memory.c 127.9 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
/*
 *  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
34
 *              Idea by Alex Bligh (alex@cconcepts.co.uk)
L
Linus Torvalds 已提交
35 36 37 38 39 40 41 42 43
 *
 * 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>
44
#include <linux/sched/mm.h>
45
#include <linux/sched/coredump.h>
46
#include <linux/sched/numa_balancing.h>
47
#include <linux/sched/task.h>
L
Linus Torvalds 已提交
48 49 50 51 52
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
53
#include <linux/memremap.h>
H
Hugh Dickins 已提交
54
#include <linux/ksm.h>
L
Linus Torvalds 已提交
55
#include <linux/rmap.h>
56
#include <linux/export.h>
57
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
58
#include <linux/init.h>
59
#include <linux/pfn_t.h>
P
Peter Zijlstra 已提交
60
#include <linux/writeback.h>
61
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
62
#include <linux/mmu_notifier.h>
63 64
#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>
73
#include <linux/numa.h>
L
Linus Torvalds 已提交
74

75 76
#include <trace/events/kmem.h>

A
Alexey Dobriyan 已提交
77
#include <asm/io.h>
78
#include <asm/mmu_context.h>
L
Linus Torvalds 已提交
79
#include <asm/pgalloc.h>
80
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
81 82 83 84
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

85 86
#include "internal.h"

87
#if defined(LAST_CPUPID_NOT_IN_PAGE_FLAGS) && !defined(CONFIG_COMPILE_TEST)
88
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
89 90
#endif

A
Andy Whitcroft 已提交
91
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
92 93 94
/* use the per-pgdat data instead for discontigmem - mbligh */
unsigned long max_mapnr;
EXPORT_SYMBOL(max_mapnr);
T
Tobin C Harding 已提交
95 96

struct page *mem_map;
L
Linus Torvalds 已提交
97 98 99 100 101 102 103 104 105 106
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
107
void *high_memory;
L
Linus Torvalds 已提交
108 109
EXPORT_SYMBOL(high_memory);

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

123 124 125 126 127 128 129 130 131 132 133 134
#ifndef arch_faults_on_old_pte
static inline bool arch_faults_on_old_pte(void)
{
	/*
	 * Those arches which don't have hw access flag feature need to
	 * implement their own helper. By default, "true" means pagefault
	 * will be hit on old pte.
	 */
	return true;
}
#endif

135 136 137
static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
138
	return 1;
139 140 141
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
142
unsigned long zero_pfn __read_mostly;
143 144
EXPORT_SYMBOL(zero_pfn);

T
Tobin C Harding 已提交
145 146
unsigned long highest_memmap_pfn __read_mostly;

H
Hugh Dickins 已提交
147 148 149 150 151 152 153 154 155
/*
 * 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);
156

157
void mm_trace_rss_stat(struct mm_struct *mm, int member, long count)
158
{
159
	trace_rss_stat(mm, member, count);
160
}
K
KAMEZAWA Hiroyuki 已提交
161

162 163
#if defined(SPLIT_RSS_COUNTING)

164
void sync_mm_rss(struct mm_struct *mm)
165 166 167 168
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
169 170 171
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
172 173
		}
	}
174
	current->rss_stat.events = 0;
175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
}

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))
196
		sync_mm_rss(task->mm);
197
}
198
#else /* SPLIT_RSS_COUNTING */
199 200 201 202 203 204 205 206

#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)
{
}

207 208
#endif /* SPLIT_RSS_COUNTING */

L
Linus Torvalds 已提交
209 210 211 212
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
213 214
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
215
{
216
	pgtable_t token = pmd_pgtable(*pmd);
217
	pmd_clear(pmd);
218
	pte_free_tlb(tlb, token, addr);
219
	mm_dec_nr_ptes(tlb->mm);
L
Linus Torvalds 已提交
220 221
}

222 223 224
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
225 226 227
{
	pmd_t *pmd;
	unsigned long next;
228
	unsigned long start;
L
Linus Torvalds 已提交
229

230
	start = addr;
L
Linus Torvalds 已提交
231 232 233 234 235
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
236
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
237 238
	} while (pmd++, addr = next, addr != end);

239 240 241 242 243 244 245
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
246
	}
247 248 249 250 251
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
252
	pmd_free_tlb(tlb, pmd, start);
253
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
254 255
}

256
static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
257 258
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
259 260 261
{
	pud_t *pud;
	unsigned long next;
262
	unsigned long start;
L
Linus Torvalds 已提交
263

264
	start = addr;
265
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
266 267 268 269
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
270
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
271 272
	} while (pud++, addr = next, addr != end);

273 274 275 276 277 278 279 280 281 282 283 284 285 286
	start &= P4D_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= P4D_MASK;
		if (!ceiling)
			return;
	}
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(p4d, start);
	p4d_clear(p4d);
	pud_free_tlb(tlb, pud, start);
K
Kirill A. Shutemov 已提交
287
	mm_dec_nr_puds(tlb->mm);
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
}

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);

307 308 309 310 311 312 313
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
314
	}
315 316 317
	if (end - 1 > ceiling - 1)
		return;

318
	p4d = p4d_offset(pgd, start);
319
	pgd_clear(pgd);
320
	p4d_free_tlb(tlb, p4d, start);
L
Linus Torvalds 已提交
321 322 323
}

/*
324
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
325
 */
326
void free_pgd_range(struct mmu_gather *tlb,
327 328
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
329 330 331
{
	pgd_t *pgd;
	unsigned long next;
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

	/*
	 * 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
358

359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
	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;
374 375 376 377
	/*
	 * We add page table cache pages with PAGE_SIZE,
	 * (see pte_free_tlb()), flush the tlb if we need
	 */
378
	tlb_change_page_size(tlb, PAGE_SIZE);
379
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
380 381 382 383
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
384
		free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
385
	} while (pgd++, addr = next, addr != end);
386 387
}

388
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
389
		unsigned long floor, unsigned long ceiling)
390 391 392 393 394
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

395
		/*
N
npiggin@suse.de 已提交
396 397
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
398
		 */
399
		unlink_anon_vmas(vma);
400 401
		unlink_file_vma(vma);

402
		if (is_vm_hugetlb_page(vma)) {
403
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
404
				floor, next ? next->vm_start : ceiling);
405 406 407 408 409
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
410
			       && !is_vm_hugetlb_page(next)) {
411 412
				vma = next;
				next = vma->vm_next;
413
				unlink_anon_vmas(vma);
414
				unlink_file_vma(vma);
415 416
			}
			free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
417
				floor, next ? next->vm_start : ceiling);
418
		}
419 420
		vma = next;
	}
L
Linus Torvalds 已提交
421 422
}

423
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd)
L
Linus Torvalds 已提交
424
{
425
	spinlock_t *ptl;
426
	pgtable_t new = pte_alloc_one(mm);
427 428 429
	if (!new)
		return -ENOMEM;

430 431 432 433 434 435 436 437 438 439 440 441 442 443 444
	/*
	 * 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 */

445
	ptl = pmd_lock(mm, pmd);
446
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
447
		mm_inc_nr_ptes(mm);
L
Linus Torvalds 已提交
448
		pmd_populate(mm, pmd, new);
449
		new = NULL;
450
	}
451
	spin_unlock(ptl);
452 453
	if (new)
		pte_free(mm, new);
454
	return 0;
L
Linus Torvalds 已提交
455 456
}

457
int __pte_alloc_kernel(pmd_t *pmd)
L
Linus Torvalds 已提交
458
{
459
	pte_t *new = pte_alloc_one_kernel(&init_mm);
460 461 462
	if (!new)
		return -ENOMEM;

463 464
	smp_wmb(); /* See comment in __pte_alloc */

465
	spin_lock(&init_mm.page_table_lock);
466
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
467
		pmd_populate_kernel(&init_mm, pmd, new);
468
		new = NULL;
469
	}
470
	spin_unlock(&init_mm.page_table_lock);
471 472
	if (new)
		pte_free_kernel(&init_mm, new);
473
	return 0;
L
Linus Torvalds 已提交
474 475
}

K
KAMEZAWA Hiroyuki 已提交
476 477 478 479 480 481
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)
482
{
K
KAMEZAWA Hiroyuki 已提交
483 484
	int i;

485
	if (current->mm == mm)
486
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
487 488 489
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
490 491
}

N
Nick Piggin 已提交
492
/*
493 494 495
 * 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
496 497 498
 *
 * The calling function must still handle the error.
 */
499 500
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
501
{
502
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
503 504
	p4d_t *p4d = p4d_offset(pgd, addr);
	pud_t *pud = pud_offset(p4d, addr);
505 506 507
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
508 509 510 511 512 513 514 515 516 517 518 519 520 521
	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) {
522 523
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
524 525 526 527 528 529
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
530 531 532 533

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

534 535 536
	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));
537
	if (page)
538
		dump_page(page, "bad pte");
539
	pr_alert("addr:%px vm_flags:%08lx anon_vma:%px mapping:%px index:%lx\n",
540
		 (void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
541
	pr_alert("file:%pD fault:%ps mmap:%ps readpage:%ps\n",
542 543 544 545
		 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
546
	dump_stack();
547
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
548 549
}

H
Hugh Dickins 已提交
550
/*
N
Nick Piggin 已提交
551
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
552
 *
N
Nick Piggin 已提交
553 554 555
 * "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
556
 *
N
Nick Piggin 已提交
557 558 559 560 561 562 563 564
 * 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.
565
 *
J
Jared Hulbert 已提交
566 567
 * 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
568 569
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
570 571 572
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
573 574 575 576 577 578
 * 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
579 580
 *
 *
N
Nick Piggin 已提交
581
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
582 583 584 585 586 587 588 589 590
 *
 * 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
591
 */
592 593
struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
			    pte_t pte)
H
Hugh Dickins 已提交
594
{
595
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
596

L
Laurent Dufour 已提交
597
	if (IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL)) {
598
		if (likely(!pte_special(pte)))
599
			goto check_pfn;
600 601
		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
602 603
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
604 605
		if (is_zero_pfn(pfn))
			return NULL;
606 607 608
		if (pte_devmap(pte))
			return NULL;

609
		print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
610 611 612
		return NULL;
	}

L
Laurent Dufour 已提交
613
	/* !CONFIG_ARCH_HAS_PTE_SPECIAL case follows: */
N
Nick Piggin 已提交
614

J
Jared Hulbert 已提交
615 616 617 618 619 620
	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
621 622
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
623 624 625 626 627
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
628 629
	}

630 631
	if (is_zero_pfn(pfn))
		return NULL;
L
Laurent Dufour 已提交
632

633 634 635 636 637
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
638 639

	/*
N
Nick Piggin 已提交
640 641
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
642
	 */
J
Jared Hulbert 已提交
643
out:
644
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
645 646
}

647 648 649 650 651 652 653 654 655
#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
L
Laurent Dufour 已提交
656
	 * !CONFIG_ARCH_HAS_PTE_SPECIAL case from vm_normal_page() here.
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672
	 */
	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;
		}
	}

673 674
	if (pmd_devmap(pmd))
		return NULL;
675 676 677 678 679 680 681 682 683 684 685 686 687 688
	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
689 690 691 692 693 694
/*
 * 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
695
static inline unsigned long
L
Linus Torvalds 已提交
696
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
697
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
698
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
699
{
N
Nick Piggin 已提交
700
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
701 702 703 704 705
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
		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);

724
			rss[mm_counter(page)]++;
725 726 727 728 729 730 731 732 733 734 735 736

			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);
737
			}
738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
		} 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
767
		}
768
		goto out_set_pte;
L
Linus Torvalds 已提交
769 770 771 772 773 774
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
775
	if (is_cow_mapping(vm_flags) && pte_write(pte)) {
L
Linus Torvalds 已提交
776
		ptep_set_wrprotect(src_mm, addr, src_pte);
777
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
778 779 780 781 782 783 784 785 786
	}

	/*
	 * 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);
787 788 789 790

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
791
		page_dup_rmap(page, false);
792
		rss[mm_counter(page)]++;
793 794
	} else if (pte_devmap(pte)) {
		page = pte_page(pte);
795
	}
796 797 798

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
799
	return 0;
L
Linus Torvalds 已提交
800 801
}

802
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
803 804
		   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
805
{
806
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
807
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
808
	spinlock_t *src_ptl, *dst_ptl;
809
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
810
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
811
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
812 813

again:
K
KAMEZAWA Hiroyuki 已提交
814 815
	init_rss_vec(rss);

H
Hugh Dickins 已提交
816
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
817 818
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
819
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
820
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
821
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
822 823
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
824
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
825 826 827 828 829 830

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
831 832 833
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
834
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
835 836
				break;
		}
L
Linus Torvalds 已提交
837 838 839 840
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
841 842 843 844
		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
845 846 847
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

848
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
849
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
850
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
851
	add_mm_rss_vec(dst_mm, rss);
852
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
853
	cond_resched();
H
Hugh Dickins 已提交
854 855 856 857 858 859

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877
	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);
878 879
		if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
			|| pmd_devmap(*src_pmd)) {
880
			int err;
881
			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
882 883 884 885 886 887 888 889
			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
890 891 892 893 894 895 896 897 898 899
		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,
900
		p4d_t *dst_p4d, p4d_t *src_p4d, struct vm_area_struct *vma,
L
Linus Torvalds 已提交
901 902 903 904 905
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

906
	dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
L
Linus Torvalds 已提交
907 908
	if (!dst_pud)
		return -ENOMEM;
909
	src_pud = pud_offset(src_p4d, addr);
L
Linus Torvalds 已提交
910 911
	do {
		next = pud_addr_end(addr, end);
912 913 914 915 916 917 918 919 920 921 922 923
		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
924 925 926 927 928 929 930 931 932
		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;
}

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954
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
955 956 957 958 959 960 961
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;
962
	struct mmu_notifier_range range;
963
	bool is_cow;
A
Andrea Arcangeli 已提交
964
	int ret;
L
Linus Torvalds 已提交
965

966 967 968 969 970 971
	/*
	 * 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.
	 */
972 973 974
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
975

L
Linus Torvalds 已提交
976 977 978
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

979
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
980 981 982 983
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
984
		ret = track_pfn_copy(vma);
985 986 987 988
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
989 990 991 992 993 994
	/*
	 * 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.
	 */
995
	is_cow = is_cow_mapping(vma->vm_flags);
996 997

	if (is_cow) {
998 999
		mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_PAGE,
					0, vma, src_mm, addr, end);
1000 1001
		mmu_notifier_invalidate_range_start(&range);
	}
A
Andrea Arcangeli 已提交
1002 1003

	ret = 0;
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009
	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;
1010
		if (unlikely(copy_p4d_range(dst_mm, src_mm, dst_pgd, src_pgd,
A
Andrea Arcangeli 已提交
1011 1012 1013 1014
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1015
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1016

1017
	if (is_cow)
1018
		mmu_notifier_invalidate_range_end(&range);
A
Andrea Arcangeli 已提交
1019
	return ret;
L
Linus Torvalds 已提交
1020 1021
}

1022
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1023
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1024
				unsigned long addr, unsigned long end,
1025
				struct zap_details *details)
L
Linus Torvalds 已提交
1026
{
N
Nick Piggin 已提交
1027
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1028
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1029
	int rss[NR_MM_COUNTERS];
1030
	spinlock_t *ptl;
1031
	pte_t *start_pte;
1032
	pte_t *pte;
1033
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1034

1035
	tlb_change_page_size(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1036
again:
1037
	init_rss_vec(rss);
1038 1039
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1040
	flush_tlb_batched_pending(mm);
1041
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1042 1043
	do {
		pte_t ptent = *pte;
T
Tobin C Harding 已提交
1044
		if (pte_none(ptent))
L
Linus Torvalds 已提交
1045
			continue;
1046

1047 1048 1049
		if (need_resched())
			break;

L
Linus Torvalds 已提交
1050
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1051
			struct page *page;
1052

1053
			page = vm_normal_page(vma, addr, ptent);
L
Linus Torvalds 已提交
1054 1055 1056 1057 1058 1059 1060
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1061
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1062 1063
					continue;
			}
N
Nick Piggin 已提交
1064
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1065
							tlb->fullmm);
L
Linus Torvalds 已提交
1066 1067 1068
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1069 1070

			if (!PageAnon(page)) {
1071 1072
				if (pte_dirty(ptent)) {
					force_flush = 1;
1073
					set_page_dirty(page);
1074
				}
1075
				if (pte_young(ptent) &&
1076
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1077
					mark_page_accessed(page);
1078
			}
1079
			rss[mm_counter(page)]--;
1080
			page_remove_rmap(page, false);
1081 1082
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1083
			if (unlikely(__tlb_remove_page(tlb, page))) {
1084
				force_flush = 1;
1085
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1086
				break;
1087
			}
L
Linus Torvalds 已提交
1088 1089
			continue;
		}
1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112

		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;
		}

1113 1114
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1115
			continue;
K
KAMEZAWA Hiroyuki 已提交
1116

1117 1118 1119 1120
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1121

1122
			page = migration_entry_to_page(entry);
1123
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1124
		}
1125 1126
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1127
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1128
	} while (pte++, addr += PAGE_SIZE, addr != end);
1129

K
KAMEZAWA Hiroyuki 已提交
1130
	add_mm_rss_vec(mm, rss);
1131
	arch_leave_lazy_mmu_mode();
1132

1133
	/* Do the actual TLB flush before dropping ptl */
1134
	if (force_flush)
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
		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;
1146
		tlb_flush_mmu(tlb);
1147 1148 1149 1150 1151
	}

	if (addr != end) {
		cond_resched();
		goto again;
P
Peter Zijlstra 已提交
1152 1153
	}

1154
	return addr;
L
Linus Torvalds 已提交
1155 1156
}

1157
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1158
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1159
				unsigned long addr, unsigned long end,
1160
				struct zap_details *details)
L
Linus Torvalds 已提交
1161 1162 1163 1164 1165 1166 1167
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1168
		if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1169
			if (next - addr != HPAGE_PMD_SIZE)
1170
				__split_huge_pmd(vma, pmd, addr, false, NULL);
1171
			else if (zap_huge_pmd(tlb, vma, pmd, addr))
1172
				goto next;
1173 1174
			/* fall through */
		}
1175 1176 1177 1178 1179 1180 1181 1182 1183
		/*
		 * 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;
1184
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1185
next:
1186 1187
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1188 1189

	return addr;
L
Linus Torvalds 已提交
1190 1191
}

1192
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
1193
				struct vm_area_struct *vma, p4d_t *p4d,
L
Linus Torvalds 已提交
1194
				unsigned long addr, unsigned long end,
1195
				struct zap_details *details)
L
Linus Torvalds 已提交
1196 1197 1198 1199
{
	pud_t *pud;
	unsigned long next;

1200
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
1201 1202
	do {
		next = pud_addr_end(addr, end);
1203 1204 1205 1206 1207 1208 1209 1210
		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 */
		}
1211
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1212
			continue;
1213
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1214 1215
next:
		cond_resched();
1216
	} while (pud++, addr = next, addr != end);
1217 1218

	return addr;
L
Linus Torvalds 已提交
1219 1220
}

1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
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
1240
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1241 1242 1243
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252
{
	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);
1253
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1254
			continue;
1255
		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
1256
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1257 1258
	tlb_end_vma(tlb, vma);
}
1259

1260 1261 1262

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1263
		unsigned long end_addr,
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
		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;

1275 1276 1277
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1278
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1279
		untrack_pfn(vma, 0, 0);
1280 1281 1282 1283 1284 1285 1286

	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
1287
			 * cleanup path of mmap_region. When
1288
			 * hugetlbfs ->mmap method fails,
1289
			 * mmap_region() nullifies vma->vm_file
1290 1291 1292 1293
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1294
			if (vma->vm_file) {
1295
				i_mmap_lock_write(vma->vm_file->f_mapping);
1296
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1297
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1298
			}
1299 1300 1301
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1302 1303 1304 1305
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1306
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1307 1308 1309 1310
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1311
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
 *
 * 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
1322
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1323
		struct vm_area_struct *vma, unsigned long start_addr,
1324
		unsigned long end_addr)
L
Linus Torvalds 已提交
1325
{
1326
	struct mmu_notifier_range range;
L
Linus Torvalds 已提交
1327

1328 1329
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, vma->vm_mm,
				start_addr, end_addr);
1330
	mmu_notifier_invalidate_range_start(&range);
1331
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1332
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
1333
	mmu_notifier_invalidate_range_end(&range);
L
Linus Torvalds 已提交
1334 1335 1336 1337 1338
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1339
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1340
 * @size: number of bytes to zap
1341 1342
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1343
 */
1344
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1345
		unsigned long size)
L
Linus Torvalds 已提交
1346
{
1347
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1348
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1349 1350

	lru_add_drain();
1351
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
1352
				start, start + size);
1353 1354 1355 1356 1357 1358 1359
	tlb_gather_mmu(&tlb, vma->vm_mm, start, range.end);
	update_hiwater_rss(vma->vm_mm);
	mmu_notifier_invalidate_range_start(&range);
	for ( ; vma && vma->vm_start < range.end; vma = vma->vm_next)
		unmap_single_vma(&tlb, vma, start, range.end, NULL);
	mmu_notifier_invalidate_range_end(&range);
	tlb_finish_mmu(&tlb, start, range.end);
L
Linus Torvalds 已提交
1360 1361
}

1362 1363 1364 1365 1366
/**
 * 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
1367
 * @details: details of shared cache invalidation
1368 1369
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1370
 */
1371
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1372 1373
		unsigned long size, struct zap_details *details)
{
1374
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1375
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1376 1377

	lru_add_drain();
1378
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
1379
				address, address + size);
1380 1381 1382 1383 1384 1385
	tlb_gather_mmu(&tlb, vma->vm_mm, address, range.end);
	update_hiwater_rss(vma->vm_mm);
	mmu_notifier_invalidate_range_start(&range);
	unmap_single_vma(&tlb, vma, address, range.end, details);
	mmu_notifier_invalidate_range_end(&range);
	tlb_finish_mmu(&tlb, address, range.end);
L
Linus Torvalds 已提交
1386 1387
}

1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
/**
 * 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.
 *
 */
1399
void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1400 1401 1402 1403
		unsigned long size)
{
	if (address < vma->vm_start || address + size > vma->vm_end ||
	    		!(vma->vm_flags & VM_PFNMAP))
1404 1405
		return;

1406
	zap_page_range_single(vma, address, size, NULL);
1407 1408 1409
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1410
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1411
			spinlock_t **ptl)
1412
{
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
	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);
1431 1432
}

1433 1434 1435 1436 1437 1438 1439
/*
 * 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
1440 1441
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1442
{
N
Nick Piggin 已提交
1443
	struct mm_struct *mm = vma->vm_mm;
1444
	int retval;
1445
	pte_t *pte;
1446 1447
	spinlock_t *ptl;

1448
	retval = -EINVAL;
1449
	if (PageAnon(page) || PageSlab(page) || page_has_type(page))
1450
		goto out;
1451 1452
	retval = -ENOMEM;
	flush_dcache_page(page);
1453
	pte = get_locked_pte(mm, addr, &ptl);
1454
	if (!pte)
1455
		goto out;
1456 1457 1458 1459 1460 1461
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1462
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1463
	page_add_file_rmap(page, false);
1464 1465 1466 1467 1468 1469 1470 1471 1472
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

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

1473 1474 1475 1476 1477 1478
/**
 * 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
 *
1479 1480 1481 1482 1483 1484
 * 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
1485
 * (see split_page()).
1486 1487 1488 1489 1490 1491 1492 1493
 *
 * 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.
1494 1495 1496 1497 1498
 *
 * 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.
1499 1500
 *
 * Return: %0 on success, negative error code otherwise.
1501
 */
N
Nick Piggin 已提交
1502 1503
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1504 1505 1506 1507 1508
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1509 1510 1511 1512 1513
	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
1514
	return insert_page(vma, addr, page, vma->vm_page_prot);
1515
}
1516
EXPORT_SYMBOL(vm_insert_page);
1517

1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536
/*
 * __vm_map_pages - maps range of kernel pages into user vma
 * @vma: user vma to map to
 * @pages: pointer to array of source kernel pages
 * @num: number of pages in page array
 * @offset: user's requested vm_pgoff
 *
 * This allows drivers to map range of kernel pages into a user vma.
 *
 * Return: 0 on success and error code otherwise.
 */
static int __vm_map_pages(struct vm_area_struct *vma, struct page **pages,
				unsigned long num, unsigned long offset)
{
	unsigned long count = vma_pages(vma);
	unsigned long uaddr = vma->vm_start;
	int ret, i;

	/* Fail if the user requested offset is beyond the end of the object */
1537
	if (offset >= num)
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
		return -ENXIO;

	/* Fail if the user requested size exceeds available object size */
	if (count > num - offset)
		return -ENXIO;

	for (i = 0; i < count; i++) {
		ret = vm_insert_page(vma, uaddr, pages[offset + i]);
		if (ret < 0)
			return ret;
		uaddr += PAGE_SIZE;
	}

	return 0;
}

/**
 * vm_map_pages - maps range of kernel pages starts with non zero offset
 * @vma: user vma to map to
 * @pages: pointer to array of source kernel pages
 * @num: number of pages in page array
 *
 * Maps an object consisting of @num pages, catering for the user's
 * requested vm_pgoff
 *
 * If we fail to insert any page into the vma, the function will return
 * immediately leaving any previously inserted pages present.  Callers
 * from the mmap handler may immediately return the error as their caller
 * will destroy the vma, removing any successfully inserted pages. Other
 * callers should make their own arrangements for calling unmap_region().
 *
 * Context: Process context. Called by mmap handlers.
 * Return: 0 on success and error code otherwise.
 */
int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
				unsigned long num)
{
	return __vm_map_pages(vma, pages, num, vma->vm_pgoff);
}
EXPORT_SYMBOL(vm_map_pages);

/**
 * vm_map_pages_zero - map range of kernel pages starts with zero offset
 * @vma: user vma to map to
 * @pages: pointer to array of source kernel pages
 * @num: number of pages in page array
 *
 * Similar to vm_map_pages(), except that it explicitly sets the offset
 * to 0. This function is intended for the drivers that did not consider
 * vm_pgoff.
 *
 * Context: Process context. Called by mmap handlers.
 * Return: 0 on success and error code otherwise.
 */
int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
				unsigned long num)
{
	return __vm_map_pages(vma, pages, num, 0);
}
EXPORT_SYMBOL(vm_map_pages_zero);

1599
static vm_fault_t insert_pfn(struct vm_area_struct *vma, unsigned long addr,
R
Ross Zwisler 已提交
1600
			pfn_t pfn, pgprot_t prot, bool mkwrite)
N
Nick Piggin 已提交
1601 1602 1603 1604 1605 1606 1607
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte, entry;
	spinlock_t *ptl;

	pte = get_locked_pte(mm, addr, &ptl);
	if (!pte)
1608
		return VM_FAULT_OOM;
R
Ross Zwisler 已提交
1609 1610 1611 1612 1613 1614 1615
	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
J
Jan Kara 已提交
1616 1617 1618 1619
			 * mapping and we expect the PFNs to match. If they
			 * don't match, we are likely racing with block
			 * allocation and mapping invalidation so just skip the
			 * update.
R
Ross Zwisler 已提交
1620
			 */
J
Jan Kara 已提交
1621 1622
			if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
				WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
R
Ross Zwisler 已提交
1623
				goto out_unlock;
J
Jan Kara 已提交
1624
			}
1625 1626 1627 1628 1629 1630
			entry = pte_mkyoung(*pte);
			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
			if (ptep_set_access_flags(vma, addr, pte, entry, 1))
				update_mmu_cache(vma, addr, pte);
		}
		goto out_unlock;
R
Ross Zwisler 已提交
1631
	}
N
Nick Piggin 已提交
1632 1633

	/* Ok, finally just insert the thing.. */
1634 1635 1636 1637
	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
1638 1639 1640 1641 1642 1643

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

N
Nick Piggin 已提交
1644
	set_pte_at(mm, addr, pte, entry);
1645
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1646 1647 1648

out_unlock:
	pte_unmap_unlock(pte, ptl);
1649
	return VM_FAULT_NOPAGE;
N
Nick Piggin 已提交
1650 1651
}

1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
/**
 * vmf_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 vmf_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;
M
Matthew Wilcox 已提交
1664
 * vmf_insert_pfn_prot should only be used if using multiple VMAs is
1665 1666
 * impractical.
 *
M
Matthew Wilcox 已提交
1667
 * Context: Process context.  May allocate using %GFP_KERNEL.
1668 1669 1670 1671 1672
 * Return: vm_fault_t value.
 */
vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn, pgprot_t pgprot)
{
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
	/*
	 * 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.
	 */
	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));

	if (addr < vma->vm_start || addr >= vma->vm_end)
		return VM_FAULT_SIGBUS;

	if (!pfn_modify_allowed(pfn, pgprot))
		return VM_FAULT_SIGBUS;

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

1693
	return insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
1694
			false);
1695 1696
}
EXPORT_SYMBOL(vmf_insert_pfn_prot);
N
Nick Piggin 已提交
1697

M
Matthew Wilcox 已提交
1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
/**
 * vmf_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
 *
 * Similar to vm_insert_page, this allows drivers to insert individual pages
 * 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 the result of this function.
 *
 * 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.
 *
 * Context: Process context.  May allocate using %GFP_KERNEL.
 * Return: vm_fault_t value.
 */
vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
			unsigned long pfn)
{
	return vmf_insert_pfn_prot(vma, addr, pfn, vma->vm_page_prot);
}
EXPORT_SYMBOL(vmf_insert_pfn);

1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
static bool vm_mixed_ok(struct vm_area_struct *vma, pfn_t pfn)
{
	/* these checks mirror the abort conditions in vm_normal_page */
	if (vma->vm_flags & VM_MIXEDMAP)
		return true;
	if (pfn_t_devmap(pfn))
		return true;
	if (pfn_t_special(pfn))
		return true;
	if (is_zero_pfn(pfn_t_to_pfn(pfn)))
		return true;
	return false;
}

1739 1740
static vm_fault_t __vm_insert_mixed(struct vm_area_struct *vma,
		unsigned long addr, pfn_t pfn, bool mkwrite)
N
Nick Piggin 已提交
1741
{
1742
	pgprot_t pgprot = vma->vm_page_prot;
1743
	int err;
1744

1745
	BUG_ON(!vm_mixed_ok(vma, pfn));
N
Nick Piggin 已提交
1746

N
Nick Piggin 已提交
1747
	if (addr < vma->vm_start || addr >= vma->vm_end)
1748
		return VM_FAULT_SIGBUS;
1749 1750

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1751

1752
	if (!pfn_modify_allowed(pfn_t_to_pfn(pfn), pgprot))
1753
		return VM_FAULT_SIGBUS;
1754

N
Nick Piggin 已提交
1755 1756 1757 1758
	/*
	 * 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
1759 1760
	 * 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
1761
	 */
L
Laurent Dufour 已提交
1762 1763
	if (!IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL) &&
	    !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1764 1765
		struct page *page;

1766 1767 1768 1769 1770 1771
		/*
		 * 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));
1772 1773
		err = insert_page(vma, addr, page, pgprot);
	} else {
1774
		return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
N
Nick Piggin 已提交
1775
	}
R
Ross Zwisler 已提交
1776

M
Matthew Wilcox 已提交
1777 1778 1779 1780 1781 1782
	if (err == -ENOMEM)
		return VM_FAULT_OOM;
	if (err < 0 && err != -EBUSY)
		return VM_FAULT_SIGBUS;

	return VM_FAULT_NOPAGE;
N
Nick Piggin 已提交
1783
}
1784 1785 1786 1787 1788 1789

vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
		pfn_t pfn)
{
	return __vm_insert_mixed(vma, addr, pfn, false);
}
M
Matthew Wilcox 已提交
1790
EXPORT_SYMBOL(vmf_insert_mixed);
N
Nick Piggin 已提交
1791

1792 1793 1794 1795 1796 1797 1798
/*
 *  If the insertion of PTE failed because someone else already added a
 *  different entry in the mean time, we treat that as success as we assume
 *  the same entry was actually inserted.
 */
vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
		unsigned long addr, pfn_t pfn)
R
Ross Zwisler 已提交
1799
{
1800
	return __vm_insert_mixed(vma, addr, pfn, true);
R
Ross Zwisler 已提交
1801
}
1802
EXPORT_SYMBOL(vmf_insert_mixed_mkwrite);
R
Ross Zwisler 已提交
1803

L
Linus Torvalds 已提交
1804 1805 1806 1807 1808 1809 1810 1811 1812 1813
/*
 * 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
1814
	spinlock_t *ptl;
1815
	int err = 0;
L
Linus Torvalds 已提交
1816

H
Hugh Dickins 已提交
1817
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1818 1819
	if (!pte)
		return -ENOMEM;
1820
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1821 1822
	do {
		BUG_ON(!pte_none(*pte));
1823 1824 1825 1826
		if (!pfn_modify_allowed(pfn, prot)) {
			err = -EACCES;
			break;
		}
N
Nick Piggin 已提交
1827
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1828 1829
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1830
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1831
	pte_unmap_unlock(pte - 1, ptl);
1832
	return err;
L
Linus Torvalds 已提交
1833 1834 1835 1836 1837 1838 1839 1840
}

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;
1841
	int err;
L
Linus Torvalds 已提交
1842 1843 1844 1845 1846

	pfn -= addr >> PAGE_SHIFT;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
1847
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1848 1849
	do {
		next = pmd_addr_end(addr, end);
1850 1851 1852 1853
		err = remap_pte_range(mm, pmd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1854 1855 1856 1857
	} while (pmd++, addr = next, addr != end);
	return 0;
}

1858
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;
1864
	int err;
L
Linus Torvalds 已提交
1865 1866

	pfn -= addr >> PAGE_SHIFT;
1867
	pud = pud_alloc(mm, p4d, addr);
L
Linus Torvalds 已提交
1868 1869 1870 1871
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
1872 1873 1874 1875
		err = remap_pmd_range(mm, pud, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1876 1877 1878 1879
	} while (pud++, addr = next, addr != end);
	return 0;
}

1880 1881 1882 1883 1884 1885
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;
1886
	int err;
1887 1888 1889 1890 1891 1892 1893

	pfn -= addr >> PAGE_SHIFT;
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
1894 1895 1896 1897
		err = remap_pud_range(mm, p4d, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
1898 1899 1900 1901
	} while (p4d++, addr = next, addr != end);
	return 0;
}

1902 1903 1904 1905 1906 1907 1908 1909
/**
 * 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
 *
1910 1911 1912
 * Note: this is only safe if the mm semaphore is held when called.
 *
 * Return: %0 on success, negative error code otherwise.
1913
 */
L
Linus Torvalds 已提交
1914 1915 1916 1917 1918
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;
1919
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1920
	struct mm_struct *mm = vma->vm_mm;
1921
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1922 1923 1924 1925 1926 1927 1928
	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).
1929 1930 1931
	 *   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.
1932 1933 1934 1935
	 *   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
1936 1937 1938 1939
	 *
	 * 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".
1940
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1941
	 */
1942 1943 1944
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1945
		vma->vm_pgoff = pfn;
1946 1947
	}

1948
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1949
	if (err)
1950
		return -EINVAL;
L
Linus Torvalds 已提交
1951

1952
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1953 1954 1955 1956 1957 1958 1959

	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);
1960
		err = remap_p4d_range(mm, pgd, addr, next,
L
Linus Torvalds 已提交
1961 1962 1963 1964
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1965 1966

	if (err)
1967
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1968

L
Linus Torvalds 已提交
1969 1970 1971 1972
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
/**
 * 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.
1985 1986
 *
 * Return: %0 on success, negative error code otherwise.
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
 */
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);

2022 2023 2024 2025 2026 2027
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;
2028
	spinlock_t *uninitialized_var(ptl);
2029 2030 2031 2032 2033 2034 2035 2036 2037

	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));

2038 2039
	arch_enter_lazy_mmu_mode();

2040
	do {
2041
		err = fn(pte++, addr, data);
2042 2043
		if (err)
			break;
2044
	} while (addr += PAGE_SIZE, addr != end);
2045

2046 2047
	arch_leave_lazy_mmu_mode();

2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
	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
2061 2062
	BUG_ON(pud_huge(*pud));

2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074
	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;
}

2075
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
2076 2077 2078 2079 2080 2081 2082
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

2083
	pud = pud_alloc(mm, p4d, addr);
2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094
	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;
}

2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114
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;
}

2115 2116 2117 2118 2119 2120 2121 2122 2123
/*
 * 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;
2124
	unsigned long end = addr + size;
2125 2126
	int err;

2127 2128 2129
	if (WARN_ON(addr >= end))
		return -EINVAL;

2130 2131 2132
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
2133
		err = apply_to_p4d_range(mm, pgd, addr, next, fn, data);
2134 2135 2136
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2137

2138 2139 2140 2141
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2142
/*
2143 2144 2145 2146 2147
 * 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;
2148
 * and do_anonymous_page can safely check later on).
2149
 */
H
Hugh Dickins 已提交
2150
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2151 2152 2153 2154 2155
				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
2156 2157
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2158
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2159
		spin_unlock(ptl);
2160 2161 2162 2163 2164 2165
	}
#endif
	pte_unmap(page_table);
	return same;
}

2166 2167
static inline bool cow_user_page(struct page *dst, struct page *src,
				 struct vm_fault *vmf)
2168
{
2169 2170 2171 2172 2173 2174 2175 2176
	bool ret;
	void *kaddr;
	void __user *uaddr;
	bool force_mkyoung;
	struct vm_area_struct *vma = vmf->vma;
	struct mm_struct *mm = vma->vm_mm;
	unsigned long addr = vmf->address;

2177 2178
	debug_dma_assert_idle(src);

2179 2180 2181 2182 2183
	if (likely(src)) {
		copy_user_highpage(dst, src, addr, vma);
		return true;
	}

2184 2185 2186 2187 2188 2189
	/*
	 * 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.
	 */
2190 2191 2192 2193 2194 2195 2196 2197 2198 2199
	kaddr = kmap_atomic(dst);
	uaddr = (void __user *)(addr & PAGE_MASK);

	/*
	 * On architectures with software "accessed" bits, we would
	 * take a double page fault, so mark it accessed here.
	 */
	force_mkyoung = arch_faults_on_old_pte() && !pte_young(vmf->orig_pte);
	if (force_mkyoung) {
		pte_t entry;
L
Linus Torvalds 已提交
2200

2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
		vmf->pte = pte_offset_map_lock(mm, vmf->pmd, addr, &vmf->ptl);
		if (!likely(pte_same(*vmf->pte, vmf->orig_pte))) {
			/*
			 * Other thread has already handled the fault
			 * and we don't need to do anything. If it's
			 * not the case, the fault will be triggered
			 * again on the same address.
			 */
			ret = false;
			goto pte_unlock;
		}

		entry = pte_mkyoung(vmf->orig_pte);
		if (ptep_set_access_flags(vma, addr, vmf->pte, entry, 0))
			update_mmu_cache(vma, addr, vmf->pte);
	}

	/*
	 * 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)) {
L
Linus Torvalds 已提交
2225
		/*
2226 2227
		 * Give a warn in case there can be some obscure
		 * use-case
L
Linus Torvalds 已提交
2228
		 */
2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241
		WARN_ON_ONCE(1);
		clear_page(kaddr);
	}

	ret = true;

pte_unlock:
	if (force_mkyoung)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	kunmap_atomic(kaddr);
	flush_dcache_page(dst);

	return ret;
2242 2243
}

2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257
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;
}

2258 2259 2260 2261 2262 2263
/*
 * 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.
 */
2264
static vm_fault_t do_page_mkwrite(struct vm_fault *vmf)
2265
{
2266
	vm_fault_t ret;
2267 2268
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2269

2270
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2271

2272 2273 2274 2275
	if (vmf->vma->vm_file &&
	    IS_SWAPFILE(vmf->vma->vm_file->f_mapping->host))
		return VM_FAULT_SIGBUS;

2276
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2277 2278
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
	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;
}

2293 2294 2295 2296 2297
/*
 * Handle dirtying of a page in shared file mapping on a write fault.
 *
 * The function expects the page to be locked and unlocks it.
 */
2298
static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
2299
{
2300
	struct vm_area_struct *vma = vmf->vma;
2301
	struct address_space *mapping;
2302
	struct page *page = vmf->page;
2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
	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);

2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328
	if (!page_mkwrite)
		file_update_time(vma->vm_file);

	/*
	 * Throttle page dirtying rate down to writeback speed.
	 *
	 * mapping may be NULL here because some device drivers do not
	 * set page.mapping but still dirty their pages
	 *
	 * Drop the mmap_sem before waiting on IO, if we can. The file
	 * is pinning the mapping, as per above.
	 */
2329
	if ((dirtied || page_mkwrite) && mapping) {
2330 2331 2332
		struct file *fpin;

		fpin = maybe_unlock_mmap_for_io(vmf, NULL);
2333
		balance_dirty_pages_ratelimited(mapping);
2334 2335 2336 2337
		if (fpin) {
			fput(fpin);
			return VM_FAULT_RETRY;
		}
2338 2339
	}

2340
	return 0;
2341 2342
}

2343 2344 2345 2346 2347 2348 2349 2350
/*
 * 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.
 */
2351
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2352
	__releases(vmf->ptl)
2353
{
J
Jan Kara 已提交
2354
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2355
	struct page *page = vmf->page;
2356 2357 2358 2359 2360 2361 2362 2363 2364
	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
2365 2366
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2367
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2368 2369 2370
	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);
2371 2372
}

2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
/*
 * 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.
 */
2389
static vm_fault_t wp_page_copy(struct vm_fault *vmf)
2390
{
J
Jan Kara 已提交
2391
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2392
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2393
	struct page *old_page = vmf->page;
2394 2395 2396 2397
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
	struct mem_cgroup *memcg;
2398
	struct mmu_notifier_range range;
2399 2400 2401 2402

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

J
Jan Kara 已提交
2403
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2404 2405
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2406 2407 2408
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2409
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2410
				vmf->address);
2411 2412
		if (!new_page)
			goto oom;
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425

		if (!cow_user_page(new_page, old_page, vmf)) {
			/*
			 * COW failed, if the fault was solved by other,
			 * it's fine. If not, userspace would re-fault on
			 * the same address and we will handle the fault
			 * from the second attempt.
			 */
			put_page(new_page);
			if (old_page)
				put_page(old_page);
			return 0;
		}
2426 2427
	}

2428
	if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
2429 2430
		goto oom_free_new;

2431 2432
	__SetPageUptodate(new_page);

2433
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm,
2434
				vmf->address & PAGE_MASK,
2435 2436
				(vmf->address & PAGE_MASK) + PAGE_SIZE);
	mmu_notifier_invalidate_range_start(&range);
2437 2438 2439 2440

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2441
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2442
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2443 2444
		if (old_page) {
			if (!PageAnon(old_page)) {
2445 2446
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2447 2448 2449 2450 2451
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2452
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2453 2454 2455 2456 2457 2458 2459 2460
		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
2461 2462
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2463
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2464 2465 2466 2467 2468 2469
		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
2470 2471
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494
		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.
			 */
2495
			page_remove_rmap(old_page, false);
2496 2497 2498 2499 2500 2501
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2502
		mem_cgroup_cancel_charge(new_page, memcg, false);
2503 2504 2505
	}

	if (new_page)
2506
		put_page(new_page);
2507

J
Jan Kara 已提交
2508
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2509 2510 2511 2512
	/*
	 * No need to double call mmu_notifier->invalidate_range() callback as
	 * the above ptep_clear_flush_notify() did already call it.
	 */
2513
	mmu_notifier_invalidate_range_only_end(&range);
2514 2515 2516 2517 2518 2519 2520
	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 */
2521 2522
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2523 2524
			unlock_page(old_page);
		}
2525
		put_page(old_page);
2526 2527 2528
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2529
	put_page(new_page);
2530 2531
oom:
	if (old_page)
2532
		put_page(old_page);
2533 2534 2535
	return VM_FAULT_OOM;
}

2536 2537 2538 2539 2540 2541 2542 2543
/**
 * 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.
2544
 * It handles locking of PTE and modifying it.
2545 2546 2547
 *
 * The function expects the page to be locked or other protection against
 * concurrent faults / writeback (such as DAX radix tree locks).
2548 2549 2550
 *
 * Return: %VM_FAULT_WRITE on success, %0 when PTE got changed before
 * we acquired PTE lock.
2551
 */
2552
vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf)
2553 2554 2555 2556 2557 2558 2559 2560 2561 2562
{
	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);
2563
		return VM_FAULT_NOPAGE;
2564 2565
	}
	wp_page_reuse(vmf);
2566
	return 0;
2567 2568
}

2569 2570 2571 2572
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
2573
static vm_fault_t wp_pfn_shared(struct vm_fault *vmf)
2574
{
J
Jan Kara 已提交
2575
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2576

2577
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
2578
		vm_fault_t ret;
2579

J
Jan Kara 已提交
2580
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2581
		vmf->flags |= FAULT_FLAG_MKWRITE;
2582
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2583
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2584
			return ret;
2585
		return finish_mkwrite_fault(vmf);
2586
	}
2587 2588
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2589 2590
}

2591
static vm_fault_t wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2592
	__releases(vmf->ptl)
2593
{
J
Jan Kara 已提交
2594
	struct vm_area_struct *vma = vmf->vma;
2595
	vm_fault_t ret = VM_FAULT_WRITE;
2596

J
Jan Kara 已提交
2597
	get_page(vmf->page);
2598 2599

	if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
2600
		vm_fault_t tmp;
2601

J
Jan Kara 已提交
2602
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2603
		tmp = do_page_mkwrite(vmf);
2604 2605
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2606
			put_page(vmf->page);
2607 2608
			return tmp;
		}
2609
		tmp = finish_mkwrite_fault(vmf);
2610
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2611 2612
			unlock_page(vmf->page);
			put_page(vmf->page);
2613
			return tmp;
2614
		}
2615 2616
	} else {
		wp_page_reuse(vmf);
2617
		lock_page(vmf->page);
2618
	}
2619
	ret |= fault_dirty_shared_page(vmf);
2620
	put_page(vmf->page);
2621

2622
	return ret;
2623 2624
}

L
Linus Torvalds 已提交
2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
/*
 * 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.
 *
2639 2640 2641
 * 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
2642
 */
2643
static vm_fault_t do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2644
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2645
{
J
Jan Kara 已提交
2646
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2647

J
Jan Kara 已提交
2648 2649
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2650
		/*
2651 2652
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2653 2654
		 *
		 * We should not cow pages in a shared writeable mapping.
2655
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2656 2657 2658
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2659
			return wp_pfn_shared(vmf);
2660

J
Jan Kara 已提交
2661
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2662
		return wp_page_copy(vmf);
2663
	}
L
Linus Torvalds 已提交
2664

2665
	/*
P
Peter Zijlstra 已提交
2666 2667
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2668
	 */
2669
	if (PageAnon(vmf->page)) {
2670
		int total_map_swapcount;
2671 2672 2673
		if (PageKsm(vmf->page) && (PageSwapCache(vmf->page) ||
					   page_count(vmf->page) != 1))
			goto copy;
J
Jan Kara 已提交
2674 2675
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2676
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2677
			lock_page(vmf->page);
J
Jan Kara 已提交
2678 2679
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2680
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2681
				unlock_page(vmf->page);
J
Jan Kara 已提交
2682
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2683
				put_page(vmf->page);
2684
				return 0;
2685
			}
J
Jan Kara 已提交
2686
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2687
		}
2688 2689 2690 2691 2692 2693 2694 2695 2696
		if (PageKsm(vmf->page)) {
			bool reused = reuse_ksm_page(vmf->page, vmf->vma,
						     vmf->address);
			unlock_page(vmf->page);
			if (!reused)
				goto copy;
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
		}
2697 2698
		if (reuse_swap_page(vmf->page, &total_map_swapcount)) {
			if (total_map_swapcount == 1) {
2699 2700 2701 2702 2703 2704 2705
				/*
				 * 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
2706
				page_move_anon_rmap(vmf->page, vma);
2707
			}
J
Jan Kara 已提交
2708
			unlock_page(vmf->page);
2709 2710
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2711
		}
J
Jan Kara 已提交
2712
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2713
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2714
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2715
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2716
	}
2717
copy:
L
Linus Torvalds 已提交
2718 2719 2720
	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2721
	get_page(vmf->page);
2722

J
Jan Kara 已提交
2723
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2724
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2725 2726
}

2727
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2728 2729 2730
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2731
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2732 2733
}

2734
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
L
Linus Torvalds 已提交
2735 2736 2737 2738 2739
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2740
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2741 2742 2743
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2744
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2745 2746 2747 2748 2749 2750 2751
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2752
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2753 2754
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2755
				details);
L
Linus Torvalds 已提交
2756 2757 2758
	}
}

M
Matthew Wilcox 已提交
2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
/**
 * unmap_mapping_pages() - Unmap pages from processes.
 * @mapping: The address space containing pages to be unmapped.
 * @start: Index of first page to be unmapped.
 * @nr: Number of pages to be unmapped.  0 to unmap to end of file.
 * @even_cows: Whether to unmap even private COWed pages.
 *
 * Unmap the pages in this address space from any userspace process which
 * has them mmaped.  Generally, you want to remove COWed pages as well when
 * a file is being truncated, but not when invalidating pages from the page
 * cache.
 */
void unmap_mapping_pages(struct address_space *mapping, pgoff_t start,
		pgoff_t nr, bool even_cows)
{
	struct zap_details details = { };

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

	i_mmap_lock_write(mapping);
	if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
		unmap_mapping_range_tree(&mapping->i_mmap, &details);
	i_mmap_unlock_write(mapping);
}

L
Linus Torvalds 已提交
2788
/**
2789
 * unmap_mapping_range - unmap the portion of all mmaps in the specified
M
Matthew Wilcox 已提交
2790
 * address_space corresponding to the specified byte range in the underlying
2791 2792
 * file.
 *
M
Martin Waitz 已提交
2793
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2794 2795
 * @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
2796
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818
 * 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)
{
	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;
	}

M
Matthew Wilcox 已提交
2819
	unmap_mapping_pages(mapping, hba, hlen, even_cows);
L
Linus Torvalds 已提交
2820 2821 2822 2823
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2824 2825
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2826 2827 2828 2829
 * 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
2830
 */
2831
vm_fault_t do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2832
{
J
Jan Kara 已提交
2833
	struct vm_area_struct *vma = vmf->vma;
M
Minchan Kim 已提交
2834
	struct page *page = NULL, *swapcache;
2835
	struct mem_cgroup *memcg;
2836
	swp_entry_t entry;
L
Linus Torvalds 已提交
2837
	pte_t pte;
2838
	int locked;
2839
	int exclusive = 0;
2840
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
2841

M
Minchan Kim 已提交
2842
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte))
2843
		goto out;
2844

J
Jan Kara 已提交
2845
	entry = pte_to_swp_entry(vmf->orig_pte);
2846 2847
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2848 2849
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2850
		} else if (is_device_private_entry(entry)) {
2851 2852
			vmf->page = device_private_entry_to_page(entry);
			ret = vmf->page->pgmap->ops->migrate_to_ram(vmf);
2853 2854 2855
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2856
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2857
			ret = VM_FAULT_SIGBUS;
2858
		}
2859 2860
		goto out;
	}
2861 2862


2863
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
M
Minchan Kim 已提交
2864 2865
	page = lookup_swap_cache(entry, vma, vmf->address);
	swapcache = page;
2866

L
Linus Torvalds 已提交
2867
	if (!page) {
2868 2869
		struct swap_info_struct *si = swp_swap_info(entry);

2870
		if (si->flags & SWP_SYNCHRONOUS_IO &&
2871
				__swap_count(entry) == 1) {
2872
			/* skip swapcache */
2873 2874
			page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
							vmf->address);
2875 2876 2877 2878 2879 2880 2881
			if (page) {
				__SetPageLocked(page);
				__SetPageSwapBacked(page);
				set_page_private(page, entry.val);
				lru_cache_add_anon(page);
				swap_readpage(page, true);
			}
2882
		} else {
2883 2884
			page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE,
						vmf);
2885
			swapcache = page;
2886 2887
		}

L
Linus Torvalds 已提交
2888 2889
		if (!page) {
			/*
2890 2891
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2892
			 */
J
Jan Kara 已提交
2893 2894
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2895
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2896
				ret = VM_FAULT_OOM;
2897
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2898
			goto unlock;
L
Linus Torvalds 已提交
2899 2900 2901 2902
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2903
		count_vm_event(PGMAJFAULT);
2904
		count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
2905
	} else if (PageHWPoison(page)) {
2906 2907 2908 2909
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2910 2911
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2912
		goto out_release;
L
Linus Torvalds 已提交
2913 2914
	}

J
Jan Kara 已提交
2915
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2916

2917
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2918 2919 2920 2921
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2922

A
Andrea Arcangeli 已提交
2923
	/*
2924 2925 2926 2927
	 * 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
2928
	 */
2929 2930
	if (unlikely((!PageSwapCache(page) ||
			page_private(page) != entry.val)) && swapcache)
A
Andrea Arcangeli 已提交
2931 2932
		goto out_page;

J
Jan Kara 已提交
2933
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2934 2935 2936 2937
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2938 2939
	}

2940 2941
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL,
					&memcg, false)) {
2942
		ret = VM_FAULT_OOM;
2943
		goto out_page;
2944 2945
	}

L
Linus Torvalds 已提交
2946
	/*
2947
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2948
	 */
J
Jan Kara 已提交
2949 2950
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2951
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2952 2953 2954 2955 2956
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2957 2958
	}

2959 2960 2961 2962 2963 2964 2965 2966 2967
	/*
	 * 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
2968

K
Kirill A. Shutemov 已提交
2969 2970
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2971
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2972
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2973
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2974
		vmf->flags &= ~FAULT_FLAG_WRITE;
2975
		ret |= VM_FAULT_WRITE;
2976
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2977 2978
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
2979
	if (pte_swp_soft_dirty(vmf->orig_pte))
2980
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2981
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
2982
	arch_do_swap_page(vma->vm_mm, vma, vmf->address, pte, vmf->orig_pte);
J
Jan Kara 已提交
2983
	vmf->orig_pte = pte;
2984 2985 2986

	/* ksm created a completely new copy */
	if (unlikely(page != swapcache && swapcache)) {
J
Jan Kara 已提交
2987
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2988
		mem_cgroup_commit_charge(page, memcg, false, false);
2989
		lru_cache_add_active_or_unevictable(page, vma);
2990 2991 2992 2993
	} else {
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
		mem_cgroup_commit_charge(page, memcg, true, false);
		activate_page(page);
2994
	}
L
Linus Torvalds 已提交
2995

2996
	swap_free(entry);
2997 2998
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2999
		try_to_free_swap(page);
3000
	unlock_page(page);
3001
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
3002 3003 3004 3005 3006 3007 3008 3009 3010
		/*
		 * 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);
3011
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3012
	}
3013

J
Jan Kara 已提交
3014
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
3015
		ret |= do_wp_page(vmf);
3016 3017
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
3018 3019 3020 3021
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3022
	update_mmu_cache(vma, vmf->address, vmf->pte);
3023
unlock:
J
Jan Kara 已提交
3024
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
3025 3026
out:
	return ret;
3027
out_nomap:
3028
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
3029
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3030
out_page:
3031
	unlock_page(page);
3032
out_release:
3033
	put_page(page);
3034
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
3035
		unlock_page(swapcache);
3036
		put_page(swapcache);
A
Andrea Arcangeli 已提交
3037
	}
3038
	return ret;
L
Linus Torvalds 已提交
3039 3040 3041
}

/*
3042 3043 3044
 * 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
3045
 */
3046
static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3047
{
J
Jan Kara 已提交
3048
	struct vm_area_struct *vma = vmf->vma;
3049
	struct mem_cgroup *memcg;
3050
	struct page *page;
3051
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
3052 3053
	pte_t entry;

3054 3055 3056 3057
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

3058 3059 3060 3061 3062 3063 3064 3065 3066 3067
	/*
	 * 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).
	 */
3068
	if (pte_alloc(vma->vm_mm, vmf->pmd))
3069 3070 3071
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3072
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
3073 3074
		return 0;

3075
	/* Use the zero-page for reads */
J
Jan Kara 已提交
3076
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3077
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
3078
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
3079
						vma->vm_page_prot));
J
Jan Kara 已提交
3080 3081 3082
		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
3083
			goto unlock;
3084 3085 3086
		ret = check_stable_address_space(vma->vm_mm);
		if (ret)
			goto unlock;
3087 3088
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3089 3090
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
3091
		}
H
Hugh Dickins 已提交
3092 3093 3094
		goto setpte;
	}

N
Nick Piggin 已提交
3095 3096 3097
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
3098
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
3099 3100
	if (!page)
		goto oom;
3101

3102 3103
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL, &memcg,
					false))
3104 3105
		goto oom_free_page;

3106 3107 3108 3109 3110
	/*
	 * 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
3111
	__SetPageUptodate(page);
3112

N
Nick Piggin 已提交
3113
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
3114 3115
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
3116

J
Jan Kara 已提交
3117 3118 3119
	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
3120
		goto release;
H
Hugh Dickins 已提交
3121

3122 3123 3124 3125
	ret = check_stable_address_space(vma->vm_mm);
	if (ret)
		goto release;

3126 3127
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3128
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3129
		mem_cgroup_cancel_charge(page, memcg, false);
3130
		put_page(page);
J
Jan Kara 已提交
3131
		return handle_userfault(vmf, VM_UFFD_MISSING);
3132 3133
	}

K
Kirill A. Shutemov 已提交
3134
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3135
	page_add_new_anon_rmap(page, vma, vmf->address, false);
3136
	mem_cgroup_commit_charge(page, memcg, false, false);
3137
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
3138
setpte:
J
Jan Kara 已提交
3139
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
3140 3141

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3142
	update_mmu_cache(vma, vmf->address, vmf->pte);
3143
unlock:
J
Jan Kara 已提交
3144
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3145
	return ret;
3146
release:
3147
	mem_cgroup_cancel_charge(page, memcg, false);
3148
	put_page(page);
3149
	goto unlock;
3150
oom_free_page:
3151
	put_page(page);
3152
oom:
L
Linus Torvalds 已提交
3153 3154 3155
	return VM_FAULT_OOM;
}

3156 3157 3158 3159 3160
/*
 * 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().
 */
3161
static vm_fault_t __do_fault(struct vm_fault *vmf)
3162
{
J
Jan Kara 已提交
3163
	struct vm_area_struct *vma = vmf->vma;
3164
	vm_fault_t ret;
3165

3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187
	/*
	 * Preallocate pte before we take page_lock because this might lead to
	 * deadlocks for memcg reclaim which waits for pages under writeback:
	 *				lock_page(A)
	 *				SetPageWriteback(A)
	 *				unlock_page(A)
	 * lock_page(B)
	 *				lock_page(B)
	 * pte_alloc_pne
	 *   shrink_page_list
	 *     wait_on_page_writeback(A)
	 *				SetPageWriteback(B)
	 *				unlock_page(B)
	 *				# flush A, B to clear the writeback
	 */
	if (pmd_none(*vmf->pmd) && !vmf->prealloc_pte) {
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
		if (!vmf->prealloc_pte)
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

3188
	ret = vma->vm_ops->fault(vmf);
3189
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
3190
			    VM_FAULT_DONE_COW)))
3191
		return ret;
3192

3193
	if (unlikely(PageHWPoison(vmf->page))) {
3194
		if (ret & VM_FAULT_LOCKED)
3195 3196
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
3197
		vmf->page = NULL;
3198 3199 3200 3201
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
3202
		lock_page(vmf->page);
3203
	else
3204
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
3205 3206 3207 3208

	return ret;
}

3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219
/*
 * 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);
}

3220
static vm_fault_t pte_alloc_one_map(struct vm_fault *vmf)
3221
{
J
Jan Kara 已提交
3222
	struct vm_area_struct *vma = vmf->vma;
3223

J
Jan Kara 已提交
3224
	if (!pmd_none(*vmf->pmd))
3225
		goto map_pte;
J
Jan Kara 已提交
3226 3227 3228 3229
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
3230 3231 3232
			goto map_pte;
		}

3233
		mm_inc_nr_ptes(vma->vm_mm);
J
Jan Kara 已提交
3234 3235
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
3236
		vmf->prealloc_pte = NULL;
3237
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd))) {
3238 3239 3240 3241 3242
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
3243 3244 3245 3246 3247 3248 3249 3250
	 * 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.
3251
	 */
3252
	if (pmd_devmap_trans_unstable(vmf->pmd))
3253 3254
		return VM_FAULT_NOPAGE;

3255 3256 3257 3258 3259 3260 3261 3262 3263
	/*
	 * 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
3264 3265
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
3266 3267 3268
	return 0;
}

3269
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
J
Jan Kara 已提交
3270
static void deposit_prealloc_pte(struct vm_fault *vmf)
3271
{
J
Jan Kara 已提交
3272
	struct vm_area_struct *vma = vmf->vma;
3273

J
Jan Kara 已提交
3274
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
3275 3276 3277 3278
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
3279
	mm_inc_nr_ptes(vma->vm_mm);
3280
	vmf->prealloc_pte = NULL;
3281 3282
}

3283
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3284
{
J
Jan Kara 已提交
3285 3286 3287
	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
3288
	pmd_t entry;
3289 3290
	int i;
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3291 3292 3293 3294 3295 3296 3297

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

3298 3299 3300 3301
	/*
	 * 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
3302
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
3303
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
J
Jan Kara 已提交
3304
		if (!vmf->prealloc_pte)
3305 3306 3307 3308
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
3309 3310
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3311 3312 3313 3314 3315 3316 3317
		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)
3318
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3319

3320
	add_mm_counter(vma->vm_mm, mm_counter_file(page), HPAGE_PMD_NR);
K
Kirill A. Shutemov 已提交
3321
	page_add_file_rmap(page, true);
3322 3323 3324 3325
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3326
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3327

J
Jan Kara 已提交
3328
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3329

J
Jan Kara 已提交
3330
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3331 3332 3333

	/* fault is handled */
	ret = 0;
3334
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3335
out:
J
Jan Kara 已提交
3336
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3337 3338 3339
	return ret;
}
#else
3340
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3341 3342 3343 3344 3345 3346
{
	BUILD_BUG();
	return 0;
}
#endif

3347
/**
3348 3349
 * 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.
3350
 *
J
Jan Kara 已提交
3351
 * @vmf: fault environment
3352
 * @memcg: memcg to charge page (only for private mappings)
3353 3354
 * @page: page to map
 *
J
Jan Kara 已提交
3355 3356
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3357 3358 3359
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
3360 3361
 *
 * Return: %0 on success, %VM_FAULT_ code in case of error.
3362
 */
3363
vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3364
		struct page *page)
3365
{
J
Jan Kara 已提交
3366 3367
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3368
	pte_t entry;
3369
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3370

J
Jan Kara 已提交
3371
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3372
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3373 3374 3375
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3376
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3377
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3378
			return ret;
K
Kirill A. Shutemov 已提交
3379
	}
3380

J
Jan Kara 已提交
3381 3382
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3383
		if (ret)
H
Hugh Dickins 已提交
3384
			return ret;
3385 3386 3387
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3388 3389
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3390

3391 3392 3393 3394
	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
3395 3396
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3397
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3398
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3399 3400
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3401
	} else {
3402
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3403
		page_add_file_rmap(page, false);
3404
	}
J
Jan Kara 已提交
3405
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3406 3407

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3408
	update_mmu_cache(vma, vmf->address, vmf->pte);
3409

H
Hugh Dickins 已提交
3410
	return 0;
3411 3412
}

3413 3414 3415 3416 3417 3418 3419 3420 3421

/**
 * 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
3422
 * addition.
3423 3424 3425
 *
 * 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).
3426 3427
 *
 * Return: %0 on success, %VM_FAULT_ code in case of error.
3428
 */
3429
vm_fault_t finish_fault(struct vm_fault *vmf)
3430 3431
{
	struct page *page;
3432
	vm_fault_t ret = 0;
3433 3434 3435 3436 3437 3438 3439

	/* 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;
3440 3441 3442 3443 3444 3445 3446 3447 3448

	/*
	 * 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);
3449 3450 3451 3452 3453
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3454 3455
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3456 3457 3458

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3459
{
3460
	*val = fault_around_bytes;
3461 3462 3463
	return 0;
}

3464
/*
3465 3466
 * fault_around_bytes must be rounded down to the nearest page order as it's
 * what do_fault_around() expects to see.
3467
 */
3468
static int fault_around_bytes_set(void *data, u64 val)
3469
{
3470
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3471
		return -EINVAL;
3472 3473 3474 3475
	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 */
3476 3477
	return 0;
}
3478
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
3479
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3480 3481 3482

static int __init fault_around_debugfs(void)
{
3483 3484
	debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
				   &fault_around_bytes_fops);
3485 3486 3487 3488
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3489

3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504
/*
 * 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.
 *
3505 3506 3507
 * fault_around_bytes defines how many bytes we'll try to map.
 * do_fault_around() expects it to be set to a power of two less than or equal
 * to PTRS_PER_PTE.
3508
 *
3509 3510 3511 3512
 * The virtual address of the area that we map is naturally aligned to
 * fault_around_bytes rounded down to the machine page size
 * (and therefore to page order).  This way it's easier to guarantee
 * that we don't cross page table boundaries.
3513
 */
3514
static vm_fault_t do_fault_around(struct vm_fault *vmf)
3515
{
J
Jan Kara 已提交
3516
	unsigned long address = vmf->address, nr_pages, mask;
3517
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3518
	pgoff_t end_pgoff;
3519 3520
	int off;
	vm_fault_t ret = 0;
3521

3522
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3523 3524
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3525 3526
	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
3527
	start_pgoff -= off;
3528 3529

	/*
3530 3531
	 *  end_pgoff is either the end of the page table, the end of
	 *  the vma or nr_pages from start_pgoff, depending what is nearest.
3532
	 */
K
Kirill A. Shutemov 已提交
3533
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3534
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3535
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3536
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3537
			start_pgoff + nr_pages - 1);
3538

J
Jan Kara 已提交
3539
	if (pmd_none(*vmf->pmd)) {
3540
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
J
Jan Kara 已提交
3541
		if (!vmf->prealloc_pte)
3542
			goto out;
3543
		smp_wmb(); /* See comment in __pte_alloc() */
3544 3545
	}

J
Jan Kara 已提交
3546
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3547 3548

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3549
	if (pmd_trans_huge(*vmf->pmd)) {
3550 3551 3552 3553 3554
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3555
	if (!vmf->pte)
3556 3557 3558
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3559 3560
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3561
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3562
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3563
out:
J
Jan Kara 已提交
3564 3565
	vmf->address = address;
	vmf->pte = NULL;
3566
	return ret;
3567 3568
}

3569
static vm_fault_t do_read_fault(struct vm_fault *vmf)
3570
{
J
Jan Kara 已提交
3571
	struct vm_area_struct *vma = vmf->vma;
3572
	vm_fault_t ret = 0;
3573 3574 3575 3576 3577 3578

	/*
	 * 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).
	 */
3579
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3580
		ret = do_fault_around(vmf);
3581 3582
		if (ret)
			return ret;
3583
	}
3584

J
Jan Kara 已提交
3585
	ret = __do_fault(vmf);
3586 3587 3588
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3589
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3590
	unlock_page(vmf->page);
3591
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3592
		put_page(vmf->page);
3593 3594 3595
	return ret;
}

3596
static vm_fault_t do_cow_fault(struct vm_fault *vmf)
3597
{
J
Jan Kara 已提交
3598
	struct vm_area_struct *vma = vmf->vma;
3599
	vm_fault_t ret;
3600 3601 3602 3603

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

J
Jan Kara 已提交
3604 3605
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3606 3607
		return VM_FAULT_OOM;

3608
	if (mem_cgroup_try_charge_delay(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3609
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3610
		put_page(vmf->cow_page);
3611 3612 3613
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3614
	ret = __do_fault(vmf);
3615 3616
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3617 3618
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3619

3620
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3621
	__SetPageUptodate(vmf->cow_page);
3622

3623
	ret |= finish_fault(vmf);
3624 3625
	unlock_page(vmf->page);
	put_page(vmf->page);
3626 3627
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3628 3629
	return ret;
uncharge_out:
3630
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3631
	put_page(vmf->cow_page);
3632 3633 3634
	return ret;
}

3635
static vm_fault_t do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3636
{
J
Jan Kara 已提交
3637
	struct vm_area_struct *vma = vmf->vma;
3638
	vm_fault_t ret, tmp;
3639

J
Jan Kara 已提交
3640
	ret = __do_fault(vmf);
3641
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3642
		return ret;
L
Linus Torvalds 已提交
3643 3644

	/*
3645 3646
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3647
	 */
3648
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3649
		unlock_page(vmf->page);
3650
		tmp = do_page_mkwrite(vmf);
3651 3652
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3653
			put_page(vmf->page);
3654
			return tmp;
3655
		}
3656 3657
	}

3658
	ret |= finish_fault(vmf);
3659 3660
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3661 3662
		unlock_page(vmf->page);
		put_page(vmf->page);
3663
		return ret;
L
Linus Torvalds 已提交
3664
	}
N
Nick Piggin 已提交
3665

3666
	ret |= fault_dirty_shared_page(vmf);
3667
	return ret;
3668
}
3669

3670 3671 3672 3673 3674
/*
 * 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().
3675 3676
 * If mmap_sem is released, vma may become invalid (for example
 * by other thread calling munmap()).
3677
 */
3678
static vm_fault_t do_fault(struct vm_fault *vmf)
3679
{
J
Jan Kara 已提交
3680
	struct vm_area_struct *vma = vmf->vma;
3681
	struct mm_struct *vm_mm = vma->vm_mm;
3682
	vm_fault_t ret;
3683

3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713
	/*
	 * The VMA was not fully populated on mmap() or missing VM_DONTEXPAND
	 */
	if (!vma->vm_ops->fault) {
		/*
		 * If we find a migration pmd entry or a none pmd entry, which
		 * should never happen, return SIGBUS
		 */
		if (unlikely(!pmd_present(*vmf->pmd)))
			ret = VM_FAULT_SIGBUS;
		else {
			vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm,
						       vmf->pmd,
						       vmf->address,
						       &vmf->ptl);
			/*
			 * Make sure this is not a temporary clearing of pte
			 * by holding ptl and checking again. A R/M/W update
			 * of pte involves: take ptl, clearing the pte so that
			 * we don't have concurrent modification by hardware
			 * followed by an update.
			 */
			if (unlikely(pte_none(*vmf->pte)))
				ret = VM_FAULT_SIGBUS;
			else
				ret = VM_FAULT_NOPAGE;

			pte_unmap_unlock(vmf->pte, vmf->ptl);
		}
	} else if (!(vmf->flags & FAULT_FLAG_WRITE))
H
Hugh Dickins 已提交
3714 3715 3716 3717 3718 3719 3720 3721
		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) {
3722
		pte_free(vm_mm, vmf->prealloc_pte);
3723
		vmf->prealloc_pte = NULL;
H
Hugh Dickins 已提交
3724 3725
	}
	return ret;
3726 3727
}

3728
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3729 3730
				unsigned long addr, int page_nid,
				int *flags)
3731 3732 3733 3734
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3735
	if (page_nid == numa_node_id()) {
3736
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3737 3738
		*flags |= TNF_FAULT_LOCAL;
	}
3739 3740 3741 3742

	return mpol_misplaced(page, vma, addr);
}

3743
static vm_fault_t do_numa_page(struct vm_fault *vmf)
3744
{
J
Jan Kara 已提交
3745
	struct vm_area_struct *vma = vmf->vma;
3746
	struct page *page = NULL;
3747
	int page_nid = NUMA_NO_NODE;
3748
	int last_cpupid;
3749
	int target_nid;
3750
	bool migrated = false;
3751
	pte_t pte, old_pte;
3752
	bool was_writable = pte_savedwrite(vmf->orig_pte);
3753
	int flags = 0;
3754 3755

	/*
T
Tobin C Harding 已提交
3756 3757 3758 3759
	 * 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
3760 3761
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
3762
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
J
Jan Kara 已提交
3763
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3764 3765 3766
		goto out;
	}

3767 3768 3769 3770
	/*
	 * Make it present again, Depending on how arch implementes non
	 * accessible ptes, some can allow access by kernel mode.
	 */
3771 3772
	old_pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
	pte = pte_modify(old_pte, vma->vm_page_prot);
3773
	pte = pte_mkyoung(pte);
3774 3775
	if (was_writable)
		pte = pte_mkwrite(pte);
3776
	ptep_modify_prot_commit(vma, vmf->address, vmf->pte, old_pte, pte);
J
Jan Kara 已提交
3777
	update_mmu_cache(vma, vmf->address, vmf->pte);
3778

J
Jan Kara 已提交
3779
	page = vm_normal_page(vma, vmf->address, pte);
3780
	if (!page) {
J
Jan Kara 已提交
3781
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3782 3783 3784
		return 0;
	}

3785 3786
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3787
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3788 3789 3790
		return 0;
	}

3791
	/*
3792 3793 3794 3795 3796 3797
	 * 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.
3798
	 */
3799
	if (!pte_write(pte))
3800 3801
		flags |= TNF_NO_GROUP;

3802 3803 3804 3805 3806 3807 3808
	/*
	 * 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;

3809
	last_cpupid = page_cpupid_last(page);
3810
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3811
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3812
			&flags);
J
Jan Kara 已提交
3813
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3814
	if (target_nid == NUMA_NO_NODE) {
3815 3816 3817 3818 3819
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3820
	migrated = migrate_misplaced_page(page, vma, target_nid);
3821
	if (migrated) {
3822
		page_nid = target_nid;
3823
		flags |= TNF_MIGRATED;
3824 3825
	} else
		flags |= TNF_MIGRATE_FAIL;
3826 3827

out:
3828
	if (page_nid != NUMA_NO_NODE)
3829
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3830 3831 3832
	return 0;
}

3833
static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3834
{
3835
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3836
		return do_huge_pmd_anonymous_page(vmf);
3837
	if (vmf->vma->vm_ops->huge_fault)
3838
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
M
Matthew Wilcox 已提交
3839 3840 3841
	return VM_FAULT_FALLBACK;
}

3842
/* `inline' is required to avoid gcc 4.1.2 build error */
3843
static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3844
{
J
Jan Kara 已提交
3845 3846
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3847
	if (vmf->vma->vm_ops->huge_fault)
3848
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
K
Kirill A. Shutemov 已提交
3849 3850

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3851 3852
	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
3853

M
Matthew Wilcox 已提交
3854 3855 3856
	return VM_FAULT_FALLBACK;
}

3857 3858 3859 3860 3861
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3862
static vm_fault_t create_huge_pud(struct vm_fault *vmf)
3863 3864 3865 3866 3867 3868
{
#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)
3869
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3870 3871 3872 3873
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

3874
static vm_fault_t wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
3875 3876 3877 3878 3879 3880
{
#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)
3881
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3882 3883 3884 3885
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3886 3887 3888 3889 3890 3891 3892 3893 3894
/*
 * 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).
 *
3895 3896
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3897
 *
3898 3899
 * 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
3900
 */
3901
static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3902 3903 3904
{
	pte_t entry;

J
Jan Kara 已提交
3905
	if (unlikely(pmd_none(*vmf->pmd))) {
3906 3907 3908 3909 3910 3911
		/*
		 * 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
3912
		vmf->pte = NULL;
3913 3914
	} else {
		/* See comment in pte_alloc_one_map() */
3915
		if (pmd_devmap_trans_unstable(vmf->pmd))
3916 3917 3918 3919 3920 3921 3922
			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
3923
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3924
		vmf->orig_pte = *vmf->pte;
3925 3926 3927 3928

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
3929 3930 3931
		 * CONFIG_32BIT=y, so READ_ONCE cannot guarantee atomic
		 * accesses.  The code below just needs a consistent view
		 * for the ifs and we later double check anyway with the
3932 3933 3934
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
J
Jan Kara 已提交
3935
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3936 3937
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3938
		}
L
Linus Torvalds 已提交
3939 3940
	}

J
Jan Kara 已提交
3941 3942 3943
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3944
		else
J
Jan Kara 已提交
3945
			return do_fault(vmf);
3946 3947
	}

J
Jan Kara 已提交
3948 3949
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3950

J
Jan Kara 已提交
3951 3952
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3953

J
Jan Kara 已提交
3954 3955
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3956
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3957
	if (unlikely(!pte_same(*vmf->pte, entry)))
3958
		goto unlock;
J
Jan Kara 已提交
3959
	if (vmf->flags & FAULT_FLAG_WRITE) {
3960
		if (!pte_write(entry))
J
Jan Kara 已提交
3961
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3962 3963 3964
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3965 3966 3967
	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);
3968 3969 3970 3971 3972 3973 3974
	} 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
3975 3976
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3977
	}
3978
unlock:
J
Jan Kara 已提交
3979
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3980
	return 0;
L
Linus Torvalds 已提交
3981 3982 3983 3984
}

/*
 * By the time we get here, we already hold the mm semaphore
3985 3986 3987
 *
 * 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
3988
 */
3989 3990
static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags)
L
Linus Torvalds 已提交
3991
{
J
Jan Kara 已提交
3992
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3993
		.vma = vma,
3994
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3995
		.flags = flags,
3996
		.pgoff = linear_page_index(vma, address),
3997
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3998
	};
3999
	unsigned int dirty = flags & FAULT_FLAG_WRITE;
4000
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
4001
	pgd_t *pgd;
4002
	p4d_t *p4d;
4003
	vm_fault_t ret;
L
Linus Torvalds 已提交
4004 4005

	pgd = pgd_offset(mm, address);
4006 4007 4008
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return VM_FAULT_OOM;
4009

4010
	vmf.pud = pud_alloc(mm, p4d, address);
4011
	if (!vmf.pud)
H
Hugh Dickins 已提交
4012
		return VM_FAULT_OOM;
4013
retry_pud:
4014
	if (pud_none(*vmf.pud) && __transparent_hugepage_enabled(vma)) {
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025
		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 */

4026
			if (dirty && !pud_write(orig_pud)) {
4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037
				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
4038
	if (!vmf.pmd)
H
Hugh Dickins 已提交
4039
		return VM_FAULT_OOM;
4040 4041 4042 4043 4044

	/* Huge pud page fault raced with pmd_alloc? */
	if (pud_trans_unstable(vmf.pud))
		goto retry_pud;

4045
	if (pmd_none(*vmf.pmd) && __transparent_hugepage_enabled(vma)) {
4046
		ret = create_huge_pmd(&vmf);
4047 4048
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
4049
	} else {
J
Jan Kara 已提交
4050
		pmd_t orig_pmd = *vmf.pmd;
4051

4052
		barrier();
4053 4054 4055 4056 4057 4058 4059
		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;
		}
4060
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
4061
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
4062
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
4063

4064
			if (dirty && !pmd_write(orig_pmd)) {
J
Jan Kara 已提交
4065
				ret = wp_huge_pmd(&vmf, orig_pmd);
4066 4067
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
4068
			} else {
J
Jan Kara 已提交
4069
				huge_pmd_set_accessed(&vmf, orig_pmd);
4070
				return 0;
4071
			}
4072 4073 4074
		}
	}

J
Jan Kara 已提交
4075
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
4076 4077
}

4078 4079 4080 4081 4082 4083
/*
 * 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().
 */
4084
vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
4085
		unsigned int flags)
4086
{
4087
	vm_fault_t ret;
4088 4089 4090 4091

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
4092
	count_memcg_event_mm(vma->vm_mm, PGFAULT);
4093 4094 4095 4096

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

4097 4098 4099 4100 4101
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

4102 4103 4104 4105 4106
	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
4107
		mem_cgroup_enter_user_fault();
4108

K
Kirill A. Shutemov 已提交
4109 4110 4111 4112
	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
4113

4114
	if (flags & FAULT_FLAG_USER) {
4115
		mem_cgroup_exit_user_fault();
T
Tobin C Harding 已提交
4116 4117 4118 4119 4120 4121 4122 4123
		/*
		 * 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);
4124
	}
4125

4126 4127
	return ret;
}
4128
EXPORT_SYMBOL_GPL(handle_mm_fault);
4129

K
Kirill A. Shutemov 已提交
4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152
#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
4153 4154 4155
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
4156
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4157
 */
4158
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
L
Linus Torvalds 已提交
4159
{
H
Hugh Dickins 已提交
4160 4161
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
4162
		return -ENOMEM;
L
Linus Torvalds 已提交
4163

4164 4165
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
4166
	spin_lock(&mm->page_table_lock);
4167
#ifndef __ARCH_HAS_5LEVEL_HACK
K
Kirill A. Shutemov 已提交
4168 4169
	if (!p4d_present(*p4d)) {
		mm_inc_nr_puds(mm);
4170
		p4d_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4171
	} else	/* Another has populated it */
4172
		pud_free(mm, new);
K
Kirill A. Shutemov 已提交
4173 4174 4175
#else
	if (!pgd_present(*p4d)) {
		mm_inc_nr_puds(mm);
4176
		pgd_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4177 4178
	} else	/* Another has populated it */
		pud_free(mm, new);
4179
#endif /* __ARCH_HAS_5LEVEL_HACK */
H
Hugh Dickins 已提交
4180
	spin_unlock(&mm->page_table_lock);
4181
	return 0;
L
Linus Torvalds 已提交
4182 4183 4184 4185 4186 4187
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
4188
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4189
 */
4190
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
4191
{
4192
	spinlock_t *ptl;
H
Hugh Dickins 已提交
4193 4194
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
4195
		return -ENOMEM;
L
Linus Torvalds 已提交
4196

4197 4198
	smp_wmb(); /* See comment in __pte_alloc */

4199
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
4200
#ifndef __ARCH_HAS_4LEVEL_HACK
4201 4202
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
4203
		pud_populate(mm, pud, new);
4204
	} else	/* Another has populated it */
4205
		pmd_free(mm, new);
4206 4207 4208
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
4209
		pgd_populate(mm, pud, new);
4210 4211
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
4212
#endif /* __ARCH_HAS_4LEVEL_HACK */
4213
	spin_unlock(ptl);
4214
	return 0;
4215
}
L
Linus Torvalds 已提交
4216 4217
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
4218
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4219
			    struct mmu_notifier_range *range,
4220
			    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
4221 4222
{
	pgd_t *pgd;
4223
	p4d_t *p4d;
J
Johannes Weiner 已提交
4224 4225 4226 4227 4228 4229 4230 4231
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

4232 4233 4234 4235 4236
	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
4237 4238 4239 4240
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
4241
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
4242

R
Ross Zwisler 已提交
4243 4244 4245 4246
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

4247
		if (range) {
4248
			mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0,
4249 4250
						NULL, mm, address & PMD_MASK,
						(address & PMD_MASK) + PMD_SIZE);
4251
			mmu_notifier_invalidate_range_start(range);
4252
		}
R
Ross Zwisler 已提交
4253 4254 4255 4256 4257 4258
		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
4259 4260
		if (range)
			mmu_notifier_invalidate_range_end(range);
R
Ross Zwisler 已提交
4261 4262 4263
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
4264 4265
		goto out;

4266
	if (range) {
4267
		mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
4268 4269
					address & PAGE_MASK,
					(address & PAGE_MASK) + PAGE_SIZE);
4270
		mmu_notifier_invalidate_range_start(range);
4271
	}
J
Johannes Weiner 已提交
4272 4273 4274 4275 4276 4277 4278
	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);
4279 4280
	if (range)
		mmu_notifier_invalidate_range_end(range);
J
Johannes Weiner 已提交
4281 4282 4283 4284
out:
	return -EINVAL;
}

4285 4286
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
4287 4288 4289 4290 4291
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4292
			   !(res = __follow_pte_pmd(mm, address, NULL,
4293
						    ptepp, NULL, ptlp)));
R
Ross Zwisler 已提交
4294 4295 4296 4297
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4298 4299
		   struct mmu_notifier_range *range,
		   pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
R
Ross Zwisler 已提交
4300 4301 4302 4303 4304
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4305
			   !(res = __follow_pte_pmd(mm, address, range,
4306
						    ptepp, pmdpp, ptlp)));
4307 4308
	return res;
}
R
Ross Zwisler 已提交
4309
EXPORT_SYMBOL(follow_pte_pmd);
4310

J
Johannes Weiner 已提交
4311 4312 4313 4314 4315 4316 4317 4318
/**
 * 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.
 *
4319
 * Return: zero and the pfn at @pfn on success, -ve otherwise.
J
Johannes Weiner 已提交
4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339
 */
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);

4340
#ifdef CONFIG_HAVE_IOREMAP_PROT
4341 4342 4343
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4344
{
4345
	int ret = -EINVAL;
4346 4347 4348
	pte_t *ptep, pte;
	spinlock_t *ptl;

4349 4350
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4351

4352
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4353
		goto out;
4354
	pte = *ptep;
4355

4356
	if ((flags & FOLL_WRITE) && !pte_write(pte))
4357 4358 4359
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4360
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4361

4362
	ret = 0;
4363 4364 4365
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4366
	return ret;
4367 4368 4369 4370 4371 4372 4373
}

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
4374
	void __iomem *maddr;
4375 4376
	int offset = addr & (PAGE_SIZE-1);

4377
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4378 4379
		return -EINVAL;

4380
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
4381 4382 4383
	if (!maddr)
		return -ENOMEM;

4384 4385 4386 4387 4388 4389 4390 4391
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
4392
EXPORT_SYMBOL_GPL(generic_access_phys);
4393 4394
#endif

4395
/*
4396 4397
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4398
 */
4399
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4400
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4401 4402 4403
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4404
	int write = gup_flags & FOLL_WRITE;
4405

4406 4407 4408
	if (down_read_killable(&mm->mmap_sem))
		return 0;

S
Simon Arlott 已提交
4409
	/* ignore errors, just check how much was successfully transferred */
4410 4411 4412
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4413
		struct page *page = NULL;
4414

4415
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4416
				gup_flags, &page, &vma, NULL);
4417
		if (ret <= 0) {
4418 4419 4420
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4421 4422 4423 4424 4425
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4426
			if (!vma || vma->vm_start > addr)
4427 4428 4429 4430 4431 4432 4433
				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;
4434
#endif
4435
		} else {
4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450
			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);
4451
			put_page(page);
4452 4453 4454 4455 4456 4457 4458 4459 4460
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4461

S
Stephen Wilson 已提交
4462
/**
4463
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4464 4465 4466 4467
 * @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
4468
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4469 4470
 *
 * The caller must hold a reference on @mm.
4471 4472
 *
 * Return: number of bytes copied from source to destination.
S
Stephen Wilson 已提交
4473 4474
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4475
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4476
{
4477
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4478 4479
}

4480 4481 4482 4483 4484 4485
/*
 * 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,
4486
		void *buf, int len, unsigned int gup_flags)
4487 4488 4489 4490 4491 4492 4493 4494
{
	struct mm_struct *mm;
	int ret;

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

4495
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4496

4497 4498 4499 4500
	mmput(mm);

	return ret;
}
4501
EXPORT_SYMBOL_GPL(access_process_vm);
4502

4503 4504 4505 4506 4507 4508 4509 4510
/*
 * 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;

4511
	/*
4512
	 * we might be running from an atomic context so we cannot sleep
4513
	 */
4514
	if (!down_read_trylock(&mm->mmap_sem))
4515 4516
		return;

4517 4518 4519
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
4520
		char *buf = (char *)__get_free_page(GFP_NOWAIT);
4521
		if (buf) {
A
Andy Shevchenko 已提交
4522
			char *p;
4523

M
Miklos Szeredi 已提交
4524
			p = file_path(f, buf, PAGE_SIZE);
4525 4526
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4527
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4528 4529 4530 4531 4532
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4533
	up_read(&mm->mmap_sem);
4534
}
4535

4536
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4537
void __might_fault(const char *file, int line)
4538
{
4539 4540 4541 4542 4543 4544
	/*
	 * 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
4545
	if (uaccess_kernel())
4546
		return;
4547
	if (pagefault_disabled())
4548
		return;
4549 4550
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4551
	if (current->mm)
4552
		might_lock_read(&current->mm->mmap_sem);
4553
#endif
4554
}
4555
EXPORT_SYMBOL(__might_fault);
4556
#endif
A
Andrea Arcangeli 已提交
4557 4558

#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
4559 4560 4561 4562 4563 4564 4565 4566 4567
/*
 * Process all subpages of the specified huge page with the specified
 * operation.  The target subpage will be processed last to keep its
 * cache lines hot.
 */
static inline void process_huge_page(
	unsigned long addr_hint, unsigned int pages_per_huge_page,
	void (*process_subpage)(unsigned long addr, int idx, void *arg),
	void *arg)
A
Andrea Arcangeli 已提交
4568
{
4569 4570 4571
	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
4572

4573
	/* Process target subpage last to keep its cache lines hot */
A
Andrea Arcangeli 已提交
4574
	might_sleep();
4575 4576
	n = (addr_hint - addr) / PAGE_SIZE;
	if (2 * n <= pages_per_huge_page) {
4577
		/* If target subpage in first half of huge page */
4578 4579
		base = 0;
		l = n;
4580
		/* Process subpages at the end of huge page */
4581 4582
		for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
			cond_resched();
4583
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4584 4585
		}
	} else {
4586
		/* If target subpage in second half of huge page */
4587 4588
		base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
		l = pages_per_huge_page - n;
4589
		/* Process subpages at the begin of huge page */
4590 4591
		for (i = 0; i < base; i++) {
			cond_resched();
4592
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4593 4594 4595
		}
	}
	/*
4596 4597
	 * Process remaining subpages in left-right-left-right pattern
	 * towards the target subpage
4598 4599 4600 4601 4602 4603
	 */
	for (i = 0; i < l; i++) {
		int left_idx = base + i;
		int right_idx = base + 2 * l - 1 - i;

		cond_resched();
4604
		process_subpage(addr + left_idx * PAGE_SIZE, left_idx, arg);
A
Andrea Arcangeli 已提交
4605
		cond_resched();
4606
		process_subpage(addr + right_idx * PAGE_SIZE, right_idx, arg);
A
Andrea Arcangeli 已提交
4607 4608 4609
	}
}

4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645
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);
	}
}

static void clear_subpage(unsigned long addr, int idx, void *arg)
{
	struct page *page = arg;

	clear_user_highpage(page + idx, addr);
}

void clear_huge_page(struct page *page,
		     unsigned long addr_hint, unsigned int pages_per_huge_page)
{
	unsigned long addr = addr_hint &
		~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);

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

	process_huge_page(addr_hint, pages_per_huge_page, clear_subpage, page);
}

A
Andrea Arcangeli 已提交
4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664
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);
	}
}

4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678
struct copy_subpage_arg {
	struct page *dst;
	struct page *src;
	struct vm_area_struct *vma;
};

static void copy_subpage(unsigned long addr, int idx, void *arg)
{
	struct copy_subpage_arg *copy_arg = arg;

	copy_user_highpage(copy_arg->dst + idx, copy_arg->src + idx,
			   addr, copy_arg->vma);
}

A
Andrea Arcangeli 已提交
4679
void copy_user_huge_page(struct page *dst, struct page *src,
4680
			 unsigned long addr_hint, struct vm_area_struct *vma,
A
Andrea Arcangeli 已提交
4681 4682
			 unsigned int pages_per_huge_page)
{
4683 4684 4685 4686 4687 4688 4689
	unsigned long addr = addr_hint &
		~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
	struct copy_subpage_arg arg = {
		.dst = dst,
		.src = src,
		.vma = vma,
	};
A
Andrea Arcangeli 已提交
4690 4691 4692 4693 4694 4695 4696

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

4697
	process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
A
Andrea Arcangeli 已提交
4698
}
4699 4700 4701

long copy_huge_page_from_user(struct page *dst_page,
				const void __user *usr_src,
4702 4703
				unsigned int pages_per_huge_page,
				bool allow_pagefault)
4704 4705 4706 4707 4708 4709 4710
{
	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++) {
4711 4712 4713 4714
		if (allow_pagefault)
			page_kaddr = kmap(dst_page + i);
		else
			page_kaddr = kmap_atomic(dst_page + i);
4715 4716 4717
		rc = copy_from_user(page_kaddr,
				(const void __user *)(src + i * PAGE_SIZE),
				PAGE_SIZE);
4718 4719 4720 4721
		if (allow_pagefault)
			kunmap(dst_page + i);
		else
			kunmap_atomic(page_kaddr);
4722 4723 4724 4725 4726 4727 4728 4729 4730

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

		cond_resched();
	}
	return ret_val;
}
A
Andrea Arcangeli 已提交
4731
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4732

4733
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4734 4735 4736 4737 4738 4739 4740 4741 4742

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);
}

4743
bool ptlock_alloc(struct page *page)
4744 4745 4746
{
	spinlock_t *ptl;

4747
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4748 4749
	if (!ptl)
		return false;
4750
	page->ptl = ptl;
4751 4752 4753
	return true;
}

4754
void ptlock_free(struct page *page)
4755
{
4756
	kmem_cache_free(page_ptl_cachep, page->ptl);
4757 4758
}
#endif