960
	 */
2961 2962
	if (unlikely((!PageSwapCache(page) ||
			page_private(page) != entry.val)) && swapcache)
A
Andrea Arcangeli 已提交
2963 2964
		goto out_page;

J
Jan Kara 已提交
2965
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2966 2967 2968 2969
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2970 2971
	}

K
Kirill A. Shutemov 已提交
2972 2973
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2974
		ret = VM_FAULT_OOM;
2975
		goto out_page;
2976 2977
	}

L
Linus Torvalds 已提交
2978
	/*
2979
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2980
	 */
J
Jan Kara 已提交
2981 2982
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2983
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2984 2985 2986 2987 2988
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2989 2990
	}

2991 2992 2993 2994 2995 2996 2997 2998 2999
	/*
	 * 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
3000

K
Kirill A. Shutemov 已提交
3001 3002
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
3003
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
3004
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
3005
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
3006
		vmf->flags &= ~FAULT_FLAG_WRITE;
3007
		ret |= VM_FAULT_WRITE;
3008
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
3009 3010
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
3011
	if (pte_swp_soft_dirty(vmf->orig_pte))
3012
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
3013
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
3014
	vmf->orig_pte = pte;
3015 3016 3017

	/* ksm created a completely new copy */
	if (unlikely(page != swapcache && swapcache)) {
J
Jan Kara 已提交
3018
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3019
		mem_cgroup_commit_charge(page, memcg, false, false);
3020
		lru_cache_add_active_or_unevictable(page, vma);
3021 3022 3023 3024
	} else {
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
		mem_cgroup_commit_charge(page, memcg, true, false);
		activate_page(page);
3025
	}
L
Linus Torvalds 已提交
3026

3027
	swap_free(entry);
3028 3029
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
3030
		try_to_free_swap(page);
3031
	unlock_page(page);
3032
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
3033 3034 3035 3036 3037 3038 3039 3040 3041
		/*
		 * 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);
3042
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3043
	}
3044

J
Jan Kara 已提交
3045
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
3046
		ret |= do_wp_page(vmf);
3047 3048
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
3049 3050 3051 3052
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3053
	update_mmu_cache(vma, vmf->address, vmf->pte);
3054
unlock:
J
Jan Kara 已提交
3055
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
3056 3057
out:
	return ret;
3058
out_nomap:
3059
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
3060
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3061
out_page:
3062
	unlock_page(page);
3063
out_release:
3064
	put_page(page);
3065
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
3066
		unlock_page(swapcache);
3067
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3068
	}
3069
	return ret;
L
Linus Torvalds 已提交
3070 3071 3072
}

/*
3073 3074 3075
 * 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
3076
 */
J
Jan Kara 已提交
3077
static int do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3078
{
J
Jan Kara 已提交
3079
	struct vm_area_struct *vma = vmf->vma;
3080
	struct mem_cgroup *memcg;
3081
	struct page *page;
3082
	int ret = 0;
L
Linus Torvalds 已提交
3083 3084
	pte_t entry;

3085 3086 3087 3088
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

3089 3090 3091 3092 3093 3094 3095 3096 3097 3098
	/*
	 * Use pte_alloc() instead of pte_alloc_map().  We can't run
	 * pte_offset_map() on pmds where a huge pmd might be created
	 * from a different thread.
	 *
	 * pte_alloc_map() is safe to use under down_write(mmap_sem) or when
	 * parallel threads are excluded by other means.
	 *
	 * Here we only have down_read(mmap_sem).
	 */
J
Jan Kara 已提交
3099
	if (pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))
3100 3101 3102
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3103
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
3104 3105
		return 0;

3106
	/* Use the zero-page for reads */
J
Jan Kara 已提交
3107
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3108
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
3109
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
3110
						vma->vm_page_prot));
J
Jan Kara 已提交
3111 3112 3113
		vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
				vmf->address, &vmf->ptl);
		if (!pte_none(*vmf->pte))
H
Hugh Dickins 已提交
3114
			goto unlock;
3115 3116 3117
		ret = check_stable_address_space(vma->vm_mm);
		if (ret)
			goto unlock;
3118 3119
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3120 3121
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
3122
		}
H
Hugh Dickins 已提交
3123 3124 3125
		goto setpte;
	}

N
Nick Piggin 已提交
3126 3127 3128
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
3129
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
3130 3131
	if (!page)
		goto oom;
3132

K
Kirill A. Shutemov 已提交
3133
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
3134 3135
		goto oom_free_page;

3136 3137 3138 3139 3140
	/*
	 * 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
3141
	__SetPageUptodate(page);
3142

N
Nick Piggin 已提交
3143
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
3144 3145
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
3146

J
Jan Kara 已提交
3147 3148 3149
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
	if (!pte_none(*vmf->pte))
N
Nick Piggin 已提交
3150
		goto release;
H
Hugh Dickins 已提交
3151

3152 3153 3154 3155
	ret = check_stable_address_space(vma->vm_mm);
	if (ret)
		goto release;

3156 3157
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3158
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3159
		mem_cgroup_cancel_charge(page, memcg, false);
3160
		put_page(page);
J
Jan Kara 已提交
3161
		return handle_userfault(vmf, VM_UFFD_MISSING);
3162 3163
	}

K
Kirill A. Shutemov 已提交
3164
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3165
	page_add_new_anon_rmap(page, vma, vmf->address, false);
3166
	mem_cgroup_commit_charge(page, memcg, false, false);
3167
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
3168
setpte:
J
Jan Kara 已提交
3169
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
3170 3171

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3172
	update_mmu_cache(vma, vmf->address, vmf->pte);
3173
unlock:
J
Jan Kara 已提交
3174
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3175
	return ret;
3176
release:
3177
	mem_cgroup_cancel_charge(page, memcg, false);
3178
	put_page(page);
3179
	goto unlock;
3180
oom_free_page:
3181
	put_page(page);
3182
oom:
L
Linus Torvalds 已提交
3183 3184 3185
	return VM_FAULT_OOM;
}

3186 3187 3188 3189 3190
/*
 * The mmap_sem must have been held on entry, and may have been
 * released depending on flags and vma->vm_ops->fault() return value.
 * See filemap_fault() and __lock_page_retry().
 */
J
Jan Kara 已提交
3191
static int __do_fault(struct vm_fault *vmf)
3192
{
J
Jan Kara 已提交
3193
	struct vm_area_struct *vma = vmf->vma;
3194 3195
	int ret;

3196
	ret = vma->vm_ops->fault(vmf);
3197
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
3198
			    VM_FAULT_DONE_COW)))
3199
		return ret;
3200

3201
	if (unlikely(PageHWPoison(vmf->page))) {
3202
		if (ret & VM_FAULT_LOCKED)
3203 3204
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
3205
		vmf->page = NULL;
3206 3207 3208 3209
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
3210
		lock_page(vmf->page);
3211
	else
3212
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
3213 3214 3215 3216

	return ret;
}

3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
/*
 * The ordering of these checks is important for pmds with _PAGE_DEVMAP set.
 * If we check pmd_trans_unstable() first we will trip the bad_pmd() check
 * inside of pmd_none_or_trans_huge_or_clear_bad(). This will end up correctly
 * returning 1 but not before it spams dmesg with the pmd_clear_bad() output.
 */
static int pmd_devmap_trans_unstable(pmd_t *pmd)
{
	return pmd_devmap(*pmd) || pmd_trans_unstable(pmd);
}

J
Jan Kara 已提交
3228
static int pte_alloc_one_map(struct vm_fault *vmf)
3229
{
J
Jan Kara 已提交
3230
	struct vm_area_struct *vma = vmf->vma;
3231

J
Jan Kara 已提交
3232
	if (!pmd_none(*vmf->pmd))
3233
		goto map_pte;
J
Jan Kara 已提交
3234 3235 3236 3237
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
3238 3239 3240
			goto map_pte;
		}

3241
		mm_inc_nr_ptes(vma->vm_mm);
J
Jan Kara 已提交
3242 3243
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
3244
		vmf->prealloc_pte = NULL;
J
Jan Kara 已提交
3245
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))) {
3246 3247 3248 3249 3250
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
3251 3252 3253 3254 3255 3256 3257 3258
	 * pmd_trans_unstable() via pmd_devmap_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.
3259
	 */
3260
	if (pmd_devmap_trans_unstable(vmf->pmd))
3261 3262
		return VM_FAULT_NOPAGE;

3263 3264 3265 3266 3267 3268 3269 3270 3271
	/*
	 * At this point we know that our vmf->pmd points to a page of ptes
	 * and it cannot become pmd_none(), pmd_devmap() or pmd_trans_huge()
	 * for the duration of the fault.  If a racing MADV_DONTNEED runs and
	 * we zap the ptes pointed to by our vmf->pmd, the vmf->ptl will still
	 * be valid and we will re-check to make sure the vmf->pte isn't
	 * pte_none() under vmf->ptl protection when we return to
	 * alloc_set_pte().
	 */
J
Jan Kara 已提交
3272 3273
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
3274 3275 3276
	return 0;
}

3277
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290

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

J
Jan Kara 已提交
3291
static void deposit_prealloc_pte(struct vm_fault *vmf)
3292
{
J
Jan Kara 已提交
3293
	struct vm_area_struct *vma = vmf->vma;
3294

J
Jan Kara 已提交
3295
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
3296 3297 3298 3299
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
3300
	mm_inc_nr_ptes(vma->vm_mm);
3301
	vmf->prealloc_pte = NULL;
3302 3303
}

J
Jan Kara 已提交
3304
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3305
{
J
Jan Kara 已提交
3306 3307 3308
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
	unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
K
Kirill A. Shutemov 已提交
3309 3310 3311 3312 3313 3314 3315 3316 3317
	pmd_t entry;
	int i, ret;

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

3318 3319 3320 3321
	/*
	 * Archs like ppc64 need additonal space to store information
	 * related to pte entry. Use the preallocated table for that.
	 */
J
Jan Kara 已提交
3322 3323 3324
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm, vmf->address);
		if (!vmf->prealloc_pte)
3325 3326 3327 3328
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
3329 3330
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3331 3332 3333 3334 3335 3336 3337
		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)
3338
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3339 3340 3341

	add_mm_counter(vma->vm_mm, MM_FILEPAGES, HPAGE_PMD_NR);
	page_add_file_rmap(page, true);
3342 3343 3344 3345
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3346
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3347

J
Jan Kara 已提交
3348
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3349

J
Jan Kara 已提交
3350
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3351 3352 3353

	/* fault is handled */
	ret = 0;
3354
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3355
out:
J
Jan Kara 已提交
3356
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3357 3358 3359
	return ret;
}
#else
J
Jan Kara 已提交
3360
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3361 3362 3363 3364 3365 3366
{
	BUILD_BUG();
	return 0;
}
#endif

3367
/**
3368 3369
 * 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.
3370
 *
J
Jan Kara 已提交
3371
 * @vmf: fault environment
3372
 * @memcg: memcg to charge page (only for private mappings)
3373 3374
 * @page: page to map
 *
J
Jan Kara 已提交
3375 3376
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3377 3378 3379 3380
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
J
Jan Kara 已提交
3381
int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3382
		struct page *page)
3383
{
J
Jan Kara 已提交
3384 3385
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3386
	pte_t entry;
K
Kirill A. Shutemov 已提交
3387 3388
	int ret;

J
Jan Kara 已提交
3389
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3390
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3391 3392 3393
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3394
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3395
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3396
			return ret;
K
Kirill A. Shutemov 已提交
3397
	}
3398

J
Jan Kara 已提交
3399 3400
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3401
		if (ret)
H
Hugh Dickins 已提交
3402
			return ret;
3403 3404 3405
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3406 3407
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3408

3409 3410 3411 3412
	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
3413 3414
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3415
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3416
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3417 3418
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3419
	} else {
3420
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3421
		page_add_file_rmap(page, false);
3422
	}
J
Jan Kara 已提交
3423
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3424 3425

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3426
	update_mmu_cache(vma, vmf->address, vmf->pte);
3427

H
Hugh Dickins 已提交
3428
	return 0;
3429 3430
}

3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448

/**
 * finish_fault - finish page fault once we have prepared the page to fault
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a page fault once the
 * page to fault in is prepared. It handles locking of PTEs, inserts PTE for
 * given page, adds reverse page mapping, handles memcg charges and LRU
 * addition. The function returns 0 on success, VM_FAULT_ code in case of
 * error.
 *
 * The function expects the page to be locked and on success it consumes a
 * reference of a page being mapped (for the PTE which maps it).
 */
int finish_fault(struct vm_fault *vmf)
{
	struct page *page;
3449
	int ret = 0;
3450 3451 3452 3453 3454 3455 3456

	/* Did we COW the page? */
	if ((vmf->flags & FAULT_FLAG_WRITE) &&
	    !(vmf->vma->vm_flags & VM_SHARED))
		page = vmf->cow_page;
	else
		page = vmf->page;
3457 3458 3459 3460 3461 3462 3463 3464 3465

	/*
	 * check even for read faults because we might have lost our CoWed
	 * page
	 */
	if (!(vmf->vma->vm_flags & VM_SHARED))
		ret = check_stable_address_space(vmf->vma->vm_mm);
	if (!ret)
		ret = alloc_set_pte(vmf, vmf->memcg, page);
3466 3467 3468 3469 3470
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3471 3472
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3473 3474 3475

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3476
{
3477
	*val = fault_around_bytes;
3478 3479 3480
	return 0;
}

3481 3482 3483 3484 3485
/*
 * 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.
 */
3486
static int fault_around_bytes_set(void *data, u64 val)
3487
{
3488
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3489
		return -EINVAL;
3490 3491 3492 3493
	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 */
3494 3495
	return 0;
}
3496
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
3497
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3498 3499 3500 3501 3502

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

3503
	ret = debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
3504
			&fault_around_bytes_fops);
3505
	if (!ret)
3506
		pr_warn("Failed to create fault_around_bytes in debugfs");
3507 3508 3509 3510
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3511

3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534
/*
 * 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.
 */
3535
static int do_fault_around(struct vm_fault *vmf)
3536
{
J
Jan Kara 已提交
3537
	unsigned long address = vmf->address, nr_pages, mask;
3538
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3539
	pgoff_t end_pgoff;
3540
	int off, ret = 0;
3541

3542
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3543 3544
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3545 3546
	vmf->address = max(address & mask, vmf->vma->vm_start);
	off = ((address - vmf->address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
K
Kirill A. Shutemov 已提交
3547
	start_pgoff -= off;
3548 3549

	/*
K
Kirill A. Shutemov 已提交
3550 3551
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3552
	 */
K
Kirill A. Shutemov 已提交
3553
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3554
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3555
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3556
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3557
			start_pgoff + nr_pages - 1);
3558

J
Jan Kara 已提交
3559 3560 3561 3562
	if (pmd_none(*vmf->pmd)) {
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm,
						  vmf->address);
		if (!vmf->prealloc_pte)
3563
			goto out;
3564
		smp_wmb(); /* See comment in __pte_alloc() */
3565 3566
	}

J
Jan Kara 已提交
3567
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3568 3569

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3570
	if (pmd_trans_huge(*vmf->pmd)) {
3571 3572 3573 3574 3575
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3576
	if (!vmf->pte)
3577 3578 3579
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3580 3581
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3582
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3583
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3584
out:
J
Jan Kara 已提交
3585 3586
	vmf->address = address;
	vmf->pte = NULL;
3587
	return ret;
3588 3589
}

3590
static int do_read_fault(struct vm_fault *vmf)
3591
{
J
Jan Kara 已提交
3592
	struct vm_area_struct *vma = vmf->vma;
3593 3594 3595 3596 3597 3598 3599
	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).
	 */
3600
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3601
		ret = do_fault_around(vmf);
3602 3603
		if (ret)
			return ret;
3604
	}
3605

J
Jan Kara 已提交
3606
	ret = __do_fault(vmf);
3607 3608 3609
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3610
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3611
	unlock_page(vmf->page);
3612
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3613
		put_page(vmf->page);
3614 3615 3616
	return ret;
}

3617
static int do_cow_fault(struct vm_fault *vmf)
3618
{
J
Jan Kara 已提交
3619
	struct vm_area_struct *vma = vmf->vma;
3620 3621 3622 3623 3624
	int ret;

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

J
Jan Kara 已提交
3625 3626
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3627 3628
		return VM_FAULT_OOM;

J
Jan Kara 已提交
3629
	if (mem_cgroup_try_charge(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3630
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3631
		put_page(vmf->cow_page);
3632 3633 3634
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3635
	ret = __do_fault(vmf);
3636 3637
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3638 3639
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3640

3641
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3642
	__SetPageUptodate(vmf->cow_page);
3643

3644
	ret |= finish_fault(vmf);
3645 3646
	unlock_page(vmf->page);
	put_page(vmf->page);
3647 3648
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3649 3650
	return ret;
uncharge_out:
3651
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3652
	put_page(vmf->cow_page);
3653 3654 3655
	return ret;
}

3656
static int do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3657
{
J
Jan Kara 已提交
3658
	struct vm_area_struct *vma = vmf->vma;
3659
	int ret, tmp;
3660

J
Jan Kara 已提交
3661
	ret = __do_fault(vmf);
3662
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3663
		return ret;
L
Linus Torvalds 已提交
3664 3665

	/*
3666 3667
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3668
	 */
3669
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3670
		unlock_page(vmf->page);
3671
		tmp = do_page_mkwrite(vmf);
3672 3673
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3674
			put_page(vmf->page);
3675
			return tmp;
3676
		}
3677 3678
	}

3679
	ret |= finish_fault(vmf);
3680 3681
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3682 3683
		unlock_page(vmf->page);
		put_page(vmf->page);
3684
		return ret;
L
Linus Torvalds 已提交
3685
	}
N
Nick Piggin 已提交
3686

3687
	fault_dirty_shared_page(vma, vmf->page);
3688
	return ret;
3689
}
3690

3691 3692 3693 3694 3695 3696
/*
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults).
 * The mmap_sem may have been released depending on flags and our
 * return value.  See filemap_fault() and __lock_page_or_retry().
 */
J
Jan Kara 已提交
3697
static int do_fault(struct vm_fault *vmf)
3698
{
J
Jan Kara 已提交
3699
	struct vm_area_struct *vma = vmf->vma;
H
Hugh Dickins 已提交
3700
	int ret;
3701

3702 3703
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
H
Hugh Dickins 已提交
3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714
		ret = VM_FAULT_SIGBUS;
	else if (!(vmf->flags & FAULT_FLAG_WRITE))
		ret = do_read_fault(vmf);
	else if (!(vma->vm_flags & VM_SHARED))
		ret = do_cow_fault(vmf);
	else
		ret = do_shared_fault(vmf);

	/* preallocated pagetable is unused: free it */
	if (vmf->prealloc_pte) {
		pte_free(vma->vm_mm, vmf->prealloc_pte);
3715
		vmf->prealloc_pte = NULL;
H
Hugh Dickins 已提交
3716 3717
	}
	return ret;
3718 3719
}

3720
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3721 3722
				unsigned long addr, int page_nid,
				int *flags)
3723 3724 3725 3726
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3727
	if (page_nid == numa_node_id()) {
3728
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3729 3730
		*flags |= TNF_FAULT_LOCAL;
	}
3731 3732 3733 3734

	return mpol_misplaced(page, vma, addr);
}

J
Jan Kara 已提交
3735
static int do_numa_page(struct vm_fault *vmf)
3736
{
J
Jan Kara 已提交
3737
	struct vm_area_struct *vma = vmf->vma;
3738
	struct page *page = NULL;
3739
	int page_nid = -1;
3740
	int last_cpupid;
3741
	int target_nid;
3742
	bool migrated = false;
3743
	pte_t pte;
3744
	bool was_writable = pte_savedwrite(vmf->orig_pte);
3745
	int flags = 0;
3746 3747

	/*
T
Tobin C Harding 已提交
3748 3749 3750 3751
	 * 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.
	 */
J
Jan Kara 已提交
3752 3753
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
3754
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
J
Jan Kara 已提交
3755
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3756 3757 3758
		goto out;
	}

3759 3760 3761 3762 3763
	/*
	 * Make it present again, Depending on how arch implementes non
	 * accessible ptes, some can allow access by kernel mode.
	 */
	pte = ptep_modify_prot_start(vma->vm_mm, vmf->address, vmf->pte);
3764 3765
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3766 3767
	if (was_writable)
		pte = pte_mkwrite(pte);
3768
	ptep_modify_prot_commit(vma->vm_mm, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
3769
	update_mmu_cache(vma, vmf->address, vmf->pte);
3770

J
Jan Kara 已提交
3771
	page = vm_normal_page(vma, vmf->address, pte);
3772
	if (!page) {
J
Jan Kara 已提交
3773
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3774 3775 3776
		return 0;
	}

3777 3778
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3779
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3780 3781 3782
		return 0;
	}

3783
	/*
3784 3785 3786 3787 3788 3789
	 * 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.
3790
	 */
3791
	if (!pte_write(pte))
3792 3793
		flags |= TNF_NO_GROUP;

3794 3795 3796 3797 3798 3799 3800
	/*
	 * 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;

3801
	last_cpupid = page_cpupid_last(page);
3802
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3803
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3804
			&flags);
J
Jan Kara 已提交
3805
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3806 3807 3808 3809 3810 3811
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3812
	migrated = migrate_misplaced_page(page, vma, target_nid);
3813
	if (migrated) {
3814
		page_nid = target_nid;
3815
		flags |= TNF_MIGRATED;
3816 3817
	} else
		flags |= TNF_MIGRATE_FAIL;
3818 3819

out:
3820
	if (page_nid != -1)
3821
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3822 3823 3824
	return 0;
}

3825
static inline int create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3826
{
3827
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3828
		return do_huge_pmd_anonymous_page(vmf);
3829
	if (vmf->vma->vm_ops->huge_fault)
3830
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
M
Matthew Wilcox 已提交
3831 3832 3833
	return VM_FAULT_FALLBACK;
}

3834 3835
/* `inline' is required to avoid gcc 4.1.2 build error */
static inline int wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3836
{
J
Jan Kara 已提交
3837 3838
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3839
	if (vmf->vma->vm_ops->huge_fault)
3840
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
K
Kirill A. Shutemov 已提交
3841 3842

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3843 3844
	VM_BUG_ON_VMA(vmf->vma->vm_flags & VM_SHARED, vmf->vma);
	__split_huge_pmd(vmf->vma, vmf->pmd, vmf->address, false, NULL);
K
Kirill A. Shutemov 已提交
3845

M
Matthew Wilcox 已提交
3846 3847 3848
	return VM_FAULT_FALLBACK;
}

3849 3850 3851 3852 3853
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3854 3855 3856 3857 3858 3859 3860
static int create_huge_pud(struct vm_fault *vmf)
{
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/* No support for anonymous transparent PUD pages yet */
	if (vma_is_anonymous(vmf->vma))
		return VM_FAULT_FALLBACK;
	if (vmf->vma->vm_ops->huge_fault)
3861
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

static int wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
{
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/* No support for anonymous transparent PUD pages yet */
	if (vma_is_anonymous(vmf->vma))
		return VM_FAULT_FALLBACK;
	if (vmf->vma->vm_ops->huge_fault)
3873
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3874 3875 3876 3877
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3878 3879 3880 3881 3882 3883 3884 3885 3886
/*
 * 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).
 *
3887 3888
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3889
 *
3890 3891
 * 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
3892
 */
J
Jan Kara 已提交
3893
static int handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3894 3895 3896
{
	pte_t entry;

J
Jan Kara 已提交
3897
	if (unlikely(pmd_none(*vmf->pmd))) {
3898 3899 3900 3901 3902 3903
		/*
		 * Leave __pte_alloc() until later: because vm_ops->fault may
		 * want to allocate huge page, and if we expose page table
		 * for an instant, it will be difficult to retract from
		 * concurrent faults and from rmap lookups.
		 */
J
Jan Kara 已提交
3904
		vmf->pte = NULL;
3905 3906
	} else {
		/* See comment in pte_alloc_one_map() */
3907
		if (pmd_devmap_trans_unstable(vmf->pmd))
3908 3909 3910 3911 3912 3913 3914
			return 0;
		/*
		 * A regular pmd is established and it can't morph into a huge
		 * pmd from under us anymore at this point because we hold the
		 * mmap_sem read mode and khugepaged takes it in write mode.
		 * So now it's safe to run pte_offset_map().
		 */
J
Jan Kara 已提交
3915
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3916
		vmf->orig_pte = *vmf->pte;
3917 3918 3919 3920

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
3921 3922 3923
		 * CONFIG_32BIT=y, so READ_ONCE cannot guarantee atomic
		 * accesses.  The code below just needs a consistent view
		 * for the ifs and we later double check anyway with the
3924 3925 3926
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
J
Jan Kara 已提交
3927
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3928 3929
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3930
		}
L
Linus Torvalds 已提交
3931 3932
	}

J
Jan Kara 已提交
3933 3934 3935
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3936
		else
J
Jan Kara 已提交
3937
			return do_fault(vmf);
3938 3939
	}

J
Jan Kara 已提交
3940 3941
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3942

J
Jan Kara 已提交
3943 3944
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3945

J
Jan Kara 已提交
3946 3947
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3948
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3949
	if (unlikely(!pte_same(*vmf->pte, entry)))
3950
		goto unlock;
J
Jan Kara 已提交
3951
	if (vmf->flags & FAULT_FLAG_WRITE) {
3952
		if (!pte_write(entry))
J
Jan Kara 已提交
3953
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3954 3955 3956
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3957 3958 3959
	if (ptep_set_access_flags(vmf->vma, vmf->address, vmf->pte, entry,
				vmf->flags & FAULT_FLAG_WRITE)) {
		update_mmu_cache(vmf->vma, vmf->address, vmf->pte);
3960 3961 3962 3963 3964 3965 3966
	} else {
		/*
		 * This is needed only for protection faults but the arch code
		 * is not yet telling us if this is a protection fault or not.
		 * This still avoids useless tlb flushes for .text page faults
		 * with threads.
		 */
J
Jan Kara 已提交
3967 3968
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3969
	}
3970
unlock:
J
Jan Kara 已提交
3971
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3972
	return 0;
L
Linus Torvalds 已提交
3973 3974 3975 3976
}

/*
 * By the time we get here, we already hold the mm semaphore
3977 3978 3979
 *
 * 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
3980
 */
3981 3982
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3983
{
J
Jan Kara 已提交
3984
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3985
		.vma = vma,
3986
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3987
		.flags = flags,
3988
		.pgoff = linear_page_index(vma, address),
3989
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3990
	};
3991
	unsigned int dirty = flags & FAULT_FLAG_WRITE;
3992
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3993
	pgd_t *pgd;
3994
	p4d_t *p4d;
3995
	int ret;
L
Linus Torvalds 已提交
3996 3997

	pgd = pgd_offset(mm, address);
3998 3999 4000
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return VM_FAULT_OOM;
4001

4002
	vmf.pud = pud_alloc(mm, p4d, address);
4003
	if (!vmf.pud)
H
Hugh Dickins 已提交
4004
		return VM_FAULT_OOM;
4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016
	if (pud_none(*vmf.pud) && transparent_hugepage_enabled(vma)) {
		ret = create_huge_pud(&vmf);
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
	} else {
		pud_t orig_pud = *vmf.pud;

		barrier();
		if (pud_trans_huge(orig_pud) || pud_devmap(orig_pud)) {

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

4017
			if (dirty && !pud_write(orig_pud)) {
4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028
				ret = wp_huge_pud(&vmf, orig_pud);
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
			} else {
				huge_pud_set_accessed(&vmf, orig_pud);
				return 0;
			}
		}
	}

	vmf.pmd = pmd_alloc(mm, vmf.pud, address);
J
Jan Kara 已提交
4029
	if (!vmf.pmd)
H
Hugh Dickins 已提交
4030
		return VM_FAULT_OOM;
J
Jan Kara 已提交
4031
	if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) {
4032
		ret = create_huge_pmd(&vmf);
4033 4034
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
4035
	} else {
J
Jan Kara 已提交
4036
		pmd_t orig_pmd = *vmf.pmd;
4037

4038
		barrier();
4039 4040 4041 4042 4043 4044 4045
		if (unlikely(is_swap_pmd(orig_pmd))) {
			VM_BUG_ON(thp_migration_supported() &&
					  !is_pmd_migration_entry(orig_pmd));
			if (is_pmd_migration_entry(orig_pmd))
				pmd_migration_entry_wait(mm, vmf.pmd);
			return 0;
		}
4046
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
4047
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
4048
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
4049

4050
			if (dirty && !pmd_write(orig_pmd)) {
J
Jan Kara 已提交
4051
				ret = wp_huge_pmd(&vmf, orig_pmd);
4052 4053
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
4054
			} else {
J
Jan Kara 已提交
4055
				huge_pmd_set_accessed(&vmf, orig_pmd);
4056
				return 0;
4057
			}
4058 4059 4060
		}
	}

J
Jan Kara 已提交
4061
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
4062 4063
}

4064 4065 4066 4067 4068 4069
/*
 * 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().
 */
4070 4071
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
4072 4073 4074 4075 4076 4077
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
4078
	count_memcg_event_mm(vma->vm_mm, PGFAULT);
4079 4080 4081 4082

	/* do counter updates before entering really critical section. */
	check_sync_rss_stat(current);

4083 4084 4085 4086 4087
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

4088 4089 4090 4091 4092
	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
4093
		mem_cgroup_oom_enable();
4094

K
Kirill A. Shutemov 已提交
4095 4096 4097 4098
	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
4099

4100 4101
	if (flags & FAULT_FLAG_USER) {
		mem_cgroup_oom_disable();
T
Tobin C Harding 已提交
4102 4103 4104 4105 4106 4107 4108 4109
		/*
		 * 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);
4110
	}
4111

4112 4113
	return ret;
}
4114
EXPORT_SYMBOL_GPL(handle_mm_fault);
4115

K
Kirill A. Shutemov 已提交
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138
#ifndef __PAGETABLE_P4D_FOLDED
/*
 * Allocate p4d page table.
 * We've already handled the fast-path in-line.
 */
int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
{
	p4d_t *new = p4d_alloc_one(mm, address);
	if (!new)
		return -ENOMEM;

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

	spin_lock(&mm->page_table_lock);
	if (pgd_present(*pgd))		/* Another has populated it */
		p4d_free(mm, new);
	else
		pgd_populate(mm, pgd, new);
	spin_unlock(&mm->page_table_lock);
	return 0;
}
#endif /* __PAGETABLE_P4D_FOLDED */

L
Linus Torvalds 已提交
4139 4140 4141
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
4142
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4143
 */
4144
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
L
Linus Torvalds 已提交
4145
{
H
Hugh Dickins 已提交
4146 4147
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
4148
		return -ENOMEM;
L
Linus Torvalds 已提交
4149

4150 4151
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
4152
	spin_lock(&mm->page_table_lock);
4153
#ifndef __ARCH_HAS_5LEVEL_HACK
K
Kirill A. Shutemov 已提交
4154 4155
	if (!p4d_present(*p4d)) {
		mm_inc_nr_puds(mm);
4156
		p4d_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4157
	} else	/* Another has populated it */
4158
		pud_free(mm, new);
K
Kirill A. Shutemov 已提交
4159 4160 4161
#else
	if (!pgd_present(*p4d)) {
		mm_inc_nr_puds(mm);
4162
		pgd_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4163 4164
	} else	/* Another has populated it */
		pud_free(mm, new);
4165
#endif /* __ARCH_HAS_5LEVEL_HACK */
H
Hugh Dickins 已提交
4166
	spin_unlock(&mm->page_table_lock);
4167
	return 0;
L
Linus Torvalds 已提交
4168 4169 4170 4171 4172 4173
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
4174
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4175
 */
4176
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
4177
{
4178
	spinlock_t *ptl;
H
Hugh Dickins 已提交
4179 4180
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
4181
		return -ENOMEM;
L
Linus Torvalds 已提交
4182

4183 4184
	smp_wmb(); /* See comment in __pte_alloc */

4185
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
4186
#ifndef __ARCH_HAS_4LEVEL_HACK
4187 4188
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
4189
		pud_populate(mm, pud, new);
4190
	} else	/* Another has populated it */
4191
		pmd_free(mm, new);
4192 4193 4194
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
4195
		pgd_populate(mm, pud, new);
4196 4197
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
4198
#endif /* __ARCH_HAS_4LEVEL_HACK */
4199
	spin_unlock(ptl);
4200
	return 0;
4201
}
L
Linus Torvalds 已提交
4202 4203
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
4204
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4205 4206
			    unsigned long *start, unsigned long *end,
			    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
4207 4208
{
	pgd_t *pgd;
4209
	p4d_t *p4d;
J
Johannes Weiner 已提交
4210 4211 4212 4213 4214 4215 4216 4217
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

4218 4219 4220 4221 4222
	p4d = p4d_offset(pgd, address);
	if (p4d_none(*p4d) || unlikely(p4d_bad(*p4d)))
		goto out;

	pud = pud_offset(p4d, address);
J
Johannes Weiner 已提交
4223 4224 4225 4226
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
4227
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
4228

R
Ross Zwisler 已提交
4229 4230 4231 4232
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

4233 4234 4235 4236 4237
		if (start && end) {
			*start = address & PMD_MASK;
			*end = *start + PMD_SIZE;
			mmu_notifier_invalidate_range_start(mm, *start, *end);
		}
R
Ross Zwisler 已提交
4238 4239 4240 4241 4242 4243
		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
4244 4245
		if (start && end)
			mmu_notifier_invalidate_range_end(mm, *start, *end);
R
Ross Zwisler 已提交
4246 4247 4248
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
4249 4250
		goto out;

4251 4252 4253 4254 4255
	if (start && end) {
		*start = address & PAGE_MASK;
		*end = *start + PAGE_SIZE;
		mmu_notifier_invalidate_range_start(mm, *start, *end);
	}
J
Johannes Weiner 已提交
4256 4257 4258 4259 4260 4261 4262
	ptep = pte_offset_map_lock(mm, pmd, address, ptlp);
	if (!pte_present(*ptep))
		goto unlock;
	*ptepp = ptep;
	return 0;
unlock:
	pte_unmap_unlock(ptep, *ptlp);
4263 4264
	if (start && end)
		mmu_notifier_invalidate_range_end(mm, *start, *end);
J
Johannes Weiner 已提交
4265 4266 4267 4268
out:
	return -EINVAL;
}

4269 4270
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
4271 4272 4273 4274 4275
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4276 4277
			   !(res = __follow_pte_pmd(mm, address, NULL, NULL,
						    ptepp, NULL, ptlp)));
R
Ross Zwisler 已提交
4278 4279 4280 4281
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4282
			     unsigned long *start, unsigned long *end,
R
Ross Zwisler 已提交
4283 4284 4285 4286 4287 4288
			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4289 4290
			   !(res = __follow_pte_pmd(mm, address, start, end,
						    ptepp, pmdpp, ptlp)));
4291 4292
	return res;
}
R
Ross Zwisler 已提交
4293
EXPORT_SYMBOL(follow_pte_pmd);
4294

J
Johannes Weiner 已提交
4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323
/**
 * 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);

4324
#ifdef CONFIG_HAVE_IOREMAP_PROT
4325 4326 4327
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4328
{
4329
	int ret = -EINVAL;
4330 4331 4332
	pte_t *ptep, pte;
	spinlock_t *ptl;

4333 4334
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4335

4336
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4337
		goto out;
4338
	pte = *ptep;
4339

4340
	if ((flags & FOLL_WRITE) && !pte_write(pte))
4341 4342 4343
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4344
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4345

4346
	ret = 0;
4347 4348 4349
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4350
	return ret;
4351 4352 4353 4354 4355 4356 4357
}

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
4358
	void __iomem *maddr;
4359 4360
	int offset = addr & (PAGE_SIZE-1);

4361
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4362 4363
		return -EINVAL;

4364
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
4365 4366 4367 4368 4369 4370 4371 4372
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
4373
EXPORT_SYMBOL_GPL(generic_access_phys);
4374 4375
#endif

4376
/*
4377 4378
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4379
 */
4380
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4381
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4382 4383 4384
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4385
	int write = gup_flags & FOLL_WRITE;
4386 4387

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4388
	/* ignore errors, just check how much was successfully transferred */
4389 4390 4391
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4392
		struct page *page = NULL;
4393

4394
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4395
				gup_flags, &page, &vma, NULL);
4396
		if (ret <= 0) {
4397 4398 4399
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4400 4401 4402 4403 4404
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4405
			if (!vma || vma->vm_start > addr)
4406 4407 4408 4409 4410 4411 4412
				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;
4413
#endif
4414
		} else {
4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429
			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);
4430
			put_page(page);
4431 4432 4433 4434 4435 4436 4437 4438 4439
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4440

S
Stephen Wilson 已提交
4441
/**
4442
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4443 4444 4445 4446
 * @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
4447
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4448 4449 4450 4451
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4452
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4453
{
4454
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4455 4456
}

4457 4458 4459 4460 4461 4462
/*
 * 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,
4463
		void *buf, int len, unsigned int gup_flags)
4464 4465 4466 4467 4468 4469 4470 4471
{
	struct mm_struct *mm;
	int ret;

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

4472
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4473

4474 4475 4476 4477
	mmput(mm);

	return ret;
}
4478
EXPORT_SYMBOL_GPL(access_process_vm);
4479

4480 4481 4482 4483 4484 4485 4486 4487
/*
 * 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;

4488
	/*
4489
	 * we might be running from an atomic context so we cannot sleep
4490
	 */
4491
	if (!down_read_trylock(&mm->mmap_sem))
4492 4493
		return;

4494 4495 4496
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
4497
		char *buf = (char *)__get_free_page(GFP_NOWAIT);
4498
		if (buf) {
A
Andy Shevchenko 已提交
4499
			char *p;
4500

M
Miklos Szeredi 已提交
4501
			p = file_path(f, buf, PAGE_SIZE);
4502 4503
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4504
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4505 4506 4507 4508 4509
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4510
	up_read(&mm->mmap_sem);
4511
}
4512

4513
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4514
void __might_fault(const char *file, int line)
4515
{
4516 4517 4518 4519 4520 4521
	/*
	 * 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.
	 */
A
Al Viro 已提交
4522
	if (uaccess_kernel())
4523
		return;
4524
	if (pagefault_disabled())
4525
		return;
4526 4527
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4528
	if (current->mm)
4529
		might_lock_read(&current->mm->mmap_sem);
4530
#endif
4531
}
4532
EXPORT_SYMBOL(__might_fault);
4533
#endif
A
Andrea Arcangeli 已提交
4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550

#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,
4551
		     unsigned long addr_hint, unsigned int pages_per_huge_page)
A
Andrea Arcangeli 已提交
4552
{
4553 4554 4555
	int i, n, base, l;
	unsigned long addr = addr_hint &
		~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
A
Andrea Arcangeli 已提交
4556 4557 4558 4559 4560 4561

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

4562
	/* Clear sub-page to access last to keep its cache lines hot */
A
Andrea Arcangeli 已提交
4563
	might_sleep();
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594
	n = (addr_hint - addr) / PAGE_SIZE;
	if (2 * n <= pages_per_huge_page) {
		/* If sub-page to access in first half of huge page */
		base = 0;
		l = n;
		/* Clear sub-pages at the end of huge page */
		for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
			cond_resched();
			clear_user_highpage(page + i, addr + i * PAGE_SIZE);
		}
	} else {
		/* If sub-page to access in second half of huge page */
		base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
		l = pages_per_huge_page - n;
		/* Clear sub-pages at the begin of huge page */
		for (i = 0; i < base; i++) {
			cond_resched();
			clear_user_highpage(page + i, addr + i * PAGE_SIZE);
		}
	}
	/*
	 * Clear remaining sub-pages in left-right-left-right pattern
	 * towards the sub-page to access
	 */
	for (i = 0; i < l; i++) {
		int left_idx = base + i;
		int right_idx = base + 2 * l - 1 - i;

		cond_resched();
		clear_user_highpage(page + left_idx,
				    addr + left_idx * PAGE_SIZE);
A
Andrea Arcangeli 已提交
4595
		cond_resched();
4596 4597
		clear_user_highpage(page + right_idx,
				    addr + right_idx * PAGE_SIZE);
A
Andrea Arcangeli 已提交
4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637
	}
}

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);
	}
}
4638 4639 4640

long copy_huge_page_from_user(struct page *dst_page,
				const void __user *usr_src,
4641 4642
				unsigned int pages_per_huge_page,
				bool allow_pagefault)
4643 4644 4645 4646 4647 4648 4649
{
	void *src = (void *)usr_src;
	void *page_kaddr;
	unsigned long i, rc = 0;
	unsigned long ret_val = pages_per_huge_page * PAGE_SIZE;

	for (i = 0; i < pages_per_huge_page; i++) {
4650 4651 4652 4653
		if (allow_pagefault)
			page_kaddr = kmap(dst_page + i);
		else
			page_kaddr = kmap_atomic(dst_page + i);
4654 4655 4656
		rc = copy_from_user(page_kaddr,
				(const void __user *)(src + i * PAGE_SIZE),
				PAGE_SIZE);
4657 4658 4659 4660
		if (allow_pagefault)
			kunmap(dst_page + i);
		else
			kunmap_atomic(page_kaddr);
4661 4662 4663 4664 4665 4666 4667 4668 4669

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

		cond_resched();
	}
	return ret_val;
}
A
Andrea Arcangeli 已提交
4670
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4671

4672
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4673 4674 4675 4676 4677 4678 4679 4680 4681

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);
}

4682
bool ptlock_alloc(struct page *page)
4683 4684 4685
{
	spinlock_t *ptl;

4686
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4687 4688
	if (!ptl)
		return false;
4689
	page->ptl = ptl;
4690 4691 4692
	return true;
}

4693
void ptlock_free(struct page *page)
4694
{
4695
	kmem_cache_free(page_ptl_cachep, page->ptl);
4696 4697
}
#endif