xref: /linux-6.15/include/linux/mm.h (revision aa65506f)
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 #ifdef CONFIG_NUMA
243 	/*
244 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
245 	 * to hold the policy upon return.  Caller should pass NULL @new to
246 	 * remove a policy and fall back to surrounding context--i.e. do not
247 	 * install a MPOL_DEFAULT policy, nor the task or system default
248 	 * mempolicy.
249 	 */
250 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
251 
252 	/*
253 	 * get_policy() op must add reference [mpol_get()] to any policy at
254 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
255 	 * in mm/mempolicy.c will do this automatically.
256 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
257 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
258 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
259 	 * must return NULL--i.e., do not "fallback" to task or system default
260 	 * policy.
261 	 */
262 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
263 					unsigned long addr);
264 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
265 		const nodemask_t *to, unsigned long flags);
266 #endif
267 	/* called by sys_remap_file_pages() to populate non-linear mapping */
268 	int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
269 			   unsigned long size, pgoff_t pgoff);
270 };
271 
272 struct mmu_gather;
273 struct inode;
274 
275 #define page_private(page)		((page)->private)
276 #define set_page_private(page, v)	((page)->private = (v))
277 
278 /* It's valid only if the page is free path or free_list */
279 static inline void set_freepage_migratetype(struct page *page, int migratetype)
280 {
281 	page->index = migratetype;
282 }
283 
284 /* It's valid only if the page is free path or free_list */
285 static inline int get_freepage_migratetype(struct page *page)
286 {
287 	return page->index;
288 }
289 
290 /*
291  * FIXME: take this include out, include page-flags.h in
292  * files which need it (119 of them)
293  */
294 #include <linux/page-flags.h>
295 #include <linux/huge_mm.h>
296 
297 /*
298  * Methods to modify the page usage count.
299  *
300  * What counts for a page usage:
301  * - cache mapping   (page->mapping)
302  * - private data    (page->private)
303  * - page mapped in a task's page tables, each mapping
304  *   is counted separately
305  *
306  * Also, many kernel routines increase the page count before a critical
307  * routine so they can be sure the page doesn't go away from under them.
308  */
309 
310 /*
311  * Drop a ref, return true if the refcount fell to zero (the page has no users)
312  */
313 static inline int put_page_testzero(struct page *page)
314 {
315 	VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
316 	return atomic_dec_and_test(&page->_count);
317 }
318 
319 /*
320  * Try to grab a ref unless the page has a refcount of zero, return false if
321  * that is the case.
322  * This can be called when MMU is off so it must not access
323  * any of the virtual mappings.
324  */
325 static inline int get_page_unless_zero(struct page *page)
326 {
327 	return atomic_inc_not_zero(&page->_count);
328 }
329 
330 /*
331  * Try to drop a ref unless the page has a refcount of one, return false if
332  * that is the case.
333  * This is to make sure that the refcount won't become zero after this drop.
334  * This can be called when MMU is off so it must not access
335  * any of the virtual mappings.
336  */
337 static inline int put_page_unless_one(struct page *page)
338 {
339 	return atomic_add_unless(&page->_count, -1, 1);
340 }
341 
342 extern int page_is_ram(unsigned long pfn);
343 
344 /* Support for virtually mapped pages */
345 struct page *vmalloc_to_page(const void *addr);
346 unsigned long vmalloc_to_pfn(const void *addr);
347 
348 /*
349  * Determine if an address is within the vmalloc range
350  *
351  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
352  * is no special casing required.
353  */
354 static inline int is_vmalloc_addr(const void *x)
355 {
356 #ifdef CONFIG_MMU
357 	unsigned long addr = (unsigned long)x;
358 
359 	return addr >= VMALLOC_START && addr < VMALLOC_END;
360 #else
361 	return 0;
362 #endif
363 }
364 #ifdef CONFIG_MMU
365 extern int is_vmalloc_or_module_addr(const void *x);
366 #else
367 static inline int is_vmalloc_or_module_addr(const void *x)
368 {
369 	return 0;
370 }
371 #endif
372 
373 extern void kvfree(const void *addr);
374 
375 static inline void compound_lock(struct page *page)
376 {
377 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
378 	VM_BUG_ON_PAGE(PageSlab(page), page);
379 	bit_spin_lock(PG_compound_lock, &page->flags);
380 #endif
381 }
382 
383 static inline void compound_unlock(struct page *page)
384 {
385 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
386 	VM_BUG_ON_PAGE(PageSlab(page), page);
387 	bit_spin_unlock(PG_compound_lock, &page->flags);
388 #endif
389 }
390 
391 static inline unsigned long compound_lock_irqsave(struct page *page)
392 {
393 	unsigned long uninitialized_var(flags);
394 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
395 	local_irq_save(flags);
396 	compound_lock(page);
397 #endif
398 	return flags;
399 }
400 
401 static inline void compound_unlock_irqrestore(struct page *page,
402 					      unsigned long flags)
403 {
404 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
405 	compound_unlock(page);
406 	local_irq_restore(flags);
407 #endif
408 }
409 
410 static inline struct page *compound_head(struct page *page)
411 {
412 	if (unlikely(PageTail(page))) {
413 		struct page *head = page->first_page;
414 
415 		/*
416 		 * page->first_page may be a dangling pointer to an old
417 		 * compound page, so recheck that it is still a tail
418 		 * page before returning.
419 		 */
420 		smp_rmb();
421 		if (likely(PageTail(page)))
422 			return head;
423 	}
424 	return page;
425 }
426 
427 /*
428  * The atomic page->_mapcount, starts from -1: so that transitions
429  * both from it and to it can be tracked, using atomic_inc_and_test
430  * and atomic_add_negative(-1).
431  */
432 static inline void page_mapcount_reset(struct page *page)
433 {
434 	atomic_set(&(page)->_mapcount, -1);
435 }
436 
437 static inline int page_mapcount(struct page *page)
438 {
439 	return atomic_read(&(page)->_mapcount) + 1;
440 }
441 
442 static inline int page_count(struct page *page)
443 {
444 	return atomic_read(&compound_head(page)->_count);
445 }
446 
447 #ifdef CONFIG_HUGETLB_PAGE
448 extern int PageHeadHuge(struct page *page_head);
449 #else /* CONFIG_HUGETLB_PAGE */
450 static inline int PageHeadHuge(struct page *page_head)
451 {
452 	return 0;
453 }
454 #endif /* CONFIG_HUGETLB_PAGE */
455 
456 static inline bool __compound_tail_refcounted(struct page *page)
457 {
458 	return !PageSlab(page) && !PageHeadHuge(page);
459 }
460 
461 /*
462  * This takes a head page as parameter and tells if the
463  * tail page reference counting can be skipped.
464  *
465  * For this to be safe, PageSlab and PageHeadHuge must remain true on
466  * any given page where they return true here, until all tail pins
467  * have been released.
468  */
469 static inline bool compound_tail_refcounted(struct page *page)
470 {
471 	VM_BUG_ON_PAGE(!PageHead(page), page);
472 	return __compound_tail_refcounted(page);
473 }
474 
475 static inline void get_huge_page_tail(struct page *page)
476 {
477 	/*
478 	 * __split_huge_page_refcount() cannot run from under us.
479 	 */
480 	VM_BUG_ON_PAGE(!PageTail(page), page);
481 	VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
482 	VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
483 	if (compound_tail_refcounted(page->first_page))
484 		atomic_inc(&page->_mapcount);
485 }
486 
487 extern bool __get_page_tail(struct page *page);
488 
489 static inline void get_page(struct page *page)
490 {
491 	if (unlikely(PageTail(page)))
492 		if (likely(__get_page_tail(page)))
493 			return;
494 	/*
495 	 * Getting a normal page or the head of a compound page
496 	 * requires to already have an elevated page->_count.
497 	 */
498 	VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
499 	atomic_inc(&page->_count);
500 }
501 
502 static inline struct page *virt_to_head_page(const void *x)
503 {
504 	struct page *page = virt_to_page(x);
505 	return compound_head(page);
506 }
507 
508 /*
509  * Setup the page count before being freed into the page allocator for
510  * the first time (boot or memory hotplug)
511  */
512 static inline void init_page_count(struct page *page)
513 {
514 	atomic_set(&page->_count, 1);
515 }
516 
517 /*
518  * PageBuddy() indicate that the page is free and in the buddy system
519  * (see mm/page_alloc.c).
520  *
521  * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
522  * -2 so that an underflow of the page_mapcount() won't be mistaken
523  * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
524  * efficiently by most CPU architectures.
525  */
526 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
527 
528 static inline int PageBuddy(struct page *page)
529 {
530 	return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
531 }
532 
533 static inline void __SetPageBuddy(struct page *page)
534 {
535 	VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
536 	atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
537 }
538 
539 static inline void __ClearPageBuddy(struct page *page)
540 {
541 	VM_BUG_ON_PAGE(!PageBuddy(page), page);
542 	atomic_set(&page->_mapcount, -1);
543 }
544 
545 void put_page(struct page *page);
546 void put_pages_list(struct list_head *pages);
547 
548 void split_page(struct page *page, unsigned int order);
549 int split_free_page(struct page *page);
550 
551 /*
552  * Compound pages have a destructor function.  Provide a
553  * prototype for that function and accessor functions.
554  * These are _only_ valid on the head of a PG_compound page.
555  */
556 typedef void compound_page_dtor(struct page *);
557 
558 static inline void set_compound_page_dtor(struct page *page,
559 						compound_page_dtor *dtor)
560 {
561 	page[1].lru.next = (void *)dtor;
562 }
563 
564 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
565 {
566 	return (compound_page_dtor *)page[1].lru.next;
567 }
568 
569 static inline int compound_order(struct page *page)
570 {
571 	if (!PageHead(page))
572 		return 0;
573 	return (unsigned long)page[1].lru.prev;
574 }
575 
576 static inline void set_compound_order(struct page *page, unsigned long order)
577 {
578 	page[1].lru.prev = (void *)order;
579 }
580 
581 #ifdef CONFIG_MMU
582 /*
583  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
584  * servicing faults for write access.  In the normal case, do always want
585  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
586  * that do not have writing enabled, when used by access_process_vm.
587  */
588 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
589 {
590 	if (likely(vma->vm_flags & VM_WRITE))
591 		pte = pte_mkwrite(pte);
592 	return pte;
593 }
594 
595 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
596 		struct page *page, pte_t *pte, bool write, bool anon);
597 #endif
598 
599 /*
600  * Multiple processes may "see" the same page. E.g. for untouched
601  * mappings of /dev/null, all processes see the same page full of
602  * zeroes, and text pages of executables and shared libraries have
603  * only one copy in memory, at most, normally.
604  *
605  * For the non-reserved pages, page_count(page) denotes a reference count.
606  *   page_count() == 0 means the page is free. page->lru is then used for
607  *   freelist management in the buddy allocator.
608  *   page_count() > 0  means the page has been allocated.
609  *
610  * Pages are allocated by the slab allocator in order to provide memory
611  * to kmalloc and kmem_cache_alloc. In this case, the management of the
612  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
613  * unless a particular usage is carefully commented. (the responsibility of
614  * freeing the kmalloc memory is the caller's, of course).
615  *
616  * A page may be used by anyone else who does a __get_free_page().
617  * In this case, page_count still tracks the references, and should only
618  * be used through the normal accessor functions. The top bits of page->flags
619  * and page->virtual store page management information, but all other fields
620  * are unused and could be used privately, carefully. The management of this
621  * page is the responsibility of the one who allocated it, and those who have
622  * subsequently been given references to it.
623  *
624  * The other pages (we may call them "pagecache pages") are completely
625  * managed by the Linux memory manager: I/O, buffers, swapping etc.
626  * The following discussion applies only to them.
627  *
628  * A pagecache page contains an opaque `private' member, which belongs to the
629  * page's address_space. Usually, this is the address of a circular list of
630  * the page's disk buffers. PG_private must be set to tell the VM to call
631  * into the filesystem to release these pages.
632  *
633  * A page may belong to an inode's memory mapping. In this case, page->mapping
634  * is the pointer to the inode, and page->index is the file offset of the page,
635  * in units of PAGE_CACHE_SIZE.
636  *
637  * If pagecache pages are not associated with an inode, they are said to be
638  * anonymous pages. These may become associated with the swapcache, and in that
639  * case PG_swapcache is set, and page->private is an offset into the swapcache.
640  *
641  * In either case (swapcache or inode backed), the pagecache itself holds one
642  * reference to the page. Setting PG_private should also increment the
643  * refcount. The each user mapping also has a reference to the page.
644  *
645  * The pagecache pages are stored in a per-mapping radix tree, which is
646  * rooted at mapping->page_tree, and indexed by offset.
647  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
648  * lists, we instead now tag pages as dirty/writeback in the radix tree.
649  *
650  * All pagecache pages may be subject to I/O:
651  * - inode pages may need to be read from disk,
652  * - inode pages which have been modified and are MAP_SHARED may need
653  *   to be written back to the inode on disk,
654  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
655  *   modified may need to be swapped out to swap space and (later) to be read
656  *   back into memory.
657  */
658 
659 /*
660  * The zone field is never updated after free_area_init_core()
661  * sets it, so none of the operations on it need to be atomic.
662  */
663 
664 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
665 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
666 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
667 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
668 #define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
669 
670 /*
671  * Define the bit shifts to access each section.  For non-existent
672  * sections we define the shift as 0; that plus a 0 mask ensures
673  * the compiler will optimise away reference to them.
674  */
675 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
676 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
677 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
678 #define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
679 
680 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
681 #ifdef NODE_NOT_IN_PAGE_FLAGS
682 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
683 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
684 						SECTIONS_PGOFF : ZONES_PGOFF)
685 #else
686 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
687 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
688 						NODES_PGOFF : ZONES_PGOFF)
689 #endif
690 
691 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
692 
693 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
694 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
695 #endif
696 
697 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
698 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
699 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
700 #define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
701 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
702 
703 static inline enum zone_type page_zonenum(const struct page *page)
704 {
705 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
706 }
707 
708 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
709 #define SECTION_IN_PAGE_FLAGS
710 #endif
711 
712 /*
713  * The identification function is mainly used by the buddy allocator for
714  * determining if two pages could be buddies. We are not really identifying
715  * the zone since we could be using the section number id if we do not have
716  * node id available in page flags.
717  * We only guarantee that it will return the same value for two combinable
718  * pages in a zone.
719  */
720 static inline int page_zone_id(struct page *page)
721 {
722 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
723 }
724 
725 static inline int zone_to_nid(struct zone *zone)
726 {
727 #ifdef CONFIG_NUMA
728 	return zone->node;
729 #else
730 	return 0;
731 #endif
732 }
733 
734 #ifdef NODE_NOT_IN_PAGE_FLAGS
735 extern int page_to_nid(const struct page *page);
736 #else
737 static inline int page_to_nid(const struct page *page)
738 {
739 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
740 }
741 #endif
742 
743 #ifdef CONFIG_NUMA_BALANCING
744 static inline int cpu_pid_to_cpupid(int cpu, int pid)
745 {
746 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
747 }
748 
749 static inline int cpupid_to_pid(int cpupid)
750 {
751 	return cpupid & LAST__PID_MASK;
752 }
753 
754 static inline int cpupid_to_cpu(int cpupid)
755 {
756 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
757 }
758 
759 static inline int cpupid_to_nid(int cpupid)
760 {
761 	return cpu_to_node(cpupid_to_cpu(cpupid));
762 }
763 
764 static inline bool cpupid_pid_unset(int cpupid)
765 {
766 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
767 }
768 
769 static inline bool cpupid_cpu_unset(int cpupid)
770 {
771 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
772 }
773 
774 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
775 {
776 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
777 }
778 
779 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
780 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
781 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
782 {
783 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
784 }
785 
786 static inline int page_cpupid_last(struct page *page)
787 {
788 	return page->_last_cpupid;
789 }
790 static inline void page_cpupid_reset_last(struct page *page)
791 {
792 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
793 }
794 #else
795 static inline int page_cpupid_last(struct page *page)
796 {
797 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
798 }
799 
800 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
801 
802 static inline void page_cpupid_reset_last(struct page *page)
803 {
804 	int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
805 
806 	page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
807 	page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
808 }
809 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
810 #else /* !CONFIG_NUMA_BALANCING */
811 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
812 {
813 	return page_to_nid(page); /* XXX */
814 }
815 
816 static inline int page_cpupid_last(struct page *page)
817 {
818 	return page_to_nid(page); /* XXX */
819 }
820 
821 static inline int cpupid_to_nid(int cpupid)
822 {
823 	return -1;
824 }
825 
826 static inline int cpupid_to_pid(int cpupid)
827 {
828 	return -1;
829 }
830 
831 static inline int cpupid_to_cpu(int cpupid)
832 {
833 	return -1;
834 }
835 
836 static inline int cpu_pid_to_cpupid(int nid, int pid)
837 {
838 	return -1;
839 }
840 
841 static inline bool cpupid_pid_unset(int cpupid)
842 {
843 	return 1;
844 }
845 
846 static inline void page_cpupid_reset_last(struct page *page)
847 {
848 }
849 
850 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
851 {
852 	return false;
853 }
854 #endif /* CONFIG_NUMA_BALANCING */
855 
856 static inline struct zone *page_zone(const struct page *page)
857 {
858 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
859 }
860 
861 #ifdef SECTION_IN_PAGE_FLAGS
862 static inline void set_page_section(struct page *page, unsigned long section)
863 {
864 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
865 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
866 }
867 
868 static inline unsigned long page_to_section(const struct page *page)
869 {
870 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
871 }
872 #endif
873 
874 static inline void set_page_zone(struct page *page, enum zone_type zone)
875 {
876 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
877 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
878 }
879 
880 static inline void set_page_node(struct page *page, unsigned long node)
881 {
882 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
883 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
884 }
885 
886 static inline void set_page_links(struct page *page, enum zone_type zone,
887 	unsigned long node, unsigned long pfn)
888 {
889 	set_page_zone(page, zone);
890 	set_page_node(page, node);
891 #ifdef SECTION_IN_PAGE_FLAGS
892 	set_page_section(page, pfn_to_section_nr(pfn));
893 #endif
894 }
895 
896 /*
897  * Some inline functions in vmstat.h depend on page_zone()
898  */
899 #include <linux/vmstat.h>
900 
901 static __always_inline void *lowmem_page_address(const struct page *page)
902 {
903 	return __va(PFN_PHYS(page_to_pfn(page)));
904 }
905 
906 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
907 #define HASHED_PAGE_VIRTUAL
908 #endif
909 
910 #if defined(WANT_PAGE_VIRTUAL)
911 static inline void *page_address(const struct page *page)
912 {
913 	return page->virtual;
914 }
915 static inline void set_page_address(struct page *page, void *address)
916 {
917 	page->virtual = address;
918 }
919 #define page_address_init()  do { } while(0)
920 #endif
921 
922 #if defined(HASHED_PAGE_VIRTUAL)
923 void *page_address(const struct page *page);
924 void set_page_address(struct page *page, void *virtual);
925 void page_address_init(void);
926 #endif
927 
928 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
929 #define page_address(page) lowmem_page_address(page)
930 #define set_page_address(page, address)  do { } while(0)
931 #define page_address_init()  do { } while(0)
932 #endif
933 
934 /*
935  * On an anonymous page mapped into a user virtual memory area,
936  * page->mapping points to its anon_vma, not to a struct address_space;
937  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
938  *
939  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
940  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
941  * and then page->mapping points, not to an anon_vma, but to a private
942  * structure which KSM associates with that merged page.  See ksm.h.
943  *
944  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
945  *
946  * Please note that, confusingly, "page_mapping" refers to the inode
947  * address_space which maps the page from disk; whereas "page_mapped"
948  * refers to user virtual address space into which the page is mapped.
949  */
950 #define PAGE_MAPPING_ANON	1
951 #define PAGE_MAPPING_KSM	2
952 #define PAGE_MAPPING_FLAGS	(PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
953 
954 extern struct address_space *page_mapping(struct page *page);
955 
956 /* Neutral page->mapping pointer to address_space or anon_vma or other */
957 static inline void *page_rmapping(struct page *page)
958 {
959 	return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
960 }
961 
962 extern struct address_space *__page_file_mapping(struct page *);
963 
964 static inline
965 struct address_space *page_file_mapping(struct page *page)
966 {
967 	if (unlikely(PageSwapCache(page)))
968 		return __page_file_mapping(page);
969 
970 	return page->mapping;
971 }
972 
973 static inline int PageAnon(struct page *page)
974 {
975 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
976 }
977 
978 /*
979  * Return the pagecache index of the passed page.  Regular pagecache pages
980  * use ->index whereas swapcache pages use ->private
981  */
982 static inline pgoff_t page_index(struct page *page)
983 {
984 	if (unlikely(PageSwapCache(page)))
985 		return page_private(page);
986 	return page->index;
987 }
988 
989 extern pgoff_t __page_file_index(struct page *page);
990 
991 /*
992  * Return the file index of the page. Regular pagecache pages use ->index
993  * whereas swapcache pages use swp_offset(->private)
994  */
995 static inline pgoff_t page_file_index(struct page *page)
996 {
997 	if (unlikely(PageSwapCache(page)))
998 		return __page_file_index(page);
999 
1000 	return page->index;
1001 }
1002 
1003 /*
1004  * Return true if this page is mapped into pagetables.
1005  */
1006 static inline int page_mapped(struct page *page)
1007 {
1008 	return atomic_read(&(page)->_mapcount) >= 0;
1009 }
1010 
1011 /*
1012  * Different kinds of faults, as returned by handle_mm_fault().
1013  * Used to decide whether a process gets delivered SIGBUS or
1014  * just gets major/minor fault counters bumped up.
1015  */
1016 
1017 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
1018 
1019 #define VM_FAULT_OOM	0x0001
1020 #define VM_FAULT_SIGBUS	0x0002
1021 #define VM_FAULT_MAJOR	0x0004
1022 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1023 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1024 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1025 
1026 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1027 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1028 #define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1029 #define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
1030 
1031 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1032 
1033 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
1034 			 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
1035 
1036 /* Encode hstate index for a hwpoisoned large page */
1037 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1038 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1039 
1040 /*
1041  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1042  */
1043 extern void pagefault_out_of_memory(void);
1044 
1045 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1046 
1047 /*
1048  * Flags passed to show_mem() and show_free_areas() to suppress output in
1049  * various contexts.
1050  */
1051 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1052 
1053 extern void show_free_areas(unsigned int flags);
1054 extern bool skip_free_areas_node(unsigned int flags, int nid);
1055 
1056 int shmem_zero_setup(struct vm_area_struct *);
1057 #ifdef CONFIG_SHMEM
1058 bool shmem_mapping(struct address_space *mapping);
1059 #else
1060 static inline bool shmem_mapping(struct address_space *mapping)
1061 {
1062 	return false;
1063 }
1064 #endif
1065 
1066 extern int can_do_mlock(void);
1067 extern int user_shm_lock(size_t, struct user_struct *);
1068 extern void user_shm_unlock(size_t, struct user_struct *);
1069 
1070 /*
1071  * Parameter block passed down to zap_pte_range in exceptional cases.
1072  */
1073 struct zap_details {
1074 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
1075 	struct address_space *check_mapping;	/* Check page->mapping if set */
1076 	pgoff_t	first_index;			/* Lowest page->index to unmap */
1077 	pgoff_t last_index;			/* Highest page->index to unmap */
1078 };
1079 
1080 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1081 		pte_t pte);
1082 
1083 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1084 		unsigned long size);
1085 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1086 		unsigned long size, struct zap_details *);
1087 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1088 		unsigned long start, unsigned long end);
1089 
1090 /**
1091  * mm_walk - callbacks for walk_page_range
1092  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1093  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1094  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1095  *	       this handler is required to be able to handle
1096  *	       pmd_trans_huge() pmds.  They may simply choose to
1097  *	       split_huge_page() instead of handling it explicitly.
1098  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1099  * @pte_hole: if set, called for each hole at all levels
1100  * @hugetlb_entry: if set, called for each hugetlb entry
1101  *		   *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1102  * 			      is used.
1103  *
1104  * (see walk_page_range for more details)
1105  */
1106 struct mm_walk {
1107 	int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1108 			 unsigned long next, struct mm_walk *walk);
1109 	int (*pud_entry)(pud_t *pud, unsigned long addr,
1110 	                 unsigned long next, struct mm_walk *walk);
1111 	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1112 			 unsigned long next, struct mm_walk *walk);
1113 	int (*pte_entry)(pte_t *pte, unsigned long addr,
1114 			 unsigned long next, struct mm_walk *walk);
1115 	int (*pte_hole)(unsigned long addr, unsigned long next,
1116 			struct mm_walk *walk);
1117 	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1118 			     unsigned long addr, unsigned long next,
1119 			     struct mm_walk *walk);
1120 	struct mm_struct *mm;
1121 	void *private;
1122 };
1123 
1124 int walk_page_range(unsigned long addr, unsigned long end,
1125 		struct mm_walk *walk);
1126 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1127 		unsigned long end, unsigned long floor, unsigned long ceiling);
1128 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1129 			struct vm_area_struct *vma);
1130 void unmap_mapping_range(struct address_space *mapping,
1131 		loff_t const holebegin, loff_t const holelen, int even_cows);
1132 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1133 	unsigned long *pfn);
1134 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1135 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1136 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1137 			void *buf, int len, int write);
1138 
1139 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1140 		loff_t const holebegin, loff_t const holelen)
1141 {
1142 	unmap_mapping_range(mapping, holebegin, holelen, 0);
1143 }
1144 
1145 extern void truncate_pagecache(struct inode *inode, loff_t new);
1146 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1147 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1148 int truncate_inode_page(struct address_space *mapping, struct page *page);
1149 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1150 int invalidate_inode_page(struct page *page);
1151 
1152 #ifdef CONFIG_MMU
1153 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1154 			unsigned long address, unsigned int flags);
1155 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1156 			    unsigned long address, unsigned int fault_flags);
1157 #else
1158 static inline int handle_mm_fault(struct mm_struct *mm,
1159 			struct vm_area_struct *vma, unsigned long address,
1160 			unsigned int flags)
1161 {
1162 	/* should never happen if there's no MMU */
1163 	BUG();
1164 	return VM_FAULT_SIGBUS;
1165 }
1166 static inline int fixup_user_fault(struct task_struct *tsk,
1167 		struct mm_struct *mm, unsigned long address,
1168 		unsigned int fault_flags)
1169 {
1170 	/* should never happen if there's no MMU */
1171 	BUG();
1172 	return -EFAULT;
1173 }
1174 #endif
1175 
1176 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1177 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1178 		void *buf, int len, int write);
1179 
1180 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1181 		      unsigned long start, unsigned long nr_pages,
1182 		      unsigned int foll_flags, struct page **pages,
1183 		      struct vm_area_struct **vmas, int *nonblocking);
1184 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1185 		    unsigned long start, unsigned long nr_pages,
1186 		    int write, int force, struct page **pages,
1187 		    struct vm_area_struct **vmas);
1188 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1189 			struct page **pages);
1190 struct kvec;
1191 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1192 			struct page **pages);
1193 int get_kernel_page(unsigned long start, int write, struct page **pages);
1194 struct page *get_dump_page(unsigned long addr);
1195 
1196 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1197 extern void do_invalidatepage(struct page *page, unsigned int offset,
1198 			      unsigned int length);
1199 
1200 int __set_page_dirty_nobuffers(struct page *page);
1201 int __set_page_dirty_no_writeback(struct page *page);
1202 int redirty_page_for_writepage(struct writeback_control *wbc,
1203 				struct page *page);
1204 void account_page_dirtied(struct page *page, struct address_space *mapping);
1205 void account_page_writeback(struct page *page);
1206 int set_page_dirty(struct page *page);
1207 int set_page_dirty_lock(struct page *page);
1208 int clear_page_dirty_for_io(struct page *page);
1209 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1210 
1211 /* Is the vma a continuation of the stack vma above it? */
1212 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1213 {
1214 	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1215 }
1216 
1217 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1218 					     unsigned long addr)
1219 {
1220 	return (vma->vm_flags & VM_GROWSDOWN) &&
1221 		(vma->vm_start == addr) &&
1222 		!vma_growsdown(vma->vm_prev, addr);
1223 }
1224 
1225 /* Is the vma a continuation of the stack vma below it? */
1226 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1227 {
1228 	return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1229 }
1230 
1231 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1232 					   unsigned long addr)
1233 {
1234 	return (vma->vm_flags & VM_GROWSUP) &&
1235 		(vma->vm_end == addr) &&
1236 		!vma_growsup(vma->vm_next, addr);
1237 }
1238 
1239 extern pid_t
1240 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1241 
1242 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1243 		unsigned long old_addr, struct vm_area_struct *new_vma,
1244 		unsigned long new_addr, unsigned long len,
1245 		bool need_rmap_locks);
1246 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1247 			      unsigned long end, pgprot_t newprot,
1248 			      int dirty_accountable, int prot_numa);
1249 extern int mprotect_fixup(struct vm_area_struct *vma,
1250 			  struct vm_area_struct **pprev, unsigned long start,
1251 			  unsigned long end, unsigned long newflags);
1252 
1253 /*
1254  * doesn't attempt to fault and will return short.
1255  */
1256 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1257 			  struct page **pages);
1258 /*
1259  * per-process(per-mm_struct) statistics.
1260  */
1261 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1262 {
1263 	long val = atomic_long_read(&mm->rss_stat.count[member]);
1264 
1265 #ifdef SPLIT_RSS_COUNTING
1266 	/*
1267 	 * counter is updated in asynchronous manner and may go to minus.
1268 	 * But it's never be expected number for users.
1269 	 */
1270 	if (val < 0)
1271 		val = 0;
1272 #endif
1273 	return (unsigned long)val;
1274 }
1275 
1276 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1277 {
1278 	atomic_long_add(value, &mm->rss_stat.count[member]);
1279 }
1280 
1281 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1282 {
1283 	atomic_long_inc(&mm->rss_stat.count[member]);
1284 }
1285 
1286 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1287 {
1288 	atomic_long_dec(&mm->rss_stat.count[member]);
1289 }
1290 
1291 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1292 {
1293 	return get_mm_counter(mm, MM_FILEPAGES) +
1294 		get_mm_counter(mm, MM_ANONPAGES);
1295 }
1296 
1297 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1298 {
1299 	return max(mm->hiwater_rss, get_mm_rss(mm));
1300 }
1301 
1302 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1303 {
1304 	return max(mm->hiwater_vm, mm->total_vm);
1305 }
1306 
1307 static inline void update_hiwater_rss(struct mm_struct *mm)
1308 {
1309 	unsigned long _rss = get_mm_rss(mm);
1310 
1311 	if ((mm)->hiwater_rss < _rss)
1312 		(mm)->hiwater_rss = _rss;
1313 }
1314 
1315 static inline void update_hiwater_vm(struct mm_struct *mm)
1316 {
1317 	if (mm->hiwater_vm < mm->total_vm)
1318 		mm->hiwater_vm = mm->total_vm;
1319 }
1320 
1321 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1322 					 struct mm_struct *mm)
1323 {
1324 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1325 
1326 	if (*maxrss < hiwater_rss)
1327 		*maxrss = hiwater_rss;
1328 }
1329 
1330 #if defined(SPLIT_RSS_COUNTING)
1331 void sync_mm_rss(struct mm_struct *mm);
1332 #else
1333 static inline void sync_mm_rss(struct mm_struct *mm)
1334 {
1335 }
1336 #endif
1337 
1338 int vma_wants_writenotify(struct vm_area_struct *vma);
1339 
1340 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1341 			       spinlock_t **ptl);
1342 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1343 				    spinlock_t **ptl)
1344 {
1345 	pte_t *ptep;
1346 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1347 	return ptep;
1348 }
1349 
1350 #ifdef __PAGETABLE_PUD_FOLDED
1351 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1352 						unsigned long address)
1353 {
1354 	return 0;
1355 }
1356 #else
1357 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1358 #endif
1359 
1360 #ifdef __PAGETABLE_PMD_FOLDED
1361 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1362 						unsigned long address)
1363 {
1364 	return 0;
1365 }
1366 #else
1367 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1368 #endif
1369 
1370 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1371 		pmd_t *pmd, unsigned long address);
1372 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1373 
1374 /*
1375  * The following ifdef needed to get the 4level-fixup.h header to work.
1376  * Remove it when 4level-fixup.h has been removed.
1377  */
1378 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1379 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1380 {
1381 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1382 		NULL: pud_offset(pgd, address);
1383 }
1384 
1385 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1386 {
1387 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1388 		NULL: pmd_offset(pud, address);
1389 }
1390 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1391 
1392 #if USE_SPLIT_PTE_PTLOCKS
1393 #if ALLOC_SPLIT_PTLOCKS
1394 void __init ptlock_cache_init(void);
1395 extern bool ptlock_alloc(struct page *page);
1396 extern void ptlock_free(struct page *page);
1397 
1398 static inline spinlock_t *ptlock_ptr(struct page *page)
1399 {
1400 	return page->ptl;
1401 }
1402 #else /* ALLOC_SPLIT_PTLOCKS */
1403 static inline void ptlock_cache_init(void)
1404 {
1405 }
1406 
1407 static inline bool ptlock_alloc(struct page *page)
1408 {
1409 	return true;
1410 }
1411 
1412 static inline void ptlock_free(struct page *page)
1413 {
1414 }
1415 
1416 static inline spinlock_t *ptlock_ptr(struct page *page)
1417 {
1418 	return &page->ptl;
1419 }
1420 #endif /* ALLOC_SPLIT_PTLOCKS */
1421 
1422 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1423 {
1424 	return ptlock_ptr(pmd_page(*pmd));
1425 }
1426 
1427 static inline bool ptlock_init(struct page *page)
1428 {
1429 	/*
1430 	 * prep_new_page() initialize page->private (and therefore page->ptl)
1431 	 * with 0. Make sure nobody took it in use in between.
1432 	 *
1433 	 * It can happen if arch try to use slab for page table allocation:
1434 	 * slab code uses page->slab_cache and page->first_page (for tail
1435 	 * pages), which share storage with page->ptl.
1436 	 */
1437 	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1438 	if (!ptlock_alloc(page))
1439 		return false;
1440 	spin_lock_init(ptlock_ptr(page));
1441 	return true;
1442 }
1443 
1444 /* Reset page->mapping so free_pages_check won't complain. */
1445 static inline void pte_lock_deinit(struct page *page)
1446 {
1447 	page->mapping = NULL;
1448 	ptlock_free(page);
1449 }
1450 
1451 #else	/* !USE_SPLIT_PTE_PTLOCKS */
1452 /*
1453  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1454  */
1455 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1456 {
1457 	return &mm->page_table_lock;
1458 }
1459 static inline void ptlock_cache_init(void) {}
1460 static inline bool ptlock_init(struct page *page) { return true; }
1461 static inline void pte_lock_deinit(struct page *page) {}
1462 #endif /* USE_SPLIT_PTE_PTLOCKS */
1463 
1464 static inline void pgtable_init(void)
1465 {
1466 	ptlock_cache_init();
1467 	pgtable_cache_init();
1468 }
1469 
1470 static inline bool pgtable_page_ctor(struct page *page)
1471 {
1472 	inc_zone_page_state(page, NR_PAGETABLE);
1473 	return ptlock_init(page);
1474 }
1475 
1476 static inline void pgtable_page_dtor(struct page *page)
1477 {
1478 	pte_lock_deinit(page);
1479 	dec_zone_page_state(page, NR_PAGETABLE);
1480 }
1481 
1482 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1483 ({							\
1484 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1485 	pte_t *__pte = pte_offset_map(pmd, address);	\
1486 	*(ptlp) = __ptl;				\
1487 	spin_lock(__ptl);				\
1488 	__pte;						\
1489 })
1490 
1491 #define pte_unmap_unlock(pte, ptl)	do {		\
1492 	spin_unlock(ptl);				\
1493 	pte_unmap(pte);					\
1494 } while (0)
1495 
1496 #define pte_alloc_map(mm, vma, pmd, address)				\
1497 	((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,	\
1498 							pmd, address))?	\
1499 	 NULL: pte_offset_map(pmd, address))
1500 
1501 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1502 	((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,	\
1503 							pmd, address))?	\
1504 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1505 
1506 #define pte_alloc_kernel(pmd, address)			\
1507 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1508 		NULL: pte_offset_kernel(pmd, address))
1509 
1510 #if USE_SPLIT_PMD_PTLOCKS
1511 
1512 static struct page *pmd_to_page(pmd_t *pmd)
1513 {
1514 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1515 	return virt_to_page((void *)((unsigned long) pmd & mask));
1516 }
1517 
1518 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1519 {
1520 	return ptlock_ptr(pmd_to_page(pmd));
1521 }
1522 
1523 static inline bool pgtable_pmd_page_ctor(struct page *page)
1524 {
1525 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1526 	page->pmd_huge_pte = NULL;
1527 #endif
1528 	return ptlock_init(page);
1529 }
1530 
1531 static inline void pgtable_pmd_page_dtor(struct page *page)
1532 {
1533 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1534 	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1535 #endif
1536 	ptlock_free(page);
1537 }
1538 
1539 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1540 
1541 #else
1542 
1543 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1544 {
1545 	return &mm->page_table_lock;
1546 }
1547 
1548 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1549 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1550 
1551 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1552 
1553 #endif
1554 
1555 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1556 {
1557 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
1558 	spin_lock(ptl);
1559 	return ptl;
1560 }
1561 
1562 extern void free_area_init(unsigned long * zones_size);
1563 extern void free_area_init_node(int nid, unsigned long * zones_size,
1564 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1565 extern void free_initmem(void);
1566 
1567 /*
1568  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1569  * into the buddy system. The freed pages will be poisoned with pattern
1570  * "poison" if it's within range [0, UCHAR_MAX].
1571  * Return pages freed into the buddy system.
1572  */
1573 extern unsigned long free_reserved_area(void *start, void *end,
1574 					int poison, char *s);
1575 
1576 #ifdef	CONFIG_HIGHMEM
1577 /*
1578  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1579  * and totalram_pages.
1580  */
1581 extern void free_highmem_page(struct page *page);
1582 #endif
1583 
1584 extern void adjust_managed_page_count(struct page *page, long count);
1585 extern void mem_init_print_info(const char *str);
1586 
1587 /* Free the reserved page into the buddy system, so it gets managed. */
1588 static inline void __free_reserved_page(struct page *page)
1589 {
1590 	ClearPageReserved(page);
1591 	init_page_count(page);
1592 	__free_page(page);
1593 }
1594 
1595 static inline void free_reserved_page(struct page *page)
1596 {
1597 	__free_reserved_page(page);
1598 	adjust_managed_page_count(page, 1);
1599 }
1600 
1601 static inline void mark_page_reserved(struct page *page)
1602 {
1603 	SetPageReserved(page);
1604 	adjust_managed_page_count(page, -1);
1605 }
1606 
1607 /*
1608  * Default method to free all the __init memory into the buddy system.
1609  * The freed pages will be poisoned with pattern "poison" if it's within
1610  * range [0, UCHAR_MAX].
1611  * Return pages freed into the buddy system.
1612  */
1613 static inline unsigned long free_initmem_default(int poison)
1614 {
1615 	extern char __init_begin[], __init_end[];
1616 
1617 	return free_reserved_area(&__init_begin, &__init_end,
1618 				  poison, "unused kernel");
1619 }
1620 
1621 static inline unsigned long get_num_physpages(void)
1622 {
1623 	int nid;
1624 	unsigned long phys_pages = 0;
1625 
1626 	for_each_online_node(nid)
1627 		phys_pages += node_present_pages(nid);
1628 
1629 	return phys_pages;
1630 }
1631 
1632 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1633 /*
1634  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1635  * zones, allocate the backing mem_map and account for memory holes in a more
1636  * architecture independent manner. This is a substitute for creating the
1637  * zone_sizes[] and zholes_size[] arrays and passing them to
1638  * free_area_init_node()
1639  *
1640  * An architecture is expected to register range of page frames backed by
1641  * physical memory with memblock_add[_node]() before calling
1642  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1643  * usage, an architecture is expected to do something like
1644  *
1645  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1646  * 							 max_highmem_pfn};
1647  * for_each_valid_physical_page_range()
1648  * 	memblock_add_node(base, size, nid)
1649  * free_area_init_nodes(max_zone_pfns);
1650  *
1651  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1652  * registered physical page range.  Similarly
1653  * sparse_memory_present_with_active_regions() calls memory_present() for
1654  * each range when SPARSEMEM is enabled.
1655  *
1656  * See mm/page_alloc.c for more information on each function exposed by
1657  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1658  */
1659 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1660 unsigned long node_map_pfn_alignment(void);
1661 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1662 						unsigned long end_pfn);
1663 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1664 						unsigned long end_pfn);
1665 extern void get_pfn_range_for_nid(unsigned int nid,
1666 			unsigned long *start_pfn, unsigned long *end_pfn);
1667 extern unsigned long find_min_pfn_with_active_regions(void);
1668 extern void free_bootmem_with_active_regions(int nid,
1669 						unsigned long max_low_pfn);
1670 extern void sparse_memory_present_with_active_regions(int nid);
1671 
1672 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1673 
1674 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1675     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1676 static inline int __early_pfn_to_nid(unsigned long pfn)
1677 {
1678 	return 0;
1679 }
1680 #else
1681 /* please see mm/page_alloc.c */
1682 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1683 /* there is a per-arch backend function. */
1684 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1685 #endif
1686 
1687 extern void set_dma_reserve(unsigned long new_dma_reserve);
1688 extern void memmap_init_zone(unsigned long, int, unsigned long,
1689 				unsigned long, enum memmap_context);
1690 extern void setup_per_zone_wmarks(void);
1691 extern int __meminit init_per_zone_wmark_min(void);
1692 extern void mem_init(void);
1693 extern void __init mmap_init(void);
1694 extern void show_mem(unsigned int flags);
1695 extern void si_meminfo(struct sysinfo * val);
1696 extern void si_meminfo_node(struct sysinfo *val, int nid);
1697 
1698 extern __printf(3, 4)
1699 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1700 
1701 extern void setup_per_cpu_pageset(void);
1702 
1703 extern void zone_pcp_update(struct zone *zone);
1704 extern void zone_pcp_reset(struct zone *zone);
1705 
1706 /* page_alloc.c */
1707 extern int min_free_kbytes;
1708 
1709 /* nommu.c */
1710 extern atomic_long_t mmap_pages_allocated;
1711 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1712 
1713 /* interval_tree.c */
1714 void vma_interval_tree_insert(struct vm_area_struct *node,
1715 			      struct rb_root *root);
1716 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1717 				    struct vm_area_struct *prev,
1718 				    struct rb_root *root);
1719 void vma_interval_tree_remove(struct vm_area_struct *node,
1720 			      struct rb_root *root);
1721 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1722 				unsigned long start, unsigned long last);
1723 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1724 				unsigned long start, unsigned long last);
1725 
1726 #define vma_interval_tree_foreach(vma, root, start, last)		\
1727 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
1728 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
1729 
1730 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1731 					struct list_head *list)
1732 {
1733 	list_add_tail(&vma->shared.nonlinear, list);
1734 }
1735 
1736 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1737 				   struct rb_root *root);
1738 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1739 				   struct rb_root *root);
1740 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1741 	struct rb_root *root, unsigned long start, unsigned long last);
1742 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1743 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
1744 #ifdef CONFIG_DEBUG_VM_RB
1745 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1746 #endif
1747 
1748 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
1749 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1750 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1751 
1752 /* mmap.c */
1753 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1754 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1755 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1756 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1757 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1758 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1759 	struct mempolicy *);
1760 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1761 extern int split_vma(struct mm_struct *,
1762 	struct vm_area_struct *, unsigned long addr, int new_below);
1763 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1764 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1765 	struct rb_node **, struct rb_node *);
1766 extern void unlink_file_vma(struct vm_area_struct *);
1767 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1768 	unsigned long addr, unsigned long len, pgoff_t pgoff,
1769 	bool *need_rmap_locks);
1770 extern void exit_mmap(struct mm_struct *);
1771 
1772 extern int mm_take_all_locks(struct mm_struct *mm);
1773 extern void mm_drop_all_locks(struct mm_struct *mm);
1774 
1775 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1776 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1777 
1778 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1779 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1780 				   unsigned long addr, unsigned long len,
1781 				   unsigned long flags, struct page **pages);
1782 extern int install_special_mapping(struct mm_struct *mm,
1783 				   unsigned long addr, unsigned long len,
1784 				   unsigned long flags, struct page **pages);
1785 
1786 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1787 
1788 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1789 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1790 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1791 	unsigned long len, unsigned long prot, unsigned long flags,
1792 	unsigned long pgoff, unsigned long *populate);
1793 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1794 
1795 #ifdef CONFIG_MMU
1796 extern int __mm_populate(unsigned long addr, unsigned long len,
1797 			 int ignore_errors);
1798 static inline void mm_populate(unsigned long addr, unsigned long len)
1799 {
1800 	/* Ignore errors */
1801 	(void) __mm_populate(addr, len, 1);
1802 }
1803 #else
1804 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1805 #endif
1806 
1807 /* These take the mm semaphore themselves */
1808 extern unsigned long vm_brk(unsigned long, unsigned long);
1809 extern int vm_munmap(unsigned long, size_t);
1810 extern unsigned long vm_mmap(struct file *, unsigned long,
1811         unsigned long, unsigned long,
1812         unsigned long, unsigned long);
1813 
1814 struct vm_unmapped_area_info {
1815 #define VM_UNMAPPED_AREA_TOPDOWN 1
1816 	unsigned long flags;
1817 	unsigned long length;
1818 	unsigned long low_limit;
1819 	unsigned long high_limit;
1820 	unsigned long align_mask;
1821 	unsigned long align_offset;
1822 };
1823 
1824 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1825 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1826 
1827 /*
1828  * Search for an unmapped address range.
1829  *
1830  * We are looking for a range that:
1831  * - does not intersect with any VMA;
1832  * - is contained within the [low_limit, high_limit) interval;
1833  * - is at least the desired size.
1834  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1835  */
1836 static inline unsigned long
1837 vm_unmapped_area(struct vm_unmapped_area_info *info)
1838 {
1839 	if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1840 		return unmapped_area(info);
1841 	else
1842 		return unmapped_area_topdown(info);
1843 }
1844 
1845 /* truncate.c */
1846 extern void truncate_inode_pages(struct address_space *, loff_t);
1847 extern void truncate_inode_pages_range(struct address_space *,
1848 				       loff_t lstart, loff_t lend);
1849 extern void truncate_inode_pages_final(struct address_space *);
1850 
1851 /* generic vm_area_ops exported for stackable file systems */
1852 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1853 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1854 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1855 
1856 /* mm/page-writeback.c */
1857 int write_one_page(struct page *page, int wait);
1858 void task_dirty_inc(struct task_struct *tsk);
1859 
1860 /* readahead.c */
1861 #define VM_MAX_READAHEAD	128	/* kbytes */
1862 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1863 
1864 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1865 			pgoff_t offset, unsigned long nr_to_read);
1866 
1867 void page_cache_sync_readahead(struct address_space *mapping,
1868 			       struct file_ra_state *ra,
1869 			       struct file *filp,
1870 			       pgoff_t offset,
1871 			       unsigned long size);
1872 
1873 void page_cache_async_readahead(struct address_space *mapping,
1874 				struct file_ra_state *ra,
1875 				struct file *filp,
1876 				struct page *pg,
1877 				pgoff_t offset,
1878 				unsigned long size);
1879 
1880 unsigned long max_sane_readahead(unsigned long nr);
1881 
1882 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1883 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1884 
1885 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1886 extern int expand_downwards(struct vm_area_struct *vma,
1887 		unsigned long address);
1888 #if VM_GROWSUP
1889 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1890 #else
1891   #define expand_upwards(vma, address) do { } while (0)
1892 #endif
1893 
1894 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1895 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1896 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1897 					     struct vm_area_struct **pprev);
1898 
1899 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1900    NULL if none.  Assume start_addr < end_addr. */
1901 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1902 {
1903 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1904 
1905 	if (vma && end_addr <= vma->vm_start)
1906 		vma = NULL;
1907 	return vma;
1908 }
1909 
1910 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1911 {
1912 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1913 }
1914 
1915 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1916 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1917 				unsigned long vm_start, unsigned long vm_end)
1918 {
1919 	struct vm_area_struct *vma = find_vma(mm, vm_start);
1920 
1921 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1922 		vma = NULL;
1923 
1924 	return vma;
1925 }
1926 
1927 #ifdef CONFIG_MMU
1928 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1929 #else
1930 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1931 {
1932 	return __pgprot(0);
1933 }
1934 #endif
1935 
1936 #ifdef CONFIG_NUMA_BALANCING
1937 unsigned long change_prot_numa(struct vm_area_struct *vma,
1938 			unsigned long start, unsigned long end);
1939 #endif
1940 
1941 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1942 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1943 			unsigned long pfn, unsigned long size, pgprot_t);
1944 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1945 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1946 			unsigned long pfn);
1947 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1948 			unsigned long pfn);
1949 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1950 
1951 
1952 struct page *follow_page_mask(struct vm_area_struct *vma,
1953 			      unsigned long address, unsigned int foll_flags,
1954 			      unsigned int *page_mask);
1955 
1956 static inline struct page *follow_page(struct vm_area_struct *vma,
1957 		unsigned long address, unsigned int foll_flags)
1958 {
1959 	unsigned int unused_page_mask;
1960 	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1961 }
1962 
1963 #define FOLL_WRITE	0x01	/* check pte is writable */
1964 #define FOLL_TOUCH	0x02	/* mark page accessed */
1965 #define FOLL_GET	0x04	/* do get_page on page */
1966 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
1967 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
1968 #define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
1969 				 * and return without waiting upon it */
1970 #define FOLL_MLOCK	0x40	/* mark page as mlocked */
1971 #define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
1972 #define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
1973 #define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
1974 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
1975 
1976 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1977 			void *data);
1978 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1979 			       unsigned long size, pte_fn_t fn, void *data);
1980 
1981 #ifdef CONFIG_PROC_FS
1982 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1983 #else
1984 static inline void vm_stat_account(struct mm_struct *mm,
1985 			unsigned long flags, struct file *file, long pages)
1986 {
1987 	mm->total_vm += pages;
1988 }
1989 #endif /* CONFIG_PROC_FS */
1990 
1991 #ifdef CONFIG_DEBUG_PAGEALLOC
1992 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1993 #ifdef CONFIG_HIBERNATION
1994 extern bool kernel_page_present(struct page *page);
1995 #endif /* CONFIG_HIBERNATION */
1996 #else
1997 static inline void
1998 kernel_map_pages(struct page *page, int numpages, int enable) {}
1999 #ifdef CONFIG_HIBERNATION
2000 static inline bool kernel_page_present(struct page *page) { return true; }
2001 #endif /* CONFIG_HIBERNATION */
2002 #endif
2003 
2004 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2005 #ifdef	__HAVE_ARCH_GATE_AREA
2006 int in_gate_area_no_mm(unsigned long addr);
2007 int in_gate_area(struct mm_struct *mm, unsigned long addr);
2008 #else
2009 int in_gate_area_no_mm(unsigned long addr);
2010 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
2011 #endif	/* __HAVE_ARCH_GATE_AREA */
2012 
2013 #ifdef CONFIG_SYSCTL
2014 extern int sysctl_drop_caches;
2015 int drop_caches_sysctl_handler(struct ctl_table *, int,
2016 					void __user *, size_t *, loff_t *);
2017 #endif
2018 
2019 unsigned long shrink_slab(struct shrink_control *shrink,
2020 			  unsigned long nr_pages_scanned,
2021 			  unsigned long lru_pages);
2022 
2023 #ifndef CONFIG_MMU
2024 #define randomize_va_space 0
2025 #else
2026 extern int randomize_va_space;
2027 #endif
2028 
2029 const char * arch_vma_name(struct vm_area_struct *vma);
2030 void print_vma_addr(char *prefix, unsigned long rip);
2031 
2032 void sparse_mem_maps_populate_node(struct page **map_map,
2033 				   unsigned long pnum_begin,
2034 				   unsigned long pnum_end,
2035 				   unsigned long map_count,
2036 				   int nodeid);
2037 
2038 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2039 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2040 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2041 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2042 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2043 void *vmemmap_alloc_block(unsigned long size, int node);
2044 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2045 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2046 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2047 			       int node);
2048 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2049 void vmemmap_populate_print_last(void);
2050 #ifdef CONFIG_MEMORY_HOTPLUG
2051 void vmemmap_free(unsigned long start, unsigned long end);
2052 #endif
2053 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2054 				  unsigned long size);
2055 
2056 enum mf_flags {
2057 	MF_COUNT_INCREASED = 1 << 0,
2058 	MF_ACTION_REQUIRED = 1 << 1,
2059 	MF_MUST_KILL = 1 << 2,
2060 	MF_SOFT_OFFLINE = 1 << 3,
2061 };
2062 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2063 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2064 extern int unpoison_memory(unsigned long pfn);
2065 extern int sysctl_memory_failure_early_kill;
2066 extern int sysctl_memory_failure_recovery;
2067 extern void shake_page(struct page *p, int access);
2068 extern atomic_long_t num_poisoned_pages;
2069 extern int soft_offline_page(struct page *page, int flags);
2070 
2071 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2072 extern void clear_huge_page(struct page *page,
2073 			    unsigned long addr,
2074 			    unsigned int pages_per_huge_page);
2075 extern void copy_user_huge_page(struct page *dst, struct page *src,
2076 				unsigned long addr, struct vm_area_struct *vma,
2077 				unsigned int pages_per_huge_page);
2078 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2079 
2080 #ifdef CONFIG_DEBUG_PAGEALLOC
2081 extern unsigned int _debug_guardpage_minorder;
2082 
2083 static inline unsigned int debug_guardpage_minorder(void)
2084 {
2085 	return _debug_guardpage_minorder;
2086 }
2087 
2088 static inline bool page_is_guard(struct page *page)
2089 {
2090 	return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2091 }
2092 #else
2093 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2094 static inline bool page_is_guard(struct page *page) { return false; }
2095 #endif /* CONFIG_DEBUG_PAGEALLOC */
2096 
2097 #if MAX_NUMNODES > 1
2098 void __init setup_nr_node_ids(void);
2099 #else
2100 static inline void setup_nr_node_ids(void) {}
2101 #endif
2102 
2103 #endif /* __KERNEL__ */
2104 #endif /* _LINUX_MM_H */
2105