memory.c 123.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 *  linux/mm/memory.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

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

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

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

/*
 * 05.04.94  -  Multi-page memory management added for v1.1.
T
Tobin C Harding 已提交
33
 *              Idea by Alex Bligh (alex@cconcepts.co.uk)
L
Linus Torvalds 已提交
34 35 36 37 38 39 40 41 42
 *
 * 16.07.99  -  Support of BIGMEM added by Gerhard Wichert, Siemens AG
 *		(Gerhard.Wichert@pdb.siemens.de)
 *
 * Aug/Sep 2004 Changed to four level page tables (Andi Kleen)
 */

#include <linux/kernel_stat.h>
#include <linux/mm.h>
43
#include <linux/sched/mm.h>
44
#include <linux/sched/coredump.h>
45
#include <linux/sched/numa_balancing.h>
46
#include <linux/sched/task.h>
L
Linus Torvalds 已提交
47 48 49 50 51
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
52
#include <linux/memremap.h>
H
Hugh Dickins 已提交
53
#include <linux/ksm.h>
L
Linus Torvalds 已提交
54
#include <linux/rmap.h>
55
#include <linux/export.h>
56
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
57
#include <linux/init.h>
58
#include <linux/pfn_t.h>
P
Peter Zijlstra 已提交
59
#include <linux/writeback.h>
60
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
61
#include <linux/mmu_notifier.h>
62 63
#include <linux/swapops.h>
#include <linux/elf.h>
64
#include <linux/gfp.h>
65
#include <linux/migrate.h>
A
Andy Shevchenko 已提交
66
#include <linux/string.h>
67
#include <linux/dma-debug.h>
68
#include <linux/debugfs.h>
69
#include <linux/userfaultfd_k.h>
70
#include <linux/dax.h>
71
#include <linux/oom.h>
72
#include <linux/numa.h>
L
Linus Torvalds 已提交
73

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

82 83
#include "internal.h"

84
#if defined(LAST_CPUPID_NOT_IN_PAGE_FLAGS) && !defined(CONFIG_COMPILE_TEST)
85
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
86 87
#endif

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

struct page *mem_map;
L
Linus Torvalds 已提交
94 95 96 97 98 99 100 101 102 103
EXPORT_SYMBOL(mem_map);
#endif

/*
 * A number of key systems in x86 including ioremap() rely on the assumption
 * that high_memory defines the upper bound on direct map memory, then end
 * of ZONE_NORMAL.  Under CONFIG_DISCONTIG this means that max_low_pfn and
 * highstart_pfn must be the same; there must be no gap between ZONE_NORMAL
 * and ZONE_HIGHMEM.
 */
T
Tobin C Harding 已提交
104
void *high_memory;
L
Linus Torvalds 已提交
105 106
EXPORT_SYMBOL(high_memory);

107 108 109 110 111 112 113 114 115 116 117 118
/*
 * Randomize the address space (stacks, mmaps, brk, etc.).
 *
 * ( When CONFIG_COMPAT_BRK=y we exclude brk from randomization,
 *   as ancient (libc5 based) binaries can segfault. )
 */
int randomize_va_space __read_mostly =
#ifdef CONFIG_COMPAT_BRK
					1;
#else
					2;
#endif
119 120 121 122

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

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

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

H
Hugh Dickins 已提交
132 133 134 135 136 137 138 139 140
/*
 * CONFIG_MMU architectures set up ZERO_PAGE in their paging_init()
 */
static int __init init_zero_pfn(void)
{
	zero_pfn = page_to_pfn(ZERO_PAGE(0));
	return 0;
}
core_initcall(init_zero_pfn);
141

K
KAMEZAWA Hiroyuki 已提交
142

143 144
#if defined(SPLIT_RSS_COUNTING)

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

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

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

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

/* sync counter once per 64 page faults */
#define TASK_RSS_EVENTS_THRESH	(64)
static void check_sync_rss_stat(struct task_struct *task)
{
	if (unlikely(task != current))
		return;
	if (unlikely(task->rss_stat.events++ > TASK_RSS_EVENTS_THRESH))
177
		sync_mm_rss(task->mm);
178
}
179
#else /* SPLIT_RSS_COUNTING */
180 181 182 183 184 185 186 187

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

static void check_sync_rss_stat(struct task_struct *task)
{
}

188 189
#endif /* SPLIT_RSS_COUNTING */

L
Linus Torvalds 已提交
190 191 192 193
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
194 195
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
196
{
197
	pgtable_t token = pmd_pgtable(*pmd);
198
	pmd_clear(pmd);
199
	pte_free_tlb(tlb, token, addr);
200
	mm_dec_nr_ptes(tlb->mm);
L
Linus Torvalds 已提交
201 202
}

203 204 205
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
206 207 208
{
	pmd_t *pmd;
	unsigned long next;
209
	unsigned long start;
L
Linus Torvalds 已提交
210

211
	start = addr;
L
Linus Torvalds 已提交
212 213 214 215 216
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
217
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
218 219
	} while (pmd++, addr = next, addr != end);

220 221 222 223 224 225 226
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
227
	}
228 229 230 231 232
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
233
	pmd_free_tlb(tlb, pmd, start);
234
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
235 236
}

237
static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
238 239
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
240 241 242
{
	pud_t *pud;
	unsigned long next;
243
	unsigned long start;
L
Linus Torvalds 已提交
244

245
	start = addr;
246
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
247 248 249 250
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
251
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
252 253
	} while (pud++, addr = next, addr != end);

254 255 256 257 258 259 260 261 262 263 264 265 266 267
	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
268
	mm_dec_nr_puds(tlb->mm);
269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
}

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);

288 289 290 291 292 293 294
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
295
	}
296 297 298
	if (end - 1 > ceiling - 1)
		return;

299
	p4d = p4d_offset(pgd, start);
300
	pgd_clear(pgd);
301
	p4d_free_tlb(tlb, p4d, start);
L
Linus Torvalds 已提交
302 303 304
}

/*
305
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
306
 */
307
void free_pgd_range(struct mmu_gather *tlb,
308 309
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
310 311 312
{
	pgd_t *pgd;
	unsigned long next;
313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338

	/*
	 * 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
339

340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
	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;
355 356 357 358 359
	/*
	 * We add page table cache pages with PAGE_SIZE,
	 * (see pte_free_tlb()), flush the tlb if we need
	 */
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
360
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
361 362 363 364
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
365
		free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
366
	} while (pgd++, addr = next, addr != end);
367 368
}

369
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
370
		unsigned long floor, unsigned long ceiling)
371 372 373 374 375
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

376
		/*
N
npiggin@suse.de 已提交
377 378
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
379
		 */
380
		unlink_anon_vmas(vma);
381 382
		unlink_file_vma(vma);

383
		if (is_vm_hugetlb_page(vma)) {
384
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
385
				floor, next ? next->vm_start : ceiling);
386 387 388 389 390
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
391
			       && !is_vm_hugetlb_page(next)) {
392 393
				vma = next;
				next = vma->vm_next;
394
				unlink_anon_vmas(vma);
395
				unlink_file_vma(vma);
396 397
			}
			free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
398
				floor, next ? next->vm_start : ceiling);
399
		}
400 401
		vma = next;
	}
L
Linus Torvalds 已提交
402 403
}

404
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd)
L
Linus Torvalds 已提交
405
{
406
	spinlock_t *ptl;
407
	pgtable_t new = pte_alloc_one(mm);
408 409 410
	if (!new)
		return -ENOMEM;

411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
	/*
	 * 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 */

426
	ptl = pmd_lock(mm, pmd);
427
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
428
		mm_inc_nr_ptes(mm);
L
Linus Torvalds 已提交
429
		pmd_populate(mm, pmd, new);
430
		new = NULL;
431
	}
432
	spin_unlock(ptl);
433 434
	if (new)
		pte_free(mm, new);
435
	return 0;
L
Linus Torvalds 已提交
436 437
}

438
int __pte_alloc_kernel(pmd_t *pmd)
L
Linus Torvalds 已提交
439
{
440
	pte_t *new = pte_alloc_one_kernel(&init_mm);
441 442 443
	if (!new)
		return -ENOMEM;

444 445
	smp_wmb(); /* See comment in __pte_alloc */

446
	spin_lock(&init_mm.page_table_lock);
447
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
448
		pmd_populate_kernel(&init_mm, pmd, new);
449
		new = NULL;
450
	}
451
	spin_unlock(&init_mm.page_table_lock);
452 453
	if (new)
		pte_free_kernel(&init_mm, new);
454
	return 0;
L
Linus Torvalds 已提交
455 456
}

K
KAMEZAWA Hiroyuki 已提交
457 458 459 460 461 462
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)
463
{
K
KAMEZAWA Hiroyuki 已提交
464 465
	int i;

466
	if (current->mm == mm)
467
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
468 469 470
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
471 472
}

N
Nick Piggin 已提交
473
/*
474 475 476
 * 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
477 478 479
 *
 * The calling function must still handle the error.
 */
480 481
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
482
{
483
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
484 485
	p4d_t *p4d = p4d_offset(pgd, addr);
	pud_t *pud = pud_offset(p4d, addr);
486 487 488
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
489 490 491 492 493 494 495 496 497 498 499 500 501 502
	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) {
503 504
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
505 506 507 508 509 510
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
511 512 513 514

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

515 516 517
	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));
518
	if (page)
519
		dump_page(page, "bad pte");
520 521
	pr_alert("addr:%p vm_flags:%08lx anon_vma:%p mapping:%p index:%lx\n",
		 (void *)addr, vma->vm_flags, vma->anon_vma, mapping, index);
522 523 524 525 526
	pr_alert("file:%pD fault:%pf mmap:%pf readpage:%pf\n",
		 vma->vm_file,
		 vma->vm_ops ? vma->vm_ops->fault : NULL,
		 vma->vm_file ? vma->vm_file->f_op->mmap : NULL,
		 mapping ? mapping->a_ops->readpage : NULL);
N
Nick Piggin 已提交
527
	dump_stack();
528
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
529 530
}

H
Hugh Dickins 已提交
531
/*
N
Nick Piggin 已提交
532
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
533
 *
N
Nick Piggin 已提交
534 535 536
 * "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
537
 *
N
Nick Piggin 已提交
538 539 540 541 542 543 544 545
 * 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.
546
 *
J
Jared Hulbert 已提交
547 548
 * 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
549 550
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
551 552 553
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
554 555 556 557 558 559
 * 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
560 561
 *
 *
N
Nick Piggin 已提交
562
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
563 564 565 566 567 568 569 570 571
 *
 * 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
572
 */
573 574
struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
			     pte_t pte, bool with_public_device)
H
Hugh Dickins 已提交
575
{
576
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
577

L
Laurent Dufour 已提交
578
	if (IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL)) {
579
		if (likely(!pte_special(pte)))
580
			goto check_pfn;
581 582
		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
583 584
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
		if (is_zero_pfn(pfn))
			return NULL;

		/*
		 * Device public pages are special pages (they are ZONE_DEVICE
		 * pages but different from persistent memory). They behave
		 * allmost like normal pages. The difference is that they are
		 * not on the lru and thus should never be involve with any-
		 * thing that involve lru manipulation (mlock, numa balancing,
		 * ...).
		 *
		 * This is why we still want to return NULL for such page from
		 * vm_normal_page() so that we do not have to special case all
		 * call site of vm_normal_page().
		 */
600
		if (likely(pfn <= highest_memmap_pfn)) {
601 602 603 604 605 606 607 608
			struct page *page = pfn_to_page(pfn);

			if (is_device_public_page(page)) {
				if (with_public_device)
					return page;
				return NULL;
			}
		}
609 610 611 612

		if (pte_devmap(pte))
			return NULL;

613
		print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
614 615 616
		return NULL;
	}

L
Laurent Dufour 已提交
617
	/* !CONFIG_ARCH_HAS_PTE_SPECIAL case follows: */
N
Nick Piggin 已提交
618

J
Jared Hulbert 已提交
619 620 621 622 623 624
	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
625 626
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
627 628 629 630 631
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
632 633
	}

634 635
	if (is_zero_pfn(pfn))
		return NULL;
L
Laurent Dufour 已提交
636

637 638 639 640 641
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
642 643

	/*
N
Nick Piggin 已提交
644 645
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
646
	 */
J
Jared Hulbert 已提交
647
out:
648
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
649 650
}

651 652 653 654 655 656 657 658 659
#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
660
	 * !CONFIG_ARCH_HAS_PTE_SPECIAL case from vm_normal_page() here.
661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676
	 */
	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;
		}
	}

677 678
	if (pmd_devmap(pmd))
		return NULL;
679 680 681 682 683 684 685 686 687 688 689 690 691 692
	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
693 694 695 696 697 698
/*
 * 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
699
static inline unsigned long
L
Linus Torvalds 已提交
700
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
701
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
702
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
703
{
N
Nick Piggin 已提交
704
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
705 706 707 708 709
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
		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);

728
			rss[mm_counter(page)]++;
729 730 731 732 733 734 735 736 737 738 739 740

			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);
741
			}
742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770
		} 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
771
		}
772
		goto out_set_pte;
L
Linus Torvalds 已提交
773 774 775 776 777 778
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
779
	if (is_cow_mapping(vm_flags) && pte_write(pte)) {
L
Linus Torvalds 已提交
780
		ptep_set_wrprotect(src_mm, addr, src_pte);
781
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
782 783 784 785 786 787 788 789 790
	}

	/*
	 * 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);
791 792 793 794

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
795
		page_dup_rmap(page, false);
796
		rss[mm_counter(page)]++;
797 798 799 800 801 802 803 804 805 806 807 808 809
	} else if (pte_devmap(pte)) {
		page = pte_page(pte);

		/*
		 * Cache coherent device memory behave like regular page and
		 * not like persistent memory page. For more informations see
		 * MEMORY_DEVICE_CACHE_COHERENT in memory_hotplug.h
		 */
		if (is_device_public_page(page)) {
			get_page(page);
			page_dup_rmap(page, false);
			rss[mm_counter(page)]++;
		}
810
	}
811 812 813

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
814
	return 0;
L
Linus Torvalds 已提交
815 816
}

817
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
818 819
		   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
820
{
821
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
822
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
823
	spinlock_t *src_ptl, *dst_ptl;
824
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
825
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
826
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
827 828

again:
K
KAMEZAWA Hiroyuki 已提交
829 830
	init_rss_vec(rss);

H
Hugh Dickins 已提交
831
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
832 833
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
834
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
835
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
836
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
837 838
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
839
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
840 841 842 843 844 845

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
846 847 848
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
849
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
850 851
				break;
		}
L
Linus Torvalds 已提交
852 853 854 855
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
856 857 858 859
		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
860 861 862
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

863
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
864
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
865
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
866
	add_mm_rss_vec(dst_mm, rss);
867
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
868
	cond_resched();
H
Hugh Dickins 已提交
869 870 871 872 873 874

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
	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);
893 894
		if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
			|| pmd_devmap(*src_pmd)) {
895
			int err;
896
			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
897 898 899 900 901 902 903 904
			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
905 906 907 908 909 910 911 912 913 914
		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,
915
		p4d_t *dst_p4d, p4d_t *src_p4d, struct vm_area_struct *vma,
L
Linus Torvalds 已提交
916 917 918 919 920
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

921
	dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
L
Linus Torvalds 已提交
922 923
	if (!dst_pud)
		return -ENOMEM;
924
	src_pud = pud_offset(src_p4d, addr);
L
Linus Torvalds 已提交
925 926
	do {
		next = pud_addr_end(addr, end);
927 928 929 930 931 932 933 934 935 936 937 938
		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
939 940 941 942 943 944 945 946 947
		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;
}

948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
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
970 971 972 973 974 975 976
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;
977
	struct mmu_notifier_range range;
978
	bool is_cow;
A
Andrea Arcangeli 已提交
979
	int ret;
L
Linus Torvalds 已提交
980

981 982 983 984 985 986
	/*
	 * 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.
	 */
987 988 989
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
990

L
Linus Torvalds 已提交
991 992 993
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

994
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
995 996 997 998
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
999
		ret = track_pfn_copy(vma);
1000 1001 1002 1003
		if (ret)
			return ret;
	}

A
Andrea Arcangeli 已提交
1004 1005 1006 1007 1008 1009
	/*
	 * 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.
	 */
1010
	is_cow = is_cow_mapping(vma->vm_flags);
1011 1012 1013 1014 1015

	if (is_cow) {
		mmu_notifier_range_init(&range, src_mm, addr, end);
		mmu_notifier_invalidate_range_start(&range);
	}
A
Andrea Arcangeli 已提交
1016 1017

	ret = 0;
L
Linus Torvalds 已提交
1018 1019 1020 1021 1022 1023
	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;
1024
		if (unlikely(copy_p4d_range(dst_mm, src_mm, dst_pgd, src_pgd,
A
Andrea Arcangeli 已提交
1025 1026 1027 1028
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1029
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1030

1031
	if (is_cow)
1032
		mmu_notifier_invalidate_range_end(&range);
A
Andrea Arcangeli 已提交
1033
	return ret;
L
Linus Torvalds 已提交
1034 1035
}

1036
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1037
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1038
				unsigned long addr, unsigned long end,
1039
				struct zap_details *details)
L
Linus Torvalds 已提交
1040
{
N
Nick Piggin 已提交
1041
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1042
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1043
	int rss[NR_MM_COUNTERS];
1044
	spinlock_t *ptl;
1045
	pte_t *start_pte;
1046
	pte_t *pte;
1047
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1048

1049
	tlb_remove_check_page_size_change(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1050
again:
1051
	init_rss_vec(rss);
1052 1053
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1054
	flush_tlb_batched_pending(mm);
1055
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1056 1057
	do {
		pte_t ptent = *pte;
T
Tobin C Harding 已提交
1058
		if (pte_none(ptent))
L
Linus Torvalds 已提交
1059
			continue;
1060

L
Linus Torvalds 已提交
1061
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1062
			struct page *page;
1063

1064
			page = _vm_normal_page(vma, addr, ptent, true);
L
Linus Torvalds 已提交
1065 1066 1067 1068 1069 1070 1071
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1072
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1073 1074
					continue;
			}
N
Nick Piggin 已提交
1075
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1076
							tlb->fullmm);
L
Linus Torvalds 已提交
1077 1078 1079
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1080 1081

			if (!PageAnon(page)) {
1082 1083
				if (pte_dirty(ptent)) {
					force_flush = 1;
1084
					set_page_dirty(page);
1085
				}
1086
				if (pte_young(ptent) &&
1087
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1088
					mark_page_accessed(page);
1089
			}
1090
			rss[mm_counter(page)]--;
1091
			page_remove_rmap(page, false);
1092 1093
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1094
			if (unlikely(__tlb_remove_page(tlb, page))) {
1095
				force_flush = 1;
1096
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1097
				break;
1098
			}
L
Linus Torvalds 已提交
1099 1100
			continue;
		}
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123

		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;
		}

1124 1125
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1126
			continue;
K
KAMEZAWA Hiroyuki 已提交
1127

1128 1129 1130 1131 1132
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1133

1134
			page = migration_entry_to_page(entry);
1135
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1136
		}
1137 1138
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1139
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1140
	} while (pte++, addr += PAGE_SIZE, addr != end);
1141

K
KAMEZAWA Hiroyuki 已提交
1142
	add_mm_rss_vec(mm, rss);
1143
	arch_leave_lazy_mmu_mode();
1144

1145
	/* Do the actual TLB flush before dropping ptl */
1146
	if (force_flush)
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		tlb_flush_mmu_tlbonly(tlb);
	pte_unmap_unlock(start_pte, ptl);

	/*
	 * If we forced a TLB flush (either due to running out of
	 * batch buffers or because we needed to flush dirty TLB
	 * entries before releasing the ptl), free the batched
	 * memory too. Restart if we didn't do everything.
	 */
	if (force_flush) {
		force_flush = 0;
		tlb_flush_mmu_free(tlb);
1159
		if (addr != end)
P
Peter Zijlstra 已提交
1160 1161 1162
			goto again;
	}

1163
	return addr;
L
Linus Torvalds 已提交
1164 1165
}

1166
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1167
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1168
				unsigned long addr, unsigned long end,
1169
				struct zap_details *details)
L
Linus Torvalds 已提交
1170 1171 1172 1173 1174 1175 1176
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1177
		if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1178
			if (next - addr != HPAGE_PMD_SIZE)
1179
				__split_huge_pmd(vma, pmd, addr, false, NULL);
1180
			else if (zap_huge_pmd(tlb, vma, pmd, addr))
1181
				goto next;
1182 1183
			/* fall through */
		}
1184 1185 1186 1187 1188 1189 1190 1191 1192
		/*
		 * 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;
1193
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1194
next:
1195 1196
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1197 1198

	return addr;
L
Linus Torvalds 已提交
1199 1200
}

1201
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
1202
				struct vm_area_struct *vma, p4d_t *p4d,
L
Linus Torvalds 已提交
1203
				unsigned long addr, unsigned long end,
1204
				struct zap_details *details)
L
Linus Torvalds 已提交
1205 1206 1207 1208
{
	pud_t *pud;
	unsigned long next;

1209
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
1210 1211
	do {
		next = pud_addr_end(addr, end);
1212 1213 1214 1215 1216 1217 1218 1219
		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 */
		}
1220
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1221
			continue;
1222
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1223 1224
next:
		cond_resched();
1225
	} while (pud++, addr = next, addr != end);
1226 1227

	return addr;
L
Linus Torvalds 已提交
1228 1229
}

1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
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
1249
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1250 1251 1252
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261
{
	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);
1262
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1263
			continue;
1264
		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
1265
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1266 1267
	tlb_end_vma(tlb, vma);
}
1268

1269 1270 1271

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1272
		unsigned long end_addr,
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283
		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;

1284 1285 1286
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1287
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1288
		untrack_pfn(vma, 0, 0);
1289 1290 1291 1292 1293 1294 1295

	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
1296
			 * cleanup path of mmap_region. When
1297
			 * hugetlbfs ->mmap method fails,
1298
			 * mmap_region() nullifies vma->vm_file
1299 1300 1301 1302
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1303
			if (vma->vm_file) {
1304
				i_mmap_lock_write(vma->vm_file->f_mapping);
1305
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1306
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1307
			}
1308 1309 1310
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1311 1312 1313 1314
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1315
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1316 1317 1318 1319
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1320
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
 *
 * 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
1331
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1332
		struct vm_area_struct *vma, unsigned long start_addr,
1333
		unsigned long end_addr)
L
Linus Torvalds 已提交
1334
{
1335
	struct mmu_notifier_range range;
L
Linus Torvalds 已提交
1336

1337 1338
	mmu_notifier_range_init(&range, vma->vm_mm, start_addr, end_addr);
	mmu_notifier_invalidate_range_start(&range);
1339
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1340
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
1341
	mmu_notifier_invalidate_range_end(&range);
L
Linus Torvalds 已提交
1342 1343 1344 1345 1346
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1347
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1348
 * @size: number of bytes to zap
1349 1350
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1351
 */
1352
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1353
		unsigned long size)
L
Linus Torvalds 已提交
1354
{
1355
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1356
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1357 1358

	lru_add_drain();
1359 1360 1361 1362 1363 1364 1365 1366
	mmu_notifier_range_init(&range, vma->vm_mm, start, start + size);
	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
1367 1368
}

1369 1370 1371 1372 1373
/**
 * 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
1374
 * @details: details of shared cache invalidation
1375 1376
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1377
 */
1378
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1379 1380
		unsigned long size, struct zap_details *details)
{
1381
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1382
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1383 1384

	lru_add_drain();
1385 1386 1387 1388 1389 1390 1391
	mmu_notifier_range_init(&range, vma->vm_mm, address, address + size);
	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
1392 1393
}

1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404
/**
 * 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.
 *
 */
1405
void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1406 1407 1408 1409
		unsigned long size)
{
	if (address < vma->vm_start || address + size > vma->vm_end ||
	    		!(vma->vm_flags & VM_PFNMAP))
1410 1411
		return;

1412
	zap_page_range_single(vma, address, size, NULL);
1413 1414 1415
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1416
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1417
			spinlock_t **ptl)
1418
{
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	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);
1437 1438
}

1439 1440 1441 1442 1443 1444 1445
/*
 * 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
1446 1447
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1448
{
N
Nick Piggin 已提交
1449
	struct mm_struct *mm = vma->vm_mm;
1450
	int retval;
1451
	pte_t *pte;
1452 1453
	spinlock_t *ptl;

1454
	retval = -EINVAL;
1455
	if (PageAnon(page) || PageSlab(page) || page_has_type(page))
1456
		goto out;
1457 1458
	retval = -ENOMEM;
	flush_dcache_page(page);
1459
	pte = get_locked_pte(mm, addr, &ptl);
1460
	if (!pte)
1461
		goto out;
1462 1463 1464 1465 1466 1467
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

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

	retval = 0;
1473 1474
	pte_unmap_unlock(pte, ptl);
	return retval;
1475 1476 1477 1478 1479 1480
out_unlock:
	pte_unmap_unlock(pte, ptl);
out:
	return retval;
}

1481 1482 1483 1484 1485 1486
/**
 * 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
 *
1487 1488 1489 1490 1491 1492
 * 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
1493
 * (see split_page()).
1494 1495 1496 1497 1498 1499 1500 1501
 *
 * 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.
1502 1503 1504 1505 1506
 *
 * 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.
1507
 */
N
Nick Piggin 已提交
1508 1509
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1510 1511 1512 1513 1514
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1515 1516 1517 1518 1519
	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
1520
	return insert_page(vma, addr, page, vma->vm_page_prot);
1521
}
1522
EXPORT_SYMBOL(vm_insert_page);
1523

1524
static vm_fault_t insert_pfn(struct vm_area_struct *vma, unsigned long addr,
R
Ross Zwisler 已提交
1525
			pfn_t pfn, pgprot_t prot, bool mkwrite)
N
Nick Piggin 已提交
1526 1527 1528 1529 1530 1531 1532
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte, entry;
	spinlock_t *ptl;

	pte = get_locked_pte(mm, addr, &ptl);
	if (!pte)
1533
		return VM_FAULT_OOM;
R
Ross Zwisler 已提交
1534 1535 1536 1537 1538 1539 1540
	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
1541 1542 1543 1544
			 * 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
1545
			 */
J
Jan Kara 已提交
1546 1547
			if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
				WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
R
Ross Zwisler 已提交
1548
				goto out_unlock;
J
Jan Kara 已提交
1549
			}
R
Ross Zwisler 已提交
1550 1551 1552 1553 1554
			entry = *pte;
			goto out_mkwrite;
		} else
			goto out_unlock;
	}
N
Nick Piggin 已提交
1555 1556

	/* Ok, finally just insert the thing.. */
1557 1558 1559 1560
	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
1561 1562 1563 1564 1565 1566 1567

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

N
Nick Piggin 已提交
1568
	set_pte_at(mm, addr, pte, entry);
1569
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1570 1571 1572

out_unlock:
	pte_unmap_unlock(pte, ptl);
1573
	return VM_FAULT_NOPAGE;
N
Nick Piggin 已提交
1574 1575
}

1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
/**
 * 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
1588
 * vmf_insert_pfn_prot should only be used if using multiple VMAs is
1589 1590
 * impractical.
 *
M
Matthew Wilcox 已提交
1591
 * Context: Process context.  May allocate using %GFP_KERNEL.
1592 1593 1594 1595 1596
 * 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)
{
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
	/*
	 * 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));

1617
	return insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
1618
			false);
1619 1620
}
EXPORT_SYMBOL(vmf_insert_pfn_prot);
N
Nick Piggin 已提交
1621

M
Matthew Wilcox 已提交
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
/**
 * 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);

1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
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;
}

1663 1664
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
1665
{
1666
	pgprot_t pgprot = vma->vm_page_prot;
1667
	int err;
1668

1669
	BUG_ON(!vm_mixed_ok(vma, pfn));
N
Nick Piggin 已提交
1670

N
Nick Piggin 已提交
1671
	if (addr < vma->vm_start || addr >= vma->vm_end)
1672
		return VM_FAULT_SIGBUS;
1673 1674

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1675

1676
	if (!pfn_modify_allowed(pfn_t_to_pfn(pfn), pgprot))
1677
		return VM_FAULT_SIGBUS;
1678

N
Nick Piggin 已提交
1679 1680 1681 1682
	/*
	 * 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
1683 1684
	 * 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
1685
	 */
L
Laurent Dufour 已提交
1686 1687
	if (!IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL) &&
	    !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1688 1689
		struct page *page;

1690 1691 1692 1693 1694 1695
		/*
		 * 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));
1696 1697
		err = insert_page(vma, addr, page, pgprot);
	} else {
1698
		return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
N
Nick Piggin 已提交
1699
	}
R
Ross Zwisler 已提交
1700

M
Matthew Wilcox 已提交
1701 1702 1703 1704 1705 1706
	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
1707
}
1708 1709 1710 1711 1712 1713

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
1714
EXPORT_SYMBOL(vmf_insert_mixed);
N
Nick Piggin 已提交
1715

1716 1717 1718 1719 1720 1721 1722
/*
 *  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
1723
{
1724
	return __vm_insert_mixed(vma, addr, pfn, true);
R
Ross Zwisler 已提交
1725
}
1726
EXPORT_SYMBOL(vmf_insert_mixed_mkwrite);
R
Ross Zwisler 已提交
1727

L
Linus Torvalds 已提交
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
/*
 * 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
1738
	spinlock_t *ptl;
1739
	int err = 0;
L
Linus Torvalds 已提交
1740

H
Hugh Dickins 已提交
1741
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1742 1743
	if (!pte)
		return -ENOMEM;
1744
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1745 1746
	do {
		BUG_ON(!pte_none(*pte));
1747 1748 1749 1750
		if (!pfn_modify_allowed(pfn, prot)) {
			err = -EACCES;
			break;
		}
N
Nick Piggin 已提交
1751
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1752 1753
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1754
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1755
	pte_unmap_unlock(pte - 1, ptl);
1756
	return err;
L
Linus Torvalds 已提交
1757 1758 1759 1760 1761 1762 1763 1764
}

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;
1765
	int err;
L
Linus Torvalds 已提交
1766 1767 1768 1769 1770

	pfn -= addr >> PAGE_SHIFT;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
1771
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1772 1773
	do {
		next = pmd_addr_end(addr, end);
1774 1775 1776 1777
		err = remap_pte_range(mm, pmd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1778 1779 1780 1781
	} while (pmd++, addr = next, addr != end);
	return 0;
}

1782
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
L
Linus Torvalds 已提交
1783 1784 1785 1786 1787
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;
1788
	int err;
L
Linus Torvalds 已提交
1789 1790

	pfn -= addr >> PAGE_SHIFT;
1791
	pud = pud_alloc(mm, p4d, addr);
L
Linus Torvalds 已提交
1792 1793 1794 1795
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
1796 1797 1798 1799
		err = remap_pmd_range(mm, pud, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1800 1801 1802 1803
	} while (pud++, addr = next, addr != end);
	return 0;
}

1804 1805 1806 1807 1808 1809
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;
1810
	int err;
1811 1812 1813 1814 1815 1816 1817

	pfn -= addr >> PAGE_SHIFT;
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
1818 1819 1820 1821
		err = remap_pud_range(mm, p4d, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
1822 1823 1824 1825
	} while (p4d++, addr = next, addr != end);
	return 0;
}

1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
/**
 * remap_pfn_range - remap kernel memory to userspace
 * @vma: user vma to map to
 * @addr: target user address to start at
 * @pfn: physical address of kernel memory
 * @size: size of map area
 * @prot: page protection flags for this mapping
 *
 *  Note: this is only safe if the mm semaphore is held when called.
 */
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840
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;
1841
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1842
	struct mm_struct *mm = vma->vm_mm;
1843
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1844 1845 1846 1847 1848 1849 1850
	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).
1851 1852 1853
	 *   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.
1854 1855 1856 1857
	 *   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
1858 1859 1860 1861
	 *
	 * 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".
1862
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1863
	 */
1864 1865 1866
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1867
		vma->vm_pgoff = pfn;
1868 1869
	}

1870
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1871
	if (err)
1872
		return -EINVAL;
L
Linus Torvalds 已提交
1873

1874
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1875 1876 1877 1878 1879 1880 1881

	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);
1882
		err = remap_p4d_range(mm, pgd, addr, next,
L
Linus Torvalds 已提交
1883 1884 1885 1886
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1887 1888

	if (err)
1889
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1890

L
Linus Torvalds 已提交
1891 1892 1893 1894
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
/**
 * vm_iomap_memory - remap memory to userspace
 * @vma: user vma to map to
 * @start: start of area
 * @len: size of area
 *
 * This is a simplified io_remap_pfn_range() for common driver use. The
 * driver just needs to give us the physical memory range to be mapped,
 * we'll figure out the rest from the vma information.
 *
 * NOTE! Some drivers might want to tweak vma->vm_page_prot first to get
 * whatever write-combining details or similar.
 */
int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
{
	unsigned long vm_len, pfn, pages;

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

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

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

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

1942 1943 1944 1945 1946 1947
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;
1948
	pgtable_t token;
1949
	spinlock_t *uninitialized_var(ptl);
1950 1951 1952 1953 1954 1955 1956 1957 1958

	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));

1959 1960
	arch_enter_lazy_mmu_mode();

1961
	token = pmd_pgtable(*pmd);
1962 1963

	do {
1964
		err = fn(pte++, token, addr, data);
1965 1966
		if (err)
			break;
1967
	} while (addr += PAGE_SIZE, addr != end);
1968

1969 1970
	arch_leave_lazy_mmu_mode();

1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
	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
1984 1985
	BUG_ON(pud_huge(*pud));

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
	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;
}

1998
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
1999 2000 2001 2002 2003 2004 2005
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

2006
	pud = pud_alloc(mm, p4d, addr);
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
	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;
}

2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
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;
}

2038 2039 2040 2041 2042 2043 2044 2045 2046
/*
 * 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;
2047
	unsigned long end = addr + size;
2048 2049
	int err;

2050 2051 2052
	if (WARN_ON(addr >= end))
		return -EINVAL;

2053 2054 2055
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
2056
		err = apply_to_p4d_range(mm, pgd, addr, next, fn, data);
2057 2058 2059
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2060

2061 2062 2063 2064
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2065
/*
2066 2067 2068 2069 2070
 * 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;
2071
 * and do_anonymous_page can safely check later on).
2072
 */
H
Hugh Dickins 已提交
2073
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2074 2075 2076 2077 2078
				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
2079 2080
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2081
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2082
		spin_unlock(ptl);
2083 2084 2085 2086 2087 2088
	}
#endif
	pte_unmap(page_table);
	return same;
}

2089
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2090
{
2091 2092
	debug_dma_assert_idle(src);

2093 2094 2095 2096 2097 2098 2099
	/*
	 * If the source page was a PFN mapping, we don't have
	 * a "struct page" for it. We do a best-effort copy by
	 * just copying from the original user address. If that
	 * fails, we just zero-fill it. Live with it.
	 */
	if (unlikely(!src)) {
2100
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2101 2102 2103 2104 2105 2106 2107 2108 2109
		void __user *uaddr = (void __user *)(va & PAGE_MASK);

		/*
		 * This really shouldn't fail, because the page is there
		 * in the page tables. But it might just be unreadable,
		 * in which case we just give up and fill the result with
		 * zeroes.
		 */
		if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE))
2110
			clear_page(kaddr);
2111
		kunmap_atomic(kaddr);
2112
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2113 2114
	} else
		copy_user_highpage(dst, src, va, vma);
2115 2116
}

2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130
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;
}

2131 2132 2133 2134 2135 2136
/*
 * 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.
 */
2137
static vm_fault_t do_page_mkwrite(struct vm_fault *vmf)
2138
{
2139
	vm_fault_t ret;
2140 2141
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2142

2143
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2144

2145
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2146 2147
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161
	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;
}

2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196
/*
 * Handle dirtying of a page in shared file mapping on a write fault.
 *
 * The function expects the page to be locked and unlocks it.
 */
static void fault_dirty_shared_page(struct vm_area_struct *vma,
				    struct page *page)
{
	struct address_space *mapping;
	bool dirtied;
	bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;

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

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

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

2197 2198 2199 2200 2201 2202 2203 2204
/*
 * 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.
 */
2205
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2206
	__releases(vmf->ptl)
2207
{
J
Jan Kara 已提交
2208
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2209
	struct page *page = vmf->page;
2210 2211 2212 2213 2214 2215 2216 2217 2218
	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
2219 2220
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2221
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2222 2223 2224
	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);
2225 2226
}

2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242
/*
 * 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.
 */
2243
static vm_fault_t wp_page_copy(struct vm_fault *vmf)
2244
{
J
Jan Kara 已提交
2245
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2246
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2247
	struct page *old_page = vmf->page;
2248 2249 2250 2251
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
	struct mem_cgroup *memcg;
2252
	struct mmu_notifier_range range;
2253 2254 2255 2256

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

J
Jan Kara 已提交
2257
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2258 2259
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2260 2261 2262
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2263
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2264
				vmf->address);
2265 2266
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2267
		cow_user_page(new_page, old_page, vmf->address, vma);
2268 2269
	}

2270
	if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
2271 2272
		goto oom_free_new;

2273 2274
	__SetPageUptodate(new_page);

2275 2276 2277
	mmu_notifier_range_init(&range, mm, vmf->address & PAGE_MASK,
				(vmf->address & PAGE_MASK) + PAGE_SIZE);
	mmu_notifier_invalidate_range_start(&range);
2278 2279 2280 2281

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2282
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2283
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2284 2285
		if (old_page) {
			if (!PageAnon(old_page)) {
2286 2287
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2288 2289 2290 2291 2292
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2293
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2294 2295 2296 2297 2298 2299 2300 2301
		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
2302 2303
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2304
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2305 2306 2307 2308 2309 2310
		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
2311 2312
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
		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.
			 */
2336
			page_remove_rmap(old_page, false);
2337 2338 2339 2340 2341 2342
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2343
		mem_cgroup_cancel_charge(new_page, memcg, false);
2344 2345 2346
	}

	if (new_page)
2347
		put_page(new_page);
2348

J
Jan Kara 已提交
2349
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2350 2351 2352 2353
	/*
	 * No need to double call mmu_notifier->invalidate_range() callback as
	 * the above ptep_clear_flush_notify() did already call it.
	 */
2354
	mmu_notifier_invalidate_range_only_end(&range);
2355 2356 2357 2358 2359 2360 2361
	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 */
2362 2363
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2364 2365
			unlock_page(old_page);
		}
2366
		put_page(old_page);
2367 2368 2369
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2370
	put_page(new_page);
2371 2372
oom:
	if (old_page)
2373
		put_page(old_page);
2374 2375 2376
	return VM_FAULT_OOM;
}

2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
/**
 * finish_mkwrite_fault - finish page fault for a shared mapping, making PTE
 *			  writeable once the page is prepared
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a write page fault in a
 * shared mapping due to PTE being read-only once the mapped page is prepared.
 * It handles locking of PTE and modifying it. The function returns
 * VM_FAULT_WRITE on success, 0 when PTE got changed before we acquired PTE
 * lock.
 *
 * The function expects the page to be locked or other protection against
 * concurrent faults / writeback (such as DAX radix tree locks).
 */
2392
vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf)
2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
{
	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);
2403
		return VM_FAULT_NOPAGE;
2404 2405
	}
	wp_page_reuse(vmf);
2406
	return 0;
2407 2408
}

2409 2410 2411 2412
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
2413
static vm_fault_t wp_pfn_shared(struct vm_fault *vmf)
2414
{
J
Jan Kara 已提交
2415
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2416

2417
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
2418
		vm_fault_t ret;
2419

J
Jan Kara 已提交
2420
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2421
		vmf->flags |= FAULT_FLAG_MKWRITE;
2422
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2423
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2424
			return ret;
2425
		return finish_mkwrite_fault(vmf);
2426
	}
2427 2428
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2429 2430
}

2431
static vm_fault_t wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2432
	__releases(vmf->ptl)
2433
{
J
Jan Kara 已提交
2434
	struct vm_area_struct *vma = vmf->vma;
2435

J
Jan Kara 已提交
2436
	get_page(vmf->page);
2437 2438

	if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
2439
		vm_fault_t tmp;
2440

J
Jan Kara 已提交
2441
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2442
		tmp = do_page_mkwrite(vmf);
2443 2444
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2445
			put_page(vmf->page);
2446 2447
			return tmp;
		}
2448
		tmp = finish_mkwrite_fault(vmf);
2449
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2450 2451
			unlock_page(vmf->page);
			put_page(vmf->page);
2452
			return tmp;
2453
		}
2454 2455
	} else {
		wp_page_reuse(vmf);
2456
		lock_page(vmf->page);
2457
	}
2458 2459
	fault_dirty_shared_page(vma, vmf->page);
	put_page(vmf->page);
2460

2461
	return VM_FAULT_WRITE;
2462 2463
}

L
Linus Torvalds 已提交
2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477
/*
 * 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.
 *
2478 2479 2480
 * 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
2481
 */
2482
static vm_fault_t do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2483
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2484
{
J
Jan Kara 已提交
2485
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2486

J
Jan Kara 已提交
2487 2488
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2489
		/*
2490 2491
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2492 2493
		 *
		 * We should not cow pages in a shared writeable mapping.
2494
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2495 2496 2497
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2498
			return wp_pfn_shared(vmf);
2499

J
Jan Kara 已提交
2500
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2501
		return wp_page_copy(vmf);
2502
	}
L
Linus Torvalds 已提交
2503

2504
	/*
P
Peter Zijlstra 已提交
2505 2506
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2507
	 */
2508
	if (PageAnon(vmf->page)) {
2509
		int total_map_swapcount;
2510 2511 2512
		if (PageKsm(vmf->page) && (PageSwapCache(vmf->page) ||
					   page_count(vmf->page) != 1))
			goto copy;
J
Jan Kara 已提交
2513 2514
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2515
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2516
			lock_page(vmf->page);
J
Jan Kara 已提交
2517 2518
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2519
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2520
				unlock_page(vmf->page);
J
Jan Kara 已提交
2521
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2522
				put_page(vmf->page);
2523
				return 0;
2524
			}
J
Jan Kara 已提交
2525
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2526
		}
2527 2528 2529 2530 2531 2532 2533 2534 2535
		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;
		}
2536 2537
		if (reuse_swap_page(vmf->page, &total_map_swapcount)) {
			if (total_map_swapcount == 1) {
2538 2539 2540 2541 2542 2543 2544
				/*
				 * 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
2545
				page_move_anon_rmap(vmf->page, vma);
2546
			}
J
Jan Kara 已提交
2547
			unlock_page(vmf->page);
2548 2549
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2550
		}
J
Jan Kara 已提交
2551
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2552
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2553
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2554
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2555
	}
2556
copy:
L
Linus Torvalds 已提交
2557 2558 2559
	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2560
	get_page(vmf->page);
2561

J
Jan Kara 已提交
2562
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2563
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2564 2565
}

2566
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2567 2568 2569
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2570
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2571 2572
}

2573
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
L
Linus Torvalds 已提交
2574 2575 2576 2577 2578
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2579
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2580 2581 2582
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2583
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2584 2585 2586 2587 2588 2589 2590
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2591
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2592 2593
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2594
				details);
L
Linus Torvalds 已提交
2595 2596 2597
	}
}

M
Matthew Wilcox 已提交
2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
/**
 * 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
2627
/**
2628
 * unmap_mapping_range - unmap the portion of all mmaps in the specified
M
Matthew Wilcox 已提交
2629
 * address_space corresponding to the specified byte range in the underlying
2630 2631
 * file.
 *
M
Martin Waitz 已提交
2632
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2633 2634
 * @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
2635
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657
 * 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
2658
	unmap_mapping_pages(mapping, hba, hlen, even_cows);
L
Linus Torvalds 已提交
2659 2660 2661 2662
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2663 2664
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2665 2666 2667 2668
 * 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
2669
 */
2670
vm_fault_t do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2671
{
J
Jan Kara 已提交
2672
	struct vm_area_struct *vma = vmf->vma;
M
Minchan Kim 已提交
2673
	struct page *page = NULL, *swapcache;
2674
	struct mem_cgroup *memcg;
2675
	swp_entry_t entry;
L
Linus Torvalds 已提交
2676
	pte_t pte;
2677
	int locked;
2678
	int exclusive = 0;
2679
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
2680

M
Minchan Kim 已提交
2681
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte))
2682
		goto out;
2683

J
Jan Kara 已提交
2684
	entry = pte_to_swp_entry(vmf->orig_pte);
2685 2686
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2687 2688
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2689 2690 2691 2692 2693 2694 2695 2696
		} else if (is_device_private_entry(entry)) {
			/*
			 * For un-addressable device memory we call the pgmap
			 * fault handler callback. The callback must migrate
			 * the page back to some CPU accessible page.
			 */
			ret = device_private_entry_fault(vma, vmf->address, entry,
						 vmf->flags, vmf->pmd);
2697 2698 2699
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2700
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2701
			ret = VM_FAULT_SIGBUS;
2702
		}
2703 2704
		goto out;
	}
2705 2706


2707
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
M
Minchan Kim 已提交
2708 2709
	page = lookup_swap_cache(entry, vma, vmf->address);
	swapcache = page;
2710

L
Linus Torvalds 已提交
2711
	if (!page) {
2712 2713
		struct swap_info_struct *si = swp_swap_info(entry);

2714 2715
		if (si->flags & SWP_SYNCHRONOUS_IO &&
				__swap_count(si, entry) == 1) {
2716
			/* skip swapcache */
2717 2718
			page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
							vmf->address);
2719 2720 2721 2722 2723 2724 2725
			if (page) {
				__SetPageLocked(page);
				__SetPageSwapBacked(page);
				set_page_private(page, entry.val);
				lru_cache_add_anon(page);
				swap_readpage(page, true);
			}
2726
		} else {
2727 2728
			page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE,
						vmf);
2729
			swapcache = page;
2730 2731
		}

L
Linus Torvalds 已提交
2732 2733
		if (!page) {
			/*
2734 2735
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2736
			 */
J
Jan Kara 已提交
2737 2738
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2739
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2740
				ret = VM_FAULT_OOM;
2741
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2742
			goto unlock;
L
Linus Torvalds 已提交
2743 2744 2745 2746
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2747
		count_vm_event(PGMAJFAULT);
2748
		count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
2749
	} else if (PageHWPoison(page)) {
2750 2751 2752 2753
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2754 2755
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2756
		goto out_release;
L
Linus Torvalds 已提交
2757 2758
	}

J
Jan Kara 已提交
2759
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2760

2761
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2762 2763 2764 2765
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2766

A
Andrea Arcangeli 已提交
2767
	/*
2768 2769 2770 2771
	 * 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
2772
	 */
2773 2774
	if (unlikely((!PageSwapCache(page) ||
			page_private(page) != entry.val)) && swapcache)
A
Andrea Arcangeli 已提交
2775 2776
		goto out_page;

J
Jan Kara 已提交
2777
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2778 2779 2780 2781
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2782 2783
	}

2784 2785
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL,
					&memcg, false)) {
2786
		ret = VM_FAULT_OOM;
2787
		goto out_page;
2788 2789
	}

L
Linus Torvalds 已提交
2790
	/*
2791
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2792
	 */
J
Jan Kara 已提交
2793 2794
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2795
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2796 2797 2798 2799 2800
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2801 2802
	}

2803 2804 2805 2806 2807 2808 2809 2810 2811
	/*
	 * 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
2812

K
Kirill A. Shutemov 已提交
2813 2814
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2815
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2816
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2817
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2818
		vmf->flags &= ~FAULT_FLAG_WRITE;
2819
		ret |= VM_FAULT_WRITE;
2820
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2821 2822
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
2823
	if (pte_swp_soft_dirty(vmf->orig_pte))
2824
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2825
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
2826
	arch_do_swap_page(vma->vm_mm, vma, vmf->address, pte, vmf->orig_pte);
J
Jan Kara 已提交
2827
	vmf->orig_pte = pte;
2828 2829 2830

	/* ksm created a completely new copy */
	if (unlikely(page != swapcache && swapcache)) {
J
Jan Kara 已提交
2831
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2832
		mem_cgroup_commit_charge(page, memcg, false, false);
2833
		lru_cache_add_active_or_unevictable(page, vma);
2834 2835 2836 2837
	} else {
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
		mem_cgroup_commit_charge(page, memcg, true, false);
		activate_page(page);
2838
	}
L
Linus Torvalds 已提交
2839

2840
	swap_free(entry);
2841 2842
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2843
		try_to_free_swap(page);
2844
	unlock_page(page);
2845
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
2846 2847 2848 2849 2850 2851 2852 2853 2854
		/*
		 * 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);
2855
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2856
	}
2857

J
Jan Kara 已提交
2858
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
2859
		ret |= do_wp_page(vmf);
2860 2861
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2862 2863 2864 2865
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2866
	update_mmu_cache(vma, vmf->address, vmf->pte);
2867
unlock:
J
Jan Kara 已提交
2868
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
2869 2870
out:
	return ret;
2871
out_nomap:
2872
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
2873
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2874
out_page:
2875
	unlock_page(page);
2876
out_release:
2877
	put_page(page);
2878
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
2879
		unlock_page(swapcache);
2880
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2881
	}
2882
	return ret;
L
Linus Torvalds 已提交
2883 2884 2885
}

/*
2886 2887 2888
 * 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
2889
 */
2890
static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2891
{
J
Jan Kara 已提交
2892
	struct vm_area_struct *vma = vmf->vma;
2893
	struct mem_cgroup *memcg;
2894
	struct page *page;
2895
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
2896 2897
	pte_t entry;

2898 2899 2900 2901
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
	/*
	 * 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).
	 */
2912
	if (pte_alloc(vma->vm_mm, vmf->pmd))
2913 2914 2915
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
2916
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
2917 2918
		return 0;

2919
	/* Use the zero-page for reads */
J
Jan Kara 已提交
2920
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
2921
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
2922
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
2923
						vma->vm_page_prot));
J
Jan Kara 已提交
2924 2925 2926
		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
2927
			goto unlock;
2928 2929 2930
		ret = check_stable_address_space(vma->vm_mm);
		if (ret)
			goto unlock;
2931 2932
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2933 2934
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
2935
		}
H
Hugh Dickins 已提交
2936 2937 2938
		goto setpte;
	}

N
Nick Piggin 已提交
2939 2940 2941
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
2942
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
2943 2944
	if (!page)
		goto oom;
2945

2946 2947
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL, &memcg,
					false))
2948 2949
		goto oom_free_page;

2950 2951 2952 2953 2954
	/*
	 * 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
2955
	__SetPageUptodate(page);
2956

N
Nick Piggin 已提交
2957
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
2958 2959
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
2960

J
Jan Kara 已提交
2961 2962 2963
	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
2964
		goto release;
H
Hugh Dickins 已提交
2965

2966 2967 2968 2969
	ret = check_stable_address_space(vma->vm_mm);
	if (ret)
		goto release;

2970 2971
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
2972
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2973
		mem_cgroup_cancel_charge(page, memcg, false);
2974
		put_page(page);
J
Jan Kara 已提交
2975
		return handle_userfault(vmf, VM_UFFD_MISSING);
2976 2977
	}

K
Kirill A. Shutemov 已提交
2978
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
2979
	page_add_new_anon_rmap(page, vma, vmf->address, false);
2980
	mem_cgroup_commit_charge(page, memcg, false, false);
2981
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
2982
setpte:
J
Jan Kara 已提交
2983
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
2984 2985

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2986
	update_mmu_cache(vma, vmf->address, vmf->pte);
2987
unlock:
J
Jan Kara 已提交
2988
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2989
	return ret;
2990
release:
2991
	mem_cgroup_cancel_charge(page, memcg, false);
2992
	put_page(page);
2993
	goto unlock;
2994
oom_free_page:
2995
	put_page(page);
2996
oom:
L
Linus Torvalds 已提交
2997 2998 2999
	return VM_FAULT_OOM;
}

3000 3001 3002 3003 3004
/*
 * 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().
 */
3005
static vm_fault_t __do_fault(struct vm_fault *vmf)
3006
{
J
Jan Kara 已提交
3007
	struct vm_area_struct *vma = vmf->vma;
3008
	vm_fault_t ret;
3009

3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
	/*
	 * 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() */
	}

3032
	ret = vma->vm_ops->fault(vmf);
3033
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
3034
			    VM_FAULT_DONE_COW)))
3035
		return ret;
3036

3037
	if (unlikely(PageHWPoison(vmf->page))) {
3038
		if (ret & VM_FAULT_LOCKED)
3039 3040
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
3041
		vmf->page = NULL;
3042 3043 3044 3045
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
3046
		lock_page(vmf->page);
3047
	else
3048
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
3049 3050 3051 3052

	return ret;
}

3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063
/*
 * 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);
}

3064
static vm_fault_t pte_alloc_one_map(struct vm_fault *vmf)
3065
{
J
Jan Kara 已提交
3066
	struct vm_area_struct *vma = vmf->vma;
3067

J
Jan Kara 已提交
3068
	if (!pmd_none(*vmf->pmd))
3069
		goto map_pte;
J
Jan Kara 已提交
3070 3071 3072 3073
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
3074 3075 3076
			goto map_pte;
		}

3077
		mm_inc_nr_ptes(vma->vm_mm);
J
Jan Kara 已提交
3078 3079
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
3080
		vmf->prealloc_pte = NULL;
3081
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd))) {
3082 3083 3084 3085 3086
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
3087 3088 3089 3090 3091 3092 3093 3094
	 * 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.
3095
	 */
3096
	if (pmd_devmap_trans_unstable(vmf->pmd))
3097 3098
		return VM_FAULT_NOPAGE;

3099 3100 3101 3102 3103 3104 3105 3106 3107
	/*
	 * 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
3108 3109
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
3110 3111 3112
	return 0;
}

3113
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126

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

J
Jan Kara 已提交
3127
static void deposit_prealloc_pte(struct vm_fault *vmf)
3128
{
J
Jan Kara 已提交
3129
	struct vm_area_struct *vma = vmf->vma;
3130

J
Jan Kara 已提交
3131
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
3132 3133 3134 3135
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
3136
	mm_inc_nr_ptes(vma->vm_mm);
3137
	vmf->prealloc_pte = NULL;
3138 3139
}

3140
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3141
{
J
Jan Kara 已提交
3142 3143 3144
	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
3145
	pmd_t entry;
3146 3147
	int i;
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3148 3149 3150 3151 3152 3153 3154

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

3155 3156 3157 3158
	/*
	 * 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
3159
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
3160
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
J
Jan Kara 已提交
3161
		if (!vmf->prealloc_pte)
3162 3163 3164 3165
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
3166 3167
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3168 3169 3170 3171 3172 3173 3174
		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)
3175
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3176

3177
	add_mm_counter(vma->vm_mm, mm_counter_file(page), HPAGE_PMD_NR);
K
Kirill A. Shutemov 已提交
3178
	page_add_file_rmap(page, true);
3179 3180 3181 3182
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3183
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3184

J
Jan Kara 已提交
3185
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3186

J
Jan Kara 已提交
3187
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3188 3189 3190

	/* fault is handled */
	ret = 0;
3191
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3192
out:
J
Jan Kara 已提交
3193
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3194 3195 3196
	return ret;
}
#else
3197
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3198 3199 3200 3201 3202 3203
{
	BUILD_BUG();
	return 0;
}
#endif

3204
/**
3205 3206
 * 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.
3207
 *
J
Jan Kara 已提交
3208
 * @vmf: fault environment
3209
 * @memcg: memcg to charge page (only for private mappings)
3210 3211
 * @page: page to map
 *
J
Jan Kara 已提交
3212 3213
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3214 3215 3216 3217
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
 */
3218
vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3219
		struct page *page)
3220
{
J
Jan Kara 已提交
3221 3222
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3223
	pte_t entry;
3224
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3225

J
Jan Kara 已提交
3226
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3227
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3228 3229 3230
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3231
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3232
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3233
			return ret;
K
Kirill A. Shutemov 已提交
3234
	}
3235

J
Jan Kara 已提交
3236 3237
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3238
		if (ret)
H
Hugh Dickins 已提交
3239
			return ret;
3240 3241 3242
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3243 3244
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3245

3246 3247 3248 3249
	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
3250 3251
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3252
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3253
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3254 3255
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3256
	} else {
3257
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3258
		page_add_file_rmap(page, false);
3259
	}
J
Jan Kara 已提交
3260
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3261 3262

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3263
	update_mmu_cache(vma, vmf->address, vmf->pte);
3264

H
Hugh Dickins 已提交
3265
	return 0;
3266 3267
}

3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282

/**
 * finish_fault - finish page fault once we have prepared the page to fault
 *
 * @vmf: structure describing the fault
 *
 * This function handles all that is needed to finish a page fault once the
 * page to fault in is prepared. It handles locking of PTEs, inserts PTE for
 * given page, adds reverse page mapping, handles memcg charges and LRU
 * addition. The function returns 0 on success, VM_FAULT_ code in case of
 * error.
 *
 * The function expects the page to be locked and on success it consumes a
 * reference of a page being mapped (for the PTE which maps it).
 */
3283
vm_fault_t finish_fault(struct vm_fault *vmf)
3284 3285
{
	struct page *page;
3286
	vm_fault_t ret = 0;
3287 3288 3289 3290 3291 3292 3293

	/* 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;
3294 3295 3296 3297 3298 3299 3300 3301 3302

	/*
	 * 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);
3303 3304 3305 3306 3307
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3308 3309
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3310 3311 3312

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3313
{
3314
	*val = fault_around_bytes;
3315 3316 3317
	return 0;
}

3318
/*
3319 3320
 * fault_around_bytes must be rounded down to the nearest page order as it's
 * what do_fault_around() expects to see.
3321
 */
3322
static int fault_around_bytes_set(void *data, u64 val)
3323
{
3324
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3325
		return -EINVAL;
3326 3327 3328 3329
	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 */
3330 3331
	return 0;
}
3332
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
3333
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3334 3335 3336

static int __init fault_around_debugfs(void)
{
3337 3338
	debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
				   &fault_around_bytes_fops);
3339 3340 3341 3342
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3343

3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358
/*
 * 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.
 *
3359 3360 3361
 * 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.
3362
 *
3363 3364 3365 3366
 * 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.
3367
 */
3368
static vm_fault_t do_fault_around(struct vm_fault *vmf)
3369
{
J
Jan Kara 已提交
3370
	unsigned long address = vmf->address, nr_pages, mask;
3371
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3372
	pgoff_t end_pgoff;
3373 3374
	int off;
	vm_fault_t ret = 0;
3375

3376
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3377 3378
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3379 3380
	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
3381
	start_pgoff -= off;
3382 3383

	/*
3384 3385
	 *  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.
3386
	 */
K
Kirill A. Shutemov 已提交
3387
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3388
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3389
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3390
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3391
			start_pgoff + nr_pages - 1);
3392

J
Jan Kara 已提交
3393
	if (pmd_none(*vmf->pmd)) {
3394
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
J
Jan Kara 已提交
3395
		if (!vmf->prealloc_pte)
3396
			goto out;
3397
		smp_wmb(); /* See comment in __pte_alloc() */
3398 3399
	}

J
Jan Kara 已提交
3400
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3401 3402

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3403
	if (pmd_trans_huge(*vmf->pmd)) {
3404 3405 3406 3407 3408
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3409
	if (!vmf->pte)
3410 3411 3412
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3413 3414
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3415
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3416
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3417
out:
J
Jan Kara 已提交
3418 3419
	vmf->address = address;
	vmf->pte = NULL;
3420
	return ret;
3421 3422
}

3423
static vm_fault_t do_read_fault(struct vm_fault *vmf)
3424
{
J
Jan Kara 已提交
3425
	struct vm_area_struct *vma = vmf->vma;
3426
	vm_fault_t ret = 0;
3427 3428 3429 3430 3431 3432

	/*
	 * 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).
	 */
3433
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3434
		ret = do_fault_around(vmf);
3435 3436
		if (ret)
			return ret;
3437
	}
3438

J
Jan Kara 已提交
3439
	ret = __do_fault(vmf);
3440 3441 3442
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3443
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3444
	unlock_page(vmf->page);
3445
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3446
		put_page(vmf->page);
3447 3448 3449
	return ret;
}

3450
static vm_fault_t do_cow_fault(struct vm_fault *vmf)
3451
{
J
Jan Kara 已提交
3452
	struct vm_area_struct *vma = vmf->vma;
3453
	vm_fault_t ret;
3454 3455 3456 3457

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

J
Jan Kara 已提交
3458 3459
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3460 3461
		return VM_FAULT_OOM;

3462
	if (mem_cgroup_try_charge_delay(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3463
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3464
		put_page(vmf->cow_page);
3465 3466 3467
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3468
	ret = __do_fault(vmf);
3469 3470
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3471 3472
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3473

3474
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3475
	__SetPageUptodate(vmf->cow_page);
3476

3477
	ret |= finish_fault(vmf);
3478 3479
	unlock_page(vmf->page);
	put_page(vmf->page);
3480 3481
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3482 3483
	return ret;
uncharge_out:
3484
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3485
	put_page(vmf->cow_page);
3486 3487 3488
	return ret;
}

3489
static vm_fault_t do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3490
{
J
Jan Kara 已提交
3491
	struct vm_area_struct *vma = vmf->vma;
3492
	vm_fault_t ret, tmp;
3493

J
Jan Kara 已提交
3494
	ret = __do_fault(vmf);
3495
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3496
		return ret;
L
Linus Torvalds 已提交
3497 3498

	/*
3499 3500
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3501
	 */
3502
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3503
		unlock_page(vmf->page);
3504
		tmp = do_page_mkwrite(vmf);
3505 3506
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3507
			put_page(vmf->page);
3508
			return tmp;
3509
		}
3510 3511
	}

3512
	ret |= finish_fault(vmf);
3513 3514
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3515 3516
		unlock_page(vmf->page);
		put_page(vmf->page);
3517
		return ret;
L
Linus Torvalds 已提交
3518
	}
N
Nick Piggin 已提交
3519

3520
	fault_dirty_shared_page(vma, vmf->page);
3521
	return ret;
3522
}
3523

3524 3525 3526 3527 3528 3529
/*
 * 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().
 */
3530
static vm_fault_t do_fault(struct vm_fault *vmf)
3531
{
J
Jan Kara 已提交
3532
	struct vm_area_struct *vma = vmf->vma;
3533
	vm_fault_t ret;
3534

3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564
	/*
	 * 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
3565 3566 3567 3568 3569 3570 3571 3572 3573
		ret = do_read_fault(vmf);
	else if (!(vma->vm_flags & VM_SHARED))
		ret = do_cow_fault(vmf);
	else
		ret = do_shared_fault(vmf);

	/* preallocated pagetable is unused: free it */
	if (vmf->prealloc_pte) {
		pte_free(vma->vm_mm, vmf->prealloc_pte);
3574
		vmf->prealloc_pte = NULL;
H
Hugh Dickins 已提交
3575 3576
	}
	return ret;
3577 3578
}

3579
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3580 3581
				unsigned long addr, int page_nid,
				int *flags)
3582 3583 3584 3585
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3586
	if (page_nid == numa_node_id()) {
3587
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3588 3589
		*flags |= TNF_FAULT_LOCAL;
	}
3590 3591 3592 3593

	return mpol_misplaced(page, vma, addr);
}

3594
static vm_fault_t do_numa_page(struct vm_fault *vmf)
3595
{
J
Jan Kara 已提交
3596
	struct vm_area_struct *vma = vmf->vma;
3597
	struct page *page = NULL;
3598
	int page_nid = NUMA_NO_NODE;
3599
	int last_cpupid;
3600
	int target_nid;
3601
	bool migrated = false;
3602
	pte_t pte;
3603
	bool was_writable = pte_savedwrite(vmf->orig_pte);
3604
	int flags = 0;
3605 3606

	/*
T
Tobin C Harding 已提交
3607 3608 3609 3610
	 * 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
3611 3612
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
3613
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
J
Jan Kara 已提交
3614
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3615 3616 3617
		goto out;
	}

3618 3619 3620 3621
	/*
	 * Make it present again, Depending on how arch implementes non
	 * accessible ptes, some can allow access by kernel mode.
	 */
3622
	pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
3623 3624
	pte = pte_modify(pte, vma->vm_page_prot);
	pte = pte_mkyoung(pte);
3625 3626
	if (was_writable)
		pte = pte_mkwrite(pte);
3627
	ptep_modify_prot_commit(vma, vmf->address, vmf->pte, pte);
J
Jan Kara 已提交
3628
	update_mmu_cache(vma, vmf->address, vmf->pte);
3629

J
Jan Kara 已提交
3630
	page = vm_normal_page(vma, vmf->address, pte);
3631
	if (!page) {
J
Jan Kara 已提交
3632
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3633 3634 3635
		return 0;
	}

3636 3637
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3638
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3639 3640 3641
		return 0;
	}

3642
	/*
3643 3644 3645 3646 3647 3648
	 * 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.
3649
	 */
3650
	if (!pte_write(pte))
3651 3652
		flags |= TNF_NO_GROUP;

3653 3654 3655 3656 3657 3658 3659
	/*
	 * 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;

3660
	last_cpupid = page_cpupid_last(page);
3661
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3662
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3663
			&flags);
J
Jan Kara 已提交
3664
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3665
	if (target_nid == NUMA_NO_NODE) {
3666 3667 3668 3669 3670
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3671
	migrated = migrate_misplaced_page(page, vma, target_nid);
3672
	if (migrated) {
3673
		page_nid = target_nid;
3674
		flags |= TNF_MIGRATED;
3675 3676
	} else
		flags |= TNF_MIGRATE_FAIL;
3677 3678

out:
3679
	if (page_nid != NUMA_NO_NODE)
3680
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3681 3682 3683
	return 0;
}

3684
static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3685
{
3686
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3687
		return do_huge_pmd_anonymous_page(vmf);
3688
	if (vmf->vma->vm_ops->huge_fault)
3689
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
M
Matthew Wilcox 已提交
3690 3691 3692
	return VM_FAULT_FALLBACK;
}

3693
/* `inline' is required to avoid gcc 4.1.2 build error */
3694
static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3695
{
J
Jan Kara 已提交
3696 3697
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3698
	if (vmf->vma->vm_ops->huge_fault)
3699
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
K
Kirill A. Shutemov 已提交
3700 3701

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3702 3703
	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
3704

M
Matthew Wilcox 已提交
3705 3706 3707
	return VM_FAULT_FALLBACK;
}

3708 3709 3710 3711 3712
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3713
static vm_fault_t create_huge_pud(struct vm_fault *vmf)
3714 3715 3716 3717 3718 3719
{
#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)
3720
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3721 3722 3723 3724
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

3725
static vm_fault_t wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
3726 3727 3728 3729 3730 3731
{
#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)
3732
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3733 3734 3735 3736
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3737 3738 3739 3740 3741 3742 3743 3744 3745
/*
 * 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).
 *
3746 3747
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3748
 *
3749 3750
 * 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
3751
 */
3752
static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3753 3754 3755
{
	pte_t entry;

J
Jan Kara 已提交
3756
	if (unlikely(pmd_none(*vmf->pmd))) {
3757 3758 3759 3760 3761 3762
		/*
		 * 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
3763
		vmf->pte = NULL;
3764 3765
	} else {
		/* See comment in pte_alloc_one_map() */
3766
		if (pmd_devmap_trans_unstable(vmf->pmd))
3767 3768 3769 3770 3771 3772 3773
			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
3774
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3775
		vmf->orig_pte = *vmf->pte;
3776 3777 3778 3779

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
3780 3781 3782
		 * 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
3783 3784 3785
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
J
Jan Kara 已提交
3786
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3787 3788
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3789
		}
L
Linus Torvalds 已提交
3790 3791
	}

J
Jan Kara 已提交
3792 3793 3794
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3795
		else
J
Jan Kara 已提交
3796
			return do_fault(vmf);
3797 3798
	}

J
Jan Kara 已提交
3799 3800
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3801

J
Jan Kara 已提交
3802 3803
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3804

J
Jan Kara 已提交
3805 3806
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3807
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3808
	if (unlikely(!pte_same(*vmf->pte, entry)))
3809
		goto unlock;
J
Jan Kara 已提交
3810
	if (vmf->flags & FAULT_FLAG_WRITE) {
3811
		if (!pte_write(entry))
J
Jan Kara 已提交
3812
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3813 3814 3815
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3816 3817 3818
	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);
3819 3820 3821 3822 3823 3824 3825
	} 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
3826 3827
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3828
	}
3829
unlock:
J
Jan Kara 已提交
3830
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3831
	return 0;
L
Linus Torvalds 已提交
3832 3833 3834 3835
}

/*
 * By the time we get here, we already hold the mm semaphore
3836 3837 3838
 *
 * 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
3839
 */
3840 3841
static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags)
L
Linus Torvalds 已提交
3842
{
J
Jan Kara 已提交
3843
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3844
		.vma = vma,
3845
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3846
		.flags = flags,
3847
		.pgoff = linear_page_index(vma, address),
3848
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3849
	};
3850
	unsigned int dirty = flags & FAULT_FLAG_WRITE;
3851
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3852
	pgd_t *pgd;
3853
	p4d_t *p4d;
3854
	vm_fault_t ret;
L
Linus Torvalds 已提交
3855 3856

	pgd = pgd_offset(mm, address);
3857 3858 3859
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return VM_FAULT_OOM;
3860

3861
	vmf.pud = pud_alloc(mm, p4d, address);
3862
	if (!vmf.pud)
H
Hugh Dickins 已提交
3863
		return VM_FAULT_OOM;
3864
	if (pud_none(*vmf.pud) && __transparent_hugepage_enabled(vma)) {
3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875
		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 */

3876
			if (dirty && !pud_write(orig_pud)) {
3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
				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
3888
	if (!vmf.pmd)
H
Hugh Dickins 已提交
3889
		return VM_FAULT_OOM;
3890
	if (pmd_none(*vmf.pmd) && __transparent_hugepage_enabled(vma)) {
3891
		ret = create_huge_pmd(&vmf);
3892 3893
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3894
	} else {
J
Jan Kara 已提交
3895
		pmd_t orig_pmd = *vmf.pmd;
3896

3897
		barrier();
3898 3899 3900 3901 3902 3903 3904
		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;
		}
3905
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
3906
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
3907
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
3908

3909
			if (dirty && !pmd_write(orig_pmd)) {
J
Jan Kara 已提交
3910
				ret = wp_huge_pmd(&vmf, orig_pmd);
3911 3912
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
3913
			} else {
J
Jan Kara 已提交
3914
				huge_pmd_set_accessed(&vmf, orig_pmd);
3915
				return 0;
3916
			}
3917 3918 3919
		}
	}

J
Jan Kara 已提交
3920
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
3921 3922
}

3923 3924 3925 3926 3927 3928
/*
 * 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().
 */
3929
vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
3930
		unsigned int flags)
3931
{
3932
	vm_fault_t ret;
3933 3934 3935 3936

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
3937
	count_memcg_event_mm(vma->vm_mm, PGFAULT);
3938 3939 3940 3941

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

3942 3943 3944 3945 3946
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

3947 3948 3949 3950 3951
	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
3952
		mem_cgroup_enter_user_fault();
3953

K
Kirill A. Shutemov 已提交
3954 3955 3956 3957
	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
3958

3959
	if (flags & FAULT_FLAG_USER) {
3960
		mem_cgroup_exit_user_fault();
T
Tobin C Harding 已提交
3961 3962 3963 3964 3965 3966 3967 3968
		/*
		 * 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);
3969
	}
3970

3971 3972
	return ret;
}
3973
EXPORT_SYMBOL_GPL(handle_mm_fault);
3974

K
Kirill A. Shutemov 已提交
3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997
#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
3998 3999 4000
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
4001
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4002
 */
4003
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
L
Linus Torvalds 已提交
4004
{
H
Hugh Dickins 已提交
4005 4006
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
4007
		return -ENOMEM;
L
Linus Torvalds 已提交
4008

4009 4010
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
4011
	spin_lock(&mm->page_table_lock);
4012
#ifndef __ARCH_HAS_5LEVEL_HACK
K
Kirill A. Shutemov 已提交
4013 4014
	if (!p4d_present(*p4d)) {
		mm_inc_nr_puds(mm);
4015
		p4d_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4016
	} else	/* Another has populated it */
4017
		pud_free(mm, new);
K
Kirill A. Shutemov 已提交
4018 4019 4020
#else
	if (!pgd_present(*p4d)) {
		mm_inc_nr_puds(mm);
4021
		pgd_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4022 4023
	} else	/* Another has populated it */
		pud_free(mm, new);
4024
#endif /* __ARCH_HAS_5LEVEL_HACK */
H
Hugh Dickins 已提交
4025
	spin_unlock(&mm->page_table_lock);
4026
	return 0;
L
Linus Torvalds 已提交
4027 4028 4029 4030 4031 4032
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
4033
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4034
 */
4035
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
4036
{
4037
	spinlock_t *ptl;
H
Hugh Dickins 已提交
4038 4039
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
4040
		return -ENOMEM;
L
Linus Torvalds 已提交
4041

4042 4043
	smp_wmb(); /* See comment in __pte_alloc */

4044
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
4045
#ifndef __ARCH_HAS_4LEVEL_HACK
4046 4047
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
4048
		pud_populate(mm, pud, new);
4049
	} else	/* Another has populated it */
4050
		pmd_free(mm, new);
4051 4052 4053
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
4054
		pgd_populate(mm, pud, new);
4055 4056
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
4057
#endif /* __ARCH_HAS_4LEVEL_HACK */
4058
	spin_unlock(ptl);
4059
	return 0;
4060
}
L
Linus Torvalds 已提交
4061 4062
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
4063
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4064
			    struct mmu_notifier_range *range,
4065
			    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
4066 4067
{
	pgd_t *pgd;
4068
	p4d_t *p4d;
J
Johannes Weiner 已提交
4069 4070 4071 4072 4073 4074 4075 4076
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

4077 4078 4079 4080 4081
	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
4082 4083 4084 4085
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
4086
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
4087

R
Ross Zwisler 已提交
4088 4089 4090 4091
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

4092 4093 4094 4095
		if (range) {
			mmu_notifier_range_init(range, mm, address & PMD_MASK,
					     (address & PMD_MASK) + PMD_SIZE);
			mmu_notifier_invalidate_range_start(range);
4096
		}
R
Ross Zwisler 已提交
4097 4098 4099 4100 4101 4102
		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
4103 4104
		if (range)
			mmu_notifier_invalidate_range_end(range);
R
Ross Zwisler 已提交
4105 4106 4107
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
4108 4109
		goto out;

4110
	if (range) {
4111 4112
		mmu_notifier_range_init(range, mm, address & PAGE_MASK,
				     (address & PAGE_MASK) + PAGE_SIZE);
4113
		mmu_notifier_invalidate_range_start(range);
4114
	}
J
Johannes Weiner 已提交
4115 4116 4117 4118 4119 4120 4121
	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);
4122 4123
	if (range)
		mmu_notifier_invalidate_range_end(range);
J
Johannes Weiner 已提交
4124 4125 4126 4127
out:
	return -EINVAL;
}

4128 4129
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
4130 4131 4132 4133 4134
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4135
			   !(res = __follow_pte_pmd(mm, address, NULL,
4136
						    ptepp, NULL, ptlp)));
R
Ross Zwisler 已提交
4137 4138 4139 4140
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4141 4142
		   struct mmu_notifier_range *range,
		   pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
R
Ross Zwisler 已提交
4143 4144 4145 4146 4147
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4148
			   !(res = __follow_pte_pmd(mm, address, range,
4149
						    ptepp, pmdpp, ptlp)));
4150 4151
	return res;
}
R
Ross Zwisler 已提交
4152
EXPORT_SYMBOL(follow_pte_pmd);
4153

J
Johannes Weiner 已提交
4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182
/**
 * follow_pfn - look up PFN at a user virtual address
 * @vma: memory mapping
 * @address: user virtual address
 * @pfn: location to store found PFN
 *
 * Only IO mappings and raw PFN mappings are allowed.
 *
 * Returns zero and the pfn at @pfn on success, -ve otherwise.
 */
int follow_pfn(struct vm_area_struct *vma, unsigned long address,
	unsigned long *pfn)
{
	int ret = -EINVAL;
	spinlock_t *ptl;
	pte_t *ptep;

	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		return ret;

	ret = follow_pte(vma->vm_mm, address, &ptep, &ptl);
	if (ret)
		return ret;
	*pfn = pte_pfn(*ptep);
	pte_unmap_unlock(ptep, ptl);
	return 0;
}
EXPORT_SYMBOL(follow_pfn);

4183
#ifdef CONFIG_HAVE_IOREMAP_PROT
4184 4185 4186
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4187
{
4188
	int ret = -EINVAL;
4189 4190 4191
	pte_t *ptep, pte;
	spinlock_t *ptl;

4192 4193
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4194

4195
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4196
		goto out;
4197
	pte = *ptep;
4198

4199
	if ((flags & FOLL_WRITE) && !pte_write(pte))
4200 4201 4202
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4203
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4204

4205
	ret = 0;
4206 4207 4208
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4209
	return ret;
4210 4211 4212 4213 4214 4215 4216
}

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
4217
	void __iomem *maddr;
4218 4219
	int offset = addr & (PAGE_SIZE-1);

4220
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4221 4222
		return -EINVAL;

4223
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
4224 4225 4226
	if (!maddr)
		return -ENOMEM;

4227 4228 4229 4230 4231 4232 4233 4234
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
4235
EXPORT_SYMBOL_GPL(generic_access_phys);
4236 4237
#endif

4238
/*
4239 4240
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4241
 */
4242
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4243
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4244 4245 4246
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4247
	int write = gup_flags & FOLL_WRITE;
4248 4249

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4250
	/* ignore errors, just check how much was successfully transferred */
4251 4252 4253
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4254
		struct page *page = NULL;
4255

4256
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4257
				gup_flags, &page, &vma, NULL);
4258
		if (ret <= 0) {
4259 4260 4261
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4262 4263 4264 4265 4266
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4267
			if (!vma || vma->vm_start > addr)
4268 4269 4270 4271 4272 4273 4274
				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;
4275
#endif
4276
		} else {
4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291
			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);
4292
			put_page(page);
4293 4294 4295 4296 4297 4298 4299 4300 4301
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4302

S
Stephen Wilson 已提交
4303
/**
4304
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4305 4306 4307 4308
 * @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
4309
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4310 4311 4312 4313
 *
 * The caller must hold a reference on @mm.
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4314
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4315
{
4316
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4317 4318
}

4319 4320 4321 4322 4323 4324
/*
 * 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,
4325
		void *buf, int len, unsigned int gup_flags)
4326 4327 4328 4329 4330 4331 4332 4333
{
	struct mm_struct *mm;
	int ret;

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

4334
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4335

4336 4337 4338 4339
	mmput(mm);

	return ret;
}
4340
EXPORT_SYMBOL_GPL(access_process_vm);
4341

4342 4343 4344 4345 4346 4347 4348 4349
/*
 * 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;

4350
	/*
4351
	 * we might be running from an atomic context so we cannot sleep
4352
	 */
4353
	if (!down_read_trylock(&mm->mmap_sem))
4354 4355
		return;

4356 4357 4358
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
4359
		char *buf = (char *)__get_free_page(GFP_NOWAIT);
4360
		if (buf) {
A
Andy Shevchenko 已提交
4361
			char *p;
4362

M
Miklos Szeredi 已提交
4363
			p = file_path(f, buf, PAGE_SIZE);
4364 4365
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4366
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4367 4368 4369 4370 4371
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4372
	up_read(&mm->mmap_sem);
4373
}
4374

4375
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4376
void __might_fault(const char *file, int line)
4377
{
4378 4379 4380 4381 4382 4383
	/*
	 * 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
4384
	if (uaccess_kernel())
4385
		return;
4386
	if (pagefault_disabled())
4387
		return;
4388 4389
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4390
	if (current->mm)
4391
		might_lock_read(&current->mm->mmap_sem);
4392
#endif
4393
}
4394
EXPORT_SYMBOL(__might_fault);
4395
#endif
A
Andrea Arcangeli 已提交
4396 4397

#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
4398 4399 4400 4401 4402 4403 4404 4405 4406
/*
 * 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
4407
{
4408 4409 4410
	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
4411

4412
	/* Process target subpage last to keep its cache lines hot */
A
Andrea Arcangeli 已提交
4413
	might_sleep();
4414 4415
	n = (addr_hint - addr) / PAGE_SIZE;
	if (2 * n <= pages_per_huge_page) {
4416
		/* If target subpage in first half of huge page */
4417 4418
		base = 0;
		l = n;
4419
		/* Process subpages at the end of huge page */
4420 4421
		for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
			cond_resched();
4422
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4423 4424
		}
	} else {
4425
		/* If target subpage in second half of huge page */
4426 4427
		base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
		l = pages_per_huge_page - n;
4428
		/* Process subpages at the begin of huge page */
4429 4430
		for (i = 0; i < base; i++) {
			cond_resched();
4431
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4432 4433 4434
		}
	}
	/*
4435 4436
	 * Process remaining subpages in left-right-left-right pattern
	 * towards the target subpage
4437 4438 4439 4440 4441 4442
	 */
	for (i = 0; i < l; i++) {
		int left_idx = base + i;
		int right_idx = base + 2 * l - 1 - i;

		cond_resched();
4443
		process_subpage(addr + left_idx * PAGE_SIZE, left_idx, arg);
A
Andrea Arcangeli 已提交
4444
		cond_resched();
4445
		process_subpage(addr + right_idx * PAGE_SIZE, right_idx, arg);
A
Andrea Arcangeli 已提交
4446 4447 4448
	}
}

4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484
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
4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503
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);
	}
}

4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517
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
4518
void copy_user_huge_page(struct page *dst, struct page *src,
4519
			 unsigned long addr_hint, struct vm_area_struct *vma,
A
Andrea Arcangeli 已提交
4520 4521
			 unsigned int pages_per_huge_page)
{
4522 4523 4524 4525 4526 4527 4528
	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
4529 4530 4531 4532 4533 4534 4535

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

4536
	process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
A
Andrea Arcangeli 已提交
4537
}
4538 4539 4540

long copy_huge_page_from_user(struct page *dst_page,
				const void __user *usr_src,
4541 4542
				unsigned int pages_per_huge_page,
				bool allow_pagefault)
4543 4544 4545 4546 4547 4548 4549
{
	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++) {
4550 4551 4552 4553
		if (allow_pagefault)
			page_kaddr = kmap(dst_page + i);
		else
			page_kaddr = kmap_atomic(dst_page + i);
4554 4555 4556
		rc = copy_from_user(page_kaddr,
				(const void __user *)(src + i * PAGE_SIZE),
				PAGE_SIZE);
4557 4558 4559 4560
		if (allow_pagefault)
			kunmap(dst_page + i);
		else
			kunmap_atomic(page_kaddr);
4561 4562 4563 4564 4565 4566 4567 4568 4569

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

		cond_resched();
	}
	return ret_val;
}
A
Andrea Arcangeli 已提交
4570
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4571

4572
#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
4573 4574 4575 4576 4577 4578 4579 4580 4581

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);
}

4582
bool ptlock_alloc(struct page *page)
4583 4584 4585
{
	spinlock_t *ptl;

4586
	ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
4587 4588
	if (!ptl)
		return false;
4589
	page->ptl = ptl;
4590 4591 4592
	return true;
}

4593
void ptlock_free(struct page *page)
4594
{
4595
	kmem_cache_free(page_ptl_cachep, page->ptl);
4596 4597
}
#endif