xref: /linux-6.15/include/linux/mm.h (revision facb4edc)
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 #define FAULT_FLAG_ALLOW_RETRY	0x08	/* Retry fault if blocking */
148 
149 /*
150  * This interface is used by x86 PAT code to identify a pfn mapping that is
151  * linear over entire vma. This is to optimize PAT code that deals with
152  * marking the physical region with a particular prot. This is not for generic
153  * mm use. Note also that this check will not work if the pfn mapping is
154  * linear for a vma starting at physical address 0. In which case PAT code
155  * falls back to slow path of reserving physical range page by page.
156  */
157 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
158 {
159 	return (vma->vm_flags & VM_PFN_AT_MMAP);
160 }
161 
162 static inline int is_pfn_mapping(struct vm_area_struct *vma)
163 {
164 	return (vma->vm_flags & VM_PFNMAP);
165 }
166 
167 /*
168  * vm_fault is filled by the the pagefault handler and passed to the vma's
169  * ->fault function. The vma's ->fault is responsible for returning a bitmask
170  * of VM_FAULT_xxx flags that give details about how the fault was handled.
171  *
172  * pgoff should be used in favour of virtual_address, if possible. If pgoff
173  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
174  * mapping support.
175  */
176 struct vm_fault {
177 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
178 	pgoff_t pgoff;			/* Logical page offset based on vma */
179 	void __user *virtual_address;	/* Faulting virtual address */
180 
181 	struct page *page;		/* ->fault handlers should return a
182 					 * page here, unless VM_FAULT_NOPAGE
183 					 * is set (which is also implied by
184 					 * VM_FAULT_ERROR).
185 					 */
186 };
187 
188 /*
189  * These are the virtual MM functions - opening of an area, closing and
190  * unmapping it (needed to keep files on disk up-to-date etc), pointer
191  * to the functions called when a no-page or a wp-page exception occurs.
192  */
193 struct vm_operations_struct {
194 	void (*open)(struct vm_area_struct * area);
195 	void (*close)(struct vm_area_struct * area);
196 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
197 
198 	/* notification that a previously read-only page is about to become
199 	 * writable, if an error is returned it will cause a SIGBUS */
200 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
201 
202 	/* called by access_process_vm when get_user_pages() fails, typically
203 	 * for use by special VMAs that can switch between memory and hardware
204 	 */
205 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
206 		      void *buf, int len, int write);
207 #ifdef CONFIG_NUMA
208 	/*
209 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
210 	 * to hold the policy upon return.  Caller should pass NULL @new to
211 	 * remove a policy and fall back to surrounding context--i.e. do not
212 	 * install a MPOL_DEFAULT policy, nor the task or system default
213 	 * mempolicy.
214 	 */
215 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
216 
217 	/*
218 	 * get_policy() op must add reference [mpol_get()] to any policy at
219 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
220 	 * in mm/mempolicy.c will do this automatically.
221 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
222 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
223 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
224 	 * must return NULL--i.e., do not "fallback" to task or system default
225 	 * policy.
226 	 */
227 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
228 					unsigned long addr);
229 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
230 		const nodemask_t *to, unsigned long flags);
231 #endif
232 };
233 
234 struct mmu_gather;
235 struct inode;
236 
237 #define page_private(page)		((page)->private)
238 #define set_page_private(page, v)	((page)->private = (v))
239 
240 /*
241  * FIXME: take this include out, include page-flags.h in
242  * files which need it (119 of them)
243  */
244 #include <linux/page-flags.h>
245 
246 /*
247  * Methods to modify the page usage count.
248  *
249  * What counts for a page usage:
250  * - cache mapping   (page->mapping)
251  * - private data    (page->private)
252  * - page mapped in a task's page tables, each mapping
253  *   is counted separately
254  *
255  * Also, many kernel routines increase the page count before a critical
256  * routine so they can be sure the page doesn't go away from under them.
257  */
258 
259 /*
260  * Drop a ref, return true if the refcount fell to zero (the page has no users)
261  */
262 static inline int put_page_testzero(struct page *page)
263 {
264 	VM_BUG_ON(atomic_read(&page->_count) == 0);
265 	return atomic_dec_and_test(&page->_count);
266 }
267 
268 /*
269  * Try to grab a ref unless the page has a refcount of zero, return false if
270  * that is the case.
271  */
272 static inline int get_page_unless_zero(struct page *page)
273 {
274 	return atomic_inc_not_zero(&page->_count);
275 }
276 
277 extern int page_is_ram(unsigned long pfn);
278 
279 /* Support for virtually mapped pages */
280 struct page *vmalloc_to_page(const void *addr);
281 unsigned long vmalloc_to_pfn(const void *addr);
282 
283 /*
284  * Determine if an address is within the vmalloc range
285  *
286  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
287  * is no special casing required.
288  */
289 static inline int is_vmalloc_addr(const void *x)
290 {
291 #ifdef CONFIG_MMU
292 	unsigned long addr = (unsigned long)x;
293 
294 	return addr >= VMALLOC_START && addr < VMALLOC_END;
295 #else
296 	return 0;
297 #endif
298 }
299 #ifdef CONFIG_MMU
300 extern int is_vmalloc_or_module_addr(const void *x);
301 #else
302 static inline int is_vmalloc_or_module_addr(const void *x)
303 {
304 	return 0;
305 }
306 #endif
307 
308 static inline struct page *compound_head(struct page *page)
309 {
310 	if (unlikely(PageTail(page)))
311 		return page->first_page;
312 	return page;
313 }
314 
315 static inline int page_count(struct page *page)
316 {
317 	return atomic_read(&compound_head(page)->_count);
318 }
319 
320 static inline void get_page(struct page *page)
321 {
322 	page = compound_head(page);
323 	VM_BUG_ON(atomic_read(&page->_count) == 0);
324 	atomic_inc(&page->_count);
325 }
326 
327 static inline struct page *virt_to_head_page(const void *x)
328 {
329 	struct page *page = virt_to_page(x);
330 	return compound_head(page);
331 }
332 
333 /*
334  * Setup the page count before being freed into the page allocator for
335  * the first time (boot or memory hotplug)
336  */
337 static inline void init_page_count(struct page *page)
338 {
339 	atomic_set(&page->_count, 1);
340 }
341 
342 void put_page(struct page *page);
343 void put_pages_list(struct list_head *pages);
344 
345 void split_page(struct page *page, unsigned int order);
346 int split_free_page(struct page *page);
347 
348 /*
349  * Compound pages have a destructor function.  Provide a
350  * prototype for that function and accessor functions.
351  * These are _only_ valid on the head of a PG_compound page.
352  */
353 typedef void compound_page_dtor(struct page *);
354 
355 static inline void set_compound_page_dtor(struct page *page,
356 						compound_page_dtor *dtor)
357 {
358 	page[1].lru.next = (void *)dtor;
359 }
360 
361 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
362 {
363 	return (compound_page_dtor *)page[1].lru.next;
364 }
365 
366 static inline int compound_order(struct page *page)
367 {
368 	if (!PageHead(page))
369 		return 0;
370 	return (unsigned long)page[1].lru.prev;
371 }
372 
373 static inline void set_compound_order(struct page *page, unsigned long order)
374 {
375 	page[1].lru.prev = (void *)order;
376 }
377 
378 /*
379  * Multiple processes may "see" the same page. E.g. for untouched
380  * mappings of /dev/null, all processes see the same page full of
381  * zeroes, and text pages of executables and shared libraries have
382  * only one copy in memory, at most, normally.
383  *
384  * For the non-reserved pages, page_count(page) denotes a reference count.
385  *   page_count() == 0 means the page is free. page->lru is then used for
386  *   freelist management in the buddy allocator.
387  *   page_count() > 0  means the page has been allocated.
388  *
389  * Pages are allocated by the slab allocator in order to provide memory
390  * to kmalloc and kmem_cache_alloc. In this case, the management of the
391  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
392  * unless a particular usage is carefully commented. (the responsibility of
393  * freeing the kmalloc memory is the caller's, of course).
394  *
395  * A page may be used by anyone else who does a __get_free_page().
396  * In this case, page_count still tracks the references, and should only
397  * be used through the normal accessor functions. The top bits of page->flags
398  * and page->virtual store page management information, but all other fields
399  * are unused and could be used privately, carefully. The management of this
400  * page is the responsibility of the one who allocated it, and those who have
401  * subsequently been given references to it.
402  *
403  * The other pages (we may call them "pagecache pages") are completely
404  * managed by the Linux memory manager: I/O, buffers, swapping etc.
405  * The following discussion applies only to them.
406  *
407  * A pagecache page contains an opaque `private' member, which belongs to the
408  * page's address_space. Usually, this is the address of a circular list of
409  * the page's disk buffers. PG_private must be set to tell the VM to call
410  * into the filesystem to release these pages.
411  *
412  * A page may belong to an inode's memory mapping. In this case, page->mapping
413  * is the pointer to the inode, and page->index is the file offset of the page,
414  * in units of PAGE_CACHE_SIZE.
415  *
416  * If pagecache pages are not associated with an inode, they are said to be
417  * anonymous pages. These may become associated with the swapcache, and in that
418  * case PG_swapcache is set, and page->private is an offset into the swapcache.
419  *
420  * In either case (swapcache or inode backed), the pagecache itself holds one
421  * reference to the page. Setting PG_private should also increment the
422  * refcount. The each user mapping also has a reference to the page.
423  *
424  * The pagecache pages are stored in a per-mapping radix tree, which is
425  * rooted at mapping->page_tree, and indexed by offset.
426  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
427  * lists, we instead now tag pages as dirty/writeback in the radix tree.
428  *
429  * All pagecache pages may be subject to I/O:
430  * - inode pages may need to be read from disk,
431  * - inode pages which have been modified and are MAP_SHARED may need
432  *   to be written back to the inode on disk,
433  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
434  *   modified may need to be swapped out to swap space and (later) to be read
435  *   back into memory.
436  */
437 
438 /*
439  * The zone field is never updated after free_area_init_core()
440  * sets it, so none of the operations on it need to be atomic.
441  */
442 
443 
444 /*
445  * page->flags layout:
446  *
447  * There are three possibilities for how page->flags get
448  * laid out.  The first is for the normal case, without
449  * sparsemem.  The second is for sparsemem when there is
450  * plenty of space for node and section.  The last is when
451  * we have run out of space and have to fall back to an
452  * alternate (slower) way of determining the node.
453  *
454  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
455  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
456  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
457  */
458 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
459 #define SECTIONS_WIDTH		SECTIONS_SHIFT
460 #else
461 #define SECTIONS_WIDTH		0
462 #endif
463 
464 #define ZONES_WIDTH		ZONES_SHIFT
465 
466 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
467 #define NODES_WIDTH		NODES_SHIFT
468 #else
469 #ifdef CONFIG_SPARSEMEM_VMEMMAP
470 #error "Vmemmap: No space for nodes field in page flags"
471 #endif
472 #define NODES_WIDTH		0
473 #endif
474 
475 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
476 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
477 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
478 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
479 
480 /*
481  * We are going to use the flags for the page to node mapping if its in
482  * there.  This includes the case where there is no node, so it is implicit.
483  */
484 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
485 #define NODE_NOT_IN_PAGE_FLAGS
486 #endif
487 
488 #ifndef PFN_SECTION_SHIFT
489 #define PFN_SECTION_SHIFT 0
490 #endif
491 
492 /*
493  * Define the bit shifts to access each section.  For non-existant
494  * sections we define the shift as 0; that plus a 0 mask ensures
495  * the compiler will optimise away reference to them.
496  */
497 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
498 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
499 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
500 
501 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
502 #ifdef NODE_NOT_IN_PAGE_FLAGS
503 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
504 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
505 						SECTIONS_PGOFF : ZONES_PGOFF)
506 #else
507 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
508 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
509 						NODES_PGOFF : ZONES_PGOFF)
510 #endif
511 
512 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
513 
514 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
515 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
516 #endif
517 
518 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
519 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
520 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
521 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
522 
523 static inline enum zone_type page_zonenum(struct page *page)
524 {
525 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
526 }
527 
528 /*
529  * The identification function is only used by the buddy allocator for
530  * determining if two pages could be buddies. We are not really
531  * identifying a zone since we could be using a the section number
532  * id if we have not node id available in page flags.
533  * We guarantee only that it will return the same value for two
534  * combinable pages in a zone.
535  */
536 static inline int page_zone_id(struct page *page)
537 {
538 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
539 }
540 
541 static inline int zone_to_nid(struct zone *zone)
542 {
543 #ifdef CONFIG_NUMA
544 	return zone->node;
545 #else
546 	return 0;
547 #endif
548 }
549 
550 #ifdef NODE_NOT_IN_PAGE_FLAGS
551 extern int page_to_nid(struct page *page);
552 #else
553 static inline int page_to_nid(struct page *page)
554 {
555 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
556 }
557 #endif
558 
559 static inline struct zone *page_zone(struct page *page)
560 {
561 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
562 }
563 
564 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
565 static inline unsigned long page_to_section(struct page *page)
566 {
567 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
568 }
569 #endif
570 
571 static inline void set_page_zone(struct page *page, enum zone_type zone)
572 {
573 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
574 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
575 }
576 
577 static inline void set_page_node(struct page *page, unsigned long node)
578 {
579 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
580 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
581 }
582 
583 static inline void set_page_section(struct page *page, unsigned long section)
584 {
585 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
586 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
587 }
588 
589 static inline void set_page_links(struct page *page, enum zone_type zone,
590 	unsigned long node, unsigned long pfn)
591 {
592 	set_page_zone(page, zone);
593 	set_page_node(page, node);
594 	set_page_section(page, pfn_to_section_nr(pfn));
595 }
596 
597 /*
598  * Some inline functions in vmstat.h depend on page_zone()
599  */
600 #include <linux/vmstat.h>
601 
602 static __always_inline void *lowmem_page_address(struct page *page)
603 {
604 	return __va(PFN_PHYS(page_to_pfn(page)));
605 }
606 
607 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
608 #define HASHED_PAGE_VIRTUAL
609 #endif
610 
611 #if defined(WANT_PAGE_VIRTUAL)
612 #define page_address(page) ((page)->virtual)
613 #define set_page_address(page, address)			\
614 	do {						\
615 		(page)->virtual = (address);		\
616 	} while(0)
617 #define page_address_init()  do { } while(0)
618 #endif
619 
620 #if defined(HASHED_PAGE_VIRTUAL)
621 void *page_address(struct page *page);
622 void set_page_address(struct page *page, void *virtual);
623 void page_address_init(void);
624 #endif
625 
626 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
627 #define page_address(page) lowmem_page_address(page)
628 #define set_page_address(page, address)  do { } while(0)
629 #define page_address_init()  do { } while(0)
630 #endif
631 
632 /*
633  * On an anonymous page mapped into a user virtual memory area,
634  * page->mapping points to its anon_vma, not to a struct address_space;
635  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
636  *
637  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
638  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
639  * and then page->mapping points, not to an anon_vma, but to a private
640  * structure which KSM associates with that merged page.  See ksm.h.
641  *
642  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
643  *
644  * Please note that, confusingly, "page_mapping" refers to the inode
645  * address_space which maps the page from disk; whereas "page_mapped"
646  * refers to user virtual address space into which the page is mapped.
647  */
648 #define PAGE_MAPPING_ANON	1
649 #define PAGE_MAPPING_KSM	2
650 #define PAGE_MAPPING_FLAGS	(PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
651 
652 extern struct address_space swapper_space;
653 static inline struct address_space *page_mapping(struct page *page)
654 {
655 	struct address_space *mapping = page->mapping;
656 
657 	VM_BUG_ON(PageSlab(page));
658 	if (unlikely(PageSwapCache(page)))
659 		mapping = &swapper_space;
660 	else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
661 		mapping = NULL;
662 	return mapping;
663 }
664 
665 /* Neutral page->mapping pointer to address_space or anon_vma or other */
666 static inline void *page_rmapping(struct page *page)
667 {
668 	return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
669 }
670 
671 static inline int PageAnon(struct page *page)
672 {
673 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
674 }
675 
676 /*
677  * Return the pagecache index of the passed page.  Regular pagecache pages
678  * use ->index whereas swapcache pages use ->private
679  */
680 static inline pgoff_t page_index(struct page *page)
681 {
682 	if (unlikely(PageSwapCache(page)))
683 		return page_private(page);
684 	return page->index;
685 }
686 
687 /*
688  * The atomic page->_mapcount, like _count, starts from -1:
689  * so that transitions both from it and to it can be tracked,
690  * using atomic_inc_and_test and atomic_add_negative(-1).
691  */
692 static inline void reset_page_mapcount(struct page *page)
693 {
694 	atomic_set(&(page)->_mapcount, -1);
695 }
696 
697 static inline int page_mapcount(struct page *page)
698 {
699 	return atomic_read(&(page)->_mapcount) + 1;
700 }
701 
702 /*
703  * Return true if this page is mapped into pagetables.
704  */
705 static inline int page_mapped(struct page *page)
706 {
707 	return atomic_read(&(page)->_mapcount) >= 0;
708 }
709 
710 /*
711  * Different kinds of faults, as returned by handle_mm_fault().
712  * Used to decide whether a process gets delivered SIGBUS or
713  * just gets major/minor fault counters bumped up.
714  */
715 
716 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
717 
718 #define VM_FAULT_OOM	0x0001
719 #define VM_FAULT_SIGBUS	0x0002
720 #define VM_FAULT_MAJOR	0x0004
721 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
722 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
723 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
724 
725 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
726 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
727 #define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
728 
729 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
730 
731 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
732 			 VM_FAULT_HWPOISON_LARGE)
733 
734 /* Encode hstate index for a hwpoisoned large page */
735 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
736 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
737 
738 /*
739  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
740  */
741 extern void pagefault_out_of_memory(void);
742 
743 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
744 
745 extern void show_free_areas(void);
746 
747 int shmem_lock(struct file *file, int lock, struct user_struct *user);
748 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
749 int shmem_zero_setup(struct vm_area_struct *);
750 
751 #ifndef CONFIG_MMU
752 extern unsigned long shmem_get_unmapped_area(struct file *file,
753 					     unsigned long addr,
754 					     unsigned long len,
755 					     unsigned long pgoff,
756 					     unsigned long flags);
757 #endif
758 
759 extern int can_do_mlock(void);
760 extern int user_shm_lock(size_t, struct user_struct *);
761 extern void user_shm_unlock(size_t, struct user_struct *);
762 
763 /*
764  * Parameter block passed down to zap_pte_range in exceptional cases.
765  */
766 struct zap_details {
767 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
768 	struct address_space *check_mapping;	/* Check page->mapping if set */
769 	pgoff_t	first_index;			/* Lowest page->index to unmap */
770 	pgoff_t last_index;			/* Highest page->index to unmap */
771 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
772 	unsigned long truncate_count;		/* Compare vm_truncate_count */
773 };
774 
775 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
776 		pte_t pte);
777 
778 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
779 		unsigned long size);
780 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
781 		unsigned long size, struct zap_details *);
782 unsigned long unmap_vmas(struct mmu_gather **tlb,
783 		struct vm_area_struct *start_vma, unsigned long start_addr,
784 		unsigned long end_addr, unsigned long *nr_accounted,
785 		struct zap_details *);
786 
787 /**
788  * mm_walk - callbacks for walk_page_range
789  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
790  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
791  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
792  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
793  * @pte_hole: if set, called for each hole at all levels
794  * @hugetlb_entry: if set, called for each hugetlb entry
795  *
796  * (see walk_page_range for more details)
797  */
798 struct mm_walk {
799 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
800 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
801 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
802 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
803 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
804 	int (*hugetlb_entry)(pte_t *, unsigned long,
805 			     unsigned long, unsigned long, struct mm_walk *);
806 	struct mm_struct *mm;
807 	void *private;
808 };
809 
810 int walk_page_range(unsigned long addr, unsigned long end,
811 		struct mm_walk *walk);
812 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
813 		unsigned long end, unsigned long floor, unsigned long ceiling);
814 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
815 			struct vm_area_struct *vma);
816 void unmap_mapping_range(struct address_space *mapping,
817 		loff_t const holebegin, loff_t const holelen, int even_cows);
818 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
819 	unsigned long *pfn);
820 int follow_phys(struct vm_area_struct *vma, unsigned long address,
821 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
822 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
823 			void *buf, int len, int write);
824 
825 static inline void unmap_shared_mapping_range(struct address_space *mapping,
826 		loff_t const holebegin, loff_t const holelen)
827 {
828 	unmap_mapping_range(mapping, holebegin, holelen, 0);
829 }
830 
831 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
832 extern void truncate_setsize(struct inode *inode, loff_t newsize);
833 extern int vmtruncate(struct inode *inode, loff_t offset);
834 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
835 
836 int truncate_inode_page(struct address_space *mapping, struct page *page);
837 int generic_error_remove_page(struct address_space *mapping, struct page *page);
838 
839 int invalidate_inode_page(struct page *page);
840 
841 #ifdef CONFIG_MMU
842 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
843 			unsigned long address, unsigned int flags);
844 #else
845 static inline int handle_mm_fault(struct mm_struct *mm,
846 			struct vm_area_struct *vma, unsigned long address,
847 			unsigned int flags)
848 {
849 	/* should never happen if there's no MMU */
850 	BUG();
851 	return VM_FAULT_SIGBUS;
852 }
853 #endif
854 
855 extern int make_pages_present(unsigned long addr, unsigned long end);
856 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
857 
858 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
859 			unsigned long start, int nr_pages, int write, int force,
860 			struct page **pages, struct vm_area_struct **vmas);
861 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
862 			struct page **pages);
863 struct page *get_dump_page(unsigned long addr);
864 
865 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
866 extern void do_invalidatepage(struct page *page, unsigned long offset);
867 
868 int __set_page_dirty_nobuffers(struct page *page);
869 int __set_page_dirty_no_writeback(struct page *page);
870 int redirty_page_for_writepage(struct writeback_control *wbc,
871 				struct page *page);
872 void account_page_dirtied(struct page *page, struct address_space *mapping);
873 void account_page_writeback(struct page *page);
874 int set_page_dirty(struct page *page);
875 int set_page_dirty_lock(struct page *page);
876 int clear_page_dirty_for_io(struct page *page);
877 
878 /* Is the vma a continuation of the stack vma above it? */
879 static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr)
880 {
881 	return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
882 }
883 
884 extern unsigned long move_page_tables(struct vm_area_struct *vma,
885 		unsigned long old_addr, struct vm_area_struct *new_vma,
886 		unsigned long new_addr, unsigned long len);
887 extern unsigned long do_mremap(unsigned long addr,
888 			       unsigned long old_len, unsigned long new_len,
889 			       unsigned long flags, unsigned long new_addr);
890 extern int mprotect_fixup(struct vm_area_struct *vma,
891 			  struct vm_area_struct **pprev, unsigned long start,
892 			  unsigned long end, unsigned long newflags);
893 
894 /*
895  * doesn't attempt to fault and will return short.
896  */
897 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
898 			  struct page **pages);
899 /*
900  * per-process(per-mm_struct) statistics.
901  */
902 #if defined(SPLIT_RSS_COUNTING)
903 /*
904  * The mm counters are not protected by its page_table_lock,
905  * so must be incremented atomically.
906  */
907 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
908 {
909 	atomic_long_set(&mm->rss_stat.count[member], value);
910 }
911 
912 unsigned long get_mm_counter(struct mm_struct *mm, int member);
913 
914 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
915 {
916 	atomic_long_add(value, &mm->rss_stat.count[member]);
917 }
918 
919 static inline void inc_mm_counter(struct mm_struct *mm, int member)
920 {
921 	atomic_long_inc(&mm->rss_stat.count[member]);
922 }
923 
924 static inline void dec_mm_counter(struct mm_struct *mm, int member)
925 {
926 	atomic_long_dec(&mm->rss_stat.count[member]);
927 }
928 
929 #else  /* !USE_SPLIT_PTLOCKS */
930 /*
931  * The mm counters are protected by its page_table_lock,
932  * so can be incremented directly.
933  */
934 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
935 {
936 	mm->rss_stat.count[member] = value;
937 }
938 
939 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
940 {
941 	return mm->rss_stat.count[member];
942 }
943 
944 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
945 {
946 	mm->rss_stat.count[member] += value;
947 }
948 
949 static inline void inc_mm_counter(struct mm_struct *mm, int member)
950 {
951 	mm->rss_stat.count[member]++;
952 }
953 
954 static inline void dec_mm_counter(struct mm_struct *mm, int member)
955 {
956 	mm->rss_stat.count[member]--;
957 }
958 
959 #endif /* !USE_SPLIT_PTLOCKS */
960 
961 static inline unsigned long get_mm_rss(struct mm_struct *mm)
962 {
963 	return get_mm_counter(mm, MM_FILEPAGES) +
964 		get_mm_counter(mm, MM_ANONPAGES);
965 }
966 
967 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
968 {
969 	return max(mm->hiwater_rss, get_mm_rss(mm));
970 }
971 
972 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
973 {
974 	return max(mm->hiwater_vm, mm->total_vm);
975 }
976 
977 static inline void update_hiwater_rss(struct mm_struct *mm)
978 {
979 	unsigned long _rss = get_mm_rss(mm);
980 
981 	if ((mm)->hiwater_rss < _rss)
982 		(mm)->hiwater_rss = _rss;
983 }
984 
985 static inline void update_hiwater_vm(struct mm_struct *mm)
986 {
987 	if (mm->hiwater_vm < mm->total_vm)
988 		mm->hiwater_vm = mm->total_vm;
989 }
990 
991 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
992 					 struct mm_struct *mm)
993 {
994 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
995 
996 	if (*maxrss < hiwater_rss)
997 		*maxrss = hiwater_rss;
998 }
999 
1000 #if defined(SPLIT_RSS_COUNTING)
1001 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
1002 #else
1003 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1004 {
1005 }
1006 #endif
1007 
1008 /*
1009  * A callback you can register to apply pressure to ageable caches.
1010  *
1011  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
1012  * look through the least-recently-used 'nr_to_scan' entries and
1013  * attempt to free them up.  It should return the number of objects
1014  * which remain in the cache.  If it returns -1, it means it cannot do
1015  * any scanning at this time (eg. there is a risk of deadlock).
1016  *
1017  * The 'gfpmask' refers to the allocation we are currently trying to
1018  * fulfil.
1019  *
1020  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1021  * querying the cache size, so a fastpath for that case is appropriate.
1022  */
1023 struct shrinker {
1024 	int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask);
1025 	int seeks;	/* seeks to recreate an obj */
1026 
1027 	/* These are for internal use */
1028 	struct list_head list;
1029 	long nr;	/* objs pending delete */
1030 };
1031 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1032 extern void register_shrinker(struct shrinker *);
1033 extern void unregister_shrinker(struct shrinker *);
1034 
1035 int vma_wants_writenotify(struct vm_area_struct *vma);
1036 
1037 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1038 			       spinlock_t **ptl);
1039 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1040 				    spinlock_t **ptl)
1041 {
1042 	pte_t *ptep;
1043 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1044 	return ptep;
1045 }
1046 
1047 #ifdef __PAGETABLE_PUD_FOLDED
1048 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1049 						unsigned long address)
1050 {
1051 	return 0;
1052 }
1053 #else
1054 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1055 #endif
1056 
1057 #ifdef __PAGETABLE_PMD_FOLDED
1058 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1059 						unsigned long address)
1060 {
1061 	return 0;
1062 }
1063 #else
1064 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1065 #endif
1066 
1067 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1068 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1069 
1070 /*
1071  * The following ifdef needed to get the 4level-fixup.h header to work.
1072  * Remove it when 4level-fixup.h has been removed.
1073  */
1074 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1075 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1076 {
1077 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1078 		NULL: pud_offset(pgd, address);
1079 }
1080 
1081 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1082 {
1083 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1084 		NULL: pmd_offset(pud, address);
1085 }
1086 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1087 
1088 #if USE_SPLIT_PTLOCKS
1089 /*
1090  * We tuck a spinlock to guard each pagetable page into its struct page,
1091  * at page->private, with BUILD_BUG_ON to make sure that this will not
1092  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1093  * When freeing, reset page->mapping so free_pages_check won't complain.
1094  */
1095 #define __pte_lockptr(page)	&((page)->ptl)
1096 #define pte_lock_init(_page)	do {					\
1097 	spin_lock_init(__pte_lockptr(_page));				\
1098 } while (0)
1099 #define pte_lock_deinit(page)	((page)->mapping = NULL)
1100 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1101 #else	/* !USE_SPLIT_PTLOCKS */
1102 /*
1103  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1104  */
1105 #define pte_lock_init(page)	do {} while (0)
1106 #define pte_lock_deinit(page)	do {} while (0)
1107 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
1108 #endif /* USE_SPLIT_PTLOCKS */
1109 
1110 static inline void pgtable_page_ctor(struct page *page)
1111 {
1112 	pte_lock_init(page);
1113 	inc_zone_page_state(page, NR_PAGETABLE);
1114 }
1115 
1116 static inline void pgtable_page_dtor(struct page *page)
1117 {
1118 	pte_lock_deinit(page);
1119 	dec_zone_page_state(page, NR_PAGETABLE);
1120 }
1121 
1122 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1123 ({							\
1124 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1125 	pte_t *__pte = pte_offset_map(pmd, address);	\
1126 	*(ptlp) = __ptl;				\
1127 	spin_lock(__ptl);				\
1128 	__pte;						\
1129 })
1130 
1131 #define pte_unmap_unlock(pte, ptl)	do {		\
1132 	spin_unlock(ptl);				\
1133 	pte_unmap(pte);					\
1134 } while (0)
1135 
1136 #define pte_alloc_map(mm, pmd, address)			\
1137 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1138 		NULL: pte_offset_map(pmd, address))
1139 
1140 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1141 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1142 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1143 
1144 #define pte_alloc_kernel(pmd, address)			\
1145 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1146 		NULL: pte_offset_kernel(pmd, address))
1147 
1148 extern void free_area_init(unsigned long * zones_size);
1149 extern void free_area_init_node(int nid, unsigned long * zones_size,
1150 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1151 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1152 /*
1153  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1154  * zones, allocate the backing mem_map and account for memory holes in a more
1155  * architecture independent manner. This is a substitute for creating the
1156  * zone_sizes[] and zholes_size[] arrays and passing them to
1157  * free_area_init_node()
1158  *
1159  * An architecture is expected to register range of page frames backed by
1160  * physical memory with add_active_range() before calling
1161  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1162  * usage, an architecture is expected to do something like
1163  *
1164  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1165  * 							 max_highmem_pfn};
1166  * for_each_valid_physical_page_range()
1167  * 	add_active_range(node_id, start_pfn, end_pfn)
1168  * free_area_init_nodes(max_zone_pfns);
1169  *
1170  * If the architecture guarantees that there are no holes in the ranges
1171  * registered with add_active_range(), free_bootmem_active_regions()
1172  * will call free_bootmem_node() for each registered physical page range.
1173  * Similarly sparse_memory_present_with_active_regions() calls
1174  * memory_present() for each range when SPARSEMEM is enabled.
1175  *
1176  * See mm/page_alloc.c for more information on each function exposed by
1177  * CONFIG_ARCH_POPULATES_NODE_MAP
1178  */
1179 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1180 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1181 					unsigned long end_pfn);
1182 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1183 					unsigned long end_pfn);
1184 extern void remove_all_active_ranges(void);
1185 void sort_node_map(void);
1186 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1187 						unsigned long end_pfn);
1188 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1189 						unsigned long end_pfn);
1190 extern void get_pfn_range_for_nid(unsigned int nid,
1191 			unsigned long *start_pfn, unsigned long *end_pfn);
1192 extern unsigned long find_min_pfn_with_active_regions(void);
1193 extern void free_bootmem_with_active_regions(int nid,
1194 						unsigned long max_low_pfn);
1195 int add_from_early_node_map(struct range *range, int az,
1196 				   int nr_range, int nid);
1197 u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1198 					u64 goal, u64 limit);
1199 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
1200 				 u64 goal, u64 limit);
1201 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1202 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1203 extern void sparse_memory_present_with_active_regions(int nid);
1204 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1205 
1206 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1207     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1208 static inline int __early_pfn_to_nid(unsigned long pfn)
1209 {
1210 	return 0;
1211 }
1212 #else
1213 /* please see mm/page_alloc.c */
1214 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1215 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1216 /* there is a per-arch backend function. */
1217 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1218 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1219 #endif
1220 
1221 extern void set_dma_reserve(unsigned long new_dma_reserve);
1222 extern void memmap_init_zone(unsigned long, int, unsigned long,
1223 				unsigned long, enum memmap_context);
1224 extern void setup_per_zone_wmarks(void);
1225 extern void calculate_zone_inactive_ratio(struct zone *zone);
1226 extern void mem_init(void);
1227 extern void __init mmap_init(void);
1228 extern void show_mem(void);
1229 extern void si_meminfo(struct sysinfo * val);
1230 extern void si_meminfo_node(struct sysinfo *val, int nid);
1231 extern int after_bootmem;
1232 
1233 extern void setup_per_cpu_pageset(void);
1234 
1235 extern void zone_pcp_update(struct zone *zone);
1236 
1237 /* nommu.c */
1238 extern atomic_long_t mmap_pages_allocated;
1239 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1240 
1241 /* prio_tree.c */
1242 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1243 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1244 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1245 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1246 	struct prio_tree_iter *iter);
1247 
1248 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1249 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1250 		(vma = vma_prio_tree_next(vma, iter)); )
1251 
1252 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1253 					struct list_head *list)
1254 {
1255 	vma->shared.vm_set.parent = NULL;
1256 	list_add_tail(&vma->shared.vm_set.list, list);
1257 }
1258 
1259 /* mmap.c */
1260 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1261 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1262 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1263 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1264 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1265 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1266 	struct mempolicy *);
1267 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1268 extern int split_vma(struct mm_struct *,
1269 	struct vm_area_struct *, unsigned long addr, int new_below);
1270 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1271 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1272 	struct rb_node **, struct rb_node *);
1273 extern void unlink_file_vma(struct vm_area_struct *);
1274 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1275 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1276 extern void exit_mmap(struct mm_struct *);
1277 
1278 extern int mm_take_all_locks(struct mm_struct *mm);
1279 extern void mm_drop_all_locks(struct mm_struct *mm);
1280 
1281 #ifdef CONFIG_PROC_FS
1282 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1283 extern void added_exe_file_vma(struct mm_struct *mm);
1284 extern void removed_exe_file_vma(struct mm_struct *mm);
1285 #else
1286 static inline void added_exe_file_vma(struct mm_struct *mm)
1287 {}
1288 
1289 static inline void removed_exe_file_vma(struct mm_struct *mm)
1290 {}
1291 #endif /* CONFIG_PROC_FS */
1292 
1293 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1294 extern int install_special_mapping(struct mm_struct *mm,
1295 				   unsigned long addr, unsigned long len,
1296 				   unsigned long flags, struct page **pages);
1297 
1298 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1299 
1300 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1301 	unsigned long len, unsigned long prot,
1302 	unsigned long flag, unsigned long pgoff);
1303 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1304 	unsigned long len, unsigned long flags,
1305 	unsigned int vm_flags, unsigned long pgoff);
1306 
1307 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1308 	unsigned long len, unsigned long prot,
1309 	unsigned long flag, unsigned long offset)
1310 {
1311 	unsigned long ret = -EINVAL;
1312 	if ((offset + PAGE_ALIGN(len)) < offset)
1313 		goto out;
1314 	if (!(offset & ~PAGE_MASK))
1315 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1316 out:
1317 	return ret;
1318 }
1319 
1320 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1321 
1322 extern unsigned long do_brk(unsigned long, unsigned long);
1323 
1324 /* filemap.c */
1325 extern unsigned long page_unuse(struct page *);
1326 extern void truncate_inode_pages(struct address_space *, loff_t);
1327 extern void truncate_inode_pages_range(struct address_space *,
1328 				       loff_t lstart, loff_t lend);
1329 
1330 /* generic vm_area_ops exported for stackable file systems */
1331 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1332 
1333 /* mm/page-writeback.c */
1334 int write_one_page(struct page *page, int wait);
1335 void task_dirty_inc(struct task_struct *tsk);
1336 
1337 /* readahead.c */
1338 #define VM_MAX_READAHEAD	128	/* kbytes */
1339 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1340 
1341 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1342 			pgoff_t offset, unsigned long nr_to_read);
1343 
1344 void page_cache_sync_readahead(struct address_space *mapping,
1345 			       struct file_ra_state *ra,
1346 			       struct file *filp,
1347 			       pgoff_t offset,
1348 			       unsigned long size);
1349 
1350 void page_cache_async_readahead(struct address_space *mapping,
1351 				struct file_ra_state *ra,
1352 				struct file *filp,
1353 				struct page *pg,
1354 				pgoff_t offset,
1355 				unsigned long size);
1356 
1357 unsigned long max_sane_readahead(unsigned long nr);
1358 unsigned long ra_submit(struct file_ra_state *ra,
1359 			struct address_space *mapping,
1360 			struct file *filp);
1361 
1362 /* Do stack extension */
1363 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1364 #if VM_GROWSUP
1365 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1366 #else
1367   #define expand_upwards(vma, address) do { } while (0)
1368 #endif
1369 extern int expand_stack_downwards(struct vm_area_struct *vma,
1370 				  unsigned long address);
1371 
1372 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1373 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1374 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1375 					     struct vm_area_struct **pprev);
1376 
1377 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1378    NULL if none.  Assume start_addr < end_addr. */
1379 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1380 {
1381 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1382 
1383 	if (vma && end_addr <= vma->vm_start)
1384 		vma = NULL;
1385 	return vma;
1386 }
1387 
1388 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1389 {
1390 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1391 }
1392 
1393 #ifdef CONFIG_MMU
1394 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1395 #else
1396 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1397 {
1398 	return __pgprot(0);
1399 }
1400 #endif
1401 
1402 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1403 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1404 			unsigned long pfn, unsigned long size, pgprot_t);
1405 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1406 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1407 			unsigned long pfn);
1408 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1409 			unsigned long pfn);
1410 
1411 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1412 			unsigned int foll_flags);
1413 #define FOLL_WRITE	0x01	/* check pte is writable */
1414 #define FOLL_TOUCH	0x02	/* mark page accessed */
1415 #define FOLL_GET	0x04	/* do get_page on page */
1416 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
1417 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
1418 
1419 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1420 			void *data);
1421 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1422 			       unsigned long size, pte_fn_t fn, void *data);
1423 
1424 #ifdef CONFIG_PROC_FS
1425 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1426 #else
1427 static inline void vm_stat_account(struct mm_struct *mm,
1428 			unsigned long flags, struct file *file, long pages)
1429 {
1430 }
1431 #endif /* CONFIG_PROC_FS */
1432 
1433 #ifdef CONFIG_DEBUG_PAGEALLOC
1434 extern int debug_pagealloc_enabled;
1435 
1436 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1437 
1438 static inline void enable_debug_pagealloc(void)
1439 {
1440 	debug_pagealloc_enabled = 1;
1441 }
1442 #ifdef CONFIG_HIBERNATION
1443 extern bool kernel_page_present(struct page *page);
1444 #endif /* CONFIG_HIBERNATION */
1445 #else
1446 static inline void
1447 kernel_map_pages(struct page *page, int numpages, int enable) {}
1448 static inline void enable_debug_pagealloc(void)
1449 {
1450 }
1451 #ifdef CONFIG_HIBERNATION
1452 static inline bool kernel_page_present(struct page *page) { return true; }
1453 #endif /* CONFIG_HIBERNATION */
1454 #endif
1455 
1456 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1457 #ifdef	__HAVE_ARCH_GATE_AREA
1458 int in_gate_area_no_task(unsigned long addr);
1459 int in_gate_area(struct task_struct *task, unsigned long addr);
1460 #else
1461 int in_gate_area_no_task(unsigned long addr);
1462 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1463 #endif	/* __HAVE_ARCH_GATE_AREA */
1464 
1465 int drop_caches_sysctl_handler(struct ctl_table *, int,
1466 					void __user *, size_t *, loff_t *);
1467 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1468 			unsigned long lru_pages);
1469 
1470 #ifndef CONFIG_MMU
1471 #define randomize_va_space 0
1472 #else
1473 extern int randomize_va_space;
1474 #endif
1475 
1476 const char * arch_vma_name(struct vm_area_struct *vma);
1477 void print_vma_addr(char *prefix, unsigned long rip);
1478 
1479 void sparse_mem_maps_populate_node(struct page **map_map,
1480 				   unsigned long pnum_begin,
1481 				   unsigned long pnum_end,
1482 				   unsigned long map_count,
1483 				   int nodeid);
1484 
1485 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1486 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1487 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1488 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1489 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1490 void *vmemmap_alloc_block(unsigned long size, int node);
1491 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1492 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1493 int vmemmap_populate_basepages(struct page *start_page,
1494 						unsigned long pages, int node);
1495 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1496 void vmemmap_populate_print_last(void);
1497 
1498 
1499 enum mf_flags {
1500 	MF_COUNT_INCREASED = 1 << 0,
1501 };
1502 extern void memory_failure(unsigned long pfn, int trapno);
1503 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1504 extern int unpoison_memory(unsigned long pfn);
1505 extern int sysctl_memory_failure_early_kill;
1506 extern int sysctl_memory_failure_recovery;
1507 extern void shake_page(struct page *p, int access);
1508 extern atomic_long_t mce_bad_pages;
1509 extern int soft_offline_page(struct page *page, int flags);
1510 #ifdef CONFIG_MEMORY_FAILURE
1511 int is_hwpoison_address(unsigned long addr);
1512 #else
1513 static inline int is_hwpoison_address(unsigned long addr)
1514 {
1515 	return 0;
1516 }
1517 #endif
1518 
1519 extern void dump_page(struct page *page);
1520 
1521 #endif /* __KERNEL__ */
1522 #endif /* _LINUX_MM_H */
1523