]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/mm.h
b3b6a7e313e947d9db18855e21b6d624419e7f8d
[mirror_ubuntu-bionic-kernel.git] / include / linux / mm.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_H
3 #define _LINUX_MM_H
4
5 #include <linux/errno.h>
6
7 #ifdef __KERNEL__
8
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28
29 struct mempolicy;
30 struct anon_vma;
31 struct anon_vma_chain;
32 struct file_ra_state;
33 struct user_struct;
34 struct writeback_control;
35 struct bdi_writeback;
36
37 void init_mm_internals(void);
38
39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr;
41
42 static inline void set_max_mapnr(unsigned long limit)
43 {
44 max_mapnr = limit;
45 }
46 #else
47 static inline void set_max_mapnr(unsigned long limit) { }
48 #endif
49
50 extern unsigned long totalram_pages;
51 extern void * high_memory;
52 extern int page_cluster;
53
54 #ifdef CONFIG_SYSCTL
55 extern int sysctl_legacy_va_layout;
56 #else
57 #define sysctl_legacy_va_layout 0
58 #endif
59
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min;
62 extern const int mmap_rnd_bits_max;
63 extern int mmap_rnd_bits __read_mostly;
64 #endif
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min;
67 extern const int mmap_rnd_compat_bits_max;
68 extern int mmap_rnd_compat_bits __read_mostly;
69 #endif
70
71 #include <asm/page.h>
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
74
75 #ifndef __pa_symbol
76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
77 #endif
78
79 #ifndef page_to_virt
80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
81 #endif
82
83 #ifndef lm_alias
84 #define lm_alias(x) __va(__pa_symbol(x))
85 #endif
86
87 /*
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
93 */
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X) (0)
96 #endif
97
98 /*
99 * On some architectures it is expensive to call memset() for small sizes.
100 * Those architectures should provide their own implementation of "struct page"
101 * zeroing by defining this macro in <asm/pgtable.h>.
102 */
103 #ifndef mm_zero_struct_page
104 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
105 #endif
106
107 /*
108 * Default maximum number of active map areas, this limits the number of vmas
109 * per mm struct. Users can overwrite this number by sysctl but there is a
110 * problem.
111 *
112 * When a program's coredump is generated as ELF format, a section is created
113 * per a vma. In ELF, the number of sections is represented in unsigned short.
114 * This means the number of sections should be smaller than 65535 at coredump.
115 * Because the kernel adds some informative sections to a image of program at
116 * generating coredump, we need some margin. The number of extra sections is
117 * 1-3 now and depends on arch. We use "5" as safe margin, here.
118 *
119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120 * not a hard limit any more. Although some userspace tools can be surprised by
121 * that.
122 */
123 #define MAPCOUNT_ELF_CORE_MARGIN (5)
124 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
125
126 extern int sysctl_max_map_count;
127
128 extern unsigned long sysctl_user_reserve_kbytes;
129 extern unsigned long sysctl_admin_reserve_kbytes;
130
131 extern int sysctl_overcommit_memory;
132 extern int sysctl_overcommit_ratio;
133 extern unsigned long sysctl_overcommit_kbytes;
134
135 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
136 size_t *, loff_t *);
137 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
138 size_t *, loff_t *);
139
140 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
141
142 /* to align the pointer to the (next) page boundary */
143 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
144
145 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
146 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
147
148 /*
149 * Linux kernel virtual memory manager primitives.
150 * The idea being to have a "virtual" mm in the same way
151 * we have a virtual fs - giving a cleaner interface to the
152 * mm details, and allowing different kinds of memory mappings
153 * (from shared memory to executable loading to arbitrary
154 * mmap() functions).
155 */
156
157 extern struct kmem_cache *vm_area_cachep;
158
159 #ifndef CONFIG_MMU
160 extern struct rb_root nommu_region_tree;
161 extern struct rw_semaphore nommu_region_sem;
162
163 extern unsigned int kobjsize(const void *objp);
164 #endif
165
166 /*
167 * vm_flags in vm_area_struct, see mm_types.h.
168 * When changing, update also include/trace/events/mmflags.h
169 */
170 #define VM_NONE 0x00000000
171
172 #define VM_READ 0x00000001 /* currently active flags */
173 #define VM_WRITE 0x00000002
174 #define VM_EXEC 0x00000004
175 #define VM_SHARED 0x00000008
176
177 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
178 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
179 #define VM_MAYWRITE 0x00000020
180 #define VM_MAYEXEC 0x00000040
181 #define VM_MAYSHARE 0x00000080
182
183 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
184 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
185 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
186 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
187 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
188
189 #define VM_LOCKED 0x00002000
190 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
191
192 /* Used by sys_madvise() */
193 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
194 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
195
196 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
197 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
198 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
199 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
200 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
201 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
202 #define VM_SYNC 0x00800000 /* Synchronous page faults */
203 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
204 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
205 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
206
207 #ifdef CONFIG_MEM_SOFT_DIRTY
208 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
209 #else
210 # define VM_SOFTDIRTY 0
211 #endif
212
213 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
214 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
215 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
216 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
217
218 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
219 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
220 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
221 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
222 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
225 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
226 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
227 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
228 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
229 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
230
231 #if defined(CONFIG_X86)
232 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
233 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
234 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
235 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
236 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
237 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
238 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
239 #endif
240 #elif defined(CONFIG_PPC)
241 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
242 #elif defined(CONFIG_PARISC)
243 # define VM_GROWSUP VM_ARCH_1
244 #elif defined(CONFIG_METAG)
245 # define VM_GROWSUP VM_ARCH_1
246 #elif defined(CONFIG_IA64)
247 # define VM_GROWSUP VM_ARCH_1
248 #elif !defined(CONFIG_MMU)
249 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
250 #endif
251
252 #if defined(CONFIG_X86_INTEL_MPX)
253 /* MPX specific bounds table or bounds directory */
254 # define VM_MPX VM_HIGH_ARCH_4
255 #else
256 # define VM_MPX VM_NONE
257 #endif
258
259 #ifndef VM_GROWSUP
260 # define VM_GROWSUP VM_NONE
261 #endif
262
263 /* Bits set in the VMA until the stack is in its final location */
264 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
265
266 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
267 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
268 #endif
269
270 #ifdef CONFIG_STACK_GROWSUP
271 #define VM_STACK VM_GROWSUP
272 #else
273 #define VM_STACK VM_GROWSDOWN
274 #endif
275
276 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
277
278 /*
279 * Special vmas that are non-mergable, non-mlock()able.
280 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
281 */
282 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
283
284 /* This mask defines which mm->def_flags a process can inherit its parent */
285 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
286
287 /* This mask is used to clear all the VMA flags used by mlock */
288 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
289
290 /*
291 * mapping from the currently active vm_flags protection bits (the
292 * low four bits) to a page protection mask..
293 */
294 extern pgprot_t protection_map[16];
295
296 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
297 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
298 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
299 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
300 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
301 #define FAULT_FLAG_TRIED 0x20 /* Second try */
302 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
303 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
304 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
305
306 #define FAULT_FLAG_TRACE \
307 { FAULT_FLAG_WRITE, "WRITE" }, \
308 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
309 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
310 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
311 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
312 { FAULT_FLAG_TRIED, "TRIED" }, \
313 { FAULT_FLAG_USER, "USER" }, \
314 { FAULT_FLAG_REMOTE, "REMOTE" }, \
315 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
316
317 /*
318 * vm_fault is filled by the the pagefault handler and passed to the vma's
319 * ->fault function. The vma's ->fault is responsible for returning a bitmask
320 * of VM_FAULT_xxx flags that give details about how the fault was handled.
321 *
322 * MM layer fills up gfp_mask for page allocations but fault handler might
323 * alter it if its implementation requires a different allocation context.
324 *
325 * pgoff should be used in favour of virtual_address, if possible.
326 */
327 struct vm_fault {
328 struct vm_area_struct *vma; /* Target VMA */
329 unsigned int flags; /* FAULT_FLAG_xxx flags */
330 gfp_t gfp_mask; /* gfp mask to be used for allocations */
331 pgoff_t pgoff; /* Logical page offset based on vma */
332 unsigned long address; /* Faulting virtual address */
333 pmd_t *pmd; /* Pointer to pmd entry matching
334 * the 'address' */
335 pud_t *pud; /* Pointer to pud entry matching
336 * the 'address'
337 */
338 pte_t orig_pte; /* Value of PTE at the time of fault */
339
340 struct page *cow_page; /* Page handler may use for COW fault */
341 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
342 struct page *page; /* ->fault handlers should return a
343 * page here, unless VM_FAULT_NOPAGE
344 * is set (which is also implied by
345 * VM_FAULT_ERROR).
346 */
347 /* These three entries are valid only while holding ptl lock */
348 pte_t *pte; /* Pointer to pte entry matching
349 * the 'address'. NULL if the page
350 * table hasn't been allocated.
351 */
352 spinlock_t *ptl; /* Page table lock.
353 * Protects pte page table if 'pte'
354 * is not NULL, otherwise pmd.
355 */
356 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
357 * vm_ops->map_pages() calls
358 * alloc_set_pte() from atomic context.
359 * do_fault_around() pre-allocates
360 * page table to avoid allocation from
361 * atomic context.
362 */
363 };
364
365 /* page entry size for vm->huge_fault() */
366 enum page_entry_size {
367 PE_SIZE_PTE = 0,
368 PE_SIZE_PMD,
369 PE_SIZE_PUD,
370 };
371
372 /*
373 * These are the virtual MM functions - opening of an area, closing and
374 * unmapping it (needed to keep files on disk up-to-date etc), pointer
375 * to the functions called when a no-page or a wp-page exception occurs.
376 */
377 struct vm_operations_struct {
378 void (*open)(struct vm_area_struct * area);
379 void (*close)(struct vm_area_struct * area);
380 int (*split)(struct vm_area_struct * area, unsigned long addr);
381 int (*mremap)(struct vm_area_struct * area);
382 int (*fault)(struct vm_fault *vmf);
383 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
384 void (*map_pages)(struct vm_fault *vmf,
385 pgoff_t start_pgoff, pgoff_t end_pgoff);
386
387 /* notification that a previously read-only page is about to become
388 * writable, if an error is returned it will cause a SIGBUS */
389 int (*page_mkwrite)(struct vm_fault *vmf);
390
391 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
392 int (*pfn_mkwrite)(struct vm_fault *vmf);
393
394 /* called by access_process_vm when get_user_pages() fails, typically
395 * for use by special VMAs that can switch between memory and hardware
396 */
397 int (*access)(struct vm_area_struct *vma, unsigned long addr,
398 void *buf, int len, int write);
399
400 /* Called by the /proc/PID/maps code to ask the vma whether it
401 * has a special name. Returning non-NULL will also cause this
402 * vma to be dumped unconditionally. */
403 const char *(*name)(struct vm_area_struct *vma);
404
405 #ifdef CONFIG_NUMA
406 /*
407 * set_policy() op must add a reference to any non-NULL @new mempolicy
408 * to hold the policy upon return. Caller should pass NULL @new to
409 * remove a policy and fall back to surrounding context--i.e. do not
410 * install a MPOL_DEFAULT policy, nor the task or system default
411 * mempolicy.
412 */
413 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
414
415 /*
416 * get_policy() op must add reference [mpol_get()] to any policy at
417 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
418 * in mm/mempolicy.c will do this automatically.
419 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
420 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
421 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
422 * must return NULL--i.e., do not "fallback" to task or system default
423 * policy.
424 */
425 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
426 unsigned long addr);
427 #endif
428 /*
429 * Called by vm_normal_page() for special PTEs to find the
430 * page for @addr. This is useful if the default behavior
431 * (using pte_page()) would not find the correct page.
432 */
433 struct page *(*find_special_page)(struct vm_area_struct *vma,
434 unsigned long addr);
435 };
436
437 struct mmu_gather;
438 struct inode;
439
440 #define page_private(page) ((page)->private)
441 #define set_page_private(page, v) ((page)->private = (v))
442
443 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
444 static inline int pmd_devmap(pmd_t pmd)
445 {
446 return 0;
447 }
448 static inline int pud_devmap(pud_t pud)
449 {
450 return 0;
451 }
452 static inline int pgd_devmap(pgd_t pgd)
453 {
454 return 0;
455 }
456 #endif
457
458 /*
459 * FIXME: take this include out, include page-flags.h in
460 * files which need it (119 of them)
461 */
462 #include <linux/page-flags.h>
463 #include <linux/huge_mm.h>
464
465 /*
466 * Methods to modify the page usage count.
467 *
468 * What counts for a page usage:
469 * - cache mapping (page->mapping)
470 * - private data (page->private)
471 * - page mapped in a task's page tables, each mapping
472 * is counted separately
473 *
474 * Also, many kernel routines increase the page count before a critical
475 * routine so they can be sure the page doesn't go away from under them.
476 */
477
478 /*
479 * Drop a ref, return true if the refcount fell to zero (the page has no users)
480 */
481 static inline int put_page_testzero(struct page *page)
482 {
483 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
484 return page_ref_dec_and_test(page);
485 }
486
487 /*
488 * Try to grab a ref unless the page has a refcount of zero, return false if
489 * that is the case.
490 * This can be called when MMU is off so it must not access
491 * any of the virtual mappings.
492 */
493 static inline int get_page_unless_zero(struct page *page)
494 {
495 return page_ref_add_unless(page, 1, 0);
496 }
497
498 extern int page_is_ram(unsigned long pfn);
499
500 enum {
501 REGION_INTERSECTS,
502 REGION_DISJOINT,
503 REGION_MIXED,
504 };
505
506 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
507 unsigned long desc);
508
509 /* Support for virtually mapped pages */
510 struct page *vmalloc_to_page(const void *addr);
511 unsigned long vmalloc_to_pfn(const void *addr);
512
513 /*
514 * Determine if an address is within the vmalloc range
515 *
516 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
517 * is no special casing required.
518 */
519 static inline bool is_vmalloc_addr(const void *x)
520 {
521 #ifdef CONFIG_MMU
522 unsigned long addr = (unsigned long)x;
523
524 return addr >= VMALLOC_START && addr < VMALLOC_END;
525 #else
526 return false;
527 #endif
528 }
529 #ifdef CONFIG_MMU
530 extern int is_vmalloc_or_module_addr(const void *x);
531 #else
532 static inline int is_vmalloc_or_module_addr(const void *x)
533 {
534 return 0;
535 }
536 #endif
537
538 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
539 static inline void *kvmalloc(size_t size, gfp_t flags)
540 {
541 return kvmalloc_node(size, flags, NUMA_NO_NODE);
542 }
543 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
544 {
545 return kvmalloc_node(size, flags | __GFP_ZERO, node);
546 }
547 static inline void *kvzalloc(size_t size, gfp_t flags)
548 {
549 return kvmalloc(size, flags | __GFP_ZERO);
550 }
551
552 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
553 {
554 if (size != 0 && n > SIZE_MAX / size)
555 return NULL;
556
557 return kvmalloc(n * size, flags);
558 }
559
560 extern void kvfree(const void *addr);
561
562 static inline atomic_t *compound_mapcount_ptr(struct page *page)
563 {
564 return &page[1].compound_mapcount;
565 }
566
567 static inline int compound_mapcount(struct page *page)
568 {
569 VM_BUG_ON_PAGE(!PageCompound(page), page);
570 page = compound_head(page);
571 return atomic_read(compound_mapcount_ptr(page)) + 1;
572 }
573
574 /*
575 * The atomic page->_mapcount, starts from -1: so that transitions
576 * both from it and to it can be tracked, using atomic_inc_and_test
577 * and atomic_add_negative(-1).
578 */
579 static inline void page_mapcount_reset(struct page *page)
580 {
581 atomic_set(&(page)->_mapcount, -1);
582 }
583
584 int __page_mapcount(struct page *page);
585
586 static inline int page_mapcount(struct page *page)
587 {
588 VM_BUG_ON_PAGE(PageSlab(page), page);
589
590 if (unlikely(PageCompound(page)))
591 return __page_mapcount(page);
592 return atomic_read(&page->_mapcount) + 1;
593 }
594
595 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
596 int total_mapcount(struct page *page);
597 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
598 #else
599 static inline int total_mapcount(struct page *page)
600 {
601 return page_mapcount(page);
602 }
603 static inline int page_trans_huge_mapcount(struct page *page,
604 int *total_mapcount)
605 {
606 int mapcount = page_mapcount(page);
607 if (total_mapcount)
608 *total_mapcount = mapcount;
609 return mapcount;
610 }
611 #endif
612
613 static inline struct page *virt_to_head_page(const void *x)
614 {
615 struct page *page = virt_to_page(x);
616
617 return compound_head(page);
618 }
619
620 void __put_page(struct page *page);
621
622 void put_pages_list(struct list_head *pages);
623
624 void split_page(struct page *page, unsigned int order);
625
626 /*
627 * Compound pages have a destructor function. Provide a
628 * prototype for that function and accessor functions.
629 * These are _only_ valid on the head of a compound page.
630 */
631 typedef void compound_page_dtor(struct page *);
632
633 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
634 enum compound_dtor_id {
635 NULL_COMPOUND_DTOR,
636 COMPOUND_PAGE_DTOR,
637 #ifdef CONFIG_HUGETLB_PAGE
638 HUGETLB_PAGE_DTOR,
639 #endif
640 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
641 TRANSHUGE_PAGE_DTOR,
642 #endif
643 NR_COMPOUND_DTORS,
644 };
645 extern compound_page_dtor * const compound_page_dtors[];
646
647 static inline void set_compound_page_dtor(struct page *page,
648 enum compound_dtor_id compound_dtor)
649 {
650 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
651 page[1].compound_dtor = compound_dtor;
652 }
653
654 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
655 {
656 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
657 return compound_page_dtors[page[1].compound_dtor];
658 }
659
660 static inline unsigned int compound_order(struct page *page)
661 {
662 if (!PageHead(page))
663 return 0;
664 return page[1].compound_order;
665 }
666
667 static inline void set_compound_order(struct page *page, unsigned int order)
668 {
669 page[1].compound_order = order;
670 }
671
672 void free_compound_page(struct page *page);
673
674 #ifdef CONFIG_MMU
675 /*
676 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
677 * servicing faults for write access. In the normal case, do always want
678 * pte_mkwrite. But get_user_pages can cause write faults for mappings
679 * that do not have writing enabled, when used by access_process_vm.
680 */
681 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
682 {
683 if (likely(vma->vm_flags & VM_WRITE))
684 pte = pte_mkwrite(pte);
685 return pte;
686 }
687
688 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
689 struct page *page);
690 int finish_fault(struct vm_fault *vmf);
691 int finish_mkwrite_fault(struct vm_fault *vmf);
692 #endif
693
694 /*
695 * Multiple processes may "see" the same page. E.g. for untouched
696 * mappings of /dev/null, all processes see the same page full of
697 * zeroes, and text pages of executables and shared libraries have
698 * only one copy in memory, at most, normally.
699 *
700 * For the non-reserved pages, page_count(page) denotes a reference count.
701 * page_count() == 0 means the page is free. page->lru is then used for
702 * freelist management in the buddy allocator.
703 * page_count() > 0 means the page has been allocated.
704 *
705 * Pages are allocated by the slab allocator in order to provide memory
706 * to kmalloc and kmem_cache_alloc. In this case, the management of the
707 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
708 * unless a particular usage is carefully commented. (the responsibility of
709 * freeing the kmalloc memory is the caller's, of course).
710 *
711 * A page may be used by anyone else who does a __get_free_page().
712 * In this case, page_count still tracks the references, and should only
713 * be used through the normal accessor functions. The top bits of page->flags
714 * and page->virtual store page management information, but all other fields
715 * are unused and could be used privately, carefully. The management of this
716 * page is the responsibility of the one who allocated it, and those who have
717 * subsequently been given references to it.
718 *
719 * The other pages (we may call them "pagecache pages") are completely
720 * managed by the Linux memory manager: I/O, buffers, swapping etc.
721 * The following discussion applies only to them.
722 *
723 * A pagecache page contains an opaque `private' member, which belongs to the
724 * page's address_space. Usually, this is the address of a circular list of
725 * the page's disk buffers. PG_private must be set to tell the VM to call
726 * into the filesystem to release these pages.
727 *
728 * A page may belong to an inode's memory mapping. In this case, page->mapping
729 * is the pointer to the inode, and page->index is the file offset of the page,
730 * in units of PAGE_SIZE.
731 *
732 * If pagecache pages are not associated with an inode, they are said to be
733 * anonymous pages. These may become associated with the swapcache, and in that
734 * case PG_swapcache is set, and page->private is an offset into the swapcache.
735 *
736 * In either case (swapcache or inode backed), the pagecache itself holds one
737 * reference to the page. Setting PG_private should also increment the
738 * refcount. The each user mapping also has a reference to the page.
739 *
740 * The pagecache pages are stored in a per-mapping radix tree, which is
741 * rooted at mapping->page_tree, and indexed by offset.
742 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
743 * lists, we instead now tag pages as dirty/writeback in the radix tree.
744 *
745 * All pagecache pages may be subject to I/O:
746 * - inode pages may need to be read from disk,
747 * - inode pages which have been modified and are MAP_SHARED may need
748 * to be written back to the inode on disk,
749 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
750 * modified may need to be swapped out to swap space and (later) to be read
751 * back into memory.
752 */
753
754 /*
755 * The zone field is never updated after free_area_init_core()
756 * sets it, so none of the operations on it need to be atomic.
757 */
758
759 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
760 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
761 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
762 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
763 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
764
765 /*
766 * Define the bit shifts to access each section. For non-existent
767 * sections we define the shift as 0; that plus a 0 mask ensures
768 * the compiler will optimise away reference to them.
769 */
770 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
771 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
772 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
773 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
774
775 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
776 #ifdef NODE_NOT_IN_PAGE_FLAGS
777 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
778 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
779 SECTIONS_PGOFF : ZONES_PGOFF)
780 #else
781 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
782 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
783 NODES_PGOFF : ZONES_PGOFF)
784 #endif
785
786 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
787
788 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
790 #endif
791
792 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
793 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
794 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
795 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
796 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
797
798 static inline enum zone_type page_zonenum(const struct page *page)
799 {
800 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
801 }
802
803 #ifdef CONFIG_ZONE_DEVICE
804 static inline bool is_zone_device_page(const struct page *page)
805 {
806 return page_zonenum(page) == ZONE_DEVICE;
807 }
808 #else
809 static inline bool is_zone_device_page(const struct page *page)
810 {
811 return false;
812 }
813 #endif
814
815 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
816 void put_zone_device_private_or_public_page(struct page *page);
817 DECLARE_STATIC_KEY_FALSE(device_private_key);
818 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
819 static inline bool is_device_private_page(const struct page *page);
820 static inline bool is_device_public_page(const struct page *page);
821 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
822 static inline void put_zone_device_private_or_public_page(struct page *page)
823 {
824 }
825 #define IS_HMM_ENABLED 0
826 static inline bool is_device_private_page(const struct page *page)
827 {
828 return false;
829 }
830 static inline bool is_device_public_page(const struct page *page)
831 {
832 return false;
833 }
834 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
835
836
837 static inline void get_page(struct page *page)
838 {
839 page = compound_head(page);
840 /*
841 * Getting a normal page or the head of a compound page
842 * requires to already have an elevated page->_refcount.
843 */
844 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
845 page_ref_inc(page);
846 }
847
848 static inline void put_page(struct page *page)
849 {
850 page = compound_head(page);
851
852 /*
853 * For private device pages we need to catch refcount transition from
854 * 2 to 1, when refcount reach one it means the private device page is
855 * free and we need to inform the device driver through callback. See
856 * include/linux/memremap.h and HMM for details.
857 */
858 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
859 unlikely(is_device_public_page(page)))) {
860 put_zone_device_private_or_public_page(page);
861 return;
862 }
863
864 if (put_page_testzero(page))
865 __put_page(page);
866 }
867
868 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
869 #define SECTION_IN_PAGE_FLAGS
870 #endif
871
872 /*
873 * The identification function is mainly used by the buddy allocator for
874 * determining if two pages could be buddies. We are not really identifying
875 * the zone since we could be using the section number id if we do not have
876 * node id available in page flags.
877 * We only guarantee that it will return the same value for two combinable
878 * pages in a zone.
879 */
880 static inline int page_zone_id(struct page *page)
881 {
882 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
883 }
884
885 static inline int zone_to_nid(struct zone *zone)
886 {
887 #ifdef CONFIG_NUMA
888 return zone->node;
889 #else
890 return 0;
891 #endif
892 }
893
894 #ifdef NODE_NOT_IN_PAGE_FLAGS
895 extern int page_to_nid(const struct page *page);
896 #else
897 static inline int page_to_nid(const struct page *page)
898 {
899 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
900 }
901 #endif
902
903 #ifdef CONFIG_NUMA_BALANCING
904 static inline int cpu_pid_to_cpupid(int cpu, int pid)
905 {
906 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
907 }
908
909 static inline int cpupid_to_pid(int cpupid)
910 {
911 return cpupid & LAST__PID_MASK;
912 }
913
914 static inline int cpupid_to_cpu(int cpupid)
915 {
916 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
917 }
918
919 static inline int cpupid_to_nid(int cpupid)
920 {
921 return cpu_to_node(cpupid_to_cpu(cpupid));
922 }
923
924 static inline bool cpupid_pid_unset(int cpupid)
925 {
926 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
927 }
928
929 static inline bool cpupid_cpu_unset(int cpupid)
930 {
931 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
932 }
933
934 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
935 {
936 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
937 }
938
939 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
940 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
941 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
942 {
943 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
944 }
945
946 static inline int page_cpupid_last(struct page *page)
947 {
948 return page->_last_cpupid;
949 }
950 static inline void page_cpupid_reset_last(struct page *page)
951 {
952 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
953 }
954 #else
955 static inline int page_cpupid_last(struct page *page)
956 {
957 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
958 }
959
960 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
961
962 static inline void page_cpupid_reset_last(struct page *page)
963 {
964 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
965 }
966 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
967 #else /* !CONFIG_NUMA_BALANCING */
968 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
969 {
970 return page_to_nid(page); /* XXX */
971 }
972
973 static inline int page_cpupid_last(struct page *page)
974 {
975 return page_to_nid(page); /* XXX */
976 }
977
978 static inline int cpupid_to_nid(int cpupid)
979 {
980 return -1;
981 }
982
983 static inline int cpupid_to_pid(int cpupid)
984 {
985 return -1;
986 }
987
988 static inline int cpupid_to_cpu(int cpupid)
989 {
990 return -1;
991 }
992
993 static inline int cpu_pid_to_cpupid(int nid, int pid)
994 {
995 return -1;
996 }
997
998 static inline bool cpupid_pid_unset(int cpupid)
999 {
1000 return 1;
1001 }
1002
1003 static inline void page_cpupid_reset_last(struct page *page)
1004 {
1005 }
1006
1007 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1008 {
1009 return false;
1010 }
1011 #endif /* CONFIG_NUMA_BALANCING */
1012
1013 static inline struct zone *page_zone(const struct page *page)
1014 {
1015 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1016 }
1017
1018 static inline pg_data_t *page_pgdat(const struct page *page)
1019 {
1020 return NODE_DATA(page_to_nid(page));
1021 }
1022
1023 #ifdef SECTION_IN_PAGE_FLAGS
1024 static inline void set_page_section(struct page *page, unsigned long section)
1025 {
1026 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1027 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1028 }
1029
1030 static inline unsigned long page_to_section(const struct page *page)
1031 {
1032 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1033 }
1034 #endif
1035
1036 static inline void set_page_zone(struct page *page, enum zone_type zone)
1037 {
1038 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1039 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1040 }
1041
1042 static inline void set_page_node(struct page *page, unsigned long node)
1043 {
1044 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1045 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1046 }
1047
1048 static inline void set_page_links(struct page *page, enum zone_type zone,
1049 unsigned long node, unsigned long pfn)
1050 {
1051 set_page_zone(page, zone);
1052 set_page_node(page, node);
1053 #ifdef SECTION_IN_PAGE_FLAGS
1054 set_page_section(page, pfn_to_section_nr(pfn));
1055 #endif
1056 }
1057
1058 #ifdef CONFIG_MEMCG
1059 static inline struct mem_cgroup *page_memcg(struct page *page)
1060 {
1061 return page->mem_cgroup;
1062 }
1063 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1064 {
1065 WARN_ON_ONCE(!rcu_read_lock_held());
1066 return READ_ONCE(page->mem_cgroup);
1067 }
1068 #else
1069 static inline struct mem_cgroup *page_memcg(struct page *page)
1070 {
1071 return NULL;
1072 }
1073 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1074 {
1075 WARN_ON_ONCE(!rcu_read_lock_held());
1076 return NULL;
1077 }
1078 #endif
1079
1080 /*
1081 * Some inline functions in vmstat.h depend on page_zone()
1082 */
1083 #include <linux/vmstat.h>
1084
1085 static __always_inline void *lowmem_page_address(const struct page *page)
1086 {
1087 return page_to_virt(page);
1088 }
1089
1090 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1091 #define HASHED_PAGE_VIRTUAL
1092 #endif
1093
1094 #if defined(WANT_PAGE_VIRTUAL)
1095 static inline void *page_address(const struct page *page)
1096 {
1097 return page->virtual;
1098 }
1099 static inline void set_page_address(struct page *page, void *address)
1100 {
1101 page->virtual = address;
1102 }
1103 #define page_address_init() do { } while(0)
1104 #endif
1105
1106 #if defined(HASHED_PAGE_VIRTUAL)
1107 void *page_address(const struct page *page);
1108 void set_page_address(struct page *page, void *virtual);
1109 void page_address_init(void);
1110 #endif
1111
1112 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1113 #define page_address(page) lowmem_page_address(page)
1114 #define set_page_address(page, address) do { } while(0)
1115 #define page_address_init() do { } while(0)
1116 #endif
1117
1118 extern void *page_rmapping(struct page *page);
1119 extern struct anon_vma *page_anon_vma(struct page *page);
1120 extern struct address_space *page_mapping(struct page *page);
1121
1122 extern struct address_space *__page_file_mapping(struct page *);
1123
1124 static inline
1125 struct address_space *page_file_mapping(struct page *page)
1126 {
1127 if (unlikely(PageSwapCache(page)))
1128 return __page_file_mapping(page);
1129
1130 return page->mapping;
1131 }
1132
1133 extern pgoff_t __page_file_index(struct page *page);
1134
1135 /*
1136 * Return the pagecache index of the passed page. Regular pagecache pages
1137 * use ->index whereas swapcache pages use swp_offset(->private)
1138 */
1139 static inline pgoff_t page_index(struct page *page)
1140 {
1141 if (unlikely(PageSwapCache(page)))
1142 return __page_file_index(page);
1143 return page->index;
1144 }
1145
1146 bool page_mapped(struct page *page);
1147 struct address_space *page_mapping(struct page *page);
1148
1149 /*
1150 * Return true only if the page has been allocated with
1151 * ALLOC_NO_WATERMARKS and the low watermark was not
1152 * met implying that the system is under some pressure.
1153 */
1154 static inline bool page_is_pfmemalloc(struct page *page)
1155 {
1156 /*
1157 * Page index cannot be this large so this must be
1158 * a pfmemalloc page.
1159 */
1160 return page->index == -1UL;
1161 }
1162
1163 /*
1164 * Only to be called by the page allocator on a freshly allocated
1165 * page.
1166 */
1167 static inline void set_page_pfmemalloc(struct page *page)
1168 {
1169 page->index = -1UL;
1170 }
1171
1172 static inline void clear_page_pfmemalloc(struct page *page)
1173 {
1174 page->index = 0;
1175 }
1176
1177 /*
1178 * Different kinds of faults, as returned by handle_mm_fault().
1179 * Used to decide whether a process gets delivered SIGBUS or
1180 * just gets major/minor fault counters bumped up.
1181 */
1182
1183 #define VM_FAULT_OOM 0x0001
1184 #define VM_FAULT_SIGBUS 0x0002
1185 #define VM_FAULT_MAJOR 0x0004
1186 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1187 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1188 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1189 #define VM_FAULT_SIGSEGV 0x0040
1190
1191 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1192 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1193 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1194 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1195 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1196 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1197 * and needs fsync() to complete (for
1198 * synchronous page faults in DAX) */
1199
1200 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1201 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1202 VM_FAULT_FALLBACK)
1203
1204 #define VM_FAULT_RESULT_TRACE \
1205 { VM_FAULT_OOM, "OOM" }, \
1206 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1207 { VM_FAULT_MAJOR, "MAJOR" }, \
1208 { VM_FAULT_WRITE, "WRITE" }, \
1209 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1210 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1211 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1212 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1213 { VM_FAULT_LOCKED, "LOCKED" }, \
1214 { VM_FAULT_RETRY, "RETRY" }, \
1215 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1216 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1217 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1218
1219 /* Encode hstate index for a hwpoisoned large page */
1220 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1221 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1222
1223 /*
1224 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1225 */
1226 extern void pagefault_out_of_memory(void);
1227
1228 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1229
1230 /*
1231 * Flags passed to show_mem() and show_free_areas() to suppress output in
1232 * various contexts.
1233 */
1234 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1235
1236 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1237
1238 extern bool can_do_mlock(void);
1239 extern int user_shm_lock(size_t, struct user_struct *);
1240 extern void user_shm_unlock(size_t, struct user_struct *);
1241
1242 /*
1243 * Parameter block passed down to zap_pte_range in exceptional cases.
1244 */
1245 struct zap_details {
1246 struct address_space *check_mapping; /* Check page->mapping if set */
1247 pgoff_t first_index; /* Lowest page->index to unmap */
1248 pgoff_t last_index; /* Highest page->index to unmap */
1249 };
1250
1251 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1252 pte_t pte, bool with_public_device);
1253 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1254
1255 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1256 pmd_t pmd);
1257
1258 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1259 unsigned long size);
1260 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1261 unsigned long size);
1262 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1263 unsigned long start, unsigned long end);
1264
1265 /**
1266 * mm_walk - callbacks for walk_page_range
1267 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1268 * this handler should only handle pud_trans_huge() puds.
1269 * the pmd_entry or pte_entry callbacks will be used for
1270 * regular PUDs.
1271 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1272 * this handler is required to be able to handle
1273 * pmd_trans_huge() pmds. They may simply choose to
1274 * split_huge_page() instead of handling it explicitly.
1275 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1276 * @pte_hole: if set, called for each hole at all levels
1277 * @hugetlb_entry: if set, called for each hugetlb entry
1278 * @test_walk: caller specific callback function to determine whether
1279 * we walk over the current vma or not. Returning 0
1280 * value means "do page table walk over the current vma,"
1281 * and a negative one means "abort current page table walk
1282 * right now." 1 means "skip the current vma."
1283 * @mm: mm_struct representing the target process of page table walk
1284 * @vma: vma currently walked (NULL if walking outside vmas)
1285 * @private: private data for callbacks' usage
1286 *
1287 * (see the comment on walk_page_range() for more details)
1288 */
1289 struct mm_walk {
1290 int (*pud_entry)(pud_t *pud, unsigned long addr,
1291 unsigned long next, struct mm_walk *walk);
1292 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1293 unsigned long next, struct mm_walk *walk);
1294 int (*pte_entry)(pte_t *pte, unsigned long addr,
1295 unsigned long next, struct mm_walk *walk);
1296 int (*pte_hole)(unsigned long addr, unsigned long next,
1297 struct mm_walk *walk);
1298 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1299 unsigned long addr, unsigned long next,
1300 struct mm_walk *walk);
1301 int (*test_walk)(unsigned long addr, unsigned long next,
1302 struct mm_walk *walk);
1303 struct mm_struct *mm;
1304 struct vm_area_struct *vma;
1305 void *private;
1306 };
1307
1308 int walk_page_range(unsigned long addr, unsigned long end,
1309 struct mm_walk *walk);
1310 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1311 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1312 unsigned long end, unsigned long floor, unsigned long ceiling);
1313 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1314 struct vm_area_struct *vma);
1315 void unmap_mapping_range(struct address_space *mapping,
1316 loff_t const holebegin, loff_t const holelen, int even_cows);
1317 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1318 unsigned long *start, unsigned long *end,
1319 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1320 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1321 unsigned long *pfn);
1322 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1323 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1324 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1325 void *buf, int len, int write);
1326
1327 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1328 loff_t const holebegin, loff_t const holelen)
1329 {
1330 unmap_mapping_range(mapping, holebegin, holelen, 0);
1331 }
1332
1333 extern void truncate_pagecache(struct inode *inode, loff_t new);
1334 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1335 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1336 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1337 int truncate_inode_page(struct address_space *mapping, struct page *page);
1338 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1339 int invalidate_inode_page(struct page *page);
1340
1341 #ifdef CONFIG_MMU
1342 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1343 unsigned int flags);
1344 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1345 unsigned long address, unsigned int fault_flags,
1346 bool *unlocked);
1347 #else
1348 static inline int handle_mm_fault(struct vm_area_struct *vma,
1349 unsigned long address, unsigned int flags)
1350 {
1351 /* should never happen if there's no MMU */
1352 BUG();
1353 return VM_FAULT_SIGBUS;
1354 }
1355 static inline int fixup_user_fault(struct task_struct *tsk,
1356 struct mm_struct *mm, unsigned long address,
1357 unsigned int fault_flags, bool *unlocked)
1358 {
1359 /* should never happen if there's no MMU */
1360 BUG();
1361 return -EFAULT;
1362 }
1363 #endif
1364
1365 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1366 unsigned int gup_flags);
1367 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1368 void *buf, int len, unsigned int gup_flags);
1369 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1370 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1371
1372 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1373 unsigned long start, unsigned long nr_pages,
1374 unsigned int gup_flags, struct page **pages,
1375 struct vm_area_struct **vmas, int *locked);
1376 long get_user_pages(unsigned long start, unsigned long nr_pages,
1377 unsigned int gup_flags, struct page **pages,
1378 struct vm_area_struct **vmas);
1379 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1380 unsigned int gup_flags, struct page **pages, int *locked);
1381 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1382 struct page **pages, unsigned int gup_flags);
1383 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1384 struct page **pages);
1385
1386 /* Container for pinned pfns / pages */
1387 struct frame_vector {
1388 unsigned int nr_allocated; /* Number of frames we have space for */
1389 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1390 bool got_ref; /* Did we pin pages by getting page ref? */
1391 bool is_pfns; /* Does array contain pages or pfns? */
1392 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1393 * pfns_vector_pages() or pfns_vector_pfns()
1394 * for access */
1395 };
1396
1397 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1398 void frame_vector_destroy(struct frame_vector *vec);
1399 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1400 unsigned int gup_flags, struct frame_vector *vec);
1401 void put_vaddr_frames(struct frame_vector *vec);
1402 int frame_vector_to_pages(struct frame_vector *vec);
1403 void frame_vector_to_pfns(struct frame_vector *vec);
1404
1405 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1406 {
1407 return vec->nr_frames;
1408 }
1409
1410 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1411 {
1412 if (vec->is_pfns) {
1413 int err = frame_vector_to_pages(vec);
1414
1415 if (err)
1416 return ERR_PTR(err);
1417 }
1418 return (struct page **)(vec->ptrs);
1419 }
1420
1421 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1422 {
1423 if (!vec->is_pfns)
1424 frame_vector_to_pfns(vec);
1425 return (unsigned long *)(vec->ptrs);
1426 }
1427
1428 struct kvec;
1429 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1430 struct page **pages);
1431 int get_kernel_page(unsigned long start, int write, struct page **pages);
1432 struct page *get_dump_page(unsigned long addr);
1433
1434 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1435 extern void do_invalidatepage(struct page *page, unsigned int offset,
1436 unsigned int length);
1437
1438 int __set_page_dirty_nobuffers(struct page *page);
1439 int __set_page_dirty_no_writeback(struct page *page);
1440 int redirty_page_for_writepage(struct writeback_control *wbc,
1441 struct page *page);
1442 void account_page_dirtied(struct page *page, struct address_space *mapping);
1443 void account_page_cleaned(struct page *page, struct address_space *mapping,
1444 struct bdi_writeback *wb);
1445 int set_page_dirty(struct page *page);
1446 int set_page_dirty_lock(struct page *page);
1447 void __cancel_dirty_page(struct page *page);
1448 static inline void cancel_dirty_page(struct page *page)
1449 {
1450 /* Avoid atomic ops, locking, etc. when not actually needed. */
1451 if (PageDirty(page))
1452 __cancel_dirty_page(page);
1453 }
1454 int clear_page_dirty_for_io(struct page *page);
1455
1456 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1457
1458 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1459 {
1460 return !vma->vm_ops;
1461 }
1462
1463 #ifdef CONFIG_SHMEM
1464 /*
1465 * The vma_is_shmem is not inline because it is used only by slow
1466 * paths in userfault.
1467 */
1468 bool vma_is_shmem(struct vm_area_struct *vma);
1469 #else
1470 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1471 #endif
1472
1473 int vma_is_stack_for_current(struct vm_area_struct *vma);
1474
1475 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1476 unsigned long old_addr, struct vm_area_struct *new_vma,
1477 unsigned long new_addr, unsigned long len,
1478 bool need_rmap_locks);
1479 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1480 unsigned long end, pgprot_t newprot,
1481 int dirty_accountable, int prot_numa);
1482 extern int mprotect_fixup(struct vm_area_struct *vma,
1483 struct vm_area_struct **pprev, unsigned long start,
1484 unsigned long end, unsigned long newflags);
1485
1486 /*
1487 * doesn't attempt to fault and will return short.
1488 */
1489 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1490 struct page **pages);
1491 /*
1492 * per-process(per-mm_struct) statistics.
1493 */
1494 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1495 {
1496 long val = atomic_long_read(&mm->rss_stat.count[member]);
1497
1498 #ifdef SPLIT_RSS_COUNTING
1499 /*
1500 * counter is updated in asynchronous manner and may go to minus.
1501 * But it's never be expected number for users.
1502 */
1503 if (val < 0)
1504 val = 0;
1505 #endif
1506 return (unsigned long)val;
1507 }
1508
1509 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1510 {
1511 atomic_long_add(value, &mm->rss_stat.count[member]);
1512 }
1513
1514 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1515 {
1516 atomic_long_inc(&mm->rss_stat.count[member]);
1517 }
1518
1519 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1520 {
1521 atomic_long_dec(&mm->rss_stat.count[member]);
1522 }
1523
1524 /* Optimized variant when page is already known not to be PageAnon */
1525 static inline int mm_counter_file(struct page *page)
1526 {
1527 if (PageSwapBacked(page))
1528 return MM_SHMEMPAGES;
1529 return MM_FILEPAGES;
1530 }
1531
1532 static inline int mm_counter(struct page *page)
1533 {
1534 if (PageAnon(page))
1535 return MM_ANONPAGES;
1536 return mm_counter_file(page);
1537 }
1538
1539 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1540 {
1541 return get_mm_counter(mm, MM_FILEPAGES) +
1542 get_mm_counter(mm, MM_ANONPAGES) +
1543 get_mm_counter(mm, MM_SHMEMPAGES);
1544 }
1545
1546 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1547 {
1548 return max(mm->hiwater_rss, get_mm_rss(mm));
1549 }
1550
1551 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1552 {
1553 return max(mm->hiwater_vm, mm->total_vm);
1554 }
1555
1556 static inline void update_hiwater_rss(struct mm_struct *mm)
1557 {
1558 unsigned long _rss = get_mm_rss(mm);
1559
1560 if ((mm)->hiwater_rss < _rss)
1561 (mm)->hiwater_rss = _rss;
1562 }
1563
1564 static inline void update_hiwater_vm(struct mm_struct *mm)
1565 {
1566 if (mm->hiwater_vm < mm->total_vm)
1567 mm->hiwater_vm = mm->total_vm;
1568 }
1569
1570 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1571 {
1572 mm->hiwater_rss = get_mm_rss(mm);
1573 }
1574
1575 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1576 struct mm_struct *mm)
1577 {
1578 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1579
1580 if (*maxrss < hiwater_rss)
1581 *maxrss = hiwater_rss;
1582 }
1583
1584 #if defined(SPLIT_RSS_COUNTING)
1585 void sync_mm_rss(struct mm_struct *mm);
1586 #else
1587 static inline void sync_mm_rss(struct mm_struct *mm)
1588 {
1589 }
1590 #endif
1591
1592 #ifndef __HAVE_ARCH_PTE_DEVMAP
1593 static inline int pte_devmap(pte_t pte)
1594 {
1595 return 0;
1596 }
1597 #endif
1598
1599 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1600
1601 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1602 spinlock_t **ptl);
1603 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1604 spinlock_t **ptl)
1605 {
1606 pte_t *ptep;
1607 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1608 return ptep;
1609 }
1610
1611 #ifdef __PAGETABLE_P4D_FOLDED
1612 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1613 unsigned long address)
1614 {
1615 return 0;
1616 }
1617 #else
1618 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1619 #endif
1620
1621 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1622 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1623 unsigned long address)
1624 {
1625 return 0;
1626 }
1627 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1628 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1629
1630 #else
1631 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1632
1633 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1634 {
1635 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1636 }
1637
1638 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1639 {
1640 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1641 }
1642 #endif
1643
1644 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1645 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1646 unsigned long address)
1647 {
1648 return 0;
1649 }
1650
1651 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1652 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1653
1654 #else
1655 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1656
1657 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1658 {
1659 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1660 }
1661
1662 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1663 {
1664 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1665 }
1666 #endif
1667
1668 #ifdef CONFIG_MMU
1669 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1670 {
1671 atomic_long_set(&mm->pgtables_bytes, 0);
1672 }
1673
1674 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1675 {
1676 return atomic_long_read(&mm->pgtables_bytes);
1677 }
1678
1679 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1680 {
1681 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1682 }
1683
1684 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1685 {
1686 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1687 }
1688 #else
1689
1690 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1691 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1692 {
1693 return 0;
1694 }
1695
1696 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1697 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1698 #endif
1699
1700 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1701 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1702
1703 /*
1704 * The following ifdef needed to get the 4level-fixup.h header to work.
1705 * Remove it when 4level-fixup.h has been removed.
1706 */
1707 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1708
1709 #ifndef __ARCH_HAS_5LEVEL_HACK
1710 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1711 unsigned long address)
1712 {
1713 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1714 NULL : p4d_offset(pgd, address);
1715 }
1716
1717 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1718 unsigned long address)
1719 {
1720 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1721 NULL : pud_offset(p4d, address);
1722 }
1723 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1724
1725 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1726 {
1727 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1728 NULL: pmd_offset(pud, address);
1729 }
1730 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1731
1732 #if USE_SPLIT_PTE_PTLOCKS
1733 #if ALLOC_SPLIT_PTLOCKS
1734 void __init ptlock_cache_init(void);
1735 extern bool ptlock_alloc(struct page *page);
1736 extern void ptlock_free(struct page *page);
1737
1738 static inline spinlock_t *ptlock_ptr(struct page *page)
1739 {
1740 return page->ptl;
1741 }
1742 #else /* ALLOC_SPLIT_PTLOCKS */
1743 static inline void ptlock_cache_init(void)
1744 {
1745 }
1746
1747 static inline bool ptlock_alloc(struct page *page)
1748 {
1749 return true;
1750 }
1751
1752 static inline void ptlock_free(struct page *page)
1753 {
1754 }
1755
1756 static inline spinlock_t *ptlock_ptr(struct page *page)
1757 {
1758 return &page->ptl;
1759 }
1760 #endif /* ALLOC_SPLIT_PTLOCKS */
1761
1762 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1763 {
1764 return ptlock_ptr(pmd_page(*pmd));
1765 }
1766
1767 static inline bool ptlock_init(struct page *page)
1768 {
1769 /*
1770 * prep_new_page() initialize page->private (and therefore page->ptl)
1771 * with 0. Make sure nobody took it in use in between.
1772 *
1773 * It can happen if arch try to use slab for page table allocation:
1774 * slab code uses page->slab_cache, which share storage with page->ptl.
1775 */
1776 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1777 if (!ptlock_alloc(page))
1778 return false;
1779 spin_lock_init(ptlock_ptr(page));
1780 return true;
1781 }
1782
1783 /* Reset page->mapping so free_pages_check won't complain. */
1784 static inline void pte_lock_deinit(struct page *page)
1785 {
1786 page->mapping = NULL;
1787 ptlock_free(page);
1788 }
1789
1790 #else /* !USE_SPLIT_PTE_PTLOCKS */
1791 /*
1792 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1793 */
1794 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1795 {
1796 return &mm->page_table_lock;
1797 }
1798 static inline void ptlock_cache_init(void) {}
1799 static inline bool ptlock_init(struct page *page) { return true; }
1800 static inline void pte_lock_deinit(struct page *page) {}
1801 #endif /* USE_SPLIT_PTE_PTLOCKS */
1802
1803 static inline void pgtable_init(void)
1804 {
1805 ptlock_cache_init();
1806 pgtable_cache_init();
1807 }
1808
1809 static inline bool pgtable_page_ctor(struct page *page)
1810 {
1811 if (!ptlock_init(page))
1812 return false;
1813 inc_zone_page_state(page, NR_PAGETABLE);
1814 return true;
1815 }
1816
1817 static inline void pgtable_page_dtor(struct page *page)
1818 {
1819 pte_lock_deinit(page);
1820 dec_zone_page_state(page, NR_PAGETABLE);
1821 }
1822
1823 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1824 ({ \
1825 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1826 pte_t *__pte = pte_offset_map(pmd, address); \
1827 *(ptlp) = __ptl; \
1828 spin_lock(__ptl); \
1829 __pte; \
1830 })
1831
1832 #define pte_unmap_unlock(pte, ptl) do { \
1833 spin_unlock(ptl); \
1834 pte_unmap(pte); \
1835 } while (0)
1836
1837 #define pte_alloc(mm, pmd, address) \
1838 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1839
1840 #define pte_alloc_map(mm, pmd, address) \
1841 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1842
1843 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1844 (pte_alloc(mm, pmd, address) ? \
1845 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1846
1847 #define pte_alloc_kernel(pmd, address) \
1848 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1849 NULL: pte_offset_kernel(pmd, address))
1850
1851 #if USE_SPLIT_PMD_PTLOCKS
1852
1853 static struct page *pmd_to_page(pmd_t *pmd)
1854 {
1855 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1856 return virt_to_page((void *)((unsigned long) pmd & mask));
1857 }
1858
1859 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1860 {
1861 return ptlock_ptr(pmd_to_page(pmd));
1862 }
1863
1864 static inline bool pgtable_pmd_page_ctor(struct page *page)
1865 {
1866 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1867 page->pmd_huge_pte = NULL;
1868 #endif
1869 return ptlock_init(page);
1870 }
1871
1872 static inline void pgtable_pmd_page_dtor(struct page *page)
1873 {
1874 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1875 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1876 #endif
1877 ptlock_free(page);
1878 }
1879
1880 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1881
1882 #else
1883
1884 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1885 {
1886 return &mm->page_table_lock;
1887 }
1888
1889 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1890 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1891
1892 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1893
1894 #endif
1895
1896 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1897 {
1898 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1899 spin_lock(ptl);
1900 return ptl;
1901 }
1902
1903 /*
1904 * No scalability reason to split PUD locks yet, but follow the same pattern
1905 * as the PMD locks to make it easier if we decide to. The VM should not be
1906 * considered ready to switch to split PUD locks yet; there may be places
1907 * which need to be converted from page_table_lock.
1908 */
1909 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1910 {
1911 return &mm->page_table_lock;
1912 }
1913
1914 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1915 {
1916 spinlock_t *ptl = pud_lockptr(mm, pud);
1917
1918 spin_lock(ptl);
1919 return ptl;
1920 }
1921
1922 extern void __init pagecache_init(void);
1923 extern void free_area_init(unsigned long * zones_size);
1924 extern void free_area_init_node(int nid, unsigned long * zones_size,
1925 unsigned long zone_start_pfn, unsigned long *zholes_size);
1926 extern void free_initmem(void);
1927
1928 /*
1929 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1930 * into the buddy system. The freed pages will be poisoned with pattern
1931 * "poison" if it's within range [0, UCHAR_MAX].
1932 * Return pages freed into the buddy system.
1933 */
1934 extern unsigned long free_reserved_area(void *start, void *end,
1935 int poison, char *s);
1936
1937 #ifdef CONFIG_HIGHMEM
1938 /*
1939 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1940 * and totalram_pages.
1941 */
1942 extern void free_highmem_page(struct page *page);
1943 #endif
1944
1945 extern void adjust_managed_page_count(struct page *page, long count);
1946 extern void mem_init_print_info(const char *str);
1947
1948 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1949
1950 /* Free the reserved page into the buddy system, so it gets managed. */
1951 static inline void __free_reserved_page(struct page *page)
1952 {
1953 ClearPageReserved(page);
1954 init_page_count(page);
1955 __free_page(page);
1956 }
1957
1958 static inline void free_reserved_page(struct page *page)
1959 {
1960 __free_reserved_page(page);
1961 adjust_managed_page_count(page, 1);
1962 }
1963
1964 static inline void mark_page_reserved(struct page *page)
1965 {
1966 SetPageReserved(page);
1967 adjust_managed_page_count(page, -1);
1968 }
1969
1970 /*
1971 * Default method to free all the __init memory into the buddy system.
1972 * The freed pages will be poisoned with pattern "poison" if it's within
1973 * range [0, UCHAR_MAX].
1974 * Return pages freed into the buddy system.
1975 */
1976 static inline unsigned long free_initmem_default(int poison)
1977 {
1978 extern char __init_begin[], __init_end[];
1979
1980 return free_reserved_area(&__init_begin, &__init_end,
1981 poison, "unused kernel");
1982 }
1983
1984 static inline unsigned long get_num_physpages(void)
1985 {
1986 int nid;
1987 unsigned long phys_pages = 0;
1988
1989 for_each_online_node(nid)
1990 phys_pages += node_present_pages(nid);
1991
1992 return phys_pages;
1993 }
1994
1995 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1996 /*
1997 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1998 * zones, allocate the backing mem_map and account for memory holes in a more
1999 * architecture independent manner. This is a substitute for creating the
2000 * zone_sizes[] and zholes_size[] arrays and passing them to
2001 * free_area_init_node()
2002 *
2003 * An architecture is expected to register range of page frames backed by
2004 * physical memory with memblock_add[_node]() before calling
2005 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2006 * usage, an architecture is expected to do something like
2007 *
2008 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2009 * max_highmem_pfn};
2010 * for_each_valid_physical_page_range()
2011 * memblock_add_node(base, size, nid)
2012 * free_area_init_nodes(max_zone_pfns);
2013 *
2014 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2015 * registered physical page range. Similarly
2016 * sparse_memory_present_with_active_regions() calls memory_present() for
2017 * each range when SPARSEMEM is enabled.
2018 *
2019 * See mm/page_alloc.c for more information on each function exposed by
2020 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2021 */
2022 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2023 unsigned long node_map_pfn_alignment(void);
2024 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2025 unsigned long end_pfn);
2026 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2027 unsigned long end_pfn);
2028 extern void get_pfn_range_for_nid(unsigned int nid,
2029 unsigned long *start_pfn, unsigned long *end_pfn);
2030 extern unsigned long find_min_pfn_with_active_regions(void);
2031 extern void free_bootmem_with_active_regions(int nid,
2032 unsigned long max_low_pfn);
2033 extern void sparse_memory_present_with_active_regions(int nid);
2034
2035 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2036
2037 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2038 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2039 static inline int __early_pfn_to_nid(unsigned long pfn,
2040 struct mminit_pfnnid_cache *state)
2041 {
2042 return 0;
2043 }
2044 #else
2045 /* please see mm/page_alloc.c */
2046 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2047 /* there is a per-arch backend function. */
2048 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2049 struct mminit_pfnnid_cache *state);
2050 #endif
2051
2052 #ifdef CONFIG_HAVE_MEMBLOCK
2053 void zero_resv_unavail(void);
2054 #else
2055 static inline void zero_resv_unavail(void) {}
2056 #endif
2057
2058 extern void set_dma_reserve(unsigned long new_dma_reserve);
2059 extern void memmap_init_zone(unsigned long, int, unsigned long,
2060 unsigned long, enum memmap_context);
2061 extern void setup_per_zone_wmarks(void);
2062 extern int __meminit init_per_zone_wmark_min(void);
2063 extern void mem_init(void);
2064 extern void __init mmap_init(void);
2065 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2066 extern long si_mem_available(void);
2067 extern void si_meminfo(struct sysinfo * val);
2068 extern void si_meminfo_node(struct sysinfo *val, int nid);
2069 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2070 extern unsigned long arch_reserved_kernel_pages(void);
2071 #endif
2072
2073 extern __printf(3, 4)
2074 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2075
2076 extern void setup_per_cpu_pageset(void);
2077
2078 extern void zone_pcp_update(struct zone *zone);
2079 extern void zone_pcp_reset(struct zone *zone);
2080
2081 /* page_alloc.c */
2082 extern int min_free_kbytes;
2083 extern int watermark_scale_factor;
2084
2085 /* nommu.c */
2086 extern atomic_long_t mmap_pages_allocated;
2087 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2088
2089 /* interval_tree.c */
2090 void vma_interval_tree_insert(struct vm_area_struct *node,
2091 struct rb_root_cached *root);
2092 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2093 struct vm_area_struct *prev,
2094 struct rb_root_cached *root);
2095 void vma_interval_tree_remove(struct vm_area_struct *node,
2096 struct rb_root_cached *root);
2097 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2098 unsigned long start, unsigned long last);
2099 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2100 unsigned long start, unsigned long last);
2101
2102 #define vma_interval_tree_foreach(vma, root, start, last) \
2103 for (vma = vma_interval_tree_iter_first(root, start, last); \
2104 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2105
2106 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2107 struct rb_root_cached *root);
2108 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2109 struct rb_root_cached *root);
2110 struct anon_vma_chain *
2111 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2112 unsigned long start, unsigned long last);
2113 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2114 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2115 #ifdef CONFIG_DEBUG_VM_RB
2116 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2117 #endif
2118
2119 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2120 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2121 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2122
2123 /* mmap.c */
2124 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2125 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2126 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2127 struct vm_area_struct *expand);
2128 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2129 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2130 {
2131 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2132 }
2133 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2134 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2135 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2136 struct mempolicy *, struct vm_userfaultfd_ctx);
2137 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2138 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2139 unsigned long addr, int new_below);
2140 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2141 unsigned long addr, int new_below);
2142 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2143 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2144 struct rb_node **, struct rb_node *);
2145 extern void unlink_file_vma(struct vm_area_struct *);
2146 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2147 unsigned long addr, unsigned long len, pgoff_t pgoff,
2148 bool *need_rmap_locks);
2149 extern void exit_mmap(struct mm_struct *);
2150
2151 static inline int check_data_rlimit(unsigned long rlim,
2152 unsigned long new,
2153 unsigned long start,
2154 unsigned long end_data,
2155 unsigned long start_data)
2156 {
2157 if (rlim < RLIM_INFINITY) {
2158 if (((new - start) + (end_data - start_data)) > rlim)
2159 return -ENOSPC;
2160 }
2161
2162 return 0;
2163 }
2164
2165 extern int mm_take_all_locks(struct mm_struct *mm);
2166 extern void mm_drop_all_locks(struct mm_struct *mm);
2167
2168 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2169 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2170 extern struct file *get_task_exe_file(struct task_struct *task);
2171
2172 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2173 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2174
2175 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2176 const struct vm_special_mapping *sm);
2177 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2178 unsigned long addr, unsigned long len,
2179 unsigned long flags,
2180 const struct vm_special_mapping *spec);
2181 /* This is an obsolete alternative to _install_special_mapping. */
2182 extern int install_special_mapping(struct mm_struct *mm,
2183 unsigned long addr, unsigned long len,
2184 unsigned long flags, struct page **pages);
2185
2186 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2187
2188 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2189 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2190 struct list_head *uf);
2191 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2192 unsigned long len, unsigned long prot, unsigned long flags,
2193 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2194 struct list_head *uf);
2195 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2196 struct list_head *uf);
2197
2198 static inline unsigned long
2199 do_mmap_pgoff(struct file *file, unsigned long addr,
2200 unsigned long len, unsigned long prot, unsigned long flags,
2201 unsigned long pgoff, unsigned long *populate,
2202 struct list_head *uf)
2203 {
2204 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2205 }
2206
2207 #ifdef CONFIG_MMU
2208 extern int __mm_populate(unsigned long addr, unsigned long len,
2209 int ignore_errors);
2210 static inline void mm_populate(unsigned long addr, unsigned long len)
2211 {
2212 /* Ignore errors */
2213 (void) __mm_populate(addr, len, 1);
2214 }
2215 #else
2216 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2217 #endif
2218
2219 /* These take the mm semaphore themselves */
2220 extern int __must_check vm_brk(unsigned long, unsigned long);
2221 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2222 extern int vm_munmap(unsigned long, size_t);
2223 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2224 unsigned long, unsigned long,
2225 unsigned long, unsigned long);
2226
2227 struct vm_unmapped_area_info {
2228 #define VM_UNMAPPED_AREA_TOPDOWN 1
2229 unsigned long flags;
2230 unsigned long length;
2231 unsigned long low_limit;
2232 unsigned long high_limit;
2233 unsigned long align_mask;
2234 unsigned long align_offset;
2235 };
2236
2237 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2238 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2239
2240 /*
2241 * Search for an unmapped address range.
2242 *
2243 * We are looking for a range that:
2244 * - does not intersect with any VMA;
2245 * - is contained within the [low_limit, high_limit) interval;
2246 * - is at least the desired size.
2247 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2248 */
2249 static inline unsigned long
2250 vm_unmapped_area(struct vm_unmapped_area_info *info)
2251 {
2252 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2253 return unmapped_area_topdown(info);
2254 else
2255 return unmapped_area(info);
2256 }
2257
2258 /* truncate.c */
2259 extern void truncate_inode_pages(struct address_space *, loff_t);
2260 extern void truncate_inode_pages_range(struct address_space *,
2261 loff_t lstart, loff_t lend);
2262 extern void truncate_inode_pages_final(struct address_space *);
2263
2264 /* generic vm_area_ops exported for stackable file systems */
2265 extern int filemap_fault(struct vm_fault *vmf);
2266 extern void filemap_map_pages(struct vm_fault *vmf,
2267 pgoff_t start_pgoff, pgoff_t end_pgoff);
2268 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2269
2270 /* mm/page-writeback.c */
2271 int __must_check write_one_page(struct page *page);
2272 void task_dirty_inc(struct task_struct *tsk);
2273
2274 /* readahead.c */
2275 #define VM_MAX_READAHEAD 128 /* kbytes */
2276 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2277
2278 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2279 pgoff_t offset, unsigned long nr_to_read);
2280
2281 void page_cache_sync_readahead(struct address_space *mapping,
2282 struct file_ra_state *ra,
2283 struct file *filp,
2284 pgoff_t offset,
2285 unsigned long size);
2286
2287 void page_cache_async_readahead(struct address_space *mapping,
2288 struct file_ra_state *ra,
2289 struct file *filp,
2290 struct page *pg,
2291 pgoff_t offset,
2292 unsigned long size);
2293
2294 extern unsigned long stack_guard_gap;
2295 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2296 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2297
2298 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2299 extern int expand_downwards(struct vm_area_struct *vma,
2300 unsigned long address);
2301 #if VM_GROWSUP
2302 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2303 #else
2304 #define expand_upwards(vma, address) (0)
2305 #endif
2306
2307 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2308 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2309 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2310 struct vm_area_struct **pprev);
2311
2312 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2313 NULL if none. Assume start_addr < end_addr. */
2314 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2315 {
2316 struct vm_area_struct * vma = find_vma(mm,start_addr);
2317
2318 if (vma && end_addr <= vma->vm_start)
2319 vma = NULL;
2320 return vma;
2321 }
2322
2323 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2324 {
2325 unsigned long vm_start = vma->vm_start;
2326
2327 if (vma->vm_flags & VM_GROWSDOWN) {
2328 vm_start -= stack_guard_gap;
2329 if (vm_start > vma->vm_start)
2330 vm_start = 0;
2331 }
2332 return vm_start;
2333 }
2334
2335 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2336 {
2337 unsigned long vm_end = vma->vm_end;
2338
2339 if (vma->vm_flags & VM_GROWSUP) {
2340 vm_end += stack_guard_gap;
2341 if (vm_end < vma->vm_end)
2342 vm_end = -PAGE_SIZE;
2343 }
2344 return vm_end;
2345 }
2346
2347 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2348 {
2349 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2350 }
2351
2352 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2353 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2354 unsigned long vm_start, unsigned long vm_end)
2355 {
2356 struct vm_area_struct *vma = find_vma(mm, vm_start);
2357
2358 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2359 vma = NULL;
2360
2361 return vma;
2362 }
2363
2364 #ifdef CONFIG_MMU
2365 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2366 void vma_set_page_prot(struct vm_area_struct *vma);
2367 #else
2368 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2369 {
2370 return __pgprot(0);
2371 }
2372 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2373 {
2374 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2375 }
2376 #endif
2377
2378 #ifdef CONFIG_NUMA_BALANCING
2379 unsigned long change_prot_numa(struct vm_area_struct *vma,
2380 unsigned long start, unsigned long end);
2381 #endif
2382
2383 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2384 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2385 unsigned long pfn, unsigned long size, pgprot_t);
2386 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2387 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2388 unsigned long pfn);
2389 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2390 unsigned long pfn, pgprot_t pgprot);
2391 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2392 pfn_t pfn);
2393 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2394 pfn_t pfn);
2395 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2396
2397
2398 struct page *follow_page_mask(struct vm_area_struct *vma,
2399 unsigned long address, unsigned int foll_flags,
2400 unsigned int *page_mask);
2401
2402 static inline struct page *follow_page(struct vm_area_struct *vma,
2403 unsigned long address, unsigned int foll_flags)
2404 {
2405 unsigned int unused_page_mask;
2406 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2407 }
2408
2409 #define FOLL_WRITE 0x01 /* check pte is writable */
2410 #define FOLL_TOUCH 0x02 /* mark page accessed */
2411 #define FOLL_GET 0x04 /* do get_page on page */
2412 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2413 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2414 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2415 * and return without waiting upon it */
2416 #define FOLL_POPULATE 0x40 /* fault in page */
2417 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2418 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2419 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2420 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2421 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2422 #define FOLL_MLOCK 0x1000 /* lock present pages */
2423 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2424 #define FOLL_COW 0x4000 /* internal GUP flag */
2425
2426 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2427 {
2428 if (vm_fault & VM_FAULT_OOM)
2429 return -ENOMEM;
2430 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2431 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2432 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2433 return -EFAULT;
2434 return 0;
2435 }
2436
2437 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2438 void *data);
2439 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2440 unsigned long size, pte_fn_t fn, void *data);
2441
2442
2443 #ifdef CONFIG_PAGE_POISONING
2444 extern bool page_poisoning_enabled(void);
2445 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2446 extern bool page_is_poisoned(struct page *page);
2447 #else
2448 static inline bool page_poisoning_enabled(void) { return false; }
2449 static inline void kernel_poison_pages(struct page *page, int numpages,
2450 int enable) { }
2451 static inline bool page_is_poisoned(struct page *page) { return false; }
2452 #endif
2453
2454 #ifdef CONFIG_DEBUG_PAGEALLOC
2455 extern bool _debug_pagealloc_enabled;
2456 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2457
2458 static inline bool debug_pagealloc_enabled(void)
2459 {
2460 return _debug_pagealloc_enabled;
2461 }
2462
2463 static inline void
2464 kernel_map_pages(struct page *page, int numpages, int enable)
2465 {
2466 if (!debug_pagealloc_enabled())
2467 return;
2468
2469 __kernel_map_pages(page, numpages, enable);
2470 }
2471 #ifdef CONFIG_HIBERNATION
2472 extern bool kernel_page_present(struct page *page);
2473 #endif /* CONFIG_HIBERNATION */
2474 #else /* CONFIG_DEBUG_PAGEALLOC */
2475 static inline void
2476 kernel_map_pages(struct page *page, int numpages, int enable) {}
2477 #ifdef CONFIG_HIBERNATION
2478 static inline bool kernel_page_present(struct page *page) { return true; }
2479 #endif /* CONFIG_HIBERNATION */
2480 static inline bool debug_pagealloc_enabled(void)
2481 {
2482 return false;
2483 }
2484 #endif /* CONFIG_DEBUG_PAGEALLOC */
2485
2486 #ifdef __HAVE_ARCH_GATE_AREA
2487 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2488 extern int in_gate_area_no_mm(unsigned long addr);
2489 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2490 #else
2491 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2492 {
2493 return NULL;
2494 }
2495 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2496 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2497 {
2498 return 0;
2499 }
2500 #endif /* __HAVE_ARCH_GATE_AREA */
2501
2502 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2503
2504 #ifdef CONFIG_SYSCTL
2505 extern int sysctl_drop_caches;
2506 int drop_caches_sysctl_handler(struct ctl_table *, int,
2507 void __user *, size_t *, loff_t *);
2508 #endif
2509
2510 void drop_slab(void);
2511 void drop_slab_node(int nid);
2512
2513 #ifndef CONFIG_MMU
2514 #define randomize_va_space 0
2515 #else
2516 extern int randomize_va_space;
2517 #endif
2518
2519 const char * arch_vma_name(struct vm_area_struct *vma);
2520 void print_vma_addr(char *prefix, unsigned long rip);
2521
2522 void sparse_mem_maps_populate_node(struct page **map_map,
2523 unsigned long pnum_begin,
2524 unsigned long pnum_end,
2525 unsigned long map_count,
2526 int nodeid);
2527
2528 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2529 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2530 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2531 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2532 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2533 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2534 void *vmemmap_alloc_block(unsigned long size, int node);
2535 struct vmem_altmap;
2536 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2537 struct vmem_altmap *altmap);
2538 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2539 {
2540 return __vmemmap_alloc_block_buf(size, node, NULL);
2541 }
2542
2543 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2544 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2545 int node);
2546 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2547 void vmemmap_populate_print_last(void);
2548 #ifdef CONFIG_MEMORY_HOTPLUG
2549 void vmemmap_free(unsigned long start, unsigned long end);
2550 #endif
2551 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2552 unsigned long nr_pages);
2553
2554 enum mf_flags {
2555 MF_COUNT_INCREASED = 1 << 0,
2556 MF_ACTION_REQUIRED = 1 << 1,
2557 MF_MUST_KILL = 1 << 2,
2558 MF_SOFT_OFFLINE = 1 << 3,
2559 };
2560 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2561 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2562 extern int unpoison_memory(unsigned long pfn);
2563 extern int get_hwpoison_page(struct page *page);
2564 #define put_hwpoison_page(page) put_page(page)
2565 extern int sysctl_memory_failure_early_kill;
2566 extern int sysctl_memory_failure_recovery;
2567 extern void shake_page(struct page *p, int access);
2568 extern atomic_long_t num_poisoned_pages;
2569 extern int soft_offline_page(struct page *page, int flags);
2570
2571
2572 /*
2573 * Error handlers for various types of pages.
2574 */
2575 enum mf_result {
2576 MF_IGNORED, /* Error: cannot be handled */
2577 MF_FAILED, /* Error: handling failed */
2578 MF_DELAYED, /* Will be handled later */
2579 MF_RECOVERED, /* Successfully recovered */
2580 };
2581
2582 enum mf_action_page_type {
2583 MF_MSG_KERNEL,
2584 MF_MSG_KERNEL_HIGH_ORDER,
2585 MF_MSG_SLAB,
2586 MF_MSG_DIFFERENT_COMPOUND,
2587 MF_MSG_POISONED_HUGE,
2588 MF_MSG_HUGE,
2589 MF_MSG_FREE_HUGE,
2590 MF_MSG_UNMAP_FAILED,
2591 MF_MSG_DIRTY_SWAPCACHE,
2592 MF_MSG_CLEAN_SWAPCACHE,
2593 MF_MSG_DIRTY_MLOCKED_LRU,
2594 MF_MSG_CLEAN_MLOCKED_LRU,
2595 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2596 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2597 MF_MSG_DIRTY_LRU,
2598 MF_MSG_CLEAN_LRU,
2599 MF_MSG_TRUNCATED_LRU,
2600 MF_MSG_BUDDY,
2601 MF_MSG_BUDDY_2ND,
2602 MF_MSG_UNKNOWN,
2603 };
2604
2605 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2606 extern void clear_huge_page(struct page *page,
2607 unsigned long addr_hint,
2608 unsigned int pages_per_huge_page);
2609 extern void copy_user_huge_page(struct page *dst, struct page *src,
2610 unsigned long addr, struct vm_area_struct *vma,
2611 unsigned int pages_per_huge_page);
2612 extern long copy_huge_page_from_user(struct page *dst_page,
2613 const void __user *usr_src,
2614 unsigned int pages_per_huge_page,
2615 bool allow_pagefault);
2616 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2617
2618 extern struct page_ext_operations debug_guardpage_ops;
2619
2620 #ifdef CONFIG_DEBUG_PAGEALLOC
2621 extern unsigned int _debug_guardpage_minorder;
2622 extern bool _debug_guardpage_enabled;
2623
2624 static inline unsigned int debug_guardpage_minorder(void)
2625 {
2626 return _debug_guardpage_minorder;
2627 }
2628
2629 static inline bool debug_guardpage_enabled(void)
2630 {
2631 return _debug_guardpage_enabled;
2632 }
2633
2634 static inline bool page_is_guard(struct page *page)
2635 {
2636 struct page_ext *page_ext;
2637
2638 if (!debug_guardpage_enabled())
2639 return false;
2640
2641 page_ext = lookup_page_ext(page);
2642 if (unlikely(!page_ext))
2643 return false;
2644
2645 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2646 }
2647 #else
2648 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2649 static inline bool debug_guardpage_enabled(void) { return false; }
2650 static inline bool page_is_guard(struct page *page) { return false; }
2651 #endif /* CONFIG_DEBUG_PAGEALLOC */
2652
2653 #if MAX_NUMNODES > 1
2654 void __init setup_nr_node_ids(void);
2655 #else
2656 static inline void setup_nr_node_ids(void) {}
2657 #endif
2658
2659 #endif /* __KERNEL__ */
2660 #endif /* _LINUX_MM_H */