capabilities: add descriptions for AUDIT_CONTROL and AUDIT_WRITE
[linux-block.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 8#include <linux/gfp.h>
187f1882 9#include <linux/bug.h>
1da177e4
LT
10#include <linux/list.h>
11#include <linux/mmzone.h>
12#include <linux/rbtree.h>
83aeeada 13#include <linux/atomic.h>
9a11b49a 14#include <linux/debug_locks.h>
5b99cd0e 15#include <linux/mm_types.h>
08677214 16#include <linux/range.h>
c6f6b596 17#include <linux/pfn.h>
e9da73d6 18#include <linux/bit_spinlock.h>
b0d40c92 19#include <linux/shrinker.h>
1da177e4
LT
20
21struct mempolicy;
22struct anon_vma;
bf181b9f 23struct anon_vma_chain;
4e950f6f 24struct file_ra_state;
e8edc6e0 25struct user_struct;
4e950f6f 26struct writeback_control;
1da177e4 27
fccc9987 28#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
1da177e4 29extern unsigned long max_mapnr;
fccc9987
JL
30
31static inline void set_max_mapnr(unsigned long limit)
32{
33 max_mapnr = limit;
34}
35#else
36static inline void set_max_mapnr(unsigned long limit) { }
1da177e4
LT
37#endif
38
4481374c 39extern unsigned long totalram_pages;
1da177e4 40extern void * high_memory;
1da177e4
LT
41extern int page_cluster;
42
43#ifdef CONFIG_SYSCTL
44extern int sysctl_legacy_va_layout;
45#else
46#define sysctl_legacy_va_layout 0
47#endif
48
49#include <asm/page.h>
50#include <asm/pgtable.h>
51#include <asm/processor.h>
1da177e4 52
79442ed1
TC
53#ifndef __pa_symbol
54#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
55#endif
56
c9b1d098 57extern unsigned long sysctl_user_reserve_kbytes;
4eeab4f5 58extern unsigned long sysctl_admin_reserve_kbytes;
c9b1d098 59
1da177e4
LT
60#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
61
27ac792c
AR
62/* to align the pointer to the (next) page boundary */
63#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
64
0fa73b86
AM
65/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
66#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
67
1da177e4
LT
68/*
69 * Linux kernel virtual memory manager primitives.
70 * The idea being to have a "virtual" mm in the same way
71 * we have a virtual fs - giving a cleaner interface to the
72 * mm details, and allowing different kinds of memory mappings
73 * (from shared memory to executable loading to arbitrary
74 * mmap() functions).
75 */
76
c43692e8
CL
77extern struct kmem_cache *vm_area_cachep;
78
1da177e4 79#ifndef CONFIG_MMU
8feae131
DH
80extern struct rb_root nommu_region_tree;
81extern struct rw_semaphore nommu_region_sem;
1da177e4
LT
82
83extern unsigned int kobjsize(const void *objp);
84#endif
85
86/*
605d9288 87 * vm_flags in vm_area_struct, see mm_types.h.
1da177e4 88 */
cc2383ec
KK
89#define VM_NONE 0x00000000
90
1da177e4
LT
91#define VM_READ 0x00000001 /* currently active flags */
92#define VM_WRITE 0x00000002
93#define VM_EXEC 0x00000004
94#define VM_SHARED 0x00000008
95
7e2cff42 96/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
1da177e4
LT
97#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
98#define VM_MAYWRITE 0x00000020
99#define VM_MAYEXEC 0x00000040
100#define VM_MAYSHARE 0x00000080
101
102#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
6aab341e 103#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
1da177e4
LT
104#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
105
1da177e4
LT
106#define VM_LOCKED 0x00002000
107#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
108
109 /* Used by sys_madvise() */
110#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
111#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
112
113#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
114#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
1da177e4 115#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
cdfd4325 116#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
1da177e4
LT
117#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
118#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
cc2383ec 119#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
0103bd16 120#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
d00806b1 121
d9104d1c
CG
122#ifdef CONFIG_MEM_SOFT_DIRTY
123# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
124#else
125# define VM_SOFTDIRTY 0
126#endif
127
b379d790 128#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
cc2383ec
KK
129#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
130#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
f8af4da3 131#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
1da177e4 132
cc2383ec
KK
133#if defined(CONFIG_X86)
134# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
135#elif defined(CONFIG_PPC)
136# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
137#elif defined(CONFIG_PARISC)
138# define VM_GROWSUP VM_ARCH_1
9ca52ed9
JH
139#elif defined(CONFIG_METAG)
140# define VM_GROWSUP VM_ARCH_1
cc2383ec
KK
141#elif defined(CONFIG_IA64)
142# define VM_GROWSUP VM_ARCH_1
143#elif !defined(CONFIG_MMU)
144# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
145#endif
146
147#ifndef VM_GROWSUP
148# define VM_GROWSUP VM_NONE
149#endif
150
a8bef8ff
MG
151/* Bits set in the VMA until the stack is in its final location */
152#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
153
1da177e4
LT
154#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
155#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
156#endif
157
158#ifdef CONFIG_STACK_GROWSUP
159#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
160#else
161#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
162#endif
163
b291f000 164/*
78f11a25
AA
165 * Special vmas that are non-mergable, non-mlock()able.
166 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
b291f000 167 */
314e51b9 168#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
b291f000 169
1da177e4
LT
170/*
171 * mapping from the currently active vm_flags protection bits (the
172 * low four bits) to a page protection mask..
173 */
174extern pgprot_t protection_map[16];
175
d0217ac0
NP
176#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
177#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
c2ec175c 178#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
d065bd81 179#define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
318b275f 180#define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
37b23e05 181#define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
45cac65b 182#define FAULT_FLAG_TRIED 0x40 /* second try */
759496ba 183#define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
d0217ac0 184
54cb8821 185/*
d0217ac0 186 * vm_fault is filled by the the pagefault handler and passed to the vma's
83c54070
NP
187 * ->fault function. The vma's ->fault is responsible for returning a bitmask
188 * of VM_FAULT_xxx flags that give details about how the fault was handled.
54cb8821 189 *
d0217ac0 190 * pgoff should be used in favour of virtual_address, if possible. If pgoff
0b173bc4 191 * is used, one may implement ->remap_pages to get nonlinear mapping support.
54cb8821 192 */
d0217ac0
NP
193struct vm_fault {
194 unsigned int flags; /* FAULT_FLAG_xxx flags */
195 pgoff_t pgoff; /* Logical page offset based on vma */
196 void __user *virtual_address; /* Faulting virtual address */
197
198 struct page *page; /* ->fault handlers should return a
83c54070 199 * page here, unless VM_FAULT_NOPAGE
d0217ac0 200 * is set (which is also implied by
83c54070 201 * VM_FAULT_ERROR).
d0217ac0 202 */
54cb8821 203};
1da177e4
LT
204
205/*
206 * These are the virtual MM functions - opening of an area, closing and
207 * unmapping it (needed to keep files on disk up-to-date etc), pointer
208 * to the functions called when a no-page or a wp-page exception occurs.
209 */
210struct vm_operations_struct {
211 void (*open)(struct vm_area_struct * area);
212 void (*close)(struct vm_area_struct * area);
d0217ac0 213 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
9637a5ef
DH
214
215 /* notification that a previously read-only page is about to become
216 * writable, if an error is returned it will cause a SIGBUS */
c2ec175c 217 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
28b2ee20
RR
218
219 /* called by access_process_vm when get_user_pages() fails, typically
220 * for use by special VMAs that can switch between memory and hardware
221 */
222 int (*access)(struct vm_area_struct *vma, unsigned long addr,
223 void *buf, int len, int write);
1da177e4 224#ifdef CONFIG_NUMA
a6020ed7
LS
225 /*
226 * set_policy() op must add a reference to any non-NULL @new mempolicy
227 * to hold the policy upon return. Caller should pass NULL @new to
228 * remove a policy and fall back to surrounding context--i.e. do not
229 * install a MPOL_DEFAULT policy, nor the task or system default
230 * mempolicy.
231 */
1da177e4 232 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
a6020ed7
LS
233
234 /*
235 * get_policy() op must add reference [mpol_get()] to any policy at
236 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
237 * in mm/mempolicy.c will do this automatically.
238 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
239 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
240 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
241 * must return NULL--i.e., do not "fallback" to task or system default
242 * policy.
243 */
1da177e4
LT
244 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
245 unsigned long addr);
7b2259b3
CL
246 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
247 const nodemask_t *to, unsigned long flags);
1da177e4 248#endif
0b173bc4
KK
249 /* called by sys_remap_file_pages() to populate non-linear mapping */
250 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
251 unsigned long size, pgoff_t pgoff);
1da177e4
LT
252};
253
254struct mmu_gather;
255struct inode;
256
349aef0b
AM
257#define page_private(page) ((page)->private)
258#define set_page_private(page, v) ((page)->private = (v))
4c21e2f2 259
b12c4ad1
MK
260/* It's valid only if the page is free path or free_list */
261static inline void set_freepage_migratetype(struct page *page, int migratetype)
262{
95e34412 263 page->index = migratetype;
b12c4ad1
MK
264}
265
266/* It's valid only if the page is free path or free_list */
267static inline int get_freepage_migratetype(struct page *page)
268{
95e34412 269 return page->index;
b12c4ad1
MK
270}
271
1da177e4
LT
272/*
273 * FIXME: take this include out, include page-flags.h in
274 * files which need it (119 of them)
275 */
276#include <linux/page-flags.h>
71e3aac0 277#include <linux/huge_mm.h>
1da177e4
LT
278
279/*
280 * Methods to modify the page usage count.
281 *
282 * What counts for a page usage:
283 * - cache mapping (page->mapping)
284 * - private data (page->private)
285 * - page mapped in a task's page tables, each mapping
286 * is counted separately
287 *
288 * Also, many kernel routines increase the page count before a critical
289 * routine so they can be sure the page doesn't go away from under them.
1da177e4
LT
290 */
291
292/*
da6052f7 293 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1da177e4 294 */
7c8ee9a8
NP
295static inline int put_page_testzero(struct page *page)
296{
725d704e 297 VM_BUG_ON(atomic_read(&page->_count) == 0);
8dc04efb 298 return atomic_dec_and_test(&page->_count);
7c8ee9a8 299}
1da177e4
LT
300
301/*
7c8ee9a8
NP
302 * Try to grab a ref unless the page has a refcount of zero, return false if
303 * that is the case.
8e0861fa
AK
304 * This can be called when MMU is off so it must not access
305 * any of the virtual mappings.
1da177e4 306 */
7c8ee9a8
NP
307static inline int get_page_unless_zero(struct page *page)
308{
8dc04efb 309 return atomic_inc_not_zero(&page->_count);
7c8ee9a8 310}
1da177e4 311
8e0861fa
AK
312/*
313 * Try to drop a ref unless the page has a refcount of one, return false if
314 * that is the case.
315 * This is to make sure that the refcount won't become zero after this drop.
316 * This can be called when MMU is off so it must not access
317 * any of the virtual mappings.
318 */
319static inline int put_page_unless_one(struct page *page)
320{
321 return atomic_add_unless(&page->_count, -1, 1);
322}
323
53df8fdc
WF
324extern int page_is_ram(unsigned long pfn);
325
48667e7a 326/* Support for virtually mapped pages */
b3bdda02
CL
327struct page *vmalloc_to_page(const void *addr);
328unsigned long vmalloc_to_pfn(const void *addr);
48667e7a 329
0738c4bb
PM
330/*
331 * Determine if an address is within the vmalloc range
332 *
333 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
334 * is no special casing required.
335 */
9e2779fa
CL
336static inline int is_vmalloc_addr(const void *x)
337{
0738c4bb 338#ifdef CONFIG_MMU
9e2779fa
CL
339 unsigned long addr = (unsigned long)x;
340
341 return addr >= VMALLOC_START && addr < VMALLOC_END;
0738c4bb
PM
342#else
343 return 0;
8ca3ed87 344#endif
0738c4bb 345}
81ac3ad9
KH
346#ifdef CONFIG_MMU
347extern int is_vmalloc_or_module_addr(const void *x);
348#else
934831d0 349static inline int is_vmalloc_or_module_addr(const void *x)
81ac3ad9
KH
350{
351 return 0;
352}
353#endif
9e2779fa 354
e9da73d6
AA
355static inline void compound_lock(struct page *page)
356{
357#ifdef CONFIG_TRANSPARENT_HUGEPAGE
5bf5f03c 358 VM_BUG_ON(PageSlab(page));
e9da73d6
AA
359 bit_spin_lock(PG_compound_lock, &page->flags);
360#endif
361}
362
363static inline void compound_unlock(struct page *page)
364{
365#ifdef CONFIG_TRANSPARENT_HUGEPAGE
5bf5f03c 366 VM_BUG_ON(PageSlab(page));
e9da73d6
AA
367 bit_spin_unlock(PG_compound_lock, &page->flags);
368#endif
369}
370
371static inline unsigned long compound_lock_irqsave(struct page *page)
372{
373 unsigned long uninitialized_var(flags);
374#ifdef CONFIG_TRANSPARENT_HUGEPAGE
375 local_irq_save(flags);
376 compound_lock(page);
377#endif
378 return flags;
379}
380
381static inline void compound_unlock_irqrestore(struct page *page,
382 unsigned long flags)
383{
384#ifdef CONFIG_TRANSPARENT_HUGEPAGE
385 compound_unlock(page);
386 local_irq_restore(flags);
387#endif
388}
389
d85f3385
CL
390static inline struct page *compound_head(struct page *page)
391{
6d777953 392 if (unlikely(PageTail(page)))
d85f3385
CL
393 return page->first_page;
394 return page;
395}
396
70b50f94
AA
397/*
398 * The atomic page->_mapcount, starts from -1: so that transitions
399 * both from it and to it can be tracked, using atomic_inc_and_test
400 * and atomic_add_negative(-1).
401 */
22b751c3 402static inline void page_mapcount_reset(struct page *page)
70b50f94
AA
403{
404 atomic_set(&(page)->_mapcount, -1);
405}
406
407static inline int page_mapcount(struct page *page)
408{
409 return atomic_read(&(page)->_mapcount) + 1;
410}
411
4c21e2f2 412static inline int page_count(struct page *page)
1da177e4 413{
d85f3385 414 return atomic_read(&compound_head(page)->_count);
1da177e4
LT
415}
416
b35a35b5
AA
417static inline void get_huge_page_tail(struct page *page)
418{
419 /*
420 * __split_huge_page_refcount() cannot run
421 * from under us.
422 */
423 VM_BUG_ON(page_mapcount(page) < 0);
424 VM_BUG_ON(atomic_read(&page->_count) != 0);
425 atomic_inc(&page->_mapcount);
426}
427
70b50f94
AA
428extern bool __get_page_tail(struct page *page);
429
1da177e4
LT
430static inline void get_page(struct page *page)
431{
70b50f94
AA
432 if (unlikely(PageTail(page)))
433 if (likely(__get_page_tail(page)))
434 return;
91807063
AA
435 /*
436 * Getting a normal page or the head of a compound page
70b50f94 437 * requires to already have an elevated page->_count.
91807063 438 */
70b50f94 439 VM_BUG_ON(atomic_read(&page->_count) <= 0);
1da177e4
LT
440 atomic_inc(&page->_count);
441}
442
b49af68f
CL
443static inline struct page *virt_to_head_page(const void *x)
444{
445 struct page *page = virt_to_page(x);
446 return compound_head(page);
447}
448
7835e98b
NP
449/*
450 * Setup the page count before being freed into the page allocator for
451 * the first time (boot or memory hotplug)
452 */
453static inline void init_page_count(struct page *page)
454{
455 atomic_set(&page->_count, 1);
456}
457
5f24ce5f
AA
458/*
459 * PageBuddy() indicate that the page is free and in the buddy system
460 * (see mm/page_alloc.c).
ef2b4b95
AA
461 *
462 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
463 * -2 so that an underflow of the page_mapcount() won't be mistaken
464 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
465 * efficiently by most CPU architectures.
5f24ce5f 466 */
ef2b4b95
AA
467#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
468
5f24ce5f
AA
469static inline int PageBuddy(struct page *page)
470{
ef2b4b95 471 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
5f24ce5f
AA
472}
473
474static inline void __SetPageBuddy(struct page *page)
475{
476 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
ef2b4b95 477 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
5f24ce5f
AA
478}
479
480static inline void __ClearPageBuddy(struct page *page)
481{
482 VM_BUG_ON(!PageBuddy(page));
483 atomic_set(&page->_mapcount, -1);
484}
485
1da177e4 486void put_page(struct page *page);
1d7ea732 487void put_pages_list(struct list_head *pages);
1da177e4 488
8dfcc9ba 489void split_page(struct page *page, unsigned int order);
748446bb 490int split_free_page(struct page *page);
8dfcc9ba 491
33f2ef89
AW
492/*
493 * Compound pages have a destructor function. Provide a
494 * prototype for that function and accessor functions.
495 * These are _only_ valid on the head of a PG_compound page.
496 */
497typedef void compound_page_dtor(struct page *);
498
499static inline void set_compound_page_dtor(struct page *page,
500 compound_page_dtor *dtor)
501{
502 page[1].lru.next = (void *)dtor;
503}
504
505static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
506{
507 return (compound_page_dtor *)page[1].lru.next;
508}
509
d85f3385
CL
510static inline int compound_order(struct page *page)
511{
6d777953 512 if (!PageHead(page))
d85f3385
CL
513 return 0;
514 return (unsigned long)page[1].lru.prev;
515}
516
517static inline void set_compound_order(struct page *page, unsigned long order)
518{
519 page[1].lru.prev = (void *)order;
520}
521
3dece370 522#ifdef CONFIG_MMU
14fd403f
AA
523/*
524 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
525 * servicing faults for write access. In the normal case, do always want
526 * pte_mkwrite. But get_user_pages can cause write faults for mappings
527 * that do not have writing enabled, when used by access_process_vm.
528 */
529static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
530{
531 if (likely(vma->vm_flags & VM_WRITE))
532 pte = pte_mkwrite(pte);
533 return pte;
534}
3dece370 535#endif
14fd403f 536
1da177e4
LT
537/*
538 * Multiple processes may "see" the same page. E.g. for untouched
539 * mappings of /dev/null, all processes see the same page full of
540 * zeroes, and text pages of executables and shared libraries have
541 * only one copy in memory, at most, normally.
542 *
543 * For the non-reserved pages, page_count(page) denotes a reference count.
7e871b6c
PBG
544 * page_count() == 0 means the page is free. page->lru is then used for
545 * freelist management in the buddy allocator.
da6052f7 546 * page_count() > 0 means the page has been allocated.
1da177e4 547 *
da6052f7
NP
548 * Pages are allocated by the slab allocator in order to provide memory
549 * to kmalloc and kmem_cache_alloc. In this case, the management of the
550 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
551 * unless a particular usage is carefully commented. (the responsibility of
552 * freeing the kmalloc memory is the caller's, of course).
1da177e4 553 *
da6052f7
NP
554 * A page may be used by anyone else who does a __get_free_page().
555 * In this case, page_count still tracks the references, and should only
556 * be used through the normal accessor functions. The top bits of page->flags
557 * and page->virtual store page management information, but all other fields
558 * are unused and could be used privately, carefully. The management of this
559 * page is the responsibility of the one who allocated it, and those who have
560 * subsequently been given references to it.
561 *
562 * The other pages (we may call them "pagecache pages") are completely
1da177e4
LT
563 * managed by the Linux memory manager: I/O, buffers, swapping etc.
564 * The following discussion applies only to them.
565 *
da6052f7
NP
566 * A pagecache page contains an opaque `private' member, which belongs to the
567 * page's address_space. Usually, this is the address of a circular list of
568 * the page's disk buffers. PG_private must be set to tell the VM to call
569 * into the filesystem to release these pages.
1da177e4 570 *
da6052f7
NP
571 * A page may belong to an inode's memory mapping. In this case, page->mapping
572 * is the pointer to the inode, and page->index is the file offset of the page,
573 * in units of PAGE_CACHE_SIZE.
1da177e4 574 *
da6052f7
NP
575 * If pagecache pages are not associated with an inode, they are said to be
576 * anonymous pages. These may become associated with the swapcache, and in that
577 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1da177e4 578 *
da6052f7
NP
579 * In either case (swapcache or inode backed), the pagecache itself holds one
580 * reference to the page. Setting PG_private should also increment the
581 * refcount. The each user mapping also has a reference to the page.
1da177e4 582 *
da6052f7
NP
583 * The pagecache pages are stored in a per-mapping radix tree, which is
584 * rooted at mapping->page_tree, and indexed by offset.
585 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
586 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1da177e4 587 *
da6052f7 588 * All pagecache pages may be subject to I/O:
1da177e4
LT
589 * - inode pages may need to be read from disk,
590 * - inode pages which have been modified and are MAP_SHARED may need
da6052f7
NP
591 * to be written back to the inode on disk,
592 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
593 * modified may need to be swapped out to swap space and (later) to be read
594 * back into memory.
1da177e4
LT
595 */
596
597/*
598 * The zone field is never updated after free_area_init_core()
599 * sets it, so none of the operations on it need to be atomic.
1da177e4 600 */
348f8b6c 601
90572890 602/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
07808b74 603#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
d41dee36
AW
604#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
605#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
90572890 606#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
d41dee36 607
348f8b6c 608/*
25985edc 609 * Define the bit shifts to access each section. For non-existent
348f8b6c
DH
610 * sections we define the shift as 0; that plus a 0 mask ensures
611 * the compiler will optimise away reference to them.
612 */
d41dee36
AW
613#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
614#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
615#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
90572890 616#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
348f8b6c 617
bce54bbf
WD
618/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
619#ifdef NODE_NOT_IN_PAGE_FLAGS
89689ae7 620#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
bd8029b6
AW
621#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
622 SECTIONS_PGOFF : ZONES_PGOFF)
d41dee36 623#else
89689ae7 624#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
bd8029b6
AW
625#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
626 NODES_PGOFF : ZONES_PGOFF)
89689ae7
CL
627#endif
628
bd8029b6 629#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
348f8b6c 630
9223b419
CL
631#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
632#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
348f8b6c
DH
633#endif
634
d41dee36
AW
635#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
636#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
637#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
90572890 638#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1)
89689ae7 639#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
348f8b6c 640
33dd4e0e 641static inline enum zone_type page_zonenum(const struct page *page)
1da177e4 642{
348f8b6c 643 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1da177e4 644}
1da177e4 645
9127ab4f
CS
646#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
647#define SECTION_IN_PAGE_FLAGS
648#endif
649
89689ae7 650/*
7a8010cd
VB
651 * The identification function is mainly used by the buddy allocator for
652 * determining if two pages could be buddies. We are not really identifying
653 * the zone since we could be using the section number id if we do not have
654 * node id available in page flags.
655 * We only guarantee that it will return the same value for two combinable
656 * pages in a zone.
89689ae7 657 */
cb2b95e1
AW
658static inline int page_zone_id(struct page *page)
659{
89689ae7 660 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
348f8b6c
DH
661}
662
25ba77c1 663static inline int zone_to_nid(struct zone *zone)
89fa3024 664{
d5f541ed
CL
665#ifdef CONFIG_NUMA
666 return zone->node;
667#else
668 return 0;
669#endif
89fa3024
CL
670}
671
89689ae7 672#ifdef NODE_NOT_IN_PAGE_FLAGS
33dd4e0e 673extern int page_to_nid(const struct page *page);
89689ae7 674#else
33dd4e0e 675static inline int page_to_nid(const struct page *page)
d41dee36 676{
89689ae7 677 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
d41dee36 678}
89689ae7
CL
679#endif
680
57e0a030 681#ifdef CONFIG_NUMA_BALANCING
90572890 682static inline int cpu_pid_to_cpupid(int cpu, int pid)
57e0a030 683{
90572890 684 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
57e0a030
MG
685}
686
90572890 687static inline int cpupid_to_pid(int cpupid)
57e0a030 688{
90572890 689 return cpupid & LAST__PID_MASK;
57e0a030 690}
b795854b 691
90572890 692static inline int cpupid_to_cpu(int cpupid)
b795854b 693{
90572890 694 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
b795854b
MG
695}
696
90572890 697static inline int cpupid_to_nid(int cpupid)
b795854b 698{
90572890 699 return cpu_to_node(cpupid_to_cpu(cpupid));
b795854b
MG
700}
701
90572890 702static inline bool cpupid_pid_unset(int cpupid)
57e0a030 703{
90572890 704 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
b795854b
MG
705}
706
90572890 707static inline bool cpupid_cpu_unset(int cpupid)
b795854b 708{
90572890 709 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
b795854b
MG
710}
711
8c8a743c
PZ
712static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
713{
714 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
715}
716
717#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
90572890
PZ
718#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
719static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
b795854b 720{
90572890 721 return xchg(&page->_last_cpupid, cpupid);
b795854b 722}
90572890
PZ
723
724static inline int page_cpupid_last(struct page *page)
725{
726 return page->_last_cpupid;
727}
728static inline void page_cpupid_reset_last(struct page *page)
b795854b 729{
90572890 730 page->_last_cpupid = -1;
57e0a030
MG
731}
732#else
90572890 733static inline int page_cpupid_last(struct page *page)
75980e97 734{
90572890 735 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
75980e97
PZ
736}
737
90572890 738extern int page_cpupid_xchg_last(struct page *page, int cpupid);
75980e97 739
90572890 740static inline void page_cpupid_reset_last(struct page *page)
75980e97 741{
90572890 742 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
4468b8f1 743
90572890
PZ
744 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
745 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
75980e97 746}
90572890
PZ
747#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
748#else /* !CONFIG_NUMA_BALANCING */
749static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
57e0a030 750{
90572890 751 return page_to_nid(page); /* XXX */
57e0a030
MG
752}
753
90572890 754static inline int page_cpupid_last(struct page *page)
57e0a030 755{
90572890 756 return page_to_nid(page); /* XXX */
57e0a030
MG
757}
758
90572890 759static inline int cpupid_to_nid(int cpupid)
b795854b
MG
760{
761 return -1;
762}
763
90572890 764static inline int cpupid_to_pid(int cpupid)
b795854b
MG
765{
766 return -1;
767}
768
90572890 769static inline int cpupid_to_cpu(int cpupid)
b795854b
MG
770{
771 return -1;
772}
773
90572890
PZ
774static inline int cpu_pid_to_cpupid(int nid, int pid)
775{
776 return -1;
777}
778
779static inline bool cpupid_pid_unset(int cpupid)
b795854b
MG
780{
781 return 1;
782}
783
90572890 784static inline void page_cpupid_reset_last(struct page *page)
57e0a030
MG
785{
786}
8c8a743c
PZ
787
788static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
789{
790 return false;
791}
90572890 792#endif /* CONFIG_NUMA_BALANCING */
57e0a030 793
33dd4e0e 794static inline struct zone *page_zone(const struct page *page)
89689ae7
CL
795{
796 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
797}
798
9127ab4f 799#ifdef SECTION_IN_PAGE_FLAGS
bf4e8902
DK
800static inline void set_page_section(struct page *page, unsigned long section)
801{
802 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
803 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
804}
805
aa462abe 806static inline unsigned long page_to_section(const struct page *page)
d41dee36
AW
807{
808 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
809}
308c05e3 810#endif
d41dee36 811
2f1b6248 812static inline void set_page_zone(struct page *page, enum zone_type zone)
348f8b6c
DH
813{
814 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
815 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
816}
2f1b6248 817
348f8b6c
DH
818static inline void set_page_node(struct page *page, unsigned long node)
819{
820 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
821 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1da177e4 822}
89689ae7 823
2f1b6248 824static inline void set_page_links(struct page *page, enum zone_type zone,
d41dee36 825 unsigned long node, unsigned long pfn)
1da177e4 826{
348f8b6c
DH
827 set_page_zone(page, zone);
828 set_page_node(page, node);
9127ab4f 829#ifdef SECTION_IN_PAGE_FLAGS
d41dee36 830 set_page_section(page, pfn_to_section_nr(pfn));
bf4e8902 831#endif
1da177e4
LT
832}
833
f6ac2354
CL
834/*
835 * Some inline functions in vmstat.h depend on page_zone()
836 */
837#include <linux/vmstat.h>
838
33dd4e0e 839static __always_inline void *lowmem_page_address(const struct page *page)
1da177e4 840{
aa462abe 841 return __va(PFN_PHYS(page_to_pfn(page)));
1da177e4
LT
842}
843
844#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
845#define HASHED_PAGE_VIRTUAL
846#endif
847
848#if defined(WANT_PAGE_VIRTUAL)
849#define page_address(page) ((page)->virtual)
850#define set_page_address(page, address) \
851 do { \
852 (page)->virtual = (address); \
853 } while(0)
854#define page_address_init() do { } while(0)
855#endif
856
857#if defined(HASHED_PAGE_VIRTUAL)
f9918794 858void *page_address(const struct page *page);
1da177e4
LT
859void set_page_address(struct page *page, void *virtual);
860void page_address_init(void);
861#endif
862
863#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
864#define page_address(page) lowmem_page_address(page)
865#define set_page_address(page, address) do { } while(0)
866#define page_address_init() do { } while(0)
867#endif
868
869/*
870 * On an anonymous page mapped into a user virtual memory area,
871 * page->mapping points to its anon_vma, not to a struct address_space;
3ca7b3c5
HD
872 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
873 *
874 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
875 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
876 * and then page->mapping points, not to an anon_vma, but to a private
877 * structure which KSM associates with that merged page. See ksm.h.
878 *
879 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
1da177e4
LT
880 *
881 * Please note that, confusingly, "page_mapping" refers to the inode
882 * address_space which maps the page from disk; whereas "page_mapped"
883 * refers to user virtual address space into which the page is mapped.
884 */
885#define PAGE_MAPPING_ANON 1
3ca7b3c5
HD
886#define PAGE_MAPPING_KSM 2
887#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
1da177e4 888
9800339b 889extern struct address_space *page_mapping(struct page *page);
1da177e4 890
3ca7b3c5
HD
891/* Neutral page->mapping pointer to address_space or anon_vma or other */
892static inline void *page_rmapping(struct page *page)
893{
894 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
895}
896
f981c595
MG
897extern struct address_space *__page_file_mapping(struct page *);
898
899static inline
900struct address_space *page_file_mapping(struct page *page)
901{
902 if (unlikely(PageSwapCache(page)))
903 return __page_file_mapping(page);
904
905 return page->mapping;
906}
907
1da177e4
LT
908static inline int PageAnon(struct page *page)
909{
910 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
911}
912
913/*
914 * Return the pagecache index of the passed page. Regular pagecache pages
915 * use ->index whereas swapcache pages use ->private
916 */
917static inline pgoff_t page_index(struct page *page)
918{
919 if (unlikely(PageSwapCache(page)))
4c21e2f2 920 return page_private(page);
1da177e4
LT
921 return page->index;
922}
923
f981c595
MG
924extern pgoff_t __page_file_index(struct page *page);
925
926/*
927 * Return the file index of the page. Regular pagecache pages use ->index
928 * whereas swapcache pages use swp_offset(->private)
929 */
930static inline pgoff_t page_file_index(struct page *page)
931{
932 if (unlikely(PageSwapCache(page)))
933 return __page_file_index(page);
934
935 return page->index;
936}
937
1da177e4
LT
938/*
939 * Return true if this page is mapped into pagetables.
940 */
941static inline int page_mapped(struct page *page)
942{
943 return atomic_read(&(page)->_mapcount) >= 0;
944}
945
1da177e4
LT
946/*
947 * Different kinds of faults, as returned by handle_mm_fault().
948 * Used to decide whether a process gets delivered SIGBUS or
949 * just gets major/minor fault counters bumped up.
950 */
d0217ac0 951
83c54070 952#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
d0217ac0 953
83c54070
NP
954#define VM_FAULT_OOM 0x0001
955#define VM_FAULT_SIGBUS 0x0002
956#define VM_FAULT_MAJOR 0x0004
957#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
aa50d3a7
AK
958#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
959#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
f33ea7f4 960
83c54070
NP
961#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
962#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
d065bd81 963#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
c0292554 964#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1da177e4 965
aa50d3a7
AK
966#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
967
968#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
c0292554 969 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
aa50d3a7
AK
970
971/* Encode hstate index for a hwpoisoned large page */
972#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
973#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
d0217ac0 974
1c0fe6e3
NP
975/*
976 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
977 */
978extern void pagefault_out_of_memory(void);
979
1da177e4
LT
980#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
981
ddd588b5 982/*
7bf02ea2 983 * Flags passed to show_mem() and show_free_areas() to suppress output in
ddd588b5
DR
984 * various contexts.
985 */
4b59e6c4
DR
986#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
987#define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
ddd588b5 988
7bf02ea2
DR
989extern void show_free_areas(unsigned int flags);
990extern bool skip_free_areas_node(unsigned int flags, int nid);
1da177e4 991
1da177e4
LT
992int shmem_zero_setup(struct vm_area_struct *);
993
e8edc6e0 994extern int can_do_mlock(void);
1da177e4
LT
995extern int user_shm_lock(size_t, struct user_struct *);
996extern void user_shm_unlock(size_t, struct user_struct *);
997
998/*
999 * Parameter block passed down to zap_pte_range in exceptional cases.
1000 */
1001struct zap_details {
1002 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
1003 struct address_space *check_mapping; /* Check page->mapping if set */
1004 pgoff_t first_index; /* Lowest page->index to unmap */
1005 pgoff_t last_index; /* Highest page->index to unmap */
1da177e4
LT
1006};
1007
7e675137
NP
1008struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1009 pte_t pte);
1010
c627f9cc
JS
1011int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1012 unsigned long size);
14f5ff5d 1013void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1da177e4 1014 unsigned long size, struct zap_details *);
4f74d2c8
LT
1015void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1016 unsigned long start, unsigned long end);
e6473092
MM
1017
1018/**
1019 * mm_walk - callbacks for walk_page_range
1020 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1021 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1022 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
03319327
DH
1023 * this handler is required to be able to handle
1024 * pmd_trans_huge() pmds. They may simply choose to
1025 * split_huge_page() instead of handling it explicitly.
e6473092
MM
1026 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1027 * @pte_hole: if set, called for each hole at all levels
5dc37642 1028 * @hugetlb_entry: if set, called for each hugetlb entry
c27fe4c8
KM
1029 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1030 * is used.
e6473092
MM
1031 *
1032 * (see walk_page_range for more details)
1033 */
1034struct mm_walk {
0f157a5b
AM
1035 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1036 unsigned long next, struct mm_walk *walk);
1037 int (*pud_entry)(pud_t *pud, unsigned long addr,
1038 unsigned long next, struct mm_walk *walk);
1039 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1040 unsigned long next, struct mm_walk *walk);
1041 int (*pte_entry)(pte_t *pte, unsigned long addr,
1042 unsigned long next, struct mm_walk *walk);
1043 int (*pte_hole)(unsigned long addr, unsigned long next,
1044 struct mm_walk *walk);
1045 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1046 unsigned long addr, unsigned long next,
1047 struct mm_walk *walk);
2165009b
DH
1048 struct mm_struct *mm;
1049 void *private;
e6473092
MM
1050};
1051
2165009b
DH
1052int walk_page_range(unsigned long addr, unsigned long end,
1053 struct mm_walk *walk);
42b77728 1054void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
3bf5ee95 1055 unsigned long end, unsigned long floor, unsigned long ceiling);
1da177e4
LT
1056int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1057 struct vm_area_struct *vma);
1da177e4
LT
1058void unmap_mapping_range(struct address_space *mapping,
1059 loff_t const holebegin, loff_t const holelen, int even_cows);
3b6748e2
JW
1060int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1061 unsigned long *pfn);
d87fe660 1062int follow_phys(struct vm_area_struct *vma, unsigned long address,
1063 unsigned int flags, unsigned long *prot, resource_size_t *phys);
28b2ee20
RR
1064int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1065 void *buf, int len, int write);
1da177e4
LT
1066
1067static inline void unmap_shared_mapping_range(struct address_space *mapping,
1068 loff_t const holebegin, loff_t const holelen)
1069{
1070 unmap_mapping_range(mapping, holebegin, holelen, 0);
1071}
1072
7caef267 1073extern void truncate_pagecache(struct inode *inode, loff_t new);
2c27c65e 1074extern void truncate_setsize(struct inode *inode, loff_t newsize);
623e3db9 1075void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
750b4987 1076int truncate_inode_page(struct address_space *mapping, struct page *page);
25718736 1077int generic_error_remove_page(struct address_space *mapping, struct page *page);
83f78668
WF
1078int invalidate_inode_page(struct page *page);
1079
7ee1dd3f 1080#ifdef CONFIG_MMU
83c54070 1081extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
d06063cc 1082 unsigned long address, unsigned int flags);
5c723ba5
PZ
1083extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1084 unsigned long address, unsigned int fault_flags);
7ee1dd3f
DH
1085#else
1086static inline int handle_mm_fault(struct mm_struct *mm,
1087 struct vm_area_struct *vma, unsigned long address,
d06063cc 1088 unsigned int flags)
7ee1dd3f
DH
1089{
1090 /* should never happen if there's no MMU */
1091 BUG();
1092 return VM_FAULT_SIGBUS;
1093}
5c723ba5
PZ
1094static inline int fixup_user_fault(struct task_struct *tsk,
1095 struct mm_struct *mm, unsigned long address,
1096 unsigned int fault_flags)
1097{
1098 /* should never happen if there's no MMU */
1099 BUG();
1100 return -EFAULT;
1101}
7ee1dd3f 1102#endif
f33ea7f4 1103
1da177e4 1104extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
5ddd36b9
SW
1105extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1106 void *buf, int len, int write);
1da177e4 1107
28a35716
ML
1108long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1109 unsigned long start, unsigned long nr_pages,
1110 unsigned int foll_flags, struct page **pages,
1111 struct vm_area_struct **vmas, int *nonblocking);
1112long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1113 unsigned long start, unsigned long nr_pages,
1114 int write, int force, struct page **pages,
1115 struct vm_area_struct **vmas);
d2bf6be8
NP
1116int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1117 struct page **pages);
18022c5d
MG
1118struct kvec;
1119int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1120 struct page **pages);
1121int get_kernel_page(unsigned long start, int write, struct page **pages);
f3e8fccd 1122struct page *get_dump_page(unsigned long addr);
1da177e4 1123
cf9a2ae8 1124extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
d47992f8
LC
1125extern void do_invalidatepage(struct page *page, unsigned int offset,
1126 unsigned int length);
cf9a2ae8 1127
1da177e4 1128int __set_page_dirty_nobuffers(struct page *page);
76719325 1129int __set_page_dirty_no_writeback(struct page *page);
1da177e4
LT
1130int redirty_page_for_writepage(struct writeback_control *wbc,
1131 struct page *page);
e3a7cca1 1132void account_page_dirtied(struct page *page, struct address_space *mapping);
f629d1c9 1133void account_page_writeback(struct page *page);
b3c97528 1134int set_page_dirty(struct page *page);
1da177e4
LT
1135int set_page_dirty_lock(struct page *page);
1136int clear_page_dirty_for_io(struct page *page);
1137
39aa3cb3 1138/* Is the vma a continuation of the stack vma above it? */
a09a79f6 1139static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
39aa3cb3
SB
1140{
1141 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1142}
1143
a09a79f6
MP
1144static inline int stack_guard_page_start(struct vm_area_struct *vma,
1145 unsigned long addr)
1146{
1147 return (vma->vm_flags & VM_GROWSDOWN) &&
1148 (vma->vm_start == addr) &&
1149 !vma_growsdown(vma->vm_prev, addr);
1150}
1151
1152/* Is the vma a continuation of the stack vma below it? */
1153static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1154{
1155 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1156}
1157
1158static inline int stack_guard_page_end(struct vm_area_struct *vma,
1159 unsigned long addr)
1160{
1161 return (vma->vm_flags & VM_GROWSUP) &&
1162 (vma->vm_end == addr) &&
1163 !vma_growsup(vma->vm_next, addr);
1164}
1165
b7643757
SP
1166extern pid_t
1167vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1168
b6a2fea3
OW
1169extern unsigned long move_page_tables(struct vm_area_struct *vma,
1170 unsigned long old_addr, struct vm_area_struct *new_vma,
38a76013
ML
1171 unsigned long new_addr, unsigned long len,
1172 bool need_rmap_locks);
7da4d641
PZ
1173extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1174 unsigned long end, pgprot_t newprot,
4b10e7d5 1175 int dirty_accountable, int prot_numa);
b6a2fea3
OW
1176extern int mprotect_fixup(struct vm_area_struct *vma,
1177 struct vm_area_struct **pprev, unsigned long start,
1178 unsigned long end, unsigned long newflags);
1da177e4 1179
465a454f
PZ
1180/*
1181 * doesn't attempt to fault and will return short.
1182 */
1183int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1184 struct page **pages);
d559db08
KH
1185/*
1186 * per-process(per-mm_struct) statistics.
1187 */
d559db08
KH
1188static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1189{
69c97823
KK
1190 long val = atomic_long_read(&mm->rss_stat.count[member]);
1191
1192#ifdef SPLIT_RSS_COUNTING
1193 /*
1194 * counter is updated in asynchronous manner and may go to minus.
1195 * But it's never be expected number for users.
1196 */
1197 if (val < 0)
1198 val = 0;
172703b0 1199#endif
69c97823
KK
1200 return (unsigned long)val;
1201}
d559db08
KH
1202
1203static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1204{
172703b0 1205 atomic_long_add(value, &mm->rss_stat.count[member]);
d559db08
KH
1206}
1207
1208static inline void inc_mm_counter(struct mm_struct *mm, int member)
1209{
172703b0 1210 atomic_long_inc(&mm->rss_stat.count[member]);
d559db08
KH
1211}
1212
1213static inline void dec_mm_counter(struct mm_struct *mm, int member)
1214{
172703b0 1215 atomic_long_dec(&mm->rss_stat.count[member]);
d559db08
KH
1216}
1217
d559db08
KH
1218static inline unsigned long get_mm_rss(struct mm_struct *mm)
1219{
1220 return get_mm_counter(mm, MM_FILEPAGES) +
1221 get_mm_counter(mm, MM_ANONPAGES);
1222}
1223
1224static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1225{
1226 return max(mm->hiwater_rss, get_mm_rss(mm));
1227}
1228
1229static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1230{
1231 return max(mm->hiwater_vm, mm->total_vm);
1232}
1233
1234static inline void update_hiwater_rss(struct mm_struct *mm)
1235{
1236 unsigned long _rss = get_mm_rss(mm);
1237
1238 if ((mm)->hiwater_rss < _rss)
1239 (mm)->hiwater_rss = _rss;
1240}
1241
1242static inline void update_hiwater_vm(struct mm_struct *mm)
1243{
1244 if (mm->hiwater_vm < mm->total_vm)
1245 mm->hiwater_vm = mm->total_vm;
1246}
1247
1248static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1249 struct mm_struct *mm)
1250{
1251 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1252
1253 if (*maxrss < hiwater_rss)
1254 *maxrss = hiwater_rss;
1255}
1256
53bddb4e 1257#if defined(SPLIT_RSS_COUNTING)
05af2e10 1258void sync_mm_rss(struct mm_struct *mm);
53bddb4e 1259#else
05af2e10 1260static inline void sync_mm_rss(struct mm_struct *mm)
53bddb4e
KH
1261{
1262}
1263#endif
465a454f 1264
4e950f6f 1265int vma_wants_writenotify(struct vm_area_struct *vma);
d08b3851 1266
25ca1d6c
NK
1267extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1268 spinlock_t **ptl);
1269static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1270 spinlock_t **ptl)
1271{
1272 pte_t *ptep;
1273 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1274 return ptep;
1275}
c9cfcddf 1276
5f22df00
NP
1277#ifdef __PAGETABLE_PUD_FOLDED
1278static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1279 unsigned long address)
1280{
1281 return 0;
1282}
1283#else
1bb3630e 1284int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
5f22df00
NP
1285#endif
1286
1287#ifdef __PAGETABLE_PMD_FOLDED
1288static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1289 unsigned long address)
1290{
1291 return 0;
1292}
1293#else
1bb3630e 1294int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
5f22df00
NP
1295#endif
1296
8ac1f832
AA
1297int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1298 pmd_t *pmd, unsigned long address);
1bb3630e
HD
1299int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1300
1da177e4
LT
1301/*
1302 * The following ifdef needed to get the 4level-fixup.h header to work.
1303 * Remove it when 4level-fixup.h has been removed.
1304 */
1bb3630e 1305#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1da177e4
LT
1306static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1307{
1bb3630e
HD
1308 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1309 NULL: pud_offset(pgd, address);
1da177e4
LT
1310}
1311
1312static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1313{
1bb3630e
HD
1314 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1315 NULL: pmd_offset(pud, address);
1da177e4 1316}
1bb3630e
HD
1317#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1318
57c1ffce 1319#if USE_SPLIT_PTE_PTLOCKS
597d795a 1320#if ALLOC_SPLIT_PTLOCKS
539edb58
PZ
1321extern bool ptlock_alloc(struct page *page);
1322extern void ptlock_free(struct page *page);
1323
1324static inline spinlock_t *ptlock_ptr(struct page *page)
1325{
1326 return page->ptl;
1327}
597d795a 1328#else /* ALLOC_SPLIT_PTLOCKS */
49076ec2
KS
1329static inline bool ptlock_alloc(struct page *page)
1330{
49076ec2
KS
1331 return true;
1332}
539edb58 1333
49076ec2
KS
1334static inline void ptlock_free(struct page *page)
1335{
49076ec2
KS
1336}
1337
1338static inline spinlock_t *ptlock_ptr(struct page *page)
1339{
539edb58 1340 return &page->ptl;
49076ec2 1341}
597d795a 1342#endif /* ALLOC_SPLIT_PTLOCKS */
49076ec2
KS
1343
1344static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1345{
1346 return ptlock_ptr(pmd_page(*pmd));
1347}
1348
1349static inline bool ptlock_init(struct page *page)
1350{
1351 /*
1352 * prep_new_page() initialize page->private (and therefore page->ptl)
1353 * with 0. Make sure nobody took it in use in between.
1354 *
1355 * It can happen if arch try to use slab for page table allocation:
1356 * slab code uses page->slab_cache and page->first_page (for tail
1357 * pages), which share storage with page->ptl.
1358 */
539edb58 1359 VM_BUG_ON(*(unsigned long *)&page->ptl);
49076ec2
KS
1360 if (!ptlock_alloc(page))
1361 return false;
1362 spin_lock_init(ptlock_ptr(page));
1363 return true;
1364}
1365
1366/* Reset page->mapping so free_pages_check won't complain. */
1367static inline void pte_lock_deinit(struct page *page)
1368{
1369 page->mapping = NULL;
1370 ptlock_free(page);
1371}
1372
57c1ffce 1373#else /* !USE_SPLIT_PTE_PTLOCKS */
4c21e2f2
HD
1374/*
1375 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1376 */
49076ec2
KS
1377static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1378{
1379 return &mm->page_table_lock;
1380}
1381static inline bool ptlock_init(struct page *page) { return true; }
1382static inline void pte_lock_deinit(struct page *page) {}
57c1ffce 1383#endif /* USE_SPLIT_PTE_PTLOCKS */
4c21e2f2 1384
390f44e2 1385static inline bool pgtable_page_ctor(struct page *page)
2f569afd 1386{
2f569afd 1387 inc_zone_page_state(page, NR_PAGETABLE);
49076ec2 1388 return ptlock_init(page);
2f569afd
MS
1389}
1390
1391static inline void pgtable_page_dtor(struct page *page)
1392{
1393 pte_lock_deinit(page);
1394 dec_zone_page_state(page, NR_PAGETABLE);
1395}
1396
c74df32c
HD
1397#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1398({ \
4c21e2f2 1399 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
c74df32c
HD
1400 pte_t *__pte = pte_offset_map(pmd, address); \
1401 *(ptlp) = __ptl; \
1402 spin_lock(__ptl); \
1403 __pte; \
1404})
1405
1406#define pte_unmap_unlock(pte, ptl) do { \
1407 spin_unlock(ptl); \
1408 pte_unmap(pte); \
1409} while (0)
1410
8ac1f832
AA
1411#define pte_alloc_map(mm, vma, pmd, address) \
1412 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1413 pmd, address))? \
1414 NULL: pte_offset_map(pmd, address))
1bb3630e 1415
c74df32c 1416#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
8ac1f832
AA
1417 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1418 pmd, address))? \
c74df32c
HD
1419 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1420
1bb3630e 1421#define pte_alloc_kernel(pmd, address) \
8ac1f832 1422 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1bb3630e 1423 NULL: pte_offset_kernel(pmd, address))
1da177e4 1424
e009bb30
KS
1425#if USE_SPLIT_PMD_PTLOCKS
1426
1427static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1428{
49076ec2 1429 return ptlock_ptr(virt_to_page(pmd));
e009bb30
KS
1430}
1431
1432static inline bool pgtable_pmd_page_ctor(struct page *page)
1433{
e009bb30
KS
1434#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1435 page->pmd_huge_pte = NULL;
1436#endif
49076ec2 1437 return ptlock_init(page);
e009bb30
KS
1438}
1439
1440static inline void pgtable_pmd_page_dtor(struct page *page)
1441{
1442#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1443 VM_BUG_ON(page->pmd_huge_pte);
1444#endif
49076ec2 1445 ptlock_free(page);
e009bb30
KS
1446}
1447
1448#define pmd_huge_pte(mm, pmd) (virt_to_page(pmd)->pmd_huge_pte)
1449
1450#else
1451
9a86cb7b
KS
1452static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1453{
1454 return &mm->page_table_lock;
1455}
1456
e009bb30
KS
1457static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1458static inline void pgtable_pmd_page_dtor(struct page *page) {}
1459
c389a250 1460#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
9a86cb7b 1461
e009bb30
KS
1462#endif
1463
9a86cb7b
KS
1464static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1465{
1466 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1467 spin_lock(ptl);
1468 return ptl;
1469}
1470
1da177e4 1471extern void free_area_init(unsigned long * zones_size);
9109fb7b
JW
1472extern void free_area_init_node(int nid, unsigned long * zones_size,
1473 unsigned long zone_start_pfn, unsigned long *zholes_size);
49a7f04a
DH
1474extern void free_initmem(void);
1475
69afade7
JL
1476/*
1477 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1478 * into the buddy system. The freed pages will be poisoned with pattern
dbe67df4 1479 * "poison" if it's within range [0, UCHAR_MAX].
69afade7
JL
1480 * Return pages freed into the buddy system.
1481 */
11199692 1482extern unsigned long free_reserved_area(void *start, void *end,
69afade7 1483 int poison, char *s);
c3d5f5f0 1484
cfa11e08
JL
1485#ifdef CONFIG_HIGHMEM
1486/*
1487 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1488 * and totalram_pages.
1489 */
1490extern void free_highmem_page(struct page *page);
1491#endif
69afade7 1492
c3d5f5f0 1493extern void adjust_managed_page_count(struct page *page, long count);
7ee3d4e8 1494extern void mem_init_print_info(const char *str);
69afade7
JL
1495
1496/* Free the reserved page into the buddy system, so it gets managed. */
1497static inline void __free_reserved_page(struct page *page)
1498{
1499 ClearPageReserved(page);
1500 init_page_count(page);
1501 __free_page(page);
1502}
1503
1504static inline void free_reserved_page(struct page *page)
1505{
1506 __free_reserved_page(page);
1507 adjust_managed_page_count(page, 1);
1508}
1509
1510static inline void mark_page_reserved(struct page *page)
1511{
1512 SetPageReserved(page);
1513 adjust_managed_page_count(page, -1);
1514}
1515
1516/*
1517 * Default method to free all the __init memory into the buddy system.
dbe67df4
JL
1518 * The freed pages will be poisoned with pattern "poison" if it's within
1519 * range [0, UCHAR_MAX].
1520 * Return pages freed into the buddy system.
69afade7
JL
1521 */
1522static inline unsigned long free_initmem_default(int poison)
1523{
1524 extern char __init_begin[], __init_end[];
1525
11199692 1526 return free_reserved_area(&__init_begin, &__init_end,
69afade7
JL
1527 poison, "unused kernel");
1528}
1529
7ee3d4e8
JL
1530static inline unsigned long get_num_physpages(void)
1531{
1532 int nid;
1533 unsigned long phys_pages = 0;
1534
1535 for_each_online_node(nid)
1536 phys_pages += node_present_pages(nid);
1537
1538 return phys_pages;
1539}
1540
0ee332c1 1541#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
c713216d 1542/*
0ee332c1 1543 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
c713216d
MG
1544 * zones, allocate the backing mem_map and account for memory holes in a more
1545 * architecture independent manner. This is a substitute for creating the
1546 * zone_sizes[] and zholes_size[] arrays and passing them to
1547 * free_area_init_node()
1548 *
1549 * An architecture is expected to register range of page frames backed by
0ee332c1 1550 * physical memory with memblock_add[_node]() before calling
c713216d
MG
1551 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1552 * usage, an architecture is expected to do something like
1553 *
1554 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1555 * max_highmem_pfn};
1556 * for_each_valid_physical_page_range()
0ee332c1 1557 * memblock_add_node(base, size, nid)
c713216d
MG
1558 * free_area_init_nodes(max_zone_pfns);
1559 *
0ee332c1
TH
1560 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1561 * registered physical page range. Similarly
1562 * sparse_memory_present_with_active_regions() calls memory_present() for
1563 * each range when SPARSEMEM is enabled.
c713216d
MG
1564 *
1565 * See mm/page_alloc.c for more information on each function exposed by
0ee332c1 1566 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
c713216d
MG
1567 */
1568extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1e01979c 1569unsigned long node_map_pfn_alignment(void);
32996250
YL
1570unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1571 unsigned long end_pfn);
c713216d
MG
1572extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1573 unsigned long end_pfn);
1574extern void get_pfn_range_for_nid(unsigned int nid,
1575 unsigned long *start_pfn, unsigned long *end_pfn);
1576extern unsigned long find_min_pfn_with_active_regions(void);
c713216d
MG
1577extern void free_bootmem_with_active_regions(int nid,
1578 unsigned long max_low_pfn);
1579extern void sparse_memory_present_with_active_regions(int nid);
f2dbcfa7 1580
0ee332c1 1581#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
f2dbcfa7 1582
0ee332c1 1583#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
f2dbcfa7
KH
1584 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1585static inline int __early_pfn_to_nid(unsigned long pfn)
1586{
1587 return 0;
1588}
1589#else
1590/* please see mm/page_alloc.c */
1591extern int __meminit early_pfn_to_nid(unsigned long pfn);
1592#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1593/* there is a per-arch backend function. */
1594extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1595#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1596#endif
1597
0e0b864e 1598extern void set_dma_reserve(unsigned long new_dma_reserve);
a2f3aa02
DH
1599extern void memmap_init_zone(unsigned long, int, unsigned long,
1600 unsigned long, enum memmap_context);
bc75d33f 1601extern void setup_per_zone_wmarks(void);
1b79acc9 1602extern int __meminit init_per_zone_wmark_min(void);
1da177e4 1603extern void mem_init(void);
8feae131 1604extern void __init mmap_init(void);
b2b755b5 1605extern void show_mem(unsigned int flags);
1da177e4
LT
1606extern void si_meminfo(struct sysinfo * val);
1607extern void si_meminfo_node(struct sysinfo *val, int nid);
1608
3ee9a4f0
JP
1609extern __printf(3, 4)
1610void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
a238ab5b 1611
e7c8d5c9 1612extern void setup_per_cpu_pageset(void);
e7c8d5c9 1613
112067f0 1614extern void zone_pcp_update(struct zone *zone);
340175b7 1615extern void zone_pcp_reset(struct zone *zone);
112067f0 1616
75f7ad8e
PS
1617/* page_alloc.c */
1618extern int min_free_kbytes;
1619
8feae131 1620/* nommu.c */
33e5d769 1621extern atomic_long_t mmap_pages_allocated;
7e660872 1622extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
8feae131 1623
6b2dbba8 1624/* interval_tree.c */
6b2dbba8
ML
1625void vma_interval_tree_insert(struct vm_area_struct *node,
1626 struct rb_root *root);
9826a516
ML
1627void vma_interval_tree_insert_after(struct vm_area_struct *node,
1628 struct vm_area_struct *prev,
1629 struct rb_root *root);
6b2dbba8
ML
1630void vma_interval_tree_remove(struct vm_area_struct *node,
1631 struct rb_root *root);
1632struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1633 unsigned long start, unsigned long last);
1634struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1635 unsigned long start, unsigned long last);
1636
1637#define vma_interval_tree_foreach(vma, root, start, last) \
1638 for (vma = vma_interval_tree_iter_first(root, start, last); \
1639 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1da177e4
LT
1640
1641static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1642 struct list_head *list)
1643{
6b2dbba8 1644 list_add_tail(&vma->shared.nonlinear, list);
1da177e4
LT
1645}
1646
bf181b9f
ML
1647void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1648 struct rb_root *root);
1649void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1650 struct rb_root *root);
1651struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1652 struct rb_root *root, unsigned long start, unsigned long last);
1653struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1654 struct anon_vma_chain *node, unsigned long start, unsigned long last);
ed8ea815
ML
1655#ifdef CONFIG_DEBUG_VM_RB
1656void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1657#endif
bf181b9f
ML
1658
1659#define anon_vma_interval_tree_foreach(avc, root, start, last) \
1660 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1661 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1662
1da177e4 1663/* mmap.c */
34b4e4aa 1664extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
5beb4930 1665extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1da177e4
LT
1666 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1667extern struct vm_area_struct *vma_merge(struct mm_struct *,
1668 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1669 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1670 struct mempolicy *);
1671extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1672extern int split_vma(struct mm_struct *,
1673 struct vm_area_struct *, unsigned long addr, int new_below);
1674extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1675extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1676 struct rb_node **, struct rb_node *);
a8fb5618 1677extern void unlink_file_vma(struct vm_area_struct *);
1da177e4 1678extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
38a76013
ML
1679 unsigned long addr, unsigned long len, pgoff_t pgoff,
1680 bool *need_rmap_locks);
1da177e4 1681extern void exit_mmap(struct mm_struct *);
925d1c40 1682
7906d00c
AA
1683extern int mm_take_all_locks(struct mm_struct *mm);
1684extern void mm_drop_all_locks(struct mm_struct *mm);
1685
38646013
JS
1686extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1687extern struct file *get_mm_exe_file(struct mm_struct *mm);
925d1c40 1688
119f657c 1689extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
fa5dc22f
RM
1690extern int install_special_mapping(struct mm_struct *mm,
1691 unsigned long addr, unsigned long len,
1692 unsigned long flags, struct page **pages);
1da177e4
LT
1693
1694extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1695
0165ab44 1696extern unsigned long mmap_region(struct file *file, unsigned long addr,
c22c0d63 1697 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
bebeb3d6
ML
1698extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1699 unsigned long len, unsigned long prot, unsigned long flags,
41badc15 1700 unsigned long pgoff, unsigned long *populate);
1da177e4
LT
1701extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1702
bebeb3d6
ML
1703#ifdef CONFIG_MMU
1704extern int __mm_populate(unsigned long addr, unsigned long len,
1705 int ignore_errors);
1706static inline void mm_populate(unsigned long addr, unsigned long len)
1707{
1708 /* Ignore errors */
1709 (void) __mm_populate(addr, len, 1);
1710}
1711#else
1712static inline void mm_populate(unsigned long addr, unsigned long len) {}
1713#endif
1714
e4eb1ff6
LT
1715/* These take the mm semaphore themselves */
1716extern unsigned long vm_brk(unsigned long, unsigned long);
bfce281c 1717extern int vm_munmap(unsigned long, size_t);
6be5ceb0
LT
1718extern unsigned long vm_mmap(struct file *, unsigned long,
1719 unsigned long, unsigned long,
1720 unsigned long, unsigned long);
1da177e4 1721
db4fbfb9
ML
1722struct vm_unmapped_area_info {
1723#define VM_UNMAPPED_AREA_TOPDOWN 1
1724 unsigned long flags;
1725 unsigned long length;
1726 unsigned long low_limit;
1727 unsigned long high_limit;
1728 unsigned long align_mask;
1729 unsigned long align_offset;
1730};
1731
1732extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1733extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1734
1735/*
1736 * Search for an unmapped address range.
1737 *
1738 * We are looking for a range that:
1739 * - does not intersect with any VMA;
1740 * - is contained within the [low_limit, high_limit) interval;
1741 * - is at least the desired size.
1742 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1743 */
1744static inline unsigned long
1745vm_unmapped_area(struct vm_unmapped_area_info *info)
1746{
1747 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1748 return unmapped_area(info);
1749 else
1750 return unmapped_area_topdown(info);
1751}
1752
85821aab 1753/* truncate.c */
1da177e4 1754extern void truncate_inode_pages(struct address_space *, loff_t);
d7339071
HR
1755extern void truncate_inode_pages_range(struct address_space *,
1756 loff_t lstart, loff_t lend);
1da177e4
LT
1757
1758/* generic vm_area_ops exported for stackable file systems */
d0217ac0 1759extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
4fcf1c62 1760extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1da177e4
LT
1761
1762/* mm/page-writeback.c */
1763int write_one_page(struct page *page, int wait);
1cf6e7d8 1764void task_dirty_inc(struct task_struct *tsk);
1da177e4
LT
1765
1766/* readahead.c */
1767#define VM_MAX_READAHEAD 128 /* kbytes */
1768#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1da177e4 1769
1da177e4 1770int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
7361f4d8 1771 pgoff_t offset, unsigned long nr_to_read);
cf914a7d
RR
1772
1773void page_cache_sync_readahead(struct address_space *mapping,
1774 struct file_ra_state *ra,
1775 struct file *filp,
1776 pgoff_t offset,
1777 unsigned long size);
1778
1779void page_cache_async_readahead(struct address_space *mapping,
1780 struct file_ra_state *ra,
1781 struct file *filp,
1782 struct page *pg,
1783 pgoff_t offset,
1784 unsigned long size);
1785
1da177e4 1786unsigned long max_sane_readahead(unsigned long nr);
d30a1100
WF
1787unsigned long ra_submit(struct file_ra_state *ra,
1788 struct address_space *mapping,
1789 struct file *filp);
1da177e4 1790
d05f3169 1791/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
46dea3d0 1792extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
d05f3169
MH
1793
1794/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1795extern int expand_downwards(struct vm_area_struct *vma,
1796 unsigned long address);
8ca3eb08 1797#if VM_GROWSUP
46dea3d0 1798extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
8ca3eb08
LT
1799#else
1800 #define expand_upwards(vma, address) do { } while (0)
9ab88515 1801#endif
1da177e4
LT
1802
1803/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1804extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1805extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1806 struct vm_area_struct **pprev);
1807
1808/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1809 NULL if none. Assume start_addr < end_addr. */
1810static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1811{
1812 struct vm_area_struct * vma = find_vma(mm,start_addr);
1813
1814 if (vma && end_addr <= vma->vm_start)
1815 vma = NULL;
1816 return vma;
1817}
1818
1819static inline unsigned long vma_pages(struct vm_area_struct *vma)
1820{
1821 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1822}
1823
640708a2
PE
1824/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1825static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1826 unsigned long vm_start, unsigned long vm_end)
1827{
1828 struct vm_area_struct *vma = find_vma(mm, vm_start);
1829
1830 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1831 vma = NULL;
1832
1833 return vma;
1834}
1835
bad849b3 1836#ifdef CONFIG_MMU
804af2cf 1837pgprot_t vm_get_page_prot(unsigned long vm_flags);
bad849b3
DH
1838#else
1839static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1840{
1841 return __pgprot(0);
1842}
1843#endif
1844
b24f53a0 1845#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
4b10e7d5 1846unsigned long change_prot_numa(struct vm_area_struct *vma,
b24f53a0
LS
1847 unsigned long start, unsigned long end);
1848#endif
1849
deceb6cd 1850struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
deceb6cd
HD
1851int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1852 unsigned long pfn, unsigned long size, pgprot_t);
a145dd41 1853int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
e0dc0d8f
NP
1854int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1855 unsigned long pfn);
423bad60
NP
1856int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1857 unsigned long pfn);
b4cbb197
LT
1858int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1859
deceb6cd 1860
240aadee
ML
1861struct page *follow_page_mask(struct vm_area_struct *vma,
1862 unsigned long address, unsigned int foll_flags,
1863 unsigned int *page_mask);
1864
1865static inline struct page *follow_page(struct vm_area_struct *vma,
1866 unsigned long address, unsigned int foll_flags)
1867{
1868 unsigned int unused_page_mask;
1869 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1870}
1871
deceb6cd
HD
1872#define FOLL_WRITE 0x01 /* check pte is writable */
1873#define FOLL_TOUCH 0x02 /* mark page accessed */
1874#define FOLL_GET 0x04 /* do get_page on page */
8e4b9a60 1875#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
58fa879e 1876#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
318b275f
GN
1877#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1878 * and return without waiting upon it */
110d74a9 1879#define FOLL_MLOCK 0x40 /* mark page as mlocked */
500d65d4 1880#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
69ebb83e 1881#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
0b9d7052 1882#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
5117b3b8 1883#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1da177e4 1884
2f569afd 1885typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
aee16b3c
JF
1886 void *data);
1887extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1888 unsigned long size, pte_fn_t fn, void *data);
1889
1da177e4 1890#ifdef CONFIG_PROC_FS
ab50b8ed 1891void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1da177e4 1892#else
ab50b8ed 1893static inline void vm_stat_account(struct mm_struct *mm,
1da177e4
LT
1894 unsigned long flags, struct file *file, long pages)
1895{
44de9d0c 1896 mm->total_vm += pages;
1da177e4
LT
1897}
1898#endif /* CONFIG_PROC_FS */
1899
12d6f21e 1900#ifdef CONFIG_DEBUG_PAGEALLOC
12d6f21e 1901extern void kernel_map_pages(struct page *page, int numpages, int enable);
8a235efa
RW
1902#ifdef CONFIG_HIBERNATION
1903extern bool kernel_page_present(struct page *page);
1904#endif /* CONFIG_HIBERNATION */
12d6f21e 1905#else
1da177e4 1906static inline void
9858db50 1907kernel_map_pages(struct page *page, int numpages, int enable) {}
8a235efa
RW
1908#ifdef CONFIG_HIBERNATION
1909static inline bool kernel_page_present(struct page *page) { return true; }
1910#endif /* CONFIG_HIBERNATION */
1da177e4
LT
1911#endif
1912
31db58b3 1913extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1da177e4 1914#ifdef __HAVE_ARCH_GATE_AREA
cae5d390 1915int in_gate_area_no_mm(unsigned long addr);
83b964bb 1916int in_gate_area(struct mm_struct *mm, unsigned long addr);
1da177e4 1917#else
cae5d390
SW
1918int in_gate_area_no_mm(unsigned long addr);
1919#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1da177e4
LT
1920#endif /* __HAVE_ARCH_GATE_AREA */
1921
146732ce
JT
1922#ifdef CONFIG_SYSCTL
1923extern int sysctl_drop_caches;
8d65af78 1924int drop_caches_sysctl_handler(struct ctl_table *, int,
9d0243bc 1925 void __user *, size_t *, loff_t *);
146732ce
JT
1926#endif
1927
a09ed5e0 1928unsigned long shrink_slab(struct shrink_control *shrink,
1495f230
YH
1929 unsigned long nr_pages_scanned,
1930 unsigned long lru_pages);
9d0243bc 1931
7a9166e3
LY
1932#ifndef CONFIG_MMU
1933#define randomize_va_space 0
1934#else
a62eaf15 1935extern int randomize_va_space;
7a9166e3 1936#endif
a62eaf15 1937
045e72ac 1938const char * arch_vma_name(struct vm_area_struct *vma);
03252919 1939void print_vma_addr(char *prefix, unsigned long rip);
e6e5494c 1940
9bdac914
YL
1941void sparse_mem_maps_populate_node(struct page **map_map,
1942 unsigned long pnum_begin,
1943 unsigned long pnum_end,
1944 unsigned long map_count,
1945 int nodeid);
1946
98f3cfc1 1947struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
29c71111
AW
1948pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1949pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1950pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1951pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
8f6aac41 1952void *vmemmap_alloc_block(unsigned long size, int node);
9bdac914 1953void *vmemmap_alloc_block_buf(unsigned long size, int node);
8f6aac41 1954void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
0aad818b
JW
1955int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1956 int node);
1957int vmemmap_populate(unsigned long start, unsigned long end, int node);
c2b91e2e 1958void vmemmap_populate_print_last(void);
0197518c 1959#ifdef CONFIG_MEMORY_HOTPLUG
0aad818b 1960void vmemmap_free(unsigned long start, unsigned long end);
0197518c 1961#endif
46723bfa
YI
1962void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1963 unsigned long size);
6a46079c 1964
82ba011b
AK
1965enum mf_flags {
1966 MF_COUNT_INCREASED = 1 << 0,
7329bbeb 1967 MF_ACTION_REQUIRED = 1 << 1,
6751ed65 1968 MF_MUST_KILL = 1 << 2,
cf870c70 1969 MF_SOFT_OFFLINE = 1 << 3,
82ba011b 1970};
cd42f4a3 1971extern int memory_failure(unsigned long pfn, int trapno, int flags);
ea8f5fb8 1972extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
847ce401 1973extern int unpoison_memory(unsigned long pfn);
6a46079c
AK
1974extern int sysctl_memory_failure_early_kill;
1975extern int sysctl_memory_failure_recovery;
facb6011 1976extern void shake_page(struct page *p, int access);
293c07e3 1977extern atomic_long_t num_poisoned_pages;
facb6011 1978extern int soft_offline_page(struct page *page, int flags);
6a46079c 1979
718a3821
WF
1980extern void dump_page(struct page *page);
1981
47ad8475
AA
1982#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1983extern void clear_huge_page(struct page *page,
1984 unsigned long addr,
1985 unsigned int pages_per_huge_page);
1986extern void copy_user_huge_page(struct page *dst, struct page *src,
1987 unsigned long addr, struct vm_area_struct *vma,
1988 unsigned int pages_per_huge_page);
1989#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1990
c0a32fc5
SG
1991#ifdef CONFIG_DEBUG_PAGEALLOC
1992extern unsigned int _debug_guardpage_minorder;
1993
1994static inline unsigned int debug_guardpage_minorder(void)
1995{
1996 return _debug_guardpage_minorder;
1997}
1998
1999static inline bool page_is_guard(struct page *page)
2000{
2001 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2002}
2003#else
2004static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2005static inline bool page_is_guard(struct page *page) { return false; }
2006#endif /* CONFIG_DEBUG_PAGEALLOC */
2007
f9872caf
CS
2008#if MAX_NUMNODES > 1
2009void __init setup_nr_node_ids(void);
2010#else
2011static inline void setup_nr_node_ids(void) {}
2012#endif
2013
1da177e4
LT
2014#endif /* __KERNEL__ */
2015#endif /* _LINUX_MM_H */