memory.c 127.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
/*
 *  linux/mm/memory.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

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

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

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

/*
 * 05.04.94  -  Multi-page memory management added for v1.1.
T
Tobin C Harding 已提交
34
 *              Idea by Alex Bligh (alex@cconcepts.co.uk)
L
Linus Torvalds 已提交
35 36 37 38 39 40 41 42 43
 *
 * 16.07.99  -  Support of BIGMEM added by Gerhard Wichert, Siemens AG
 *		(Gerhard.Wichert@pdb.siemens.de)
 *
 * Aug/Sep 2004 Changed to four level page tables (Andi Kleen)
 */

#include <linux/kernel_stat.h>
#include <linux/mm.h>
44
#include <linux/sched/mm.h>
45
#include <linux/sched/coredump.h>
46
#include <linux/sched/numa_balancing.h>
47
#include <linux/sched/task.h>
L
Linus Torvalds 已提交
48 49 50 51 52
#include <linux/hugetlb.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
53
#include <linux/memremap.h>
H
Hugh Dickins 已提交
54
#include <linux/ksm.h>
L
Linus Torvalds 已提交
55
#include <linux/rmap.h>
56
#include <linux/export.h>
57
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
58
#include <linux/init.h>
59
#include <linux/pfn_t.h>
P
Peter Zijlstra 已提交
60
#include <linux/writeback.h>
61
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
62
#include <linux/mmu_notifier.h>
63 64
#include <linux/swapops.h>
#include <linux/elf.h>
65
#include <linux/gfp.h>
66
#include <linux/migrate.h>
A
Andy Shevchenko 已提交
67
#include <linux/string.h>
68
#include <linux/dma-debug.h>
69
#include <linux/debugfs.h>
70
#include <linux/userfaultfd_k.h>
71
#include <linux/dax.h>
72
#include <linux/oom.h>
73
#include <linux/numa.h>
L
Linus Torvalds 已提交
74

A
Alexey Dobriyan 已提交
75
#include <asm/io.h>
76
#include <asm/mmu_context.h>
L
Linus Torvalds 已提交
77
#include <asm/pgalloc.h>
78
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
79 80 81 82
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgtable.h>

83 84
#include "internal.h"

85
#if defined(LAST_CPUPID_NOT_IN_PAGE_FLAGS) && !defined(CONFIG_COMPILE_TEST)
86
#warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
87 88
#endif

A
Andy Whitcroft 已提交
89
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
90 91 92
/* use the per-pgdat data instead for discontigmem - mbligh */
unsigned long max_mapnr;
EXPORT_SYMBOL(max_mapnr);
T
Tobin C Harding 已提交
93 94

struct page *mem_map;
L
Linus Torvalds 已提交
95 96 97 98 99 100 101 102 103 104
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
105
void *high_memory;
L
Linus Torvalds 已提交
106 107
EXPORT_SYMBOL(high_memory);

108 109 110 111 112 113 114 115 116 117 118 119
/*
 * 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
120 121 122 123

static int __init disable_randmaps(char *s)
{
	randomize_va_space = 0;
124
	return 1;
125 126 127
}
__setup("norandmaps", disable_randmaps);

H
Hugh Dickins 已提交
128
unsigned long zero_pfn __read_mostly;
129 130
EXPORT_SYMBOL(zero_pfn);

T
Tobin C Harding 已提交
131 132
unsigned long highest_memmap_pfn __read_mostly;

H
Hugh Dickins 已提交
133 134 135 136 137 138 139 140 141
/*
 * 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);
142

K
KAMEZAWA Hiroyuki 已提交
143

144 145
#if defined(SPLIT_RSS_COUNTING)

146
void sync_mm_rss(struct mm_struct *mm)
147 148 149 150
{
	int i;

	for (i = 0; i < NR_MM_COUNTERS; i++) {
151 152 153
		if (current->rss_stat.count[i]) {
			add_mm_counter(mm, i, current->rss_stat.count[i]);
			current->rss_stat.count[i] = 0;
154 155
		}
	}
156
	current->rss_stat.events = 0;
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
}

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))
178
		sync_mm_rss(task->mm);
179
}
180
#else /* SPLIT_RSS_COUNTING */
181 182 183 184 185 186 187 188

#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)
{
}

189 190
#endif /* SPLIT_RSS_COUNTING */

L
Linus Torvalds 已提交
191 192 193 194
/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
195 196
static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
			   unsigned long addr)
L
Linus Torvalds 已提交
197
{
198
	pgtable_t token = pmd_pgtable(*pmd);
199
	pmd_clear(pmd);
200
	pte_free_tlb(tlb, token, addr);
201
	mm_dec_nr_ptes(tlb->mm);
L
Linus Torvalds 已提交
202 203
}

204 205 206
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
207 208 209
{
	pmd_t *pmd;
	unsigned long next;
210
	unsigned long start;
L
Linus Torvalds 已提交
211

212
	start = addr;
L
Linus Torvalds 已提交
213 214 215 216 217
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
218
		free_pte_range(tlb, pmd, addr);
L
Linus Torvalds 已提交
219 220
	} while (pmd++, addr = next, addr != end);

221 222 223 224 225 226 227
	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
228
	}
229 230 231 232 233
	if (end - 1 > ceiling - 1)
		return;

	pmd = pmd_offset(pud, start);
	pud_clear(pud);
234
	pmd_free_tlb(tlb, pmd, start);
235
	mm_dec_nr_pmds(tlb->mm);
L
Linus Torvalds 已提交
236 237
}

238
static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
239 240
				unsigned long addr, unsigned long end,
				unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
241 242 243
{
	pud_t *pud;
	unsigned long next;
244
	unsigned long start;
L
Linus Torvalds 已提交
245

246
	start = addr;
247
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
248 249 250 251
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
252
		free_pmd_range(tlb, pud, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
253 254
	} while (pud++, addr = next, addr != end);

255 256 257 258 259 260 261 262 263 264 265 266 267 268
	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
269
	mm_dec_nr_puds(tlb->mm);
270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
}

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);

289 290 291 292 293 294 295
	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
296
	}
297 298 299
	if (end - 1 > ceiling - 1)
		return;

300
	p4d = p4d_offset(pgd, start);
301
	pgd_clear(pgd);
302
	p4d_free_tlb(tlb, p4d, start);
L
Linus Torvalds 已提交
303 304 305
}

/*
306
 * This function frees user-level page tables of a process.
L
Linus Torvalds 已提交
307
 */
308
void free_pgd_range(struct mmu_gather *tlb,
309 310
			unsigned long addr, unsigned long end,
			unsigned long floor, unsigned long ceiling)
L
Linus Torvalds 已提交
311 312 313
{
	pgd_t *pgd;
	unsigned long next;
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339

	/*
	 * 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
340

341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
	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;
356 357 358 359
	/*
	 * We add page table cache pages with PAGE_SIZE,
	 * (see pte_free_tlb()), flush the tlb if we need
	 */
360
	tlb_change_page_size(tlb, PAGE_SIZE);
361
	pgd = pgd_offset(tlb->mm, addr);
L
Linus Torvalds 已提交
362 363 364 365
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
366
		free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
L
Linus Torvalds 已提交
367
	} while (pgd++, addr = next, addr != end);
368 369
}

370
void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
371
		unsigned long floor, unsigned long ceiling)
372 373 374 375 376
{
	while (vma) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long addr = vma->vm_start;

377
		/*
N
npiggin@suse.de 已提交
378 379
		 * Hide vma from rmap and truncate_pagecache before freeing
		 * pgtables
380
		 */
381
		unlink_anon_vmas(vma);
382 383
		unlink_file_vma(vma);

384
		if (is_vm_hugetlb_page(vma)) {
385
			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
386
				floor, next ? next->vm_start : ceiling);
387 388 389 390 391
		} else {
			/*
			 * Optimization: gather nearby vmas into one call down
			 */
			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
392
			       && !is_vm_hugetlb_page(next)) {
393 394
				vma = next;
				next = vma->vm_next;
395
				unlink_anon_vmas(vma);
396
				unlink_file_vma(vma);
397 398
			}
			free_pgd_range(tlb, addr, vma->vm_end,
T
Tobin C Harding 已提交
399
				floor, next ? next->vm_start : ceiling);
400
		}
401 402
		vma = next;
	}
L
Linus Torvalds 已提交
403 404
}

405
int __pte_alloc(struct mm_struct *mm, pmd_t *pmd)
L
Linus Torvalds 已提交
406
{
407
	spinlock_t *ptl;
408
	pgtable_t new = pte_alloc_one(mm);
409 410 411
	if (!new)
		return -ENOMEM;

412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
	/*
	 * 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 */

427
	ptl = pmd_lock(mm, pmd);
428
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
429
		mm_inc_nr_ptes(mm);
L
Linus Torvalds 已提交
430
		pmd_populate(mm, pmd, new);
431
		new = NULL;
432
	}
433
	spin_unlock(ptl);
434 435
	if (new)
		pte_free(mm, new);
436
	return 0;
L
Linus Torvalds 已提交
437 438
}

439
int __pte_alloc_kernel(pmd_t *pmd)
L
Linus Torvalds 已提交
440
{
441
	pte_t *new = pte_alloc_one_kernel(&init_mm);
442 443 444
	if (!new)
		return -ENOMEM;

445 446
	smp_wmb(); /* See comment in __pte_alloc */

447
	spin_lock(&init_mm.page_table_lock);
448
	if (likely(pmd_none(*pmd))) {	/* Has another populated it ? */
449
		pmd_populate_kernel(&init_mm, pmd, new);
450
		new = NULL;
451
	}
452
	spin_unlock(&init_mm.page_table_lock);
453 454
	if (new)
		pte_free_kernel(&init_mm, new);
455
	return 0;
L
Linus Torvalds 已提交
456 457
}

K
KAMEZAWA Hiroyuki 已提交
458 459 460 461 462 463
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)
464
{
K
KAMEZAWA Hiroyuki 已提交
465 466
	int i;

467
	if (current->mm == mm)
468
		sync_mm_rss(mm);
K
KAMEZAWA Hiroyuki 已提交
469 470 471
	for (i = 0; i < NR_MM_COUNTERS; i++)
		if (rss[i])
			add_mm_counter(mm, i, rss[i]);
472 473
}

N
Nick Piggin 已提交
474
/*
475 476 477
 * 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
478 479 480
 *
 * The calling function must still handle the error.
 */
481 482
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
483
{
484
	pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
485 486
	p4d_t *p4d = p4d_offset(pgd, addr);
	pud_t *pud = pud_offset(p4d, addr);
487 488 489
	pmd_t *pmd = pmd_offset(pud, addr);
	struct address_space *mapping;
	pgoff_t index;
490 491 492 493 494 495 496 497 498 499 500 501 502 503
	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) {
504 505
			pr_alert("BUG: Bad page map: %lu messages suppressed\n",
				 nr_unshown);
506 507 508 509 510 511
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;
512 513 514 515

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

516 517 518
	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));
519
	if (page)
520
		dump_page(page, "bad pte");
521 522
	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);
523
	pr_alert("file:%pD fault:%ps mmap:%ps readpage:%ps\n",
524 525 526 527
		 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
528
	dump_stack();
529
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
N
Nick Piggin 已提交
530 531
}

H
Hugh Dickins 已提交
532
/*
N
Nick Piggin 已提交
533
 * vm_normal_page -- This function gets the "struct page" associated with a pte.
534
 *
N
Nick Piggin 已提交
535 536 537
 * "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
538
 *
N
Nick Piggin 已提交
539 540 541 542 543 544 545 546
 * 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.
547
 *
J
Jared Hulbert 已提交
548 549
 * 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
550 551
 * set, and the vm_pgoff will point to the first PFN mapped: thus every special
 * mapping will always honor the rule
552 553 554
 *
 *	pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
 *
N
Nick Piggin 已提交
555 556 557 558 559 560
 * 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
561 562
 *
 *
N
Nick Piggin 已提交
563
 * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
J
Jared Hulbert 已提交
564 565 566 567 568 569 570 571 572
 *
 * 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
573
 */
574 575
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
576
{
577
	unsigned long pfn = pte_pfn(pte);
N
Nick Piggin 已提交
578

L
Laurent Dufour 已提交
579
	if (IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL)) {
580
		if (likely(!pte_special(pte)))
581
			goto check_pfn;
582 583
		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
584 585
		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			return NULL;
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600
		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().
		 */
601
		if (likely(pfn <= highest_memmap_pfn)) {
602 603 604 605 606 607 608 609
			struct page *page = pfn_to_page(pfn);

			if (is_device_public_page(page)) {
				if (with_public_device)
					return page;
				return NULL;
			}
		}
610 611 612 613

		if (pte_devmap(pte))
			return NULL;

614
		print_bad_pte(vma, addr, pte, NULL);
N
Nick Piggin 已提交
615 616 617
		return NULL;
	}

L
Laurent Dufour 已提交
618
	/* !CONFIG_ARCH_HAS_PTE_SPECIAL case follows: */
N
Nick Piggin 已提交
619

J
Jared Hulbert 已提交
620 621 622 623 624 625
	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
626 627
			unsigned long off;
			off = (addr - vma->vm_start) >> PAGE_SHIFT;
J
Jared Hulbert 已提交
628 629 630 631 632
			if (pfn == vma->vm_pgoff + off)
				return NULL;
			if (!is_cow_mapping(vma->vm_flags))
				return NULL;
		}
633 634
	}

635 636
	if (is_zero_pfn(pfn))
		return NULL;
L
Laurent Dufour 已提交
637

638 639 640 641 642
check_pfn:
	if (unlikely(pfn > highest_memmap_pfn)) {
		print_bad_pte(vma, addr, pte, NULL);
		return NULL;
	}
643 644

	/*
N
Nick Piggin 已提交
645 646
	 * NOTE! We still have PageReserved() pages in the page tables.
	 * eg. VDSO mappings can cause them to exist.
647
	 */
J
Jared Hulbert 已提交
648
out:
649
	return pfn_to_page(pfn);
H
Hugh Dickins 已提交
650 651
}

652 653 654 655 656 657 658 659 660
#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
661
	 * !CONFIG_ARCH_HAS_PTE_SPECIAL case from vm_normal_page() here.
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
	 */
	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;
		}
	}

678 679
	if (pmd_devmap(pmd))
		return NULL;
680 681 682 683 684 685 686 687 688 689 690 691 692 693
	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
694 695 696 697 698 699
/*
 * 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
700
static inline unsigned long
L
Linus Torvalds 已提交
701
copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
N
Nick Piggin 已提交
702
		pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
H
Hugh Dickins 已提交
703
		unsigned long addr, int *rss)
L
Linus Torvalds 已提交
704
{
N
Nick Piggin 已提交
705
	unsigned long vm_flags = vma->vm_flags;
L
Linus Torvalds 已提交
706 707 708 709 710
	pte_t pte = *src_pte;
	struct page *page;

	/* pte contains position in swap or file, so copy. */
	if (unlikely(!pte_present(pte))) {
711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728
		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);

729
			rss[mm_counter(page)]++;
730 731 732 733 734 735 736 737 738 739 740 741

			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);
742
			}
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
		} 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
772
		}
773
		goto out_set_pte;
L
Linus Torvalds 已提交
774 775 776 777 778 779
	}

	/*
	 * If it's a COW mapping, write protect it both
	 * in the parent and the child
	 */
780
	if (is_cow_mapping(vm_flags) && pte_write(pte)) {
L
Linus Torvalds 已提交
781
		ptep_set_wrprotect(src_mm, addr, src_pte);
782
		pte = pte_wrprotect(pte);
L
Linus Torvalds 已提交
783 784 785 786 787 788 789 790 791
	}

	/*
	 * 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);
792 793 794 795

	page = vm_normal_page(vma, addr, pte);
	if (page) {
		get_page(page);
796
		page_dup_rmap(page, false);
797
		rss[mm_counter(page)]++;
798 799 800 801 802 803 804 805 806 807 808 809 810
	} 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)]++;
		}
811
	}
812 813 814

out_set_pte:
	set_pte_at(dst_mm, addr, dst_pte, pte);
H
Hugh Dickins 已提交
815
	return 0;
L
Linus Torvalds 已提交
816 817
}

818
static int copy_pte_range(struct mm_struct *dst_mm, struct mm_struct *src_mm,
819 820
		   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
821
{
822
	pte_t *orig_src_pte, *orig_dst_pte;
L
Linus Torvalds 已提交
823
	pte_t *src_pte, *dst_pte;
H
Hugh Dickins 已提交
824
	spinlock_t *src_ptl, *dst_ptl;
825
	int progress = 0;
K
KAMEZAWA Hiroyuki 已提交
826
	int rss[NR_MM_COUNTERS];
H
Hugh Dickins 已提交
827
	swp_entry_t entry = (swp_entry_t){0};
L
Linus Torvalds 已提交
828 829

again:
K
KAMEZAWA Hiroyuki 已提交
830 831
	init_rss_vec(rss);

H
Hugh Dickins 已提交
832
	dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
L
Linus Torvalds 已提交
833 834
	if (!dst_pte)
		return -ENOMEM;
P
Peter Zijlstra 已提交
835
	src_pte = pte_offset_map(src_pmd, addr);
H
Hugh Dickins 已提交
836
	src_ptl = pte_lockptr(src_mm, src_pmd);
I
Ingo Molnar 已提交
837
	spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
838 839
	orig_src_pte = src_pte;
	orig_dst_pte = dst_pte;
840
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
841 842 843 844 845 846

	do {
		/*
		 * We are holding two locks at this point - either of them
		 * could generate latencies in another task on another CPU.
		 */
847 848 849
		if (progress >= 32) {
			progress = 0;
			if (need_resched() ||
N
Nick Piggin 已提交
850
			    spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
851 852
				break;
		}
L
Linus Torvalds 已提交
853 854 855 856
		if (pte_none(*src_pte)) {
			progress++;
			continue;
		}
H
Hugh Dickins 已提交
857 858 859 860
		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
861 862 863
		progress += 8;
	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);

864
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
865
	spin_unlock(src_ptl);
P
Peter Zijlstra 已提交
866
	pte_unmap(orig_src_pte);
K
KAMEZAWA Hiroyuki 已提交
867
	add_mm_rss_vec(dst_mm, rss);
868
	pte_unmap_unlock(orig_dst_pte, dst_ptl);
H
Hugh Dickins 已提交
869
	cond_resched();
H
Hugh Dickins 已提交
870 871 872 873 874 875

	if (entry.val) {
		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
			return -ENOMEM;
		progress = 0;
	}
L
Linus Torvalds 已提交
876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
	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);
894 895
		if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
			|| pmd_devmap(*src_pmd)) {
896
			int err;
897
			VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, vma);
898 899 900 901 902 903 904 905
			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
906 907 908 909 910 911 912 913 914 915
		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,
916
		p4d_t *dst_p4d, p4d_t *src_p4d, struct vm_area_struct *vma,
L
Linus Torvalds 已提交
917 918 919 920 921
		unsigned long addr, unsigned long end)
{
	pud_t *src_pud, *dst_pud;
	unsigned long next;

922
	dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
L
Linus Torvalds 已提交
923 924
	if (!dst_pud)
		return -ENOMEM;
925
	src_pud = pud_offset(src_p4d, addr);
L
Linus Torvalds 已提交
926 927
	do {
		next = pud_addr_end(addr, end);
928 929 930 931 932 933 934 935 936 937 938 939
		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
940 941 942 943 944 945 946 947 948
		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;
}

949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970
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
971 972 973 974 975 976 977
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;
978
	struct mmu_notifier_range range;
979
	bool is_cow;
A
Andrea Arcangeli 已提交
980
	int ret;
L
Linus Torvalds 已提交
981

982 983 984 985 986 987
	/*
	 * 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.
	 */
988 989 990
	if (!(vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
			!vma->anon_vma)
		return 0;
991

L
Linus Torvalds 已提交
992 993 994
	if (is_vm_hugetlb_page(vma))
		return copy_hugetlb_page_range(dst_mm, src_mm, vma);

995
	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
996 997 998 999
		/*
		 * We do not free on error cases below as remove_vma
		 * gets called on error from higher level routine
		 */
1000
		ret = track_pfn_copy(vma);
1001 1002 1003 1004
		if (ret)
			return ret;
	}

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

	if (is_cow) {
1014 1015
		mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_PAGE,
					0, vma, src_mm, addr, end);
1016 1017
		mmu_notifier_invalidate_range_start(&range);
	}
A
Andrea Arcangeli 已提交
1018 1019

	ret = 0;
L
Linus Torvalds 已提交
1020 1021 1022 1023 1024 1025
	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;
1026
		if (unlikely(copy_p4d_range(dst_mm, src_mm, dst_pgd, src_pgd,
A
Andrea Arcangeli 已提交
1027 1028 1029 1030
					    vma, addr, next))) {
			ret = -ENOMEM;
			break;
		}
L
Linus Torvalds 已提交
1031
	} while (dst_pgd++, src_pgd++, addr = next, addr != end);
A
Andrea Arcangeli 已提交
1032

1033
	if (is_cow)
1034
		mmu_notifier_invalidate_range_end(&range);
A
Andrea Arcangeli 已提交
1035
	return ret;
L
Linus Torvalds 已提交
1036 1037
}

1038
static unsigned long zap_pte_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1039
				struct vm_area_struct *vma, pmd_t *pmd,
L
Linus Torvalds 已提交
1040
				unsigned long addr, unsigned long end,
1041
				struct zap_details *details)
L
Linus Torvalds 已提交
1042
{
N
Nick Piggin 已提交
1043
	struct mm_struct *mm = tlb->mm;
P
Peter Zijlstra 已提交
1044
	int force_flush = 0;
K
KAMEZAWA Hiroyuki 已提交
1045
	int rss[NR_MM_COUNTERS];
1046
	spinlock_t *ptl;
1047
	pte_t *start_pte;
1048
	pte_t *pte;
1049
	swp_entry_t entry;
K
KAMEZAWA Hiroyuki 已提交
1050

1051
	tlb_change_page_size(tlb, PAGE_SIZE);
P
Peter Zijlstra 已提交
1052
again:
1053
	init_rss_vec(rss);
1054 1055
	start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	pte = start_pte;
1056
	flush_tlb_batched_pending(mm);
1057
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1058 1059
	do {
		pte_t ptent = *pte;
T
Tobin C Harding 已提交
1060
		if (pte_none(ptent))
L
Linus Torvalds 已提交
1061
			continue;
1062

L
Linus Torvalds 已提交
1063
		if (pte_present(ptent)) {
H
Hugh Dickins 已提交
1064
			struct page *page;
1065

1066
			page = _vm_normal_page(vma, addr, ptent, true);
L
Linus Torvalds 已提交
1067 1068 1069 1070 1071 1072 1073
			if (unlikely(details) && page) {
				/*
				 * unmap_shared_mapping_pages() wants to
				 * invalidate cache without truncating:
				 * unmap shared but keep private pages.
				 */
				if (details->check_mapping &&
1074
				    details->check_mapping != page_rmapping(page))
L
Linus Torvalds 已提交
1075 1076
					continue;
			}
N
Nick Piggin 已提交
1077
			ptent = ptep_get_and_clear_full(mm, addr, pte,
1078
							tlb->fullmm);
L
Linus Torvalds 已提交
1079 1080 1081
			tlb_remove_tlb_entry(tlb, pte, addr);
			if (unlikely(!page))
				continue;
1082 1083

			if (!PageAnon(page)) {
1084 1085
				if (pte_dirty(ptent)) {
					force_flush = 1;
1086
					set_page_dirty(page);
1087
				}
1088
				if (pte_young(ptent) &&
1089
				    likely(!(vma->vm_flags & VM_SEQ_READ)))
1090
					mark_page_accessed(page);
1091
			}
1092
			rss[mm_counter(page)]--;
1093
			page_remove_rmap(page, false);
1094 1095
			if (unlikely(page_mapcount(page) < 0))
				print_bad_pte(vma, addr, ptent, page);
1096
			if (unlikely(__tlb_remove_page(tlb, page))) {
1097
				force_flush = 1;
1098
				addr += PAGE_SIZE;
P
Peter Zijlstra 已提交
1099
				break;
1100
			}
L
Linus Torvalds 已提交
1101 1102
			continue;
		}
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125

		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;
		}

1126 1127
		/* If details->check_mapping, we leave swap entries. */
		if (unlikely(details))
L
Linus Torvalds 已提交
1128
			continue;
K
KAMEZAWA Hiroyuki 已提交
1129

1130 1131 1132 1133 1134
		entry = pte_to_swp_entry(ptent);
		if (!non_swap_entry(entry))
			rss[MM_SWAPENTS]--;
		else if (is_migration_entry(entry)) {
			struct page *page;
1135

1136
			page = migration_entry_to_page(entry);
1137
			rss[mm_counter(page)]--;
K
KAMEZAWA Hiroyuki 已提交
1138
		}
1139 1140
		if (unlikely(!free_swap_and_cache(entry)))
			print_bad_pte(vma, addr, ptent, NULL);
1141
		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
1142
	} while (pte++, addr += PAGE_SIZE, addr != end);
1143

K
KAMEZAWA Hiroyuki 已提交
1144
	add_mm_rss_vec(mm, rss);
1145
	arch_leave_lazy_mmu_mode();
1146

1147
	/* Do the actual TLB flush before dropping ptl */
1148
	if (force_flush)
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
		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;
1160
		tlb_flush_mmu(tlb);
1161
		if (addr != end)
P
Peter Zijlstra 已提交
1162 1163 1164
			goto again;
	}

1165
	return addr;
L
Linus Torvalds 已提交
1166 1167
}

1168
static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
N
Nick Piggin 已提交
1169
				struct vm_area_struct *vma, pud_t *pud,
L
Linus Torvalds 已提交
1170
				unsigned long addr, unsigned long end,
1171
				struct zap_details *details)
L
Linus Torvalds 已提交
1172 1173 1174 1175 1176 1177 1178
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
1179
		if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
1180
			if (next - addr != HPAGE_PMD_SIZE)
1181
				__split_huge_pmd(vma, pmd, addr, false, NULL);
1182
			else if (zap_huge_pmd(tlb, vma, pmd, addr))
1183
				goto next;
1184 1185
			/* fall through */
		}
1186 1187 1188 1189 1190 1191 1192 1193 1194
		/*
		 * 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;
1195
		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
1196
next:
1197 1198
		cond_resched();
	} while (pmd++, addr = next, addr != end);
1199 1200

	return addr;
L
Linus Torvalds 已提交
1201 1202
}

1203
static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
1204
				struct vm_area_struct *vma, p4d_t *p4d,
L
Linus Torvalds 已提交
1205
				unsigned long addr, unsigned long end,
1206
				struct zap_details *details)
L
Linus Torvalds 已提交
1207 1208 1209 1210
{
	pud_t *pud;
	unsigned long next;

1211
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
1212 1213
	do {
		next = pud_addr_end(addr, end);
1214 1215 1216 1217 1218 1219 1220 1221
		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 */
		}
1222
		if (pud_none_or_clear_bad(pud))
L
Linus Torvalds 已提交
1223
			continue;
1224
		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
1225 1226
next:
		cond_resched();
1227
	} while (pud++, addr = next, addr != end);
1228 1229

	return addr;
L
Linus Torvalds 已提交
1230 1231
}

1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
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
1251
void unmap_page_range(struct mmu_gather *tlb,
A
Al Viro 已提交
1252 1253 1254
			     struct vm_area_struct *vma,
			     unsigned long addr, unsigned long end,
			     struct zap_details *details)
L
Linus Torvalds 已提交
1255 1256 1257 1258 1259 1260 1261 1262 1263
{
	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);
1264
		if (pgd_none_or_clear_bad(pgd))
L
Linus Torvalds 已提交
1265
			continue;
1266
		next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
1267
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
1268 1269
	tlb_end_vma(tlb, vma);
}
1270

1271 1272 1273

static void unmap_single_vma(struct mmu_gather *tlb,
		struct vm_area_struct *vma, unsigned long start_addr,
1274
		unsigned long end_addr,
1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
		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;

1286 1287 1288
	if (vma->vm_file)
		uprobe_munmap(vma, start, end);

1289
	if (unlikely(vma->vm_flags & VM_PFNMAP))
1290
		untrack_pfn(vma, 0, 0);
1291 1292 1293 1294 1295 1296 1297

	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
1298
			 * cleanup path of mmap_region. When
1299
			 * hugetlbfs ->mmap method fails,
1300
			 * mmap_region() nullifies vma->vm_file
1301 1302 1303 1304
			 * before calling this function to clean up.
			 * Since no pte has actually been setup, it is
			 * safe to do nothing in this case.
			 */
1305
			if (vma->vm_file) {
1306
				i_mmap_lock_write(vma->vm_file->f_mapping);
1307
				__unmap_hugepage_range_final(tlb, vma, start, end, NULL);
1308
				i_mmap_unlock_write(vma->vm_file->f_mapping);
1309
			}
1310 1311 1312
		} else
			unmap_page_range(tlb, vma, start, end, details);
	}
L
Linus Torvalds 已提交
1313 1314 1315 1316
}

/**
 * unmap_vmas - unmap a range of memory covered by a list of vma's
1317
 * @tlb: address of the caller's struct mmu_gather
L
Linus Torvalds 已提交
1318 1319 1320 1321
 * @vma: the starting vma
 * @start_addr: virtual address at which to start unmapping
 * @end_addr: virtual address at which to end unmapping
 *
1322
 * Unmap all pages in the vma list.
L
Linus Torvalds 已提交
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
 *
 * 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
1333
void unmap_vmas(struct mmu_gather *tlb,
L
Linus Torvalds 已提交
1334
		struct vm_area_struct *vma, unsigned long start_addr,
1335
		unsigned long end_addr)
L
Linus Torvalds 已提交
1336
{
1337
	struct mmu_notifier_range range;
L
Linus Torvalds 已提交
1338

1339 1340
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, vma->vm_mm,
				start_addr, end_addr);
1341
	mmu_notifier_invalidate_range_start(&range);
1342
	for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
1343
		unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
1344
	mmu_notifier_invalidate_range_end(&range);
L
Linus Torvalds 已提交
1345 1346 1347 1348 1349
}

/**
 * zap_page_range - remove user pages in a given range
 * @vma: vm_area_struct holding the applicable pages
1350
 * @start: starting address of pages to zap
L
Linus Torvalds 已提交
1351
 * @size: number of bytes to zap
1352 1353
 *
 * Caller must protect the VMA list
L
Linus Torvalds 已提交
1354
 */
1355
void zap_page_range(struct vm_area_struct *vma, unsigned long start,
1356
		unsigned long size)
L
Linus Torvalds 已提交
1357
{
1358
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1359
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1360 1361

	lru_add_drain();
1362
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
1363
				start, start + size);
1364 1365 1366 1367 1368 1369 1370
	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
1371 1372
}

1373 1374 1375 1376 1377
/**
 * 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
1378
 * @details: details of shared cache invalidation
1379 1380
 *
 * The range must fit into one VMA.
L
Linus Torvalds 已提交
1381
 */
1382
static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
L
Linus Torvalds 已提交
1383 1384
		unsigned long size, struct zap_details *details)
{
1385
	struct mmu_notifier_range range;
P
Peter Zijlstra 已提交
1386
	struct mmu_gather tlb;
L
Linus Torvalds 已提交
1387 1388

	lru_add_drain();
1389
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
1390
				address, address + size);
1391 1392 1393 1394 1395 1396
	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
1397 1398
}

1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409
/**
 * 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.
 *
 */
1410
void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1411 1412 1413 1414
		unsigned long size)
{
	if (address < vma->vm_start || address + size > vma->vm_end ||
	    		!(vma->vm_flags & VM_PFNMAP))
1415 1416
		return;

1417
	zap_page_range_single(vma, address, size, NULL);
1418 1419 1420
}
EXPORT_SYMBOL_GPL(zap_vma_ptes);

1421
pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
H
Harvey Harrison 已提交
1422
			spinlock_t **ptl)
1423
{
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
	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);
1442 1443
}

1444 1445 1446 1447 1448 1449 1450
/*
 * 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
1451 1452
static int insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page, pgprot_t prot)
1453
{
N
Nick Piggin 已提交
1454
	struct mm_struct *mm = vma->vm_mm;
1455
	int retval;
1456
	pte_t *pte;
1457 1458
	spinlock_t *ptl;

1459
	retval = -EINVAL;
1460
	if (PageAnon(page) || PageSlab(page) || page_has_type(page))
1461
		goto out;
1462 1463
	retval = -ENOMEM;
	flush_dcache_page(page);
1464
	pte = get_locked_pte(mm, addr, &ptl);
1465
	if (!pte)
1466
		goto out;
1467 1468 1469 1470 1471 1472
	retval = -EBUSY;
	if (!pte_none(*pte))
		goto out_unlock;

	/* Ok, finally just insert the thing.. */
	get_page(page);
1473
	inc_mm_counter_fast(mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
1474
	page_add_file_rmap(page, false);
1475 1476 1477 1478 1479 1480 1481 1482 1483
	set_pte_at(mm, addr, pte, mk_pte(page, prot));

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

1484 1485 1486 1487 1488 1489
/**
 * 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
 *
1490 1491 1492 1493 1494 1495
 * 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
1496
 * (see split_page()).
1497 1498 1499 1500 1501 1502 1503 1504
 *
 * 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.
1505 1506 1507 1508 1509
 *
 * 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.
1510 1511
 *
 * Return: %0 on success, negative error code otherwise.
1512
 */
N
Nick Piggin 已提交
1513 1514
int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
			struct page *page)
1515 1516 1517 1518 1519
{
	if (addr < vma->vm_start || addr >= vma->vm_end)
		return -EFAULT;
	if (!page_count(page))
		return -EINVAL;
1520 1521 1522 1523 1524
	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
1525
	return insert_page(vma, addr, page, vma->vm_page_prot);
1526
}
1527
EXPORT_SYMBOL(vm_insert_page);
1528

1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
/*
 * __vm_map_pages - maps range of kernel pages into user vma
 * @vma: user vma to map to
 * @pages: pointer to array of source kernel pages
 * @num: number of pages in page array
 * @offset: user's requested vm_pgoff
 *
 * This allows drivers to map range of kernel pages into a user vma.
 *
 * Return: 0 on success and error code otherwise.
 */
static int __vm_map_pages(struct vm_area_struct *vma, struct page **pages,
				unsigned long num, unsigned long offset)
{
	unsigned long count = vma_pages(vma);
	unsigned long uaddr = vma->vm_start;
	int ret, i;

	/* Fail if the user requested offset is beyond the end of the object */
1548
	if (offset >= num)
1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
		return -ENXIO;

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

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

	return 0;
}

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

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

1610
static vm_fault_t insert_pfn(struct vm_area_struct *vma, unsigned long addr,
R
Ross Zwisler 已提交
1611
			pfn_t pfn, pgprot_t prot, bool mkwrite)
N
Nick Piggin 已提交
1612 1613 1614 1615 1616 1617 1618
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte, entry;
	spinlock_t *ptl;

	pte = get_locked_pte(mm, addr, &ptl);
	if (!pte)
1619
		return VM_FAULT_OOM;
R
Ross Zwisler 已提交
1620 1621 1622 1623 1624 1625 1626
	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
1627 1628 1629 1630
			 * 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
1631
			 */
J
Jan Kara 已提交
1632 1633
			if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
				WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
R
Ross Zwisler 已提交
1634
				goto out_unlock;
J
Jan Kara 已提交
1635
			}
1636 1637 1638 1639 1640 1641
			entry = pte_mkyoung(*pte);
			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
			if (ptep_set_access_flags(vma, addr, pte, entry, 1))
				update_mmu_cache(vma, addr, pte);
		}
		goto out_unlock;
R
Ross Zwisler 已提交
1642
	}
N
Nick Piggin 已提交
1643 1644

	/* Ok, finally just insert the thing.. */
1645 1646 1647 1648
	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
1649 1650 1651 1652 1653 1654

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

N
Nick Piggin 已提交
1655
	set_pte_at(mm, addr, pte, entry);
1656
	update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
N
Nick Piggin 已提交
1657 1658 1659

out_unlock:
	pte_unmap_unlock(pte, ptl);
1660
	return VM_FAULT_NOPAGE;
N
Nick Piggin 已提交
1661 1662
}

1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674
/**
 * 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
1675
 * vmf_insert_pfn_prot should only be used if using multiple VMAs is
1676 1677
 * impractical.
 *
M
Matthew Wilcox 已提交
1678
 * Context: Process context.  May allocate using %GFP_KERNEL.
1679 1680 1681 1682 1683
 * 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)
{
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703
	/*
	 * 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));

1704
	return insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
1705
			false);
1706 1707
}
EXPORT_SYMBOL(vmf_insert_pfn_prot);
N
Nick Piggin 已提交
1708

M
Matthew Wilcox 已提交
1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735
/**
 * 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);

1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
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;
}

1750 1751
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
1752
{
1753
	pgprot_t pgprot = vma->vm_page_prot;
1754
	int err;
1755

1756
	BUG_ON(!vm_mixed_ok(vma, pfn));
N
Nick Piggin 已提交
1757

N
Nick Piggin 已提交
1758
	if (addr < vma->vm_start || addr >= vma->vm_end)
1759
		return VM_FAULT_SIGBUS;
1760 1761

	track_pfn_insert(vma, &pgprot, pfn);
N
Nick Piggin 已提交
1762

1763
	if (!pfn_modify_allowed(pfn_t_to_pfn(pfn), pgprot))
1764
		return VM_FAULT_SIGBUS;
1765

N
Nick Piggin 已提交
1766 1767 1768 1769
	/*
	 * 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
1770 1771
	 * 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
1772
	 */
L
Laurent Dufour 已提交
1773 1774
	if (!IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL) &&
	    !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
N
Nick Piggin 已提交
1775 1776
		struct page *page;

1777 1778 1779 1780 1781 1782
		/*
		 * 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));
1783 1784
		err = insert_page(vma, addr, page, pgprot);
	} else {
1785
		return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
N
Nick Piggin 已提交
1786
	}
R
Ross Zwisler 已提交
1787

M
Matthew Wilcox 已提交
1788 1789 1790 1791 1792 1793
	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
1794
}
1795 1796 1797 1798 1799 1800

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
1801
EXPORT_SYMBOL(vmf_insert_mixed);
N
Nick Piggin 已提交
1802

1803 1804 1805 1806 1807 1808 1809
/*
 *  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
1810
{
1811
	return __vm_insert_mixed(vma, addr, pfn, true);
R
Ross Zwisler 已提交
1812
}
1813
EXPORT_SYMBOL(vmf_insert_mixed_mkwrite);
R
Ross Zwisler 已提交
1814

L
Linus Torvalds 已提交
1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
/*
 * 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
1825
	spinlock_t *ptl;
1826
	int err = 0;
L
Linus Torvalds 已提交
1827

H
Hugh Dickins 已提交
1828
	pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
L
Linus Torvalds 已提交
1829 1830
	if (!pte)
		return -ENOMEM;
1831
	arch_enter_lazy_mmu_mode();
L
Linus Torvalds 已提交
1832 1833
	do {
		BUG_ON(!pte_none(*pte));
1834 1835 1836 1837
		if (!pfn_modify_allowed(pfn, prot)) {
			err = -EACCES;
			break;
		}
N
Nick Piggin 已提交
1838
		set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
L
Linus Torvalds 已提交
1839 1840
		pfn++;
	} while (pte++, addr += PAGE_SIZE, addr != end);
1841
	arch_leave_lazy_mmu_mode();
H
Hugh Dickins 已提交
1842
	pte_unmap_unlock(pte - 1, ptl);
1843
	return err;
L
Linus Torvalds 已提交
1844 1845 1846 1847 1848 1849 1850 1851
}

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;
1852
	int err;
L
Linus Torvalds 已提交
1853 1854 1855 1856 1857

	pfn -= addr >> PAGE_SHIFT;
	pmd = pmd_alloc(mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
1858
	VM_BUG_ON(pmd_trans_huge(*pmd));
L
Linus Torvalds 已提交
1859 1860
	do {
		next = pmd_addr_end(addr, end);
1861 1862 1863 1864
		err = remap_pte_range(mm, pmd, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1865 1866 1867 1868
	} while (pmd++, addr = next, addr != end);
	return 0;
}

1869
static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
L
Linus Torvalds 已提交
1870 1871 1872 1873 1874
			unsigned long addr, unsigned long end,
			unsigned long pfn, pgprot_t prot)
{
	pud_t *pud;
	unsigned long next;
1875
	int err;
L
Linus Torvalds 已提交
1876 1877

	pfn -= addr >> PAGE_SHIFT;
1878
	pud = pud_alloc(mm, p4d, addr);
L
Linus Torvalds 已提交
1879 1880 1881 1882
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
1883 1884 1885 1886
		err = remap_pmd_range(mm, pud, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
L
Linus Torvalds 已提交
1887 1888 1889 1890
	} while (pud++, addr = next, addr != end);
	return 0;
}

1891 1892 1893 1894 1895 1896
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;
1897
	int err;
1898 1899 1900 1901 1902 1903 1904

	pfn -= addr >> PAGE_SHIFT;
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
1905 1906 1907 1908
		err = remap_pud_range(mm, p4d, addr, next,
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			return err;
1909 1910 1911 1912
	} while (p4d++, addr = next, addr != end);
	return 0;
}

1913 1914 1915 1916 1917 1918 1919 1920
/**
 * 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
 *
1921 1922 1923
 * Note: this is only safe if the mm semaphore is held when called.
 *
 * Return: %0 on success, negative error code otherwise.
1924
 */
L
Linus Torvalds 已提交
1925 1926 1927 1928 1929
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;
1930
	unsigned long end = addr + PAGE_ALIGN(size);
L
Linus Torvalds 已提交
1931
	struct mm_struct *mm = vma->vm_mm;
1932
	unsigned long remap_pfn = pfn;
L
Linus Torvalds 已提交
1933 1934 1935 1936 1937 1938 1939
	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).
1940 1941 1942
	 *   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.
1943 1944 1945 1946
	 *   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
1947 1948 1949 1950
	 *
	 * 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".
1951
	 * See vm_normal_page() for details.
L
Linus Torvalds 已提交
1952
	 */
1953 1954 1955
	if (is_cow_mapping(vma->vm_flags)) {
		if (addr != vma->vm_start || end != vma->vm_end)
			return -EINVAL;
L
Linus Torvalds 已提交
1956
		vma->vm_pgoff = pfn;
1957 1958
	}

1959
	err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
1960
	if (err)
1961
		return -EINVAL;
L
Linus Torvalds 已提交
1962

1963
	vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
L
Linus Torvalds 已提交
1964 1965 1966 1967 1968 1969 1970

	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);
1971
		err = remap_p4d_range(mm, pgd, addr, next,
L
Linus Torvalds 已提交
1972 1973 1974 1975
				pfn + (addr >> PAGE_SHIFT), prot);
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
1976 1977

	if (err)
1978
		untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
1979

L
Linus Torvalds 已提交
1980 1981 1982 1983
	return err;
}
EXPORT_SYMBOL(remap_pfn_range);

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
/**
 * 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.
1996 1997
 *
 * Return: %0 on success, negative error code otherwise.
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
 */
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);

2033 2034 2035 2036 2037 2038
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;
2039
	spinlock_t *uninitialized_var(ptl);
2040 2041 2042 2043 2044 2045 2046 2047 2048

	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));

2049 2050
	arch_enter_lazy_mmu_mode();

2051
	do {
2052
		err = fn(pte++, addr, data);
2053 2054
		if (err)
			break;
2055
	} while (addr += PAGE_SIZE, addr != end);
2056

2057 2058
	arch_leave_lazy_mmu_mode();

2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
	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
2072 2073
	BUG_ON(pud_huge(*pud));

2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085
	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;
}

2086
static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
2087 2088 2089 2090 2091 2092 2093
				     unsigned long addr, unsigned long end,
				     pte_fn_t fn, void *data)
{
	pud_t *pud;
	unsigned long next;
	int err;

2094
	pud = pud_alloc(mm, p4d, addr);
2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105
	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;
}

2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
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;
}

2126 2127 2128 2129 2130 2131 2132 2133 2134
/*
 * 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;
2135
	unsigned long end = addr + size;
2136 2137
	int err;

2138 2139 2140
	if (WARN_ON(addr >= end))
		return -EINVAL;

2141 2142 2143
	pgd = pgd_offset(mm, addr);
	do {
		next = pgd_addr_end(addr, end);
2144
		err = apply_to_p4d_range(mm, pgd, addr, next, fn, data);
2145 2146 2147
		if (err)
			break;
	} while (pgd++, addr = next, addr != end);
2148

2149 2150 2151 2152
	return err;
}
EXPORT_SYMBOL_GPL(apply_to_page_range);

2153
/*
2154 2155 2156 2157 2158
 * 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;
2159
 * and do_anonymous_page can safely check later on).
2160
 */
H
Hugh Dickins 已提交
2161
static inline int pte_unmap_same(struct mm_struct *mm, pmd_t *pmd,
2162 2163 2164 2165 2166
				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
2167 2168
		spinlock_t *ptl = pte_lockptr(mm, pmd);
		spin_lock(ptl);
2169
		same = pte_same(*page_table, orig_pte);
H
Hugh Dickins 已提交
2170
		spin_unlock(ptl);
2171 2172 2173 2174 2175 2176
	}
#endif
	pte_unmap(page_table);
	return same;
}

2177
static inline void cow_user_page(struct page *dst, struct page *src, unsigned long va, struct vm_area_struct *vma)
2178
{
2179 2180
	debug_dma_assert_idle(src);

2181 2182 2183 2184 2185 2186 2187
	/*
	 * 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)) {
2188
		void *kaddr = kmap_atomic(dst);
L
Linus Torvalds 已提交
2189 2190 2191 2192 2193 2194 2195 2196 2197
		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))
2198
			clear_page(kaddr);
2199
		kunmap_atomic(kaddr);
2200
		flush_dcache_page(dst);
N
Nick Piggin 已提交
2201 2202
	} else
		copy_user_highpage(dst, src, va, vma);
2203 2204
}

2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218
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;
}

2219 2220 2221 2222 2223 2224
/*
 * 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.
 */
2225
static vm_fault_t do_page_mkwrite(struct vm_fault *vmf)
2226
{
2227
	vm_fault_t ret;
2228 2229
	struct page *page = vmf->page;
	unsigned int old_flags = vmf->flags;
2230

2231
	vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
2232

2233
	ret = vmf->vma->vm_ops->page_mkwrite(vmf);
2234 2235
	/* Restore original flags so that caller is not surprised */
	vmf->flags = old_flags;
2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249
	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;
}

2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
/*
 * 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);
}

2285 2286 2287 2288 2289 2290 2291 2292
/*
 * 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.
 */
2293
static inline void wp_page_reuse(struct vm_fault *vmf)
J
Jan Kara 已提交
2294
	__releases(vmf->ptl)
2295
{
J
Jan Kara 已提交
2296
	struct vm_area_struct *vma = vmf->vma;
J
Jan Kara 已提交
2297
	struct page *page = vmf->page;
2298 2299 2300 2301 2302 2303 2304 2305 2306
	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
2307 2308
	flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
	entry = pte_mkyoung(vmf->orig_pte);
2309
	entry = maybe_mkwrite(pte_mkdirty(entry), vma);
J
Jan Kara 已提交
2310 2311 2312
	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);
2313 2314
}

2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
/*
 * 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.
 */
2331
static vm_fault_t wp_page_copy(struct vm_fault *vmf)
2332
{
J
Jan Kara 已提交
2333
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2334
	struct mm_struct *mm = vma->vm_mm;
J
Jan Kara 已提交
2335
	struct page *old_page = vmf->page;
2336 2337 2338 2339
	struct page *new_page = NULL;
	pte_t entry;
	int page_copied = 0;
	struct mem_cgroup *memcg;
2340
	struct mmu_notifier_range range;
2341 2342 2343 2344

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

J
Jan Kara 已提交
2345
	if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
J
Jan Kara 已提交
2346 2347
		new_page = alloc_zeroed_user_highpage_movable(vma,
							      vmf->address);
2348 2349 2350
		if (!new_page)
			goto oom;
	} else {
K
Kirill A. Shutemov 已提交
2351
		new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
J
Jan Kara 已提交
2352
				vmf->address);
2353 2354
		if (!new_page)
			goto oom;
J
Jan Kara 已提交
2355
		cow_user_page(new_page, old_page, vmf->address, vma);
2356 2357
	}

2358
	if (mem_cgroup_try_charge_delay(new_page, mm, GFP_KERNEL, &memcg, false))
2359 2360
		goto oom_free_new;

2361 2362
	__SetPageUptodate(new_page);

2363
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm,
2364
				vmf->address & PAGE_MASK,
2365 2366
				(vmf->address & PAGE_MASK) + PAGE_SIZE);
	mmu_notifier_invalidate_range_start(&range);
2367 2368 2369 2370

	/*
	 * Re-check the pte - we dropped the lock
	 */
J
Jan Kara 已提交
2371
	vmf->pte = pte_offset_map_lock(mm, vmf->pmd, vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2372
	if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
2373 2374
		if (old_page) {
			if (!PageAnon(old_page)) {
2375 2376
				dec_mm_counter_fast(mm,
						mm_counter_file(old_page));
2377 2378 2379 2380 2381
				inc_mm_counter_fast(mm, MM_ANONPAGES);
			}
		} else {
			inc_mm_counter_fast(mm, MM_ANONPAGES);
		}
J
Jan Kara 已提交
2382
		flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
2383 2384 2385 2386 2387 2388 2389 2390
		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
2391 2392
		ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
		page_add_new_anon_rmap(new_page, vma, vmf->address, false);
2393
		mem_cgroup_commit_charge(new_page, memcg, false, false);
2394 2395 2396 2397 2398 2399
		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
2400 2401
		set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
		update_mmu_cache(vma, vmf->address, vmf->pte);
2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424
		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.
			 */
2425
			page_remove_rmap(old_page, false);
2426 2427 2428 2429 2430 2431
		}

		/* Free the old page.. */
		new_page = old_page;
		page_copied = 1;
	} else {
2432
		mem_cgroup_cancel_charge(new_page, memcg, false);
2433 2434 2435
	}

	if (new_page)
2436
		put_page(new_page);
2437

J
Jan Kara 已提交
2438
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2439 2440 2441 2442
	/*
	 * No need to double call mmu_notifier->invalidate_range() callback as
	 * the above ptep_clear_flush_notify() did already call it.
	 */
2443
	mmu_notifier_invalidate_range_only_end(&range);
2444 2445 2446 2447 2448 2449 2450
	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 */
2451 2452
			if (PageMlocked(old_page))
				munlock_vma_page(old_page);
2453 2454
			unlock_page(old_page);
		}
2455
		put_page(old_page);
2456 2457 2458
	}
	return page_copied ? VM_FAULT_WRITE : 0;
oom_free_new:
2459
	put_page(new_page);
2460 2461
oom:
	if (old_page)
2462
		put_page(old_page);
2463 2464 2465
	return VM_FAULT_OOM;
}

2466 2467 2468 2469 2470 2471 2472 2473
/**
 * 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.
2474
 * It handles locking of PTE and modifying it.
2475 2476 2477
 *
 * The function expects the page to be locked or other protection against
 * concurrent faults / writeback (such as DAX radix tree locks).
2478 2479 2480
 *
 * Return: %VM_FAULT_WRITE on success, %0 when PTE got changed before
 * we acquired PTE lock.
2481
 */
2482
vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf)
2483 2484 2485 2486 2487 2488 2489 2490 2491 2492
{
	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);
2493
		return VM_FAULT_NOPAGE;
2494 2495
	}
	wp_page_reuse(vmf);
2496
	return 0;
2497 2498
}

2499 2500 2501 2502
/*
 * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
 * mapping
 */
2503
static vm_fault_t wp_pfn_shared(struct vm_fault *vmf)
2504
{
J
Jan Kara 已提交
2505
	struct vm_area_struct *vma = vmf->vma;
K
Kirill A. Shutemov 已提交
2506

2507
	if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
2508
		vm_fault_t ret;
2509

J
Jan Kara 已提交
2510
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2511
		vmf->flags |= FAULT_FLAG_MKWRITE;
2512
		ret = vma->vm_ops->pfn_mkwrite(vmf);
2513
		if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
2514
			return ret;
2515
		return finish_mkwrite_fault(vmf);
2516
	}
2517 2518
	wp_page_reuse(vmf);
	return VM_FAULT_WRITE;
2519 2520
}

2521
static vm_fault_t wp_page_shared(struct vm_fault *vmf)
J
Jan Kara 已提交
2522
	__releases(vmf->ptl)
2523
{
J
Jan Kara 已提交
2524
	struct vm_area_struct *vma = vmf->vma;
2525

J
Jan Kara 已提交
2526
	get_page(vmf->page);
2527 2528

	if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
2529
		vm_fault_t tmp;
2530

J
Jan Kara 已提交
2531
		pte_unmap_unlock(vmf->pte, vmf->ptl);
2532
		tmp = do_page_mkwrite(vmf);
2533 2534
		if (unlikely(!tmp || (tmp &
				      (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
2535
			put_page(vmf->page);
2536 2537
			return tmp;
		}
2538
		tmp = finish_mkwrite_fault(vmf);
2539
		if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
J
Jan Kara 已提交
2540 2541
			unlock_page(vmf->page);
			put_page(vmf->page);
2542
			return tmp;
2543
		}
2544 2545
	} else {
		wp_page_reuse(vmf);
2546
		lock_page(vmf->page);
2547
	}
2548 2549
	fault_dirty_shared_page(vma, vmf->page);
	put_page(vmf->page);
2550

2551
	return VM_FAULT_WRITE;
2552 2553
}

L
Linus Torvalds 已提交
2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567
/*
 * 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.
 *
2568 2569 2570
 * 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
2571
 */
2572
static vm_fault_t do_wp_page(struct vm_fault *vmf)
J
Jan Kara 已提交
2573
	__releases(vmf->ptl)
L
Linus Torvalds 已提交
2574
{
J
Jan Kara 已提交
2575
	struct vm_area_struct *vma = vmf->vma;
L
Linus Torvalds 已提交
2576

J
Jan Kara 已提交
2577 2578
	vmf->page = vm_normal_page(vma, vmf->address, vmf->orig_pte);
	if (!vmf->page) {
2579
		/*
2580 2581
		 * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
		 * VM_PFNMAP VMA.
2582 2583
		 *
		 * We should not cow pages in a shared writeable mapping.
2584
		 * Just mark the pages writable and/or call ops->pfn_mkwrite.
2585 2586 2587
		 */
		if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
				     (VM_WRITE|VM_SHARED))
J
Jan Kara 已提交
2588
			return wp_pfn_shared(vmf);
2589

J
Jan Kara 已提交
2590
		pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2591
		return wp_page_copy(vmf);
2592
	}
L
Linus Torvalds 已提交
2593

2594
	/*
P
Peter Zijlstra 已提交
2595 2596
	 * Take out anonymous pages first, anonymous shared vmas are
	 * not dirty accountable.
2597
	 */
2598
	if (PageAnon(vmf->page)) {
2599
		int total_map_swapcount;
2600 2601 2602
		if (PageKsm(vmf->page) && (PageSwapCache(vmf->page) ||
					   page_count(vmf->page) != 1))
			goto copy;
J
Jan Kara 已提交
2603 2604
		if (!trylock_page(vmf->page)) {
			get_page(vmf->page);
J
Jan Kara 已提交
2605
			pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2606
			lock_page(vmf->page);
J
Jan Kara 已提交
2607 2608
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2609
			if (!pte_same(*vmf->pte, vmf->orig_pte)) {
J
Jan Kara 已提交
2610
				unlock_page(vmf->page);
J
Jan Kara 已提交
2611
				pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2612
				put_page(vmf->page);
2613
				return 0;
2614
			}
J
Jan Kara 已提交
2615
			put_page(vmf->page);
P
Peter Zijlstra 已提交
2616
		}
2617 2618 2619 2620 2621 2622 2623 2624 2625
		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;
		}
2626 2627
		if (reuse_swap_page(vmf->page, &total_map_swapcount)) {
			if (total_map_swapcount == 1) {
2628 2629 2630 2631 2632 2633 2634
				/*
				 * 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
2635
				page_move_anon_rmap(vmf->page, vma);
2636
			}
J
Jan Kara 已提交
2637
			unlock_page(vmf->page);
2638 2639
			wp_page_reuse(vmf);
			return VM_FAULT_WRITE;
2640
		}
J
Jan Kara 已提交
2641
		unlock_page(vmf->page);
P
Peter Zijlstra 已提交
2642
	} else if (unlikely((vma->vm_flags & (VM_WRITE|VM_SHARED)) ==
2643
					(VM_WRITE|VM_SHARED))) {
J
Jan Kara 已提交
2644
		return wp_page_shared(vmf);
L
Linus Torvalds 已提交
2645
	}
2646
copy:
L
Linus Torvalds 已提交
2647 2648 2649
	/*
	 * Ok, we need to copy. Oh, well..
	 */
J
Jan Kara 已提交
2650
	get_page(vmf->page);
2651

J
Jan Kara 已提交
2652
	pte_unmap_unlock(vmf->pte, vmf->ptl);
J
Jan Kara 已提交
2653
	return wp_page_copy(vmf);
L
Linus Torvalds 已提交
2654 2655
}

2656
static void unmap_mapping_range_vma(struct vm_area_struct *vma,
L
Linus Torvalds 已提交
2657 2658 2659
		unsigned long start_addr, unsigned long end_addr,
		struct zap_details *details)
{
2660
	zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
L
Linus Torvalds 已提交
2661 2662
}

2663
static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
L
Linus Torvalds 已提交
2664 2665 2666 2667 2668
					    struct zap_details *details)
{
	struct vm_area_struct *vma;
	pgoff_t vba, vea, zba, zea;

2669
	vma_interval_tree_foreach(vma, root,
L
Linus Torvalds 已提交
2670 2671 2672
			details->first_index, details->last_index) {

		vba = vma->vm_pgoff;
2673
		vea = vba + vma_pages(vma) - 1;
L
Linus Torvalds 已提交
2674 2675 2676 2677 2678 2679 2680
		zba = details->first_index;
		if (zba < vba)
			zba = vba;
		zea = details->last_index;
		if (zea > vea)
			zea = vea;

2681
		unmap_mapping_range_vma(vma,
L
Linus Torvalds 已提交
2682 2683
			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
2684
				details);
L
Linus Torvalds 已提交
2685 2686 2687
	}
}

M
Matthew Wilcox 已提交
2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716
/**
 * 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
2717
/**
2718
 * unmap_mapping_range - unmap the portion of all mmaps in the specified
M
Matthew Wilcox 已提交
2719
 * address_space corresponding to the specified byte range in the underlying
2720 2721
 * file.
 *
M
Martin Waitz 已提交
2722
 * @mapping: the address space containing mmaps to be unmapped.
L
Linus Torvalds 已提交
2723 2724
 * @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
2725
 * boundary.  Note that this is different from truncate_pagecache(), which
L
Linus Torvalds 已提交
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
 * 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
2748
	unmap_mapping_pages(mapping, hba, hlen, even_cows);
L
Linus Torvalds 已提交
2749 2750 2751 2752
}
EXPORT_SYMBOL(unmap_mapping_range);

/*
2753 2754
 * We enter with non-exclusive mmap_sem (to exclude vma changes,
 * but allow concurrent faults), and pte mapped but not yet locked.
2755 2756 2757 2758
 * 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
2759
 */
2760
vm_fault_t do_swap_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2761
{
J
Jan Kara 已提交
2762
	struct vm_area_struct *vma = vmf->vma;
M
Minchan Kim 已提交
2763
	struct page *page = NULL, *swapcache;
2764
	struct mem_cgroup *memcg;
2765
	swp_entry_t entry;
L
Linus Torvalds 已提交
2766
	pte_t pte;
2767
	int locked;
2768
	int exclusive = 0;
2769
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
2770

M
Minchan Kim 已提交
2771
	if (!pte_unmap_same(vma->vm_mm, vmf->pmd, vmf->pte, vmf->orig_pte))
2772
		goto out;
2773

J
Jan Kara 已提交
2774
	entry = pte_to_swp_entry(vmf->orig_pte);
2775 2776
	if (unlikely(non_swap_entry(entry))) {
		if (is_migration_entry(entry)) {
J
Jan Kara 已提交
2777 2778
			migration_entry_wait(vma->vm_mm, vmf->pmd,
					     vmf->address);
2779 2780 2781 2782 2783 2784 2785 2786
		} 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);
2787 2788 2789
		} else if (is_hwpoison_entry(entry)) {
			ret = VM_FAULT_HWPOISON;
		} else {
J
Jan Kara 已提交
2790
			print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
H
Hugh Dickins 已提交
2791
			ret = VM_FAULT_SIGBUS;
2792
		}
2793 2794
		goto out;
	}
2795 2796


2797
	delayacct_set_flag(DELAYACCT_PF_SWAPIN);
M
Minchan Kim 已提交
2798 2799
	page = lookup_swap_cache(entry, vma, vmf->address);
	swapcache = page;
2800

L
Linus Torvalds 已提交
2801
	if (!page) {
2802 2803
		struct swap_info_struct *si = swp_swap_info(entry);

2804
		if (si->flags & SWP_SYNCHRONOUS_IO &&
2805
				__swap_count(entry) == 1) {
2806
			/* skip swapcache */
2807 2808
			page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
							vmf->address);
2809 2810 2811 2812 2813 2814 2815
			if (page) {
				__SetPageLocked(page);
				__SetPageSwapBacked(page);
				set_page_private(page, entry.val);
				lru_cache_add_anon(page);
				swap_readpage(page, true);
			}
2816
		} else {
2817 2818
			page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE,
						vmf);
2819
			swapcache = page;
2820 2821
		}

L
Linus Torvalds 已提交
2822 2823
		if (!page) {
			/*
2824 2825
			 * Back out if somebody else faulted in this pte
			 * while we released the pte lock.
L
Linus Torvalds 已提交
2826
			 */
J
Jan Kara 已提交
2827 2828
			vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
					vmf->address, &vmf->ptl);
J
Jan Kara 已提交
2829
			if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
L
Linus Torvalds 已提交
2830
				ret = VM_FAULT_OOM;
2831
			delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2832
			goto unlock;
L
Linus Torvalds 已提交
2833 2834 2835 2836
		}

		/* Had to read the page from swap area: Major fault */
		ret = VM_FAULT_MAJOR;
2837
		count_vm_event(PGMAJFAULT);
2838
		count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
2839
	} else if (PageHWPoison(page)) {
2840 2841 2842 2843
		/*
		 * hwpoisoned dirty swapcache pages are kept for killing
		 * owner processes (which may be unknown at hwpoison time)
		 */
2844 2845
		ret = VM_FAULT_HWPOISON;
		delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2846
		goto out_release;
L
Linus Torvalds 已提交
2847 2848
	}

J
Jan Kara 已提交
2849
	locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
R
Rik van Riel 已提交
2850

2851
	delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
2852 2853 2854 2855
	if (!locked) {
		ret |= VM_FAULT_RETRY;
		goto out_release;
	}
2856

A
Andrea Arcangeli 已提交
2857
	/*
2858 2859 2860 2861
	 * 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
2862
	 */
2863 2864
	if (unlikely((!PageSwapCache(page) ||
			page_private(page) != entry.val)) && swapcache)
A
Andrea Arcangeli 已提交
2865 2866
		goto out_page;

J
Jan Kara 已提交
2867
	page = ksm_might_need_to_copy(page, vma, vmf->address);
2868 2869 2870 2871
	if (unlikely(!page)) {
		ret = VM_FAULT_OOM;
		page = swapcache;
		goto out_page;
H
Hugh Dickins 已提交
2872 2873
	}

2874 2875
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL,
					&memcg, false)) {
2876
		ret = VM_FAULT_OOM;
2877
		goto out_page;
2878 2879
	}

L
Linus Torvalds 已提交
2880
	/*
2881
	 * Back out if somebody else already faulted in this pte.
L
Linus Torvalds 已提交
2882
	 */
J
Jan Kara 已提交
2883 2884
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
J
Jan Kara 已提交
2885
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
2886 2887 2888 2889 2890
		goto out_nomap;

	if (unlikely(!PageUptodate(page))) {
		ret = VM_FAULT_SIGBUS;
		goto out_nomap;
L
Linus Torvalds 已提交
2891 2892
	}

2893 2894 2895 2896 2897 2898 2899 2900 2901
	/*
	 * 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
2902

K
Kirill A. Shutemov 已提交
2903 2904
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
	dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
L
Linus Torvalds 已提交
2905
	pte = mk_pte(page, vma->vm_page_prot);
J
Jan Kara 已提交
2906
	if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
L
Linus Torvalds 已提交
2907
		pte = maybe_mkwrite(pte_mkdirty(pte), vma);
J
Jan Kara 已提交
2908
		vmf->flags &= ~FAULT_FLAG_WRITE;
2909
		ret |= VM_FAULT_WRITE;
2910
		exclusive = RMAP_EXCLUSIVE;
L
Linus Torvalds 已提交
2911 2912
	}
	flush_icache_page(vma, page);
J
Jan Kara 已提交
2913
	if (pte_swp_soft_dirty(vmf->orig_pte))
2914
		pte = pte_mksoft_dirty(pte);
J
Jan Kara 已提交
2915
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
2916
	arch_do_swap_page(vma->vm_mm, vma, vmf->address, pte, vmf->orig_pte);
J
Jan Kara 已提交
2917
	vmf->orig_pte = pte;
2918 2919 2920

	/* ksm created a completely new copy */
	if (unlikely(page != swapcache && swapcache)) {
J
Jan Kara 已提交
2921
		page_add_new_anon_rmap(page, vma, vmf->address, false);
2922
		mem_cgroup_commit_charge(page, memcg, false, false);
2923
		lru_cache_add_active_or_unevictable(page, vma);
2924 2925 2926 2927
	} else {
		do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
		mem_cgroup_commit_charge(page, memcg, true, false);
		activate_page(page);
2928
	}
L
Linus Torvalds 已提交
2929

2930
	swap_free(entry);
2931 2932
	if (mem_cgroup_swap_full(page) ||
	    (vma->vm_flags & VM_LOCKED) || PageMlocked(page))
2933
		try_to_free_swap(page);
2934
	unlock_page(page);
2935
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
2936 2937 2938 2939 2940 2941 2942 2943 2944
		/*
		 * 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);
2945
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2946
	}
2947

J
Jan Kara 已提交
2948
	if (vmf->flags & FAULT_FLAG_WRITE) {
J
Jan Kara 已提交
2949
		ret |= do_wp_page(vmf);
2950 2951
		if (ret & VM_FAULT_ERROR)
			ret &= VM_FAULT_ERROR;
L
Linus Torvalds 已提交
2952 2953 2954 2955
		goto out;
	}

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
2956
	update_mmu_cache(vma, vmf->address, vmf->pte);
2957
unlock:
J
Jan Kara 已提交
2958
	pte_unmap_unlock(vmf->pte, vmf->ptl);
L
Linus Torvalds 已提交
2959 2960
out:
	return ret;
2961
out_nomap:
2962
	mem_cgroup_cancel_charge(page, memcg, false);
J
Jan Kara 已提交
2963
	pte_unmap_unlock(vmf->pte, vmf->ptl);
2964
out_page:
2965
	unlock_page(page);
2966
out_release:
2967
	put_page(page);
2968
	if (page != swapcache && swapcache) {
A
Andrea Arcangeli 已提交
2969
		unlock_page(swapcache);
2970
		put_page(swapcache);
A
Andrea Arcangeli 已提交
2971
	}
2972
	return ret;
L
Linus Torvalds 已提交
2973 2974 2975
}

/*
2976 2977 2978
 * 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
2979
 */
2980
static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
L
Linus Torvalds 已提交
2981
{
J
Jan Kara 已提交
2982
	struct vm_area_struct *vma = vmf->vma;
2983
	struct mem_cgroup *memcg;
2984
	struct page *page;
2985
	vm_fault_t ret = 0;
L
Linus Torvalds 已提交
2986 2987
	pte_t entry;

2988 2989 2990 2991
	/* File mapping without ->vm_ops ? */
	if (vma->vm_flags & VM_SHARED)
		return VM_FAULT_SIGBUS;

2992 2993 2994 2995 2996 2997 2998 2999 3000 3001
	/*
	 * 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).
	 */
3002
	if (pte_alloc(vma->vm_mm, vmf->pmd))
3003 3004 3005
		return VM_FAULT_OOM;

	/* See the comment in pte_alloc_one_map() */
J
Jan Kara 已提交
3006
	if (unlikely(pmd_trans_unstable(vmf->pmd)))
3007 3008
		return 0;

3009
	/* Use the zero-page for reads */
J
Jan Kara 已提交
3010
	if (!(vmf->flags & FAULT_FLAG_WRITE) &&
K
Kirill A. Shutemov 已提交
3011
			!mm_forbids_zeropage(vma->vm_mm)) {
J
Jan Kara 已提交
3012
		entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
H
Hugh Dickins 已提交
3013
						vma->vm_page_prot));
J
Jan Kara 已提交
3014 3015 3016
		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
3017
			goto unlock;
3018 3019 3020
		ret = check_stable_address_space(vma->vm_mm);
		if (ret)
			goto unlock;
3021 3022
		/* Deliver the page fault to userland, check inside PT lock */
		if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3023 3024
			pte_unmap_unlock(vmf->pte, vmf->ptl);
			return handle_userfault(vmf, VM_UFFD_MISSING);
3025
		}
H
Hugh Dickins 已提交
3026 3027 3028
		goto setpte;
	}

N
Nick Piggin 已提交
3029 3030 3031
	/* Allocate our own private page. */
	if (unlikely(anon_vma_prepare(vma)))
		goto oom;
J
Jan Kara 已提交
3032
	page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
N
Nick Piggin 已提交
3033 3034
	if (!page)
		goto oom;
3035

3036 3037
	if (mem_cgroup_try_charge_delay(page, vma->vm_mm, GFP_KERNEL, &memcg,
					false))
3038 3039
		goto oom_free_page;

3040 3041 3042 3043 3044
	/*
	 * 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
3045
	__SetPageUptodate(page);
3046

N
Nick Piggin 已提交
3047
	entry = mk_pte(page, vma->vm_page_prot);
H
Hugh Dickins 已提交
3048 3049
	if (vma->vm_flags & VM_WRITE)
		entry = pte_mkwrite(pte_mkdirty(entry));
L
Linus Torvalds 已提交
3050

J
Jan Kara 已提交
3051 3052 3053
	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
3054
		goto release;
H
Hugh Dickins 已提交
3055

3056 3057 3058 3059
	ret = check_stable_address_space(vma->vm_mm);
	if (ret)
		goto release;

3060 3061
	/* Deliver the page fault to userland, check inside PT lock */
	if (userfaultfd_missing(vma)) {
J
Jan Kara 已提交
3062
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3063
		mem_cgroup_cancel_charge(page, memcg, false);
3064
		put_page(page);
J
Jan Kara 已提交
3065
		return handle_userfault(vmf, VM_UFFD_MISSING);
3066 3067
	}

K
Kirill A. Shutemov 已提交
3068
	inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3069
	page_add_new_anon_rmap(page, vma, vmf->address, false);
3070
	mem_cgroup_commit_charge(page, memcg, false, false);
3071
	lru_cache_add_active_or_unevictable(page, vma);
H
Hugh Dickins 已提交
3072
setpte:
J
Jan Kara 已提交
3073
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
L
Linus Torvalds 已提交
3074 3075

	/* No need to invalidate - it was non-present before */
J
Jan Kara 已提交
3076
	update_mmu_cache(vma, vmf->address, vmf->pte);
3077
unlock:
J
Jan Kara 已提交
3078
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3079
	return ret;
3080
release:
3081
	mem_cgroup_cancel_charge(page, memcg, false);
3082
	put_page(page);
3083
	goto unlock;
3084
oom_free_page:
3085
	put_page(page);
3086
oom:
L
Linus Torvalds 已提交
3087 3088 3089
	return VM_FAULT_OOM;
}

3090 3091 3092 3093 3094
/*
 * 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().
 */
3095
static vm_fault_t __do_fault(struct vm_fault *vmf)
3096
{
J
Jan Kara 已提交
3097
	struct vm_area_struct *vma = vmf->vma;
3098
	vm_fault_t ret;
3099

3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121
	/*
	 * 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() */
	}

3122
	ret = vma->vm_ops->fault(vmf);
3123
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
3124
			    VM_FAULT_DONE_COW)))
3125
		return ret;
3126

3127
	if (unlikely(PageHWPoison(vmf->page))) {
3128
		if (ret & VM_FAULT_LOCKED)
3129 3130
			unlock_page(vmf->page);
		put_page(vmf->page);
J
Jan Kara 已提交
3131
		vmf->page = NULL;
3132 3133 3134 3135
		return VM_FAULT_HWPOISON;
	}

	if (unlikely(!(ret & VM_FAULT_LOCKED)))
3136
		lock_page(vmf->page);
3137
	else
3138
		VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
3139 3140 3141 3142

	return ret;
}

3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153
/*
 * 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);
}

3154
static vm_fault_t pte_alloc_one_map(struct vm_fault *vmf)
3155
{
J
Jan Kara 已提交
3156
	struct vm_area_struct *vma = vmf->vma;
3157

J
Jan Kara 已提交
3158
	if (!pmd_none(*vmf->pmd))
3159
		goto map_pte;
J
Jan Kara 已提交
3160 3161 3162 3163
	if (vmf->prealloc_pte) {
		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
		if (unlikely(!pmd_none(*vmf->pmd))) {
			spin_unlock(vmf->ptl);
3164 3165 3166
			goto map_pte;
		}

3167
		mm_inc_nr_ptes(vma->vm_mm);
J
Jan Kara 已提交
3168 3169
		pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
		spin_unlock(vmf->ptl);
3170
		vmf->prealloc_pte = NULL;
3171
	} else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd))) {
3172 3173 3174 3175 3176
		return VM_FAULT_OOM;
	}
map_pte:
	/*
	 * If a huge pmd materialized under us just retry later.  Use
3177 3178 3179 3180 3181 3182 3183 3184
	 * 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.
3185
	 */
3186
	if (pmd_devmap_trans_unstable(vmf->pmd))
3187 3188
		return VM_FAULT_NOPAGE;

3189 3190 3191 3192 3193 3194 3195 3196 3197
	/*
	 * 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
3198 3199
	vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, vmf->address,
			&vmf->ptl);
3200 3201 3202
	return 0;
}

3203
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
K
Kirill A. Shutemov 已提交
3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216

#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
3217
static void deposit_prealloc_pte(struct vm_fault *vmf)
3218
{
J
Jan Kara 已提交
3219
	struct vm_area_struct *vma = vmf->vma;
3220

J
Jan Kara 已提交
3221
	pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
3222 3223 3224 3225
	/*
	 * We are going to consume the prealloc table,
	 * count that as nr_ptes.
	 */
3226
	mm_inc_nr_ptes(vma->vm_mm);
3227
	vmf->prealloc_pte = NULL;
3228 3229
}

3230
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3231
{
J
Jan Kara 已提交
3232 3233 3234
	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
3235
	pmd_t entry;
3236 3237
	int i;
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3238 3239 3240 3241 3242 3243 3244

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

	ret = VM_FAULT_FALLBACK;
	page = compound_head(page);

3245 3246 3247 3248
	/*
	 * 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
3249
	if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
3250
		vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
J
Jan Kara 已提交
3251
		if (!vmf->prealloc_pte)
3252 3253 3254 3255
			return VM_FAULT_OOM;
		smp_wmb(); /* See comment in __pte_alloc() */
	}

J
Jan Kara 已提交
3256 3257
	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
	if (unlikely(!pmd_none(*vmf->pmd)))
K
Kirill A. Shutemov 已提交
3258 3259 3260 3261 3262 3263 3264
		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)
3265
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
K
Kirill A. Shutemov 已提交
3266

3267
	add_mm_counter(vma->vm_mm, mm_counter_file(page), HPAGE_PMD_NR);
K
Kirill A. Shutemov 已提交
3268
	page_add_file_rmap(page, true);
3269 3270 3271 3272
	/*
	 * deposit and withdraw with pmd lock held
	 */
	if (arch_needs_pgtable_deposit())
J
Jan Kara 已提交
3273
		deposit_prealloc_pte(vmf);
K
Kirill A. Shutemov 已提交
3274

J
Jan Kara 已提交
3275
	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
K
Kirill A. Shutemov 已提交
3276

J
Jan Kara 已提交
3277
	update_mmu_cache_pmd(vma, haddr, vmf->pmd);
K
Kirill A. Shutemov 已提交
3278 3279 3280

	/* fault is handled */
	ret = 0;
3281
	count_vm_event(THP_FILE_MAPPED);
K
Kirill A. Shutemov 已提交
3282
out:
J
Jan Kara 已提交
3283
	spin_unlock(vmf->ptl);
K
Kirill A. Shutemov 已提交
3284 3285 3286
	return ret;
}
#else
3287
static vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
K
Kirill A. Shutemov 已提交
3288 3289 3290 3291 3292 3293
{
	BUILD_BUG();
	return 0;
}
#endif

3294
/**
3295 3296
 * 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.
3297
 *
J
Jan Kara 已提交
3298
 * @vmf: fault environment
3299
 * @memcg: memcg to charge page (only for private mappings)
3300 3301
 * @page: page to map
 *
J
Jan Kara 已提交
3302 3303
 * Caller must take care of unlocking vmf->ptl, if vmf->pte is non-NULL on
 * return.
3304 3305 3306
 *
 * Target users are page handler itself and implementations of
 * vm_ops->map_pages.
3307 3308
 *
 * Return: %0 on success, %VM_FAULT_ code in case of error.
3309
 */
3310
vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
3311
		struct page *page)
3312
{
J
Jan Kara 已提交
3313 3314
	struct vm_area_struct *vma = vmf->vma;
	bool write = vmf->flags & FAULT_FLAG_WRITE;
3315
	pte_t entry;
3316
	vm_fault_t ret;
K
Kirill A. Shutemov 已提交
3317

J
Jan Kara 已提交
3318
	if (pmd_none(*vmf->pmd) && PageTransCompound(page) &&
3319
			IS_ENABLED(CONFIG_TRANSPARENT_HUGE_PAGECACHE)) {
K
Kirill A. Shutemov 已提交
3320 3321 3322
		/* THP on COW? */
		VM_BUG_ON_PAGE(memcg, page);

J
Jan Kara 已提交
3323
		ret = do_set_pmd(vmf, page);
K
Kirill A. Shutemov 已提交
3324
		if (ret != VM_FAULT_FALLBACK)
H
Hugh Dickins 已提交
3325
			return ret;
K
Kirill A. Shutemov 已提交
3326
	}
3327

J
Jan Kara 已提交
3328 3329
	if (!vmf->pte) {
		ret = pte_alloc_one_map(vmf);
3330
		if (ret)
H
Hugh Dickins 已提交
3331
			return ret;
3332 3333 3334
	}

	/* Re-check under ptl */
H
Hugh Dickins 已提交
3335 3336
	if (unlikely(!pte_none(*vmf->pte)))
		return VM_FAULT_NOPAGE;
3337

3338 3339 3340 3341
	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
3342 3343
	/* copy-on-write page */
	if (write && !(vma->vm_flags & VM_SHARED)) {
3344
		inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
J
Jan Kara 已提交
3345
		page_add_new_anon_rmap(page, vma, vmf->address, false);
3346 3347
		mem_cgroup_commit_charge(page, memcg, false, false);
		lru_cache_add_active_or_unevictable(page, vma);
3348
	} else {
3349
		inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
K
Kirill A. Shutemov 已提交
3350
		page_add_file_rmap(page, false);
3351
	}
J
Jan Kara 已提交
3352
	set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
3353 3354

	/* no need to invalidate: a not-present page won't be cached */
J
Jan Kara 已提交
3355
	update_mmu_cache(vma, vmf->address, vmf->pte);
3356

H
Hugh Dickins 已提交
3357
	return 0;
3358 3359
}

3360 3361 3362 3363 3364 3365 3366 3367 3368

/**
 * 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
3369
 * addition.
3370 3371 3372
 *
 * 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).
3373 3374
 *
 * Return: %0 on success, %VM_FAULT_ code in case of error.
3375
 */
3376
vm_fault_t finish_fault(struct vm_fault *vmf)
3377 3378
{
	struct page *page;
3379
	vm_fault_t ret = 0;
3380 3381 3382 3383 3384 3385 3386

	/* 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;
3387 3388 3389 3390 3391 3392 3393 3394 3395

	/*
	 * 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);
3396 3397 3398 3399 3400
	if (vmf->pte)
		pte_unmap_unlock(vmf->pte, vmf->ptl);
	return ret;
}

3401 3402
static unsigned long fault_around_bytes __read_mostly =
	rounddown_pow_of_two(65536);
3403 3404 3405

#ifdef CONFIG_DEBUG_FS
static int fault_around_bytes_get(void *data, u64 *val)
3406
{
3407
	*val = fault_around_bytes;
3408 3409 3410
	return 0;
}

3411
/*
3412 3413
 * fault_around_bytes must be rounded down to the nearest page order as it's
 * what do_fault_around() expects to see.
3414
 */
3415
static int fault_around_bytes_set(void *data, u64 val)
3416
{
3417
	if (val / PAGE_SIZE > PTRS_PER_PTE)
3418
		return -EINVAL;
3419 3420 3421 3422
	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 */
3423 3424
	return 0;
}
3425
DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
3426
		fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
3427 3428 3429

static int __init fault_around_debugfs(void)
{
3430 3431
	debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
				   &fault_around_bytes_fops);
3432 3433 3434 3435
	return 0;
}
late_initcall(fault_around_debugfs);
#endif
3436

3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451
/*
 * 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.
 *
3452 3453 3454
 * 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.
3455
 *
3456 3457 3458 3459
 * 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.
3460
 */
3461
static vm_fault_t do_fault_around(struct vm_fault *vmf)
3462
{
J
Jan Kara 已提交
3463
	unsigned long address = vmf->address, nr_pages, mask;
3464
	pgoff_t start_pgoff = vmf->pgoff;
K
Kirill A. Shutemov 已提交
3465
	pgoff_t end_pgoff;
3466 3467
	int off;
	vm_fault_t ret = 0;
3468

3469
	nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
3470 3471
	mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;

J
Jan Kara 已提交
3472 3473
	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
3474
	start_pgoff -= off;
3475 3476

	/*
3477 3478
	 *  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.
3479
	 */
K
Kirill A. Shutemov 已提交
3480
	end_pgoff = start_pgoff -
J
Jan Kara 已提交
3481
		((vmf->address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
3482
		PTRS_PER_PTE - 1;
J
Jan Kara 已提交
3483
	end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
K
Kirill A. Shutemov 已提交
3484
			start_pgoff + nr_pages - 1);
3485

J
Jan Kara 已提交
3486
	if (pmd_none(*vmf->pmd)) {
3487
		vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
J
Jan Kara 已提交
3488
		if (!vmf->prealloc_pte)
3489
			goto out;
3490
		smp_wmb(); /* See comment in __pte_alloc() */
3491 3492
	}

J
Jan Kara 已提交
3493
	vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
3494 3495

	/* Huge page is mapped? Page fault is solved */
J
Jan Kara 已提交
3496
	if (pmd_trans_huge(*vmf->pmd)) {
3497 3498 3499 3500 3501
		ret = VM_FAULT_NOPAGE;
		goto out;
	}

	/* ->map_pages() haven't done anything useful. Cold page cache? */
J
Jan Kara 已提交
3502
	if (!vmf->pte)
3503 3504 3505
		goto out;

	/* check if the page fault is solved */
J
Jan Kara 已提交
3506 3507
	vmf->pte -= (vmf->address >> PAGE_SHIFT) - (address >> PAGE_SHIFT);
	if (!pte_none(*vmf->pte))
3508
		ret = VM_FAULT_NOPAGE;
J
Jan Kara 已提交
3509
	pte_unmap_unlock(vmf->pte, vmf->ptl);
K
Kirill A. Shutemov 已提交
3510
out:
J
Jan Kara 已提交
3511 3512
	vmf->address = address;
	vmf->pte = NULL;
3513
	return ret;
3514 3515
}

3516
static vm_fault_t do_read_fault(struct vm_fault *vmf)
3517
{
J
Jan Kara 已提交
3518
	struct vm_area_struct *vma = vmf->vma;
3519
	vm_fault_t ret = 0;
3520 3521 3522 3523 3524 3525

	/*
	 * 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).
	 */
3526
	if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
3527
		ret = do_fault_around(vmf);
3528 3529
		if (ret)
			return ret;
3530
	}
3531

J
Jan Kara 已提交
3532
	ret = __do_fault(vmf);
3533 3534 3535
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		return ret;

3536
	ret |= finish_fault(vmf);
J
Jan Kara 已提交
3537
	unlock_page(vmf->page);
3538
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
J
Jan Kara 已提交
3539
		put_page(vmf->page);
3540 3541 3542
	return ret;
}

3543
static vm_fault_t do_cow_fault(struct vm_fault *vmf)
3544
{
J
Jan Kara 已提交
3545
	struct vm_area_struct *vma = vmf->vma;
3546
	vm_fault_t ret;
3547 3548 3549 3550

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

J
Jan Kara 已提交
3551 3552
	vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
	if (!vmf->cow_page)
3553 3554
		return VM_FAULT_OOM;

3555
	if (mem_cgroup_try_charge_delay(vmf->cow_page, vma->vm_mm, GFP_KERNEL,
3556
				&vmf->memcg, false)) {
J
Jan Kara 已提交
3557
		put_page(vmf->cow_page);
3558 3559 3560
		return VM_FAULT_OOM;
	}

J
Jan Kara 已提交
3561
	ret = __do_fault(vmf);
3562 3563
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3564 3565
	if (ret & VM_FAULT_DONE_COW)
		return ret;
3566

3567
	copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
J
Jan Kara 已提交
3568
	__SetPageUptodate(vmf->cow_page);
3569

3570
	ret |= finish_fault(vmf);
3571 3572
	unlock_page(vmf->page);
	put_page(vmf->page);
3573 3574
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
		goto uncharge_out;
3575 3576
	return ret;
uncharge_out:
3577
	mem_cgroup_cancel_charge(vmf->cow_page, vmf->memcg, false);
J
Jan Kara 已提交
3578
	put_page(vmf->cow_page);
3579 3580 3581
	return ret;
}

3582
static vm_fault_t do_shared_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3583
{
J
Jan Kara 已提交
3584
	struct vm_area_struct *vma = vmf->vma;
3585
	vm_fault_t ret, tmp;
3586

J
Jan Kara 已提交
3587
	ret = __do_fault(vmf);
3588
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
3589
		return ret;
L
Linus Torvalds 已提交
3590 3591

	/*
3592 3593
	 * Check if the backing address space wants to know that the page is
	 * about to become writable
L
Linus Torvalds 已提交
3594
	 */
3595
	if (vma->vm_ops->page_mkwrite) {
J
Jan Kara 已提交
3596
		unlock_page(vmf->page);
3597
		tmp = do_page_mkwrite(vmf);
3598 3599
		if (unlikely(!tmp ||
				(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
J
Jan Kara 已提交
3600
			put_page(vmf->page);
3601
			return tmp;
3602
		}
3603 3604
	}

3605
	ret |= finish_fault(vmf);
3606 3607
	if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
					VM_FAULT_RETRY))) {
J
Jan Kara 已提交
3608 3609
		unlock_page(vmf->page);
		put_page(vmf->page);
3610
		return ret;
L
Linus Torvalds 已提交
3611
	}
N
Nick Piggin 已提交
3612

3613
	fault_dirty_shared_page(vma, vmf->page);
3614
	return ret;
3615
}
3616

3617 3618 3619 3620 3621
/*
 * 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().
3622 3623
 * If mmap_sem is released, vma may become invalid (for example
 * by other thread calling munmap()).
3624
 */
3625
static vm_fault_t do_fault(struct vm_fault *vmf)
3626
{
J
Jan Kara 已提交
3627
	struct vm_area_struct *vma = vmf->vma;
3628
	struct mm_struct *vm_mm = vma->vm_mm;
3629
	vm_fault_t ret;
3630

3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660
	/*
	 * 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
3661 3662 3663 3664 3665 3666 3667 3668
		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) {
3669
		pte_free(vm_mm, vmf->prealloc_pte);
3670
		vmf->prealloc_pte = NULL;
H
Hugh Dickins 已提交
3671 3672
	}
	return ret;
3673 3674
}

3675
static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
3676 3677
				unsigned long addr, int page_nid,
				int *flags)
3678 3679 3680 3681
{
	get_page(page);

	count_vm_numa_event(NUMA_HINT_FAULTS);
3682
	if (page_nid == numa_node_id()) {
3683
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
3684 3685
		*flags |= TNF_FAULT_LOCAL;
	}
3686 3687 3688 3689

	return mpol_misplaced(page, vma, addr);
}

3690
static vm_fault_t do_numa_page(struct vm_fault *vmf)
3691
{
J
Jan Kara 已提交
3692
	struct vm_area_struct *vma = vmf->vma;
3693
	struct page *page = NULL;
3694
	int page_nid = NUMA_NO_NODE;
3695
	int last_cpupid;
3696
	int target_nid;
3697
	bool migrated = false;
3698
	pte_t pte, old_pte;
3699
	bool was_writable = pte_savedwrite(vmf->orig_pte);
3700
	int flags = 0;
3701 3702

	/*
T
Tobin C Harding 已提交
3703 3704 3705 3706
	 * 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
3707 3708
	vmf->ptl = pte_lockptr(vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
3709
	if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
J
Jan Kara 已提交
3710
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3711 3712 3713
		goto out;
	}

3714 3715 3716 3717
	/*
	 * Make it present again, Depending on how arch implementes non
	 * accessible ptes, some can allow access by kernel mode.
	 */
3718 3719
	old_pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
	pte = pte_modify(old_pte, vma->vm_page_prot);
3720
	pte = pte_mkyoung(pte);
3721 3722
	if (was_writable)
		pte = pte_mkwrite(pte);
3723
	ptep_modify_prot_commit(vma, vmf->address, vmf->pte, old_pte, pte);
J
Jan Kara 已提交
3724
	update_mmu_cache(vma, vmf->address, vmf->pte);
3725

J
Jan Kara 已提交
3726
	page = vm_normal_page(vma, vmf->address, pte);
3727
	if (!page) {
J
Jan Kara 已提交
3728
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3729 3730 3731
		return 0;
	}

3732 3733
	/* TODO: handle PTE-mapped THP */
	if (PageCompound(page)) {
J
Jan Kara 已提交
3734
		pte_unmap_unlock(vmf->pte, vmf->ptl);
3735 3736 3737
		return 0;
	}

3738
	/*
3739 3740 3741 3742 3743 3744
	 * 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.
3745
	 */
3746
	if (!pte_write(pte))
3747 3748
		flags |= TNF_NO_GROUP;

3749 3750 3751 3752 3753 3754 3755
	/*
	 * 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;

3756
	last_cpupid = page_cpupid_last(page);
3757
	page_nid = page_to_nid(page);
J
Jan Kara 已提交
3758
	target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
K
Kirill A. Shutemov 已提交
3759
			&flags);
J
Jan Kara 已提交
3760
	pte_unmap_unlock(vmf->pte, vmf->ptl);
3761
	if (target_nid == NUMA_NO_NODE) {
3762 3763 3764 3765 3766
		put_page(page);
		goto out;
	}

	/* Migrate to the requested node */
3767
	migrated = migrate_misplaced_page(page, vma, target_nid);
3768
	if (migrated) {
3769
		page_nid = target_nid;
3770
		flags |= TNF_MIGRATED;
3771 3772
	} else
		flags |= TNF_MIGRATE_FAIL;
3773 3774

out:
3775
	if (page_nid != NUMA_NO_NODE)
3776
		task_numa_fault(last_cpupid, page_nid, 1, flags);
3777 3778 3779
	return 0;
}

3780
static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
M
Matthew Wilcox 已提交
3781
{
3782
	if (vma_is_anonymous(vmf->vma))
J
Jan Kara 已提交
3783
		return do_huge_pmd_anonymous_page(vmf);
3784
	if (vmf->vma->vm_ops->huge_fault)
3785
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
M
Matthew Wilcox 已提交
3786 3787 3788
	return VM_FAULT_FALLBACK;
}

3789
/* `inline' is required to avoid gcc 4.1.2 build error */
3790
static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
M
Matthew Wilcox 已提交
3791
{
J
Jan Kara 已提交
3792 3793
	if (vma_is_anonymous(vmf->vma))
		return do_huge_pmd_wp_page(vmf, orig_pmd);
3794
	if (vmf->vma->vm_ops->huge_fault)
3795
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
K
Kirill A. Shutemov 已提交
3796 3797

	/* COW handled on pte level: split pmd */
J
Jan Kara 已提交
3798 3799
	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
3800

M
Matthew Wilcox 已提交
3801 3802 3803
	return VM_FAULT_FALLBACK;
}

3804 3805 3806 3807 3808
static inline bool vma_is_accessible(struct vm_area_struct *vma)
{
	return vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE);
}

3809
static vm_fault_t create_huge_pud(struct vm_fault *vmf)
3810 3811 3812 3813 3814 3815
{
#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)
3816
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3817 3818 3819 3820
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

3821
static vm_fault_t wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
3822 3823 3824 3825 3826 3827
{
#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)
3828
		return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
3829 3830 3831 3832
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
	return VM_FAULT_FALLBACK;
}

L
Linus Torvalds 已提交
3833 3834 3835 3836 3837 3838 3839 3840 3841
/*
 * 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).
 *
3842 3843
 * We enter with non-exclusive mmap_sem (to exclude vma changes, but allow
 * concurrent faults).
3844
 *
3845 3846
 * 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
3847
 */
3848
static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3849 3850 3851
{
	pte_t entry;

J
Jan Kara 已提交
3852
	if (unlikely(pmd_none(*vmf->pmd))) {
3853 3854 3855 3856 3857 3858
		/*
		 * 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
3859
		vmf->pte = NULL;
3860 3861
	} else {
		/* See comment in pte_alloc_one_map() */
3862
		if (pmd_devmap_trans_unstable(vmf->pmd))
3863 3864 3865 3866 3867 3868 3869
			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
3870
		vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
J
Jan Kara 已提交
3871
		vmf->orig_pte = *vmf->pte;
3872 3873 3874 3875

		/*
		 * some architectures can have larger ptes than wordsize,
		 * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
3876 3877 3878
		 * 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
3879 3880 3881
		 * ptl lock held. So here a barrier will do.
		 */
		barrier();
J
Jan Kara 已提交
3882
		if (pte_none(vmf->orig_pte)) {
J
Jan Kara 已提交
3883 3884
			pte_unmap(vmf->pte);
			vmf->pte = NULL;
3885
		}
L
Linus Torvalds 已提交
3886 3887
	}

J
Jan Kara 已提交
3888 3889 3890
	if (!vmf->pte) {
		if (vma_is_anonymous(vmf->vma))
			return do_anonymous_page(vmf);
3891
		else
J
Jan Kara 已提交
3892
			return do_fault(vmf);
3893 3894
	}

J
Jan Kara 已提交
3895 3896
	if (!pte_present(vmf->orig_pte))
		return do_swap_page(vmf);
3897

J
Jan Kara 已提交
3898 3899
	if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
		return do_numa_page(vmf);
3900

J
Jan Kara 已提交
3901 3902
	vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
	spin_lock(vmf->ptl);
J
Jan Kara 已提交
3903
	entry = vmf->orig_pte;
J
Jan Kara 已提交
3904
	if (unlikely(!pte_same(*vmf->pte, entry)))
3905
		goto unlock;
J
Jan Kara 已提交
3906
	if (vmf->flags & FAULT_FLAG_WRITE) {
3907
		if (!pte_write(entry))
J
Jan Kara 已提交
3908
			return do_wp_page(vmf);
L
Linus Torvalds 已提交
3909 3910 3911
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
J
Jan Kara 已提交
3912 3913 3914
	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);
3915 3916 3917 3918 3919 3920 3921
	} 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
3922 3923
		if (vmf->flags & FAULT_FLAG_WRITE)
			flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
3924
	}
3925
unlock:
J
Jan Kara 已提交
3926
	pte_unmap_unlock(vmf->pte, vmf->ptl);
N
Nick Piggin 已提交
3927
	return 0;
L
Linus Torvalds 已提交
3928 3929 3930 3931
}

/*
 * By the time we get here, we already hold the mm semaphore
3932 3933 3934
 *
 * 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
3935
 */
3936 3937
static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags)
L
Linus Torvalds 已提交
3938
{
J
Jan Kara 已提交
3939
	struct vm_fault vmf = {
K
Kirill A. Shutemov 已提交
3940
		.vma = vma,
3941
		.address = address & PAGE_MASK,
K
Kirill A. Shutemov 已提交
3942
		.flags = flags,
3943
		.pgoff = linear_page_index(vma, address),
3944
		.gfp_mask = __get_fault_gfp_mask(vma),
K
Kirill A. Shutemov 已提交
3945
	};
3946
	unsigned int dirty = flags & FAULT_FLAG_WRITE;
3947
	struct mm_struct *mm = vma->vm_mm;
L
Linus Torvalds 已提交
3948
	pgd_t *pgd;
3949
	p4d_t *p4d;
3950
	vm_fault_t ret;
L
Linus Torvalds 已提交
3951 3952

	pgd = pgd_offset(mm, address);
3953 3954 3955
	p4d = p4d_alloc(mm, pgd, address);
	if (!p4d)
		return VM_FAULT_OOM;
3956

3957
	vmf.pud = pud_alloc(mm, p4d, address);
3958
	if (!vmf.pud)
H
Hugh Dickins 已提交
3959
		return VM_FAULT_OOM;
3960
	if (pud_none(*vmf.pud) && __transparent_hugepage_enabled(vma)) {
3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971
		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 */

3972
			if (dirty && !pud_write(orig_pud)) {
3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983
				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
3984
	if (!vmf.pmd)
H
Hugh Dickins 已提交
3985
		return VM_FAULT_OOM;
3986
	if (pmd_none(*vmf.pmd) && __transparent_hugepage_enabled(vma)) {
3987
		ret = create_huge_pmd(&vmf);
3988 3989
		if (!(ret & VM_FAULT_FALLBACK))
			return ret;
3990
	} else {
J
Jan Kara 已提交
3991
		pmd_t orig_pmd = *vmf.pmd;
3992

3993
		barrier();
3994 3995 3996 3997 3998 3999 4000
		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;
		}
4001
		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
4002
			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
J
Jan Kara 已提交
4003
				return do_huge_pmd_numa_page(&vmf, orig_pmd);
4004

4005
			if (dirty && !pmd_write(orig_pmd)) {
J
Jan Kara 已提交
4006
				ret = wp_huge_pmd(&vmf, orig_pmd);
4007 4008
				if (!(ret & VM_FAULT_FALLBACK))
					return ret;
4009
			} else {
J
Jan Kara 已提交
4010
				huge_pmd_set_accessed(&vmf, orig_pmd);
4011
				return 0;
4012
			}
4013 4014 4015
		}
	}

J
Jan Kara 已提交
4016
	return handle_pte_fault(&vmf);
L
Linus Torvalds 已提交
4017 4018
}

4019 4020 4021 4022 4023 4024
/*
 * 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().
 */
4025
vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
4026
		unsigned int flags)
4027
{
4028
	vm_fault_t ret;
4029 4030 4031 4032

	__set_current_state(TASK_RUNNING);

	count_vm_event(PGFAULT);
4033
	count_memcg_event_mm(vma->vm_mm, PGFAULT);
4034 4035 4036 4037

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

4038 4039 4040 4041 4042
	if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
					    flags & FAULT_FLAG_INSTRUCTION,
					    flags & FAULT_FLAG_REMOTE))
		return VM_FAULT_SIGSEGV;

4043 4044 4045 4046 4047
	/*
	 * Enable the memcg OOM handling for faults triggered in user
	 * space.  Kernel faults are handled more gracefully.
	 */
	if (flags & FAULT_FLAG_USER)
4048
		mem_cgroup_enter_user_fault();
4049

K
Kirill A. Shutemov 已提交
4050 4051 4052 4053
	if (unlikely(is_vm_hugetlb_page(vma)))
		ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
	else
		ret = __handle_mm_fault(vma, address, flags);
4054

4055
	if (flags & FAULT_FLAG_USER) {
4056
		mem_cgroup_exit_user_fault();
T
Tobin C Harding 已提交
4057 4058 4059 4060 4061 4062 4063 4064
		/*
		 * 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);
4065
	}
4066

4067 4068
	return ret;
}
4069
EXPORT_SYMBOL_GPL(handle_mm_fault);
4070

K
Kirill A. Shutemov 已提交
4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
#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
4094 4095 4096
#ifndef __PAGETABLE_PUD_FOLDED
/*
 * Allocate page upper directory.
H
Hugh Dickins 已提交
4097
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4098
 */
4099
int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
L
Linus Torvalds 已提交
4100
{
H
Hugh Dickins 已提交
4101 4102
	pud_t *new = pud_alloc_one(mm, address);
	if (!new)
4103
		return -ENOMEM;
L
Linus Torvalds 已提交
4104

4105 4106
	smp_wmb(); /* See comment in __pte_alloc */

H
Hugh Dickins 已提交
4107
	spin_lock(&mm->page_table_lock);
4108
#ifndef __ARCH_HAS_5LEVEL_HACK
K
Kirill A. Shutemov 已提交
4109 4110
	if (!p4d_present(*p4d)) {
		mm_inc_nr_puds(mm);
4111
		p4d_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4112
	} else	/* Another has populated it */
4113
		pud_free(mm, new);
K
Kirill A. Shutemov 已提交
4114 4115 4116
#else
	if (!pgd_present(*p4d)) {
		mm_inc_nr_puds(mm);
4117
		pgd_populate(mm, p4d, new);
K
Kirill A. Shutemov 已提交
4118 4119
	} else	/* Another has populated it */
		pud_free(mm, new);
4120
#endif /* __ARCH_HAS_5LEVEL_HACK */
H
Hugh Dickins 已提交
4121
	spin_unlock(&mm->page_table_lock);
4122
	return 0;
L
Linus Torvalds 已提交
4123 4124 4125 4126 4127 4128
}
#endif /* __PAGETABLE_PUD_FOLDED */

#ifndef __PAGETABLE_PMD_FOLDED
/*
 * Allocate page middle directory.
H
Hugh Dickins 已提交
4129
 * We've already handled the fast-path in-line.
L
Linus Torvalds 已提交
4130
 */
4131
int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
L
Linus Torvalds 已提交
4132
{
4133
	spinlock_t *ptl;
H
Hugh Dickins 已提交
4134 4135
	pmd_t *new = pmd_alloc_one(mm, address);
	if (!new)
4136
		return -ENOMEM;
L
Linus Torvalds 已提交
4137

4138 4139
	smp_wmb(); /* See comment in __pte_alloc */

4140
	ptl = pud_lock(mm, pud);
L
Linus Torvalds 已提交
4141
#ifndef __ARCH_HAS_4LEVEL_HACK
4142 4143
	if (!pud_present(*pud)) {
		mm_inc_nr_pmds(mm);
4144
		pud_populate(mm, pud, new);
4145
	} else	/* Another has populated it */
4146
		pmd_free(mm, new);
4147 4148 4149
#else
	if (!pgd_present(*pud)) {
		mm_inc_nr_pmds(mm);
4150
		pgd_populate(mm, pud, new);
4151 4152
	} else /* Another has populated it */
		pmd_free(mm, new);
L
Linus Torvalds 已提交
4153
#endif /* __ARCH_HAS_4LEVEL_HACK */
4154
	spin_unlock(ptl);
4155
	return 0;
4156
}
L
Linus Torvalds 已提交
4157 4158
#endif /* __PAGETABLE_PMD_FOLDED */

R
Ross Zwisler 已提交
4159
static int __follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4160
			    struct mmu_notifier_range *range,
4161
			    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
J
Johannes Weiner 已提交
4162 4163
{
	pgd_t *pgd;
4164
	p4d_t *p4d;
J
Johannes Weiner 已提交
4165 4166 4167 4168 4169 4170 4171 4172
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;

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

4173 4174 4175 4176 4177
	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
4178 4179 4180 4181
	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
		goto out;

	pmd = pmd_offset(pud, address);
4182
	VM_BUG_ON(pmd_trans_huge(*pmd));
J
Johannes Weiner 已提交
4183

R
Ross Zwisler 已提交
4184 4185 4186 4187
	if (pmd_huge(*pmd)) {
		if (!pmdpp)
			goto out;

4188
		if (range) {
4189
			mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0,
4190 4191
						NULL, mm, address & PMD_MASK,
						(address & PMD_MASK) + PMD_SIZE);
4192
			mmu_notifier_invalidate_range_start(range);
4193
		}
R
Ross Zwisler 已提交
4194 4195 4196 4197 4198 4199
		*ptlp = pmd_lock(mm, pmd);
		if (pmd_huge(*pmd)) {
			*pmdpp = pmd;
			return 0;
		}
		spin_unlock(*ptlp);
4200 4201
		if (range)
			mmu_notifier_invalidate_range_end(range);
R
Ross Zwisler 已提交
4202 4203 4204
	}

	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
J
Johannes Weiner 已提交
4205 4206
		goto out;

4207
	if (range) {
4208
		mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
4209 4210
					address & PAGE_MASK,
					(address & PAGE_MASK) + PAGE_SIZE);
4211
		mmu_notifier_invalidate_range_start(range);
4212
	}
J
Johannes Weiner 已提交
4213 4214 4215 4216 4217 4218 4219
	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);
4220 4221
	if (range)
		mmu_notifier_invalidate_range_end(range);
J
Johannes Weiner 已提交
4222 4223 4224 4225
out:
	return -EINVAL;
}

4226 4227
static inline int follow_pte(struct mm_struct *mm, unsigned long address,
			     pte_t **ptepp, spinlock_t **ptlp)
4228 4229 4230 4231 4232
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4233
			   !(res = __follow_pte_pmd(mm, address, NULL,
4234
						    ptepp, NULL, ptlp)));
R
Ross Zwisler 已提交
4235 4236 4237 4238
	return res;
}

int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
4239 4240
		   struct mmu_notifier_range *range,
		   pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp)
R
Ross Zwisler 已提交
4241 4242 4243 4244 4245
{
	int res;

	/* (void) is needed to make gcc happy */
	(void) __cond_lock(*ptlp,
4246
			   !(res = __follow_pte_pmd(mm, address, range,
4247
						    ptepp, pmdpp, ptlp)));
4248 4249
	return res;
}
R
Ross Zwisler 已提交
4250
EXPORT_SYMBOL(follow_pte_pmd);
4251

J
Johannes Weiner 已提交
4252 4253 4254 4255 4256 4257 4258 4259
/**
 * 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.
 *
4260
 * Return: zero and the pfn at @pfn on success, -ve otherwise.
J
Johannes Weiner 已提交
4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280
 */
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);

4281
#ifdef CONFIG_HAVE_IOREMAP_PROT
4282 4283 4284
int follow_phys(struct vm_area_struct *vma,
		unsigned long address, unsigned int flags,
		unsigned long *prot, resource_size_t *phys)
4285
{
4286
	int ret = -EINVAL;
4287 4288 4289
	pte_t *ptep, pte;
	spinlock_t *ptl;

4290 4291
	if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
		goto out;
4292

4293
	if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
4294
		goto out;
4295
	pte = *ptep;
4296

4297
	if ((flags & FOLL_WRITE) && !pte_write(pte))
4298 4299 4300
		goto unlock;

	*prot = pgprot_val(pte_pgprot(pte));
4301
	*phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
4302

4303
	ret = 0;
4304 4305 4306
unlock:
	pte_unmap_unlock(ptep, ptl);
out:
4307
	return ret;
4308 4309 4310 4311 4312 4313 4314
}

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
4315
	void __iomem *maddr;
4316 4317
	int offset = addr & (PAGE_SIZE-1);

4318
	if (follow_phys(vma, addr, write, &prot, &phys_addr))
4319 4320
		return -EINVAL;

4321
	maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
4322 4323 4324
	if (!maddr)
		return -ENOMEM;

4325 4326 4327 4328 4329 4330 4331 4332
	if (write)
		memcpy_toio(maddr + offset, buf, len);
	else
		memcpy_fromio(buf, maddr + offset, len);
	iounmap(maddr);

	return len;
}
4333
EXPORT_SYMBOL_GPL(generic_access_phys);
4334 4335
#endif

4336
/*
4337 4338
 * Access another process' address space as given in mm.  If non-NULL, use the
 * given task for page fault accounting.
4339
 */
4340
int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
4341
		unsigned long addr, void *buf, int len, unsigned int gup_flags)
4342 4343 4344
{
	struct vm_area_struct *vma;
	void *old_buf = buf;
4345
	int write = gup_flags & FOLL_WRITE;
4346 4347

	down_read(&mm->mmap_sem);
S
Simon Arlott 已提交
4348
	/* ignore errors, just check how much was successfully transferred */
4349 4350 4351
	while (len) {
		int bytes, ret, offset;
		void *maddr;
4352
		struct page *page = NULL;
4353

4354
		ret = get_user_pages_remote(tsk, mm, addr, 1,
4355
				gup_flags, &page, &vma, NULL);
4356
		if (ret <= 0) {
4357 4358 4359
#ifndef CONFIG_HAVE_IOREMAP_PROT
			break;
#else
4360 4361 4362 4363 4364
			/*
			 * Check if this is a VM_IO | VM_PFNMAP VMA, which
			 * we can access using slightly different code.
			 */
			vma = find_vma(mm, addr);
4365
			if (!vma || vma->vm_start > addr)
4366 4367 4368 4369 4370 4371 4372
				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;
4373
#endif
4374
		} else {
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389
			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);
4390
			put_page(page);
4391 4392 4393 4394 4395 4396 4397 4398 4399
		}
		len -= bytes;
		buf += bytes;
		addr += bytes;
	}
	up_read(&mm->mmap_sem);

	return buf - old_buf;
}
4400

S
Stephen Wilson 已提交
4401
/**
4402
 * access_remote_vm - access another process' address space
S
Stephen Wilson 已提交
4403 4404 4405 4406
 * @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
4407
 * @gup_flags:	flags modifying lookup behaviour
S
Stephen Wilson 已提交
4408 4409
 *
 * The caller must hold a reference on @mm.
4410 4411
 *
 * Return: number of bytes copied from source to destination.
S
Stephen Wilson 已提交
4412 4413
 */
int access_remote_vm(struct mm_struct *mm, unsigned long addr,
4414
		void *buf, int len, unsigned int gup_flags)
S
Stephen Wilson 已提交
4415
{
4416
	return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
S
Stephen Wilson 已提交
4417 4418
}

4419 4420 4421 4422 4423 4424
/*
 * 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,
4425
		void *buf, int len, unsigned int gup_flags)
4426 4427 4428 4429 4430 4431 4432 4433
{
	struct mm_struct *mm;
	int ret;

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

4434
	ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
4435

4436 4437 4438 4439
	mmput(mm);

	return ret;
}
4440
EXPORT_SYMBOL_GPL(access_process_vm);
4441

4442 4443 4444 4445 4446 4447 4448 4449
/*
 * 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;

4450
	/*
4451
	 * we might be running from an atomic context so we cannot sleep
4452
	 */
4453
	if (!down_read_trylock(&mm->mmap_sem))
4454 4455
		return;

4456 4457 4458
	vma = find_vma(mm, ip);
	if (vma && vma->vm_file) {
		struct file *f = vma->vm_file;
4459
		char *buf = (char *)__get_free_page(GFP_NOWAIT);
4460
		if (buf) {
A
Andy Shevchenko 已提交
4461
			char *p;
4462

M
Miklos Szeredi 已提交
4463
			p = file_path(f, buf, PAGE_SIZE);
4464 4465
			if (IS_ERR(p))
				p = "?";
A
Andy Shevchenko 已提交
4466
			printk("%s%s[%lx+%lx]", prefix, kbasename(p),
4467 4468 4469 4470 4471
					vma->vm_start,
					vma->vm_end - vma->vm_start);
			free_page((unsigned long)buf);
		}
	}
4472
	up_read(&mm->mmap_sem);
4473
}
4474

4475
#if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4476
void __might_fault(const char *file, int line)
4477
{
4478 4479 4480 4481 4482 4483
	/*
	 * 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
4484
	if (uaccess_kernel())
4485
		return;
4486
	if (pagefault_disabled())
4487
		return;
4488 4489
	__might_sleep(file, line, 0);
#if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
4490
	if (current->mm)
4491
		might_lock_read(&current->mm->mmap_sem);
4492
#endif
4493
}
4494
EXPORT_SYMBOL(__might_fault);
4495
#endif
A
Andrea Arcangeli 已提交
4496 4497

#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
4498 4499 4500 4501 4502 4503 4504 4505 4506
/*
 * 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
4507
{
4508 4509 4510
	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
4511

4512
	/* Process target subpage last to keep its cache lines hot */
A
Andrea Arcangeli 已提交
4513
	might_sleep();
4514 4515
	n = (addr_hint - addr) / PAGE_SIZE;
	if (2 * n <= pages_per_huge_page) {
4516
		/* If target subpage in first half of huge page */
4517 4518
		base = 0;
		l = n;
4519
		/* Process subpages at the end of huge page */
4520 4521
		for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
			cond_resched();
4522
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4523 4524
		}
	} else {
4525
		/* If target subpage in second half of huge page */
4526 4527
		base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
		l = pages_per_huge_page - n;
4528
		/* Process subpages at the begin of huge page */
4529 4530
		for (i = 0; i < base; i++) {
			cond_resched();
4531
			process_subpage(addr + i * PAGE_SIZE, i, arg);
4532 4533 4534
		}
	}
	/*
4535 4536
	 * Process remaining subpages in left-right-left-right pattern
	 * towards the target subpage
4537 4538 4539 4540 4541 4542
	 */
	for (i = 0; i < l; i++) {
		int left_idx = base + i;
		int right_idx = base + 2 * l - 1 - i;

		cond_resched();
4543
		process_subpage(addr + left_idx * PAGE_SIZE, left_idx, arg);
A
Andrea Arcangeli 已提交
4544
		cond_resched();
4545
		process_subpage(addr + right_idx * PAGE_SIZE, right_idx, arg);
A
Andrea Arcangeli 已提交
4546 4547 4548
	}
}

4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584
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
4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603
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);
	}
}

4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617
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
4618
void copy_user_huge_page(struct page *dst, struct page *src,
4619
			 unsigned long addr_hint, struct vm_area_struct *vma,
A
Andrea Arcangeli 已提交
4620 4621
			 unsigned int pages_per_huge_page)
{
4622 4623 4624 4625 4626 4627 4628
	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
4629 4630 4631 4632 4633 4634 4635

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

4636
	process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
A
Andrea Arcangeli 已提交
4637
}
4638 4639 4640

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

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

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

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

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

static struct kmem_cache *page_ptl_cachep;

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

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

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

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