memory.c 124.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 *  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
	atomic_long_dec(&tlb->mm->nr_ptes);
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);

495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
	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);
}

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);

528 529 530 531 532 533 534
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
535
	}
536 537 538
	if (end - 1 > ceiling - 1)
		return;

539
	p4d = p4d_offset(pgd, start);
540
	pgd_clear(pgd);
541
	p4d_free_tlb(tlb, p4d, start);
L
Linus Torvalds 已提交
542 543 544
}

/*
545
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
546
 */
547
void free_pgd_range(struct mmu_gather *tlb,
548 549
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
550 551 552
{
	pgd_t *pgd;
	unsigned long next;
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578

	/*
	 * 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
579

580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
	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;
595 596 597 598 599
	/*
	 * 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);
600
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
601 602 603 604
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
605
		free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
606
	} while (pgd++, addr = next, addr != end);
607 608
}

609
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
610
		unsigned long floor, unsigned long ceiling)
611 612 613 614 615
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

616
		/*
N
npiggin@suse.de 已提交
617 618
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
619
		 */
620
		unlink_anon_vmas(vma);
621 622
		unlink_file_vma(vma);

623
		if (is_vm_hugetlb_page(vma)) {
624
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
625
				floor, next ? next->vm_start : ceiling);
626 627 628 629 630
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
631
			       && !is_vm_hugetlb_page(next)) {
632 633
				vma = next;
				next = vma->vm_next;
634
				unlink_anon_vmas(vma);
635
				unlink_file_vma(vma);
636 637
			}
			free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
638
				floor, next ? next->vm_start : ceiling);
639
		}
640 641
		vma = next;
	}
L
Linus Torvalds 已提交
642 643
}

644
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
645
{
646
	spinlock_t *ptl;
647
	pgtable_t new = pte_alloc_one(mm, address);
648 649 650
	if (!new)
		return -ENOMEM;

651 652 653 654 655 656 657 658 659 660 661 662 663 664 665
	/*
	 * 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 */

666
	ptl = pmd_lock(mm, pmd);
667
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
668
		atomic_long_inc(&mm->nr_ptes);
L
Linus Torvalds 已提交
669
		pmd_populate(mm, pmd, new);
670
		new = NULL;
671
	}
672
	spin_unlock(ptl);
673 674
	if (new)
		pte_free(mm, new);
675
	return 0;
L
Linus Torvalds 已提交
676 677
}

678
int __pte_alloc_kernel(pmd_t *pmd, unsigned long address)
L
Linus Torvalds 已提交
679
{
680 681 682 683
	pte_t *new = pte_alloc_one_kernel(&init_mm, address);
	if (!new)
		return -ENOMEM;

684 685
	smp_wmb(); /* See comment in __pte_alloc */

686
	spin_lock(&init_mm.page_table_lock);
687
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
688
		pmd_populate_kernel(&init_mm, pmd, new);
689
		new = NULL;
690
	}
691
	spin_unlock(&init_mm.page_table_lock);
692 693
	if (new)
		pte_free_kernel(&init_mm, new);
694
	return 0;
L
Linus Torvalds 已提交
695 696
}

K
KAMEZAWA Hiroyuki 已提交
697 698 699 700 701 702
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)
703
{
K
KAMEZAWA Hiroyuki 已提交
704 705
	int i;

706
	if (current->mm == mm)
707
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
708 709 710
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
711 712
}

N
Nick Piggin 已提交
713
/*
714 715 716
 * 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
717 718 719
 *
 * The calling function must still handle the error.
 */
720 721
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
722
{
723
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
724 725
	p4d_t *p4d = p4d_offset(pgd, addr);
	pud_t *pud = pud_offset(p4d, addr);
726 727 728
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
729 730 731 732 733 734 735 736 737 738 739 740 741 742
	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) {
743 744
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
745 746 747 748 749 750
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
751 752 753 754

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

755 756 757
	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));
758
	if (page)
759
		dump_page(page, "bad pte");
760 761
	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);
762 763 764
	/*
	 * Choose text because data symbols depend on CONFIG_KALLSYMS_ALL=y
	 */
765 766 767 768 769
	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
770
	dump_stack();
771
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
772 773
}

H
Hugh Dickins 已提交
774
/*
N
Nick Piggin 已提交
775
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
776
 *
N
Nick Piggin 已提交
777 778 779
 * "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
780
 *
N
Nick Piggin 已提交
781 782 783 784 785 786 787 788
 * 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.
789
 *
J
Jared Hulbert 已提交
790 791
 * 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
792 793
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
794 795 796
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
797 798 799 800 801 802
 * 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
803 804
 *
 *
N
Nick Piggin 已提交
805
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
806 807 808 809 810 811 812 813 814
 *
 * 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
815
 */
N
Nick Piggin 已提交
816 817 818 819 820
#ifdef __HAVE_ARCH_PTE_SPECIAL
# define HAVE_PTE_SPECIAL 1
#else
# define HAVE_PTE_SPECIAL 0
#endif
821 822
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
823
{
824
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
825 826

	if (HAVE_PTE_SPECIAL) {
827
		if (likely(!pte_special(pte)))
828
			goto check_pfn;
829 830
		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
831 832
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
		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().
		 */
		if (likely(pfn < highest_memmap_pfn)) {
			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
858 859 860 861 862
		return NULL;
	}

	/* !HAVE_PTE_SPECIAL case follows: */

J
Jared Hulbert 已提交
863 864 865 866 867 868
	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
869 870
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
871 872 873 874 875
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
876 877
	}

878 879
	if (is_zero_pfn(pfn))
		return NULL;
880 881 882 883 884
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
885 886

	/*
N
Nick Piggin 已提交
887 888
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
889
	 */
J
Jared Hulbert 已提交
890
out:
891
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
892 893
}

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
#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
934 935 936 937 938 939
/*
 * 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
940
static inline unsigned long
L
Linus Torvalds 已提交
941
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
942
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
943
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
944
{
N
Nick Piggin 已提交
945
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
946 947 948 949 950
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968
		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);

969
			rss[mm_counter(page)]++;
970 971 972 973 974 975 976 977 978 979 980 981

			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);
982
			}
983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
		} 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
1012
		}
1013
		goto out_set_pte;
L
Linus Torvalds 已提交
1014 1015 1016 1017 1018 1019
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
1020
	if (is_cow_mapping(vm_flags)) {
L
Linus Torvalds 已提交
1021
		ptep_set_wrprotect(src_mm, addr, src_pte);
1022
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
1023 1024 1025 1026 1027 1028 1029 1030 1031
	}

	/*
	 * 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);
1032 1033 1034 1035

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
1036
		page_dup_rmap(page, false);
1037
		rss[mm_counter(page)]++;
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
	} 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)]++;
		}
1051
	}
1052 1053 1054

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
1055
	return 0;
L
Linus Torvalds 已提交
1056 1057
}

1058
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
1059 1060
		   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
1061
{
1062
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
1063
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
1064
	spinlock_t *src_ptl, *dst_ptl;
1065
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
1066
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
1067
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
1068 1069

again:
K
KAMEZAWA Hiroyuki 已提交
1070 1071
	init_rss_vec(rss);

H
Hugh Dickins 已提交
1072
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
1073 1074
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
1075
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
1076
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
1077
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
1078 1079
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
1080
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
1087 1088 1089
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
1090
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
1091 1092
				break;
		}
L
Linus Torvalds 已提交
1093 1094 1095 1096
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
1097 1098 1099 1100
		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
1101 1102 1103
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

1104
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1105
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
1106
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
1107
	add_mm_rss_vec(dst_mm, rss);
1108
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
1109
	cond_resched();
H
Hugh Dickins 已提交
1110 1111 1112 1113 1114 1115

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
	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);
1134 1135
		if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
			|| pmd_devmap(*src_pmd)) {
1136
			int err;
1137
			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
1138 1139 1140 1141 1142 1143 1144 1145
			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
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
		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,
1156
		p4d_t *dst_p4d, p4d_t *src_p4d, struct vm_area_struct *vma,
L
Linus Torvalds 已提交
1157 1158 1159 1160 1161
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

1162
	dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
L
Linus Torvalds 已提交
1163 1164
	if (!dst_pud)
		return -ENOMEM;
1165
	src_pud = pud_offset(src_p4d, addr);
L
Linus Torvalds 已提交
1166 1167
	do {
		next = pud_addr_end(addr, end);
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
		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
1180 1181 1182 1183 1184 1185 1186 1187 1188
		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;
}

1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
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
1211 1212 1213 1214 1215 1216 1217
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;
1218 1219 1220
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	bool is_cow;
A
Andrea Arcangeli 已提交
1221
	int ret;
L
Linus Torvalds 已提交
1222

1223 1224 1225 1226 1227 1228
	/*
	 * 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.
	 */
1229 1230 1231
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
1232

L
Linus Torvalds 已提交
1233 1234 1235
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

1236
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
1237 1238 1239 1240
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1241
		ret = track_pfn_copy(vma);
1242 1243 1244 1245
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1246 1247 1248 1249 1250 1251
	/*
	 * 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.
	 */
1252 1253 1254 1255 1256 1257
	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
1258 1259

	ret = 0;
L
Linus Torvalds 已提交
1260 1261 1262 1263 1264 1265
	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;
1266
		if (unlikely(copy_p4d_range(dst_mm, src_mm, dst_pgd, src_pgd,
A
Andrea Arcangeli 已提交
1267 1268 1269 1270
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1271
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1272

1273 1274
	if (is_cow)
		mmu_notifier_invalidate_range_end(src_mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1275
	return ret;
L
Linus Torvalds 已提交
1276 1277
}

1278
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1279
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1280
				unsigned long addr, unsigned long end,
1281
				struct zap_details *details)
L
Linus Torvalds 已提交
1282
{
N
Nick Piggin 已提交
1283
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1284
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1285
	int rss[NR_MM_COUNTERS];
1286
	spinlock_t *ptl;
1287
	pte_t *start_pte;
1288
	pte_t *pte;
1289
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1290

1291
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1292
again:
1293
	init_rss_vec(rss);
1294 1295
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1296
	flush_tlb_batched_pending(mm);
1297
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1298 1299
	do {
		pte_t ptent = *pte;
T
Tobin C Harding 已提交
1300
		if (pte_none(ptent))
L
Linus Torvalds 已提交
1301
			continue;
1302

L
Linus Torvalds 已提交
1303
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1304
			struct page *page;
1305

1306
			page = _vm_normal_page(vma, addr, ptent, true);
L
Linus Torvalds 已提交
1307 1308 1309 1310 1311 1312 1313
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1314
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1315 1316
					continue;
			}
N
Nick Piggin 已提交
1317
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1318
							tlb->fullmm);
L
Linus Torvalds 已提交
1319 1320 1321
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1322 1323

			if (!PageAnon(page)) {
1324 1325
				if (pte_dirty(ptent)) {
					force_flush = 1;
1326
					set_page_dirty(page);
1327
				}
1328
				if (pte_young(ptent) &&
1329
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1330
					mark_page_accessed(page);
1331
			}
1332
			rss[mm_counter(page)]--;
1333
			page_remove_rmap(page, false);
1334 1335
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1336
			if (unlikely(__tlb_remove_page(tlb, page))) {
1337
				force_flush = 1;
1338
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1339
				break;
1340
			}
L
Linus Torvalds 已提交
1341 1342
			continue;
		}
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365

		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;
		}

1366 1367
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1368
			continue;
K
KAMEZAWA Hiroyuki 已提交
1369

1370 1371 1372 1373 1374
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1375

1376
			page = migration_entry_to_page(entry);
1377
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1378
		}
1379 1380
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1381
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1382
	} while (pte++, addr += PAGE_SIZE, addr != end);
1383

K
KAMEZAWA Hiroyuki 已提交
1384
	add_mm_rss_vec(mm, rss);
1385
	arch_leave_lazy_mmu_mode();
1386

1387
	/* Do the actual TLB flush before dropping ptl */
1388
	if (force_flush)
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400
		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);
1401
		if (addr != end)
P
Peter Zijlstra 已提交
1402 1403 1404
			goto again;
	}

1405
	return addr;
L
Linus Torvalds 已提交
1406 1407
}

1408
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1409
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1410
				unsigned long addr, unsigned long end,
1411
				struct zap_details *details)
L
Linus Torvalds 已提交
1412 1413 1414 1415 1416 1417 1418
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1419
		if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1420
			if (next - addr != HPAGE_PMD_SIZE) {
1421 1422
				VM_BUG_ON_VMA(vma_is_anonymous(vma) &&
				    !rwsem_is_locked(&tlb->mm->mmap_sem), vma);
1423
				__split_huge_pmd(vma, pmd, addr, false, NULL);
S
Shaohua Li 已提交
1424
			} else if (zap_huge_pmd(tlb, vma, pmd, addr))
1425
				goto next;
1426 1427
			/* fall through */
		}
1428 1429 1430 1431 1432 1433 1434 1435 1436
		/*
		 * 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;
1437
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1438
next:
1439 1440
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1441 1442

	return addr;
L
Linus Torvalds 已提交
1443 1444
}

1445
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
1446
				struct vm_area_struct *vma, p4d_t *p4d,
L
Linus Torvalds 已提交
1447
				unsigned long addr, unsigned long end,
1448
				struct zap_details *details)
L
Linus Torvalds 已提交
1449 1450 1451 1452
{
	pud_t *pud;
	unsigned long next;

1453
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
1454 1455
	do {
		next = pud_addr_end(addr, end);
1456 1457 1458 1459 1460 1461 1462 1463
		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 */
		}
1464
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1465
			continue;
1466
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1467 1468
next:
		cond_resched();
1469
	} while (pud++, addr = next, addr != end);
1470 1471

	return addr;
L
Linus Torvalds 已提交
1472 1473
}

1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
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
1493
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1494 1495 1496
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1497 1498 1499 1500 1501 1502 1503 1504 1505
{
	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);
1506
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1507
			continue;
1508
		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
1509
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1510 1511
	tlb_end_vma(tlb, vma);
}
1512

1513 1514 1515

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1516
		unsigned long end_addr,
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
		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;

1528 1529 1530
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1531
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1532
		untrack_pfn(vma, 0, 0);
1533 1534 1535 1536 1537 1538 1539

	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
1540
			 * cleanup path of mmap_region. When
1541
			 * hugetlbfs ->mmap method fails,
1542
			 * mmap_region() nullifies vma->vm_file
1543 1544 1545 1546
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1547
			if (vma->vm_file) {
1548
				i_mmap_lock_write(vma->vm_file->f_mapping);
1549
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1550
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1551
			}
1552 1553 1554
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1555 1556 1557 1558
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1559
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1560 1561 1562 1563
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1564
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
 *
 * 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
1575
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1576
		struct vm_area_struct *vma, unsigned long start_addr,
1577
		unsigned long end_addr)
L
Linus Torvalds 已提交
1578
{
A
Andrea Arcangeli 已提交
1579
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
1580

A
Andrea Arcangeli 已提交
1581
	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
1582
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1583
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
A
Andrea Arcangeli 已提交
1584
	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
L
Linus Torvalds 已提交
1585 1586 1587 1588 1589
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1590
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1591
 * @size: number of bytes to zap
1592 1593
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1594
 */
1595
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1596
		unsigned long size)
L
Linus Torvalds 已提交
1597 1598
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1599
	struct mmu_gather tlb;
1600
	unsigned long end = start + size;
L
Linus Torvalds 已提交
1601 1602

	lru_add_drain();
1603
	tlb_gather_mmu(&tlb, mm, start, end);
1604
	update_hiwater_rss(mm);
1605
	mmu_notifier_invalidate_range_start(mm, start, end);
1606
	for ( ; vma && vma->vm_start < end; vma = vma->vm_next) {
1607
		unmap_single_vma(&tlb, vma, start, end, NULL);
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619

		/*
		 * 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);
	}

1620 1621
	mmu_notifier_invalidate_range_end(mm, start, end);
	tlb_finish_mmu(&tlb, start, end);
L
Linus Torvalds 已提交
1622 1623
}

1624 1625 1626 1627 1628
/**
 * 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
1629
 * @details: details of shared cache invalidation
1630 1631
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1632
 */
1633
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1634 1635 1636
		unsigned long size, struct zap_details *details)
{
	struct mm_struct *mm = vma->vm_mm;
P
Peter Zijlstra 已提交
1637
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1638 1639 1640
	unsigned long end = address + size;

	lru_add_drain();
1641
	tlb_gather_mmu(&tlb, mm, address, end);
1642
	update_hiwater_rss(mm);
1643
	mmu_notifier_invalidate_range_start(mm, address, end);
1644
	unmap_single_vma(&tlb, vma, address, end, details);
1645
	mmu_notifier_invalidate_range_end(mm, address, end);
P
Peter Zijlstra 已提交
1646
	tlb_finish_mmu(&tlb, address, end);
L
Linus Torvalds 已提交
1647 1648
}

1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
/**
 * 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;
1667
	zap_page_range_single(vma, address, size, NULL);
1668 1669 1670 1671
	return 0;
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1672
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1673
			spinlock_t **ptl)
1674
{
1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
	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);
1693 1694
}

1695 1696 1697 1698 1699 1700 1701
/*
 * 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
1702 1703
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1704
{
N
Nick Piggin 已提交
1705
	struct mm_struct *mm = vma->vm_mm;
1706
	int retval;
1707
	pte_t *pte;
1708 1709
	spinlock_t *ptl;

1710
	retval = -EINVAL;
1711
	if (PageAnon(page))
1712
		goto out;
1713 1714
	retval = -ENOMEM;
	flush_dcache_page(page);
1715
	pte = get_locked_pte(mm, addr, &ptl);
1716
	if (!pte)
1717
		goto out;
1718 1719 1720 1721 1722 1723
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1724
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1725
	page_add_file_rmap(page, false);
1726 1727 1728
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

	retval = 0;
1729 1730
	pte_unmap_unlock(pte, ptl);
	return retval;
1731 1732 1733 1734 1735 1736
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1737 1738 1739 1740 1741 1742
/**
 * 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
 *
1743 1744 1745 1746 1747 1748
 * 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
1749
 * (see split_page()).
1750 1751 1752 1753 1754 1755 1756 1757
 *
 * 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.
1758 1759 1760 1761 1762
 *
 * 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.
1763
 */
N
Nick Piggin 已提交
1764 1765
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1766 1767 1768 1769 1770
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1771 1772 1773 1774 1775
	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
1776
	return insert_page(vma, addr, page, vma->vm_page_prot);
1777
}
1778
EXPORT_SYMBOL(vm_insert_page);
1779

N
Nick Piggin 已提交
1780
static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
R
Ross Zwisler 已提交
1781
			pfn_t pfn, pgprot_t prot, bool mkwrite)
N
Nick Piggin 已提交
1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
{
	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
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808
	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
1809 1810

	/* Ok, finally just insert the thing.. */
1811 1812 1813 1814
	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
1815 1816 1817 1818 1819 1820 1821

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

N
Nick Piggin 已提交
1822
	set_pte_at(mm, addr, pte, entry);
1823
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1824 1825 1826 1827 1828 1829 1830 1831

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

N
Nick Piggin 已提交
1832 1833 1834 1835 1836 1837
/**
 * 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
 *
1838
 * Similar to vm_insert_page, this allows drivers to insert individual pages
N
Nick Piggin 已提交
1839 1840 1841 1842
 * 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
1843 1844 1845 1846 1847
 *
 * 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
1848 1849
 */
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
N
Nick Piggin 已提交
1850
			unsigned long pfn)
A
Andy Lutomirski 已提交
1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872
{
	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
1873
{
1874
	int ret;
N
Nick Piggin 已提交
1875 1876 1877 1878 1879 1880
	/*
	 * 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
1881 1882 1883 1884 1885
	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
1886

N
Nick Piggin 已提交
1887 1888
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1889 1890

	track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
1891

R
Ross Zwisler 已提交
1892 1893
	ret = insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
			false);
1894 1895

	return ret;
N
Nick Piggin 已提交
1896
}
A
Andy Lutomirski 已提交
1897
EXPORT_SYMBOL(vm_insert_pfn_prot);
N
Nick Piggin 已提交
1898

R
Ross Zwisler 已提交
1899 1900
static int __vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
			pfn_t pfn, bool mkwrite)
N
Nick Piggin 已提交
1901
{
1902 1903
	pgprot_t pgprot = vma->vm_page_prot;

N
Nick Piggin 已提交
1904
	BUG_ON(!(vma->vm_flags & VM_MIXEDMAP));
N
Nick Piggin 已提交
1905

N
Nick Piggin 已提交
1906 1907
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
1908 1909

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1910

N
Nick Piggin 已提交
1911 1912 1913 1914
	/*
	 * 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
1915 1916
	 * 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
1917
	 */
1918
	if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1919 1920
		struct page *page;

1921 1922 1923 1924 1925 1926
		/*
		 * 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));
1927
		return insert_page(vma, addr, page, pgprot);
N
Nick Piggin 已提交
1928
	}
R
Ross Zwisler 已提交
1929 1930 1931 1932 1933 1934 1935 1936
	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
1937
}
N
Nick Piggin 已提交
1938
EXPORT_SYMBOL(vm_insert_mixed);
N
Nick Piggin 已提交
1939

R
Ross Zwisler 已提交
1940 1941 1942 1943 1944 1945 1946
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
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
/*
 * 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
1957
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1958

H
Hugh Dickins 已提交
1959
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1960 1961
	if (!pte)
		return -ENOMEM;
1962
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1963 1964
	do {
		BUG_ON(!pte_none(*pte));
N
Nick Piggin 已提交
1965
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1966 1967
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1968
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1969
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
	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;
1984
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1985 1986 1987 1988 1989 1990 1991 1992 1993
	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;
}

1994
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
L
Linus Torvalds 已提交
1995 1996 1997 1998 1999 2000 2001
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;

	pfn -= addr >> PAGE_SHIFT;
2002
	pud = pud_alloc(mm, p4d, addr);
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
	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;
}

2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033
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;
}

2034 2035 2036 2037 2038 2039 2040 2041 2042 2043
/**
 * 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
2044 2045 2046 2047 2048
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;
2049
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
2050
	struct mm_struct *mm = vma->vm_mm;
2051
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
2052 2053 2054 2055 2056 2057 2058
	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).
2059 2060 2061
	 *   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.
2062 2063 2064 2065
	 *   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
2066 2067 2068 2069
	 *
	 * 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".
2070
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
2071
	 */
2072 2073 2074
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
2075
		vma->vm_pgoff = pfn;
2076 2077
	}

2078
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
2079
	if (err)
2080
		return -EINVAL;
L
Linus Torvalds 已提交
2081

2082
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
2083 2084 2085 2086 2087 2088 2089

	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);
2090
		err = remap_p4d_range(mm, pgd, addr, next,
L
Linus Torvalds 已提交
2091 2092 2093 2094
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2095 2096

	if (err)
2097
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
2098

L
Linus Torvalds 已提交
2099 2100 2101 2102
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

2103 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
/**
 * 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);

2150 2151 2152 2153 2154 2155
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;
2156
	pgtable_t token;
2157
	spinlock_t *uninitialized_var(ptl);
2158 2159 2160 2161 2162 2163 2164 2165 2166

	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));

2167 2168
	arch_enter_lazy_mmu_mode();

2169
	token = pmd_pgtable(*pmd);
2170 2171

	do {
2172
		err = fn(pte++, token, addr, data);
2173 2174
		if (err)
			break;
2175
	} while (addr += PAGE_SIZE, addr != end);
2176

2177 2178
	arch_leave_lazy_mmu_mode();

2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
	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
2192 2193
	BUG_ON(pud_huge(*pud));

2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205
	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;
}

2206
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
2207 2208 2209 2210 2211 2212 2213
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

2214
	pud = pud_alloc(mm, p4d, addr);
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225
	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;
}

2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
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;
}

2246 2247 2248 2249 2250 2251 2252 2253 2254
/*
 * 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;
2255
	unsigned long end = addr + size;
2256 2257
	int err;

2258 2259 2260
	if (WARN_ON(addr >= end))
		return -EINVAL;

2261 2262 2263
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
2264
		err = apply_to_p4d_range(mm, pgd, addr, next, fn, data);
2265 2266 2267
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2268

2269 2270 2271 2272
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2273
/*
2274 2275 2276 2277 2278
 * 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;
2279
 * and do_anonymous_page can safely check later on).
2280
 */
H
Hugh Dickins 已提交
2281
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2282 2283 2284 2285 2286
				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
2287 2288
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2289
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2290
		spin_unlock(ptl);
2291 2292 2293 2294 2295 2296
	}
#endif
	pte_unmap(page_table);
	return same;
}

2297
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2298
{
2299 2300
	debug_dma_assert_idle(src);

2301 2302 2303 2304 2305 2306 2307
	/*
	 * 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)) {
2308
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2309 2310 2311 2312 2313 2314 2315 2316 2317
		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))
2318
			clear_page(kaddr);
2319
		kunmap_atomic(kaddr);
2320
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2321 2322
	} else
		copy_user_highpage(dst, src, va, vma);
2323 2324
}

2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338
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;
}

2339 2340 2341 2342 2343 2344
/*
 * 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.
 */
2345
static int do_page_mkwrite(struct vm_fault *vmf)
2346 2347
{
	int ret;
2348 2349
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2350

2351
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2352

2353
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2354 2355
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
	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;
}

2370 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
/*
 * 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);
}

2405 2406 2407 2408 2409 2410 2411 2412
/*
 * 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.
 */
2413
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2414
	__releases(vmf->ptl)
2415
{
J
Jan Kara 已提交
2416
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2417
	struct page *page = vmf->page;
2418 2419 2420 2421 2422 2423 2424 2425 2426
	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
2427 2428
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2429
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2430 2431 2432
	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);
2433 2434
}

2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450
/*
 * 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
2451
static int wp_page_copy(struct vm_fault *vmf)
2452
{
J
Jan Kara 已提交
2453
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2454
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2455
	struct page *old_page = vmf->page;
2456 2457 2458
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
J
Jan Kara 已提交
2459
	const unsigned long mmun_start = vmf->address & PAGE_MASK;
K
Kirill A. Shutemov 已提交
2460
	const unsigned long mmun_end = mmun_start + PAGE_SIZE;
2461 2462 2463 2464 2465
	struct mem_cgroup *memcg;

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

J
Jan Kara 已提交
2466
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2467 2468
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2469 2470 2471
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2472
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2473
				vmf->address);
2474 2475
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2476
		cow_user_page(new_page, old_page, vmf->address, vma);
2477 2478
	}

2479
	if (mem_cgroup_try_charge(new_page, mm, GFP_KERNEL, &memcg, false))
2480 2481
		goto oom_free_new;

2482 2483
	__SetPageUptodate(new_page);

2484 2485 2486 2487 2488
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2489
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2490
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2491 2492
		if (old_page) {
			if (!PageAnon(old_page)) {
2493 2494
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2495 2496 2497 2498 2499
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2500
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2501 2502 2503 2504 2505 2506 2507 2508
		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
2509 2510
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2511
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2512 2513 2514 2515 2516 2517
		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
2518 2519
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
		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.
			 */
2543
			page_remove_rmap(old_page, false);
2544 2545 2546 2547 2548 2549
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2550
		mem_cgroup_cancel_charge(new_page, memcg, false);
2551 2552 2553
	}

	if (new_page)
2554
		put_page(new_page);
2555

J
Jan Kara 已提交
2556
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2557 2558 2559 2560 2561 2562 2563 2564
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
	if (old_page) {
		/*
		 * Don't let another task, with possibly unlocked vma,
		 * keep the mlocked page.
		 */
		if (page_copied && (vma->vm_flags & VM_LOCKED)) {
			lock_page(old_page);	/* LRU manipulation */
2565 2566
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2567 2568
			unlock_page(old_page);
		}
2569
		put_page(old_page);
2570 2571 2572
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2573
	put_page(new_page);
2574 2575
oom:
	if (old_page)
2576
		put_page(old_page);
2577 2578 2579
	return VM_FAULT_OOM;
}

2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
/**
 * 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);
2606
		return VM_FAULT_NOPAGE;
2607 2608
	}
	wp_page_reuse(vmf);
2609
	return 0;
2610 2611
}

2612 2613 2614 2615
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
J
Jan Kara 已提交
2616
static int wp_pfn_shared(struct vm_fault *vmf)
2617
{
J
Jan Kara 已提交
2618
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2619

2620 2621 2622
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
		int ret;

J
Jan Kara 已提交
2623
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2624
		vmf->flags |= FAULT_FLAG_MKWRITE;
2625
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2626
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2627
			return ret;
2628
		return finish_mkwrite_fault(vmf);
2629
	}
2630 2631
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2632 2633
}

J
Jan Kara 已提交
2634
static int wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2635
	__releases(vmf->ptl)
2636
{
J
Jan Kara 已提交
2637
	struct vm_area_struct *vma = vmf->vma;
2638

J
Jan Kara 已提交
2639
	get_page(vmf->page);
2640 2641 2642 2643

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

J
Jan Kara 已提交
2644
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2645
		tmp = do_page_mkwrite(vmf);
2646 2647
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2648
			put_page(vmf->page);
2649 2650
			return tmp;
		}
2651
		tmp = finish_mkwrite_fault(vmf);
2652
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2653 2654
			unlock_page(vmf->page);
			put_page(vmf->page);
2655
			return tmp;
2656
		}
2657 2658
	} else {
		wp_page_reuse(vmf);
2659
		lock_page(vmf->page);
2660
	}
2661 2662
	fault_dirty_shared_page(vma, vmf->page);
	put_page(vmf->page);
2663

2664
	return VM_FAULT_WRITE;
2665 2666
}

L
Linus Torvalds 已提交
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680
/*
 * 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.
 *
2681 2682 2683
 * 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
2684
 */
J
Jan Kara 已提交
2685
static int do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2686
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2687
{
J
Jan Kara 已提交
2688
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2689

J
Jan Kara 已提交
2690 2691
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2692
		/*
2693 2694
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2695 2696
		 *
		 * We should not cow pages in a shared writeable mapping.
2697
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2698 2699 2700
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2701
			return wp_pfn_shared(vmf);
2702

J
Jan Kara 已提交
2703
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2704
		return wp_page_copy(vmf);
2705
	}
L
Linus Torvalds 已提交
2706

2707
	/*
P
Peter Zijlstra 已提交
2708 2709
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2710
	 */
J
Jan Kara 已提交
2711
	if (PageAnon(vmf->page) && !PageKsm(vmf->page)) {
2712
		int total_map_swapcount;
J
Jan Kara 已提交
2713 2714
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2715
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2716
			lock_page(vmf->page);
J
Jan Kara 已提交
2717 2718
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2719
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2720
				unlock_page(vmf->page);
J
Jan Kara 已提交
2721
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2722
				put_page(vmf->page);
2723
				return 0;
2724
			}
J
Jan Kara 已提交
2725
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2726
		}
2727 2728
		if (reuse_swap_page(vmf->page, &total_map_swapcount)) {
			if (total_map_swapcount == 1) {
2729 2730 2731 2732 2733 2734 2735
				/*
				 * 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
2736
				page_move_anon_rmap(vmf->page, vma);
2737
			}
J
Jan Kara 已提交
2738
			unlock_page(vmf->page);
2739 2740
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2741
		}
J
Jan Kara 已提交
2742
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2743
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2744
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2745
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2746 2747 2748 2749 2750
	}

	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2751
	get_page(vmf->page);
2752

J
Jan Kara 已提交
2753
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2754
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2755 2756
}

2757
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2758 2759 2760
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2761
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2762 2763
}

2764
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
L
Linus Torvalds 已提交
2765 2766 2767 2768 2769
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2770
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2771 2772 2773
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2774
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2775 2776 2777 2778 2779 2780 2781
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2782
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2783 2784
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2785
				details);
L
Linus Torvalds 已提交
2786 2787 2788 2789
	}
}

/**
2790 2791 2792 2793
 * 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
2794
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2795 2796
 * @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
2797
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808
 * 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
2809
	struct zap_details details = { };
L
Linus Torvalds 已提交
2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820
	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
2821
	details.check_mapping = even_cows ? NULL : mapping;
L
Linus Torvalds 已提交
2822 2823 2824 2825 2826
	details.first_index = hba;
	details.last_index = hba + hlen - 1;
	if (details.last_index < details.first_index)
		details.last_index = ULONG_MAX;

2827
	i_mmap_lock_write(mapping);
2828
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
L
Linus Torvalds 已提交
2829
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
2830
	i_mmap_unlock_write(mapping);
L
Linus Torvalds 已提交
2831 2832 2833 2834
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2835 2836
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2837 2838 2839 2840
 * 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
2841
 */
J
Jan Kara 已提交
2842
int do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2843
{
J
Jan Kara 已提交
2844
	struct vm_area_struct *vma = vmf->vma;
H
Huang Ying 已提交
2845
	struct page *page = NULL, *swapcache;
2846
	struct mem_cgroup *memcg;
H
Huang Ying 已提交
2847
	struct vma_swap_readahead swap_ra;
2848
	swp_entry_t entry;
L
Linus Torvalds 已提交
2849
	pte_t pte;
2850
	int locked;
2851
	int exclusive = 0;
N
Nick Piggin 已提交
2852
	int ret = 0;
H
Huang Ying 已提交
2853
	bool vma_readahead = swap_use_vma_readahead();
L
Linus Torvalds 已提交
2854

H
Huang Ying 已提交
2855 2856 2857 2858 2859
	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);
2860
		goto out;
H
Huang Ying 已提交
2861
	}
2862

J
Jan Kara 已提交
2863
	entry = pte_to_swp_entry(vmf->orig_pte);
2864 2865
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2866 2867
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2868 2869 2870 2871 2872 2873 2874 2875
		} 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);
2876 2877 2878
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2879
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2880
			ret = VM_FAULT_SIGBUS;
2881
		}
2882 2883
		goto out;
	}
2884
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
H
Huang Ying 已提交
2885 2886 2887
	if (!page)
		page = lookup_swap_cache(entry, vma_readahead ? vma : NULL,
					 vmf->address);
L
Linus Torvalds 已提交
2888
	if (!page) {
H
Huang Ying 已提交
2889 2890 2891 2892 2893 2894
		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);
L
Linus Torvalds 已提交
2895 2896
		if (!page) {
			/*
2897 2898
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2899
			 */
J
Jan Kara 已提交
2900 2901
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2902
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2903
				ret = VM_FAULT_OOM;
2904
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2905
			goto unlock;
L
Linus Torvalds 已提交
2906 2907 2908 2909
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2910
		count_vm_event(PGMAJFAULT);
2911
		count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
2912
	} else if (PageHWPoison(page)) {
2913 2914 2915 2916
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2917 2918
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2919
		swapcache = page;
2920
		goto out_release;
L
Linus Torvalds 已提交
2921 2922
	}

2923
	swapcache = page;
J
Jan Kara 已提交
2924
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2925

2926
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2927 2928 2929 2930
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2931

A
Andrea Arcangeli 已提交
2932
	/*
2933 2934 2935 2936
	 * 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
2937
	 */
2938
	if (unlikely(!PageSwapCache(page) || page_private(page) != entry.val))
A
Andrea Arcangeli 已提交
2939 2940
		goto out_page;

J
Jan Kara 已提交
2941
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2942 2943 2944 2945
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2946 2947
	}

K
Kirill A. Shutemov 已提交
2948 2949
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL,
				&memcg, false)) {
2950
		ret = VM_FAULT_OOM;
2951
		goto out_page;
2952 2953
	}

L
Linus Torvalds 已提交
2954
	/*
2955
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2956
	 */
J
Jan Kara 已提交
2957 2958
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2959
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2960 2961 2962 2963 2964
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2965 2966
	}

2967 2968 2969 2970 2971 2972 2973 2974 2975
	/*
	 * 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
2976

K
Kirill A. Shutemov 已提交
2977 2978
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2979
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2980
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2981
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2982
		vmf->flags &= ~FAULT_FLAG_WRITE;
2983
		ret |= VM_FAULT_WRITE;
2984
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2985 2986
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
2987
	if (pte_swp_soft_dirty(vmf->orig_pte))
2988
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2989
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
2990
	vmf->orig_pte = pte;
2991
	if (page == swapcache) {
J
Jan Kara 已提交
2992
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
2993
		mem_cgroup_commit_charge(page, memcg, true, false);
2994
		activate_page(page);
2995
	} else { /* ksm created a completely new copy */
J
Jan Kara 已提交
2996
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2997
		mem_cgroup_commit_charge(page, memcg, false, false);
2998 2999
		lru_cache_add_active_or_unevictable(page, vma);
	}
L
Linus Torvalds 已提交
3000

3001
	swap_free(entry);
3002 3003
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
3004
		try_to_free_swap(page);
3005
	unlock_page(page);
3006
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
3007 3008 3009 3010 3011 3012 3013 3014 3015
		/*
		 * 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);
3016
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3017
	}
3018

J
Jan Kara 已提交
3019
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
3020
		ret |= do_wp_page(vmf);
3021 3022
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
3023 3024 3025 3026
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3027
	update_mmu_cache(vma, vmf->address, vmf->pte);
3028
unlock:
J
Jan Kara 已提交
3029
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
3030 3031
out:
	return ret;
3032
out_nomap:
3033
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
3034
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3035
out_page:
3036
	unlock_page(page);
3037
out_release:
3038
	put_page(page);
3039
	if (page != swapcache) {
A
Andrea Arcangeli 已提交
3040
		unlock_page(swapcache);
3041
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3042
	}
3043
	return ret;
L
Linus Torvalds 已提交
3044 3045 3046
}

/*
3047 3048 3049
 * 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
3050
 */
J
Jan Kara 已提交
3051
static int do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3052
{
J
Jan Kara 已提交
3053
	struct vm_area_struct *vma = vmf->vma;
3054
	struct mem_cgroup *memcg;
3055
	struct page *page;
3056
	int ret = 0;
L
Linus Torvalds 已提交
3057 3058
	pte_t entry;

3059 3060 3061 3062
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

3063 3064 3065 3066 3067 3068 3069 3070 3071 3072
	/*
	 * 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
3073
	if (pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))
3074 3075 3076
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3077
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
3078 3079
		return 0;

3080
	/* Use the zero-page for reads */
J
Jan Kara 已提交
3081
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3082
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
3083
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
3084
						vma->vm_page_prot));
J
Jan Kara 已提交
3085 3086 3087
		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
3088
			goto unlock;
3089 3090 3091
		ret = check_stable_address_space(vma->vm_mm);
		if (ret)
			goto unlock;
3092 3093
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3094 3095
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
3096
		}
H
Hugh Dickins 已提交
3097 3098 3099
		goto setpte;
	}

N
Nick Piggin 已提交
3100 3101 3102
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
3103
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
3104 3105
	if (!page)
		goto oom;
3106

K
Kirill A. Shutemov 已提交
3107
	if (mem_cgroup_try_charge(page, vma->vm_mm, GFP_KERNEL, &memcg, false))
3108 3109
		goto oom_free_page;

3110 3111 3112 3113 3114
	/*
	 * 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
3115
	__SetPageUptodate(page);
3116

N
Nick Piggin 已提交
3117
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
3118 3119
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
3120

J
Jan Kara 已提交
3121 3122 3123
	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
3124
		goto release;
H
Hugh Dickins 已提交
3125

3126 3127 3128 3129
	ret = check_stable_address_space(vma->vm_mm);
	if (ret)
		goto release;

3130 3131
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3132
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3133
		mem_cgroup_cancel_charge(page, memcg, false);
3134
		put_page(page);
J
Jan Kara 已提交
3135
		return handle_userfault(vmf, VM_UFFD_MISSING);
3136 3137
	}

K
Kirill A. Shutemov 已提交
3138
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3139
	page_add_new_anon_rmap(page, vma, vmf->address, false);
3140
	mem_cgroup_commit_charge(page, memcg, false, false);
3141
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
3142
setpte:
J
Jan Kara 已提交
3143
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
3144 3145

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3146
	update_mmu_cache(vma, vmf->address, vmf->pte);
3147
unlock:
J
Jan Kara 已提交
3148
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3149
	return ret;
3150
release:
3151
	mem_cgroup_cancel_charge(page, memcg, false);
3152
	put_page(page);
3153
	goto unlock;
3154
oom_free_page:
3155
	put_page(page);
3156
oom:
L
Linus Torvalds 已提交
3157 3158 3159
	return VM_FAULT_OOM;
}

3160 3161 3162 3163 3164
/*
 * 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
3165
static int __do_fault(struct vm_fault *vmf)
3166
{
J
Jan Kara 已提交
3167
	struct vm_area_struct *vma = vmf->vma;
3168 3169
	int ret;

3170
	ret = vma->vm_ops->fault(vmf);
3171
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
3172
			    VM_FAULT_DONE_COW)))
3173
		return ret;
3174

3175
	if (unlikely(PageHWPoison(vmf->page))) {
3176
		if (ret & VM_FAULT_LOCKED)
3177 3178
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
3179
		vmf->page = NULL;
3180 3181 3182 3183
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
3184
		lock_page(vmf->page);
3185
	else
3186
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
3187 3188 3189 3190

	return ret;
}

3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201
/*
 * 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
3202
static int pte_alloc_one_map(struct vm_fault *vmf)
3203
{
J
Jan Kara 已提交
3204
	struct vm_area_struct *vma = vmf->vma;
3205

J
Jan Kara 已提交
3206
	if (!pmd_none(*vmf->pmd))
3207
		goto map_pte;
J
Jan Kara 已提交
3208 3209 3210 3211
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
3212 3213 3214 3215
			goto map_pte;
		}

		atomic_long_inc(&vma->vm_mm->nr_ptes);
J
Jan Kara 已提交
3216 3217
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
3218
		vmf->prealloc_pte = NULL;
J
Jan Kara 已提交
3219
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd, vmf->address))) {
3220 3221 3222 3223 3224
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
3225 3226 3227 3228 3229 3230 3231 3232
	 * 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.
3233
	 */
3234
	if (pmd_devmap_trans_unstable(vmf->pmd))
3235 3236
		return VM_FAULT_NOPAGE;

3237 3238 3239 3240 3241 3242 3243 3244 3245
	/*
	 * 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
3246 3247
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
3248 3249 3250
	return 0;
}

3251
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264

#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
3265
static void deposit_prealloc_pte(struct vm_fault *vmf)
3266
{
J
Jan Kara 已提交
3267
	struct vm_area_struct *vma = vmf->vma;
3268

J
Jan Kara 已提交
3269
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
3270 3271 3272 3273 3274
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
	atomic_long_inc(&vma->vm_mm->nr_ptes);
3275
	vmf->prealloc_pte = NULL;
3276 3277
}

J
Jan Kara 已提交
3278
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3279
{
J
Jan Kara 已提交
3280 3281 3282
	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
3283 3284 3285 3286 3287 3288 3289 3290 3291
	pmd_t entry;
	int i, ret;

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

3292 3293 3294 3295
	/*
	 * 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
3296 3297 3298
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm, vmf->address);
		if (!vmf->prealloc_pte)
3299 3300 3301 3302
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
3303 3304
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315
		goto out;

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

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

	add_mm_counter(vma->vm_mm, MM_FILEPAGES, HPAGE_PMD_NR);
	page_add_file_rmap(page, true);
3316 3317 3318 3319
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3320
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3321

J
Jan Kara 已提交
3322
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3323

J
Jan Kara 已提交
3324
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3325 3326 3327

	/* fault is handled */
	ret = 0;
3328
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3329
out:
J
Jan Kara 已提交
3330
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3331 3332 3333
	return ret;
}
#else
J
Jan Kara 已提交
3334
static int do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3335 3336 3337 3338 3339 3340
{
	BUILD_BUG();
	return 0;
}
#endif

3341
/**
3342 3343
 * 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.
3344
 *
J
Jan Kara 已提交
3345
 * @vmf: fault environment
3346
 * @memcg: memcg to charge page (only for private mappings)
3347 3348
 * @page: page to map
 *
J
Jan Kara 已提交
3349 3350
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3351 3352 3353 3354
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
J
Jan Kara 已提交
3355
int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3356
		struct page *page)
3357
{
J
Jan Kara 已提交
3358 3359
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3360
	pte_t entry;
K
Kirill A. Shutemov 已提交
3361 3362
	int ret;

J
Jan Kara 已提交
3363
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3364
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3365 3366 3367
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3368
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3369
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3370
			return ret;
K
Kirill A. Shutemov 已提交
3371
	}
3372

J
Jan Kara 已提交
3373 3374
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3375
		if (ret)
H
Hugh Dickins 已提交
3376
			return ret;
3377 3378 3379
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3380 3381
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3382

3383 3384 3385 3386
	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
3387 3388
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3389
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3390
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3391 3392
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3393
	} else {
3394
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3395
		page_add_file_rmap(page, false);
3396
	}
J
Jan Kara 已提交
3397
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3398 3399

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3400
	update_mmu_cache(vma, vmf->address, vmf->pte);
3401

H
Hugh Dickins 已提交
3402
	return 0;
3403 3404
}

3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422

/**
 * 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;
3423
	int ret = 0;
3424 3425 3426 3427 3428 3429 3430

	/* 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;
3431 3432 3433 3434 3435 3436 3437 3438 3439

	/*
	 * 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);
3440 3441 3442 3443 3444
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3445 3446
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3447 3448 3449

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3450
{
3451
	*val = fault_around_bytes;
3452 3453 3454
	return 0;
}

3455 3456 3457 3458 3459
/*
 * 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.
 */
3460
static int fault_around_bytes_set(void *data, u64 val)
3461
{
3462
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3463
		return -EINVAL;
3464 3465 3466 3467
	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 */
3468 3469
	return 0;
}
3470
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
3471
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3472 3473 3474 3475 3476

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

3477
	ret = debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
3478
			&fault_around_bytes_fops);
3479
	if (!ret)
3480
		pr_warn("Failed to create fault_around_bytes in debugfs");
3481 3482 3483 3484
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3485

3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508
/*
 * 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.
 */
3509
static int do_fault_around(struct vm_fault *vmf)
3510
{
J
Jan Kara 已提交
3511
	unsigned long address = vmf->address, nr_pages, mask;
3512
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3513
	pgoff_t end_pgoff;
3514
	int off, ret = 0;
3515

3516
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3517 3518
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3519 3520
	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
3521
	start_pgoff -= off;
3522 3523

	/*
K
Kirill A. Shutemov 已提交
3524 3525
	 *  end_pgoff is either end of page table or end of vma
	 *  or fault_around_pages() from start_pgoff, depending what is nearest.
3526
	 */
K
Kirill A. Shutemov 已提交
3527
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3528
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3529
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3530
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3531
			start_pgoff + nr_pages - 1);
3532

J
Jan Kara 已提交
3533 3534 3535 3536
	if (pmd_none(*vmf->pmd)) {
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm,
						  vmf->address);
		if (!vmf->prealloc_pte)
3537
			goto out;
3538
		smp_wmb(); /* See comment in __pte_alloc() */
3539 3540
	}

J
Jan Kara 已提交
3541
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3542 3543

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3544
	if (pmd_trans_huge(*vmf->pmd)) {
3545 3546 3547 3548 3549
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3550
	if (!vmf->pte)
3551 3552 3553
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3554 3555
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3556
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3557
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3558
out:
J
Jan Kara 已提交
3559 3560
	vmf->address = address;
	vmf->pte = NULL;
3561
	return ret;
3562 3563
}

3564
static int do_read_fault(struct vm_fault *vmf)
3565
{
J
Jan Kara 已提交
3566
	struct vm_area_struct *vma = vmf->vma;
3567 3568 3569 3570 3571 3572 3573
	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).
	 */
3574
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3575
		ret = do_fault_around(vmf);
3576 3577
		if (ret)
			return ret;
3578
	}
3579

J
Jan Kara 已提交
3580
	ret = __do_fault(vmf);
3581 3582 3583
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3584
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3585
	unlock_page(vmf->page);
3586
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3587
		put_page(vmf->page);
3588 3589 3590
	return ret;
}

3591
static int do_cow_fault(struct vm_fault *vmf)
3592
{
J
Jan Kara 已提交
3593
	struct vm_area_struct *vma = vmf->vma;
3594 3595 3596 3597 3598
	int ret;

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

J
Jan Kara 已提交
3599 3600
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3601 3602
		return VM_FAULT_OOM;

J
Jan Kara 已提交
3603
	if (mem_cgroup_try_charge(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3604
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3605
		put_page(vmf->cow_page);
3606 3607 3608
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3609
	ret = __do_fault(vmf);
3610 3611
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3612 3613
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3614

3615
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3616
	__SetPageUptodate(vmf->cow_page);
3617

3618
	ret |= finish_fault(vmf);
3619 3620
	unlock_page(vmf->page);
	put_page(vmf->page);
3621 3622
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3623 3624
	return ret;
uncharge_out:
3625
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3626
	put_page(vmf->cow_page);
3627 3628 3629
	return ret;
}

3630
static int do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3631
{
J
Jan Kara 已提交
3632
	struct vm_area_struct *vma = vmf->vma;
3633
	int ret, tmp;
3634

J
Jan Kara 已提交
3635
	ret = __do_fault(vmf);
3636
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3637
		return ret;
L
Linus Torvalds 已提交
3638 3639

	/*
3640 3641
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3642
	 */
3643
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3644
		unlock_page(vmf->page);
3645
		tmp = do_page_mkwrite(vmf);
3646 3647
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3648
			put_page(vmf->page);
3649
			return tmp;
3650
		}
3651 3652
	}

3653
	ret |= finish_fault(vmf);
3654 3655
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3656 3657
		unlock_page(vmf->page);
		put_page(vmf->page);
3658
		return ret;
L
Linus Torvalds 已提交
3659
	}
N
Nick Piggin 已提交
3660

3661
	fault_dirty_shared_page(vma, vmf->page);
3662
	return ret;
3663
}
3664

3665 3666 3667 3668 3669 3670
/*
 * 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
3671
static int do_fault(struct vm_fault *vmf)
3672
{
J
Jan Kara 已提交
3673
	struct vm_area_struct *vma = vmf->vma;
H
Hugh Dickins 已提交
3674
	int ret;
3675

3676 3677
	/* The VMA was not fully populated on mmap() or missing VM_DONTEXPAND */
	if (!vma->vm_ops->fault)
H
Hugh Dickins 已提交
3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688
		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);
3689
		vmf->prealloc_pte = NULL;
H
Hugh Dickins 已提交
3690 3691
	}
	return ret;
3692 3693
}

3694
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3695 3696
				unsigned long addr, int page_nid,
				int *flags)
3697 3698 3699 3700
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3701
	if (page_nid == numa_node_id()) {
3702
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3703 3704
		*flags |= TNF_FAULT_LOCAL;
	}
3705 3706 3707 3708

	return mpol_misplaced(page, vma, addr);
}

J
Jan Kara 已提交
3709
static int do_numa_page(struct vm_fault *vmf)
3710
{
J
Jan Kara 已提交
3711
	struct vm_area_struct *vma = vmf->vma;
3712
	struct page *page = NULL;
3713
	int page_nid = -1;
3714
	int last_cpupid;
3715
	int target_nid;
3716
	bool migrated = false;
3717
	pte_t pte;
3718
	bool was_writable = pte_savedwrite(vmf->orig_pte);
3719
	int flags = 0;
3720 3721

	/*
T
Tobin C Harding 已提交
3722 3723 3724 3725
	 * 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
3726 3727
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
3728
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
J
Jan Kara 已提交
3729
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3730 3731 3732
		goto out;
	}

3733 3734 3735 3736 3737
	/*
	 * 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);
3738 3739
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3740 3741
	if (was_writable)
		pte = pte_mkwrite(pte);
3742
	ptep_modify_prot_commit(vma->vm_mm, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
3743
	update_mmu_cache(vma, vmf->address, vmf->pte);
3744

J
Jan Kara 已提交
3745
	page = vm_normal_page(vma, vmf->address, pte);
3746
	if (!page) {
J
Jan Kara 已提交
3747
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3748 3749 3750
		return 0;
	}

3751 3752
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3753
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3754 3755 3756
		return 0;
	}

3757
	/*
3758 3759 3760 3761 3762 3763
	 * 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.
3764
	 */
3765
	if (!pte_write(pte))
3766 3767
		flags |= TNF_NO_GROUP;

3768 3769 3770 3771 3772 3773 3774
	/*
	 * 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;

3775
	last_cpupid = page_cpupid_last(page);
3776
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3777
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3778
			&flags);
J
Jan Kara 已提交
3779
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3780 3781 3782 3783 3784 3785
	if (target_nid == -1) {
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3786
	migrated = migrate_misplaced_page(page, vma, target_nid);
3787
	if (migrated) {
3788
		page_nid = target_nid;
3789
		flags |= TNF_MIGRATED;
3790 3791
	} else
		flags |= TNF_MIGRATE_FAIL;
3792 3793

out:
3794
	if (page_nid != -1)
3795
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3796 3797 3798
	return 0;
}

3799
static inline int create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3800
{
3801
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3802
		return do_huge_pmd_anonymous_page(vmf);
3803
	if (vmf->vma->vm_ops->huge_fault)
3804
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
M
Matthew Wilcox 已提交
3805 3806 3807
	return VM_FAULT_FALLBACK;
}

J
Jan Kara 已提交
3808
static int wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3809
{
J
Jan Kara 已提交
3810 3811
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3812
	if (vmf->vma->vm_ops->huge_fault)
3813
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
K
Kirill A. Shutemov 已提交
3814 3815

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3816 3817
	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
3818

M
Matthew Wilcox 已提交
3819 3820 3821
	return VM_FAULT_FALLBACK;
}

3822 3823 3824 3825 3826
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3827 3828 3829 3830 3831 3832 3833
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)
3834
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845
#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)
3846
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3847 3848 3849 3850
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3851 3852 3853 3854 3855 3856 3857 3858 3859
/*
 * 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).
 *
3860 3861
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3862
 *
3863 3864
 * 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
3865
 */
J
Jan Kara 已提交
3866
static int handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3867 3868 3869
{
	pte_t entry;

J
Jan Kara 已提交
3870
	if (unlikely(pmd_none(*vmf->pmd))) {
3871 3872 3873 3874 3875 3876
		/*
		 * 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
3877
		vmf->pte = NULL;
3878 3879
	} else {
		/* See comment in pte_alloc_one_map() */
3880
		if (pmd_devmap_trans_unstable(vmf->pmd))
3881 3882 3883 3884 3885 3886 3887
			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
3888
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3889
		vmf->orig_pte = *vmf->pte;
3890 3891 3892 3893 3894 3895 3896 3897 3898 3899

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
		 * CONFIG_32BIT=y, so READ_ONCE or ACCESS_ONCE cannot guarantee
		 * atomic accesses.  The code below just needs a consistent
		 * view for the ifs and we later double check anyway with the
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
J
Jan Kara 已提交
3900
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3901 3902
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3903
		}
L
Linus Torvalds 已提交
3904 3905
	}

J
Jan Kara 已提交
3906 3907 3908
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3909
		else
J
Jan Kara 已提交
3910
			return do_fault(vmf);
3911 3912
	}

J
Jan Kara 已提交
3913 3914
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3915

J
Jan Kara 已提交
3916 3917
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3918

J
Jan Kara 已提交
3919 3920
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3921
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3922
	if (unlikely(!pte_same(*vmf->pte, entry)))
3923
		goto unlock;
J
Jan Kara 已提交
3924
	if (vmf->flags & FAULT_FLAG_WRITE) {
L
Linus Torvalds 已提交
3925
		if (!pte_write(entry))
J
Jan Kara 已提交
3926
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3927 3928 3929
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3930 3931 3932
	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);
3933 3934 3935 3936 3937 3938 3939
	} 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
3940 3941
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3942
	}
3943
unlock:
J
Jan Kara 已提交
3944
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3945
	return 0;
L
Linus Torvalds 已提交
3946 3947 3948 3949
}

/*
 * By the time we get here, we already hold the mm semaphore
3950 3951 3952
 *
 * 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
3953
 */
3954 3955
static int __handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
L
Linus Torvalds 已提交
3956
{
J
Jan Kara 已提交
3957
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3958
		.vma = vma,
3959
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3960
		.flags = flags,
3961
		.pgoff = linear_page_index(vma, address),
3962
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3963
	};
3964
	unsigned int dirty = flags & FAULT_FLAG_WRITE;
3965
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3966
	pgd_t *pgd;
3967
	p4d_t *p4d;
3968
	int ret;
L
Linus Torvalds 已提交
3969 3970

	pgd = pgd_offset(mm, address);
3971 3972 3973
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return VM_FAULT_OOM;
3974

3975
	vmf.pud = pud_alloc(mm, p4d, address);
3976
	if (!vmf.pud)
H
Hugh Dickins 已提交
3977
		return VM_FAULT_OOM;
3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001
	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 */

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

	vmf.pmd = pmd_alloc(mm, vmf.pud, address);
J
Jan Kara 已提交
4002
	if (!vmf.pmd)
H
Hugh Dickins 已提交
4003
		return VM_FAULT_OOM;
J
Jan Kara 已提交
4004
	if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) {
4005
		ret = create_huge_pmd(&vmf);
4006 4007
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
4008
	} else {
J
Jan Kara 已提交
4009
		pmd_t orig_pmd = *vmf.pmd;
4010

4011
		barrier();
4012 4013 4014 4015 4016 4017 4018
		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;
		}
4019
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
4020
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
4021
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
4022

4023
			if (dirty && !pmd_write(orig_pmd)) {
J
Jan Kara 已提交
4024
				ret = wp_huge_pmd(&vmf, orig_pmd);
4025 4026
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
4027
			} else {
J
Jan Kara 已提交
4028
				huge_pmd_set_accessed(&vmf, orig_pmd);
4029
				return 0;
4030
			}
4031 4032 4033
		}
	}

J
Jan Kara 已提交
4034
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
4035 4036
}

4037 4038 4039 4040 4041 4042
/*
 * 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().
 */
4043 4044
int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
		unsigned int flags)
4045 4046 4047 4048 4049 4050
{
	int ret;

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
4051
	count_memcg_event_mm(vma->vm_mm, PGFAULT);
4052 4053 4054 4055

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

4056 4057 4058 4059 4060
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

4061 4062 4063 4064 4065
	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
4066
		mem_cgroup_oom_enable();
4067

K
Kirill A. Shutemov 已提交
4068 4069 4070 4071
	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
4072

4073 4074
	if (flags & FAULT_FLAG_USER) {
		mem_cgroup_oom_disable();
T
Tobin C Harding 已提交
4075 4076 4077 4078 4079 4080 4081 4082
		/*
		 * 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);
4083
	}
4084

4085 4086
	return ret;
}
4087
EXPORT_SYMBOL_GPL(handle_mm_fault);
4088

K
Kirill A. Shutemov 已提交
4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111
#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
4112 4113 4114
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
4115
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4116
 */
4117
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
L
Linus Torvalds 已提交
4118
{
H
Hugh Dickins 已提交
4119 4120
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
4121
		return -ENOMEM;
L
Linus Torvalds 已提交
4122

4123 4124
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
4125
	spin_lock(&mm->page_table_lock);
4126 4127
#ifndef __ARCH_HAS_5LEVEL_HACK
	if (p4d_present(*p4d))		/* Another has populated it */
4128
		pud_free(mm, new);
4129
	else
4130 4131 4132
		p4d_populate(mm, p4d, new);
#else
	if (pgd_present(*p4d))		/* Another has populated it */
4133
		pud_free(mm, new);
4134
	else
4135 4136
		pgd_populate(mm, p4d, new);
#endif /* __ARCH_HAS_5LEVEL_HACK */
H
Hugh Dickins 已提交
4137
	spin_unlock(&mm->page_table_lock);
4138
	return 0;
L
Linus Torvalds 已提交
4139 4140 4141 4142 4143 4144
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
4145
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4146
 */
4147
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
4148
{
4149
	spinlock_t *ptl;
H
Hugh Dickins 已提交
4150 4151
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
4152
		return -ENOMEM;
L
Linus Torvalds 已提交
4153

4154 4155
	smp_wmb(); /* See comment in __pte_alloc */

4156
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
4157
#ifndef __ARCH_HAS_4LEVEL_HACK
4158 4159
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
4160
		pud_populate(mm, pud, new);
4161
	} else	/* Another has populated it */
4162
		pmd_free(mm, new);
4163 4164 4165
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
4166
		pgd_populate(mm, pud, new);
4167 4168
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
4169
#endif /* __ARCH_HAS_4LEVEL_HACK */
4170
	spin_unlock(ptl);
4171
	return 0;
4172
}
L
Linus Torvalds 已提交
4173 4174
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
4175
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4176 4177
			    unsigned long *start, unsigned long *end,
			    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
4178 4179
{
	pgd_t *pgd;
4180
	p4d_t *p4d;
J
Johannes Weiner 已提交
4181 4182 4183 4184 4185 4186 4187 4188
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

4189 4190 4191 4192 4193
	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
4194 4195 4196 4197
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
4198
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
4199

R
Ross Zwisler 已提交
4200 4201 4202 4203
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

4204 4205 4206 4207 4208
		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
4209 4210 4211 4212 4213 4214
		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
4215 4216
		if (start && end)
			mmu_notifier_invalidate_range_end(mm, *start, *end);
R
Ross Zwisler 已提交
4217 4218 4219
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
4220 4221
		goto out;

4222 4223 4224 4225 4226
	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
4227 4228 4229 4230 4231 4232 4233
	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);
4234 4235
	if (start && end)
		mmu_notifier_invalidate_range_end(mm, *start, *end);
J
Johannes Weiner 已提交
4236 4237 4238 4239
out:
	return -EINVAL;
}

4240 4241
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
4242 4243 4244 4245 4246
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4247 4248
			   !(res = __follow_pte_pmd(mm, address, NULL, NULL,
						    ptepp, NULL, ptlp)));
R
Ross Zwisler 已提交
4249 4250 4251 4252
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4253
			     unsigned long *start, unsigned long *end,
R
Ross Zwisler 已提交
4254 4255 4256 4257 4258 4259
			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4260 4261
			   !(res = __follow_pte_pmd(mm, address, start, end,
						    ptepp, pmdpp, ptlp)));
4262 4263
	return res;
}
R
Ross Zwisler 已提交
4264
EXPORT_SYMBOL(follow_pte_pmd);
4265

J
Johannes Weiner 已提交
4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294
/**
 * 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);

4295
#ifdef CONFIG_HAVE_IOREMAP_PROT
4296 4297 4298
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4299
{
4300
	int ret = -EINVAL;
4301 4302 4303
	pte_t *ptep, pte;
	spinlock_t *ptl;

4304 4305
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4306

4307
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4308
		goto out;
4309
	pte = *ptep;
4310

4311 4312 4313 4314
	if ((flags & FOLL_WRITE) && !pte_write(pte))
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4315
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4316

4317
	ret = 0;
4318 4319 4320
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4321
	return ret;
4322 4323 4324 4325 4326 4327 4328
}

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
4329
	void __iomem *maddr;
4330 4331
	int offset = addr & (PAGE_SIZE-1);

4332
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4333 4334
		return -EINVAL;

4335
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
4336 4337 4338 4339 4340 4341 4342 4343
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
4344
EXPORT_SYMBOL_GPL(generic_access_phys);
4345 4346
#endif

4347
/*
4348 4349
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4350
 */
4351
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4352
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4353 4354 4355
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4356
	int write = gup_flags & FOLL_WRITE;
4357 4358

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4359
	/* ignore errors, just check how much was successfully transferred */
4360 4361 4362
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4363
		struct page *page = NULL;
4364

4365
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4366
				gup_flags, &page, &vma, NULL);
4367
		if (ret <= 0) {
4368 4369 4370
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4371 4372 4373 4374 4375
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4376
			if (!vma || vma->vm_start > addr)
4377 4378 4379 4380 4381 4382 4383
				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;
4384
#endif
4385
		} else {
4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400
			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);
4401
			put_page(page);
4402 4403 4404 4405 4406 4407 4408 4409 4410
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4411

S
Stephen Wilson 已提交
4412
/**
4413
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4414 4415 4416 4417
 * @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
4418
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4419 4420 4421 4422
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4423
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4424
{
4425
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4426 4427
}

4428 4429 4430 4431 4432 4433
/*
 * 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,
4434
		void *buf, int len, unsigned int gup_flags)
4435 4436 4437 4438 4439 4440 4441 4442
{
	struct mm_struct *mm;
	int ret;

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

4443
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4444

4445 4446 4447 4448
	mmput(mm);

	return ret;
}
4449
EXPORT_SYMBOL_GPL(access_process_vm);
4450

4451 4452 4453 4454 4455 4456 4457 4458
/*
 * 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;

4459 4460 4461 4462 4463 4464 4465
	/*
	 * Do not print if we are in atomic
	 * contexts (in exception stacks, etc.):
	 */
	if (preempt_count())
		return;

4466 4467 4468 4469 4470 4471
	down_read(&mm->mmap_sem);
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
		char *buf = (char *)__get_free_page(GFP_KERNEL);
		if (buf) {
A
Andy Shevchenko 已提交
4472
			char *p;
4473

M
Miklos Szeredi 已提交
4474
			p = file_path(f, buf, PAGE_SIZE);
4475 4476
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4477
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4478 4479 4480 4481 4482
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4483
	up_read(&mm->mmap_sem);
4484
}
4485

4486
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4487
void __might_fault(const char *file, int line)
4488
{
4489 4490 4491 4492 4493 4494
	/*
	 * 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
4495
	if (uaccess_kernel())
4496
		return;
4497
	if (pagefault_disabled())
4498
		return;
4499 4500
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4501
	if (current->mm)
4502
		might_lock_read(&current->mm->mmap_sem);
4503
#endif
4504
}
4505
EXPORT_SYMBOL(__might_fault);
4506
#endif
A
Andrea Arcangeli 已提交
4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523

#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,
4524
		     unsigned long addr_hint, unsigned int pages_per_huge_page)
A
Andrea Arcangeli 已提交
4525
{
4526 4527 4528
	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
4529 4530 4531 4532 4533 4534

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

4535
	/* Clear sub-page to access last to keep its cache lines hot */
A
Andrea Arcangeli 已提交
4536
	might_sleep();
4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567
	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
4568
		cond_resched();
4569 4570
		clear_user_highpage(page + right_idx,
				    addr + right_idx * PAGE_SIZE);
A
Andrea Arcangeli 已提交
4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610
	}
}

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);
	}
}
4611 4612 4613

long copy_huge_page_from_user(struct page *dst_page,
				const void __user *usr_src,
4614 4615
				unsigned int pages_per_huge_page,
				bool allow_pagefault)
4616 4617 4618 4619 4620 4621 4622
{
	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++) {
4623 4624 4625 4626
		if (allow_pagefault)
			page_kaddr = kmap(dst_page + i);
		else
			page_kaddr = kmap_atomic(dst_page + i);
4627 4628 4629
		rc = copy_from_user(page_kaddr,
				(const void __user *)(src + i * PAGE_SIZE),
				PAGE_SIZE);
4630 4631 4632 4633
		if (allow_pagefault)
			kunmap(dst_page + i);
		else
			kunmap_atomic(page_kaddr);
4634 4635 4636 4637 4638 4639 4640 4641 4642

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

		cond_resched();
	}
	return ret_val;
}
A
Andrea Arcangeli 已提交
4643
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4644

4645
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4646 4647 4648 4649 4650 4651 4652 4653 4654

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);
}

4655
bool ptlock_alloc(struct page *page)
4656 4657 4658
{
	spinlock_t *ptl;

4659
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4660 4661
	if (!ptl)
		return false;
4662
	page->ptl = ptl;
4663 4664 4665
	return true;
}

4666
void ptlock_free(struct page *page)
4667
{
4668
	kmem_cache_free(page_ptl_cachep, page->ptl);
4669 4670
}
#endif