xref: /linux-6.15/include/linux/mm.h (revision db181ce0)
1 #ifndef _LINUX_MM_H
2 #define _LINUX_MM_H
3 
4 #include <linux/errno.h>
5 
6 #ifdef __KERNEL__
7 
8 #include <linux/mmdebug.h>
9 #include <linux/gfp.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 
22 struct mempolicy;
23 struct anon_vma;
24 struct anon_vma_chain;
25 struct file_ra_state;
26 struct user_struct;
27 struct writeback_control;
28 
29 #ifndef CONFIG_NEED_MULTIPLE_NODES	/* Don't use mapnrs, do it properly */
30 extern unsigned long max_mapnr;
31 
32 static inline void set_max_mapnr(unsigned long limit)
33 {
34 	max_mapnr = limit;
35 }
36 #else
37 static inline void set_max_mapnr(unsigned long limit) { }
38 #endif
39 
40 extern unsigned long totalram_pages;
41 extern void * high_memory;
42 extern int page_cluster;
43 
44 #ifdef CONFIG_SYSCTL
45 extern int sysctl_legacy_va_layout;
46 #else
47 #define sysctl_legacy_va_layout 0
48 #endif
49 
50 #include <asm/page.h>
51 #include <asm/pgtable.h>
52 #include <asm/processor.h>
53 
54 #ifndef __pa_symbol
55 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
56 #endif
57 
58 extern unsigned long sysctl_user_reserve_kbytes;
59 extern unsigned long sysctl_admin_reserve_kbytes;
60 
61 extern int sysctl_overcommit_memory;
62 extern int sysctl_overcommit_ratio;
63 extern unsigned long sysctl_overcommit_kbytes;
64 
65 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
66 				    size_t *, loff_t *);
67 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
68 				    size_t *, loff_t *);
69 
70 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
71 
72 /* to align the pointer to the (next) page boundary */
73 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
74 
75 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
76 #define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
77 
78 /*
79  * Linux kernel virtual memory manager primitives.
80  * The idea being to have a "virtual" mm in the same way
81  * we have a virtual fs - giving a cleaner interface to the
82  * mm details, and allowing different kinds of memory mappings
83  * (from shared memory to executable loading to arbitrary
84  * mmap() functions).
85  */
86 
87 extern struct kmem_cache *vm_area_cachep;
88 
89 #ifndef CONFIG_MMU
90 extern struct rb_root nommu_region_tree;
91 extern struct rw_semaphore nommu_region_sem;
92 
93 extern unsigned int kobjsize(const void *objp);
94 #endif
95 
96 /*
97  * vm_flags in vm_area_struct, see mm_types.h.
98  */
99 #define VM_NONE		0x00000000
100 
101 #define VM_READ		0x00000001	/* currently active flags */
102 #define VM_WRITE	0x00000002
103 #define VM_EXEC		0x00000004
104 #define VM_SHARED	0x00000008
105 
106 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
107 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
108 #define VM_MAYWRITE	0x00000020
109 #define VM_MAYEXEC	0x00000040
110 #define VM_MAYSHARE	0x00000080
111 
112 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
113 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
114 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
115 
116 #define VM_LOCKED	0x00002000
117 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
118 
119 					/* Used by sys_madvise() */
120 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
121 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
122 
123 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
124 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
125 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
126 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
127 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
128 #define VM_NONLINEAR	0x00800000	/* Is non-linear (remap_file_pages) */
129 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
130 #define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
131 
132 #ifdef CONFIG_MEM_SOFT_DIRTY
133 # define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
134 #else
135 # define VM_SOFTDIRTY	0
136 #endif
137 
138 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
139 #define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
140 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
141 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
142 
143 #if defined(CONFIG_X86)
144 # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
145 #elif defined(CONFIG_PPC)
146 # define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
147 #elif defined(CONFIG_PARISC)
148 # define VM_GROWSUP	VM_ARCH_1
149 #elif defined(CONFIG_METAG)
150 # define VM_GROWSUP	VM_ARCH_1
151 #elif defined(CONFIG_IA64)
152 # define VM_GROWSUP	VM_ARCH_1
153 #elif !defined(CONFIG_MMU)
154 # define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
155 #endif
156 
157 #ifndef VM_GROWSUP
158 # define VM_GROWSUP	VM_NONE
159 #endif
160 
161 /* Bits set in the VMA until the stack is in its final location */
162 #define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
163 
164 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
165 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
166 #endif
167 
168 #ifdef CONFIG_STACK_GROWSUP
169 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
170 #else
171 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
172 #endif
173 
174 /*
175  * Special vmas that are non-mergable, non-mlock()able.
176  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
177  */
178 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
179 
180 /* This mask defines which mm->def_flags a process can inherit its parent */
181 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
182 
183 /*
184  * mapping from the currently active vm_flags protection bits (the
185  * low four bits) to a page protection mask..
186  */
187 extern pgprot_t protection_map[16];
188 
189 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
190 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
191 #define FAULT_FLAG_MKWRITE	0x04	/* Fault was mkwrite of existing pte */
192 #define FAULT_FLAG_ALLOW_RETRY	0x08	/* Retry fault if blocking */
193 #define FAULT_FLAG_RETRY_NOWAIT	0x10	/* Don't drop mmap_sem and wait when retrying */
194 #define FAULT_FLAG_KILLABLE	0x20	/* The fault task is in SIGKILL killable region */
195 #define FAULT_FLAG_TRIED	0x40	/* second try */
196 #define FAULT_FLAG_USER		0x80	/* The fault originated in userspace */
197 
198 /*
199  * vm_fault is filled by the the pagefault handler and passed to the vma's
200  * ->fault function. The vma's ->fault is responsible for returning a bitmask
201  * of VM_FAULT_xxx flags that give details about how the fault was handled.
202  *
203  * pgoff should be used in favour of virtual_address, if possible. If pgoff
204  * is used, one may implement ->remap_pages to get nonlinear mapping support.
205  */
206 struct vm_fault {
207 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
208 	pgoff_t pgoff;			/* Logical page offset based on vma */
209 	void __user *virtual_address;	/* Faulting virtual address */
210 
211 	struct page *page;		/* ->fault handlers should return a
212 					 * page here, unless VM_FAULT_NOPAGE
213 					 * is set (which is also implied by
214 					 * VM_FAULT_ERROR).
215 					 */
216 	/* for ->map_pages() only */
217 	pgoff_t max_pgoff;		/* map pages for offset from pgoff till
218 					 * max_pgoff inclusive */
219 	pte_t *pte;			/* pte entry associated with ->pgoff */
220 };
221 
222 /*
223  * These are the virtual MM functions - opening of an area, closing and
224  * unmapping it (needed to keep files on disk up-to-date etc), pointer
225  * to the functions called when a no-page or a wp-page exception occurs.
226  */
227 struct vm_operations_struct {
228 	void (*open)(struct vm_area_struct * area);
229 	void (*close)(struct vm_area_struct * area);
230 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
231 	void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
232 
233 	/* notification that a previously read-only page is about to become
234 	 * writable, if an error is returned it will cause a SIGBUS */
235 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
236 
237 	/* called by access_process_vm when get_user_pages() fails, typically
238 	 * for use by special VMAs that can switch between memory and hardware
239 	 */
240 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
241 		      void *buf, int len, int write);
242 
243 	/* Called by the /proc/PID/maps code to ask the vma whether it
244 	 * has a special name.  Returning non-NULL will also cause this
245 	 * vma to be dumped unconditionally. */
246 	const char *(*name)(struct vm_area_struct *vma);
247 
248 #ifdef CONFIG_NUMA
249 	/*
250 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
251 	 * to hold the policy upon return.  Caller should pass NULL @new to
252 	 * remove a policy and fall back to surrounding context--i.e. do not
253 	 * install a MPOL_DEFAULT policy, nor the task or system default
254 	 * mempolicy.
255 	 */
256 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
257 
258 	/*
259 	 * get_policy() op must add reference [mpol_get()] to any policy at
260 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
261 	 * in mm/mempolicy.c will do this automatically.
262 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
263 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
264 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
265 	 * must return NULL--i.e., do not "fallback" to task or system default
266 	 * policy.
267 	 */
268 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
269 					unsigned long addr);
270 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
271 		const nodemask_t *to, unsigned long flags);
272 #endif
273 	/* called by sys_remap_file_pages() to populate non-linear mapping */
274 	int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
275 			   unsigned long size, pgoff_t pgoff);
276 };
277 
278 struct mmu_gather;
279 struct inode;
280 
281 #define page_private(page)		((page)->private)
282 #define set_page_private(page, v)	((page)->private = (v))
283 
284 /* It's valid only if the page is free path or free_list */
285 static inline void set_freepage_migratetype(struct page *page, int migratetype)
286 {
287 	page->index = migratetype;
288 }
289 
290 /* It's valid only if the page is free path or free_list */
291 static inline int get_freepage_migratetype(struct page *page)
292 {
293 	return page->index;
294 }
295 
296 /*
297  * FIXME: take this include out, include page-flags.h in
298  * files which need it (119 of them)
299  */
300 #include <linux/page-flags.h>
301 #include <linux/huge_mm.h>
302 
303 /*
304  * Methods to modify the page usage count.
305  *
306  * What counts for a page usage:
307  * - cache mapping   (page->mapping)
308  * - private data    (page->private)
309  * - page mapped in a task's page tables, each mapping
310  *   is counted separately
311  *
312  * Also, many kernel routines increase the page count before a critical
313  * routine so they can be sure the page doesn't go away from under them.
314  */
315 
316 /*
317  * Drop a ref, return true if the refcount fell to zero (the page has no users)
318  */
319 static inline int put_page_testzero(struct page *page)
320 {
321 	VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
322 	return atomic_dec_and_test(&page->_count);
323 }
324 
325 /*
326  * Try to grab a ref unless the page has a refcount of zero, return false if
327  * that is the case.
328  * This can be called when MMU is off so it must not access
329  * any of the virtual mappings.
330  */
331 static inline int get_page_unless_zero(struct page *page)
332 {
333 	return atomic_inc_not_zero(&page->_count);
334 }
335 
336 /*
337  * Try to drop a ref unless the page has a refcount of one, return false if
338  * that is the case.
339  * This is to make sure that the refcount won't become zero after this drop.
340  * This can be called when MMU is off so it must not access
341  * any of the virtual mappings.
342  */
343 static inline int put_page_unless_one(struct page *page)
344 {
345 	return atomic_add_unless(&page->_count, -1, 1);
346 }
347 
348 extern int page_is_ram(unsigned long pfn);
349 
350 /* Support for virtually mapped pages */
351 struct page *vmalloc_to_page(const void *addr);
352 unsigned long vmalloc_to_pfn(const void *addr);
353 
354 /*
355  * Determine if an address is within the vmalloc range
356  *
357  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
358  * is no special casing required.
359  */
360 static inline int is_vmalloc_addr(const void *x)
361 {
362 #ifdef CONFIG_MMU
363 	unsigned long addr = (unsigned long)x;
364 
365 	return addr >= VMALLOC_START && addr < VMALLOC_END;
366 #else
367 	return 0;
368 #endif
369 }
370 #ifdef CONFIG_MMU
371 extern int is_vmalloc_or_module_addr(const void *x);
372 #else
373 static inline int is_vmalloc_or_module_addr(const void *x)
374 {
375 	return 0;
376 }
377 #endif
378 
379 extern void kvfree(const void *addr);
380 
381 static inline void compound_lock(struct page *page)
382 {
383 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
384 	VM_BUG_ON_PAGE(PageSlab(page), page);
385 	bit_spin_lock(PG_compound_lock, &page->flags);
386 #endif
387 }
388 
389 static inline void compound_unlock(struct page *page)
390 {
391 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
392 	VM_BUG_ON_PAGE(PageSlab(page), page);
393 	bit_spin_unlock(PG_compound_lock, &page->flags);
394 #endif
395 }
396 
397 static inline unsigned long compound_lock_irqsave(struct page *page)
398 {
399 	unsigned long uninitialized_var(flags);
400 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
401 	local_irq_save(flags);
402 	compound_lock(page);
403 #endif
404 	return flags;
405 }
406 
407 static inline void compound_unlock_irqrestore(struct page *page,
408 					      unsigned long flags)
409 {
410 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
411 	compound_unlock(page);
412 	local_irq_restore(flags);
413 #endif
414 }
415 
416 static inline struct page *compound_head_by_tail(struct page *tail)
417 {
418 	struct page *head = tail->first_page;
419 
420 	/*
421 	 * page->first_page may be a dangling pointer to an old
422 	 * compound page, so recheck that it is still a tail
423 	 * page before returning.
424 	 */
425 	smp_rmb();
426 	if (likely(PageTail(tail)))
427 		return head;
428 	return tail;
429 }
430 
431 static inline struct page *compound_head(struct page *page)
432 {
433 	if (unlikely(PageTail(page)))
434 		return compound_head_by_tail(page);
435 	return page;
436 }
437 
438 /*
439  * The atomic page->_mapcount, starts from -1: so that transitions
440  * both from it and to it can be tracked, using atomic_inc_and_test
441  * and atomic_add_negative(-1).
442  */
443 static inline void page_mapcount_reset(struct page *page)
444 {
445 	atomic_set(&(page)->_mapcount, -1);
446 }
447 
448 static inline int page_mapcount(struct page *page)
449 {
450 	return atomic_read(&(page)->_mapcount) + 1;
451 }
452 
453 static inline int page_count(struct page *page)
454 {
455 	return atomic_read(&compound_head(page)->_count);
456 }
457 
458 #ifdef CONFIG_HUGETLB_PAGE
459 extern int PageHeadHuge(struct page *page_head);
460 #else /* CONFIG_HUGETLB_PAGE */
461 static inline int PageHeadHuge(struct page *page_head)
462 {
463 	return 0;
464 }
465 #endif /* CONFIG_HUGETLB_PAGE */
466 
467 static inline bool __compound_tail_refcounted(struct page *page)
468 {
469 	return !PageSlab(page) && !PageHeadHuge(page);
470 }
471 
472 /*
473  * This takes a head page as parameter and tells if the
474  * tail page reference counting can be skipped.
475  *
476  * For this to be safe, PageSlab and PageHeadHuge must remain true on
477  * any given page where they return true here, until all tail pins
478  * have been released.
479  */
480 static inline bool compound_tail_refcounted(struct page *page)
481 {
482 	VM_BUG_ON_PAGE(!PageHead(page), page);
483 	return __compound_tail_refcounted(page);
484 }
485 
486 static inline void get_huge_page_tail(struct page *page)
487 {
488 	/*
489 	 * __split_huge_page_refcount() cannot run from under us.
490 	 */
491 	VM_BUG_ON_PAGE(!PageTail(page), page);
492 	VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
493 	VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
494 	if (compound_tail_refcounted(page->first_page))
495 		atomic_inc(&page->_mapcount);
496 }
497 
498 extern bool __get_page_tail(struct page *page);
499 
500 static inline void get_page(struct page *page)
501 {
502 	if (unlikely(PageTail(page)))
503 		if (likely(__get_page_tail(page)))
504 			return;
505 	/*
506 	 * Getting a normal page or the head of a compound page
507 	 * requires to already have an elevated page->_count.
508 	 */
509 	VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
510 	atomic_inc(&page->_count);
511 }
512 
513 static inline struct page *virt_to_head_page(const void *x)
514 {
515 	struct page *page = virt_to_page(x);
516 	return compound_head(page);
517 }
518 
519 /*
520  * Setup the page count before being freed into the page allocator for
521  * the first time (boot or memory hotplug)
522  */
523 static inline void init_page_count(struct page *page)
524 {
525 	atomic_set(&page->_count, 1);
526 }
527 
528 /*
529  * PageBuddy() indicate that the page is free and in the buddy system
530  * (see mm/page_alloc.c).
531  *
532  * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
533  * -2 so that an underflow of the page_mapcount() won't be mistaken
534  * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
535  * efficiently by most CPU architectures.
536  */
537 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
538 
539 static inline int PageBuddy(struct page *page)
540 {
541 	return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
542 }
543 
544 static inline void __SetPageBuddy(struct page *page)
545 {
546 	VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
547 	atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
548 }
549 
550 static inline void __ClearPageBuddy(struct page *page)
551 {
552 	VM_BUG_ON_PAGE(!PageBuddy(page), page);
553 	atomic_set(&page->_mapcount, -1);
554 }
555 
556 void put_page(struct page *page);
557 void put_pages_list(struct list_head *pages);
558 
559 void split_page(struct page *page, unsigned int order);
560 int split_free_page(struct page *page);
561 
562 /*
563  * Compound pages have a destructor function.  Provide a
564  * prototype for that function and accessor functions.
565  * These are _only_ valid on the head of a PG_compound page.
566  */
567 typedef void compound_page_dtor(struct page *);
568 
569 static inline void set_compound_page_dtor(struct page *page,
570 						compound_page_dtor *dtor)
571 {
572 	page[1].lru.next = (void *)dtor;
573 }
574 
575 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
576 {
577 	return (compound_page_dtor *)page[1].lru.next;
578 }
579 
580 static inline int compound_order(struct page *page)
581 {
582 	if (!PageHead(page))
583 		return 0;
584 	return (unsigned long)page[1].lru.prev;
585 }
586 
587 static inline void set_compound_order(struct page *page, unsigned long order)
588 {
589 	page[1].lru.prev = (void *)order;
590 }
591 
592 #ifdef CONFIG_MMU
593 /*
594  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
595  * servicing faults for write access.  In the normal case, do always want
596  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
597  * that do not have writing enabled, when used by access_process_vm.
598  */
599 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
600 {
601 	if (likely(vma->vm_flags & VM_WRITE))
602 		pte = pte_mkwrite(pte);
603 	return pte;
604 }
605 
606 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
607 		struct page *page, pte_t *pte, bool write, bool anon);
608 #endif
609 
610 /*
611  * Multiple processes may "see" the same page. E.g. for untouched
612  * mappings of /dev/null, all processes see the same page full of
613  * zeroes, and text pages of executables and shared libraries have
614  * only one copy in memory, at most, normally.
615  *
616  * For the non-reserved pages, page_count(page) denotes a reference count.
617  *   page_count() == 0 means the page is free. page->lru is then used for
618  *   freelist management in the buddy allocator.
619  *   page_count() > 0  means the page has been allocated.
620  *
621  * Pages are allocated by the slab allocator in order to provide memory
622  * to kmalloc and kmem_cache_alloc. In this case, the management of the
623  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
624  * unless a particular usage is carefully commented. (the responsibility of
625  * freeing the kmalloc memory is the caller's, of course).
626  *
627  * A page may be used by anyone else who does a __get_free_page().
628  * In this case, page_count still tracks the references, and should only
629  * be used through the normal accessor functions. The top bits of page->flags
630  * and page->virtual store page management information, but all other fields
631  * are unused and could be used privately, carefully. The management of this
632  * page is the responsibility of the one who allocated it, and those who have
633  * subsequently been given references to it.
634  *
635  * The other pages (we may call them "pagecache pages") are completely
636  * managed by the Linux memory manager: I/O, buffers, swapping etc.
637  * The following discussion applies only to them.
638  *
639  * A pagecache page contains an opaque `private' member, which belongs to the
640  * page's address_space. Usually, this is the address of a circular list of
641  * the page's disk buffers. PG_private must be set to tell the VM to call
642  * into the filesystem to release these pages.
643  *
644  * A page may belong to an inode's memory mapping. In this case, page->mapping
645  * is the pointer to the inode, and page->index is the file offset of the page,
646  * in units of PAGE_CACHE_SIZE.
647  *
648  * If pagecache pages are not associated with an inode, they are said to be
649  * anonymous pages. These may become associated with the swapcache, and in that
650  * case PG_swapcache is set, and page->private is an offset into the swapcache.
651  *
652  * In either case (swapcache or inode backed), the pagecache itself holds one
653  * reference to the page. Setting PG_private should also increment the
654  * refcount. The each user mapping also has a reference to the page.
655  *
656  * The pagecache pages are stored in a per-mapping radix tree, which is
657  * rooted at mapping->page_tree, and indexed by offset.
658  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
659  * lists, we instead now tag pages as dirty/writeback in the radix tree.
660  *
661  * All pagecache pages may be subject to I/O:
662  * - inode pages may need to be read from disk,
663  * - inode pages which have been modified and are MAP_SHARED may need
664  *   to be written back to the inode on disk,
665  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
666  *   modified may need to be swapped out to swap space and (later) to be read
667  *   back into memory.
668  */
669 
670 /*
671  * The zone field is never updated after free_area_init_core()
672  * sets it, so none of the operations on it need to be atomic.
673  */
674 
675 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
676 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
677 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
678 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
679 #define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
680 
681 /*
682  * Define the bit shifts to access each section.  For non-existent
683  * sections we define the shift as 0; that plus a 0 mask ensures
684  * the compiler will optimise away reference to them.
685  */
686 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
687 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
688 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
689 #define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
690 
691 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
692 #ifdef NODE_NOT_IN_PAGE_FLAGS
693 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
694 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
695 						SECTIONS_PGOFF : ZONES_PGOFF)
696 #else
697 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
698 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
699 						NODES_PGOFF : ZONES_PGOFF)
700 #endif
701 
702 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
703 
704 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
705 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
706 #endif
707 
708 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
709 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
710 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
711 #define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
712 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
713 
714 static inline enum zone_type page_zonenum(const struct page *page)
715 {
716 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
717 }
718 
719 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
720 #define SECTION_IN_PAGE_FLAGS
721 #endif
722 
723 /*
724  * The identification function is mainly used by the buddy allocator for
725  * determining if two pages could be buddies. We are not really identifying
726  * the zone since we could be using the section number id if we do not have
727  * node id available in page flags.
728  * We only guarantee that it will return the same value for two combinable
729  * pages in a zone.
730  */
731 static inline int page_zone_id(struct page *page)
732 {
733 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
734 }
735 
736 static inline int zone_to_nid(struct zone *zone)
737 {
738 #ifdef CONFIG_NUMA
739 	return zone->node;
740 #else
741 	return 0;
742 #endif
743 }
744 
745 #ifdef NODE_NOT_IN_PAGE_FLAGS
746 extern int page_to_nid(const struct page *page);
747 #else
748 static inline int page_to_nid(const struct page *page)
749 {
750 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
751 }
752 #endif
753 
754 #ifdef CONFIG_NUMA_BALANCING
755 static inline int cpu_pid_to_cpupid(int cpu, int pid)
756 {
757 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
758 }
759 
760 static inline int cpupid_to_pid(int cpupid)
761 {
762 	return cpupid & LAST__PID_MASK;
763 }
764 
765 static inline int cpupid_to_cpu(int cpupid)
766 {
767 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
768 }
769 
770 static inline int cpupid_to_nid(int cpupid)
771 {
772 	return cpu_to_node(cpupid_to_cpu(cpupid));
773 }
774 
775 static inline bool cpupid_pid_unset(int cpupid)
776 {
777 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
778 }
779 
780 static inline bool cpupid_cpu_unset(int cpupid)
781 {
782 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
783 }
784 
785 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
786 {
787 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
788 }
789 
790 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
791 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
792 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
793 {
794 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
795 }
796 
797 static inline int page_cpupid_last(struct page *page)
798 {
799 	return page->_last_cpupid;
800 }
801 static inline void page_cpupid_reset_last(struct page *page)
802 {
803 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
804 }
805 #else
806 static inline int page_cpupid_last(struct page *page)
807 {
808 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
809 }
810 
811 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
812 
813 static inline void page_cpupid_reset_last(struct page *page)
814 {
815 	int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
816 
817 	page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
818 	page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
819 }
820 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
821 #else /* !CONFIG_NUMA_BALANCING */
822 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
823 {
824 	return page_to_nid(page); /* XXX */
825 }
826 
827 static inline int page_cpupid_last(struct page *page)
828 {
829 	return page_to_nid(page); /* XXX */
830 }
831 
832 static inline int cpupid_to_nid(int cpupid)
833 {
834 	return -1;
835 }
836 
837 static inline int cpupid_to_pid(int cpupid)
838 {
839 	return -1;
840 }
841 
842 static inline int cpupid_to_cpu(int cpupid)
843 {
844 	return -1;
845 }
846 
847 static inline int cpu_pid_to_cpupid(int nid, int pid)
848 {
849 	return -1;
850 }
851 
852 static inline bool cpupid_pid_unset(int cpupid)
853 {
854 	return 1;
855 }
856 
857 static inline void page_cpupid_reset_last(struct page *page)
858 {
859 }
860 
861 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
862 {
863 	return false;
864 }
865 #endif /* CONFIG_NUMA_BALANCING */
866 
867 static inline struct zone *page_zone(const struct page *page)
868 {
869 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
870 }
871 
872 #ifdef SECTION_IN_PAGE_FLAGS
873 static inline void set_page_section(struct page *page, unsigned long section)
874 {
875 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
876 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
877 }
878 
879 static inline unsigned long page_to_section(const struct page *page)
880 {
881 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
882 }
883 #endif
884 
885 static inline void set_page_zone(struct page *page, enum zone_type zone)
886 {
887 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
888 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
889 }
890 
891 static inline void set_page_node(struct page *page, unsigned long node)
892 {
893 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
894 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
895 }
896 
897 static inline void set_page_links(struct page *page, enum zone_type zone,
898 	unsigned long node, unsigned long pfn)
899 {
900 	set_page_zone(page, zone);
901 	set_page_node(page, node);
902 #ifdef SECTION_IN_PAGE_FLAGS
903 	set_page_section(page, pfn_to_section_nr(pfn));
904 #endif
905 }
906 
907 /*
908  * Some inline functions in vmstat.h depend on page_zone()
909  */
910 #include <linux/vmstat.h>
911 
912 static __always_inline void *lowmem_page_address(const struct page *page)
913 {
914 	return __va(PFN_PHYS(page_to_pfn(page)));
915 }
916 
917 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
918 #define HASHED_PAGE_VIRTUAL
919 #endif
920 
921 #if defined(WANT_PAGE_VIRTUAL)
922 static inline void *page_address(const struct page *page)
923 {
924 	return page->virtual;
925 }
926 static inline void set_page_address(struct page *page, void *address)
927 {
928 	page->virtual = address;
929 }
930 #define page_address_init()  do { } while(0)
931 #endif
932 
933 #if defined(HASHED_PAGE_VIRTUAL)
934 void *page_address(const struct page *page);
935 void set_page_address(struct page *page, void *virtual);
936 void page_address_init(void);
937 #endif
938 
939 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
940 #define page_address(page) lowmem_page_address(page)
941 #define set_page_address(page, address)  do { } while(0)
942 #define page_address_init()  do { } while(0)
943 #endif
944 
945 /*
946  * On an anonymous page mapped into a user virtual memory area,
947  * page->mapping points to its anon_vma, not to a struct address_space;
948  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
949  *
950  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
951  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
952  * and then page->mapping points, not to an anon_vma, but to a private
953  * structure which KSM associates with that merged page.  See ksm.h.
954  *
955  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
956  *
957  * Please note that, confusingly, "page_mapping" refers to the inode
958  * address_space which maps the page from disk; whereas "page_mapped"
959  * refers to user virtual address space into which the page is mapped.
960  */
961 #define PAGE_MAPPING_ANON	1
962 #define PAGE_MAPPING_KSM	2
963 #define PAGE_MAPPING_FLAGS	(PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
964 
965 extern struct address_space *page_mapping(struct page *page);
966 
967 /* Neutral page->mapping pointer to address_space or anon_vma or other */
968 static inline void *page_rmapping(struct page *page)
969 {
970 	return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
971 }
972 
973 extern struct address_space *__page_file_mapping(struct page *);
974 
975 static inline
976 struct address_space *page_file_mapping(struct page *page)
977 {
978 	if (unlikely(PageSwapCache(page)))
979 		return __page_file_mapping(page);
980 
981 	return page->mapping;
982 }
983 
984 static inline int PageAnon(struct page *page)
985 {
986 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
987 }
988 
989 /*
990  * Return the pagecache index of the passed page.  Regular pagecache pages
991  * use ->index whereas swapcache pages use ->private
992  */
993 static inline pgoff_t page_index(struct page *page)
994 {
995 	if (unlikely(PageSwapCache(page)))
996 		return page_private(page);
997 	return page->index;
998 }
999 
1000 extern pgoff_t __page_file_index(struct page *page);
1001 
1002 /*
1003  * Return the file index of the page. Regular pagecache pages use ->index
1004  * whereas swapcache pages use swp_offset(->private)
1005  */
1006 static inline pgoff_t page_file_index(struct page *page)
1007 {
1008 	if (unlikely(PageSwapCache(page)))
1009 		return __page_file_index(page);
1010 
1011 	return page->index;
1012 }
1013 
1014 /*
1015  * Return true if this page is mapped into pagetables.
1016  */
1017 static inline int page_mapped(struct page *page)
1018 {
1019 	return atomic_read(&(page)->_mapcount) >= 0;
1020 }
1021 
1022 /*
1023  * Different kinds of faults, as returned by handle_mm_fault().
1024  * Used to decide whether a process gets delivered SIGBUS or
1025  * just gets major/minor fault counters bumped up.
1026  */
1027 
1028 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
1029 
1030 #define VM_FAULT_OOM	0x0001
1031 #define VM_FAULT_SIGBUS	0x0002
1032 #define VM_FAULT_MAJOR	0x0004
1033 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1034 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1035 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1036 
1037 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1038 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1039 #define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1040 #define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
1041 
1042 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1043 
1044 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
1045 			 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
1046 
1047 /* Encode hstate index for a hwpoisoned large page */
1048 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1049 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1050 
1051 /*
1052  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1053  */
1054 extern void pagefault_out_of_memory(void);
1055 
1056 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1057 
1058 /*
1059  * Flags passed to show_mem() and show_free_areas() to suppress output in
1060  * various contexts.
1061  */
1062 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1063 
1064 extern void show_free_areas(unsigned int flags);
1065 extern bool skip_free_areas_node(unsigned int flags, int nid);
1066 
1067 int shmem_zero_setup(struct vm_area_struct *);
1068 #ifdef CONFIG_SHMEM
1069 bool shmem_mapping(struct address_space *mapping);
1070 #else
1071 static inline bool shmem_mapping(struct address_space *mapping)
1072 {
1073 	return false;
1074 }
1075 #endif
1076 
1077 extern int can_do_mlock(void);
1078 extern int user_shm_lock(size_t, struct user_struct *);
1079 extern void user_shm_unlock(size_t, struct user_struct *);
1080 
1081 /*
1082  * Parameter block passed down to zap_pte_range in exceptional cases.
1083  */
1084 struct zap_details {
1085 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
1086 	struct address_space *check_mapping;	/* Check page->mapping if set */
1087 	pgoff_t	first_index;			/* Lowest page->index to unmap */
1088 	pgoff_t last_index;			/* Highest page->index to unmap */
1089 };
1090 
1091 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1092 		pte_t pte);
1093 
1094 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1095 		unsigned long size);
1096 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1097 		unsigned long size, struct zap_details *);
1098 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1099 		unsigned long start, unsigned long end);
1100 
1101 /**
1102  * mm_walk - callbacks for walk_page_range
1103  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1104  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1105  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1106  *	       this handler is required to be able to handle
1107  *	       pmd_trans_huge() pmds.  They may simply choose to
1108  *	       split_huge_page() instead of handling it explicitly.
1109  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1110  * @pte_hole: if set, called for each hole at all levels
1111  * @hugetlb_entry: if set, called for each hugetlb entry
1112  *		   *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1113  * 			      is used.
1114  *
1115  * (see walk_page_range for more details)
1116  */
1117 struct mm_walk {
1118 	int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1119 			 unsigned long next, struct mm_walk *walk);
1120 	int (*pud_entry)(pud_t *pud, unsigned long addr,
1121 	                 unsigned long next, struct mm_walk *walk);
1122 	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1123 			 unsigned long next, struct mm_walk *walk);
1124 	int (*pte_entry)(pte_t *pte, unsigned long addr,
1125 			 unsigned long next, struct mm_walk *walk);
1126 	int (*pte_hole)(unsigned long addr, unsigned long next,
1127 			struct mm_walk *walk);
1128 	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1129 			     unsigned long addr, unsigned long next,
1130 			     struct mm_walk *walk);
1131 	struct mm_struct *mm;
1132 	void *private;
1133 };
1134 
1135 int walk_page_range(unsigned long addr, unsigned long end,
1136 		struct mm_walk *walk);
1137 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1138 		unsigned long end, unsigned long floor, unsigned long ceiling);
1139 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1140 			struct vm_area_struct *vma);
1141 void unmap_mapping_range(struct address_space *mapping,
1142 		loff_t const holebegin, loff_t const holelen, int even_cows);
1143 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1144 	unsigned long *pfn);
1145 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1146 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1147 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1148 			void *buf, int len, int write);
1149 
1150 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1151 		loff_t const holebegin, loff_t const holelen)
1152 {
1153 	unmap_mapping_range(mapping, holebegin, holelen, 0);
1154 }
1155 
1156 extern void truncate_pagecache(struct inode *inode, loff_t new);
1157 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1158 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1159 int truncate_inode_page(struct address_space *mapping, struct page *page);
1160 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1161 int invalidate_inode_page(struct page *page);
1162 
1163 #ifdef CONFIG_MMU
1164 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1165 			unsigned long address, unsigned int flags);
1166 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1167 			    unsigned long address, unsigned int fault_flags);
1168 #else
1169 static inline int handle_mm_fault(struct mm_struct *mm,
1170 			struct vm_area_struct *vma, unsigned long address,
1171 			unsigned int flags)
1172 {
1173 	/* should never happen if there's no MMU */
1174 	BUG();
1175 	return VM_FAULT_SIGBUS;
1176 }
1177 static inline int fixup_user_fault(struct task_struct *tsk,
1178 		struct mm_struct *mm, unsigned long address,
1179 		unsigned int fault_flags)
1180 {
1181 	/* should never happen if there's no MMU */
1182 	BUG();
1183 	return -EFAULT;
1184 }
1185 #endif
1186 
1187 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1188 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1189 		void *buf, int len, int write);
1190 
1191 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1192 		      unsigned long start, unsigned long nr_pages,
1193 		      unsigned int foll_flags, struct page **pages,
1194 		      struct vm_area_struct **vmas, int *nonblocking);
1195 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1196 		    unsigned long start, unsigned long nr_pages,
1197 		    int write, int force, struct page **pages,
1198 		    struct vm_area_struct **vmas);
1199 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1200 			struct page **pages);
1201 struct kvec;
1202 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1203 			struct page **pages);
1204 int get_kernel_page(unsigned long start, int write, struct page **pages);
1205 struct page *get_dump_page(unsigned long addr);
1206 
1207 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1208 extern void do_invalidatepage(struct page *page, unsigned int offset,
1209 			      unsigned int length);
1210 
1211 int __set_page_dirty_nobuffers(struct page *page);
1212 int __set_page_dirty_no_writeback(struct page *page);
1213 int redirty_page_for_writepage(struct writeback_control *wbc,
1214 				struct page *page);
1215 void account_page_dirtied(struct page *page, struct address_space *mapping);
1216 void account_page_writeback(struct page *page);
1217 int set_page_dirty(struct page *page);
1218 int set_page_dirty_lock(struct page *page);
1219 int clear_page_dirty_for_io(struct page *page);
1220 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1221 
1222 /* Is the vma a continuation of the stack vma above it? */
1223 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1224 {
1225 	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1226 }
1227 
1228 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1229 					     unsigned long addr)
1230 {
1231 	return (vma->vm_flags & VM_GROWSDOWN) &&
1232 		(vma->vm_start == addr) &&
1233 		!vma_growsdown(vma->vm_prev, addr);
1234 }
1235 
1236 /* Is the vma a continuation of the stack vma below it? */
1237 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1238 {
1239 	return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1240 }
1241 
1242 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1243 					   unsigned long addr)
1244 {
1245 	return (vma->vm_flags & VM_GROWSUP) &&
1246 		(vma->vm_end == addr) &&
1247 		!vma_growsup(vma->vm_next, addr);
1248 }
1249 
1250 extern pid_t
1251 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1252 
1253 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1254 		unsigned long old_addr, struct vm_area_struct *new_vma,
1255 		unsigned long new_addr, unsigned long len,
1256 		bool need_rmap_locks);
1257 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1258 			      unsigned long end, pgprot_t newprot,
1259 			      int dirty_accountable, int prot_numa);
1260 extern int mprotect_fixup(struct vm_area_struct *vma,
1261 			  struct vm_area_struct **pprev, unsigned long start,
1262 			  unsigned long end, unsigned long newflags);
1263 
1264 /*
1265  * doesn't attempt to fault and will return short.
1266  */
1267 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1268 			  struct page **pages);
1269 /*
1270  * per-process(per-mm_struct) statistics.
1271  */
1272 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1273 {
1274 	long val = atomic_long_read(&mm->rss_stat.count[member]);
1275 
1276 #ifdef SPLIT_RSS_COUNTING
1277 	/*
1278 	 * counter is updated in asynchronous manner and may go to minus.
1279 	 * But it's never be expected number for users.
1280 	 */
1281 	if (val < 0)
1282 		val = 0;
1283 #endif
1284 	return (unsigned long)val;
1285 }
1286 
1287 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1288 {
1289 	atomic_long_add(value, &mm->rss_stat.count[member]);
1290 }
1291 
1292 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1293 {
1294 	atomic_long_inc(&mm->rss_stat.count[member]);
1295 }
1296 
1297 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1298 {
1299 	atomic_long_dec(&mm->rss_stat.count[member]);
1300 }
1301 
1302 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1303 {
1304 	return get_mm_counter(mm, MM_FILEPAGES) +
1305 		get_mm_counter(mm, MM_ANONPAGES);
1306 }
1307 
1308 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1309 {
1310 	return max(mm->hiwater_rss, get_mm_rss(mm));
1311 }
1312 
1313 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1314 {
1315 	return max(mm->hiwater_vm, mm->total_vm);
1316 }
1317 
1318 static inline void update_hiwater_rss(struct mm_struct *mm)
1319 {
1320 	unsigned long _rss = get_mm_rss(mm);
1321 
1322 	if ((mm)->hiwater_rss < _rss)
1323 		(mm)->hiwater_rss = _rss;
1324 }
1325 
1326 static inline void update_hiwater_vm(struct mm_struct *mm)
1327 {
1328 	if (mm->hiwater_vm < mm->total_vm)
1329 		mm->hiwater_vm = mm->total_vm;
1330 }
1331 
1332 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1333 					 struct mm_struct *mm)
1334 {
1335 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1336 
1337 	if (*maxrss < hiwater_rss)
1338 		*maxrss = hiwater_rss;
1339 }
1340 
1341 #if defined(SPLIT_RSS_COUNTING)
1342 void sync_mm_rss(struct mm_struct *mm);
1343 #else
1344 static inline void sync_mm_rss(struct mm_struct *mm)
1345 {
1346 }
1347 #endif
1348 
1349 int vma_wants_writenotify(struct vm_area_struct *vma);
1350 
1351 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1352 			       spinlock_t **ptl);
1353 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1354 				    spinlock_t **ptl)
1355 {
1356 	pte_t *ptep;
1357 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1358 	return ptep;
1359 }
1360 
1361 #ifdef __PAGETABLE_PUD_FOLDED
1362 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1363 						unsigned long address)
1364 {
1365 	return 0;
1366 }
1367 #else
1368 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1369 #endif
1370 
1371 #ifdef __PAGETABLE_PMD_FOLDED
1372 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1373 						unsigned long address)
1374 {
1375 	return 0;
1376 }
1377 #else
1378 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1379 #endif
1380 
1381 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1382 		pmd_t *pmd, unsigned long address);
1383 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1384 
1385 /*
1386  * The following ifdef needed to get the 4level-fixup.h header to work.
1387  * Remove it when 4level-fixup.h has been removed.
1388  */
1389 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1390 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1391 {
1392 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1393 		NULL: pud_offset(pgd, address);
1394 }
1395 
1396 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1397 {
1398 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1399 		NULL: pmd_offset(pud, address);
1400 }
1401 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1402 
1403 #if USE_SPLIT_PTE_PTLOCKS
1404 #if ALLOC_SPLIT_PTLOCKS
1405 void __init ptlock_cache_init(void);
1406 extern bool ptlock_alloc(struct page *page);
1407 extern void ptlock_free(struct page *page);
1408 
1409 static inline spinlock_t *ptlock_ptr(struct page *page)
1410 {
1411 	return page->ptl;
1412 }
1413 #else /* ALLOC_SPLIT_PTLOCKS */
1414 static inline void ptlock_cache_init(void)
1415 {
1416 }
1417 
1418 static inline bool ptlock_alloc(struct page *page)
1419 {
1420 	return true;
1421 }
1422 
1423 static inline void ptlock_free(struct page *page)
1424 {
1425 }
1426 
1427 static inline spinlock_t *ptlock_ptr(struct page *page)
1428 {
1429 	return &page->ptl;
1430 }
1431 #endif /* ALLOC_SPLIT_PTLOCKS */
1432 
1433 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1434 {
1435 	return ptlock_ptr(pmd_page(*pmd));
1436 }
1437 
1438 static inline bool ptlock_init(struct page *page)
1439 {
1440 	/*
1441 	 * prep_new_page() initialize page->private (and therefore page->ptl)
1442 	 * with 0. Make sure nobody took it in use in between.
1443 	 *
1444 	 * It can happen if arch try to use slab for page table allocation:
1445 	 * slab code uses page->slab_cache and page->first_page (for tail
1446 	 * pages), which share storage with page->ptl.
1447 	 */
1448 	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1449 	if (!ptlock_alloc(page))
1450 		return false;
1451 	spin_lock_init(ptlock_ptr(page));
1452 	return true;
1453 }
1454 
1455 /* Reset page->mapping so free_pages_check won't complain. */
1456 static inline void pte_lock_deinit(struct page *page)
1457 {
1458 	page->mapping = NULL;
1459 	ptlock_free(page);
1460 }
1461 
1462 #else	/* !USE_SPLIT_PTE_PTLOCKS */
1463 /*
1464  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1465  */
1466 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1467 {
1468 	return &mm->page_table_lock;
1469 }
1470 static inline void ptlock_cache_init(void) {}
1471 static inline bool ptlock_init(struct page *page) { return true; }
1472 static inline void pte_lock_deinit(struct page *page) {}
1473 #endif /* USE_SPLIT_PTE_PTLOCKS */
1474 
1475 static inline void pgtable_init(void)
1476 {
1477 	ptlock_cache_init();
1478 	pgtable_cache_init();
1479 }
1480 
1481 static inline bool pgtable_page_ctor(struct page *page)
1482 {
1483 	inc_zone_page_state(page, NR_PAGETABLE);
1484 	return ptlock_init(page);
1485 }
1486 
1487 static inline void pgtable_page_dtor(struct page *page)
1488 {
1489 	pte_lock_deinit(page);
1490 	dec_zone_page_state(page, NR_PAGETABLE);
1491 }
1492 
1493 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1494 ({							\
1495 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1496 	pte_t *__pte = pte_offset_map(pmd, address);	\
1497 	*(ptlp) = __ptl;				\
1498 	spin_lock(__ptl);				\
1499 	__pte;						\
1500 })
1501 
1502 #define pte_unmap_unlock(pte, ptl)	do {		\
1503 	spin_unlock(ptl);				\
1504 	pte_unmap(pte);					\
1505 } while (0)
1506 
1507 #define pte_alloc_map(mm, vma, pmd, address)				\
1508 	((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,	\
1509 							pmd, address))?	\
1510 	 NULL: pte_offset_map(pmd, address))
1511 
1512 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1513 	((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,	\
1514 							pmd, address))?	\
1515 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1516 
1517 #define pte_alloc_kernel(pmd, address)			\
1518 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1519 		NULL: pte_offset_kernel(pmd, address))
1520 
1521 #if USE_SPLIT_PMD_PTLOCKS
1522 
1523 static struct page *pmd_to_page(pmd_t *pmd)
1524 {
1525 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1526 	return virt_to_page((void *)((unsigned long) pmd & mask));
1527 }
1528 
1529 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1530 {
1531 	return ptlock_ptr(pmd_to_page(pmd));
1532 }
1533 
1534 static inline bool pgtable_pmd_page_ctor(struct page *page)
1535 {
1536 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1537 	page->pmd_huge_pte = NULL;
1538 #endif
1539 	return ptlock_init(page);
1540 }
1541 
1542 static inline void pgtable_pmd_page_dtor(struct page *page)
1543 {
1544 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1545 	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1546 #endif
1547 	ptlock_free(page);
1548 }
1549 
1550 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1551 
1552 #else
1553 
1554 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1555 {
1556 	return &mm->page_table_lock;
1557 }
1558 
1559 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1560 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1561 
1562 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1563 
1564 #endif
1565 
1566 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1567 {
1568 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
1569 	spin_lock(ptl);
1570 	return ptl;
1571 }
1572 
1573 extern void free_area_init(unsigned long * zones_size);
1574 extern void free_area_init_node(int nid, unsigned long * zones_size,
1575 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1576 extern void free_initmem(void);
1577 
1578 /*
1579  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1580  * into the buddy system. The freed pages will be poisoned with pattern
1581  * "poison" if it's within range [0, UCHAR_MAX].
1582  * Return pages freed into the buddy system.
1583  */
1584 extern unsigned long free_reserved_area(void *start, void *end,
1585 					int poison, char *s);
1586 
1587 #ifdef	CONFIG_HIGHMEM
1588 /*
1589  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1590  * and totalram_pages.
1591  */
1592 extern void free_highmem_page(struct page *page);
1593 #endif
1594 
1595 extern void adjust_managed_page_count(struct page *page, long count);
1596 extern void mem_init_print_info(const char *str);
1597 
1598 /* Free the reserved page into the buddy system, so it gets managed. */
1599 static inline void __free_reserved_page(struct page *page)
1600 {
1601 	ClearPageReserved(page);
1602 	init_page_count(page);
1603 	__free_page(page);
1604 }
1605 
1606 static inline void free_reserved_page(struct page *page)
1607 {
1608 	__free_reserved_page(page);
1609 	adjust_managed_page_count(page, 1);
1610 }
1611 
1612 static inline void mark_page_reserved(struct page *page)
1613 {
1614 	SetPageReserved(page);
1615 	adjust_managed_page_count(page, -1);
1616 }
1617 
1618 /*
1619  * Default method to free all the __init memory into the buddy system.
1620  * The freed pages will be poisoned with pattern "poison" if it's within
1621  * range [0, UCHAR_MAX].
1622  * Return pages freed into the buddy system.
1623  */
1624 static inline unsigned long free_initmem_default(int poison)
1625 {
1626 	extern char __init_begin[], __init_end[];
1627 
1628 	return free_reserved_area(&__init_begin, &__init_end,
1629 				  poison, "unused kernel");
1630 }
1631 
1632 static inline unsigned long get_num_physpages(void)
1633 {
1634 	int nid;
1635 	unsigned long phys_pages = 0;
1636 
1637 	for_each_online_node(nid)
1638 		phys_pages += node_present_pages(nid);
1639 
1640 	return phys_pages;
1641 }
1642 
1643 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1644 /*
1645  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1646  * zones, allocate the backing mem_map and account for memory holes in a more
1647  * architecture independent manner. This is a substitute for creating the
1648  * zone_sizes[] and zholes_size[] arrays and passing them to
1649  * free_area_init_node()
1650  *
1651  * An architecture is expected to register range of page frames backed by
1652  * physical memory with memblock_add[_node]() before calling
1653  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1654  * usage, an architecture is expected to do something like
1655  *
1656  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1657  * 							 max_highmem_pfn};
1658  * for_each_valid_physical_page_range()
1659  * 	memblock_add_node(base, size, nid)
1660  * free_area_init_nodes(max_zone_pfns);
1661  *
1662  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1663  * registered physical page range.  Similarly
1664  * sparse_memory_present_with_active_regions() calls memory_present() for
1665  * each range when SPARSEMEM is enabled.
1666  *
1667  * See mm/page_alloc.c for more information on each function exposed by
1668  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1669  */
1670 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1671 unsigned long node_map_pfn_alignment(void);
1672 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1673 						unsigned long end_pfn);
1674 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1675 						unsigned long end_pfn);
1676 extern void get_pfn_range_for_nid(unsigned int nid,
1677 			unsigned long *start_pfn, unsigned long *end_pfn);
1678 extern unsigned long find_min_pfn_with_active_regions(void);
1679 extern void free_bootmem_with_active_regions(int nid,
1680 						unsigned long max_low_pfn);
1681 extern void sparse_memory_present_with_active_regions(int nid);
1682 
1683 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1684 
1685 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1686     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1687 static inline int __early_pfn_to_nid(unsigned long pfn)
1688 {
1689 	return 0;
1690 }
1691 #else
1692 /* please see mm/page_alloc.c */
1693 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1694 /* there is a per-arch backend function. */
1695 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1696 #endif
1697 
1698 extern void set_dma_reserve(unsigned long new_dma_reserve);
1699 extern void memmap_init_zone(unsigned long, int, unsigned long,
1700 				unsigned long, enum memmap_context);
1701 extern void setup_per_zone_wmarks(void);
1702 extern int __meminit init_per_zone_wmark_min(void);
1703 extern void mem_init(void);
1704 extern void __init mmap_init(void);
1705 extern void show_mem(unsigned int flags);
1706 extern void si_meminfo(struct sysinfo * val);
1707 extern void si_meminfo_node(struct sysinfo *val, int nid);
1708 
1709 extern __printf(3, 4)
1710 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1711 
1712 extern void setup_per_cpu_pageset(void);
1713 
1714 extern void zone_pcp_update(struct zone *zone);
1715 extern void zone_pcp_reset(struct zone *zone);
1716 
1717 /* page_alloc.c */
1718 extern int min_free_kbytes;
1719 
1720 /* nommu.c */
1721 extern atomic_long_t mmap_pages_allocated;
1722 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1723 
1724 /* interval_tree.c */
1725 void vma_interval_tree_insert(struct vm_area_struct *node,
1726 			      struct rb_root *root);
1727 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1728 				    struct vm_area_struct *prev,
1729 				    struct rb_root *root);
1730 void vma_interval_tree_remove(struct vm_area_struct *node,
1731 			      struct rb_root *root);
1732 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1733 				unsigned long start, unsigned long last);
1734 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1735 				unsigned long start, unsigned long last);
1736 
1737 #define vma_interval_tree_foreach(vma, root, start, last)		\
1738 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
1739 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
1740 
1741 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1742 					struct list_head *list)
1743 {
1744 	list_add_tail(&vma->shared.nonlinear, list);
1745 }
1746 
1747 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1748 				   struct rb_root *root);
1749 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1750 				   struct rb_root *root);
1751 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1752 	struct rb_root *root, unsigned long start, unsigned long last);
1753 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1754 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
1755 #ifdef CONFIG_DEBUG_VM_RB
1756 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1757 #endif
1758 
1759 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
1760 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1761 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1762 
1763 /* mmap.c */
1764 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1765 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1766 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1767 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1768 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1769 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1770 	struct mempolicy *);
1771 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1772 extern int split_vma(struct mm_struct *,
1773 	struct vm_area_struct *, unsigned long addr, int new_below);
1774 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1775 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1776 	struct rb_node **, struct rb_node *);
1777 extern void unlink_file_vma(struct vm_area_struct *);
1778 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1779 	unsigned long addr, unsigned long len, pgoff_t pgoff,
1780 	bool *need_rmap_locks);
1781 extern void exit_mmap(struct mm_struct *);
1782 
1783 extern int mm_take_all_locks(struct mm_struct *mm);
1784 extern void mm_drop_all_locks(struct mm_struct *mm);
1785 
1786 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1787 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1788 
1789 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1790 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1791 				   unsigned long addr, unsigned long len,
1792 				   unsigned long flags,
1793 				   const struct vm_special_mapping *spec);
1794 /* This is an obsolete alternative to _install_special_mapping. */
1795 extern int install_special_mapping(struct mm_struct *mm,
1796 				   unsigned long addr, unsigned long len,
1797 				   unsigned long flags, struct page **pages);
1798 
1799 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1800 
1801 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1802 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1803 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1804 	unsigned long len, unsigned long prot, unsigned long flags,
1805 	unsigned long pgoff, unsigned long *populate);
1806 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1807 
1808 #ifdef CONFIG_MMU
1809 extern int __mm_populate(unsigned long addr, unsigned long len,
1810 			 int ignore_errors);
1811 static inline void mm_populate(unsigned long addr, unsigned long len)
1812 {
1813 	/* Ignore errors */
1814 	(void) __mm_populate(addr, len, 1);
1815 }
1816 #else
1817 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1818 #endif
1819 
1820 /* These take the mm semaphore themselves */
1821 extern unsigned long vm_brk(unsigned long, unsigned long);
1822 extern int vm_munmap(unsigned long, size_t);
1823 extern unsigned long vm_mmap(struct file *, unsigned long,
1824         unsigned long, unsigned long,
1825         unsigned long, unsigned long);
1826 
1827 struct vm_unmapped_area_info {
1828 #define VM_UNMAPPED_AREA_TOPDOWN 1
1829 	unsigned long flags;
1830 	unsigned long length;
1831 	unsigned long low_limit;
1832 	unsigned long high_limit;
1833 	unsigned long align_mask;
1834 	unsigned long align_offset;
1835 };
1836 
1837 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1838 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1839 
1840 /*
1841  * Search for an unmapped address range.
1842  *
1843  * We are looking for a range that:
1844  * - does not intersect with any VMA;
1845  * - is contained within the [low_limit, high_limit) interval;
1846  * - is at least the desired size.
1847  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1848  */
1849 static inline unsigned long
1850 vm_unmapped_area(struct vm_unmapped_area_info *info)
1851 {
1852 	if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1853 		return unmapped_area(info);
1854 	else
1855 		return unmapped_area_topdown(info);
1856 }
1857 
1858 /* truncate.c */
1859 extern void truncate_inode_pages(struct address_space *, loff_t);
1860 extern void truncate_inode_pages_range(struct address_space *,
1861 				       loff_t lstart, loff_t lend);
1862 extern void truncate_inode_pages_final(struct address_space *);
1863 
1864 /* generic vm_area_ops exported for stackable file systems */
1865 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1866 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1867 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1868 
1869 /* mm/page-writeback.c */
1870 int write_one_page(struct page *page, int wait);
1871 void task_dirty_inc(struct task_struct *tsk);
1872 
1873 /* readahead.c */
1874 #define VM_MAX_READAHEAD	128	/* kbytes */
1875 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1876 
1877 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1878 			pgoff_t offset, unsigned long nr_to_read);
1879 
1880 void page_cache_sync_readahead(struct address_space *mapping,
1881 			       struct file_ra_state *ra,
1882 			       struct file *filp,
1883 			       pgoff_t offset,
1884 			       unsigned long size);
1885 
1886 void page_cache_async_readahead(struct address_space *mapping,
1887 				struct file_ra_state *ra,
1888 				struct file *filp,
1889 				struct page *pg,
1890 				pgoff_t offset,
1891 				unsigned long size);
1892 
1893 unsigned long max_sane_readahead(unsigned long nr);
1894 
1895 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1896 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1897 
1898 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1899 extern int expand_downwards(struct vm_area_struct *vma,
1900 		unsigned long address);
1901 #if VM_GROWSUP
1902 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1903 #else
1904   #define expand_upwards(vma, address) do { } while (0)
1905 #endif
1906 
1907 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1908 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1909 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1910 					     struct vm_area_struct **pprev);
1911 
1912 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1913    NULL if none.  Assume start_addr < end_addr. */
1914 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1915 {
1916 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1917 
1918 	if (vma && end_addr <= vma->vm_start)
1919 		vma = NULL;
1920 	return vma;
1921 }
1922 
1923 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1924 {
1925 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1926 }
1927 
1928 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1929 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1930 				unsigned long vm_start, unsigned long vm_end)
1931 {
1932 	struct vm_area_struct *vma = find_vma(mm, vm_start);
1933 
1934 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1935 		vma = NULL;
1936 
1937 	return vma;
1938 }
1939 
1940 #ifdef CONFIG_MMU
1941 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1942 #else
1943 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1944 {
1945 	return __pgprot(0);
1946 }
1947 #endif
1948 
1949 #ifdef CONFIG_NUMA_BALANCING
1950 unsigned long change_prot_numa(struct vm_area_struct *vma,
1951 			unsigned long start, unsigned long end);
1952 #endif
1953 
1954 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1955 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1956 			unsigned long pfn, unsigned long size, pgprot_t);
1957 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1958 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1959 			unsigned long pfn);
1960 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1961 			unsigned long pfn);
1962 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1963 
1964 
1965 struct page *follow_page_mask(struct vm_area_struct *vma,
1966 			      unsigned long address, unsigned int foll_flags,
1967 			      unsigned int *page_mask);
1968 
1969 static inline struct page *follow_page(struct vm_area_struct *vma,
1970 		unsigned long address, unsigned int foll_flags)
1971 {
1972 	unsigned int unused_page_mask;
1973 	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1974 }
1975 
1976 #define FOLL_WRITE	0x01	/* check pte is writable */
1977 #define FOLL_TOUCH	0x02	/* mark page accessed */
1978 #define FOLL_GET	0x04	/* do get_page on page */
1979 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
1980 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
1981 #define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
1982 				 * and return without waiting upon it */
1983 #define FOLL_MLOCK	0x40	/* mark page as mlocked */
1984 #define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
1985 #define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
1986 #define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
1987 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
1988 
1989 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1990 			void *data);
1991 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1992 			       unsigned long size, pte_fn_t fn, void *data);
1993 
1994 #ifdef CONFIG_PROC_FS
1995 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1996 #else
1997 static inline void vm_stat_account(struct mm_struct *mm,
1998 			unsigned long flags, struct file *file, long pages)
1999 {
2000 	mm->total_vm += pages;
2001 }
2002 #endif /* CONFIG_PROC_FS */
2003 
2004 #ifdef CONFIG_DEBUG_PAGEALLOC
2005 extern void kernel_map_pages(struct page *page, int numpages, int enable);
2006 #ifdef CONFIG_HIBERNATION
2007 extern bool kernel_page_present(struct page *page);
2008 #endif /* CONFIG_HIBERNATION */
2009 #else
2010 static inline void
2011 kernel_map_pages(struct page *page, int numpages, int enable) {}
2012 #ifdef CONFIG_HIBERNATION
2013 static inline bool kernel_page_present(struct page *page) { return true; }
2014 #endif /* CONFIG_HIBERNATION */
2015 #endif
2016 
2017 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2018 #ifdef	__HAVE_ARCH_GATE_AREA
2019 int in_gate_area_no_mm(unsigned long addr);
2020 int in_gate_area(struct mm_struct *mm, unsigned long addr);
2021 #else
2022 int in_gate_area_no_mm(unsigned long addr);
2023 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
2024 #endif	/* __HAVE_ARCH_GATE_AREA */
2025 
2026 #ifdef CONFIG_SYSCTL
2027 extern int sysctl_drop_caches;
2028 int drop_caches_sysctl_handler(struct ctl_table *, int,
2029 					void __user *, size_t *, loff_t *);
2030 #endif
2031 
2032 unsigned long shrink_slab(struct shrink_control *shrink,
2033 			  unsigned long nr_pages_scanned,
2034 			  unsigned long lru_pages);
2035 
2036 #ifndef CONFIG_MMU
2037 #define randomize_va_space 0
2038 #else
2039 extern int randomize_va_space;
2040 #endif
2041 
2042 const char * arch_vma_name(struct vm_area_struct *vma);
2043 void print_vma_addr(char *prefix, unsigned long rip);
2044 
2045 void sparse_mem_maps_populate_node(struct page **map_map,
2046 				   unsigned long pnum_begin,
2047 				   unsigned long pnum_end,
2048 				   unsigned long map_count,
2049 				   int nodeid);
2050 
2051 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2052 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2053 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2054 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2055 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2056 void *vmemmap_alloc_block(unsigned long size, int node);
2057 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2058 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2059 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2060 			       int node);
2061 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2062 void vmemmap_populate_print_last(void);
2063 #ifdef CONFIG_MEMORY_HOTPLUG
2064 void vmemmap_free(unsigned long start, unsigned long end);
2065 #endif
2066 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2067 				  unsigned long size);
2068 
2069 enum mf_flags {
2070 	MF_COUNT_INCREASED = 1 << 0,
2071 	MF_ACTION_REQUIRED = 1 << 1,
2072 	MF_MUST_KILL = 1 << 2,
2073 	MF_SOFT_OFFLINE = 1 << 3,
2074 };
2075 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2076 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2077 extern int unpoison_memory(unsigned long pfn);
2078 extern int sysctl_memory_failure_early_kill;
2079 extern int sysctl_memory_failure_recovery;
2080 extern void shake_page(struct page *p, int access);
2081 extern atomic_long_t num_poisoned_pages;
2082 extern int soft_offline_page(struct page *page, int flags);
2083 
2084 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2085 extern void clear_huge_page(struct page *page,
2086 			    unsigned long addr,
2087 			    unsigned int pages_per_huge_page);
2088 extern void copy_user_huge_page(struct page *dst, struct page *src,
2089 				unsigned long addr, struct vm_area_struct *vma,
2090 				unsigned int pages_per_huge_page);
2091 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2092 
2093 #ifdef CONFIG_DEBUG_PAGEALLOC
2094 extern unsigned int _debug_guardpage_minorder;
2095 
2096 static inline unsigned int debug_guardpage_minorder(void)
2097 {
2098 	return _debug_guardpage_minorder;
2099 }
2100 
2101 static inline bool page_is_guard(struct page *page)
2102 {
2103 	return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2104 }
2105 #else
2106 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2107 static inline bool page_is_guard(struct page *page) { return false; }
2108 #endif /* CONFIG_DEBUG_PAGEALLOC */
2109 
2110 #if MAX_NUMNODES > 1
2111 void __init setup_nr_node_ids(void);
2112 #else
2113 static inline void setup_nr_node_ids(void) {}
2114 #endif
2115 
2116 #endif /* __KERNEL__ */
2117 #endif /* _LINUX_MM_H */
2118