xref: /linux-6.15/include/linux/mm.h (revision e11da5b4)
1 #ifndef _LINUX_MM_H
2 #define _LINUX_MM_H
3 
4 #include <linux/errno.h>
5 
6 #ifdef __KERNEL__
7 
8 #include <linux/gfp.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/range.h>
16 #include <linux/pfn.h>
17 
18 struct mempolicy;
19 struct anon_vma;
20 struct file_ra_state;
21 struct user_struct;
22 struct writeback_control;
23 
24 #ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
25 extern unsigned long max_mapnr;
26 #endif
27 
28 extern unsigned long num_physpages;
29 extern unsigned long totalram_pages;
30 extern void * high_memory;
31 extern int page_cluster;
32 
33 #ifdef CONFIG_SYSCTL
34 extern int sysctl_legacy_va_layout;
35 #else
36 #define sysctl_legacy_va_layout 0
37 #endif
38 
39 #include <asm/page.h>
40 #include <asm/pgtable.h>
41 #include <asm/processor.h>
42 
43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44 
45 /* to align the pointer to the (next) page boundary */
46 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
47 
48 /*
49  * Linux kernel virtual memory manager primitives.
50  * The idea being to have a "virtual" mm in the same way
51  * we have a virtual fs - giving a cleaner interface to the
52  * mm details, and allowing different kinds of memory mappings
53  * (from shared memory to executable loading to arbitrary
54  * mmap() functions).
55  */
56 
57 extern struct kmem_cache *vm_area_cachep;
58 
59 #ifndef CONFIG_MMU
60 extern struct rb_root nommu_region_tree;
61 extern struct rw_semaphore nommu_region_sem;
62 
63 extern unsigned int kobjsize(const void *objp);
64 #endif
65 
66 /*
67  * vm_flags in vm_area_struct, see mm_types.h.
68  */
69 #define VM_READ		0x00000001	/* currently active flags */
70 #define VM_WRITE	0x00000002
71 #define VM_EXEC		0x00000004
72 #define VM_SHARED	0x00000008
73 
74 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
75 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
76 #define VM_MAYWRITE	0x00000020
77 #define VM_MAYEXEC	0x00000040
78 #define VM_MAYSHARE	0x00000080
79 
80 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
81 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
82 #define VM_GROWSUP	0x00000200
83 #else
84 #define VM_GROWSUP	0x00000000
85 #endif
86 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
87 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
88 
89 #define VM_EXECUTABLE	0x00001000
90 #define VM_LOCKED	0x00002000
91 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
92 
93 					/* Used by sys_madvise() */
94 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
95 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
96 
97 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
98 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
99 #define VM_RESERVED	0x00080000	/* Count as reserved_vm like IO */
100 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
101 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
102 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
103 #define VM_NONLINEAR	0x00800000	/* Is non-linear (remap_file_pages) */
104 #define VM_MAPPED_COPY	0x01000000	/* T if mapped copy of data (nommu mmap) */
105 #define VM_INSERTPAGE	0x02000000	/* The vma has had "vm_insert_page()" done on it */
106 #define VM_ALWAYSDUMP	0x04000000	/* Always include in core dumps */
107 
108 #define VM_CAN_NONLINEAR 0x08000000	/* Has ->fault & does nonlinear pages */
109 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
110 #define VM_SAO		0x20000000	/* Strong Access Ordering (powerpc) */
111 #define VM_PFN_AT_MMAP	0x40000000	/* PFNMAP vma that is fully mapped at mmap time */
112 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
113 
114 /* Bits set in the VMA until the stack is in its final location */
115 #define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
116 
117 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
118 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
119 #endif
120 
121 #ifdef CONFIG_STACK_GROWSUP
122 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
123 #else
124 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
125 #endif
126 
127 #define VM_READHINTMASK			(VM_SEQ_READ | VM_RAND_READ)
128 #define VM_ClearReadHint(v)		(v)->vm_flags &= ~VM_READHINTMASK
129 #define VM_NormalReadHint(v)		(!((v)->vm_flags & VM_READHINTMASK))
130 #define VM_SequentialReadHint(v)	((v)->vm_flags & VM_SEQ_READ)
131 #define VM_RandomReadHint(v)		((v)->vm_flags & VM_RAND_READ)
132 
133 /*
134  * special vmas that are non-mergable, non-mlock()able
135  */
136 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
137 
138 /*
139  * mapping from the currently active vm_flags protection bits (the
140  * low four bits) to a page protection mask..
141  */
142 extern pgprot_t protection_map[16];
143 
144 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
145 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
146 #define FAULT_FLAG_MKWRITE	0x04	/* Fault was mkwrite of existing pte */
147 
148 /*
149  * This interface is used by x86 PAT code to identify a pfn mapping that is
150  * linear over entire vma. This is to optimize PAT code that deals with
151  * marking the physical region with a particular prot. This is not for generic
152  * mm use. Note also that this check will not work if the pfn mapping is
153  * linear for a vma starting at physical address 0. In which case PAT code
154  * falls back to slow path of reserving physical range page by page.
155  */
156 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
157 {
158 	return (vma->vm_flags & VM_PFN_AT_MMAP);
159 }
160 
161 static inline int is_pfn_mapping(struct vm_area_struct *vma)
162 {
163 	return (vma->vm_flags & VM_PFNMAP);
164 }
165 
166 /*
167  * vm_fault is filled by the the pagefault handler and passed to the vma's
168  * ->fault function. The vma's ->fault is responsible for returning a bitmask
169  * of VM_FAULT_xxx flags that give details about how the fault was handled.
170  *
171  * pgoff should be used in favour of virtual_address, if possible. If pgoff
172  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
173  * mapping support.
174  */
175 struct vm_fault {
176 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
177 	pgoff_t pgoff;			/* Logical page offset based on vma */
178 	void __user *virtual_address;	/* Faulting virtual address */
179 
180 	struct page *page;		/* ->fault handlers should return a
181 					 * page here, unless VM_FAULT_NOPAGE
182 					 * is set (which is also implied by
183 					 * VM_FAULT_ERROR).
184 					 */
185 };
186 
187 /*
188  * These are the virtual MM functions - opening of an area, closing and
189  * unmapping it (needed to keep files on disk up-to-date etc), pointer
190  * to the functions called when a no-page or a wp-page exception occurs.
191  */
192 struct vm_operations_struct {
193 	void (*open)(struct vm_area_struct * area);
194 	void (*close)(struct vm_area_struct * area);
195 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
196 
197 	/* notification that a previously read-only page is about to become
198 	 * writable, if an error is returned it will cause a SIGBUS */
199 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
200 
201 	/* called by access_process_vm when get_user_pages() fails, typically
202 	 * for use by special VMAs that can switch between memory and hardware
203 	 */
204 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
205 		      void *buf, int len, int write);
206 #ifdef CONFIG_NUMA
207 	/*
208 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
209 	 * to hold the policy upon return.  Caller should pass NULL @new to
210 	 * remove a policy and fall back to surrounding context--i.e. do not
211 	 * install a MPOL_DEFAULT policy, nor the task or system default
212 	 * mempolicy.
213 	 */
214 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
215 
216 	/*
217 	 * get_policy() op must add reference [mpol_get()] to any policy at
218 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
219 	 * in mm/mempolicy.c will do this automatically.
220 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
221 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
222 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
223 	 * must return NULL--i.e., do not "fallback" to task or system default
224 	 * policy.
225 	 */
226 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
227 					unsigned long addr);
228 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
229 		const nodemask_t *to, unsigned long flags);
230 #endif
231 };
232 
233 struct mmu_gather;
234 struct inode;
235 
236 #define page_private(page)		((page)->private)
237 #define set_page_private(page, v)	((page)->private = (v))
238 
239 /*
240  * FIXME: take this include out, include page-flags.h in
241  * files which need it (119 of them)
242  */
243 #include <linux/page-flags.h>
244 
245 /*
246  * Methods to modify the page usage count.
247  *
248  * What counts for a page usage:
249  * - cache mapping   (page->mapping)
250  * - private data    (page->private)
251  * - page mapped in a task's page tables, each mapping
252  *   is counted separately
253  *
254  * Also, many kernel routines increase the page count before a critical
255  * routine so they can be sure the page doesn't go away from under them.
256  */
257 
258 /*
259  * Drop a ref, return true if the refcount fell to zero (the page has no users)
260  */
261 static inline int put_page_testzero(struct page *page)
262 {
263 	VM_BUG_ON(atomic_read(&page->_count) == 0);
264 	return atomic_dec_and_test(&page->_count);
265 }
266 
267 /*
268  * Try to grab a ref unless the page has a refcount of zero, return false if
269  * that is the case.
270  */
271 static inline int get_page_unless_zero(struct page *page)
272 {
273 	return atomic_inc_not_zero(&page->_count);
274 }
275 
276 extern int page_is_ram(unsigned long pfn);
277 
278 /* Support for virtually mapped pages */
279 struct page *vmalloc_to_page(const void *addr);
280 unsigned long vmalloc_to_pfn(const void *addr);
281 
282 /*
283  * Determine if an address is within the vmalloc range
284  *
285  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
286  * is no special casing required.
287  */
288 static inline int is_vmalloc_addr(const void *x)
289 {
290 #ifdef CONFIG_MMU
291 	unsigned long addr = (unsigned long)x;
292 
293 	return addr >= VMALLOC_START && addr < VMALLOC_END;
294 #else
295 	return 0;
296 #endif
297 }
298 #ifdef CONFIG_MMU
299 extern int is_vmalloc_or_module_addr(const void *x);
300 #else
301 static inline int is_vmalloc_or_module_addr(const void *x)
302 {
303 	return 0;
304 }
305 #endif
306 
307 static inline struct page *compound_head(struct page *page)
308 {
309 	if (unlikely(PageTail(page)))
310 		return page->first_page;
311 	return page;
312 }
313 
314 static inline int page_count(struct page *page)
315 {
316 	return atomic_read(&compound_head(page)->_count);
317 }
318 
319 static inline void get_page(struct page *page)
320 {
321 	page = compound_head(page);
322 	VM_BUG_ON(atomic_read(&page->_count) == 0);
323 	atomic_inc(&page->_count);
324 }
325 
326 static inline struct page *virt_to_head_page(const void *x)
327 {
328 	struct page *page = virt_to_page(x);
329 	return compound_head(page);
330 }
331 
332 /*
333  * Setup the page count before being freed into the page allocator for
334  * the first time (boot or memory hotplug)
335  */
336 static inline void init_page_count(struct page *page)
337 {
338 	atomic_set(&page->_count, 1);
339 }
340 
341 void put_page(struct page *page);
342 void put_pages_list(struct list_head *pages);
343 
344 void split_page(struct page *page, unsigned int order);
345 int split_free_page(struct page *page);
346 
347 /*
348  * Compound pages have a destructor function.  Provide a
349  * prototype for that function and accessor functions.
350  * These are _only_ valid on the head of a PG_compound page.
351  */
352 typedef void compound_page_dtor(struct page *);
353 
354 static inline void set_compound_page_dtor(struct page *page,
355 						compound_page_dtor *dtor)
356 {
357 	page[1].lru.next = (void *)dtor;
358 }
359 
360 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
361 {
362 	return (compound_page_dtor *)page[1].lru.next;
363 }
364 
365 static inline int compound_order(struct page *page)
366 {
367 	if (!PageHead(page))
368 		return 0;
369 	return (unsigned long)page[1].lru.prev;
370 }
371 
372 static inline void set_compound_order(struct page *page, unsigned long order)
373 {
374 	page[1].lru.prev = (void *)order;
375 }
376 
377 /*
378  * Multiple processes may "see" the same page. E.g. for untouched
379  * mappings of /dev/null, all processes see the same page full of
380  * zeroes, and text pages of executables and shared libraries have
381  * only one copy in memory, at most, normally.
382  *
383  * For the non-reserved pages, page_count(page) denotes a reference count.
384  *   page_count() == 0 means the page is free. page->lru is then used for
385  *   freelist management in the buddy allocator.
386  *   page_count() > 0  means the page has been allocated.
387  *
388  * Pages are allocated by the slab allocator in order to provide memory
389  * to kmalloc and kmem_cache_alloc. In this case, the management of the
390  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
391  * unless a particular usage is carefully commented. (the responsibility of
392  * freeing the kmalloc memory is the caller's, of course).
393  *
394  * A page may be used by anyone else who does a __get_free_page().
395  * In this case, page_count still tracks the references, and should only
396  * be used through the normal accessor functions. The top bits of page->flags
397  * and page->virtual store page management information, but all other fields
398  * are unused and could be used privately, carefully. The management of this
399  * page is the responsibility of the one who allocated it, and those who have
400  * subsequently been given references to it.
401  *
402  * The other pages (we may call them "pagecache pages") are completely
403  * managed by the Linux memory manager: I/O, buffers, swapping etc.
404  * The following discussion applies only to them.
405  *
406  * A pagecache page contains an opaque `private' member, which belongs to the
407  * page's address_space. Usually, this is the address of a circular list of
408  * the page's disk buffers. PG_private must be set to tell the VM to call
409  * into the filesystem to release these pages.
410  *
411  * A page may belong to an inode's memory mapping. In this case, page->mapping
412  * is the pointer to the inode, and page->index is the file offset of the page,
413  * in units of PAGE_CACHE_SIZE.
414  *
415  * If pagecache pages are not associated with an inode, they are said to be
416  * anonymous pages. These may become associated with the swapcache, and in that
417  * case PG_swapcache is set, and page->private is an offset into the swapcache.
418  *
419  * In either case (swapcache or inode backed), the pagecache itself holds one
420  * reference to the page. Setting PG_private should also increment the
421  * refcount. The each user mapping also has a reference to the page.
422  *
423  * The pagecache pages are stored in a per-mapping radix tree, which is
424  * rooted at mapping->page_tree, and indexed by offset.
425  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
426  * lists, we instead now tag pages as dirty/writeback in the radix tree.
427  *
428  * All pagecache pages may be subject to I/O:
429  * - inode pages may need to be read from disk,
430  * - inode pages which have been modified and are MAP_SHARED may need
431  *   to be written back to the inode on disk,
432  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
433  *   modified may need to be swapped out to swap space and (later) to be read
434  *   back into memory.
435  */
436 
437 /*
438  * The zone field is never updated after free_area_init_core()
439  * sets it, so none of the operations on it need to be atomic.
440  */
441 
442 
443 /*
444  * page->flags layout:
445  *
446  * There are three possibilities for how page->flags get
447  * laid out.  The first is for the normal case, without
448  * sparsemem.  The second is for sparsemem when there is
449  * plenty of space for node and section.  The last is when
450  * we have run out of space and have to fall back to an
451  * alternate (slower) way of determining the node.
452  *
453  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
454  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
455  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
456  */
457 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
458 #define SECTIONS_WIDTH		SECTIONS_SHIFT
459 #else
460 #define SECTIONS_WIDTH		0
461 #endif
462 
463 #define ZONES_WIDTH		ZONES_SHIFT
464 
465 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
466 #define NODES_WIDTH		NODES_SHIFT
467 #else
468 #ifdef CONFIG_SPARSEMEM_VMEMMAP
469 #error "Vmemmap: No space for nodes field in page flags"
470 #endif
471 #define NODES_WIDTH		0
472 #endif
473 
474 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
475 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
476 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
477 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
478 
479 /*
480  * We are going to use the flags for the page to node mapping if its in
481  * there.  This includes the case where there is no node, so it is implicit.
482  */
483 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
484 #define NODE_NOT_IN_PAGE_FLAGS
485 #endif
486 
487 #ifndef PFN_SECTION_SHIFT
488 #define PFN_SECTION_SHIFT 0
489 #endif
490 
491 /*
492  * Define the bit shifts to access each section.  For non-existant
493  * sections we define the shift as 0; that plus a 0 mask ensures
494  * the compiler will optimise away reference to them.
495  */
496 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
497 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
498 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
499 
500 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
501 #ifdef NODE_NOT_IN_PAGE_FLAGS
502 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
503 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
504 						SECTIONS_PGOFF : ZONES_PGOFF)
505 #else
506 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
507 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
508 						NODES_PGOFF : ZONES_PGOFF)
509 #endif
510 
511 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
512 
513 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
514 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
515 #endif
516 
517 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
518 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
519 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
520 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
521 
522 static inline enum zone_type page_zonenum(struct page *page)
523 {
524 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
525 }
526 
527 /*
528  * The identification function is only used by the buddy allocator for
529  * determining if two pages could be buddies. We are not really
530  * identifying a zone since we could be using a the section number
531  * id if we have not node id available in page flags.
532  * We guarantee only that it will return the same value for two
533  * combinable pages in a zone.
534  */
535 static inline int page_zone_id(struct page *page)
536 {
537 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
538 }
539 
540 static inline int zone_to_nid(struct zone *zone)
541 {
542 #ifdef CONFIG_NUMA
543 	return zone->node;
544 #else
545 	return 0;
546 #endif
547 }
548 
549 #ifdef NODE_NOT_IN_PAGE_FLAGS
550 extern int page_to_nid(struct page *page);
551 #else
552 static inline int page_to_nid(struct page *page)
553 {
554 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
555 }
556 #endif
557 
558 static inline struct zone *page_zone(struct page *page)
559 {
560 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
561 }
562 
563 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
564 static inline unsigned long page_to_section(struct page *page)
565 {
566 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
567 }
568 #endif
569 
570 static inline void set_page_zone(struct page *page, enum zone_type zone)
571 {
572 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
573 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
574 }
575 
576 static inline void set_page_node(struct page *page, unsigned long node)
577 {
578 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
579 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
580 }
581 
582 static inline void set_page_section(struct page *page, unsigned long section)
583 {
584 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
585 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
586 }
587 
588 static inline void set_page_links(struct page *page, enum zone_type zone,
589 	unsigned long node, unsigned long pfn)
590 {
591 	set_page_zone(page, zone);
592 	set_page_node(page, node);
593 	set_page_section(page, pfn_to_section_nr(pfn));
594 }
595 
596 /*
597  * Some inline functions in vmstat.h depend on page_zone()
598  */
599 #include <linux/vmstat.h>
600 
601 static __always_inline void *lowmem_page_address(struct page *page)
602 {
603 	return __va(PFN_PHYS(page_to_pfn(page)));
604 }
605 
606 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
607 #define HASHED_PAGE_VIRTUAL
608 #endif
609 
610 #if defined(WANT_PAGE_VIRTUAL)
611 #define page_address(page) ((page)->virtual)
612 #define set_page_address(page, address)			\
613 	do {						\
614 		(page)->virtual = (address);		\
615 	} while(0)
616 #define page_address_init()  do { } while(0)
617 #endif
618 
619 #if defined(HASHED_PAGE_VIRTUAL)
620 void *page_address(struct page *page);
621 void set_page_address(struct page *page, void *virtual);
622 void page_address_init(void);
623 #endif
624 
625 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
626 #define page_address(page) lowmem_page_address(page)
627 #define set_page_address(page, address)  do { } while(0)
628 #define page_address_init()  do { } while(0)
629 #endif
630 
631 /*
632  * On an anonymous page mapped into a user virtual memory area,
633  * page->mapping points to its anon_vma, not to a struct address_space;
634  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
635  *
636  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
637  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
638  * and then page->mapping points, not to an anon_vma, but to a private
639  * structure which KSM associates with that merged page.  See ksm.h.
640  *
641  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
642  *
643  * Please note that, confusingly, "page_mapping" refers to the inode
644  * address_space which maps the page from disk; whereas "page_mapped"
645  * refers to user virtual address space into which the page is mapped.
646  */
647 #define PAGE_MAPPING_ANON	1
648 #define PAGE_MAPPING_KSM	2
649 #define PAGE_MAPPING_FLAGS	(PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
650 
651 extern struct address_space swapper_space;
652 static inline struct address_space *page_mapping(struct page *page)
653 {
654 	struct address_space *mapping = page->mapping;
655 
656 	VM_BUG_ON(PageSlab(page));
657 	if (unlikely(PageSwapCache(page)))
658 		mapping = &swapper_space;
659 	else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
660 		mapping = NULL;
661 	return mapping;
662 }
663 
664 /* Neutral page->mapping pointer to address_space or anon_vma or other */
665 static inline void *page_rmapping(struct page *page)
666 {
667 	return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
668 }
669 
670 static inline int PageAnon(struct page *page)
671 {
672 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
673 }
674 
675 /*
676  * Return the pagecache index of the passed page.  Regular pagecache pages
677  * use ->index whereas swapcache pages use ->private
678  */
679 static inline pgoff_t page_index(struct page *page)
680 {
681 	if (unlikely(PageSwapCache(page)))
682 		return page_private(page);
683 	return page->index;
684 }
685 
686 /*
687  * The atomic page->_mapcount, like _count, starts from -1:
688  * so that transitions both from it and to it can be tracked,
689  * using atomic_inc_and_test and atomic_add_negative(-1).
690  */
691 static inline void reset_page_mapcount(struct page *page)
692 {
693 	atomic_set(&(page)->_mapcount, -1);
694 }
695 
696 static inline int page_mapcount(struct page *page)
697 {
698 	return atomic_read(&(page)->_mapcount) + 1;
699 }
700 
701 /*
702  * Return true if this page is mapped into pagetables.
703  */
704 static inline int page_mapped(struct page *page)
705 {
706 	return atomic_read(&(page)->_mapcount) >= 0;
707 }
708 
709 /*
710  * Different kinds of faults, as returned by handle_mm_fault().
711  * Used to decide whether a process gets delivered SIGBUS or
712  * just gets major/minor fault counters bumped up.
713  */
714 
715 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
716 
717 #define VM_FAULT_OOM	0x0001
718 #define VM_FAULT_SIGBUS	0x0002
719 #define VM_FAULT_MAJOR	0x0004
720 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
721 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
722 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
723 
724 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
725 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
726 
727 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
728 
729 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
730 			 VM_FAULT_HWPOISON_LARGE)
731 
732 /* Encode hstate index for a hwpoisoned large page */
733 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
734 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
735 
736 /*
737  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
738  */
739 extern void pagefault_out_of_memory(void);
740 
741 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
742 
743 extern void show_free_areas(void);
744 
745 int shmem_lock(struct file *file, int lock, struct user_struct *user);
746 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
747 int shmem_zero_setup(struct vm_area_struct *);
748 
749 #ifndef CONFIG_MMU
750 extern unsigned long shmem_get_unmapped_area(struct file *file,
751 					     unsigned long addr,
752 					     unsigned long len,
753 					     unsigned long pgoff,
754 					     unsigned long flags);
755 #endif
756 
757 extern int can_do_mlock(void);
758 extern int user_shm_lock(size_t, struct user_struct *);
759 extern void user_shm_unlock(size_t, struct user_struct *);
760 
761 /*
762  * Parameter block passed down to zap_pte_range in exceptional cases.
763  */
764 struct zap_details {
765 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
766 	struct address_space *check_mapping;	/* Check page->mapping if set */
767 	pgoff_t	first_index;			/* Lowest page->index to unmap */
768 	pgoff_t last_index;			/* Highest page->index to unmap */
769 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
770 	unsigned long truncate_count;		/* Compare vm_truncate_count */
771 };
772 
773 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
774 		pte_t pte);
775 
776 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
777 		unsigned long size);
778 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
779 		unsigned long size, struct zap_details *);
780 unsigned long unmap_vmas(struct mmu_gather **tlb,
781 		struct vm_area_struct *start_vma, unsigned long start_addr,
782 		unsigned long end_addr, unsigned long *nr_accounted,
783 		struct zap_details *);
784 
785 /**
786  * mm_walk - callbacks for walk_page_range
787  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
788  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
789  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
790  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
791  * @pte_hole: if set, called for each hole at all levels
792  * @hugetlb_entry: if set, called for each hugetlb entry
793  *
794  * (see walk_page_range for more details)
795  */
796 struct mm_walk {
797 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
798 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
799 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
800 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
801 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
802 	int (*hugetlb_entry)(pte_t *, unsigned long,
803 			     unsigned long, unsigned long, struct mm_walk *);
804 	struct mm_struct *mm;
805 	void *private;
806 };
807 
808 int walk_page_range(unsigned long addr, unsigned long end,
809 		struct mm_walk *walk);
810 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
811 		unsigned long end, unsigned long floor, unsigned long ceiling);
812 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
813 			struct vm_area_struct *vma);
814 void unmap_mapping_range(struct address_space *mapping,
815 		loff_t const holebegin, loff_t const holelen, int even_cows);
816 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
817 	unsigned long *pfn);
818 int follow_phys(struct vm_area_struct *vma, unsigned long address,
819 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
820 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
821 			void *buf, int len, int write);
822 
823 static inline void unmap_shared_mapping_range(struct address_space *mapping,
824 		loff_t const holebegin, loff_t const holelen)
825 {
826 	unmap_mapping_range(mapping, holebegin, holelen, 0);
827 }
828 
829 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
830 extern void truncate_setsize(struct inode *inode, loff_t newsize);
831 extern int vmtruncate(struct inode *inode, loff_t offset);
832 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
833 
834 int truncate_inode_page(struct address_space *mapping, struct page *page);
835 int generic_error_remove_page(struct address_space *mapping, struct page *page);
836 
837 int invalidate_inode_page(struct page *page);
838 
839 #ifdef CONFIG_MMU
840 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
841 			unsigned long address, unsigned int flags);
842 #else
843 static inline int handle_mm_fault(struct mm_struct *mm,
844 			struct vm_area_struct *vma, unsigned long address,
845 			unsigned int flags)
846 {
847 	/* should never happen if there's no MMU */
848 	BUG();
849 	return VM_FAULT_SIGBUS;
850 }
851 #endif
852 
853 extern int make_pages_present(unsigned long addr, unsigned long end);
854 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
855 
856 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
857 			unsigned long start, int nr_pages, int write, int force,
858 			struct page **pages, struct vm_area_struct **vmas);
859 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
860 			struct page **pages);
861 struct page *get_dump_page(unsigned long addr);
862 
863 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
864 extern void do_invalidatepage(struct page *page, unsigned long offset);
865 
866 int __set_page_dirty_nobuffers(struct page *page);
867 int __set_page_dirty_no_writeback(struct page *page);
868 int redirty_page_for_writepage(struct writeback_control *wbc,
869 				struct page *page);
870 void account_page_dirtied(struct page *page, struct address_space *mapping);
871 void account_page_writeback(struct page *page);
872 int set_page_dirty(struct page *page);
873 int set_page_dirty_lock(struct page *page);
874 int clear_page_dirty_for_io(struct page *page);
875 
876 /* Is the vma a continuation of the stack vma above it? */
877 static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr)
878 {
879 	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
880 }
881 
882 extern unsigned long move_page_tables(struct vm_area_struct *vma,
883 		unsigned long old_addr, struct vm_area_struct *new_vma,
884 		unsigned long new_addr, unsigned long len);
885 extern unsigned long do_mremap(unsigned long addr,
886 			       unsigned long old_len, unsigned long new_len,
887 			       unsigned long flags, unsigned long new_addr);
888 extern int mprotect_fixup(struct vm_area_struct *vma,
889 			  struct vm_area_struct **pprev, unsigned long start,
890 			  unsigned long end, unsigned long newflags);
891 
892 /*
893  * doesn't attempt to fault and will return short.
894  */
895 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
896 			  struct page **pages);
897 /*
898  * per-process(per-mm_struct) statistics.
899  */
900 #if defined(SPLIT_RSS_COUNTING)
901 /*
902  * The mm counters are not protected by its page_table_lock,
903  * so must be incremented atomically.
904  */
905 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
906 {
907 	atomic_long_set(&mm->rss_stat.count[member], value);
908 }
909 
910 unsigned long get_mm_counter(struct mm_struct *mm, int member);
911 
912 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
913 {
914 	atomic_long_add(value, &mm->rss_stat.count[member]);
915 }
916 
917 static inline void inc_mm_counter(struct mm_struct *mm, int member)
918 {
919 	atomic_long_inc(&mm->rss_stat.count[member]);
920 }
921 
922 static inline void dec_mm_counter(struct mm_struct *mm, int member)
923 {
924 	atomic_long_dec(&mm->rss_stat.count[member]);
925 }
926 
927 #else  /* !USE_SPLIT_PTLOCKS */
928 /*
929  * The mm counters are protected by its page_table_lock,
930  * so can be incremented directly.
931  */
932 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
933 {
934 	mm->rss_stat.count[member] = value;
935 }
936 
937 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
938 {
939 	return mm->rss_stat.count[member];
940 }
941 
942 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
943 {
944 	mm->rss_stat.count[member] += value;
945 }
946 
947 static inline void inc_mm_counter(struct mm_struct *mm, int member)
948 {
949 	mm->rss_stat.count[member]++;
950 }
951 
952 static inline void dec_mm_counter(struct mm_struct *mm, int member)
953 {
954 	mm->rss_stat.count[member]--;
955 }
956 
957 #endif /* !USE_SPLIT_PTLOCKS */
958 
959 static inline unsigned long get_mm_rss(struct mm_struct *mm)
960 {
961 	return get_mm_counter(mm, MM_FILEPAGES) +
962 		get_mm_counter(mm, MM_ANONPAGES);
963 }
964 
965 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
966 {
967 	return max(mm->hiwater_rss, get_mm_rss(mm));
968 }
969 
970 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
971 {
972 	return max(mm->hiwater_vm, mm->total_vm);
973 }
974 
975 static inline void update_hiwater_rss(struct mm_struct *mm)
976 {
977 	unsigned long _rss = get_mm_rss(mm);
978 
979 	if ((mm)->hiwater_rss < _rss)
980 		(mm)->hiwater_rss = _rss;
981 }
982 
983 static inline void update_hiwater_vm(struct mm_struct *mm)
984 {
985 	if (mm->hiwater_vm < mm->total_vm)
986 		mm->hiwater_vm = mm->total_vm;
987 }
988 
989 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
990 					 struct mm_struct *mm)
991 {
992 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
993 
994 	if (*maxrss < hiwater_rss)
995 		*maxrss = hiwater_rss;
996 }
997 
998 #if defined(SPLIT_RSS_COUNTING)
999 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
1000 #else
1001 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1002 {
1003 }
1004 #endif
1005 
1006 /*
1007  * A callback you can register to apply pressure to ageable caches.
1008  *
1009  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
1010  * look through the least-recently-used 'nr_to_scan' entries and
1011  * attempt to free them up.  It should return the number of objects
1012  * which remain in the cache.  If it returns -1, it means it cannot do
1013  * any scanning at this time (eg. there is a risk of deadlock).
1014  *
1015  * The 'gfpmask' refers to the allocation we are currently trying to
1016  * fulfil.
1017  *
1018  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1019  * querying the cache size, so a fastpath for that case is appropriate.
1020  */
1021 struct shrinker {
1022 	int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask);
1023 	int seeks;	/* seeks to recreate an obj */
1024 
1025 	/* These are for internal use */
1026 	struct list_head list;
1027 	long nr;	/* objs pending delete */
1028 };
1029 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1030 extern void register_shrinker(struct shrinker *);
1031 extern void unregister_shrinker(struct shrinker *);
1032 
1033 int vma_wants_writenotify(struct vm_area_struct *vma);
1034 
1035 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
1036 
1037 #ifdef __PAGETABLE_PUD_FOLDED
1038 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1039 						unsigned long address)
1040 {
1041 	return 0;
1042 }
1043 #else
1044 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1045 #endif
1046 
1047 #ifdef __PAGETABLE_PMD_FOLDED
1048 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1049 						unsigned long address)
1050 {
1051 	return 0;
1052 }
1053 #else
1054 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1055 #endif
1056 
1057 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1058 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1059 
1060 /*
1061  * The following ifdef needed to get the 4level-fixup.h header to work.
1062  * Remove it when 4level-fixup.h has been removed.
1063  */
1064 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1065 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1066 {
1067 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1068 		NULL: pud_offset(pgd, address);
1069 }
1070 
1071 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1072 {
1073 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1074 		NULL: pmd_offset(pud, address);
1075 }
1076 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1077 
1078 #if USE_SPLIT_PTLOCKS
1079 /*
1080  * We tuck a spinlock to guard each pagetable page into its struct page,
1081  * at page->private, with BUILD_BUG_ON to make sure that this will not
1082  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1083  * When freeing, reset page->mapping so free_pages_check won't complain.
1084  */
1085 #define __pte_lockptr(page)	&((page)->ptl)
1086 #define pte_lock_init(_page)	do {					\
1087 	spin_lock_init(__pte_lockptr(_page));				\
1088 } while (0)
1089 #define pte_lock_deinit(page)	((page)->mapping = NULL)
1090 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1091 #else	/* !USE_SPLIT_PTLOCKS */
1092 /*
1093  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1094  */
1095 #define pte_lock_init(page)	do {} while (0)
1096 #define pte_lock_deinit(page)	do {} while (0)
1097 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
1098 #endif /* USE_SPLIT_PTLOCKS */
1099 
1100 static inline void pgtable_page_ctor(struct page *page)
1101 {
1102 	pte_lock_init(page);
1103 	inc_zone_page_state(page, NR_PAGETABLE);
1104 }
1105 
1106 static inline void pgtable_page_dtor(struct page *page)
1107 {
1108 	pte_lock_deinit(page);
1109 	dec_zone_page_state(page, NR_PAGETABLE);
1110 }
1111 
1112 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1113 ({							\
1114 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1115 	pte_t *__pte = pte_offset_map(pmd, address);	\
1116 	*(ptlp) = __ptl;				\
1117 	spin_lock(__ptl);				\
1118 	__pte;						\
1119 })
1120 
1121 #define pte_unmap_unlock(pte, ptl)	do {		\
1122 	spin_unlock(ptl);				\
1123 	pte_unmap(pte);					\
1124 } while (0)
1125 
1126 #define pte_alloc_map(mm, pmd, address)			\
1127 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1128 		NULL: pte_offset_map(pmd, address))
1129 
1130 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1131 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1132 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1133 
1134 #define pte_alloc_kernel(pmd, address)			\
1135 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1136 		NULL: pte_offset_kernel(pmd, address))
1137 
1138 extern void free_area_init(unsigned long * zones_size);
1139 extern void free_area_init_node(int nid, unsigned long * zones_size,
1140 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1141 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1142 /*
1143  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1144  * zones, allocate the backing mem_map and account for memory holes in a more
1145  * architecture independent manner. This is a substitute for creating the
1146  * zone_sizes[] and zholes_size[] arrays and passing them to
1147  * free_area_init_node()
1148  *
1149  * An architecture is expected to register range of page frames backed by
1150  * physical memory with add_active_range() before calling
1151  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1152  * usage, an architecture is expected to do something like
1153  *
1154  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1155  * 							 max_highmem_pfn};
1156  * for_each_valid_physical_page_range()
1157  * 	add_active_range(node_id, start_pfn, end_pfn)
1158  * free_area_init_nodes(max_zone_pfns);
1159  *
1160  * If the architecture guarantees that there are no holes in the ranges
1161  * registered with add_active_range(), free_bootmem_active_regions()
1162  * will call free_bootmem_node() for each registered physical page range.
1163  * Similarly sparse_memory_present_with_active_regions() calls
1164  * memory_present() for each range when SPARSEMEM is enabled.
1165  *
1166  * See mm/page_alloc.c for more information on each function exposed by
1167  * CONFIG_ARCH_POPULATES_NODE_MAP
1168  */
1169 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1170 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1171 					unsigned long end_pfn);
1172 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1173 					unsigned long end_pfn);
1174 extern void remove_all_active_ranges(void);
1175 void sort_node_map(void);
1176 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1177 						unsigned long end_pfn);
1178 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1179 						unsigned long end_pfn);
1180 extern void get_pfn_range_for_nid(unsigned int nid,
1181 			unsigned long *start_pfn, unsigned long *end_pfn);
1182 extern unsigned long find_min_pfn_with_active_regions(void);
1183 extern void free_bootmem_with_active_regions(int nid,
1184 						unsigned long max_low_pfn);
1185 int add_from_early_node_map(struct range *range, int az,
1186 				   int nr_range, int nid);
1187 u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1188 					u64 goal, u64 limit);
1189 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
1190 				 u64 goal, u64 limit);
1191 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1192 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1193 extern void sparse_memory_present_with_active_regions(int nid);
1194 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1195 
1196 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1197     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1198 static inline int __early_pfn_to_nid(unsigned long pfn)
1199 {
1200 	return 0;
1201 }
1202 #else
1203 /* please see mm/page_alloc.c */
1204 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1205 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1206 /* there is a per-arch backend function. */
1207 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1208 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1209 #endif
1210 
1211 extern void set_dma_reserve(unsigned long new_dma_reserve);
1212 extern void memmap_init_zone(unsigned long, int, unsigned long,
1213 				unsigned long, enum memmap_context);
1214 extern void setup_per_zone_wmarks(void);
1215 extern void calculate_zone_inactive_ratio(struct zone *zone);
1216 extern void mem_init(void);
1217 extern void __init mmap_init(void);
1218 extern void show_mem(void);
1219 extern void si_meminfo(struct sysinfo * val);
1220 extern void si_meminfo_node(struct sysinfo *val, int nid);
1221 extern int after_bootmem;
1222 
1223 extern void setup_per_cpu_pageset(void);
1224 
1225 extern void zone_pcp_update(struct zone *zone);
1226 
1227 /* nommu.c */
1228 extern atomic_long_t mmap_pages_allocated;
1229 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1230 
1231 /* prio_tree.c */
1232 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1233 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1234 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1235 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1236 	struct prio_tree_iter *iter);
1237 
1238 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1239 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1240 		(vma = vma_prio_tree_next(vma, iter)); )
1241 
1242 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1243 					struct list_head *list)
1244 {
1245 	vma->shared.vm_set.parent = NULL;
1246 	list_add_tail(&vma->shared.vm_set.list, list);
1247 }
1248 
1249 /* mmap.c */
1250 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1251 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1252 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1253 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1254 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1255 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1256 	struct mempolicy *);
1257 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1258 extern int split_vma(struct mm_struct *,
1259 	struct vm_area_struct *, unsigned long addr, int new_below);
1260 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1261 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1262 	struct rb_node **, struct rb_node *);
1263 extern void unlink_file_vma(struct vm_area_struct *);
1264 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1265 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1266 extern void exit_mmap(struct mm_struct *);
1267 
1268 extern int mm_take_all_locks(struct mm_struct *mm);
1269 extern void mm_drop_all_locks(struct mm_struct *mm);
1270 
1271 #ifdef CONFIG_PROC_FS
1272 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1273 extern void added_exe_file_vma(struct mm_struct *mm);
1274 extern void removed_exe_file_vma(struct mm_struct *mm);
1275 #else
1276 static inline void added_exe_file_vma(struct mm_struct *mm)
1277 {}
1278 
1279 static inline void removed_exe_file_vma(struct mm_struct *mm)
1280 {}
1281 #endif /* CONFIG_PROC_FS */
1282 
1283 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1284 extern int install_special_mapping(struct mm_struct *mm,
1285 				   unsigned long addr, unsigned long len,
1286 				   unsigned long flags, struct page **pages);
1287 
1288 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1289 
1290 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1291 	unsigned long len, unsigned long prot,
1292 	unsigned long flag, unsigned long pgoff);
1293 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1294 	unsigned long len, unsigned long flags,
1295 	unsigned int vm_flags, unsigned long pgoff);
1296 
1297 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1298 	unsigned long len, unsigned long prot,
1299 	unsigned long flag, unsigned long offset)
1300 {
1301 	unsigned long ret = -EINVAL;
1302 	if ((offset + PAGE_ALIGN(len)) < offset)
1303 		goto out;
1304 	if (!(offset & ~PAGE_MASK))
1305 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1306 out:
1307 	return ret;
1308 }
1309 
1310 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1311 
1312 extern unsigned long do_brk(unsigned long, unsigned long);
1313 
1314 /* filemap.c */
1315 extern unsigned long page_unuse(struct page *);
1316 extern void truncate_inode_pages(struct address_space *, loff_t);
1317 extern void truncate_inode_pages_range(struct address_space *,
1318 				       loff_t lstart, loff_t lend);
1319 
1320 /* generic vm_area_ops exported for stackable file systems */
1321 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1322 
1323 /* mm/page-writeback.c */
1324 int write_one_page(struct page *page, int wait);
1325 void task_dirty_inc(struct task_struct *tsk);
1326 
1327 /* readahead.c */
1328 #define VM_MAX_READAHEAD	128	/* kbytes */
1329 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1330 
1331 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1332 			pgoff_t offset, unsigned long nr_to_read);
1333 
1334 void page_cache_sync_readahead(struct address_space *mapping,
1335 			       struct file_ra_state *ra,
1336 			       struct file *filp,
1337 			       pgoff_t offset,
1338 			       unsigned long size);
1339 
1340 void page_cache_async_readahead(struct address_space *mapping,
1341 				struct file_ra_state *ra,
1342 				struct file *filp,
1343 				struct page *pg,
1344 				pgoff_t offset,
1345 				unsigned long size);
1346 
1347 unsigned long max_sane_readahead(unsigned long nr);
1348 unsigned long ra_submit(struct file_ra_state *ra,
1349 			struct address_space *mapping,
1350 			struct file *filp);
1351 
1352 /* Do stack extension */
1353 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1354 #if VM_GROWSUP
1355 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1356 #else
1357   #define expand_upwards(vma, address) do { } while (0)
1358 #endif
1359 extern int expand_stack_downwards(struct vm_area_struct *vma,
1360 				  unsigned long address);
1361 
1362 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1363 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1364 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1365 					     struct vm_area_struct **pprev);
1366 
1367 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1368    NULL if none.  Assume start_addr < end_addr. */
1369 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1370 {
1371 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1372 
1373 	if (vma && end_addr <= vma->vm_start)
1374 		vma = NULL;
1375 	return vma;
1376 }
1377 
1378 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1379 {
1380 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1381 }
1382 
1383 #ifdef CONFIG_MMU
1384 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1385 #else
1386 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1387 {
1388 	return __pgprot(0);
1389 }
1390 #endif
1391 
1392 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1393 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1394 			unsigned long pfn, unsigned long size, pgprot_t);
1395 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1396 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1397 			unsigned long pfn);
1398 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1399 			unsigned long pfn);
1400 
1401 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1402 			unsigned int foll_flags);
1403 #define FOLL_WRITE	0x01	/* check pte is writable */
1404 #define FOLL_TOUCH	0x02	/* mark page accessed */
1405 #define FOLL_GET	0x04	/* do get_page on page */
1406 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
1407 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
1408 
1409 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1410 			void *data);
1411 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1412 			       unsigned long size, pte_fn_t fn, void *data);
1413 
1414 #ifdef CONFIG_PROC_FS
1415 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1416 #else
1417 static inline void vm_stat_account(struct mm_struct *mm,
1418 			unsigned long flags, struct file *file, long pages)
1419 {
1420 }
1421 #endif /* CONFIG_PROC_FS */
1422 
1423 #ifdef CONFIG_DEBUG_PAGEALLOC
1424 extern int debug_pagealloc_enabled;
1425 
1426 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1427 
1428 static inline void enable_debug_pagealloc(void)
1429 {
1430 	debug_pagealloc_enabled = 1;
1431 }
1432 #ifdef CONFIG_HIBERNATION
1433 extern bool kernel_page_present(struct page *page);
1434 #endif /* CONFIG_HIBERNATION */
1435 #else
1436 static inline void
1437 kernel_map_pages(struct page *page, int numpages, int enable) {}
1438 static inline void enable_debug_pagealloc(void)
1439 {
1440 }
1441 #ifdef CONFIG_HIBERNATION
1442 static inline bool kernel_page_present(struct page *page) { return true; }
1443 #endif /* CONFIG_HIBERNATION */
1444 #endif
1445 
1446 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1447 #ifdef	__HAVE_ARCH_GATE_AREA
1448 int in_gate_area_no_task(unsigned long addr);
1449 int in_gate_area(struct task_struct *task, unsigned long addr);
1450 #else
1451 int in_gate_area_no_task(unsigned long addr);
1452 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1453 #endif	/* __HAVE_ARCH_GATE_AREA */
1454 
1455 int drop_caches_sysctl_handler(struct ctl_table *, int,
1456 					void __user *, size_t *, loff_t *);
1457 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1458 			unsigned long lru_pages);
1459 
1460 #ifndef CONFIG_MMU
1461 #define randomize_va_space 0
1462 #else
1463 extern int randomize_va_space;
1464 #endif
1465 
1466 const char * arch_vma_name(struct vm_area_struct *vma);
1467 void print_vma_addr(char *prefix, unsigned long rip);
1468 
1469 void sparse_mem_maps_populate_node(struct page **map_map,
1470 				   unsigned long pnum_begin,
1471 				   unsigned long pnum_end,
1472 				   unsigned long map_count,
1473 				   int nodeid);
1474 
1475 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1476 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1477 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1478 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1479 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1480 void *vmemmap_alloc_block(unsigned long size, int node);
1481 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1482 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1483 int vmemmap_populate_basepages(struct page *start_page,
1484 						unsigned long pages, int node);
1485 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1486 void vmemmap_populate_print_last(void);
1487 
1488 
1489 enum mf_flags {
1490 	MF_COUNT_INCREASED = 1 << 0,
1491 };
1492 extern void memory_failure(unsigned long pfn, int trapno);
1493 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1494 extern int unpoison_memory(unsigned long pfn);
1495 extern int sysctl_memory_failure_early_kill;
1496 extern int sysctl_memory_failure_recovery;
1497 extern void shake_page(struct page *p, int access);
1498 extern atomic_long_t mce_bad_pages;
1499 extern int soft_offline_page(struct page *page, int flags);
1500 #ifdef CONFIG_MEMORY_FAILURE
1501 int is_hwpoison_address(unsigned long addr);
1502 #else
1503 static inline int is_hwpoison_address(unsigned long addr)
1504 {
1505 	return 0;
1506 }
1507 #endif
1508 
1509 extern void dump_page(struct page *page);
1510 
1511 #endif /* __KERNEL__ */
1512 #endif /* _LINUX_MM_H */
1513