xref: /linux-6.15/include/linux/mm.h (revision c9d05cfc)
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 #include <asm/page.h>
38 #include <asm/pgtable.h>
39 #include <asm/processor.h>
40 
41 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
42 
43 /* to align the pointer to the (next) page boundary */
44 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
45 
46 /*
47  * Linux kernel virtual memory manager primitives.
48  * The idea being to have a "virtual" mm in the same way
49  * we have a virtual fs - giving a cleaner interface to the
50  * mm details, and allowing different kinds of memory mappings
51  * (from shared memory to executable loading to arbitrary
52  * mmap() functions).
53  */
54 
55 extern struct kmem_cache *vm_area_cachep;
56 
57 #ifndef CONFIG_MMU
58 extern struct rb_root nommu_region_tree;
59 extern struct rw_semaphore nommu_region_sem;
60 
61 extern unsigned int kobjsize(const void *objp);
62 #endif
63 
64 /*
65  * vm_flags in vm_area_struct, see mm_types.h.
66  */
67 #define VM_READ		0x00000001	/* currently active flags */
68 #define VM_WRITE	0x00000002
69 #define VM_EXEC		0x00000004
70 #define VM_SHARED	0x00000008
71 
72 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
73 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
74 #define VM_MAYWRITE	0x00000020
75 #define VM_MAYEXEC	0x00000040
76 #define VM_MAYSHARE	0x00000080
77 
78 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
79 #define VM_GROWSUP	0x00000200
80 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
81 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
82 
83 #define VM_EXECUTABLE	0x00001000
84 #define VM_LOCKED	0x00002000
85 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
86 
87 					/* Used by sys_madvise() */
88 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
89 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
90 
91 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
92 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
93 #define VM_RESERVED	0x00080000	/* Count as reserved_vm like IO */
94 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
95 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
96 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
97 #define VM_NONLINEAR	0x00800000	/* Is non-linear (remap_file_pages) */
98 #define VM_MAPPED_COPY	0x01000000	/* T if mapped copy of data (nommu mmap) */
99 #define VM_INSERTPAGE	0x02000000	/* The vma has had "vm_insert_page()" done on it */
100 #define VM_ALWAYSDUMP	0x04000000	/* Always include in core dumps */
101 
102 #define VM_CAN_NONLINEAR 0x08000000	/* Has ->fault & does nonlinear pages */
103 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
104 #define VM_SAO		0x20000000	/* Strong Access Ordering (powerpc) */
105 #define VM_PFN_AT_MMAP	0x40000000	/* PFNMAP vma that is fully mapped at mmap time */
106 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
107 
108 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
109 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
110 #endif
111 
112 #ifdef CONFIG_STACK_GROWSUP
113 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
114 #else
115 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
116 #endif
117 
118 #define VM_READHINTMASK			(VM_SEQ_READ | VM_RAND_READ)
119 #define VM_ClearReadHint(v)		(v)->vm_flags &= ~VM_READHINTMASK
120 #define VM_NormalReadHint(v)		(!((v)->vm_flags & VM_READHINTMASK))
121 #define VM_SequentialReadHint(v)	((v)->vm_flags & VM_SEQ_READ)
122 #define VM_RandomReadHint(v)		((v)->vm_flags & VM_RAND_READ)
123 
124 /*
125  * special vmas that are non-mergable, non-mlock()able
126  */
127 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
128 
129 /*
130  * mapping from the currently active vm_flags protection bits (the
131  * low four bits) to a page protection mask..
132  */
133 extern pgprot_t protection_map[16];
134 
135 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
136 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
137 #define FAULT_FLAG_MKWRITE	0x04	/* Fault was mkwrite of existing pte */
138 
139 /*
140  * This interface is used by x86 PAT code to identify a pfn mapping that is
141  * linear over entire vma. This is to optimize PAT code that deals with
142  * marking the physical region with a particular prot. This is not for generic
143  * mm use. Note also that this check will not work if the pfn mapping is
144  * linear for a vma starting at physical address 0. In which case PAT code
145  * falls back to slow path of reserving physical range page by page.
146  */
147 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
148 {
149 	return (vma->vm_flags & VM_PFN_AT_MMAP);
150 }
151 
152 static inline int is_pfn_mapping(struct vm_area_struct *vma)
153 {
154 	return (vma->vm_flags & VM_PFNMAP);
155 }
156 
157 /*
158  * vm_fault is filled by the the pagefault handler and passed to the vma's
159  * ->fault function. The vma's ->fault is responsible for returning a bitmask
160  * of VM_FAULT_xxx flags that give details about how the fault was handled.
161  *
162  * pgoff should be used in favour of virtual_address, if possible. If pgoff
163  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
164  * mapping support.
165  */
166 struct vm_fault {
167 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
168 	pgoff_t pgoff;			/* Logical page offset based on vma */
169 	void __user *virtual_address;	/* Faulting virtual address */
170 
171 	struct page *page;		/* ->fault handlers should return a
172 					 * page here, unless VM_FAULT_NOPAGE
173 					 * is set (which is also implied by
174 					 * VM_FAULT_ERROR).
175 					 */
176 };
177 
178 /*
179  * These are the virtual MM functions - opening of an area, closing and
180  * unmapping it (needed to keep files on disk up-to-date etc), pointer
181  * to the functions called when a no-page or a wp-page exception occurs.
182  */
183 struct vm_operations_struct {
184 	void (*open)(struct vm_area_struct * area);
185 	void (*close)(struct vm_area_struct * area);
186 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
187 
188 	/* notification that a previously read-only page is about to become
189 	 * writable, if an error is returned it will cause a SIGBUS */
190 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
191 
192 	/* called by access_process_vm when get_user_pages() fails, typically
193 	 * for use by special VMAs that can switch between memory and hardware
194 	 */
195 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
196 		      void *buf, int len, int write);
197 #ifdef CONFIG_NUMA
198 	/*
199 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
200 	 * to hold the policy upon return.  Caller should pass NULL @new to
201 	 * remove a policy and fall back to surrounding context--i.e. do not
202 	 * install a MPOL_DEFAULT policy, nor the task or system default
203 	 * mempolicy.
204 	 */
205 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
206 
207 	/*
208 	 * get_policy() op must add reference [mpol_get()] to any policy at
209 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
210 	 * in mm/mempolicy.c will do this automatically.
211 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
212 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
213 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
214 	 * must return NULL--i.e., do not "fallback" to task or system default
215 	 * policy.
216 	 */
217 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
218 					unsigned long addr);
219 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
220 		const nodemask_t *to, unsigned long flags);
221 #endif
222 };
223 
224 struct mmu_gather;
225 struct inode;
226 
227 #define page_private(page)		((page)->private)
228 #define set_page_private(page, v)	((page)->private = (v))
229 
230 /*
231  * FIXME: take this include out, include page-flags.h in
232  * files which need it (119 of them)
233  */
234 #include <linux/page-flags.h>
235 
236 /*
237  * Methods to modify the page usage count.
238  *
239  * What counts for a page usage:
240  * - cache mapping   (page->mapping)
241  * - private data    (page->private)
242  * - page mapped in a task's page tables, each mapping
243  *   is counted separately
244  *
245  * Also, many kernel routines increase the page count before a critical
246  * routine so they can be sure the page doesn't go away from under them.
247  */
248 
249 /*
250  * Drop a ref, return true if the refcount fell to zero (the page has no users)
251  */
252 static inline int put_page_testzero(struct page *page)
253 {
254 	VM_BUG_ON(atomic_read(&page->_count) == 0);
255 	return atomic_dec_and_test(&page->_count);
256 }
257 
258 /*
259  * Try to grab a ref unless the page has a refcount of zero, return false if
260  * that is the case.
261  */
262 static inline int get_page_unless_zero(struct page *page)
263 {
264 	return atomic_inc_not_zero(&page->_count);
265 }
266 
267 /* Support for virtually mapped pages */
268 struct page *vmalloc_to_page(const void *addr);
269 unsigned long vmalloc_to_pfn(const void *addr);
270 
271 /*
272  * Determine if an address is within the vmalloc range
273  *
274  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
275  * is no special casing required.
276  */
277 static inline int is_vmalloc_addr(const void *x)
278 {
279 #ifdef CONFIG_MMU
280 	unsigned long addr = (unsigned long)x;
281 
282 	return addr >= VMALLOC_START && addr < VMALLOC_END;
283 #else
284 	return 0;
285 #endif
286 }
287 
288 static inline struct page *compound_head(struct page *page)
289 {
290 	if (unlikely(PageTail(page)))
291 		return page->first_page;
292 	return page;
293 }
294 
295 static inline int page_count(struct page *page)
296 {
297 	return atomic_read(&compound_head(page)->_count);
298 }
299 
300 static inline void get_page(struct page *page)
301 {
302 	page = compound_head(page);
303 	VM_BUG_ON(atomic_read(&page->_count) == 0);
304 	atomic_inc(&page->_count);
305 }
306 
307 static inline struct page *virt_to_head_page(const void *x)
308 {
309 	struct page *page = virt_to_page(x);
310 	return compound_head(page);
311 }
312 
313 /*
314  * Setup the page count before being freed into the page allocator for
315  * the first time (boot or memory hotplug)
316  */
317 static inline void init_page_count(struct page *page)
318 {
319 	atomic_set(&page->_count, 1);
320 }
321 
322 void put_page(struct page *page);
323 void put_pages_list(struct list_head *pages);
324 
325 void split_page(struct page *page, unsigned int order);
326 
327 /*
328  * Compound pages have a destructor function.  Provide a
329  * prototype for that function and accessor functions.
330  * These are _only_ valid on the head of a PG_compound page.
331  */
332 typedef void compound_page_dtor(struct page *);
333 
334 static inline void set_compound_page_dtor(struct page *page,
335 						compound_page_dtor *dtor)
336 {
337 	page[1].lru.next = (void *)dtor;
338 }
339 
340 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
341 {
342 	return (compound_page_dtor *)page[1].lru.next;
343 }
344 
345 static inline int compound_order(struct page *page)
346 {
347 	if (!PageHead(page))
348 		return 0;
349 	return (unsigned long)page[1].lru.prev;
350 }
351 
352 static inline void set_compound_order(struct page *page, unsigned long order)
353 {
354 	page[1].lru.prev = (void *)order;
355 }
356 
357 /*
358  * Multiple processes may "see" the same page. E.g. for untouched
359  * mappings of /dev/null, all processes see the same page full of
360  * zeroes, and text pages of executables and shared libraries have
361  * only one copy in memory, at most, normally.
362  *
363  * For the non-reserved pages, page_count(page) denotes a reference count.
364  *   page_count() == 0 means the page is free. page->lru is then used for
365  *   freelist management in the buddy allocator.
366  *   page_count() > 0  means the page has been allocated.
367  *
368  * Pages are allocated by the slab allocator in order to provide memory
369  * to kmalloc and kmem_cache_alloc. In this case, the management of the
370  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
371  * unless a particular usage is carefully commented. (the responsibility of
372  * freeing the kmalloc memory is the caller's, of course).
373  *
374  * A page may be used by anyone else who does a __get_free_page().
375  * In this case, page_count still tracks the references, and should only
376  * be used through the normal accessor functions. The top bits of page->flags
377  * and page->virtual store page management information, but all other fields
378  * are unused and could be used privately, carefully. The management of this
379  * page is the responsibility of the one who allocated it, and those who have
380  * subsequently been given references to it.
381  *
382  * The other pages (we may call them "pagecache pages") are completely
383  * managed by the Linux memory manager: I/O, buffers, swapping etc.
384  * The following discussion applies only to them.
385  *
386  * A pagecache page contains an opaque `private' member, which belongs to the
387  * page's address_space. Usually, this is the address of a circular list of
388  * the page's disk buffers. PG_private must be set to tell the VM to call
389  * into the filesystem to release these pages.
390  *
391  * A page may belong to an inode's memory mapping. In this case, page->mapping
392  * is the pointer to the inode, and page->index is the file offset of the page,
393  * in units of PAGE_CACHE_SIZE.
394  *
395  * If pagecache pages are not associated with an inode, they are said to be
396  * anonymous pages. These may become associated with the swapcache, and in that
397  * case PG_swapcache is set, and page->private is an offset into the swapcache.
398  *
399  * In either case (swapcache or inode backed), the pagecache itself holds one
400  * reference to the page. Setting PG_private should also increment the
401  * refcount. The each user mapping also has a reference to the page.
402  *
403  * The pagecache pages are stored in a per-mapping radix tree, which is
404  * rooted at mapping->page_tree, and indexed by offset.
405  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
406  * lists, we instead now tag pages as dirty/writeback in the radix tree.
407  *
408  * All pagecache pages may be subject to I/O:
409  * - inode pages may need to be read from disk,
410  * - inode pages which have been modified and are MAP_SHARED may need
411  *   to be written back to the inode on disk,
412  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
413  *   modified may need to be swapped out to swap space and (later) to be read
414  *   back into memory.
415  */
416 
417 /*
418  * The zone field is never updated after free_area_init_core()
419  * sets it, so none of the operations on it need to be atomic.
420  */
421 
422 
423 /*
424  * page->flags layout:
425  *
426  * There are three possibilities for how page->flags get
427  * laid out.  The first is for the normal case, without
428  * sparsemem.  The second is for sparsemem when there is
429  * plenty of space for node and section.  The last is when
430  * we have run out of space and have to fall back to an
431  * alternate (slower) way of determining the node.
432  *
433  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
434  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
435  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
436  */
437 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
438 #define SECTIONS_WIDTH		SECTIONS_SHIFT
439 #else
440 #define SECTIONS_WIDTH		0
441 #endif
442 
443 #define ZONES_WIDTH		ZONES_SHIFT
444 
445 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
446 #define NODES_WIDTH		NODES_SHIFT
447 #else
448 #ifdef CONFIG_SPARSEMEM_VMEMMAP
449 #error "Vmemmap: No space for nodes field in page flags"
450 #endif
451 #define NODES_WIDTH		0
452 #endif
453 
454 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
455 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
456 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
457 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
458 
459 /*
460  * We are going to use the flags for the page to node mapping if its in
461  * there.  This includes the case where there is no node, so it is implicit.
462  */
463 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
464 #define NODE_NOT_IN_PAGE_FLAGS
465 #endif
466 
467 #ifndef PFN_SECTION_SHIFT
468 #define PFN_SECTION_SHIFT 0
469 #endif
470 
471 /*
472  * Define the bit shifts to access each section.  For non-existant
473  * sections we define the shift as 0; that plus a 0 mask ensures
474  * the compiler will optimise away reference to them.
475  */
476 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
477 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
478 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
479 
480 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
481 #ifdef NODE_NOT_IN_PAGEFLAGS
482 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
483 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
484 						SECTIONS_PGOFF : ZONES_PGOFF)
485 #else
486 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
487 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
488 						NODES_PGOFF : ZONES_PGOFF)
489 #endif
490 
491 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
492 
493 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
494 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
495 #endif
496 
497 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
498 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
499 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
500 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
501 
502 static inline enum zone_type page_zonenum(struct page *page)
503 {
504 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
505 }
506 
507 /*
508  * The identification function is only used by the buddy allocator for
509  * determining if two pages could be buddies. We are not really
510  * identifying a zone since we could be using a the section number
511  * id if we have not node id available in page flags.
512  * We guarantee only that it will return the same value for two
513  * combinable pages in a zone.
514  */
515 static inline int page_zone_id(struct page *page)
516 {
517 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
518 }
519 
520 static inline int zone_to_nid(struct zone *zone)
521 {
522 #ifdef CONFIG_NUMA
523 	return zone->node;
524 #else
525 	return 0;
526 #endif
527 }
528 
529 #ifdef NODE_NOT_IN_PAGE_FLAGS
530 extern int page_to_nid(struct page *page);
531 #else
532 static inline int page_to_nid(struct page *page)
533 {
534 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
535 }
536 #endif
537 
538 static inline struct zone *page_zone(struct page *page)
539 {
540 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
541 }
542 
543 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
544 static inline unsigned long page_to_section(struct page *page)
545 {
546 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
547 }
548 #endif
549 
550 static inline void set_page_zone(struct page *page, enum zone_type zone)
551 {
552 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
553 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
554 }
555 
556 static inline void set_page_node(struct page *page, unsigned long node)
557 {
558 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
559 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
560 }
561 
562 static inline void set_page_section(struct page *page, unsigned long section)
563 {
564 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
565 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
566 }
567 
568 static inline void set_page_links(struct page *page, enum zone_type zone,
569 	unsigned long node, unsigned long pfn)
570 {
571 	set_page_zone(page, zone);
572 	set_page_node(page, node);
573 	set_page_section(page, pfn_to_section_nr(pfn));
574 }
575 
576 /*
577  * Some inline functions in vmstat.h depend on page_zone()
578  */
579 #include <linux/vmstat.h>
580 
581 static __always_inline void *lowmem_page_address(struct page *page)
582 {
583 	return __va(page_to_pfn(page) << PAGE_SHIFT);
584 }
585 
586 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
587 #define HASHED_PAGE_VIRTUAL
588 #endif
589 
590 #if defined(WANT_PAGE_VIRTUAL)
591 #define page_address(page) ((page)->virtual)
592 #define set_page_address(page, address)			\
593 	do {						\
594 		(page)->virtual = (address);		\
595 	} while(0)
596 #define page_address_init()  do { } while(0)
597 #endif
598 
599 #if defined(HASHED_PAGE_VIRTUAL)
600 void *page_address(struct page *page);
601 void set_page_address(struct page *page, void *virtual);
602 void page_address_init(void);
603 #endif
604 
605 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
606 #define page_address(page) lowmem_page_address(page)
607 #define set_page_address(page, address)  do { } while(0)
608 #define page_address_init()  do { } while(0)
609 #endif
610 
611 /*
612  * On an anonymous page mapped into a user virtual memory area,
613  * page->mapping points to its anon_vma, not to a struct address_space;
614  * with the PAGE_MAPPING_ANON bit set to distinguish it.
615  *
616  * Please note that, confusingly, "page_mapping" refers to the inode
617  * address_space which maps the page from disk; whereas "page_mapped"
618  * refers to user virtual address space into which the page is mapped.
619  */
620 #define PAGE_MAPPING_ANON	1
621 
622 extern struct address_space swapper_space;
623 static inline struct address_space *page_mapping(struct page *page)
624 {
625 	struct address_space *mapping = page->mapping;
626 
627 	VM_BUG_ON(PageSlab(page));
628 #ifdef CONFIG_SWAP
629 	if (unlikely(PageSwapCache(page)))
630 		mapping = &swapper_space;
631 	else
632 #endif
633 	if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
634 		mapping = NULL;
635 	return mapping;
636 }
637 
638 static inline int PageAnon(struct page *page)
639 {
640 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
641 }
642 
643 /*
644  * Return the pagecache index of the passed page.  Regular pagecache pages
645  * use ->index whereas swapcache pages use ->private
646  */
647 static inline pgoff_t page_index(struct page *page)
648 {
649 	if (unlikely(PageSwapCache(page)))
650 		return page_private(page);
651 	return page->index;
652 }
653 
654 /*
655  * The atomic page->_mapcount, like _count, starts from -1:
656  * so that transitions both from it and to it can be tracked,
657  * using atomic_inc_and_test and atomic_add_negative(-1).
658  */
659 static inline void reset_page_mapcount(struct page *page)
660 {
661 	atomic_set(&(page)->_mapcount, -1);
662 }
663 
664 static inline int page_mapcount(struct page *page)
665 {
666 	return atomic_read(&(page)->_mapcount) + 1;
667 }
668 
669 /*
670  * Return true if this page is mapped into pagetables.
671  */
672 static inline int page_mapped(struct page *page)
673 {
674 	return atomic_read(&(page)->_mapcount) >= 0;
675 }
676 
677 /*
678  * Different kinds of faults, as returned by handle_mm_fault().
679  * Used to decide whether a process gets delivered SIGBUS or
680  * just gets major/minor fault counters bumped up.
681  */
682 
683 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
684 
685 #define VM_FAULT_OOM	0x0001
686 #define VM_FAULT_SIGBUS	0x0002
687 #define VM_FAULT_MAJOR	0x0004
688 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
689 
690 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
691 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
692 
693 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS)
694 
695 /*
696  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
697  */
698 extern void pagefault_out_of_memory(void);
699 
700 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
701 
702 extern void show_free_areas(void);
703 
704 #ifdef CONFIG_SHMEM
705 extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
706 #else
707 static inline int shmem_lock(struct file *file, int lock,
708 			    struct user_struct *user)
709 {
710 	return 0;
711 }
712 #endif
713 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
714 
715 int shmem_zero_setup(struct vm_area_struct *);
716 
717 #ifndef CONFIG_MMU
718 extern unsigned long shmem_get_unmapped_area(struct file *file,
719 					     unsigned long addr,
720 					     unsigned long len,
721 					     unsigned long pgoff,
722 					     unsigned long flags);
723 #endif
724 
725 extern int can_do_mlock(void);
726 extern int user_shm_lock(size_t, struct user_struct *);
727 extern void user_shm_unlock(size_t, struct user_struct *);
728 
729 /*
730  * Parameter block passed down to zap_pte_range in exceptional cases.
731  */
732 struct zap_details {
733 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
734 	struct address_space *check_mapping;	/* Check page->mapping if set */
735 	pgoff_t	first_index;			/* Lowest page->index to unmap */
736 	pgoff_t last_index;			/* Highest page->index to unmap */
737 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
738 	unsigned long truncate_count;		/* Compare vm_truncate_count */
739 };
740 
741 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
742 		pte_t pte);
743 
744 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
745 		unsigned long size);
746 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
747 		unsigned long size, struct zap_details *);
748 unsigned long unmap_vmas(struct mmu_gather **tlb,
749 		struct vm_area_struct *start_vma, unsigned long start_addr,
750 		unsigned long end_addr, unsigned long *nr_accounted,
751 		struct zap_details *);
752 
753 /**
754  * mm_walk - callbacks for walk_page_range
755  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
756  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
757  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
758  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
759  * @pte_hole: if set, called for each hole at all levels
760  *
761  * (see walk_page_range for more details)
762  */
763 struct mm_walk {
764 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
765 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
766 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
767 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
768 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
769 	struct mm_struct *mm;
770 	void *private;
771 };
772 
773 int walk_page_range(unsigned long addr, unsigned long end,
774 		struct mm_walk *walk);
775 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
776 		unsigned long end, unsigned long floor, unsigned long ceiling);
777 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
778 			struct vm_area_struct *vma);
779 void unmap_mapping_range(struct address_space *mapping,
780 		loff_t const holebegin, loff_t const holelen, int even_cows);
781 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
782 	unsigned long *pfn);
783 int follow_phys(struct vm_area_struct *vma, unsigned long address,
784 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
785 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
786 			void *buf, int len, int write);
787 
788 static inline void unmap_shared_mapping_range(struct address_space *mapping,
789 		loff_t const holebegin, loff_t const holelen)
790 {
791 	unmap_mapping_range(mapping, holebegin, holelen, 0);
792 }
793 
794 extern int vmtruncate(struct inode * inode, loff_t offset);
795 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
796 
797 #ifdef CONFIG_MMU
798 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
799 			unsigned long address, unsigned int flags);
800 #else
801 static inline int handle_mm_fault(struct mm_struct *mm,
802 			struct vm_area_struct *vma, unsigned long address,
803 			unsigned int flags)
804 {
805 	/* should never happen if there's no MMU */
806 	BUG();
807 	return VM_FAULT_SIGBUS;
808 }
809 #endif
810 
811 extern int make_pages_present(unsigned long addr, unsigned long end);
812 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
813 
814 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
815 			unsigned long start, int nr_pages, int write, int force,
816 			struct page **pages, struct vm_area_struct **vmas);
817 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
818 			struct page **pages);
819 
820 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
821 extern void do_invalidatepage(struct page *page, unsigned long offset);
822 
823 int __set_page_dirty_nobuffers(struct page *page);
824 int __set_page_dirty_no_writeback(struct page *page);
825 int redirty_page_for_writepage(struct writeback_control *wbc,
826 				struct page *page);
827 void account_page_dirtied(struct page *page, struct address_space *mapping);
828 int set_page_dirty(struct page *page);
829 int set_page_dirty_lock(struct page *page);
830 int clear_page_dirty_for_io(struct page *page);
831 
832 extern unsigned long move_page_tables(struct vm_area_struct *vma,
833 		unsigned long old_addr, struct vm_area_struct *new_vma,
834 		unsigned long new_addr, unsigned long len);
835 extern unsigned long do_mremap(unsigned long addr,
836 			       unsigned long old_len, unsigned long new_len,
837 			       unsigned long flags, unsigned long new_addr);
838 extern int mprotect_fixup(struct vm_area_struct *vma,
839 			  struct vm_area_struct **pprev, unsigned long start,
840 			  unsigned long end, unsigned long newflags);
841 
842 /*
843  * doesn't attempt to fault and will return short.
844  */
845 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
846 			  struct page **pages);
847 
848 /*
849  * A callback you can register to apply pressure to ageable caches.
850  *
851  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
852  * look through the least-recently-used 'nr_to_scan' entries and
853  * attempt to free them up.  It should return the number of objects
854  * which remain in the cache.  If it returns -1, it means it cannot do
855  * any scanning at this time (eg. there is a risk of deadlock).
856  *
857  * The 'gfpmask' refers to the allocation we are currently trying to
858  * fulfil.
859  *
860  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
861  * querying the cache size, so a fastpath for that case is appropriate.
862  */
863 struct shrinker {
864 	int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
865 	int seeks;	/* seeks to recreate an obj */
866 
867 	/* These are for internal use */
868 	struct list_head list;
869 	long nr;	/* objs pending delete */
870 };
871 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
872 extern void register_shrinker(struct shrinker *);
873 extern void unregister_shrinker(struct shrinker *);
874 
875 int vma_wants_writenotify(struct vm_area_struct *vma);
876 
877 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
878 
879 #ifdef __PAGETABLE_PUD_FOLDED
880 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
881 						unsigned long address)
882 {
883 	return 0;
884 }
885 #else
886 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
887 #endif
888 
889 #ifdef __PAGETABLE_PMD_FOLDED
890 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
891 						unsigned long address)
892 {
893 	return 0;
894 }
895 #else
896 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
897 #endif
898 
899 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
900 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
901 
902 /*
903  * The following ifdef needed to get the 4level-fixup.h header to work.
904  * Remove it when 4level-fixup.h has been removed.
905  */
906 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
907 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
908 {
909 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
910 		NULL: pud_offset(pgd, address);
911 }
912 
913 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
914 {
915 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
916 		NULL: pmd_offset(pud, address);
917 }
918 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
919 
920 #if USE_SPLIT_PTLOCKS
921 /*
922  * We tuck a spinlock to guard each pagetable page into its struct page,
923  * at page->private, with BUILD_BUG_ON to make sure that this will not
924  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
925  * When freeing, reset page->mapping so free_pages_check won't complain.
926  */
927 #define __pte_lockptr(page)	&((page)->ptl)
928 #define pte_lock_init(_page)	do {					\
929 	spin_lock_init(__pte_lockptr(_page));				\
930 } while (0)
931 #define pte_lock_deinit(page)	((page)->mapping = NULL)
932 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
933 #else	/* !USE_SPLIT_PTLOCKS */
934 /*
935  * We use mm->page_table_lock to guard all pagetable pages of the mm.
936  */
937 #define pte_lock_init(page)	do {} while (0)
938 #define pte_lock_deinit(page)	do {} while (0)
939 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
940 #endif /* USE_SPLIT_PTLOCKS */
941 
942 static inline void pgtable_page_ctor(struct page *page)
943 {
944 	pte_lock_init(page);
945 	inc_zone_page_state(page, NR_PAGETABLE);
946 }
947 
948 static inline void pgtable_page_dtor(struct page *page)
949 {
950 	pte_lock_deinit(page);
951 	dec_zone_page_state(page, NR_PAGETABLE);
952 }
953 
954 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
955 ({							\
956 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
957 	pte_t *__pte = pte_offset_map(pmd, address);	\
958 	*(ptlp) = __ptl;				\
959 	spin_lock(__ptl);				\
960 	__pte;						\
961 })
962 
963 #define pte_unmap_unlock(pte, ptl)	do {		\
964 	spin_unlock(ptl);				\
965 	pte_unmap(pte);					\
966 } while (0)
967 
968 #define pte_alloc_map(mm, pmd, address)			\
969 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
970 		NULL: pte_offset_map(pmd, address))
971 
972 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
973 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
974 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
975 
976 #define pte_alloc_kernel(pmd, address)			\
977 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
978 		NULL: pte_offset_kernel(pmd, address))
979 
980 extern void free_area_init(unsigned long * zones_size);
981 extern void free_area_init_node(int nid, unsigned long * zones_size,
982 		unsigned long zone_start_pfn, unsigned long *zholes_size);
983 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
984 /*
985  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
986  * zones, allocate the backing mem_map and account for memory holes in a more
987  * architecture independent manner. This is a substitute for creating the
988  * zone_sizes[] and zholes_size[] arrays and passing them to
989  * free_area_init_node()
990  *
991  * An architecture is expected to register range of page frames backed by
992  * physical memory with add_active_range() before calling
993  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
994  * usage, an architecture is expected to do something like
995  *
996  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
997  * 							 max_highmem_pfn};
998  * for_each_valid_physical_page_range()
999  * 	add_active_range(node_id, start_pfn, end_pfn)
1000  * free_area_init_nodes(max_zone_pfns);
1001  *
1002  * If the architecture guarantees that there are no holes in the ranges
1003  * registered with add_active_range(), free_bootmem_active_regions()
1004  * will call free_bootmem_node() for each registered physical page range.
1005  * Similarly sparse_memory_present_with_active_regions() calls
1006  * memory_present() for each range when SPARSEMEM is enabled.
1007  *
1008  * See mm/page_alloc.c for more information on each function exposed by
1009  * CONFIG_ARCH_POPULATES_NODE_MAP
1010  */
1011 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1012 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1013 					unsigned long end_pfn);
1014 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1015 					unsigned long end_pfn);
1016 extern void remove_all_active_ranges(void);
1017 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1018 						unsigned long end_pfn);
1019 extern void get_pfn_range_for_nid(unsigned int nid,
1020 			unsigned long *start_pfn, unsigned long *end_pfn);
1021 extern unsigned long find_min_pfn_with_active_regions(void);
1022 extern void free_bootmem_with_active_regions(int nid,
1023 						unsigned long max_low_pfn);
1024 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1025 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1026 extern void sparse_memory_present_with_active_regions(int nid);
1027 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1028 
1029 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1030     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1031 static inline int __early_pfn_to_nid(unsigned long pfn)
1032 {
1033 	return 0;
1034 }
1035 #else
1036 /* please see mm/page_alloc.c */
1037 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1038 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1039 /* there is a per-arch backend function. */
1040 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1041 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1042 #endif
1043 
1044 extern void set_dma_reserve(unsigned long new_dma_reserve);
1045 extern void memmap_init_zone(unsigned long, int, unsigned long,
1046 				unsigned long, enum memmap_context);
1047 extern void setup_per_zone_wmarks(void);
1048 extern void calculate_zone_inactive_ratio(struct zone *zone);
1049 extern void mem_init(void);
1050 extern void __init mmap_init(void);
1051 extern void show_mem(void);
1052 extern void si_meminfo(struct sysinfo * val);
1053 extern void si_meminfo_node(struct sysinfo *val, int nid);
1054 extern int after_bootmem;
1055 
1056 #ifdef CONFIG_NUMA
1057 extern void setup_per_cpu_pageset(void);
1058 #else
1059 static inline void setup_per_cpu_pageset(void) {}
1060 #endif
1061 
1062 extern void zone_pcp_update(struct zone *zone);
1063 
1064 /* nommu.c */
1065 extern atomic_long_t mmap_pages_allocated;
1066 
1067 /* prio_tree.c */
1068 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1069 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1070 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1071 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1072 	struct prio_tree_iter *iter);
1073 
1074 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1075 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1076 		(vma = vma_prio_tree_next(vma, iter)); )
1077 
1078 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1079 					struct list_head *list)
1080 {
1081 	vma->shared.vm_set.parent = NULL;
1082 	list_add_tail(&vma->shared.vm_set.list, list);
1083 }
1084 
1085 /* mmap.c */
1086 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1087 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1088 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1089 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1090 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1091 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1092 	struct mempolicy *);
1093 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1094 extern int split_vma(struct mm_struct *,
1095 	struct vm_area_struct *, unsigned long addr, int new_below);
1096 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1097 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1098 	struct rb_node **, struct rb_node *);
1099 extern void unlink_file_vma(struct vm_area_struct *);
1100 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1101 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1102 extern void exit_mmap(struct mm_struct *);
1103 
1104 extern int mm_take_all_locks(struct mm_struct *mm);
1105 extern void mm_drop_all_locks(struct mm_struct *mm);
1106 
1107 #ifdef CONFIG_PROC_FS
1108 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1109 extern void added_exe_file_vma(struct mm_struct *mm);
1110 extern void removed_exe_file_vma(struct mm_struct *mm);
1111 #else
1112 static inline void added_exe_file_vma(struct mm_struct *mm)
1113 {}
1114 
1115 static inline void removed_exe_file_vma(struct mm_struct *mm)
1116 {}
1117 #endif /* CONFIG_PROC_FS */
1118 
1119 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1120 extern int install_special_mapping(struct mm_struct *mm,
1121 				   unsigned long addr, unsigned long len,
1122 				   unsigned long flags, struct page **pages);
1123 
1124 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1125 
1126 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1127 	unsigned long len, unsigned long prot,
1128 	unsigned long flag, unsigned long pgoff);
1129 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1130 	unsigned long len, unsigned long flags,
1131 	unsigned int vm_flags, unsigned long pgoff);
1132 
1133 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1134 	unsigned long len, unsigned long prot,
1135 	unsigned long flag, unsigned long offset)
1136 {
1137 	unsigned long ret = -EINVAL;
1138 	if ((offset + PAGE_ALIGN(len)) < offset)
1139 		goto out;
1140 	if (!(offset & ~PAGE_MASK))
1141 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1142 out:
1143 	return ret;
1144 }
1145 
1146 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1147 
1148 extern unsigned long do_brk(unsigned long, unsigned long);
1149 
1150 /* filemap.c */
1151 extern unsigned long page_unuse(struct page *);
1152 extern void truncate_inode_pages(struct address_space *, loff_t);
1153 extern void truncate_inode_pages_range(struct address_space *,
1154 				       loff_t lstart, loff_t lend);
1155 
1156 /* generic vm_area_ops exported for stackable file systems */
1157 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1158 
1159 /* mm/page-writeback.c */
1160 int write_one_page(struct page *page, int wait);
1161 void task_dirty_inc(struct task_struct *tsk);
1162 
1163 /* readahead.c */
1164 #define VM_MAX_READAHEAD	128	/* kbytes */
1165 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1166 
1167 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1168 			pgoff_t offset, unsigned long nr_to_read);
1169 
1170 void page_cache_sync_readahead(struct address_space *mapping,
1171 			       struct file_ra_state *ra,
1172 			       struct file *filp,
1173 			       pgoff_t offset,
1174 			       unsigned long size);
1175 
1176 void page_cache_async_readahead(struct address_space *mapping,
1177 				struct file_ra_state *ra,
1178 				struct file *filp,
1179 				struct page *pg,
1180 				pgoff_t offset,
1181 				unsigned long size);
1182 
1183 unsigned long max_sane_readahead(unsigned long nr);
1184 unsigned long ra_submit(struct file_ra_state *ra,
1185 			struct address_space *mapping,
1186 			struct file *filp);
1187 
1188 /* Do stack extension */
1189 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1190 #ifdef CONFIG_IA64
1191 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1192 #endif
1193 extern int expand_stack_downwards(struct vm_area_struct *vma,
1194 				  unsigned long address);
1195 
1196 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1197 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1198 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1199 					     struct vm_area_struct **pprev);
1200 
1201 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1202    NULL if none.  Assume start_addr < end_addr. */
1203 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1204 {
1205 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1206 
1207 	if (vma && end_addr <= vma->vm_start)
1208 		vma = NULL;
1209 	return vma;
1210 }
1211 
1212 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1213 {
1214 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1215 }
1216 
1217 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1218 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1219 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1220 			unsigned long pfn, unsigned long size, pgprot_t);
1221 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1222 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1223 			unsigned long pfn);
1224 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1225 			unsigned long pfn);
1226 
1227 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1228 			unsigned int foll_flags);
1229 #define FOLL_WRITE	0x01	/* check pte is writable */
1230 #define FOLL_TOUCH	0x02	/* mark page accessed */
1231 #define FOLL_GET	0x04	/* do get_page on page */
1232 #define FOLL_ANON	0x08	/* give ZERO_PAGE if no pgtable */
1233 
1234 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1235 			void *data);
1236 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1237 			       unsigned long size, pte_fn_t fn, void *data);
1238 
1239 #ifdef CONFIG_PROC_FS
1240 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1241 #else
1242 static inline void vm_stat_account(struct mm_struct *mm,
1243 			unsigned long flags, struct file *file, long pages)
1244 {
1245 }
1246 #endif /* CONFIG_PROC_FS */
1247 
1248 #ifdef CONFIG_DEBUG_PAGEALLOC
1249 extern int debug_pagealloc_enabled;
1250 
1251 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1252 
1253 static inline void enable_debug_pagealloc(void)
1254 {
1255 	debug_pagealloc_enabled = 1;
1256 }
1257 #ifdef CONFIG_HIBERNATION
1258 extern bool kernel_page_present(struct page *page);
1259 #endif /* CONFIG_HIBERNATION */
1260 #else
1261 static inline void
1262 kernel_map_pages(struct page *page, int numpages, int enable) {}
1263 static inline void enable_debug_pagealloc(void)
1264 {
1265 }
1266 #ifdef CONFIG_HIBERNATION
1267 static inline bool kernel_page_present(struct page *page) { return true; }
1268 #endif /* CONFIG_HIBERNATION */
1269 #endif
1270 
1271 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1272 #ifdef	__HAVE_ARCH_GATE_AREA
1273 int in_gate_area_no_task(unsigned long addr);
1274 int in_gate_area(struct task_struct *task, unsigned long addr);
1275 #else
1276 int in_gate_area_no_task(unsigned long addr);
1277 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1278 #endif	/* __HAVE_ARCH_GATE_AREA */
1279 
1280 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1281 					void __user *, size_t *, loff_t *);
1282 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1283 			unsigned long lru_pages);
1284 
1285 #ifndef CONFIG_MMU
1286 #define randomize_va_space 0
1287 #else
1288 extern int randomize_va_space;
1289 #endif
1290 
1291 const char * arch_vma_name(struct vm_area_struct *vma);
1292 void print_vma_addr(char *prefix, unsigned long rip);
1293 
1294 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1295 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1296 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1297 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1298 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1299 void *vmemmap_alloc_block(unsigned long size, int node);
1300 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1301 int vmemmap_populate_basepages(struct page *start_page,
1302 						unsigned long pages, int node);
1303 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1304 void vmemmap_populate_print_last(void);
1305 
1306 extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1307 				 size_t size);
1308 extern void refund_locked_memory(struct mm_struct *mm, size_t size);
1309 #endif /* __KERNEL__ */
1310 #endif /* _LINUX_MM_H */
1311