xref: /linux-6.15/include/linux/mm.h (revision 652afc27)
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/mmdebug.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/prio_tree.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 
17 struct mempolicy;
18 struct anon_vma;
19 struct file_ra_state;
20 struct user_struct;
21 struct writeback_control;
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 /*
60  * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
61  * disabled, then there's a single shared list of VMAs maintained by the
62  * system, and mm's subscribe to these individually
63  */
64 struct vm_list_struct {
65 	struct vm_list_struct	*next;
66 	struct vm_area_struct	*vma;
67 };
68 
69 #ifndef CONFIG_MMU
70 extern struct rb_root nommu_vma_tree;
71 extern struct rw_semaphore nommu_vma_sem;
72 
73 extern unsigned int kobjsize(const void *objp);
74 #endif
75 
76 /*
77  * vm_flags in vm_area_struct, see mm_types.h.
78  */
79 #define VM_READ		0x00000001	/* currently active flags */
80 #define VM_WRITE	0x00000002
81 #define VM_EXEC		0x00000004
82 #define VM_SHARED	0x00000008
83 
84 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
85 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
86 #define VM_MAYWRITE	0x00000020
87 #define VM_MAYEXEC	0x00000040
88 #define VM_MAYSHARE	0x00000080
89 
90 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
91 #define VM_GROWSUP	0x00000200
92 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
93 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
94 
95 #define VM_EXECUTABLE	0x00001000
96 #define VM_LOCKED	0x00002000
97 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
98 
99 					/* Used by sys_madvise() */
100 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
101 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
102 
103 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
104 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
105 #define VM_RESERVED	0x00080000	/* Count as reserved_vm like IO */
106 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
107 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
108 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
109 #define VM_NONLINEAR	0x00800000	/* Is non-linear (remap_file_pages) */
110 #define VM_MAPPED_COPY	0x01000000	/* T if mapped copy of data (nommu mmap) */
111 #define VM_INSERTPAGE	0x02000000	/* The vma has had "vm_insert_page()" done on it */
112 #define VM_ALWAYSDUMP	0x04000000	/* Always include in core dumps */
113 
114 #define VM_CAN_NONLINEAR 0x08000000	/* Has ->fault & does nonlinear pages */
115 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
116 #define VM_SAO		0x20000000	/* Strong Access Ordering (powerpc) */
117 
118 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
119 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
120 #endif
121 
122 #ifdef CONFIG_STACK_GROWSUP
123 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
124 #else
125 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
126 #endif
127 
128 #define VM_READHINTMASK			(VM_SEQ_READ | VM_RAND_READ)
129 #define VM_ClearReadHint(v)		(v)->vm_flags &= ~VM_READHINTMASK
130 #define VM_NormalReadHint(v)		(!((v)->vm_flags & VM_READHINTMASK))
131 #define VM_SequentialReadHint(v)	((v)->vm_flags & VM_SEQ_READ)
132 #define VM_RandomReadHint(v)		((v)->vm_flags & VM_RAND_READ)
133 
134 /*
135  * special vmas that are non-mergable, non-mlock()able
136  */
137 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
138 
139 /*
140  * mapping from the currently active vm_flags protection bits (the
141  * low four bits) to a page protection mask..
142  */
143 extern pgprot_t protection_map[16];
144 
145 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
146 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
147 
148 /*
149  * This interface is used by x86 PAT code to identify a pfn mapping that is
150  * linear over entire vma. This is to optimize PAT code that deals with
151  * marking the physical region with a particular prot. This is not for generic
152  * mm use. Note also that this check will not work if the pfn mapping is
153  * linear for a vma starting at physical address 0. In which case PAT code
154  * falls back to slow path of reserving physical range page by page.
155  */
156 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
157 {
158 	return ((vma->vm_flags & VM_PFNMAP) && vma->vm_pgoff);
159 }
160 
161 static inline int is_pfn_mapping(struct vm_area_struct *vma)
162 {
163 	return (vma->vm_flags & VM_PFNMAP);
164 }
165 
166 /*
167  * vm_fault is filled by the the pagefault handler and passed to the vma's
168  * ->fault function. The vma's ->fault is responsible for returning a bitmask
169  * of VM_FAULT_xxx flags that give details about how the fault was handled.
170  *
171  * pgoff should be used in favour of virtual_address, if possible. If pgoff
172  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
173  * mapping support.
174  */
175 struct vm_fault {
176 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
177 	pgoff_t pgoff;			/* Logical page offset based on vma */
178 	void __user *virtual_address;	/* Faulting virtual address */
179 
180 	struct page *page;		/* ->fault handlers should return a
181 					 * page here, unless VM_FAULT_NOPAGE
182 					 * is set (which is also implied by
183 					 * VM_FAULT_ERROR).
184 					 */
185 };
186 
187 /*
188  * These are the virtual MM functions - opening of an area, closing and
189  * unmapping it (needed to keep files on disk up-to-date etc), pointer
190  * to the functions called when a no-page or a wp-page exception occurs.
191  */
192 struct vm_operations_struct {
193 	void (*open)(struct vm_area_struct * area);
194 	void (*close)(struct vm_area_struct * area);
195 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
196 
197 	/* notification that a previously read-only page is about to become
198 	 * writable, if an error is returned it will cause a SIGBUS */
199 	int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
200 
201 	/* called by access_process_vm when get_user_pages() fails, typically
202 	 * for use by special VMAs that can switch between memory and hardware
203 	 */
204 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
205 		      void *buf, int len, int write);
206 #ifdef CONFIG_NUMA
207 	/*
208 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
209 	 * to hold the policy upon return.  Caller should pass NULL @new to
210 	 * remove a policy and fall back to surrounding context--i.e. do not
211 	 * install a MPOL_DEFAULT policy, nor the task or system default
212 	 * mempolicy.
213 	 */
214 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
215 
216 	/*
217 	 * get_policy() op must add reference [mpol_get()] to any policy at
218 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
219 	 * in mm/mempolicy.c will do this automatically.
220 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
221 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
222 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
223 	 * must return NULL--i.e., do not "fallback" to task or system default
224 	 * policy.
225 	 */
226 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
227 					unsigned long addr);
228 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
229 		const nodemask_t *to, unsigned long flags);
230 #endif
231 };
232 
233 struct mmu_gather;
234 struct inode;
235 
236 #define page_private(page)		((page)->private)
237 #define set_page_private(page, v)	((page)->private = (v))
238 
239 /*
240  * FIXME: take this include out, include page-flags.h in
241  * files which need it (119 of them)
242  */
243 #include <linux/page-flags.h>
244 
245 /*
246  * Methods to modify the page usage count.
247  *
248  * What counts for a page usage:
249  * - cache mapping   (page->mapping)
250  * - private data    (page->private)
251  * - page mapped in a task's page tables, each mapping
252  *   is counted separately
253  *
254  * Also, many kernel routines increase the page count before a critical
255  * routine so they can be sure the page doesn't go away from under them.
256  */
257 
258 /*
259  * Drop a ref, return true if the refcount fell to zero (the page has no users)
260  */
261 static inline int put_page_testzero(struct page *page)
262 {
263 	VM_BUG_ON(atomic_read(&page->_count) == 0);
264 	return atomic_dec_and_test(&page->_count);
265 }
266 
267 /*
268  * Try to grab a ref unless the page has a refcount of zero, return false if
269  * that is the case.
270  */
271 static inline int get_page_unless_zero(struct page *page)
272 {
273 	VM_BUG_ON(PageTail(page));
274 	return atomic_inc_not_zero(&page->_count);
275 }
276 
277 /* Support for virtually mapped pages */
278 struct page *vmalloc_to_page(const void *addr);
279 unsigned long vmalloc_to_pfn(const void *addr);
280 
281 /*
282  * Determine if an address is within the vmalloc range
283  *
284  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
285  * is no special casing required.
286  */
287 static inline int is_vmalloc_addr(const void *x)
288 {
289 #ifdef CONFIG_MMU
290 	unsigned long addr = (unsigned long)x;
291 
292 	return addr >= VMALLOC_START && addr < VMALLOC_END;
293 #else
294 	return 0;
295 #endif
296 }
297 
298 static inline struct page *compound_head(struct page *page)
299 {
300 	if (unlikely(PageTail(page)))
301 		return page->first_page;
302 	return page;
303 }
304 
305 static inline int page_count(struct page *page)
306 {
307 	return atomic_read(&compound_head(page)->_count);
308 }
309 
310 static inline void get_page(struct page *page)
311 {
312 	page = compound_head(page);
313 	VM_BUG_ON(atomic_read(&page->_count) == 0);
314 	atomic_inc(&page->_count);
315 }
316 
317 static inline struct page *virt_to_head_page(const void *x)
318 {
319 	struct page *page = virt_to_page(x);
320 	return compound_head(page);
321 }
322 
323 /*
324  * Setup the page count before being freed into the page allocator for
325  * the first time (boot or memory hotplug)
326  */
327 static inline void init_page_count(struct page *page)
328 {
329 	atomic_set(&page->_count, 1);
330 }
331 
332 void put_page(struct page *page);
333 void put_pages_list(struct list_head *pages);
334 
335 void split_page(struct page *page, unsigned int order);
336 
337 /*
338  * Compound pages have a destructor function.  Provide a
339  * prototype for that function and accessor functions.
340  * These are _only_ valid on the head of a PG_compound page.
341  */
342 typedef void compound_page_dtor(struct page *);
343 
344 static inline void set_compound_page_dtor(struct page *page,
345 						compound_page_dtor *dtor)
346 {
347 	page[1].lru.next = (void *)dtor;
348 }
349 
350 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
351 {
352 	return (compound_page_dtor *)page[1].lru.next;
353 }
354 
355 static inline int compound_order(struct page *page)
356 {
357 	if (!PageHead(page))
358 		return 0;
359 	return (unsigned long)page[1].lru.prev;
360 }
361 
362 static inline void set_compound_order(struct page *page, unsigned long order)
363 {
364 	page[1].lru.prev = (void *)order;
365 }
366 
367 /*
368  * Multiple processes may "see" the same page. E.g. for untouched
369  * mappings of /dev/null, all processes see the same page full of
370  * zeroes, and text pages of executables and shared libraries have
371  * only one copy in memory, at most, normally.
372  *
373  * For the non-reserved pages, page_count(page) denotes a reference count.
374  *   page_count() == 0 means the page is free. page->lru is then used for
375  *   freelist management in the buddy allocator.
376  *   page_count() > 0  means the page has been allocated.
377  *
378  * Pages are allocated by the slab allocator in order to provide memory
379  * to kmalloc and kmem_cache_alloc. In this case, the management of the
380  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
381  * unless a particular usage is carefully commented. (the responsibility of
382  * freeing the kmalloc memory is the caller's, of course).
383  *
384  * A page may be used by anyone else who does a __get_free_page().
385  * In this case, page_count still tracks the references, and should only
386  * be used through the normal accessor functions. The top bits of page->flags
387  * and page->virtual store page management information, but all other fields
388  * are unused and could be used privately, carefully. The management of this
389  * page is the responsibility of the one who allocated it, and those who have
390  * subsequently been given references to it.
391  *
392  * The other pages (we may call them "pagecache pages") are completely
393  * managed by the Linux memory manager: I/O, buffers, swapping etc.
394  * The following discussion applies only to them.
395  *
396  * A pagecache page contains an opaque `private' member, which belongs to the
397  * page's address_space. Usually, this is the address of a circular list of
398  * the page's disk buffers. PG_private must be set to tell the VM to call
399  * into the filesystem to release these pages.
400  *
401  * A page may belong to an inode's memory mapping. In this case, page->mapping
402  * is the pointer to the inode, and page->index is the file offset of the page,
403  * in units of PAGE_CACHE_SIZE.
404  *
405  * If pagecache pages are not associated with an inode, they are said to be
406  * anonymous pages. These may become associated with the swapcache, and in that
407  * case PG_swapcache is set, and page->private is an offset into the swapcache.
408  *
409  * In either case (swapcache or inode backed), the pagecache itself holds one
410  * reference to the page. Setting PG_private should also increment the
411  * refcount. The each user mapping also has a reference to the page.
412  *
413  * The pagecache pages are stored in a per-mapping radix tree, which is
414  * rooted at mapping->page_tree, and indexed by offset.
415  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
416  * lists, we instead now tag pages as dirty/writeback in the radix tree.
417  *
418  * All pagecache pages may be subject to I/O:
419  * - inode pages may need to be read from disk,
420  * - inode pages which have been modified and are MAP_SHARED may need
421  *   to be written back to the inode on disk,
422  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
423  *   modified may need to be swapped out to swap space and (later) to be read
424  *   back into memory.
425  */
426 
427 /*
428  * The zone field is never updated after free_area_init_core()
429  * sets it, so none of the operations on it need to be atomic.
430  */
431 
432 
433 /*
434  * page->flags layout:
435  *
436  * There are three possibilities for how page->flags get
437  * laid out.  The first is for the normal case, without
438  * sparsemem.  The second is for sparsemem when there is
439  * plenty of space for node and section.  The last is when
440  * we have run out of space and have to fall back to an
441  * alternate (slower) way of determining the node.
442  *
443  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
444  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
445  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
446  */
447 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
448 #define SECTIONS_WIDTH		SECTIONS_SHIFT
449 #else
450 #define SECTIONS_WIDTH		0
451 #endif
452 
453 #define ZONES_WIDTH		ZONES_SHIFT
454 
455 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
456 #define NODES_WIDTH		NODES_SHIFT
457 #else
458 #ifdef CONFIG_SPARSEMEM_VMEMMAP
459 #error "Vmemmap: No space for nodes field in page flags"
460 #endif
461 #define NODES_WIDTH		0
462 #endif
463 
464 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
465 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
466 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
467 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
468 
469 /*
470  * We are going to use the flags for the page to node mapping if its in
471  * there.  This includes the case where there is no node, so it is implicit.
472  */
473 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
474 #define NODE_NOT_IN_PAGE_FLAGS
475 #endif
476 
477 #ifndef PFN_SECTION_SHIFT
478 #define PFN_SECTION_SHIFT 0
479 #endif
480 
481 /*
482  * Define the bit shifts to access each section.  For non-existant
483  * sections we define the shift as 0; that plus a 0 mask ensures
484  * the compiler will optimise away reference to them.
485  */
486 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
487 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
488 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
489 
490 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
491 #ifdef NODE_NOT_IN_PAGEFLAGS
492 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
493 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
494 						SECTIONS_PGOFF : ZONES_PGOFF)
495 #else
496 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
497 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
498 						NODES_PGOFF : ZONES_PGOFF)
499 #endif
500 
501 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
502 
503 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
504 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
505 #endif
506 
507 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
508 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
509 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
510 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
511 
512 static inline enum zone_type page_zonenum(struct page *page)
513 {
514 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
515 }
516 
517 /*
518  * The identification function is only used by the buddy allocator for
519  * determining if two pages could be buddies. We are not really
520  * identifying a zone since we could be using a the section number
521  * id if we have not node id available in page flags.
522  * We guarantee only that it will return the same value for two
523  * combinable pages in a zone.
524  */
525 static inline int page_zone_id(struct page *page)
526 {
527 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
528 }
529 
530 static inline int zone_to_nid(struct zone *zone)
531 {
532 #ifdef CONFIG_NUMA
533 	return zone->node;
534 #else
535 	return 0;
536 #endif
537 }
538 
539 #ifdef NODE_NOT_IN_PAGE_FLAGS
540 extern int page_to_nid(struct page *page);
541 #else
542 static inline int page_to_nid(struct page *page)
543 {
544 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
545 }
546 #endif
547 
548 static inline struct zone *page_zone(struct page *page)
549 {
550 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
551 }
552 
553 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
554 static inline unsigned long page_to_section(struct page *page)
555 {
556 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
557 }
558 #endif
559 
560 static inline void set_page_zone(struct page *page, enum zone_type zone)
561 {
562 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
563 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
564 }
565 
566 static inline void set_page_node(struct page *page, unsigned long node)
567 {
568 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
569 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
570 }
571 
572 static inline void set_page_section(struct page *page, unsigned long section)
573 {
574 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
575 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
576 }
577 
578 static inline void set_page_links(struct page *page, enum zone_type zone,
579 	unsigned long node, unsigned long pfn)
580 {
581 	set_page_zone(page, zone);
582 	set_page_node(page, node);
583 	set_page_section(page, pfn_to_section_nr(pfn));
584 }
585 
586 /*
587  * If a hint addr is less than mmap_min_addr change hint to be as
588  * low as possible but still greater than mmap_min_addr
589  */
590 static inline unsigned long round_hint_to_min(unsigned long hint)
591 {
592 #ifdef CONFIG_SECURITY
593 	hint &= PAGE_MASK;
594 	if (((void *)hint != NULL) &&
595 	    (hint < mmap_min_addr))
596 		return PAGE_ALIGN(mmap_min_addr);
597 #endif
598 	return hint;
599 }
600 
601 /*
602  * Some inline functions in vmstat.h depend on page_zone()
603  */
604 #include <linux/vmstat.h>
605 
606 static __always_inline void *lowmem_page_address(struct page *page)
607 {
608 	return __va(page_to_pfn(page) << PAGE_SHIFT);
609 }
610 
611 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
612 #define HASHED_PAGE_VIRTUAL
613 #endif
614 
615 #if defined(WANT_PAGE_VIRTUAL)
616 #define page_address(page) ((page)->virtual)
617 #define set_page_address(page, address)			\
618 	do {						\
619 		(page)->virtual = (address);		\
620 	} while(0)
621 #define page_address_init()  do { } while(0)
622 #endif
623 
624 #if defined(HASHED_PAGE_VIRTUAL)
625 void *page_address(struct page *page);
626 void set_page_address(struct page *page, void *virtual);
627 void page_address_init(void);
628 #endif
629 
630 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
631 #define page_address(page) lowmem_page_address(page)
632 #define set_page_address(page, address)  do { } while(0)
633 #define page_address_init()  do { } while(0)
634 #endif
635 
636 /*
637  * On an anonymous page mapped into a user virtual memory area,
638  * page->mapping points to its anon_vma, not to a struct address_space;
639  * with the PAGE_MAPPING_ANON bit set to distinguish it.
640  *
641  * Please note that, confusingly, "page_mapping" refers to the inode
642  * address_space which maps the page from disk; whereas "page_mapped"
643  * refers to user virtual address space into which the page is mapped.
644  */
645 #define PAGE_MAPPING_ANON	1
646 
647 extern struct address_space swapper_space;
648 static inline struct address_space *page_mapping(struct page *page)
649 {
650 	struct address_space *mapping = page->mapping;
651 
652 	VM_BUG_ON(PageSlab(page));
653 #ifdef CONFIG_SWAP
654 	if (unlikely(PageSwapCache(page)))
655 		mapping = &swapper_space;
656 	else
657 #endif
658 	if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
659 		mapping = NULL;
660 	return mapping;
661 }
662 
663 static inline int PageAnon(struct page *page)
664 {
665 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
666 }
667 
668 /*
669  * Return the pagecache index of the passed page.  Regular pagecache pages
670  * use ->index whereas swapcache pages use ->private
671  */
672 static inline pgoff_t page_index(struct page *page)
673 {
674 	if (unlikely(PageSwapCache(page)))
675 		return page_private(page);
676 	return page->index;
677 }
678 
679 /*
680  * The atomic page->_mapcount, like _count, starts from -1:
681  * so that transitions both from it and to it can be tracked,
682  * using atomic_inc_and_test and atomic_add_negative(-1).
683  */
684 static inline void reset_page_mapcount(struct page *page)
685 {
686 	atomic_set(&(page)->_mapcount, -1);
687 }
688 
689 static inline int page_mapcount(struct page *page)
690 {
691 	return atomic_read(&(page)->_mapcount) + 1;
692 }
693 
694 /*
695  * Return true if this page is mapped into pagetables.
696  */
697 static inline int page_mapped(struct page *page)
698 {
699 	return atomic_read(&(page)->_mapcount) >= 0;
700 }
701 
702 /*
703  * Different kinds of faults, as returned by handle_mm_fault().
704  * Used to decide whether a process gets delivered SIGBUS or
705  * just gets major/minor fault counters bumped up.
706  */
707 
708 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
709 
710 #define VM_FAULT_OOM	0x0001
711 #define VM_FAULT_SIGBUS	0x0002
712 #define VM_FAULT_MAJOR	0x0004
713 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
714 
715 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
716 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
717 
718 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS)
719 
720 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
721 
722 extern void show_free_areas(void);
723 
724 #ifdef CONFIG_SHMEM
725 extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
726 #else
727 static inline int shmem_lock(struct file *file, int lock,
728 			    struct user_struct *user)
729 {
730 	return 0;
731 }
732 #endif
733 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
734 
735 int shmem_zero_setup(struct vm_area_struct *);
736 
737 #ifndef CONFIG_MMU
738 extern unsigned long shmem_get_unmapped_area(struct file *file,
739 					     unsigned long addr,
740 					     unsigned long len,
741 					     unsigned long pgoff,
742 					     unsigned long flags);
743 #endif
744 
745 extern int can_do_mlock(void);
746 extern int user_shm_lock(size_t, struct user_struct *);
747 extern void user_shm_unlock(size_t, struct user_struct *);
748 
749 /*
750  * Parameter block passed down to zap_pte_range in exceptional cases.
751  */
752 struct zap_details {
753 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
754 	struct address_space *check_mapping;	/* Check page->mapping if set */
755 	pgoff_t	first_index;			/* Lowest page->index to unmap */
756 	pgoff_t last_index;			/* Highest page->index to unmap */
757 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
758 	unsigned long truncate_count;		/* Compare vm_truncate_count */
759 };
760 
761 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
762 		pte_t pte);
763 
764 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
765 		unsigned long size);
766 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
767 		unsigned long size, struct zap_details *);
768 unsigned long unmap_vmas(struct mmu_gather **tlb,
769 		struct vm_area_struct *start_vma, unsigned long start_addr,
770 		unsigned long end_addr, unsigned long *nr_accounted,
771 		struct zap_details *);
772 
773 /**
774  * mm_walk - callbacks for walk_page_range
775  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
776  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
777  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
778  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
779  * @pte_hole: if set, called for each hole at all levels
780  *
781  * (see walk_page_range for more details)
782  */
783 struct mm_walk {
784 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
785 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
786 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
787 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
788 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
789 	struct mm_struct *mm;
790 	void *private;
791 };
792 
793 int walk_page_range(unsigned long addr, unsigned long end,
794 		struct mm_walk *walk);
795 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
796 		unsigned long end, unsigned long floor, unsigned long ceiling);
797 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
798 			struct vm_area_struct *vma);
799 void unmap_mapping_range(struct address_space *mapping,
800 		loff_t const holebegin, loff_t const holelen, int even_cows);
801 int follow_phys(struct vm_area_struct *vma, unsigned long address,
802 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
803 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
804 			void *buf, int len, int write);
805 
806 static inline void unmap_shared_mapping_range(struct address_space *mapping,
807 		loff_t const holebegin, loff_t const holelen)
808 {
809 	unmap_mapping_range(mapping, holebegin, holelen, 0);
810 }
811 
812 extern int vmtruncate(struct inode * inode, loff_t offset);
813 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
814 
815 #ifdef CONFIG_MMU
816 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
817 			unsigned long address, int write_access);
818 #else
819 static inline int handle_mm_fault(struct mm_struct *mm,
820 			struct vm_area_struct *vma, unsigned long address,
821 			int write_access)
822 {
823 	/* should never happen if there's no MMU */
824 	BUG();
825 	return VM_FAULT_SIGBUS;
826 }
827 #endif
828 
829 extern int make_pages_present(unsigned long addr, unsigned long end);
830 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
831 
832 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
833 		int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
834 
835 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
836 extern void do_invalidatepage(struct page *page, unsigned long offset);
837 
838 int __set_page_dirty_nobuffers(struct page *page);
839 int __set_page_dirty_no_writeback(struct page *page);
840 int redirty_page_for_writepage(struct writeback_control *wbc,
841 				struct page *page);
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  * get_user_pages_fast provides equivalent functionality to get_user_pages,
858  * operating on current and current->mm (force=0 and doesn't return any vmas).
859  *
860  * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
861  * can be made about locking. get_user_pages_fast is to be implemented in a
862  * way that is advantageous (vs get_user_pages()) when the user memory area is
863  * already faulted in and present in ptes. However if the pages have to be
864  * faulted in, it may turn out to be slightly slower).
865  */
866 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
867 			struct page **pages);
868 
869 /*
870  * A callback you can register to apply pressure to ageable caches.
871  *
872  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
873  * look through the least-recently-used 'nr_to_scan' entries and
874  * attempt to free them up.  It should return the number of objects
875  * which remain in the cache.  If it returns -1, it means it cannot do
876  * any scanning at this time (eg. there is a risk of deadlock).
877  *
878  * The 'gfpmask' refers to the allocation we are currently trying to
879  * fulfil.
880  *
881  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
882  * querying the cache size, so a fastpath for that case is appropriate.
883  */
884 struct shrinker {
885 	int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
886 	int seeks;	/* seeks to recreate an obj */
887 
888 	/* These are for internal use */
889 	struct list_head list;
890 	long nr;	/* objs pending delete */
891 };
892 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
893 extern void register_shrinker(struct shrinker *);
894 extern void unregister_shrinker(struct shrinker *);
895 
896 int vma_wants_writenotify(struct vm_area_struct *vma);
897 
898 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
899 
900 #ifdef __PAGETABLE_PUD_FOLDED
901 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
902 						unsigned long address)
903 {
904 	return 0;
905 }
906 #else
907 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
908 #endif
909 
910 #ifdef __PAGETABLE_PMD_FOLDED
911 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
912 						unsigned long address)
913 {
914 	return 0;
915 }
916 #else
917 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
918 #endif
919 
920 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
921 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
922 
923 /*
924  * The following ifdef needed to get the 4level-fixup.h header to work.
925  * Remove it when 4level-fixup.h has been removed.
926  */
927 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
928 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
929 {
930 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
931 		NULL: pud_offset(pgd, address);
932 }
933 
934 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
935 {
936 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
937 		NULL: pmd_offset(pud, address);
938 }
939 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
940 
941 #if USE_SPLIT_PTLOCKS
942 /*
943  * We tuck a spinlock to guard each pagetable page into its struct page,
944  * at page->private, with BUILD_BUG_ON to make sure that this will not
945  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
946  * When freeing, reset page->mapping so free_pages_check won't complain.
947  */
948 #define __pte_lockptr(page)	&((page)->ptl)
949 #define pte_lock_init(_page)	do {					\
950 	spin_lock_init(__pte_lockptr(_page));				\
951 } while (0)
952 #define pte_lock_deinit(page)	((page)->mapping = NULL)
953 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
954 #else	/* !USE_SPLIT_PTLOCKS */
955 /*
956  * We use mm->page_table_lock to guard all pagetable pages of the mm.
957  */
958 #define pte_lock_init(page)	do {} while (0)
959 #define pte_lock_deinit(page)	do {} while (0)
960 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
961 #endif /* USE_SPLIT_PTLOCKS */
962 
963 static inline void pgtable_page_ctor(struct page *page)
964 {
965 	pte_lock_init(page);
966 	inc_zone_page_state(page, NR_PAGETABLE);
967 }
968 
969 static inline void pgtable_page_dtor(struct page *page)
970 {
971 	pte_lock_deinit(page);
972 	dec_zone_page_state(page, NR_PAGETABLE);
973 }
974 
975 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
976 ({							\
977 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
978 	pte_t *__pte = pte_offset_map(pmd, address);	\
979 	*(ptlp) = __ptl;				\
980 	spin_lock(__ptl);				\
981 	__pte;						\
982 })
983 
984 #define pte_unmap_unlock(pte, ptl)	do {		\
985 	spin_unlock(ptl);				\
986 	pte_unmap(pte);					\
987 } while (0)
988 
989 #define pte_alloc_map(mm, pmd, address)			\
990 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
991 		NULL: pte_offset_map(pmd, address))
992 
993 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
994 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
995 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
996 
997 #define pte_alloc_kernel(pmd, address)			\
998 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
999 		NULL: pte_offset_kernel(pmd, address))
1000 
1001 extern void free_area_init(unsigned long * zones_size);
1002 extern void free_area_init_node(int nid, unsigned long * zones_size,
1003 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1004 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1005 /*
1006  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1007  * zones, allocate the backing mem_map and account for memory holes in a more
1008  * architecture independent manner. This is a substitute for creating the
1009  * zone_sizes[] and zholes_size[] arrays and passing them to
1010  * free_area_init_node()
1011  *
1012  * An architecture is expected to register range of page frames backed by
1013  * physical memory with add_active_range() before calling
1014  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1015  * usage, an architecture is expected to do something like
1016  *
1017  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1018  * 							 max_highmem_pfn};
1019  * for_each_valid_physical_page_range()
1020  * 	add_active_range(node_id, start_pfn, end_pfn)
1021  * free_area_init_nodes(max_zone_pfns);
1022  *
1023  * If the architecture guarantees that there are no holes in the ranges
1024  * registered with add_active_range(), free_bootmem_active_regions()
1025  * will call free_bootmem_node() for each registered physical page range.
1026  * Similarly sparse_memory_present_with_active_regions() calls
1027  * memory_present() for each range when SPARSEMEM is enabled.
1028  *
1029  * See mm/page_alloc.c for more information on each function exposed by
1030  * CONFIG_ARCH_POPULATES_NODE_MAP
1031  */
1032 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1033 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1034 					unsigned long end_pfn);
1035 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1036 					unsigned long end_pfn);
1037 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1038 					unsigned long end_pfn);
1039 extern void remove_all_active_ranges(void);
1040 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1041 						unsigned long end_pfn);
1042 extern void get_pfn_range_for_nid(unsigned int nid,
1043 			unsigned long *start_pfn, unsigned long *end_pfn);
1044 extern unsigned long find_min_pfn_with_active_regions(void);
1045 extern void free_bootmem_with_active_regions(int nid,
1046 						unsigned long max_low_pfn);
1047 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1048 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1049 extern void sparse_memory_present_with_active_regions(int nid);
1050 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1051 extern int early_pfn_to_nid(unsigned long pfn);
1052 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1053 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1054 extern void set_dma_reserve(unsigned long new_dma_reserve);
1055 extern void memmap_init_zone(unsigned long, int, unsigned long,
1056 				unsigned long, enum memmap_context);
1057 extern void setup_per_zone_pages_min(void);
1058 extern void mem_init(void);
1059 extern void show_mem(void);
1060 extern void si_meminfo(struct sysinfo * val);
1061 extern void si_meminfo_node(struct sysinfo *val, int nid);
1062 extern int after_bootmem;
1063 
1064 #ifdef CONFIG_NUMA
1065 extern void setup_per_cpu_pageset(void);
1066 #else
1067 static inline void setup_per_cpu_pageset(void) {}
1068 #endif
1069 
1070 /* prio_tree.c */
1071 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1072 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1073 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1074 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1075 	struct prio_tree_iter *iter);
1076 
1077 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1078 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1079 		(vma = vma_prio_tree_next(vma, iter)); )
1080 
1081 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1082 					struct list_head *list)
1083 {
1084 	vma->shared.vm_set.parent = NULL;
1085 	list_add_tail(&vma->shared.vm_set.list, list);
1086 }
1087 
1088 /* mmap.c */
1089 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1090 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1091 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1092 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1093 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1094 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1095 	struct mempolicy *);
1096 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1097 extern int split_vma(struct mm_struct *,
1098 	struct vm_area_struct *, unsigned long addr, int new_below);
1099 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1100 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1101 	struct rb_node **, struct rb_node *);
1102 extern void unlink_file_vma(struct vm_area_struct *);
1103 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1104 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1105 extern void exit_mmap(struct mm_struct *);
1106 
1107 extern int mm_take_all_locks(struct mm_struct *mm);
1108 extern void mm_drop_all_locks(struct mm_struct *mm);
1109 
1110 #ifdef CONFIG_PROC_FS
1111 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1112 extern void added_exe_file_vma(struct mm_struct *mm);
1113 extern void removed_exe_file_vma(struct mm_struct *mm);
1114 #else
1115 static inline void added_exe_file_vma(struct mm_struct *mm)
1116 {}
1117 
1118 static inline void removed_exe_file_vma(struct mm_struct *mm)
1119 {}
1120 #endif /* CONFIG_PROC_FS */
1121 
1122 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1123 extern int install_special_mapping(struct mm_struct *mm,
1124 				   unsigned long addr, unsigned long len,
1125 				   unsigned long flags, struct page **pages);
1126 
1127 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1128 
1129 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1130 	unsigned long len, unsigned long prot,
1131 	unsigned long flag, unsigned long pgoff);
1132 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1133 	unsigned long len, unsigned long flags,
1134 	unsigned int vm_flags, unsigned long pgoff,
1135 	int accountable);
1136 
1137 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1138 	unsigned long len, unsigned long prot,
1139 	unsigned long flag, unsigned long offset)
1140 {
1141 	unsigned long ret = -EINVAL;
1142 	if ((offset + PAGE_ALIGN(len)) < offset)
1143 		goto out;
1144 	if (!(offset & ~PAGE_MASK))
1145 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1146 out:
1147 	return ret;
1148 }
1149 
1150 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1151 
1152 extern unsigned long do_brk(unsigned long, unsigned long);
1153 
1154 /* filemap.c */
1155 extern unsigned long page_unuse(struct page *);
1156 extern void truncate_inode_pages(struct address_space *, loff_t);
1157 extern void truncate_inode_pages_range(struct address_space *,
1158 				       loff_t lstart, loff_t lend);
1159 
1160 /* generic vm_area_ops exported for stackable file systems */
1161 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1162 
1163 /* mm/page-writeback.c */
1164 int write_one_page(struct page *page, int wait);
1165 
1166 /* readahead.c */
1167 #define VM_MAX_READAHEAD	128	/* kbytes */
1168 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1169 
1170 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1171 			pgoff_t offset, unsigned long nr_to_read);
1172 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1173 			pgoff_t offset, unsigned long nr_to_read);
1174 
1175 void page_cache_sync_readahead(struct address_space *mapping,
1176 			       struct file_ra_state *ra,
1177 			       struct file *filp,
1178 			       pgoff_t offset,
1179 			       unsigned long size);
1180 
1181 void page_cache_async_readahead(struct address_space *mapping,
1182 				struct file_ra_state *ra,
1183 				struct file *filp,
1184 				struct page *pg,
1185 				pgoff_t offset,
1186 				unsigned long size);
1187 
1188 unsigned long max_sane_readahead(unsigned long nr);
1189 
1190 /* Do stack extension */
1191 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1192 #ifdef CONFIG_IA64
1193 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1194 #endif
1195 extern int expand_stack_downwards(struct vm_area_struct *vma,
1196 				  unsigned long address);
1197 
1198 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1199 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1200 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1201 					     struct vm_area_struct **pprev);
1202 
1203 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1204    NULL if none.  Assume start_addr < end_addr. */
1205 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1206 {
1207 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1208 
1209 	if (vma && end_addr <= vma->vm_start)
1210 		vma = NULL;
1211 	return vma;
1212 }
1213 
1214 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1215 {
1216 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1217 }
1218 
1219 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1220 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1221 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1222 			unsigned long pfn, unsigned long size, pgprot_t);
1223 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1224 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1225 			unsigned long pfn);
1226 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1227 			unsigned long pfn);
1228 
1229 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1230 			unsigned int foll_flags);
1231 #define FOLL_WRITE	0x01	/* check pte is writable */
1232 #define FOLL_TOUCH	0x02	/* mark page accessed */
1233 #define FOLL_GET	0x04	/* do get_page on page */
1234 #define FOLL_ANON	0x08	/* give ZERO_PAGE if no pgtable */
1235 
1236 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1237 			void *data);
1238 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1239 			       unsigned long size, pte_fn_t fn, void *data);
1240 
1241 #ifdef CONFIG_PROC_FS
1242 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1243 #else
1244 static inline void vm_stat_account(struct mm_struct *mm,
1245 			unsigned long flags, struct file *file, long pages)
1246 {
1247 }
1248 #endif /* CONFIG_PROC_FS */
1249 
1250 #ifdef CONFIG_DEBUG_PAGEALLOC
1251 extern int debug_pagealloc_enabled;
1252 
1253 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1254 
1255 static inline void enable_debug_pagealloc(void)
1256 {
1257 	debug_pagealloc_enabled = 1;
1258 }
1259 #ifdef CONFIG_HIBERNATION
1260 extern bool kernel_page_present(struct page *page);
1261 #endif /* CONFIG_HIBERNATION */
1262 #else
1263 static inline void
1264 kernel_map_pages(struct page *page, int numpages, int enable) {}
1265 static inline void enable_debug_pagealloc(void)
1266 {
1267 }
1268 #ifdef CONFIG_HIBERNATION
1269 static inline bool kernel_page_present(struct page *page) { return true; }
1270 #endif /* CONFIG_HIBERNATION */
1271 #endif
1272 
1273 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1274 #ifdef	__HAVE_ARCH_GATE_AREA
1275 int in_gate_area_no_task(unsigned long addr);
1276 int in_gate_area(struct task_struct *task, unsigned long addr);
1277 #else
1278 int in_gate_area_no_task(unsigned long addr);
1279 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1280 #endif	/* __HAVE_ARCH_GATE_AREA */
1281 
1282 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1283 					void __user *, size_t *, loff_t *);
1284 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1285 			unsigned long lru_pages);
1286 
1287 #ifndef CONFIG_MMU
1288 #define randomize_va_space 0
1289 #else
1290 extern int randomize_va_space;
1291 #endif
1292 
1293 const char * arch_vma_name(struct vm_area_struct *vma);
1294 void print_vma_addr(char *prefix, unsigned long rip);
1295 
1296 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1297 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1298 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1299 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1300 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1301 void *vmemmap_alloc_block(unsigned long size, int node);
1302 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1303 int vmemmap_populate_basepages(struct page *start_page,
1304 						unsigned long pages, int node);
1305 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1306 void vmemmap_populate_print_last(void);
1307 
1308 extern void *alloc_locked_buffer(size_t size);
1309 extern void free_locked_buffer(void *buffer, size_t size);
1310 #endif /* __KERNEL__ */
1311 #endif /* _LINUX_MM_H */
1312