xref: /linux-6.15/include/linux/mm.h (revision e79d5747)
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 
107 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
108 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
109 #endif
110 
111 #ifdef CONFIG_STACK_GROWSUP
112 #define VM_STACK_FLAGS	(VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
113 #else
114 #define VM_STACK_FLAGS	(VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
115 #endif
116 
117 #define VM_READHINTMASK			(VM_SEQ_READ | VM_RAND_READ)
118 #define VM_ClearReadHint(v)		(v)->vm_flags &= ~VM_READHINTMASK
119 #define VM_NormalReadHint(v)		(!((v)->vm_flags & VM_READHINTMASK))
120 #define VM_SequentialReadHint(v)	((v)->vm_flags & VM_SEQ_READ)
121 #define VM_RandomReadHint(v)		((v)->vm_flags & VM_RAND_READ)
122 
123 /*
124  * special vmas that are non-mergable, non-mlock()able
125  */
126 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
127 
128 /*
129  * mapping from the currently active vm_flags protection bits (the
130  * low four bits) to a page protection mask..
131  */
132 extern pgprot_t protection_map[16];
133 
134 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
135 #define FAULT_FLAG_NONLINEAR	0x02	/* Fault was via a nonlinear mapping */
136 #define FAULT_FLAG_MKWRITE	0x04	/* Fault was mkwrite of existing pte */
137 
138 /*
139  * This interface is used by x86 PAT code to identify a pfn mapping that is
140  * linear over entire vma. This is to optimize PAT code that deals with
141  * marking the physical region with a particular prot. This is not for generic
142  * mm use. Note also that this check will not work if the pfn mapping is
143  * linear for a vma starting at physical address 0. In which case PAT code
144  * falls back to slow path of reserving physical range page by page.
145  */
146 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
147 {
148 	return (vma->vm_flags & VM_PFN_AT_MMAP);
149 }
150 
151 static inline int is_pfn_mapping(struct vm_area_struct *vma)
152 {
153 	return (vma->vm_flags & VM_PFNMAP);
154 }
155 
156 /*
157  * vm_fault is filled by the the pagefault handler and passed to the vma's
158  * ->fault function. The vma's ->fault is responsible for returning a bitmask
159  * of VM_FAULT_xxx flags that give details about how the fault was handled.
160  *
161  * pgoff should be used in favour of virtual_address, if possible. If pgoff
162  * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
163  * mapping support.
164  */
165 struct vm_fault {
166 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
167 	pgoff_t pgoff;			/* Logical page offset based on vma */
168 	void __user *virtual_address;	/* Faulting virtual address */
169 
170 	struct page *page;		/* ->fault handlers should return a
171 					 * page here, unless VM_FAULT_NOPAGE
172 					 * is set (which is also implied by
173 					 * VM_FAULT_ERROR).
174 					 */
175 };
176 
177 /*
178  * These are the virtual MM functions - opening of an area, closing and
179  * unmapping it (needed to keep files on disk up-to-date etc), pointer
180  * to the functions called when a no-page or a wp-page exception occurs.
181  */
182 struct vm_operations_struct {
183 	void (*open)(struct vm_area_struct * area);
184 	void (*close)(struct vm_area_struct * area);
185 	int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
186 
187 	/* notification that a previously read-only page is about to become
188 	 * writable, if an error is returned it will cause a SIGBUS */
189 	int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
190 
191 	/* called by access_process_vm when get_user_pages() fails, typically
192 	 * for use by special VMAs that can switch between memory and hardware
193 	 */
194 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
195 		      void *buf, int len, int write);
196 #ifdef CONFIG_NUMA
197 	/*
198 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
199 	 * to hold the policy upon return.  Caller should pass NULL @new to
200 	 * remove a policy and fall back to surrounding context--i.e. do not
201 	 * install a MPOL_DEFAULT policy, nor the task or system default
202 	 * mempolicy.
203 	 */
204 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
205 
206 	/*
207 	 * get_policy() op must add reference [mpol_get()] to any policy at
208 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
209 	 * in mm/mempolicy.c will do this automatically.
210 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
211 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
212 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
213 	 * must return NULL--i.e., do not "fallback" to task or system default
214 	 * policy.
215 	 */
216 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
217 					unsigned long addr);
218 	int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
219 		const nodemask_t *to, unsigned long flags);
220 #endif
221 };
222 
223 struct mmu_gather;
224 struct inode;
225 
226 #define page_private(page)		((page)->private)
227 #define set_page_private(page, v)	((page)->private = (v))
228 
229 /*
230  * FIXME: take this include out, include page-flags.h in
231  * files which need it (119 of them)
232  */
233 #include <linux/page-flags.h>
234 
235 /*
236  * Methods to modify the page usage count.
237  *
238  * What counts for a page usage:
239  * - cache mapping   (page->mapping)
240  * - private data    (page->private)
241  * - page mapped in a task's page tables, each mapping
242  *   is counted separately
243  *
244  * Also, many kernel routines increase the page count before a critical
245  * routine so they can be sure the page doesn't go away from under them.
246  */
247 
248 /*
249  * Drop a ref, return true if the refcount fell to zero (the page has no users)
250  */
251 static inline int put_page_testzero(struct page *page)
252 {
253 	VM_BUG_ON(atomic_read(&page->_count) == 0);
254 	return atomic_dec_and_test(&page->_count);
255 }
256 
257 /*
258  * Try to grab a ref unless the page has a refcount of zero, return false if
259  * that is the case.
260  */
261 static inline int get_page_unless_zero(struct page *page)
262 {
263 	return atomic_inc_not_zero(&page->_count);
264 }
265 
266 /* Support for virtually mapped pages */
267 struct page *vmalloc_to_page(const void *addr);
268 unsigned long vmalloc_to_pfn(const void *addr);
269 
270 /*
271  * Determine if an address is within the vmalloc range
272  *
273  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
274  * is no special casing required.
275  */
276 static inline int is_vmalloc_addr(const void *x)
277 {
278 #ifdef CONFIG_MMU
279 	unsigned long addr = (unsigned long)x;
280 
281 	return addr >= VMALLOC_START && addr < VMALLOC_END;
282 #else
283 	return 0;
284 #endif
285 }
286 
287 static inline struct page *compound_head(struct page *page)
288 {
289 	if (unlikely(PageTail(page)))
290 		return page->first_page;
291 	return page;
292 }
293 
294 static inline int page_count(struct page *page)
295 {
296 	return atomic_read(&compound_head(page)->_count);
297 }
298 
299 static inline void get_page(struct page *page)
300 {
301 	page = compound_head(page);
302 	VM_BUG_ON(atomic_read(&page->_count) == 0);
303 	atomic_inc(&page->_count);
304 }
305 
306 static inline struct page *virt_to_head_page(const void *x)
307 {
308 	struct page *page = virt_to_page(x);
309 	return compound_head(page);
310 }
311 
312 /*
313  * Setup the page count before being freed into the page allocator for
314  * the first time (boot or memory hotplug)
315  */
316 static inline void init_page_count(struct page *page)
317 {
318 	atomic_set(&page->_count, 1);
319 }
320 
321 void put_page(struct page *page);
322 void put_pages_list(struct list_head *pages);
323 
324 void split_page(struct page *page, unsigned int order);
325 
326 /*
327  * Compound pages have a destructor function.  Provide a
328  * prototype for that function and accessor functions.
329  * These are _only_ valid on the head of a PG_compound page.
330  */
331 typedef void compound_page_dtor(struct page *);
332 
333 static inline void set_compound_page_dtor(struct page *page,
334 						compound_page_dtor *dtor)
335 {
336 	page[1].lru.next = (void *)dtor;
337 }
338 
339 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
340 {
341 	return (compound_page_dtor *)page[1].lru.next;
342 }
343 
344 static inline int compound_order(struct page *page)
345 {
346 	if (!PageHead(page))
347 		return 0;
348 	return (unsigned long)page[1].lru.prev;
349 }
350 
351 static inline void set_compound_order(struct page *page, unsigned long order)
352 {
353 	page[1].lru.prev = (void *)order;
354 }
355 
356 /*
357  * Multiple processes may "see" the same page. E.g. for untouched
358  * mappings of /dev/null, all processes see the same page full of
359  * zeroes, and text pages of executables and shared libraries have
360  * only one copy in memory, at most, normally.
361  *
362  * For the non-reserved pages, page_count(page) denotes a reference count.
363  *   page_count() == 0 means the page is free. page->lru is then used for
364  *   freelist management in the buddy allocator.
365  *   page_count() > 0  means the page has been allocated.
366  *
367  * Pages are allocated by the slab allocator in order to provide memory
368  * to kmalloc and kmem_cache_alloc. In this case, the management of the
369  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
370  * unless a particular usage is carefully commented. (the responsibility of
371  * freeing the kmalloc memory is the caller's, of course).
372  *
373  * A page may be used by anyone else who does a __get_free_page().
374  * In this case, page_count still tracks the references, and should only
375  * be used through the normal accessor functions. The top bits of page->flags
376  * and page->virtual store page management information, but all other fields
377  * are unused and could be used privately, carefully. The management of this
378  * page is the responsibility of the one who allocated it, and those who have
379  * subsequently been given references to it.
380  *
381  * The other pages (we may call them "pagecache pages") are completely
382  * managed by the Linux memory manager: I/O, buffers, swapping etc.
383  * The following discussion applies only to them.
384  *
385  * A pagecache page contains an opaque `private' member, which belongs to the
386  * page's address_space. Usually, this is the address of a circular list of
387  * the page's disk buffers. PG_private must be set to tell the VM to call
388  * into the filesystem to release these pages.
389  *
390  * A page may belong to an inode's memory mapping. In this case, page->mapping
391  * is the pointer to the inode, and page->index is the file offset of the page,
392  * in units of PAGE_CACHE_SIZE.
393  *
394  * If pagecache pages are not associated with an inode, they are said to be
395  * anonymous pages. These may become associated with the swapcache, and in that
396  * case PG_swapcache is set, and page->private is an offset into the swapcache.
397  *
398  * In either case (swapcache or inode backed), the pagecache itself holds one
399  * reference to the page. Setting PG_private should also increment the
400  * refcount. The each user mapping also has a reference to the page.
401  *
402  * The pagecache pages are stored in a per-mapping radix tree, which is
403  * rooted at mapping->page_tree, and indexed by offset.
404  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
405  * lists, we instead now tag pages as dirty/writeback in the radix tree.
406  *
407  * All pagecache pages may be subject to I/O:
408  * - inode pages may need to be read from disk,
409  * - inode pages which have been modified and are MAP_SHARED may need
410  *   to be written back to the inode on disk,
411  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
412  *   modified may need to be swapped out to swap space and (later) to be read
413  *   back into memory.
414  */
415 
416 /*
417  * The zone field is never updated after free_area_init_core()
418  * sets it, so none of the operations on it need to be atomic.
419  */
420 
421 
422 /*
423  * page->flags layout:
424  *
425  * There are three possibilities for how page->flags get
426  * laid out.  The first is for the normal case, without
427  * sparsemem.  The second is for sparsemem when there is
428  * plenty of space for node and section.  The last is when
429  * we have run out of space and have to fall back to an
430  * alternate (slower) way of determining the node.
431  *
432  * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
433  * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
434  * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
435  */
436 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
437 #define SECTIONS_WIDTH		SECTIONS_SHIFT
438 #else
439 #define SECTIONS_WIDTH		0
440 #endif
441 
442 #define ZONES_WIDTH		ZONES_SHIFT
443 
444 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
445 #define NODES_WIDTH		NODES_SHIFT
446 #else
447 #ifdef CONFIG_SPARSEMEM_VMEMMAP
448 #error "Vmemmap: No space for nodes field in page flags"
449 #endif
450 #define NODES_WIDTH		0
451 #endif
452 
453 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
454 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
455 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
456 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
457 
458 /*
459  * We are going to use the flags for the page to node mapping if its in
460  * there.  This includes the case where there is no node, so it is implicit.
461  */
462 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
463 #define NODE_NOT_IN_PAGE_FLAGS
464 #endif
465 
466 #ifndef PFN_SECTION_SHIFT
467 #define PFN_SECTION_SHIFT 0
468 #endif
469 
470 /*
471  * Define the bit shifts to access each section.  For non-existant
472  * sections we define the shift as 0; that plus a 0 mask ensures
473  * the compiler will optimise away reference to them.
474  */
475 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
476 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
477 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
478 
479 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
480 #ifdef NODE_NOT_IN_PAGEFLAGS
481 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
482 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
483 						SECTIONS_PGOFF : ZONES_PGOFF)
484 #else
485 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
486 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
487 						NODES_PGOFF : ZONES_PGOFF)
488 #endif
489 
490 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
491 
492 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
493 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
494 #endif
495 
496 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
497 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
498 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
499 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
500 
501 static inline enum zone_type page_zonenum(struct page *page)
502 {
503 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
504 }
505 
506 /*
507  * The identification function is only used by the buddy allocator for
508  * determining if two pages could be buddies. We are not really
509  * identifying a zone since we could be using a the section number
510  * id if we have not node id available in page flags.
511  * We guarantee only that it will return the same value for two
512  * combinable pages in a zone.
513  */
514 static inline int page_zone_id(struct page *page)
515 {
516 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
517 }
518 
519 static inline int zone_to_nid(struct zone *zone)
520 {
521 #ifdef CONFIG_NUMA
522 	return zone->node;
523 #else
524 	return 0;
525 #endif
526 }
527 
528 #ifdef NODE_NOT_IN_PAGE_FLAGS
529 extern int page_to_nid(struct page *page);
530 #else
531 static inline int page_to_nid(struct page *page)
532 {
533 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
534 }
535 #endif
536 
537 static inline struct zone *page_zone(struct page *page)
538 {
539 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
540 }
541 
542 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
543 static inline unsigned long page_to_section(struct page *page)
544 {
545 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
546 }
547 #endif
548 
549 static inline void set_page_zone(struct page *page, enum zone_type zone)
550 {
551 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
552 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
553 }
554 
555 static inline void set_page_node(struct page *page, unsigned long node)
556 {
557 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
558 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
559 }
560 
561 static inline void set_page_section(struct page *page, unsigned long section)
562 {
563 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
564 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
565 }
566 
567 static inline void set_page_links(struct page *page, enum zone_type zone,
568 	unsigned long node, unsigned long pfn)
569 {
570 	set_page_zone(page, zone);
571 	set_page_node(page, node);
572 	set_page_section(page, pfn_to_section_nr(pfn));
573 }
574 
575 /*
576  * Some inline functions in vmstat.h depend on page_zone()
577  */
578 #include <linux/vmstat.h>
579 
580 static __always_inline void *lowmem_page_address(struct page *page)
581 {
582 	return __va(page_to_pfn(page) << PAGE_SHIFT);
583 }
584 
585 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
586 #define HASHED_PAGE_VIRTUAL
587 #endif
588 
589 #if defined(WANT_PAGE_VIRTUAL)
590 #define page_address(page) ((page)->virtual)
591 #define set_page_address(page, address)			\
592 	do {						\
593 		(page)->virtual = (address);		\
594 	} while(0)
595 #define page_address_init()  do { } while(0)
596 #endif
597 
598 #if defined(HASHED_PAGE_VIRTUAL)
599 void *page_address(struct page *page);
600 void set_page_address(struct page *page, void *virtual);
601 void page_address_init(void);
602 #endif
603 
604 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
605 #define page_address(page) lowmem_page_address(page)
606 #define set_page_address(page, address)  do { } while(0)
607 #define page_address_init()  do { } while(0)
608 #endif
609 
610 /*
611  * On an anonymous page mapped into a user virtual memory area,
612  * page->mapping points to its anon_vma, not to a struct address_space;
613  * with the PAGE_MAPPING_ANON bit set to distinguish it.
614  *
615  * Please note that, confusingly, "page_mapping" refers to the inode
616  * address_space which maps the page from disk; whereas "page_mapped"
617  * refers to user virtual address space into which the page is mapped.
618  */
619 #define PAGE_MAPPING_ANON	1
620 
621 extern struct address_space swapper_space;
622 static inline struct address_space *page_mapping(struct page *page)
623 {
624 	struct address_space *mapping = page->mapping;
625 
626 	VM_BUG_ON(PageSlab(page));
627 #ifdef CONFIG_SWAP
628 	if (unlikely(PageSwapCache(page)))
629 		mapping = &swapper_space;
630 	else
631 #endif
632 	if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
633 		mapping = NULL;
634 	return mapping;
635 }
636 
637 static inline int PageAnon(struct page *page)
638 {
639 	return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
640 }
641 
642 /*
643  * Return the pagecache index of the passed page.  Regular pagecache pages
644  * use ->index whereas swapcache pages use ->private
645  */
646 static inline pgoff_t page_index(struct page *page)
647 {
648 	if (unlikely(PageSwapCache(page)))
649 		return page_private(page);
650 	return page->index;
651 }
652 
653 /*
654  * The atomic page->_mapcount, like _count, starts from -1:
655  * so that transitions both from it and to it can be tracked,
656  * using atomic_inc_and_test and atomic_add_negative(-1).
657  */
658 static inline void reset_page_mapcount(struct page *page)
659 {
660 	atomic_set(&(page)->_mapcount, -1);
661 }
662 
663 static inline int page_mapcount(struct page *page)
664 {
665 	return atomic_read(&(page)->_mapcount) + 1;
666 }
667 
668 /*
669  * Return true if this page is mapped into pagetables.
670  */
671 static inline int page_mapped(struct page *page)
672 {
673 	return atomic_read(&(page)->_mapcount) >= 0;
674 }
675 
676 /*
677  * Different kinds of faults, as returned by handle_mm_fault().
678  * Used to decide whether a process gets delivered SIGBUS or
679  * just gets major/minor fault counters bumped up.
680  */
681 
682 #define VM_FAULT_MINOR	0 /* For backwards compat. Remove me quickly. */
683 
684 #define VM_FAULT_OOM	0x0001
685 #define VM_FAULT_SIGBUS	0x0002
686 #define VM_FAULT_MAJOR	0x0004
687 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
688 
689 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
690 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
691 
692 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS)
693 
694 /*
695  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
696  */
697 extern void pagefault_out_of_memory(void);
698 
699 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
700 
701 extern void show_free_areas(void);
702 
703 #ifdef CONFIG_SHMEM
704 extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
705 #else
706 static inline int shmem_lock(struct file *file, int lock,
707 			    struct user_struct *user)
708 {
709 	return 0;
710 }
711 #endif
712 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
713 
714 int shmem_zero_setup(struct vm_area_struct *);
715 
716 #ifndef CONFIG_MMU
717 extern unsigned long shmem_get_unmapped_area(struct file *file,
718 					     unsigned long addr,
719 					     unsigned long len,
720 					     unsigned long pgoff,
721 					     unsigned long flags);
722 #endif
723 
724 extern int can_do_mlock(void);
725 extern int user_shm_lock(size_t, struct user_struct *);
726 extern void user_shm_unlock(size_t, struct user_struct *);
727 
728 /*
729  * Parameter block passed down to zap_pte_range in exceptional cases.
730  */
731 struct zap_details {
732 	struct vm_area_struct *nonlinear_vma;	/* Check page->index if set */
733 	struct address_space *check_mapping;	/* Check page->mapping if set */
734 	pgoff_t	first_index;			/* Lowest page->index to unmap */
735 	pgoff_t last_index;			/* Highest page->index to unmap */
736 	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
737 	unsigned long truncate_count;		/* Compare vm_truncate_count */
738 };
739 
740 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
741 		pte_t pte);
742 
743 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
744 		unsigned long size);
745 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
746 		unsigned long size, struct zap_details *);
747 unsigned long unmap_vmas(struct mmu_gather **tlb,
748 		struct vm_area_struct *start_vma, unsigned long start_addr,
749 		unsigned long end_addr, unsigned long *nr_accounted,
750 		struct zap_details *);
751 
752 /**
753  * mm_walk - callbacks for walk_page_range
754  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
755  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
756  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
757  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
758  * @pte_hole: if set, called for each hole at all levels
759  *
760  * (see walk_page_range for more details)
761  */
762 struct mm_walk {
763 	int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
764 	int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
765 	int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
766 	int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
767 	int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
768 	struct mm_struct *mm;
769 	void *private;
770 };
771 
772 int walk_page_range(unsigned long addr, unsigned long end,
773 		struct mm_walk *walk);
774 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
775 		unsigned long end, unsigned long floor, unsigned long ceiling);
776 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
777 			struct vm_area_struct *vma);
778 void unmap_mapping_range(struct address_space *mapping,
779 		loff_t const holebegin, loff_t const holelen, int even_cows);
780 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
781 	unsigned long *pfn);
782 int follow_phys(struct vm_area_struct *vma, unsigned long address,
783 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
784 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
785 			void *buf, int len, int write);
786 
787 static inline void unmap_shared_mapping_range(struct address_space *mapping,
788 		loff_t const holebegin, loff_t const holelen)
789 {
790 	unmap_mapping_range(mapping, holebegin, holelen, 0);
791 }
792 
793 extern int vmtruncate(struct inode * inode, loff_t offset);
794 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
795 
796 #ifdef CONFIG_MMU
797 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
798 			unsigned long address, unsigned int flags);
799 #else
800 static inline int handle_mm_fault(struct mm_struct *mm,
801 			struct vm_area_struct *vma, unsigned long address,
802 			unsigned int flags)
803 {
804 	/* should never happen if there's no MMU */
805 	BUG();
806 	return VM_FAULT_SIGBUS;
807 }
808 #endif
809 
810 extern int make_pages_present(unsigned long addr, unsigned long end);
811 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
812 
813 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
814 			unsigned long start, int nr_pages, int write, int force,
815 			struct page **pages, struct vm_area_struct **vmas);
816 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
817 			struct page **pages);
818 
819 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
820 extern void do_invalidatepage(struct page *page, unsigned long offset);
821 
822 int __set_page_dirty_nobuffers(struct page *page);
823 int __set_page_dirty_no_writeback(struct page *page);
824 int redirty_page_for_writepage(struct writeback_control *wbc,
825 				struct page *page);
826 void account_page_dirtied(struct page *page, struct address_space *mapping);
827 int set_page_dirty(struct page *page);
828 int set_page_dirty_lock(struct page *page);
829 int clear_page_dirty_for_io(struct page *page);
830 
831 extern unsigned long move_page_tables(struct vm_area_struct *vma,
832 		unsigned long old_addr, struct vm_area_struct *new_vma,
833 		unsigned long new_addr, unsigned long len);
834 extern unsigned long do_mremap(unsigned long addr,
835 			       unsigned long old_len, unsigned long new_len,
836 			       unsigned long flags, unsigned long new_addr);
837 extern int mprotect_fixup(struct vm_area_struct *vma,
838 			  struct vm_area_struct **pprev, unsigned long start,
839 			  unsigned long end, unsigned long newflags);
840 
841 /*
842  * doesn't attempt to fault and will return short.
843  */
844 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
845 			  struct page **pages);
846 
847 /*
848  * A callback you can register to apply pressure to ageable caches.
849  *
850  * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
851  * look through the least-recently-used 'nr_to_scan' entries and
852  * attempt to free them up.  It should return the number of objects
853  * which remain in the cache.  If it returns -1, it means it cannot do
854  * any scanning at this time (eg. there is a risk of deadlock).
855  *
856  * The 'gfpmask' refers to the allocation we are currently trying to
857  * fulfil.
858  *
859  * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
860  * querying the cache size, so a fastpath for that case is appropriate.
861  */
862 struct shrinker {
863 	int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
864 	int seeks;	/* seeks to recreate an obj */
865 
866 	/* These are for internal use */
867 	struct list_head list;
868 	long nr;	/* objs pending delete */
869 };
870 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
871 extern void register_shrinker(struct shrinker *);
872 extern void unregister_shrinker(struct shrinker *);
873 
874 int vma_wants_writenotify(struct vm_area_struct *vma);
875 
876 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
877 
878 #ifdef __PAGETABLE_PUD_FOLDED
879 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
880 						unsigned long address)
881 {
882 	return 0;
883 }
884 #else
885 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
886 #endif
887 
888 #ifdef __PAGETABLE_PMD_FOLDED
889 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
890 						unsigned long address)
891 {
892 	return 0;
893 }
894 #else
895 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
896 #endif
897 
898 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
899 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
900 
901 /*
902  * The following ifdef needed to get the 4level-fixup.h header to work.
903  * Remove it when 4level-fixup.h has been removed.
904  */
905 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
906 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
907 {
908 	return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
909 		NULL: pud_offset(pgd, address);
910 }
911 
912 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
913 {
914 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
915 		NULL: pmd_offset(pud, address);
916 }
917 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
918 
919 #if USE_SPLIT_PTLOCKS
920 /*
921  * We tuck a spinlock to guard each pagetable page into its struct page,
922  * at page->private, with BUILD_BUG_ON to make sure that this will not
923  * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
924  * When freeing, reset page->mapping so free_pages_check won't complain.
925  */
926 #define __pte_lockptr(page)	&((page)->ptl)
927 #define pte_lock_init(_page)	do {					\
928 	spin_lock_init(__pte_lockptr(_page));				\
929 } while (0)
930 #define pte_lock_deinit(page)	((page)->mapping = NULL)
931 #define pte_lockptr(mm, pmd)	({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
932 #else	/* !USE_SPLIT_PTLOCKS */
933 /*
934  * We use mm->page_table_lock to guard all pagetable pages of the mm.
935  */
936 #define pte_lock_init(page)	do {} while (0)
937 #define pte_lock_deinit(page)	do {} while (0)
938 #define pte_lockptr(mm, pmd)	({(void)(pmd); &(mm)->page_table_lock;})
939 #endif /* USE_SPLIT_PTLOCKS */
940 
941 static inline void pgtable_page_ctor(struct page *page)
942 {
943 	pte_lock_init(page);
944 	inc_zone_page_state(page, NR_PAGETABLE);
945 }
946 
947 static inline void pgtable_page_dtor(struct page *page)
948 {
949 	pte_lock_deinit(page);
950 	dec_zone_page_state(page, NR_PAGETABLE);
951 }
952 
953 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
954 ({							\
955 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
956 	pte_t *__pte = pte_offset_map(pmd, address);	\
957 	*(ptlp) = __ptl;				\
958 	spin_lock(__ptl);				\
959 	__pte;						\
960 })
961 
962 #define pte_unmap_unlock(pte, ptl)	do {		\
963 	spin_unlock(ptl);				\
964 	pte_unmap(pte);					\
965 } while (0)
966 
967 #define pte_alloc_map(mm, pmd, address)			\
968 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
969 		NULL: pte_offset_map(pmd, address))
970 
971 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
972 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
973 		NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
974 
975 #define pte_alloc_kernel(pmd, address)			\
976 	((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
977 		NULL: pte_offset_kernel(pmd, address))
978 
979 extern void free_area_init(unsigned long * zones_size);
980 extern void free_area_init_node(int nid, unsigned long * zones_size,
981 		unsigned long zone_start_pfn, unsigned long *zholes_size);
982 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
983 /*
984  * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
985  * zones, allocate the backing mem_map and account for memory holes in a more
986  * architecture independent manner. This is a substitute for creating the
987  * zone_sizes[] and zholes_size[] arrays and passing them to
988  * free_area_init_node()
989  *
990  * An architecture is expected to register range of page frames backed by
991  * physical memory with add_active_range() before calling
992  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
993  * usage, an architecture is expected to do something like
994  *
995  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
996  * 							 max_highmem_pfn};
997  * for_each_valid_physical_page_range()
998  * 	add_active_range(node_id, start_pfn, end_pfn)
999  * free_area_init_nodes(max_zone_pfns);
1000  *
1001  * If the architecture guarantees that there are no holes in the ranges
1002  * registered with add_active_range(), free_bootmem_active_regions()
1003  * will call free_bootmem_node() for each registered physical page range.
1004  * Similarly sparse_memory_present_with_active_regions() calls
1005  * memory_present() for each range when SPARSEMEM is enabled.
1006  *
1007  * See mm/page_alloc.c for more information on each function exposed by
1008  * CONFIG_ARCH_POPULATES_NODE_MAP
1009  */
1010 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1011 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1012 					unsigned long end_pfn);
1013 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1014 					unsigned long end_pfn);
1015 extern void remove_all_active_ranges(void);
1016 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1017 						unsigned long end_pfn);
1018 extern void get_pfn_range_for_nid(unsigned int nid,
1019 			unsigned long *start_pfn, unsigned long *end_pfn);
1020 extern unsigned long find_min_pfn_with_active_regions(void);
1021 extern void free_bootmem_with_active_regions(int nid,
1022 						unsigned long max_low_pfn);
1023 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1024 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1025 extern void sparse_memory_present_with_active_regions(int nid);
1026 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1027 
1028 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1029     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1030 static inline int __early_pfn_to_nid(unsigned long pfn)
1031 {
1032 	return 0;
1033 }
1034 #else
1035 /* please see mm/page_alloc.c */
1036 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1037 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1038 /* there is a per-arch backend function. */
1039 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1040 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1041 #endif
1042 
1043 extern void set_dma_reserve(unsigned long new_dma_reserve);
1044 extern void memmap_init_zone(unsigned long, int, unsigned long,
1045 				unsigned long, enum memmap_context);
1046 extern void setup_per_zone_wmarks(void);
1047 extern void calculate_zone_inactive_ratio(struct zone *zone);
1048 extern void mem_init(void);
1049 extern void __init mmap_init(void);
1050 extern void show_mem(void);
1051 extern void si_meminfo(struct sysinfo * val);
1052 extern void si_meminfo_node(struct sysinfo *val, int nid);
1053 extern int after_bootmem;
1054 
1055 #ifdef CONFIG_NUMA
1056 extern void setup_per_cpu_pageset(void);
1057 #else
1058 static inline void setup_per_cpu_pageset(void) {}
1059 #endif
1060 
1061 /* nommu.c */
1062 extern atomic_long_t mmap_pages_allocated;
1063 
1064 /* prio_tree.c */
1065 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1066 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1067 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1068 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1069 	struct prio_tree_iter *iter);
1070 
1071 #define vma_prio_tree_foreach(vma, iter, root, begin, end)	\
1072 	for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;	\
1073 		(vma = vma_prio_tree_next(vma, iter)); )
1074 
1075 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1076 					struct list_head *list)
1077 {
1078 	vma->shared.vm_set.parent = NULL;
1079 	list_add_tail(&vma->shared.vm_set.list, list);
1080 }
1081 
1082 /* mmap.c */
1083 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1084 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1085 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1086 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1087 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1088 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1089 	struct mempolicy *);
1090 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1091 extern int split_vma(struct mm_struct *,
1092 	struct vm_area_struct *, unsigned long addr, int new_below);
1093 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1094 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1095 	struct rb_node **, struct rb_node *);
1096 extern void unlink_file_vma(struct vm_area_struct *);
1097 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1098 	unsigned long addr, unsigned long len, pgoff_t pgoff);
1099 extern void exit_mmap(struct mm_struct *);
1100 
1101 extern int mm_take_all_locks(struct mm_struct *mm);
1102 extern void mm_drop_all_locks(struct mm_struct *mm);
1103 
1104 #ifdef CONFIG_PROC_FS
1105 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1106 extern void added_exe_file_vma(struct mm_struct *mm);
1107 extern void removed_exe_file_vma(struct mm_struct *mm);
1108 #else
1109 static inline void added_exe_file_vma(struct mm_struct *mm)
1110 {}
1111 
1112 static inline void removed_exe_file_vma(struct mm_struct *mm)
1113 {}
1114 #endif /* CONFIG_PROC_FS */
1115 
1116 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1117 extern int install_special_mapping(struct mm_struct *mm,
1118 				   unsigned long addr, unsigned long len,
1119 				   unsigned long flags, struct page **pages);
1120 
1121 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1122 
1123 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1124 	unsigned long len, unsigned long prot,
1125 	unsigned long flag, unsigned long pgoff);
1126 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1127 	unsigned long len, unsigned long flags,
1128 	unsigned int vm_flags, unsigned long pgoff);
1129 
1130 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1131 	unsigned long len, unsigned long prot,
1132 	unsigned long flag, unsigned long offset)
1133 {
1134 	unsigned long ret = -EINVAL;
1135 	if ((offset + PAGE_ALIGN(len)) < offset)
1136 		goto out;
1137 	if (!(offset & ~PAGE_MASK))
1138 		ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1139 out:
1140 	return ret;
1141 }
1142 
1143 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1144 
1145 extern unsigned long do_brk(unsigned long, unsigned long);
1146 
1147 /* filemap.c */
1148 extern unsigned long page_unuse(struct page *);
1149 extern void truncate_inode_pages(struct address_space *, loff_t);
1150 extern void truncate_inode_pages_range(struct address_space *,
1151 				       loff_t lstart, loff_t lend);
1152 
1153 /* generic vm_area_ops exported for stackable file systems */
1154 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1155 
1156 /* mm/page-writeback.c */
1157 int write_one_page(struct page *page, int wait);
1158 void task_dirty_inc(struct task_struct *tsk);
1159 
1160 /* readahead.c */
1161 #define VM_MAX_READAHEAD	128	/* kbytes */
1162 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
1163 
1164 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1165 			pgoff_t offset, unsigned long nr_to_read);
1166 
1167 void page_cache_sync_readahead(struct address_space *mapping,
1168 			       struct file_ra_state *ra,
1169 			       struct file *filp,
1170 			       pgoff_t offset,
1171 			       unsigned long size);
1172 
1173 void page_cache_async_readahead(struct address_space *mapping,
1174 				struct file_ra_state *ra,
1175 				struct file *filp,
1176 				struct page *pg,
1177 				pgoff_t offset,
1178 				unsigned long size);
1179 
1180 unsigned long max_sane_readahead(unsigned long nr);
1181 unsigned long ra_submit(struct file_ra_state *ra,
1182 			struct address_space *mapping,
1183 			struct file *filp);
1184 
1185 /* Do stack extension */
1186 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1187 #ifdef CONFIG_IA64
1188 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1189 #endif
1190 extern int expand_stack_downwards(struct vm_area_struct *vma,
1191 				  unsigned long address);
1192 
1193 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1194 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1195 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1196 					     struct vm_area_struct **pprev);
1197 
1198 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1199    NULL if none.  Assume start_addr < end_addr. */
1200 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1201 {
1202 	struct vm_area_struct * vma = find_vma(mm,start_addr);
1203 
1204 	if (vma && end_addr <= vma->vm_start)
1205 		vma = NULL;
1206 	return vma;
1207 }
1208 
1209 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1210 {
1211 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1212 }
1213 
1214 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1215 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1216 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1217 			unsigned long pfn, unsigned long size, pgprot_t);
1218 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1219 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1220 			unsigned long pfn);
1221 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1222 			unsigned long pfn);
1223 
1224 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1225 			unsigned int foll_flags);
1226 #define FOLL_WRITE	0x01	/* check pte is writable */
1227 #define FOLL_TOUCH	0x02	/* mark page accessed */
1228 #define FOLL_GET	0x04	/* do get_page on page */
1229 #define FOLL_ANON	0x08	/* give ZERO_PAGE if no pgtable */
1230 
1231 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1232 			void *data);
1233 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1234 			       unsigned long size, pte_fn_t fn, void *data);
1235 
1236 #ifdef CONFIG_PROC_FS
1237 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1238 #else
1239 static inline void vm_stat_account(struct mm_struct *mm,
1240 			unsigned long flags, struct file *file, long pages)
1241 {
1242 }
1243 #endif /* CONFIG_PROC_FS */
1244 
1245 #ifdef CONFIG_DEBUG_PAGEALLOC
1246 extern int debug_pagealloc_enabled;
1247 
1248 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1249 
1250 static inline void enable_debug_pagealloc(void)
1251 {
1252 	debug_pagealloc_enabled = 1;
1253 }
1254 #ifdef CONFIG_HIBERNATION
1255 extern bool kernel_page_present(struct page *page);
1256 #endif /* CONFIG_HIBERNATION */
1257 #else
1258 static inline void
1259 kernel_map_pages(struct page *page, int numpages, int enable) {}
1260 static inline void enable_debug_pagealloc(void)
1261 {
1262 }
1263 #ifdef CONFIG_HIBERNATION
1264 static inline bool kernel_page_present(struct page *page) { return true; }
1265 #endif /* CONFIG_HIBERNATION */
1266 #endif
1267 
1268 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1269 #ifdef	__HAVE_ARCH_GATE_AREA
1270 int in_gate_area_no_task(unsigned long addr);
1271 int in_gate_area(struct task_struct *task, unsigned long addr);
1272 #else
1273 int in_gate_area_no_task(unsigned long addr);
1274 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1275 #endif	/* __HAVE_ARCH_GATE_AREA */
1276 
1277 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1278 					void __user *, size_t *, loff_t *);
1279 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1280 			unsigned long lru_pages);
1281 
1282 #ifndef CONFIG_MMU
1283 #define randomize_va_space 0
1284 #else
1285 extern int randomize_va_space;
1286 #endif
1287 
1288 const char * arch_vma_name(struct vm_area_struct *vma);
1289 void print_vma_addr(char *prefix, unsigned long rip);
1290 
1291 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1292 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1293 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1294 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1295 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1296 void *vmemmap_alloc_block(unsigned long size, int node);
1297 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1298 int vmemmap_populate_basepages(struct page *start_page,
1299 						unsigned long pages, int node);
1300 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1301 void vmemmap_populate_print_last(void);
1302 
1303 extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1304 				 size_t size);
1305 extern void refund_locked_memory(struct mm_struct *mm, size_t size);
1306 #endif /* __KERNEL__ */
1307 #endif /* _LINUX_MM_H */
1308