huge page private reservation review cleanups
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / include / linux / mm.h
CommitLineData
1da177e4
LT
1#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
1da177e4
LT
4#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
1da177e4
LT
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>
9a11b49a 13#include <linux/debug_locks.h>
5b99cd0e 14#include <linux/mm_types.h>
1da177e4
LT
15
16struct mempolicy;
17struct anon_vma;
4e950f6f 18struct file_ra_state;
e8edc6e0 19struct user_struct;
4e950f6f 20struct writeback_control;
1da177e4
LT
21
22#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23extern unsigned long max_mapnr;
24#endif
25
26extern unsigned long num_physpages;
27extern void * high_memory;
1da177e4
LT
28extern int page_cluster;
29
30#ifdef CONFIG_SYSCTL
31extern int sysctl_legacy_va_layout;
32#else
33#define sysctl_legacy_va_layout 0
34#endif
35
42d7896e
JM
36extern unsigned long mmap_min_addr;
37
1da177e4
LT
38#include <asm/page.h>
39#include <asm/pgtable.h>
40#include <asm/processor.h>
1da177e4 41
1da177e4
LT
42#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
43
44/*
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
50 * mmap() functions).
51 */
52
c43692e8
CL
53extern struct kmem_cache *vm_area_cachep;
54
1da177e4
LT
55/*
56 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
57 * disabled, then there's a single shared list of VMAs maintained by the
58 * system, and mm's subscribe to these individually
59 */
60struct vm_list_struct {
61 struct vm_list_struct *next;
62 struct vm_area_struct *vma;
63};
64
65#ifndef CONFIG_MMU
66extern struct rb_root nommu_vma_tree;
67extern struct rw_semaphore nommu_vma_sem;
68
69extern unsigned int kobjsize(const void *objp);
70#endif
71
72/*
73 * vm_flags..
74 */
75#define VM_READ 0x00000001 /* currently active flags */
76#define VM_WRITE 0x00000002
77#define VM_EXEC 0x00000004
78#define VM_SHARED 0x00000008
79
7e2cff42 80/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
1da177e4
LT
81#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82#define VM_MAYWRITE 0x00000020
83#define VM_MAYEXEC 0x00000040
84#define VM_MAYSHARE 0x00000080
85
86#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87#define VM_GROWSUP 0x00000200
6aab341e 88#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
1da177e4
LT
89#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90
91#define VM_EXECUTABLE 0x00001000
92#define VM_LOCKED 0x00002000
93#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94
95 /* Used by sys_madvise() */
96#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98
99#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
0b14c179 101#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
1da177e4
LT
102#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
104#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
105#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
4d7672b4 106#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
e5b97dde 107#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
d00806b1 108
d0217ac0 109#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
b379d790 110#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
aba46c50 111#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
1da177e4
LT
112
113#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
114#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
115#endif
116
117#ifdef CONFIG_STACK_GROWSUP
118#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
119#else
120#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
121#endif
122
123#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
124#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
125#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
126#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
127#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
128
129/*
130 * mapping from the currently active vm_flags protection bits (the
131 * low four bits) to a page protection mask..
132 */
133extern pgprot_t protection_map[16];
134
d0217ac0
NP
135#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
136#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
137
138
54cb8821 139/*
d0217ac0 140 * vm_fault is filled by the the pagefault handler and passed to the vma's
83c54070
NP
141 * ->fault function. The vma's ->fault is responsible for returning a bitmask
142 * of VM_FAULT_xxx flags that give details about how the fault was handled.
54cb8821 143 *
d0217ac0
NP
144 * pgoff should be used in favour of virtual_address, if possible. If pgoff
145 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
146 * mapping support.
54cb8821 147 */
d0217ac0
NP
148struct vm_fault {
149 unsigned int flags; /* FAULT_FLAG_xxx flags */
150 pgoff_t pgoff; /* Logical page offset based on vma */
151 void __user *virtual_address; /* Faulting virtual address */
152
153 struct page *page; /* ->fault handlers should return a
83c54070 154 * page here, unless VM_FAULT_NOPAGE
d0217ac0 155 * is set (which is also implied by
83c54070 156 * VM_FAULT_ERROR).
d0217ac0 157 */
54cb8821 158};
1da177e4
LT
159
160/*
161 * These are the virtual MM functions - opening of an area, closing and
162 * unmapping it (needed to keep files on disk up-to-date etc), pointer
163 * to the functions called when a no-page or a wp-page exception occurs.
164 */
165struct vm_operations_struct {
166 void (*open)(struct vm_area_struct * area);
167 void (*close)(struct vm_area_struct * area);
d0217ac0 168 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
9637a5ef
DH
169
170 /* notification that a previously read-only page is about to become
171 * writable, if an error is returned it will cause a SIGBUS */
172 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
28b2ee20
RR
173
174 /* called by access_process_vm when get_user_pages() fails, typically
175 * for use by special VMAs that can switch between memory and hardware
176 */
177 int (*access)(struct vm_area_struct *vma, unsigned long addr,
178 void *buf, int len, int write);
1da177e4 179#ifdef CONFIG_NUMA
a6020ed7
LS
180 /*
181 * set_policy() op must add a reference to any non-NULL @new mempolicy
182 * to hold the policy upon return. Caller should pass NULL @new to
183 * remove a policy and fall back to surrounding context--i.e. do not
184 * install a MPOL_DEFAULT policy, nor the task or system default
185 * mempolicy.
186 */
1da177e4 187 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
a6020ed7
LS
188
189 /*
190 * get_policy() op must add reference [mpol_get()] to any policy at
191 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
192 * in mm/mempolicy.c will do this automatically.
193 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
194 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
195 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
196 * must return NULL--i.e., do not "fallback" to task or system default
197 * policy.
198 */
1da177e4
LT
199 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
200 unsigned long addr);
7b2259b3
CL
201 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
202 const nodemask_t *to, unsigned long flags);
1da177e4
LT
203#endif
204};
205
206struct mmu_gather;
207struct inode;
208
349aef0b
AM
209#define page_private(page) ((page)->private)
210#define set_page_private(page, v) ((page)->private = (v))
4c21e2f2 211
1da177e4
LT
212/*
213 * FIXME: take this include out, include page-flags.h in
214 * files which need it (119 of them)
215 */
216#include <linux/page-flags.h>
217
725d704e
NP
218#ifdef CONFIG_DEBUG_VM
219#define VM_BUG_ON(cond) BUG_ON(cond)
220#else
221#define VM_BUG_ON(condition) do { } while(0)
222#endif
223
1da177e4
LT
224/*
225 * Methods to modify the page usage count.
226 *
227 * What counts for a page usage:
228 * - cache mapping (page->mapping)
229 * - private data (page->private)
230 * - page mapped in a task's page tables, each mapping
231 * is counted separately
232 *
233 * Also, many kernel routines increase the page count before a critical
234 * routine so they can be sure the page doesn't go away from under them.
1da177e4
LT
235 */
236
237/*
da6052f7 238 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1da177e4 239 */
7c8ee9a8
NP
240static inline int put_page_testzero(struct page *page)
241{
725d704e 242 VM_BUG_ON(atomic_read(&page->_count) == 0);
8dc04efb 243 return atomic_dec_and_test(&page->_count);
7c8ee9a8 244}
1da177e4
LT
245
246/*
7c8ee9a8
NP
247 * Try to grab a ref unless the page has a refcount of zero, return false if
248 * that is the case.
1da177e4 249 */
7c8ee9a8
NP
250static inline int get_page_unless_zero(struct page *page)
251{
aec2c3ed 252 VM_BUG_ON(PageTail(page));
8dc04efb 253 return atomic_inc_not_zero(&page->_count);
7c8ee9a8 254}
1da177e4 255
48667e7a 256/* Support for virtually mapped pages */
b3bdda02
CL
257struct page *vmalloc_to_page(const void *addr);
258unsigned long vmalloc_to_pfn(const void *addr);
48667e7a 259
0738c4bb
PM
260/*
261 * Determine if an address is within the vmalloc range
262 *
263 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
264 * is no special casing required.
265 */
9e2779fa
CL
266static inline int is_vmalloc_addr(const void *x)
267{
0738c4bb 268#ifdef CONFIG_MMU
9e2779fa
CL
269 unsigned long addr = (unsigned long)x;
270
271 return addr >= VMALLOC_START && addr < VMALLOC_END;
0738c4bb
PM
272#else
273 return 0;
8ca3ed87 274#endif
0738c4bb 275}
9e2779fa 276
d85f3385
CL
277static inline struct page *compound_head(struct page *page)
278{
6d777953 279 if (unlikely(PageTail(page)))
d85f3385
CL
280 return page->first_page;
281 return page;
282}
283
4c21e2f2 284static inline int page_count(struct page *page)
1da177e4 285{
d85f3385 286 return atomic_read(&compound_head(page)->_count);
1da177e4
LT
287}
288
289static inline void get_page(struct page *page)
290{
d85f3385 291 page = compound_head(page);
725d704e 292 VM_BUG_ON(atomic_read(&page->_count) == 0);
1da177e4
LT
293 atomic_inc(&page->_count);
294}
295
b49af68f
CL
296static inline struct page *virt_to_head_page(const void *x)
297{
298 struct page *page = virt_to_page(x);
299 return compound_head(page);
300}
301
7835e98b
NP
302/*
303 * Setup the page count before being freed into the page allocator for
304 * the first time (boot or memory hotplug)
305 */
306static inline void init_page_count(struct page *page)
307{
308 atomic_set(&page->_count, 1);
309}
310
1da177e4 311void put_page(struct page *page);
1d7ea732 312void put_pages_list(struct list_head *pages);
1da177e4 313
8dfcc9ba 314void split_page(struct page *page, unsigned int order);
8dfcc9ba 315
33f2ef89
AW
316/*
317 * Compound pages have a destructor function. Provide a
318 * prototype for that function and accessor functions.
319 * These are _only_ valid on the head of a PG_compound page.
320 */
321typedef void compound_page_dtor(struct page *);
322
323static inline void set_compound_page_dtor(struct page *page,
324 compound_page_dtor *dtor)
325{
326 page[1].lru.next = (void *)dtor;
327}
328
329static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
330{
331 return (compound_page_dtor *)page[1].lru.next;
332}
333
d85f3385
CL
334static inline int compound_order(struct page *page)
335{
6d777953 336 if (!PageHead(page))
d85f3385
CL
337 return 0;
338 return (unsigned long)page[1].lru.prev;
339}
340
341static inline void set_compound_order(struct page *page, unsigned long order)
342{
343 page[1].lru.prev = (void *)order;
344}
345
1da177e4
LT
346/*
347 * Multiple processes may "see" the same page. E.g. for untouched
348 * mappings of /dev/null, all processes see the same page full of
349 * zeroes, and text pages of executables and shared libraries have
350 * only one copy in memory, at most, normally.
351 *
352 * For the non-reserved pages, page_count(page) denotes a reference count.
7e871b6c
PBG
353 * page_count() == 0 means the page is free. page->lru is then used for
354 * freelist management in the buddy allocator.
da6052f7 355 * page_count() > 0 means the page has been allocated.
1da177e4 356 *
da6052f7
NP
357 * Pages are allocated by the slab allocator in order to provide memory
358 * to kmalloc and kmem_cache_alloc. In this case, the management of the
359 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
360 * unless a particular usage is carefully commented. (the responsibility of
361 * freeing the kmalloc memory is the caller's, of course).
1da177e4 362 *
da6052f7
NP
363 * A page may be used by anyone else who does a __get_free_page().
364 * In this case, page_count still tracks the references, and should only
365 * be used through the normal accessor functions. The top bits of page->flags
366 * and page->virtual store page management information, but all other fields
367 * are unused and could be used privately, carefully. The management of this
368 * page is the responsibility of the one who allocated it, and those who have
369 * subsequently been given references to it.
370 *
371 * The other pages (we may call them "pagecache pages") are completely
1da177e4
LT
372 * managed by the Linux memory manager: I/O, buffers, swapping etc.
373 * The following discussion applies only to them.
374 *
da6052f7
NP
375 * A pagecache page contains an opaque `private' member, which belongs to the
376 * page's address_space. Usually, this is the address of a circular list of
377 * the page's disk buffers. PG_private must be set to tell the VM to call
378 * into the filesystem to release these pages.
1da177e4 379 *
da6052f7
NP
380 * A page may belong to an inode's memory mapping. In this case, page->mapping
381 * is the pointer to the inode, and page->index is the file offset of the page,
382 * in units of PAGE_CACHE_SIZE.
1da177e4 383 *
da6052f7
NP
384 * If pagecache pages are not associated with an inode, they are said to be
385 * anonymous pages. These may become associated with the swapcache, and in that
386 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1da177e4 387 *
da6052f7
NP
388 * In either case (swapcache or inode backed), the pagecache itself holds one
389 * reference to the page. Setting PG_private should also increment the
390 * refcount. The each user mapping also has a reference to the page.
1da177e4 391 *
da6052f7
NP
392 * The pagecache pages are stored in a per-mapping radix tree, which is
393 * rooted at mapping->page_tree, and indexed by offset.
394 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
395 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1da177e4 396 *
da6052f7 397 * All pagecache pages may be subject to I/O:
1da177e4
LT
398 * - inode pages may need to be read from disk,
399 * - inode pages which have been modified and are MAP_SHARED may need
da6052f7
NP
400 * to be written back to the inode on disk,
401 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
402 * modified may need to be swapped out to swap space and (later) to be read
403 * back into memory.
1da177e4
LT
404 */
405
406/*
407 * The zone field is never updated after free_area_init_core()
408 * sets it, so none of the operations on it need to be atomic.
1da177e4 409 */
348f8b6c 410
d41dee36
AW
411
412/*
413 * page->flags layout:
414 *
415 * There are three possibilities for how page->flags get
416 * laid out. The first is for the normal case, without
417 * sparsemem. The second is for sparsemem when there is
418 * plenty of space for node and section. The last is when
419 * we have run out of space and have to fall back to an
420 * alternate (slower) way of determining the node.
421 *
308c05e3
CL
422 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
423 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
424 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
d41dee36 425 */
308c05e3 426#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
d41dee36
AW
427#define SECTIONS_WIDTH SECTIONS_SHIFT
428#else
429#define SECTIONS_WIDTH 0
430#endif
431
432#define ZONES_WIDTH ZONES_SHIFT
433
9223b419 434#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
d41dee36
AW
435#define NODES_WIDTH NODES_SHIFT
436#else
308c05e3
CL
437#ifdef CONFIG_SPARSEMEM_VMEMMAP
438#error "Vmemmap: No space for nodes field in page flags"
439#endif
d41dee36
AW
440#define NODES_WIDTH 0
441#endif
442
443/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
07808b74 444#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
d41dee36
AW
445#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
446#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
447
448/*
449 * We are going to use the flags for the page to node mapping if its in
450 * there. This includes the case where there is no node, so it is implicit.
451 */
89689ae7
CL
452#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
453#define NODE_NOT_IN_PAGE_FLAGS
454#endif
d41dee36
AW
455
456#ifndef PFN_SECTION_SHIFT
457#define PFN_SECTION_SHIFT 0
458#endif
348f8b6c
DH
459
460/*
461 * Define the bit shifts to access each section. For non-existant
462 * sections we define the shift as 0; that plus a 0 mask ensures
463 * the compiler will optimise away reference to them.
464 */
d41dee36
AW
465#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
466#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
467#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
348f8b6c 468
89689ae7
CL
469/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
470#ifdef NODE_NOT_IN_PAGEFLAGS
471#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
bd8029b6
AW
472#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
473 SECTIONS_PGOFF : ZONES_PGOFF)
d41dee36 474#else
89689ae7 475#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
bd8029b6
AW
476#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
477 NODES_PGOFF : ZONES_PGOFF)
89689ae7
CL
478#endif
479
bd8029b6 480#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
348f8b6c 481
9223b419
CL
482#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
483#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
348f8b6c
DH
484#endif
485
d41dee36
AW
486#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
487#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
488#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
89689ae7 489#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
348f8b6c 490
2f1b6248 491static inline enum zone_type page_zonenum(struct page *page)
1da177e4 492{
348f8b6c 493 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1da177e4 494}
1da177e4 495
89689ae7
CL
496/*
497 * The identification function is only used by the buddy allocator for
498 * determining if two pages could be buddies. We are not really
499 * identifying a zone since we could be using a the section number
500 * id if we have not node id available in page flags.
501 * We guarantee only that it will return the same value for two
502 * combinable pages in a zone.
503 */
cb2b95e1
AW
504static inline int page_zone_id(struct page *page)
505{
89689ae7 506 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
348f8b6c
DH
507}
508
25ba77c1 509static inline int zone_to_nid(struct zone *zone)
89fa3024 510{
d5f541ed
CL
511#ifdef CONFIG_NUMA
512 return zone->node;
513#else
514 return 0;
515#endif
89fa3024
CL
516}
517
89689ae7 518#ifdef NODE_NOT_IN_PAGE_FLAGS
25ba77c1 519extern int page_to_nid(struct page *page);
89689ae7 520#else
25ba77c1 521static inline int page_to_nid(struct page *page)
d41dee36 522{
89689ae7 523 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
d41dee36 524}
89689ae7
CL
525#endif
526
527static inline struct zone *page_zone(struct page *page)
528{
529 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
530}
531
308c05e3 532#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
d41dee36
AW
533static inline unsigned long page_to_section(struct page *page)
534{
535 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
536}
308c05e3 537#endif
d41dee36 538
2f1b6248 539static inline void set_page_zone(struct page *page, enum zone_type zone)
348f8b6c
DH
540{
541 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
542 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
543}
2f1b6248 544
348f8b6c
DH
545static inline void set_page_node(struct page *page, unsigned long node)
546{
547 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
548 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1da177e4 549}
89689ae7 550
d41dee36
AW
551static inline void set_page_section(struct page *page, unsigned long section)
552{
553 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
554 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
555}
1da177e4 556
2f1b6248 557static inline void set_page_links(struct page *page, enum zone_type zone,
d41dee36 558 unsigned long node, unsigned long pfn)
1da177e4 559{
348f8b6c
DH
560 set_page_zone(page, zone);
561 set_page_node(page, node);
d41dee36 562 set_page_section(page, pfn_to_section_nr(pfn));
1da177e4
LT
563}
564
7cd94146
EP
565/*
566 * If a hint addr is less than mmap_min_addr change hint to be as
567 * low as possible but still greater than mmap_min_addr
568 */
569static inline unsigned long round_hint_to_min(unsigned long hint)
570{
571#ifdef CONFIG_SECURITY
572 hint &= PAGE_MASK;
573 if (((void *)hint != NULL) &&
574 (hint < mmap_min_addr))
575 return PAGE_ALIGN(mmap_min_addr);
576#endif
577 return hint;
578}
579
f6ac2354
CL
580/*
581 * Some inline functions in vmstat.h depend on page_zone()
582 */
583#include <linux/vmstat.h>
584
652050ae 585static __always_inline void *lowmem_page_address(struct page *page)
1da177e4
LT
586{
587 return __va(page_to_pfn(page) << PAGE_SHIFT);
588}
589
590#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
591#define HASHED_PAGE_VIRTUAL
592#endif
593
594#if defined(WANT_PAGE_VIRTUAL)
595#define page_address(page) ((page)->virtual)
596#define set_page_address(page, address) \
597 do { \
598 (page)->virtual = (address); \
599 } while(0)
600#define page_address_init() do { } while(0)
601#endif
602
603#if defined(HASHED_PAGE_VIRTUAL)
604void *page_address(struct page *page);
605void set_page_address(struct page *page, void *virtual);
606void page_address_init(void);
607#endif
608
609#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
610#define page_address(page) lowmem_page_address(page)
611#define set_page_address(page, address) do { } while(0)
612#define page_address_init() do { } while(0)
613#endif
614
615/*
616 * On an anonymous page mapped into a user virtual memory area,
617 * page->mapping points to its anon_vma, not to a struct address_space;
618 * with the PAGE_MAPPING_ANON bit set to distinguish it.
619 *
620 * Please note that, confusingly, "page_mapping" refers to the inode
621 * address_space which maps the page from disk; whereas "page_mapped"
622 * refers to user virtual address space into which the page is mapped.
623 */
624#define PAGE_MAPPING_ANON 1
625
626extern struct address_space swapper_space;
627static inline struct address_space *page_mapping(struct page *page)
628{
629 struct address_space *mapping = page->mapping;
630
b5fab14e 631 VM_BUG_ON(PageSlab(page));
726b8012 632#ifdef CONFIG_SWAP
1da177e4
LT
633 if (unlikely(PageSwapCache(page)))
634 mapping = &swapper_space;
726b8012
AM
635 else
636#endif
637 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
1da177e4
LT
638 mapping = NULL;
639 return mapping;
640}
641
642static inline int PageAnon(struct page *page)
643{
644 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
645}
646
647/*
648 * Return the pagecache index of the passed page. Regular pagecache pages
649 * use ->index whereas swapcache pages use ->private
650 */
651static inline pgoff_t page_index(struct page *page)
652{
653 if (unlikely(PageSwapCache(page)))
4c21e2f2 654 return page_private(page);
1da177e4
LT
655 return page->index;
656}
657
658/*
659 * The atomic page->_mapcount, like _count, starts from -1:
660 * so that transitions both from it and to it can be tracked,
661 * using atomic_inc_and_test and atomic_add_negative(-1).
662 */
663static inline void reset_page_mapcount(struct page *page)
664{
665 atomic_set(&(page)->_mapcount, -1);
666}
667
668static inline int page_mapcount(struct page *page)
669{
670 return atomic_read(&(page)->_mapcount) + 1;
671}
672
673/*
674 * Return true if this page is mapped into pagetables.
675 */
676static inline int page_mapped(struct page *page)
677{
678 return atomic_read(&(page)->_mapcount) >= 0;
679}
680
1da177e4
LT
681/*
682 * Different kinds of faults, as returned by handle_mm_fault().
683 * Used to decide whether a process gets delivered SIGBUS or
684 * just gets major/minor fault counters bumped up.
685 */
d0217ac0 686
83c54070 687#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
d0217ac0 688
83c54070
NP
689#define VM_FAULT_OOM 0x0001
690#define VM_FAULT_SIGBUS 0x0002
691#define VM_FAULT_MAJOR 0x0004
692#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
f33ea7f4 693
83c54070
NP
694#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
695#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1da177e4 696
83c54070 697#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
d0217ac0 698
1da177e4
LT
699#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
700
701extern void show_free_areas(void);
702
703#ifdef CONFIG_SHMEM
1da177e4
LT
704int shmem_lock(struct file *file, int lock, struct user_struct *user);
705#else
03b00ebc
RK
706static inline int shmem_lock(struct file *file, int lock,
707 struct user_struct *user)
708{
709 return 0;
710}
1da177e4
LT
711#endif
712struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
713
714int shmem_zero_setup(struct vm_area_struct *);
715
b0e15190
DH
716#ifndef CONFIG_MMU
717extern 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
e8edc6e0 724extern int can_do_mlock(void);
1da177e4
LT
725extern int user_shm_lock(size_t, struct user_struct *);
726extern void user_shm_unlock(size_t, struct user_struct *);
727
728/*
729 * Parameter block passed down to zap_pte_range in exceptional cases.
730 */
731struct 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: */
1da177e4
LT
737 unsigned long truncate_count; /* Compare vm_truncate_count */
738};
739
7e675137
NP
740struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
741 pte_t pte);
742
ee39b37b 743unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
1da177e4 744 unsigned long size, struct zap_details *);
508034a3 745unsigned long unmap_vmas(struct mmu_gather **tlb,
1da177e4
LT
746 struct vm_area_struct *start_vma, unsigned long start_addr,
747 unsigned long end_addr, unsigned long *nr_accounted,
748 struct zap_details *);
e6473092
MM
749
750/**
751 * mm_walk - callbacks for walk_page_range
752 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
753 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
754 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
755 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
756 * @pte_hole: if set, called for each hole at all levels
757 *
758 * (see walk_page_range for more details)
759 */
760struct mm_walk {
2165009b
DH
761 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
762 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
763 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
764 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
765 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
766 struct mm_struct *mm;
767 void *private;
e6473092
MM
768};
769
2165009b
DH
770int walk_page_range(unsigned long addr, unsigned long end,
771 struct mm_walk *walk);
42b77728 772void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
3bf5ee95 773 unsigned long end, unsigned long floor, unsigned long ceiling);
1da177e4
LT
774int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
775 struct vm_area_struct *vma);
1da177e4
LT
776void unmap_mapping_range(struct address_space *mapping,
777 loff_t const holebegin, loff_t const holelen, int even_cows);
28b2ee20
RR
778int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
779 void *buf, int len, int write);
1da177e4
LT
780
781static inline void unmap_shared_mapping_range(struct address_space *mapping,
782 loff_t const holebegin, loff_t const holelen)
783{
784 unmap_mapping_range(mapping, holebegin, holelen, 0);
785}
786
787extern int vmtruncate(struct inode * inode, loff_t offset);
f6b3ec23 788extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
f33ea7f4 789
7ee1dd3f 790#ifdef CONFIG_MMU
83c54070 791extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
7ee1dd3f 792 unsigned long address, int write_access);
7ee1dd3f
DH
793#else
794static inline int handle_mm_fault(struct mm_struct *mm,
795 struct vm_area_struct *vma, unsigned long address,
796 int write_access)
797{
798 /* should never happen if there's no MMU */
799 BUG();
800 return VM_FAULT_SIGBUS;
801}
802#endif
f33ea7f4 803
1da177e4
LT
804extern int make_pages_present(unsigned long addr, unsigned long end);
805extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1da177e4
LT
806
807int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
808 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
b5810039 809void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
1da177e4 810
cf9a2ae8
DH
811extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
812extern void do_invalidatepage(struct page *page, unsigned long offset);
813
1da177e4 814int __set_page_dirty_nobuffers(struct page *page);
76719325 815int __set_page_dirty_no_writeback(struct page *page);
1da177e4
LT
816int redirty_page_for_writepage(struct writeback_control *wbc,
817 struct page *page);
b3c97528 818int set_page_dirty(struct page *page);
1da177e4
LT
819int set_page_dirty_lock(struct page *page);
820int clear_page_dirty_for_io(struct page *page);
821
b6a2fea3
OW
822extern unsigned long move_page_tables(struct vm_area_struct *vma,
823 unsigned long old_addr, struct vm_area_struct *new_vma,
824 unsigned long new_addr, unsigned long len);
1da177e4
LT
825extern unsigned long do_mremap(unsigned long addr,
826 unsigned long old_len, unsigned long new_len,
827 unsigned long flags, unsigned long new_addr);
b6a2fea3
OW
828extern int mprotect_fixup(struct vm_area_struct *vma,
829 struct vm_area_struct **pprev, unsigned long start,
830 unsigned long end, unsigned long newflags);
1da177e4
LT
831
832/*
8e1f936b 833 * A callback you can register to apply pressure to ageable caches.
1da177e4 834 *
8e1f936b
RR
835 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
836 * look through the least-recently-used 'nr_to_scan' entries and
837 * attempt to free them up. It should return the number of objects
838 * which remain in the cache. If it returns -1, it means it cannot do
839 * any scanning at this time (eg. there is a risk of deadlock).
1da177e4 840 *
8e1f936b
RR
841 * The 'gfpmask' refers to the allocation we are currently trying to
842 * fulfil.
843 *
844 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
845 * querying the cache size, so a fastpath for that case is appropriate.
1da177e4 846 */
8e1f936b
RR
847struct shrinker {
848 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
849 int seeks; /* seeks to recreate an obj */
1da177e4 850
8e1f936b
RR
851 /* These are for internal use */
852 struct list_head list;
853 long nr; /* objs pending delete */
854};
855#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
856extern void register_shrinker(struct shrinker *);
857extern void unregister_shrinker(struct shrinker *);
1da177e4 858
4e950f6f 859int vma_wants_writenotify(struct vm_area_struct *vma);
d08b3851 860
b3c97528 861extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
c9cfcddf 862
5f22df00
NP
863#ifdef __PAGETABLE_PUD_FOLDED
864static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
865 unsigned long address)
866{
867 return 0;
868}
869#else
1bb3630e 870int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
5f22df00
NP
871#endif
872
873#ifdef __PAGETABLE_PMD_FOLDED
874static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
875 unsigned long address)
876{
877 return 0;
878}
879#else
1bb3630e 880int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
5f22df00
NP
881#endif
882
1bb3630e
HD
883int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
884int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
885
1da177e4
LT
886/*
887 * The following ifdef needed to get the 4level-fixup.h header to work.
888 * Remove it when 4level-fixup.h has been removed.
889 */
1bb3630e 890#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1da177e4
LT
891static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
892{
1bb3630e
HD
893 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
894 NULL: pud_offset(pgd, address);
1da177e4
LT
895}
896
897static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
898{
1bb3630e
HD
899 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
900 NULL: pmd_offset(pud, address);
1da177e4 901}
1bb3630e
HD
902#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
903
4c21e2f2
HD
904#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
905/*
906 * We tuck a spinlock to guard each pagetable page into its struct page,
907 * at page->private, with BUILD_BUG_ON to make sure that this will not
908 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
909 * When freeing, reset page->mapping so free_pages_check won't complain.
910 */
349aef0b 911#define __pte_lockptr(page) &((page)->ptl)
4c21e2f2
HD
912#define pte_lock_init(_page) do { \
913 spin_lock_init(__pte_lockptr(_page)); \
914} while (0)
915#define pte_lock_deinit(page) ((page)->mapping = NULL)
916#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
917#else
918/*
919 * We use mm->page_table_lock to guard all pagetable pages of the mm.
920 */
921#define pte_lock_init(page) do {} while (0)
922#define pte_lock_deinit(page) do {} while (0)
923#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
924#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
925
2f569afd
MS
926static inline void pgtable_page_ctor(struct page *page)
927{
928 pte_lock_init(page);
929 inc_zone_page_state(page, NR_PAGETABLE);
930}
931
932static inline void pgtable_page_dtor(struct page *page)
933{
934 pte_lock_deinit(page);
935 dec_zone_page_state(page, NR_PAGETABLE);
936}
937
c74df32c
HD
938#define pte_offset_map_lock(mm, pmd, address, ptlp) \
939({ \
4c21e2f2 940 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
c74df32c
HD
941 pte_t *__pte = pte_offset_map(pmd, address); \
942 *(ptlp) = __ptl; \
943 spin_lock(__ptl); \
944 __pte; \
945})
946
947#define pte_unmap_unlock(pte, ptl) do { \
948 spin_unlock(ptl); \
949 pte_unmap(pte); \
950} while (0)
951
1bb3630e
HD
952#define pte_alloc_map(mm, pmd, address) \
953 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
954 NULL: pte_offset_map(pmd, address))
955
c74df32c
HD
956#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
957 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
958 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
959
1bb3630e
HD
960#define pte_alloc_kernel(pmd, address) \
961 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
962 NULL: pte_offset_kernel(pmd, address))
1da177e4
LT
963
964extern void free_area_init(unsigned long * zones_size);
9109fb7b
JW
965extern void free_area_init_node(int nid, unsigned long * zones_size,
966 unsigned long zone_start_pfn, unsigned long *zholes_size);
c713216d
MG
967#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
968/*
969 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
970 * zones, allocate the backing mem_map and account for memory holes in a more
971 * architecture independent manner. This is a substitute for creating the
972 * zone_sizes[] and zholes_size[] arrays and passing them to
973 * free_area_init_node()
974 *
975 * An architecture is expected to register range of page frames backed by
976 * physical memory with add_active_range() before calling
977 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
978 * usage, an architecture is expected to do something like
979 *
980 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
981 * max_highmem_pfn};
982 * for_each_valid_physical_page_range()
983 * add_active_range(node_id, start_pfn, end_pfn)
984 * free_area_init_nodes(max_zone_pfns);
985 *
986 * If the architecture guarantees that there are no holes in the ranges
987 * registered with add_active_range(), free_bootmem_active_regions()
988 * will call free_bootmem_node() for each registered physical page range.
989 * Similarly sparse_memory_present_with_active_regions() calls
990 * memory_present() for each range when SPARSEMEM is enabled.
991 *
992 * See mm/page_alloc.c for more information on each function exposed by
993 * CONFIG_ARCH_POPULATES_NODE_MAP
994 */
995extern void free_area_init_nodes(unsigned long *max_zone_pfn);
996extern void add_active_range(unsigned int nid, unsigned long start_pfn,
997 unsigned long end_pfn);
cc1050ba
YL
998extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
999 unsigned long end_pfn);
fb01439c
MG
1000extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1001 unsigned long end_pfn);
c713216d
MG
1002extern void remove_all_active_ranges(void);
1003extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1004 unsigned long end_pfn);
1005extern void get_pfn_range_for_nid(unsigned int nid,
1006 unsigned long *start_pfn, unsigned long *end_pfn);
1007extern unsigned long find_min_pfn_with_active_regions(void);
1008extern unsigned long find_max_pfn_with_active_regions(void);
1009extern void free_bootmem_with_active_regions(int nid,
1010 unsigned long max_low_pfn);
d52d53b8 1011typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
b5bc6c0e 1012extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
c713216d
MG
1013extern void sparse_memory_present_with_active_regions(int nid);
1014#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1015extern int early_pfn_to_nid(unsigned long pfn);
1016#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1017#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
0e0b864e 1018extern void set_dma_reserve(unsigned long new_dma_reserve);
a2f3aa02
DH
1019extern void memmap_init_zone(unsigned long, int, unsigned long,
1020 unsigned long, enum memmap_context);
3947be19 1021extern void setup_per_zone_pages_min(void);
1da177e4
LT
1022extern void mem_init(void);
1023extern void show_mem(void);
1024extern void si_meminfo(struct sysinfo * val);
1025extern void si_meminfo_node(struct sysinfo *val, int nid);
3461b0af 1026extern int after_bootmem;
1da177e4 1027
e7c8d5c9
CL
1028#ifdef CONFIG_NUMA
1029extern void setup_per_cpu_pageset(void);
1030#else
1031static inline void setup_per_cpu_pageset(void) {}
1032#endif
1033
1da177e4
LT
1034/* prio_tree.c */
1035void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1036void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1037void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1038struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1039 struct prio_tree_iter *iter);
1040
1041#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1042 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1043 (vma = vma_prio_tree_next(vma, iter)); )
1044
1045static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1046 struct list_head *list)
1047{
1048 vma->shared.vm_set.parent = NULL;
1049 list_add_tail(&vma->shared.vm_set.list, list);
1050}
1051
1052/* mmap.c */
34b4e4aa 1053extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1da177e4
LT
1054extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1055 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1056extern struct vm_area_struct *vma_merge(struct mm_struct *,
1057 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1058 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1059 struct mempolicy *);
1060extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1061extern int split_vma(struct mm_struct *,
1062 struct vm_area_struct *, unsigned long addr, int new_below);
1063extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1064extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1065 struct rb_node **, struct rb_node *);
a8fb5618 1066extern void unlink_file_vma(struct vm_area_struct *);
1da177e4
LT
1067extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1068 unsigned long addr, unsigned long len, pgoff_t pgoff);
1069extern void exit_mmap(struct mm_struct *);
925d1c40
MH
1070
1071#ifdef CONFIG_PROC_FS
1072/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1073extern void added_exe_file_vma(struct mm_struct *mm);
1074extern void removed_exe_file_vma(struct mm_struct *mm);
1075#else
1076static inline void added_exe_file_vma(struct mm_struct *mm)
1077{}
1078
1079static inline void removed_exe_file_vma(struct mm_struct *mm)
1080{}
1081#endif /* CONFIG_PROC_FS */
1082
119f657c 1083extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
fa5dc22f
RM
1084extern int install_special_mapping(struct mm_struct *mm,
1085 unsigned long addr, unsigned long len,
1086 unsigned long flags, struct page **pages);
1da177e4
LT
1087
1088extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1089
1090extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1091 unsigned long len, unsigned long prot,
1092 unsigned long flag, unsigned long pgoff);
0165ab44
MS
1093extern unsigned long mmap_region(struct file *file, unsigned long addr,
1094 unsigned long len, unsigned long flags,
1095 unsigned int vm_flags, unsigned long pgoff,
1096 int accountable);
1da177e4
LT
1097
1098static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1099 unsigned long len, unsigned long prot,
1100 unsigned long flag, unsigned long offset)
1101{
1102 unsigned long ret = -EINVAL;
1103 if ((offset + PAGE_ALIGN(len)) < offset)
1104 goto out;
1105 if (!(offset & ~PAGE_MASK))
1106 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1107out:
1108 return ret;
1109}
1110
1111extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1112
1113extern unsigned long do_brk(unsigned long, unsigned long);
1114
1115/* filemap.c */
1116extern unsigned long page_unuse(struct page *);
1117extern void truncate_inode_pages(struct address_space *, loff_t);
d7339071
HR
1118extern void truncate_inode_pages_range(struct address_space *,
1119 loff_t lstart, loff_t lend);
1da177e4
LT
1120
1121/* generic vm_area_ops exported for stackable file systems */
d0217ac0 1122extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1da177e4
LT
1123
1124/* mm/page-writeback.c */
1125int write_one_page(struct page *page, int wait);
1126
1127/* readahead.c */
1128#define VM_MAX_READAHEAD 128 /* kbytes */
1129#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1da177e4
LT
1130
1131int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
7361f4d8 1132 pgoff_t offset, unsigned long nr_to_read);
1da177e4 1133int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
7361f4d8 1134 pgoff_t offset, unsigned long nr_to_read);
cf914a7d
RR
1135
1136void page_cache_sync_readahead(struct address_space *mapping,
1137 struct file_ra_state *ra,
1138 struct file *filp,
1139 pgoff_t offset,
1140 unsigned long size);
1141
1142void page_cache_async_readahead(struct address_space *mapping,
1143 struct file_ra_state *ra,
1144 struct file *filp,
1145 struct page *pg,
1146 pgoff_t offset,
1147 unsigned long size);
1148
1da177e4
LT
1149unsigned long max_sane_readahead(unsigned long nr);
1150
1151/* Do stack extension */
46dea3d0 1152extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
9ab88515 1153#ifdef CONFIG_IA64
46dea3d0 1154extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
9ab88515 1155#endif
b6a2fea3
OW
1156extern int expand_stack_downwards(struct vm_area_struct *vma,
1157 unsigned long address);
1da177e4
LT
1158
1159/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1160extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1161extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1162 struct vm_area_struct **pprev);
1163
1164/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1165 NULL if none. Assume start_addr < end_addr. */
1166static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1167{
1168 struct vm_area_struct * vma = find_vma(mm,start_addr);
1169
1170 if (vma && end_addr <= vma->vm_start)
1171 vma = NULL;
1172 return vma;
1173}
1174
1175static inline unsigned long vma_pages(struct vm_area_struct *vma)
1176{
1177 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1178}
1179
804af2cf 1180pgprot_t vm_get_page_prot(unsigned long vm_flags);
deceb6cd 1181struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
deceb6cd
HD
1182int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1183 unsigned long pfn, unsigned long size, pgprot_t);
a145dd41 1184int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
e0dc0d8f
NP
1185int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1186 unsigned long pfn);
423bad60
NP
1187int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1188 unsigned long pfn);
deceb6cd 1189
6aab341e 1190struct page *follow_page(struct vm_area_struct *, unsigned long address,
deceb6cd
HD
1191 unsigned int foll_flags);
1192#define FOLL_WRITE 0x01 /* check pte is writable */
1193#define FOLL_TOUCH 0x02 /* mark page accessed */
1194#define FOLL_GET 0x04 /* do get_page on page */
1195#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1da177e4 1196
2f569afd 1197typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
aee16b3c
JF
1198 void *data);
1199extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1200 unsigned long size, pte_fn_t fn, void *data);
1201
1da177e4 1202#ifdef CONFIG_PROC_FS
ab50b8ed 1203void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1da177e4 1204#else
ab50b8ed 1205static inline void vm_stat_account(struct mm_struct *mm,
1da177e4
LT
1206 unsigned long flags, struct file *file, long pages)
1207{
1208}
1209#endif /* CONFIG_PROC_FS */
1210
12d6f21e
IM
1211#ifdef CONFIG_DEBUG_PAGEALLOC
1212extern int debug_pagealloc_enabled;
1213
1214extern void kernel_map_pages(struct page *page, int numpages, int enable);
1215
1216static inline void enable_debug_pagealloc(void)
1217{
1218 debug_pagealloc_enabled = 1;
1219}
8a235efa
RW
1220#ifdef CONFIG_HIBERNATION
1221extern bool kernel_page_present(struct page *page);
1222#endif /* CONFIG_HIBERNATION */
12d6f21e 1223#else
1da177e4 1224static inline void
9858db50 1225kernel_map_pages(struct page *page, int numpages, int enable) {}
12d6f21e
IM
1226static inline void enable_debug_pagealloc(void)
1227{
1228}
8a235efa
RW
1229#ifdef CONFIG_HIBERNATION
1230static inline bool kernel_page_present(struct page *page) { return true; }
1231#endif /* CONFIG_HIBERNATION */
1da177e4
LT
1232#endif
1233
1234extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1235#ifdef __HAVE_ARCH_GATE_AREA
1236int in_gate_area_no_task(unsigned long addr);
1237int in_gate_area(struct task_struct *task, unsigned long addr);
1238#else
1239int in_gate_area_no_task(unsigned long addr);
1240#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1241#endif /* __HAVE_ARCH_GATE_AREA */
1242
9d0243bc
AM
1243int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1244 void __user *, size_t *, loff_t *);
69e05944 1245unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
9d0243bc 1246 unsigned long lru_pages);
9d0243bc 1247
7a9166e3
LY
1248#ifndef CONFIG_MMU
1249#define randomize_va_space 0
1250#else
a62eaf15 1251extern int randomize_va_space;
7a9166e3 1252#endif
a62eaf15 1253
045e72ac 1254const char * arch_vma_name(struct vm_area_struct *vma);
03252919 1255void print_vma_addr(char *prefix, unsigned long rip);
e6e5494c 1256
98f3cfc1 1257struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
29c71111
AW
1258pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1259pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1260pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1261pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
8f6aac41
CL
1262void *vmemmap_alloc_block(unsigned long size, int node);
1263void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
29c71111
AW
1264int vmemmap_populate_basepages(struct page *start_page,
1265 unsigned long pages, int node);
1266int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
c2b91e2e 1267void vmemmap_populate_print_last(void);
8f6aac41 1268
1da177e4
LT
1269#endif /* __KERNEL__ */
1270#endif /* _LINUX_MM_H */