xref: /linux-6.15/include/linux/mm.h (revision 2277ab4a)
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 
16 struct mempolicy;
17 struct anon_vma;
18 struct file_ra_state;
19 struct user_struct;
20 struct writeback_control;
21 struct rlimit;
22 
23 #ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
24 extern unsigned long max_mapnr;
25 #endif
26 
27 extern unsigned long num_physpages;
28 extern void * high_memory;
29 extern int page_cluster;
30 
31 #ifdef CONFIG_SYSCTL
32 extern int sysctl_legacy_va_layout;
33 #else
34 #define sysctl_legacy_va_layout 0
35 #endif
36 
37 extern unsigned long mmap_min_addr;
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 #define VM_GROWSUP	0x00000200
82 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
83 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
84 
85 #define VM_EXECUTABLE	0x00001000
86 #define VM_LOCKED	0x00002000
87 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
88 
89 					/* Used by sys_madvise() */
90 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
91 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
92 
93 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
94 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
95 #define VM_RESERVED	0x00080000	/* Count as reserved_vm like IO */
96 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
97 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
98 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
99 #define VM_NONLINEAR	0x00800000	/* Is non-linear (remap_file_pages) */
100 #define VM_MAPPED_COPY	0x01000000	/* T if mapped copy of data (nommu mmap) */
101 #define VM_INSERTPAGE	0x02000000	/* The vma has had "vm_insert_page()" done on it */
102 #define VM_ALWAYSDUMP	0x04000000	/* Always include in core dumps */
103 
104 #define VM_CAN_NONLINEAR 0x08000000	/* Has ->fault & does nonlinear pages */
105 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
106 #define VM_SAO		0x20000000	/* Strong Access Ordering (powerpc) */
107 #define VM_PFN_AT_MMAP	0x40000000	/* PFNMAP vma that is fully mapped at mmap time */
108 
109 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
110 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
111 #endif
112 
113 #ifdef CONFIG_STACK_GROWSUP
114 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
115 #else
116 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
117 #endif
118 
119 #define VM_READHINTMASK			(VM_SEQ_READ | VM_RAND_READ)
120 #define VM_ClearReadHint(v)		(v)->vm_flags &= ~VM_READHINTMASK
121 #define VM_NormalReadHint(v)		(!((v)->vm_flags & VM_READHINTMASK))
122 #define VM_SequentialReadHint(v)	((v)->vm_flags & VM_SEQ_READ)
123 #define VM_RandomReadHint(v)		((v)->vm_flags & VM_RAND_READ)
124 
125 /*
126  * special vmas that are non-mergable, non-mlock()able
127  */
128 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
129 
130 /*
131  * mapping from the currently active vm_flags protection bits (the
132  * low four bits) to a page protection mask..
133  */
134 extern pgprot_t protection_map[16];
135 
136 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
137 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
138 #define FAULT_FLAG_MKWRITE	0x04	/* Fault was mkwrite of existing pte */
139 
140 /*
141  * This interface is used by x86 PAT code to identify a pfn mapping that is
142  * linear over entire vma. This is to optimize PAT code that deals with
143  * marking the physical region with a particular prot. This is not for generic
144  * mm use. Note also that this check will not work if the pfn mapping is
145  * linear for a vma starting at physical address 0. In which case PAT code
146  * falls back to slow path of reserving physical range page by page.
147  */
148 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
149 {
150 	return (vma->vm_flags & VM_PFN_AT_MMAP);
151 }
152 
153 static inline int is_pfn_mapping(struct vm_area_struct *vma)
154 {
155 	return (vma->vm_flags & VM_PFNMAP);
156 }
157 
158 /*
159  * vm_fault is filled by the the pagefault handler and passed to the vma's
160  * ->fault function. The vma's ->fault is responsible for returning a bitmask
161  * of VM_FAULT_xxx flags that give details about how the fault was handled.
162  *
163  * pgoff should be used in favour of virtual_address, if possible. If pgoff
164  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
165  * mapping support.
166  */
167 struct vm_fault {
168 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
169 	pgoff_t pgoff;			/* Logical page offset based on vma */
170 	void __user *virtual_address;	/* Faulting virtual address */
171 
172 	struct page *page;		/* ->fault handlers should return a
173 					 * page here, unless VM_FAULT_NOPAGE
174 					 * is set (which is also implied by
175 					 * VM_FAULT_ERROR).
176 					 */
177 };
178 
179 /*
180  * These are the virtual MM functions - opening of an area, closing and
181  * unmapping it (needed to keep files on disk up-to-date etc), pointer
182  * to the functions called when a no-page or a wp-page exception occurs.
183  */
184 struct vm_operations_struct {
185 	void (*open)(struct vm_area_struct * area);
186 	void (*close)(struct vm_area_struct * area);
187 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
188 
189 	/* notification that a previously read-only page is about to become
190 	 * writable, if an error is returned it will cause a SIGBUS */
191 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
192 
193 	/* called by access_process_vm when get_user_pages() fails, typically
194 	 * for use by special VMAs that can switch between memory and hardware
195 	 */
196 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
197 		      void *buf, int len, int write);
198 #ifdef CONFIG_NUMA
199 	/*
200 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
201 	 * to hold the policy upon return.  Caller should pass NULL @new to
202 	 * remove a policy and fall back to surrounding context--i.e. do not
203 	 * install a MPOL_DEFAULT policy, nor the task or system default
204 	 * mempolicy.
205 	 */
206 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
207 
208 	/*
209 	 * get_policy() op must add reference [mpol_get()] to any policy at
210 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
211 	 * in mm/mempolicy.c will do this automatically.
212 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
213 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
214 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
215 	 * must return NULL--i.e., do not "fallback" to task or system default
216 	 * policy.
217 	 */
218 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
219 					unsigned long addr);
220 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
221 		const nodemask_t *to, unsigned long flags);
222 #endif
223 };
224 
225 struct mmu_gather;
226 struct inode;
227 
228 #define page_private(page)		((page)->private)
229 #define set_page_private(page, v)	((page)->private = (v))
230 
231 /*
232  * FIXME: take this include out, include page-flags.h in
233  * files which need it (119 of them)
234  */
235 #include <linux/page-flags.h>
236 
237 /*
238  * Methods to modify the page usage count.
239  *
240  * What counts for a page usage:
241  * - cache mapping   (page->mapping)
242  * - private data    (page->private)
243  * - page mapped in a task's page tables, each mapping
244  *   is counted separately
245  *
246  * Also, many kernel routines increase the page count before a critical
247  * routine so they can be sure the page doesn't go away from under them.
248  */
249 
250 /*
251  * Drop a ref, return true if the refcount fell to zero (the page has no users)
252  */
253 static inline int put_page_testzero(struct page *page)
254 {
255 	VM_BUG_ON(atomic_read(&page->_count) == 0);
256 	return atomic_dec_and_test(&page->_count);
257 }
258 
259 /*
260  * Try to grab a ref unless the page has a refcount of zero, return false if
261  * that is the case.
262  */
263 static inline int get_page_unless_zero(struct page *page)
264 {
265 	return atomic_inc_not_zero(&page->_count);
266 }
267 
268 /* Support for virtually mapped pages */
269 struct page *vmalloc_to_page(const void *addr);
270 unsigned long vmalloc_to_pfn(const void *addr);
271 
272 /*
273  * Determine if an address is within the vmalloc range
274  *
275  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
276  * is no special casing required.
277  */
278 static inline int is_vmalloc_addr(const void *x)
279 {
280 #ifdef CONFIG_MMU
281 	unsigned long addr = (unsigned long)x;
282 
283 	return addr >= VMALLOC_START && addr < VMALLOC_END;
284 #else
285 	return 0;
286 #endif
287 }
288 
289 static inline struct page *compound_head(struct page *page)
290 {
291 	if (unlikely(PageTail(page)))
292 		return page->first_page;
293 	return page;
294 }
295 
296 static inline int page_count(struct page *page)
297 {
298 	return atomic_read(&compound_head(page)->_count);
299 }
300 
301 static inline void get_page(struct page *page)
302 {
303 	page = compound_head(page);
304 	VM_BUG_ON(atomic_read(&page->_count) == 0);
305 	atomic_inc(&page->_count);
306 }
307 
308 static inline struct page *virt_to_head_page(const void *x)
309 {
310 	struct page *page = virt_to_page(x);
311 	return compound_head(page);
312 }
313 
314 /*
315  * Setup the page count before being freed into the page allocator for
316  * the first time (boot or memory hotplug)
317  */
318 static inline void init_page_count(struct page *page)
319 {
320 	atomic_set(&page->_count, 1);
321 }
322 
323 void put_page(struct page *page);
324 void put_pages_list(struct list_head *pages);
325 
326 void split_page(struct page *page, unsigned int order);
327 
328 /*
329  * Compound pages have a destructor function.  Provide a
330  * prototype for that function and accessor functions.
331  * These are _only_ valid on the head of a PG_compound page.
332  */
333 typedef void compound_page_dtor(struct page *);
334 
335 static inline void set_compound_page_dtor(struct page *page,
336 						compound_page_dtor *dtor)
337 {
338 	page[1].lru.next = (void *)dtor;
339 }
340 
341 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
342 {
343 	return (compound_page_dtor *)page[1].lru.next;
344 }
345 
346 static inline int compound_order(struct page *page)
347 {
348 	if (!PageHead(page))
349 		return 0;
350 	return (unsigned long)page[1].lru.prev;
351 }
352 
353 static inline void set_compound_order(struct page *page, unsigned long order)
354 {
355 	page[1].lru.prev = (void *)order;
356 }
357 
358 /*
359  * Multiple processes may "see" the same page. E.g. for untouched
360  * mappings of /dev/null, all processes see the same page full of
361  * zeroes, and text pages of executables and shared libraries have
362  * only one copy in memory, at most, normally.
363  *
364  * For the non-reserved pages, page_count(page) denotes a reference count.
365  *   page_count() == 0 means the page is free. page->lru is then used for
366  *   freelist management in the buddy allocator.
367  *   page_count() > 0  means the page has been allocated.
368  *
369  * Pages are allocated by the slab allocator in order to provide memory
370  * to kmalloc and kmem_cache_alloc. In this case, the management of the
371  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
372  * unless a particular usage is carefully commented. (the responsibility of
373  * freeing the kmalloc memory is the caller's, of course).
374  *
375  * A page may be used by anyone else who does a __get_free_page().
376  * In this case, page_count still tracks the references, and should only
377  * be used through the normal accessor functions. The top bits of page->flags
378  * and page->virtual store page management information, but all other fields
379  * are unused and could be used privately, carefully. The management of this
380  * page is the responsibility of the one who allocated it, and those who have
381  * subsequently been given references to it.
382  *
383  * The other pages (we may call them "pagecache pages") are completely
384  * managed by the Linux memory manager: I/O, buffers, swapping etc.
385  * The following discussion applies only to them.
386  *
387  * A pagecache page contains an opaque `private' member, which belongs to the
388  * page's address_space. Usually, this is the address of a circular list of
389  * the page's disk buffers. PG_private must be set to tell the VM to call
390  * into the filesystem to release these pages.
391  *
392  * A page may belong to an inode's memory mapping. In this case, page->mapping
393  * is the pointer to the inode, and page->index is the file offset of the page,
394  * in units of PAGE_CACHE_SIZE.
395  *
396  * If pagecache pages are not associated with an inode, they are said to be
397  * anonymous pages. These may become associated with the swapcache, and in that
398  * case PG_swapcache is set, and page->private is an offset into the swapcache.
399  *
400  * In either case (swapcache or inode backed), the pagecache itself holds one
401  * reference to the page. Setting PG_private should also increment the
402  * refcount. The each user mapping also has a reference to the page.
403  *
404  * The pagecache pages are stored in a per-mapping radix tree, which is
405  * rooted at mapping->page_tree, and indexed by offset.
406  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
407  * lists, we instead now tag pages as dirty/writeback in the radix tree.
408  *
409  * All pagecache pages may be subject to I/O:
410  * - inode pages may need to be read from disk,
411  * - inode pages which have been modified and are MAP_SHARED may need
412  *   to be written back to the inode on disk,
413  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
414  *   modified may need to be swapped out to swap space and (later) to be read
415  *   back into memory.
416  */
417 
418 /*
419  * The zone field is never updated after free_area_init_core()
420  * sets it, so none of the operations on it need to be atomic.
421  */
422 
423 
424 /*
425  * page->flags layout:
426  *
427  * There are three possibilities for how page->flags get
428  * laid out.  The first is for the normal case, without
429  * sparsemem.  The second is for sparsemem when there is
430  * plenty of space for node and section.  The last is when
431  * we have run out of space and have to fall back to an
432  * alternate (slower) way of determining the node.
433  *
434  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
435  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
436  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
437  */
438 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
439 #define SECTIONS_WIDTH		SECTIONS_SHIFT
440 #else
441 #define SECTIONS_WIDTH		0
442 #endif
443 
444 #define ZONES_WIDTH		ZONES_SHIFT
445 
446 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
447 #define NODES_WIDTH		NODES_SHIFT
448 #else
449 #ifdef CONFIG_SPARSEMEM_VMEMMAP
450 #error "Vmemmap: No space for nodes field in page flags"
451 #endif
452 #define NODES_WIDTH		0
453 #endif
454 
455 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
456 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
457 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
458 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
459 
460 /*
461  * We are going to use the flags for the page to node mapping if its in
462  * there.  This includes the case where there is no node, so it is implicit.
463  */
464 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
465 #define NODE_NOT_IN_PAGE_FLAGS
466 #endif
467 
468 #ifndef PFN_SECTION_SHIFT
469 #define PFN_SECTION_SHIFT 0
470 #endif
471 
472 /*
473  * Define the bit shifts to access each section.  For non-existant
474  * sections we define the shift as 0; that plus a 0 mask ensures
475  * the compiler will optimise away reference to them.
476  */
477 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
478 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
479 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
480 
481 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
482 #ifdef NODE_NOT_IN_PAGEFLAGS
483 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
484 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
485 						SECTIONS_PGOFF : ZONES_PGOFF)
486 #else
487 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
488 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
489 						NODES_PGOFF : ZONES_PGOFF)
490 #endif
491 
492 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
493 
494 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
495 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
496 #endif
497 
498 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
499 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
500 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
501 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
502 
503 static inline enum zone_type page_zonenum(struct page *page)
504 {
505 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
506 }
507 
508 /*
509  * The identification function is only used by the buddy allocator for
510  * determining if two pages could be buddies. We are not really
511  * identifying a zone since we could be using a the section number
512  * id if we have not node id available in page flags.
513  * We guarantee only that it will return the same value for two
514  * combinable pages in a zone.
515  */
516 static inline int page_zone_id(struct page *page)
517 {
518 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
519 }
520 
521 static inline int zone_to_nid(struct zone *zone)
522 {
523 #ifdef CONFIG_NUMA
524 	return zone->node;
525 #else
526 	return 0;
527 #endif
528 }
529 
530 #ifdef NODE_NOT_IN_PAGE_FLAGS
531 extern int page_to_nid(struct page *page);
532 #else
533 static inline int page_to_nid(struct page *page)
534 {
535 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
536 }
537 #endif
538 
539 static inline struct zone *page_zone(struct page *page)
540 {
541 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
542 }
543 
544 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
545 static inline unsigned long page_to_section(struct page *page)
546 {
547 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
548 }
549 #endif
550 
551 static inline void set_page_zone(struct page *page, enum zone_type zone)
552 {
553 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
554 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
555 }
556 
557 static inline void set_page_node(struct page *page, unsigned long node)
558 {
559 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
560 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
561 }
562 
563 static inline void set_page_section(struct page *page, unsigned long section)
564 {
565 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
566 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
567 }
568 
569 static inline void set_page_links(struct page *page, enum zone_type zone,
570 	unsigned long node, unsigned long pfn)
571 {
572 	set_page_zone(page, zone);
573 	set_page_node(page, node);
574 	set_page_section(page, pfn_to_section_nr(pfn));
575 }
576 
577 /*
578  * If a hint addr is less than mmap_min_addr change hint to be as
579  * low as possible but still greater than mmap_min_addr
580  */
581 static inline unsigned long round_hint_to_min(unsigned long hint)
582 {
583 	hint &= PAGE_MASK;
584 	if (((void *)hint != NULL) &&
585 	    (hint < mmap_min_addr))
586 		return PAGE_ALIGN(mmap_min_addr);
587 	return hint;
588 }
589 
590 /*
591  * Some inline functions in vmstat.h depend on page_zone()
592  */
593 #include <linux/vmstat.h>
594 
595 static __always_inline void *lowmem_page_address(struct page *page)
596 {
597 	return __va(page_to_pfn(page) << PAGE_SHIFT);
598 }
599 
600 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
601 #define HASHED_PAGE_VIRTUAL
602 #endif
603 
604 #if defined(WANT_PAGE_VIRTUAL)
605 #define page_address(page) ((page)->virtual)
606 #define set_page_address(page, address)			\
607 	do {						\
608 		(page)->virtual = (address);		\
609 	} while(0)
610 #define page_address_init()  do { } while(0)
611 #endif
612 
613 #if defined(HASHED_PAGE_VIRTUAL)
614 void *page_address(struct page *page);
615 void set_page_address(struct page *page, void *virtual);
616 void page_address_init(void);
617 #endif
618 
619 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
620 #define page_address(page) lowmem_page_address(page)
621 #define set_page_address(page, address)  do { } while(0)
622 #define page_address_init()  do { } while(0)
623 #endif
624 
625 /*
626  * On an anonymous page mapped into a user virtual memory area,
627  * page->mapping points to its anon_vma, not to a struct address_space;
628  * with the PAGE_MAPPING_ANON bit set to distinguish it.
629  *
630  * Please note that, confusingly, "page_mapping" refers to the inode
631  * address_space which maps the page from disk; whereas "page_mapped"
632  * refers to user virtual address space into which the page is mapped.
633  */
634 #define PAGE_MAPPING_ANON	1
635 
636 extern struct address_space swapper_space;
637 static inline struct address_space *page_mapping(struct page *page)
638 {
639 	struct address_space *mapping = page->mapping;
640 
641 	VM_BUG_ON(PageSlab(page));
642 #ifdef CONFIG_SWAP
643 	if (unlikely(PageSwapCache(page)))
644 		mapping = &swapper_space;
645 	else
646 #endif
647 	if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
648 		mapping = NULL;
649 	return mapping;
650 }
651 
652 static inline int PageAnon(struct page *page)
653 {
654 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
655 }
656 
657 /*
658  * Return the pagecache index of the passed page.  Regular pagecache pages
659  * use ->index whereas swapcache pages use ->private
660  */
661 static inline pgoff_t page_index(struct page *page)
662 {
663 	if (unlikely(PageSwapCache(page)))
664 		return page_private(page);
665 	return page->index;
666 }
667 
668 /*
669  * The atomic page->_mapcount, like _count, starts from -1:
670  * so that transitions both from it and to it can be tracked,
671  * using atomic_inc_and_test and atomic_add_negative(-1).
672  */
673 static inline void reset_page_mapcount(struct page *page)
674 {
675 	atomic_set(&(page)->_mapcount, -1);
676 }
677 
678 static inline int page_mapcount(struct page *page)
679 {
680 	return atomic_read(&(page)->_mapcount) + 1;
681 }
682 
683 /*
684  * Return true if this page is mapped into pagetables.
685  */
686 static inline int page_mapped(struct page *page)
687 {
688 	return atomic_read(&(page)->_mapcount) >= 0;
689 }
690 
691 /*
692  * Different kinds of faults, as returned by handle_mm_fault().
693  * Used to decide whether a process gets delivered SIGBUS or
694  * just gets major/minor fault counters bumped up.
695  */
696 
697 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
698 
699 #define VM_FAULT_OOM	0x0001
700 #define VM_FAULT_SIGBUS	0x0002
701 #define VM_FAULT_MAJOR	0x0004
702 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
703 
704 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
705 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
706 
707 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS)
708 
709 /*
710  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
711  */
712 extern void pagefault_out_of_memory(void);
713 
714 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
715 
716 extern void show_free_areas(void);
717 
718 #ifdef CONFIG_SHMEM
719 extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
720 #else
721 static inline int shmem_lock(struct file *file, int lock,
722 			    struct user_struct *user)
723 {
724 	return 0;
725 }
726 #endif
727 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
728 
729 int shmem_zero_setup(struct vm_area_struct *);
730 
731 #ifndef CONFIG_MMU
732 extern unsigned long shmem_get_unmapped_area(struct file *file,
733 					     unsigned long addr,
734 					     unsigned long len,
735 					     unsigned long pgoff,
736 					     unsigned long flags);
737 #endif
738 
739 extern int can_do_mlock(void);
740 extern int user_shm_lock(size_t, struct user_struct *);
741 extern void user_shm_unlock(size_t, struct user_struct *);
742 
743 /*
744  * Parameter block passed down to zap_pte_range in exceptional cases.
745  */
746 struct zap_details {
747 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
748 	struct address_space *check_mapping;	/* Check page->mapping if set */
749 	pgoff_t	first_index;			/* Lowest page->index to unmap */
750 	pgoff_t last_index;			/* Highest page->index to unmap */
751 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
752 	unsigned long truncate_count;		/* Compare vm_truncate_count */
753 };
754 
755 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
756 		pte_t pte);
757 
758 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
759 		unsigned long size);
760 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
761 		unsigned long size, struct zap_details *);
762 unsigned long unmap_vmas(struct mmu_gather **tlb,
763 		struct vm_area_struct *start_vma, unsigned long start_addr,
764 		unsigned long end_addr, unsigned long *nr_accounted,
765 		struct zap_details *);
766 
767 /**
768  * mm_walk - callbacks for walk_page_range
769  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
770  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
771  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
772  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
773  * @pte_hole: if set, called for each hole at all levels
774  *
775  * (see walk_page_range for more details)
776  */
777 struct mm_walk {
778 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
779 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
780 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
781 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
782 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
783 	struct mm_struct *mm;
784 	void *private;
785 };
786 
787 int walk_page_range(unsigned long addr, unsigned long end,
788 		struct mm_walk *walk);
789 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
790 		unsigned long end, unsigned long floor, unsigned long ceiling);
791 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
792 			struct vm_area_struct *vma);
793 void unmap_mapping_range(struct address_space *mapping,
794 		loff_t const holebegin, loff_t const holelen, int even_cows);
795 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
796 	unsigned long *pfn);
797 int follow_phys(struct vm_area_struct *vma, unsigned long address,
798 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
799 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
800 			void *buf, int len, int write);
801 
802 static inline void unmap_shared_mapping_range(struct address_space *mapping,
803 		loff_t const holebegin, loff_t const holelen)
804 {
805 	unmap_mapping_range(mapping, holebegin, holelen, 0);
806 }
807 
808 extern int vmtruncate(struct inode * inode, loff_t offset);
809 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
810 
811 #ifdef CONFIG_MMU
812 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
813 			unsigned long address, unsigned int flags);
814 #else
815 static inline int handle_mm_fault(struct mm_struct *mm,
816 			struct vm_area_struct *vma, unsigned long address,
817 			unsigned int flags)
818 {
819 	/* should never happen if there's no MMU */
820 	BUG();
821 	return VM_FAULT_SIGBUS;
822 }
823 #endif
824 
825 extern int make_pages_present(unsigned long addr, unsigned long end);
826 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
827 
828 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
829 			unsigned long start, int nr_pages, int write, int force,
830 			struct page **pages, struct vm_area_struct **vmas);
831 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
832 			struct page **pages);
833 
834 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
835 extern void do_invalidatepage(struct page *page, unsigned long offset);
836 
837 int __set_page_dirty_nobuffers(struct page *page);
838 int __set_page_dirty_no_writeback(struct page *page);
839 int redirty_page_for_writepage(struct writeback_control *wbc,
840 				struct page *page);
841 void account_page_dirtied(struct page *page, struct address_space *mapping);
842 int set_page_dirty(struct page *page);
843 int set_page_dirty_lock(struct page *page);
844 int clear_page_dirty_for_io(struct page *page);
845 
846 extern unsigned long move_page_tables(struct vm_area_struct *vma,
847 		unsigned long old_addr, struct vm_area_struct *new_vma,
848 		unsigned long new_addr, unsigned long len);
849 extern unsigned long do_mremap(unsigned long addr,
850 			       unsigned long old_len, unsigned long new_len,
851 			       unsigned long flags, unsigned long new_addr);
852 extern int mprotect_fixup(struct vm_area_struct *vma,
853 			  struct vm_area_struct **pprev, unsigned long start,
854 			  unsigned long end, unsigned long newflags);
855 
856 /*
857  * doesn't attempt to fault and will return short.
858  */
859 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
860 			  struct page **pages);
861 
862 /*
863  * A callback you can register to apply pressure to ageable caches.
864  *
865  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
866  * look through the least-recently-used 'nr_to_scan' entries and
867  * attempt to free them up.  It should return the number of objects
868  * which remain in the cache.  If it returns -1, it means it cannot do
869  * any scanning at this time (eg. there is a risk of deadlock).
870  *
871  * The 'gfpmask' refers to the allocation we are currently trying to
872  * fulfil.
873  *
874  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
875  * querying the cache size, so a fastpath for that case is appropriate.
876  */
877 struct shrinker {
878 	int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
879 	int seeks;	/* seeks to recreate an obj */
880 
881 	/* These are for internal use */
882 	struct list_head list;
883 	long nr;	/* objs pending delete */
884 };
885 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
886 extern void register_shrinker(struct shrinker *);
887 extern void unregister_shrinker(struct shrinker *);
888 
889 int vma_wants_writenotify(struct vm_area_struct *vma);
890 
891 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
892 
893 #ifdef __PAGETABLE_PUD_FOLDED
894 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
895 						unsigned long address)
896 {
897 	return 0;
898 }
899 #else
900 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
901 #endif
902 
903 #ifdef __PAGETABLE_PMD_FOLDED
904 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
905 						unsigned long address)
906 {
907 	return 0;
908 }
909 #else
910 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
911 #endif
912 
913 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
914 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
915 
916 /*
917  * The following ifdef needed to get the 4level-fixup.h header to work.
918  * Remove it when 4level-fixup.h has been removed.
919  */
920 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
921 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
922 {
923 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
924 		NULL: pud_offset(pgd, address);
925 }
926 
927 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
928 {
929 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
930 		NULL: pmd_offset(pud, address);
931 }
932 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
933 
934 #if USE_SPLIT_PTLOCKS
935 /*
936  * We tuck a spinlock to guard each pagetable page into its struct page,
937  * at page->private, with BUILD_BUG_ON to make sure that this will not
938  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
939  * When freeing, reset page->mapping so free_pages_check won't complain.
940  */
941 #define __pte_lockptr(page)	&((page)->ptl)
942 #define pte_lock_init(_page)	do {					\
943 	spin_lock_init(__pte_lockptr(_page));				\
944 } while (0)
945 #define pte_lock_deinit(page)	((page)->mapping = NULL)
946 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
947 #else	/* !USE_SPLIT_PTLOCKS */
948 /*
949  * We use mm->page_table_lock to guard all pagetable pages of the mm.
950  */
951 #define pte_lock_init(page)	do {} while (0)
952 #define pte_lock_deinit(page)	do {} while (0)
953 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
954 #endif /* USE_SPLIT_PTLOCKS */
955 
956 static inline void pgtable_page_ctor(struct page *page)
957 {
958 	pte_lock_init(page);
959 	inc_zone_page_state(page, NR_PAGETABLE);
960 }
961 
962 static inline void pgtable_page_dtor(struct page *page)
963 {
964 	pte_lock_deinit(page);
965 	dec_zone_page_state(page, NR_PAGETABLE);
966 }
967 
968 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
969 ({							\
970 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
971 	pte_t *__pte = pte_offset_map(pmd, address);	\
972 	*(ptlp) = __ptl;				\
973 	spin_lock(__ptl);				\
974 	__pte;						\
975 })
976 
977 #define pte_unmap_unlock(pte, ptl)	do {		\
978 	spin_unlock(ptl);				\
979 	pte_unmap(pte);					\
980 } while (0)
981 
982 #define pte_alloc_map(mm, pmd, address)			\
983 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
984 		NULL: pte_offset_map(pmd, address))
985 
986 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
987 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
988 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
989 
990 #define pte_alloc_kernel(pmd, address)			\
991 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
992 		NULL: pte_offset_kernel(pmd, address))
993 
994 extern void free_area_init(unsigned long * zones_size);
995 extern void free_area_init_node(int nid, unsigned long * zones_size,
996 		unsigned long zone_start_pfn, unsigned long *zholes_size);
997 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
998 /*
999  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1000  * zones, allocate the backing mem_map and account for memory holes in a more
1001  * architecture independent manner. This is a substitute for creating the
1002  * zone_sizes[] and zholes_size[] arrays and passing them to
1003  * free_area_init_node()
1004  *
1005  * An architecture is expected to register range of page frames backed by
1006  * physical memory with add_active_range() before calling
1007  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1008  * usage, an architecture is expected to do something like
1009  *
1010  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1011  * 							 max_highmem_pfn};
1012  * for_each_valid_physical_page_range()
1013  * 	add_active_range(node_id, start_pfn, end_pfn)
1014  * free_area_init_nodes(max_zone_pfns);
1015  *
1016  * If the architecture guarantees that there are no holes in the ranges
1017  * registered with add_active_range(), free_bootmem_active_regions()
1018  * will call free_bootmem_node() for each registered physical page range.
1019  * Similarly sparse_memory_present_with_active_regions() calls
1020  * memory_present() for each range when SPARSEMEM is enabled.
1021  *
1022  * See mm/page_alloc.c for more information on each function exposed by
1023  * CONFIG_ARCH_POPULATES_NODE_MAP
1024  */
1025 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1026 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1027 					unsigned long end_pfn);
1028 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1029 					unsigned long end_pfn);
1030 extern void remove_all_active_ranges(void);
1031 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1032 						unsigned long end_pfn);
1033 extern void get_pfn_range_for_nid(unsigned int nid,
1034 			unsigned long *start_pfn, unsigned long *end_pfn);
1035 extern unsigned long find_min_pfn_with_active_regions(void);
1036 extern void free_bootmem_with_active_regions(int nid,
1037 						unsigned long max_low_pfn);
1038 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1039 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1040 extern void sparse_memory_present_with_active_regions(int nid);
1041 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1042 
1043 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1044     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1045 static inline int __early_pfn_to_nid(unsigned long pfn)
1046 {
1047 	return 0;
1048 }
1049 #else
1050 /* please see mm/page_alloc.c */
1051 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1052 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1053 /* there is a per-arch backend function. */
1054 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1055 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1056 #endif
1057 
1058 extern void set_dma_reserve(unsigned long new_dma_reserve);
1059 extern void memmap_init_zone(unsigned long, int, unsigned long,
1060 				unsigned long, enum memmap_context);
1061 extern void setup_per_zone_wmarks(void);
1062 extern void calculate_zone_inactive_ratio(struct zone *zone);
1063 extern void mem_init(void);
1064 extern void __init mmap_init(void);
1065 extern void show_mem(void);
1066 extern void si_meminfo(struct sysinfo * val);
1067 extern void si_meminfo_node(struct sysinfo *val, int nid);
1068 extern int after_bootmem;
1069 
1070 #ifdef CONFIG_NUMA
1071 extern void setup_per_cpu_pageset(void);
1072 #else
1073 static inline void setup_per_cpu_pageset(void) {}
1074 #endif
1075 
1076 /* nommu.c */
1077 extern atomic_long_t mmap_pages_allocated;
1078 
1079 /* prio_tree.c */
1080 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1081 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1082 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1083 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1084 	struct prio_tree_iter *iter);
1085 
1086 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1087 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1088 		(vma = vma_prio_tree_next(vma, iter)); )
1089 
1090 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1091 					struct list_head *list)
1092 {
1093 	vma->shared.vm_set.parent = NULL;
1094 	list_add_tail(&vma->shared.vm_set.list, list);
1095 }
1096 
1097 /* mmap.c */
1098 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1099 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1100 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1101 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1102 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1103 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1104 	struct mempolicy *);
1105 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1106 extern int split_vma(struct mm_struct *,
1107 	struct vm_area_struct *, unsigned long addr, int new_below);
1108 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1109 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1110 	struct rb_node **, struct rb_node *);
1111 extern void unlink_file_vma(struct vm_area_struct *);
1112 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1113 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1114 extern void exit_mmap(struct mm_struct *);
1115 
1116 extern int mm_take_all_locks(struct mm_struct *mm);
1117 extern void mm_drop_all_locks(struct mm_struct *mm);
1118 
1119 #ifdef CONFIG_PROC_FS
1120 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1121 extern void added_exe_file_vma(struct mm_struct *mm);
1122 extern void removed_exe_file_vma(struct mm_struct *mm);
1123 #else
1124 static inline void added_exe_file_vma(struct mm_struct *mm)
1125 {}
1126 
1127 static inline void removed_exe_file_vma(struct mm_struct *mm)
1128 {}
1129 #endif /* CONFIG_PROC_FS */
1130 
1131 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1132 extern int install_special_mapping(struct mm_struct *mm,
1133 				   unsigned long addr, unsigned long len,
1134 				   unsigned long flags, struct page **pages);
1135 
1136 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1137 
1138 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1139 	unsigned long len, unsigned long prot,
1140 	unsigned long flag, unsigned long pgoff);
1141 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1142 	unsigned long len, unsigned long flags,
1143 	unsigned int vm_flags, unsigned long pgoff);
1144 
1145 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1146 	unsigned long len, unsigned long prot,
1147 	unsigned long flag, unsigned long offset)
1148 {
1149 	unsigned long ret = -EINVAL;
1150 	if ((offset + PAGE_ALIGN(len)) < offset)
1151 		goto out;
1152 	if (!(offset & ~PAGE_MASK))
1153 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1154 out:
1155 	return ret;
1156 }
1157 
1158 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1159 
1160 extern unsigned long do_brk(unsigned long, unsigned long);
1161 
1162 /* filemap.c */
1163 extern unsigned long page_unuse(struct page *);
1164 extern void truncate_inode_pages(struct address_space *, loff_t);
1165 extern void truncate_inode_pages_range(struct address_space *,
1166 				       loff_t lstart, loff_t lend);
1167 
1168 /* generic vm_area_ops exported for stackable file systems */
1169 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1170 
1171 /* mm/page-writeback.c */
1172 int write_one_page(struct page *page, int wait);
1173 void task_dirty_inc(struct task_struct *tsk);
1174 
1175 /* readahead.c */
1176 #define VM_MAX_READAHEAD	128	/* kbytes */
1177 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1178 
1179 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1180 			pgoff_t offset, unsigned long nr_to_read);
1181 
1182 void page_cache_sync_readahead(struct address_space *mapping,
1183 			       struct file_ra_state *ra,
1184 			       struct file *filp,
1185 			       pgoff_t offset,
1186 			       unsigned long size);
1187 
1188 void page_cache_async_readahead(struct address_space *mapping,
1189 				struct file_ra_state *ra,
1190 				struct file *filp,
1191 				struct page *pg,
1192 				pgoff_t offset,
1193 				unsigned long size);
1194 
1195 unsigned long max_sane_readahead(unsigned long nr);
1196 unsigned long ra_submit(struct file_ra_state *ra,
1197 			struct address_space *mapping,
1198 			struct file *filp);
1199 
1200 /* Do stack extension */
1201 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1202 #ifdef CONFIG_IA64
1203 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1204 #endif
1205 extern int expand_stack_downwards(struct vm_area_struct *vma,
1206 				  unsigned long address);
1207 
1208 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1209 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1210 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1211 					     struct vm_area_struct **pprev);
1212 
1213 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1214    NULL if none.  Assume start_addr < end_addr. */
1215 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1216 {
1217 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1218 
1219 	if (vma && end_addr <= vma->vm_start)
1220 		vma = NULL;
1221 	return vma;
1222 }
1223 
1224 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1225 {
1226 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1227 }
1228 
1229 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1230 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1231 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1232 			unsigned long pfn, unsigned long size, pgprot_t);
1233 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1234 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1235 			unsigned long pfn);
1236 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1237 			unsigned long pfn);
1238 
1239 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1240 			unsigned int foll_flags);
1241 #define FOLL_WRITE	0x01	/* check pte is writable */
1242 #define FOLL_TOUCH	0x02	/* mark page accessed */
1243 #define FOLL_GET	0x04	/* do get_page on page */
1244 #define FOLL_ANON	0x08	/* give ZERO_PAGE if no pgtable */
1245 
1246 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1247 			void *data);
1248 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1249 			       unsigned long size, pte_fn_t fn, void *data);
1250 
1251 #ifdef CONFIG_PROC_FS
1252 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1253 #else
1254 static inline void vm_stat_account(struct mm_struct *mm,
1255 			unsigned long flags, struct file *file, long pages)
1256 {
1257 }
1258 #endif /* CONFIG_PROC_FS */
1259 
1260 #ifdef CONFIG_DEBUG_PAGEALLOC
1261 extern int debug_pagealloc_enabled;
1262 
1263 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1264 
1265 static inline void enable_debug_pagealloc(void)
1266 {
1267 	debug_pagealloc_enabled = 1;
1268 }
1269 #ifdef CONFIG_HIBERNATION
1270 extern bool kernel_page_present(struct page *page);
1271 #endif /* CONFIG_HIBERNATION */
1272 #else
1273 static inline void
1274 kernel_map_pages(struct page *page, int numpages, int enable) {}
1275 static inline void enable_debug_pagealloc(void)
1276 {
1277 }
1278 #ifdef CONFIG_HIBERNATION
1279 static inline bool kernel_page_present(struct page *page) { return true; }
1280 #endif /* CONFIG_HIBERNATION */
1281 #endif
1282 
1283 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1284 #ifdef	__HAVE_ARCH_GATE_AREA
1285 int in_gate_area_no_task(unsigned long addr);
1286 int in_gate_area(struct task_struct *task, unsigned long addr);
1287 #else
1288 int in_gate_area_no_task(unsigned long addr);
1289 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1290 #endif	/* __HAVE_ARCH_GATE_AREA */
1291 
1292 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1293 					void __user *, size_t *, loff_t *);
1294 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1295 			unsigned long lru_pages);
1296 
1297 #ifndef CONFIG_MMU
1298 #define randomize_va_space 0
1299 #else
1300 extern int randomize_va_space;
1301 #endif
1302 
1303 const char * arch_vma_name(struct vm_area_struct *vma);
1304 void print_vma_addr(char *prefix, unsigned long rip);
1305 
1306 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1307 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1308 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1309 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1310 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1311 void *vmemmap_alloc_block(unsigned long size, int node);
1312 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1313 int vmemmap_populate_basepages(struct page *start_page,
1314 						unsigned long pages, int node);
1315 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1316 void vmemmap_populate_print_last(void);
1317 
1318 extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1319 				 size_t size);
1320 extern void refund_locked_memory(struct mm_struct *mm, size_t size);
1321 #endif /* __KERNEL__ */
1322 #endif /* _LINUX_MM_H */
1323