blob: 065cdf8c09fb50fcce6ff67bcae848ac555c6b7a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/gfp.h>
9#include <linux/list.h>
Jiri Slaby59ea7462008-06-12 13:56:40 +020010#include <linux/mmdebug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/mmzone.h>
12#include <linux/rbtree.h>
13#include <linux/prio_tree.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070014#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070015#include <linux/mm_types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016
17struct mempolicy;
18struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040019struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040020struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040021struct writeback_control;
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
23#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24extern unsigned long max_mapnr;
25#endif
26
27extern unsigned long num_physpages;
28extern void * high_memory;
Linus Torvalds1da177e2005-04-16 15:20:36 -070029extern int page_cluster;
30
31#ifdef CONFIG_SYSCTL
32extern int sysctl_legacy_va_layout;
33#else
34#define sysctl_legacy_va_layout 0
35#endif
36
James Morris42d78962007-12-26 15:12:37 +110037extern unsigned long mmap_min_addr;
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <asm/page.h>
40#include <asm/pgtable.h>
41#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44
Andrea Righi27ac7922008-07-23 21:28:13 -070045/* to align the pointer to the (next) page boundary */
46#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
47
Linus Torvalds1da177e2005-04-16 15:20:36 -070048/*
49 * Linux kernel virtual memory manager primitives.
50 * The idea being to have a "virtual" mm in the same way
51 * we have a virtual fs - giving a cleaner interface to the
52 * mm details, and allowing different kinds of memory mappings
53 * (from shared memory to executable loading to arbitrary
54 * mmap() functions).
55 */
56
Christoph Lameterc43692e2006-12-06 20:32:48 -080057extern struct kmem_cache *vm_area_cachep;
58
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000060extern struct rb_root nommu_region_tree;
61extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63extern unsigned int kobjsize(const void *objp);
64#endif
65
66/*
Hugh Dickins605d9282008-08-16 11:07:21 +010067 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 */
69#define VM_READ 0x00000001 /* currently active flags */
70#define VM_WRITE 0x00000002
71#define VM_EXEC 0x00000004
72#define VM_SHARED 0x00000008
73
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070074/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
76#define VM_MAYWRITE 0x00000020
77#define VM_MAYEXEC 0x00000040
78#define VM_MAYSHARE 0x00000080
79
80#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
81#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080082#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
84
85#define VM_EXECUTABLE 0x00001000
86#define VM_LOCKED 0x00002000
87#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
88
89 /* Used by sys_madvise() */
90#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
91#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
92
93#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
94#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -080095#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -070097#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
99#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
100#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800101#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800102#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
Nick Piggind0217ac2007-07-19 01:47:03 -0700104#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700105#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000106#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Nick Piggind00806b2007-07-19 01:46:57 -0700107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
109#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
110#endif
111
112#ifdef CONFIG_STACK_GROWSUP
113#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
114#else
115#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
116#endif
117
118#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
119#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
120#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
121#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
122#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
123
124/*
Nick Pigginb291f002008-10-18 20:26:44 -0700125 * special vmas that are non-mergable, non-mlock()able
126 */
127#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
128
129/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 * 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
Nick Piggind0217ac2007-07-19 01:47:03 -0700135#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
136#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
137
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800138/*
139 * This interface is used by x86 PAT code to identify a pfn mapping that is
140 * linear over entire vma. This is to optimize PAT code that deals with
141 * marking the physical region with a particular prot. This is not for generic
142 * mm use. Note also that this check will not work if the pfn mapping is
143 * linear for a vma starting at physical address 0. In which case PAT code
144 * falls back to slow path of reserving physical range page by page.
145 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800146static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
147{
148 return ((vma->vm_flags & VM_PFNMAP) && vma->vm_pgoff);
149}
150
151static inline int is_pfn_mapping(struct vm_area_struct *vma)
152{
153 return (vma->vm_flags & VM_PFNMAP);
154}
Nick Piggind0217ac2007-07-19 01:47:03 -0700155
Nick Piggin54cb8822007-07-19 01:46:59 -0700156/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700157 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700158 * ->fault function. The vma's ->fault is responsible for returning a bitmask
159 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700160 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700161 * pgoff should be used in favour of virtual_address, if possible. If pgoff
162 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
163 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700164 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700165struct vm_fault {
166 unsigned int flags; /* FAULT_FLAG_xxx flags */
167 pgoff_t pgoff; /* Logical page offset based on vma */
168 void __user *virtual_address; /* Faulting virtual address */
169
170 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700171 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700172 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700173 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700174 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700175};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176
177/*
178 * These are the virtual MM functions - opening of an area, closing and
179 * unmapping it (needed to keep files on disk up-to-date etc), pointer
180 * to the functions called when a no-page or a wp-page exception occurs.
181 */
182struct vm_operations_struct {
183 void (*open)(struct vm_area_struct * area);
184 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700185 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700186
187 /* notification that a previously read-only page is about to become
188 * writable, if an error is returned it will cause a SIGBUS */
189 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700190
191 /* called by access_process_vm when get_user_pages() fails, typically
192 * for use by special VMAs that can switch between memory and hardware
193 */
194 int (*access)(struct vm_area_struct *vma, unsigned long addr,
195 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700197 /*
198 * set_policy() op must add a reference to any non-NULL @new mempolicy
199 * to hold the policy upon return. Caller should pass NULL @new to
200 * remove a policy and fall back to surrounding context--i.e. do not
201 * install a MPOL_DEFAULT policy, nor the task or system default
202 * mempolicy.
203 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700205
206 /*
207 * get_policy() op must add reference [mpol_get()] to any policy at
208 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
209 * in mm/mempolicy.c will do this automatically.
210 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
211 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
212 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
213 * must return NULL--i.e., do not "fallback" to task or system default
214 * policy.
215 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
217 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700218 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
219 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220#endif
221};
222
223struct mmu_gather;
224struct inode;
225
Andrew Morton349aef02006-01-08 01:04:36 -0800226#define page_private(page) ((page)->private)
227#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700228
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229/*
230 * FIXME: take this include out, include page-flags.h in
231 * files which need it (119 of them)
232 */
233#include <linux/page-flags.h>
234
235/*
236 * Methods to modify the page usage count.
237 *
238 * What counts for a page usage:
239 * - cache mapping (page->mapping)
240 * - private data (page->private)
241 * - page mapped in a task's page tables, each mapping
242 * is counted separately
243 *
244 * Also, many kernel routines increase the page count before a critical
245 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 */
247
248/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700249 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800251static inline int put_page_testzero(struct page *page)
252{
Nick Piggin725d7042006-09-25 23:30:55 -0700253 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800254 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800255}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
257/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800258 * Try to grab a ref unless the page has a refcount of zero, return false if
259 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800261static inline int get_page_unless_zero(struct page *page)
262{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800263 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800264}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Christoph Lameter48667e72008-02-04 22:28:31 -0800266/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800267struct page *vmalloc_to_page(const void *addr);
268unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800269
Paul Mundt0738c4b2008-03-12 16:51:31 +0900270/*
271 * Determine if an address is within the vmalloc range
272 *
273 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
274 * is no special casing required.
275 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800276static inline int is_vmalloc_addr(const void *x)
277{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900278#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800279 unsigned long addr = (unsigned long)x;
280
281 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900282#else
283 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800284#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900285}
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800286
Christoph Lameterd85f3382007-05-06 14:49:39 -0700287static inline struct page *compound_head(struct page *page)
288{
Christoph Lameter6d777952007-05-06 14:49:40 -0700289 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700290 return page->first_page;
291 return page;
292}
293
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700294static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700296 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297}
298
299static inline void get_page(struct page *page)
300{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700301 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700302 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 atomic_inc(&page->_count);
304}
305
Christoph Lameterb49af682007-05-06 14:49:41 -0700306static inline struct page *virt_to_head_page(const void *x)
307{
308 struct page *page = virt_to_page(x);
309 return compound_head(page);
310}
311
Nick Piggin7835e982006-03-22 00:08:40 -0800312/*
313 * Setup the page count before being freed into the page allocator for
314 * the first time (boot or memory hotplug)
315 */
316static inline void init_page_count(struct page *page)
317{
318 atomic_set(&page->_count, 1);
319}
320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700322void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800324void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800325
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800327 * Compound pages have a destructor function. Provide a
328 * prototype for that function and accessor functions.
329 * These are _only_ valid on the head of a PG_compound page.
330 */
331typedef void compound_page_dtor(struct page *);
332
333static inline void set_compound_page_dtor(struct page *page,
334 compound_page_dtor *dtor)
335{
336 page[1].lru.next = (void *)dtor;
337}
338
339static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
340{
341 return (compound_page_dtor *)page[1].lru.next;
342}
343
Christoph Lameterd85f3382007-05-06 14:49:39 -0700344static inline int compound_order(struct page *page)
345{
Christoph Lameter6d777952007-05-06 14:49:40 -0700346 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700347 return 0;
348 return (unsigned long)page[1].lru.prev;
349}
350
351static inline void set_compound_order(struct page *page, unsigned long order)
352{
353 page[1].lru.prev = (void *)order;
354}
355
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800356/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 * Multiple processes may "see" the same page. E.g. for untouched
358 * mappings of /dev/null, all processes see the same page full of
359 * zeroes, and text pages of executables and shared libraries have
360 * only one copy in memory, at most, normally.
361 *
362 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700363 * page_count() == 0 means the page is free. page->lru is then used for
364 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700365 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700367 * Pages are allocated by the slab allocator in order to provide memory
368 * to kmalloc and kmem_cache_alloc. In this case, the management of the
369 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
370 * unless a particular usage is carefully commented. (the responsibility of
371 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700373 * A page may be used by anyone else who does a __get_free_page().
374 * In this case, page_count still tracks the references, and should only
375 * be used through the normal accessor functions. The top bits of page->flags
376 * and page->virtual store page management information, but all other fields
377 * are unused and could be used privately, carefully. The management of this
378 * page is the responsibility of the one who allocated it, and those who have
379 * subsequently been given references to it.
380 *
381 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 * managed by the Linux memory manager: I/O, buffers, swapping etc.
383 * The following discussion applies only to them.
384 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700385 * A pagecache page contains an opaque `private' member, which belongs to the
386 * page's address_space. Usually, this is the address of a circular list of
387 * the page's disk buffers. PG_private must be set to tell the VM to call
388 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700390 * A page may belong to an inode's memory mapping. In this case, page->mapping
391 * is the pointer to the inode, and page->index is the file offset of the page,
392 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700394 * If pagecache pages are not associated with an inode, they are said to be
395 * anonymous pages. These may become associated with the swapcache, and in that
396 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700398 * In either case (swapcache or inode backed), the pagecache itself holds one
399 * reference to the page. Setting PG_private should also increment the
400 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700402 * The pagecache pages are stored in a per-mapping radix tree, which is
403 * rooted at mapping->page_tree, and indexed by offset.
404 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
405 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700407 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 * - inode pages may need to be read from disk,
409 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700410 * to be written back to the inode on disk,
411 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
412 * modified may need to be swapped out to swap space and (later) to be read
413 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 */
415
416/*
417 * The zone field is never updated after free_area_init_core()
418 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700420
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700421
422/*
423 * page->flags layout:
424 *
425 * There are three possibilities for how page->flags get
426 * laid out. The first is for the normal case, without
427 * sparsemem. The second is for sparsemem when there is
428 * plenty of space for node and section. The last is when
429 * we have run out of space and have to fall back to an
430 * alternate (slower) way of determining the node.
431 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700432 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
433 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
434 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700435 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700436#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700437#define SECTIONS_WIDTH SECTIONS_SHIFT
438#else
439#define SECTIONS_WIDTH 0
440#endif
441
442#define ZONES_WIDTH ZONES_SHIFT
443
Christoph Lameter9223b412008-04-28 02:12:48 -0700444#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700445#define NODES_WIDTH NODES_SHIFT
446#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700447#ifdef CONFIG_SPARSEMEM_VMEMMAP
448#error "Vmemmap: No space for nodes field in page flags"
449#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700450#define NODES_WIDTH 0
451#endif
452
453/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100454#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700455#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
456#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
457
458/*
459 * We are going to use the flags for the page to node mapping if its in
460 * there. This includes the case where there is no node, so it is implicit.
461 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800462#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
463#define NODE_NOT_IN_PAGE_FLAGS
464#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700465
466#ifndef PFN_SECTION_SHIFT
467#define PFN_SECTION_SHIFT 0
468#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700469
470/*
471 * Define the bit shifts to access each section. For non-existant
472 * sections we define the shift as 0; that plus a 0 mask ensures
473 * the compiler will optimise away reference to them.
474 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700475#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
476#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
477#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700478
Christoph Lameter89689ae2006-12-06 20:31:45 -0800479/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
480#ifdef NODE_NOT_IN_PAGEFLAGS
481#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800482#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
483 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700484#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800485#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800486#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
487 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700488#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800489
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800490#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700491
Christoph Lameter9223b412008-04-28 02:12:48 -0700492#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
493#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700494#endif
495
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700496#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
497#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
498#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800499#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700500
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700501static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
Dave Hansen348f8b62005-06-23 00:07:40 -0700503 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505
Christoph Lameter89689ae2006-12-06 20:31:45 -0800506/*
507 * The identification function is only used by the buddy allocator for
508 * determining if two pages could be buddies. We are not really
509 * identifying a zone since we could be using a the section number
510 * id if we have not node id available in page flags.
511 * We guarantee only that it will return the same value for two
512 * combinable pages in a zone.
513 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700514static inline int page_zone_id(struct page *page)
515{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800516 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517}
518
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800519static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700520{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700521#ifdef CONFIG_NUMA
522 return zone->node;
523#else
524 return 0;
525#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700526}
527
Christoph Lameter89689ae2006-12-06 20:31:45 -0800528#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800529extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800530#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800531static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700532{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800533 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700534}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800535#endif
536
537static inline struct zone *page_zone(struct page *page)
538{
539 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
540}
541
Christoph Lameter308c05e2008-04-28 02:12:43 -0700542#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700543static inline unsigned long page_to_section(struct page *page)
544{
545 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
546}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700547#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700548
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700549static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
Dave Hansen348f8b62005-06-23 00:07:40 -0700551 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
552 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
553}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700554
Dave Hansen348f8b62005-06-23 00:07:40 -0700555static inline void set_page_node(struct page *page, unsigned long node)
556{
557 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
558 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
559}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800560
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700561static inline void set_page_section(struct page *page, unsigned long section)
562{
563 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
564 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
565}
Dave Hansen348f8b62005-06-23 00:07:40 -0700566
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700567static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700568 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700569{
570 set_page_zone(page, zone);
571 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700572 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573}
574
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700575/*
Eric Paris7cd94142007-11-26 18:47:40 -0500576 * If a hint addr is less than mmap_min_addr change hint to be as
577 * low as possible but still greater than mmap_min_addr
578 */
579static inline unsigned long round_hint_to_min(unsigned long hint)
580{
581#ifdef CONFIG_SECURITY
582 hint &= PAGE_MASK;
583 if (((void *)hint != NULL) &&
584 (hint < mmap_min_addr))
585 return PAGE_ALIGN(mmap_min_addr);
586#endif
587 return hint;
588}
589
590/*
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700591 * Some inline functions in vmstat.h depend on page_zone()
592 */
593#include <linux/vmstat.h>
594
Ingo Molnar652050a2006-01-14 13:21:30 -0800595static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596{
597 return __va(page_to_pfn(page) << PAGE_SHIFT);
598}
599
600#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
601#define HASHED_PAGE_VIRTUAL
602#endif
603
604#if defined(WANT_PAGE_VIRTUAL)
605#define page_address(page) ((page)->virtual)
606#define set_page_address(page, address) \
607 do { \
608 (page)->virtual = (address); \
609 } while(0)
610#define page_address_init() do { } while(0)
611#endif
612
613#if defined(HASHED_PAGE_VIRTUAL)
614void *page_address(struct page *page);
615void set_page_address(struct page *page, void *virtual);
616void page_address_init(void);
617#endif
618
619#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
620#define page_address(page) lowmem_page_address(page)
621#define set_page_address(page, address) do { } while(0)
622#define page_address_init() do { } while(0)
623#endif
624
625/*
626 * On an anonymous page mapped into a user virtual memory area,
627 * page->mapping points to its anon_vma, not to a struct address_space;
628 * with the PAGE_MAPPING_ANON bit set to distinguish it.
629 *
630 * Please note that, confusingly, "page_mapping" refers to the inode
631 * address_space which maps the page from disk; whereas "page_mapped"
632 * refers to user virtual address space into which the page is mapped.
633 */
634#define PAGE_MAPPING_ANON 1
635
636extern struct address_space swapper_space;
637static inline struct address_space *page_mapping(struct page *page)
638{
639 struct address_space *mapping = page->mapping;
640
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700641 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700642#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 if (unlikely(PageSwapCache(page)))
644 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700645 else
646#endif
647 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 mapping = NULL;
649 return mapping;
650}
651
652static inline int PageAnon(struct page *page)
653{
654 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
655}
656
657/*
658 * Return the pagecache index of the passed page. Regular pagecache pages
659 * use ->index whereas swapcache pages use ->private
660 */
661static inline pgoff_t page_index(struct page *page)
662{
663 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700664 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 return page->index;
666}
667
668/*
669 * The atomic page->_mapcount, like _count, starts from -1:
670 * so that transitions both from it and to it can be tracked,
671 * using atomic_inc_and_test and atomic_add_negative(-1).
672 */
673static inline void reset_page_mapcount(struct page *page)
674{
675 atomic_set(&(page)->_mapcount, -1);
676}
677
678static inline int page_mapcount(struct page *page)
679{
680 return atomic_read(&(page)->_mapcount) + 1;
681}
682
683/*
684 * Return true if this page is mapped into pagetables.
685 */
686static inline int page_mapped(struct page *page)
687{
688 return atomic_read(&(page)->_mapcount) >= 0;
689}
690
691/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 * Different kinds of faults, as returned by handle_mm_fault().
693 * Used to decide whether a process gets delivered SIGBUS or
694 * just gets major/minor fault counters bumped up.
695 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700696
Nick Piggin83c54072007-07-19 01:47:05 -0700697#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700698
Nick Piggin83c54072007-07-19 01:47:05 -0700699#define VM_FAULT_OOM 0x0001
700#define VM_FAULT_SIGBUS 0x0002
701#define VM_FAULT_MAJOR 0x0004
702#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000703
Nick Piggin83c54072007-07-19 01:47:05 -0700704#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
705#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
Nick Piggin83c54072007-07-19 01:47:05 -0700707#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700708
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800709/*
710 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
711 */
712extern void pagefault_out_of_memory(void);
713
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
715
716extern void show_free_areas(void);
717
718#ifdef CONFIG_SHMEM
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700719extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720#else
Russell King03b00eb2006-01-06 00:10:52 -0800721static inline int shmem_lock(struct file *file, int lock,
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700722 struct user_struct *user)
Russell King03b00eb2006-01-06 00:10:52 -0800723{
724 return 0;
725}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726#endif
727struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
728
729int shmem_zero_setup(struct vm_area_struct *);
730
David Howellsb0e15192006-01-06 00:11:42 -0800731#ifndef CONFIG_MMU
732extern unsigned long shmem_get_unmapped_area(struct file *file,
733 unsigned long addr,
734 unsigned long len,
735 unsigned long pgoff,
736 unsigned long flags);
737#endif
738
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400739extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740extern int user_shm_lock(size_t, struct user_struct *);
741extern void user_shm_unlock(size_t, struct user_struct *);
742
743/*
744 * Parameter block passed down to zap_pte_range in exceptional cases.
745 */
746struct zap_details {
747 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
748 struct address_space *check_mapping; /* Check page->mapping if set */
749 pgoff_t first_index; /* Lowest page->index to unmap */
750 pgoff_t last_index; /* Highest page->index to unmap */
751 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 unsigned long truncate_count; /* Compare vm_truncate_count */
753};
754
Nick Piggin7e675132008-04-28 02:13:00 -0700755struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
756 pte_t pte);
757
Jack Steinerc627f9c2008-07-29 22:33:53 -0700758int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
759 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700760unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700762unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 struct vm_area_struct *start_vma, unsigned long start_addr,
764 unsigned long end_addr, unsigned long *nr_accounted,
765 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800766
767/**
768 * mm_walk - callbacks for walk_page_range
769 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
770 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
771 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
772 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
773 * @pte_hole: if set, called for each hole at all levels
774 *
775 * (see walk_page_range for more details)
776 */
777struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700778 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
779 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
780 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
781 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
782 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
783 struct mm_struct *mm;
784 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800785};
786
Dave Hansen21650092008-06-12 15:21:47 -0700787int walk_page_range(unsigned long addr, unsigned long end,
788 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700789void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700790 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
792 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793void unmap_mapping_range(struct address_space *mapping,
794 loff_t const holebegin, loff_t const holelen, int even_cows);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800795int follow_phys(struct vm_area_struct *vma, unsigned long address,
796 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700797int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
798 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799
800static inline void unmap_shared_mapping_range(struct address_space *mapping,
801 loff_t const holebegin, loff_t const holelen)
802{
803 unmap_mapping_range(mapping, holebegin, holelen, 0);
804}
805
806extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800807extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000808
David Howells7ee1dd32006-01-06 00:11:44 -0800809#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700810extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
David Howells7ee1dd32006-01-06 00:11:44 -0800811 unsigned long address, int write_access);
David Howells7ee1dd32006-01-06 00:11:44 -0800812#else
813static inline int handle_mm_fault(struct mm_struct *mm,
814 struct vm_area_struct *vma, unsigned long address,
815 int write_access)
816{
817 /* should never happen if there's no MMU */
818 BUG();
819 return VM_FAULT_SIGBUS;
820}
821#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823extern int make_pages_present(unsigned long addr, unsigned long end);
824extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
826int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
827 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
828
David Howellscf9a2ae2006-08-29 19:05:54 +0100829extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
830extern void do_invalidatepage(struct page *page, unsigned long offset);
831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800833int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834int redirty_page_for_writepage(struct writeback_control *wbc,
835 struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800836int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837int set_page_dirty_lock(struct page *page);
838int clear_page_dirty_for_io(struct page *page);
839
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700840extern unsigned long move_page_tables(struct vm_area_struct *vma,
841 unsigned long old_addr, struct vm_area_struct *new_vma,
842 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843extern unsigned long do_mremap(unsigned long addr,
844 unsigned long old_len, unsigned long new_len,
845 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700846extern int mprotect_fixup(struct vm_area_struct *vma,
847 struct vm_area_struct **pprev, unsigned long start,
848 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Nick Piggin21cc1992008-07-25 19:45:22 -0700850/*
851 * get_user_pages_fast provides equivalent functionality to get_user_pages,
852 * operating on current and current->mm (force=0 and doesn't return any vmas).
853 *
854 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
855 * can be made about locking. get_user_pages_fast is to be implemented in a
856 * way that is advantageous (vs get_user_pages()) when the user memory area is
857 * already faulted in and present in ptes. However if the pages have to be
858 * faulted in, it may turn out to be slightly slower).
859 */
860int get_user_pages_fast(unsigned long start, int nr_pages, int write,
861 struct page **pages);
862
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700864 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700866 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
867 * look through the least-recently-used 'nr_to_scan' entries and
868 * attempt to free them up. It should return the number of objects
869 * which remain in the cache. If it returns -1, it means it cannot do
870 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700872 * The 'gfpmask' refers to the allocation we are currently trying to
873 * fulfil.
874 *
875 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
876 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700878struct shrinker {
879 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
880 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
Rusty Russell8e1f9362007-07-17 04:03:17 -0700882 /* These are for internal use */
883 struct list_head list;
884 long nr; /* objs pending delete */
885};
886#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
887extern void register_shrinker(struct shrinker *);
888extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400890int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700891
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800892extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800893
Nick Piggin5f22df02007-05-06 14:49:02 -0700894#ifdef __PAGETABLE_PUD_FOLDED
895static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
896 unsigned long address)
897{
898 return 0;
899}
900#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700901int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700902#endif
903
904#ifdef __PAGETABLE_PMD_FOLDED
905static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
906 unsigned long address)
907{
908 return 0;
909}
910#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700911int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700912#endif
913
Hugh Dickins1bb36302005-10-29 18:16:22 -0700914int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
915int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
916
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917/*
918 * The following ifdef needed to get the 4level-fixup.h header to work.
919 * Remove it when 4level-fixup.h has been removed.
920 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700921#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
923{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700924 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
925 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926}
927
928static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
929{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700930 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
931 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700933#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
934
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700935#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700936/*
937 * We tuck a spinlock to guard each pagetable page into its struct page,
938 * at page->private, with BUILD_BUG_ON to make sure that this will not
939 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
940 * When freeing, reset page->mapping so free_pages_check won't complain.
941 */
Andrew Morton349aef02006-01-08 01:04:36 -0800942#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700943#define pte_lock_init(_page) do { \
944 spin_lock_init(__pte_lockptr(_page)); \
945} while (0)
946#define pte_lock_deinit(page) ((page)->mapping = NULL)
947#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700948#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700949/*
950 * We use mm->page_table_lock to guard all pagetable pages of the mm.
951 */
952#define pte_lock_init(page) do {} while (0)
953#define pte_lock_deinit(page) do {} while (0)
954#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700955#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700956
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800957static inline void pgtable_page_ctor(struct page *page)
958{
959 pte_lock_init(page);
960 inc_zone_page_state(page, NR_PAGETABLE);
961}
962
963static inline void pgtable_page_dtor(struct page *page)
964{
965 pte_lock_deinit(page);
966 dec_zone_page_state(page, NR_PAGETABLE);
967}
968
Hugh Dickinsc74df322005-10-29 18:16:23 -0700969#define pte_offset_map_lock(mm, pmd, address, ptlp) \
970({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700971 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700972 pte_t *__pte = pte_offset_map(pmd, address); \
973 *(ptlp) = __ptl; \
974 spin_lock(__ptl); \
975 __pte; \
976})
977
978#define pte_unmap_unlock(pte, ptl) do { \
979 spin_unlock(ptl); \
980 pte_unmap(pte); \
981} while (0)
982
Hugh Dickins1bb36302005-10-29 18:16:22 -0700983#define pte_alloc_map(mm, pmd, address) \
984 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
985 NULL: pte_offset_map(pmd, address))
986
Hugh Dickinsc74df322005-10-29 18:16:23 -0700987#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
988 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
989 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
990
Hugh Dickins1bb36302005-10-29 18:16:22 -0700991#define pte_alloc_kernel(pmd, address) \
992 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
993 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
995extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -0700996extern void free_area_init_node(int nid, unsigned long * zones_size,
997 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700998#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
999/*
1000 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1001 * zones, allocate the backing mem_map and account for memory holes in a more
1002 * architecture independent manner. This is a substitute for creating the
1003 * zone_sizes[] and zholes_size[] arrays and passing them to
1004 * free_area_init_node()
1005 *
1006 * An architecture is expected to register range of page frames backed by
1007 * physical memory with add_active_range() before calling
1008 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1009 * usage, an architecture is expected to do something like
1010 *
1011 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1012 * max_highmem_pfn};
1013 * for_each_valid_physical_page_range()
1014 * add_active_range(node_id, start_pfn, end_pfn)
1015 * free_area_init_nodes(max_zone_pfns);
1016 *
1017 * If the architecture guarantees that there are no holes in the ranges
1018 * registered with add_active_range(), free_bootmem_active_regions()
1019 * will call free_bootmem_node() for each registered physical page range.
1020 * Similarly sparse_memory_present_with_active_regions() calls
1021 * memory_present() for each range when SPARSEMEM is enabled.
1022 *
1023 * See mm/page_alloc.c for more information on each function exposed by
1024 * CONFIG_ARCH_POPULATES_NODE_MAP
1025 */
1026extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1027extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1028 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001029extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1030 unsigned long end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -07001031extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1032 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001033extern void remove_all_active_ranges(void);
1034extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1035 unsigned long end_pfn);
1036extern void get_pfn_range_for_nid(unsigned int nid,
1037 unsigned long *start_pfn, unsigned long *end_pfn);
1038extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001039extern void free_bootmem_with_active_regions(int nid,
1040 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001041typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001042extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001043extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001044#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001045
1046#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1047 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1048static inline int __early_pfn_to_nid(unsigned long pfn)
1049{
1050 return 0;
1051}
1052#else
1053/* please see mm/page_alloc.c */
1054extern int __meminit early_pfn_to_nid(unsigned long pfn);
1055#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1056/* there is a per-arch backend function. */
1057extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1058#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1059#endif
1060
Mel Gorman0e0b8642006-09-27 01:49:56 -07001061extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001062extern void memmap_init_zone(unsigned long, int, unsigned long,
1063 unsigned long, enum memmap_context);
Dave Hansen3947be12005-10-29 18:16:54 -07001064extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001066extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067extern void show_mem(void);
1068extern void si_meminfo(struct sysinfo * val);
1069extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001070extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001072#ifdef CONFIG_NUMA
1073extern void setup_per_cpu_pageset(void);
1074#else
1075static inline void setup_per_cpu_pageset(void) {}
1076#endif
1077
David Howells8feae132009-01-08 12:04:47 +00001078/* nommu.c */
1079extern atomic_t mmap_pages_allocated;
1080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081/* prio_tree.c */
1082void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1083void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1084void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1085struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1086 struct prio_tree_iter *iter);
1087
1088#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1089 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1090 (vma = vma_prio_tree_next(vma, iter)); )
1091
1092static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1093 struct list_head *list)
1094{
1095 vma->shared.vm_set.parent = NULL;
1096 list_add_tail(&vma->shared.vm_set.list, list);
1097}
1098
1099/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001100extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1102 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1103extern struct vm_area_struct *vma_merge(struct mm_struct *,
1104 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1105 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1106 struct mempolicy *);
1107extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1108extern int split_vma(struct mm_struct *,
1109 struct vm_area_struct *, unsigned long addr, int new_below);
1110extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1111extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1112 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001113extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1115 unsigned long addr, unsigned long len, pgoff_t pgoff);
1116extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001117
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001118extern int mm_take_all_locks(struct mm_struct *mm);
1119extern void mm_drop_all_locks(struct mm_struct *mm);
1120
Matt Helsley925d1c42008-04-29 01:01:36 -07001121#ifdef CONFIG_PROC_FS
1122/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1123extern void added_exe_file_vma(struct mm_struct *mm);
1124extern void removed_exe_file_vma(struct mm_struct *mm);
1125#else
1126static inline void added_exe_file_vma(struct mm_struct *mm)
1127{}
1128
1129static inline void removed_exe_file_vma(struct mm_struct *mm)
1130{}
1131#endif /* CONFIG_PROC_FS */
1132
akpm@osdl.org119f6572005-05-01 08:58:35 -07001133extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001134extern int install_special_mapping(struct mm_struct *mm,
1135 unsigned long addr, unsigned long len,
1136 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137
1138extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1139
1140extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1141 unsigned long len, unsigned long prot,
1142 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001143extern unsigned long mmap_region(struct file *file, unsigned long addr,
1144 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001145 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
1147static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1148 unsigned long len, unsigned long prot,
1149 unsigned long flag, unsigned long offset)
1150{
1151 unsigned long ret = -EINVAL;
1152 if ((offset + PAGE_ALIGN(len)) < offset)
1153 goto out;
1154 if (!(offset & ~PAGE_MASK))
1155 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1156out:
1157 return ret;
1158}
1159
1160extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1161
1162extern unsigned long do_brk(unsigned long, unsigned long);
1163
1164/* filemap.c */
1165extern unsigned long page_unuse(struct page *);
1166extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001167extern void truncate_inode_pages_range(struct address_space *,
1168 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
1170/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001171extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
1173/* mm/page-writeback.c */
1174int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001175void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176
1177/* readahead.c */
1178#define VM_MAX_READAHEAD 128 /* kbytes */
1179#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
1181int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001182 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001184 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001185
1186void page_cache_sync_readahead(struct address_space *mapping,
1187 struct file_ra_state *ra,
1188 struct file *filp,
1189 pgoff_t offset,
1190 unsigned long size);
1191
1192void page_cache_async_readahead(struct address_space *mapping,
1193 struct file_ra_state *ra,
1194 struct file *filp,
1195 struct page *pg,
1196 pgoff_t offset,
1197 unsigned long size);
1198
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199unsigned long max_sane_readahead(unsigned long nr);
1200
1201/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001202extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001203#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001204extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001205#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001206extern int expand_stack_downwards(struct vm_area_struct *vma,
1207 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
1209/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1210extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1211extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1212 struct vm_area_struct **pprev);
1213
1214/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1215 NULL if none. Assume start_addr < end_addr. */
1216static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1217{
1218 struct vm_area_struct * vma = find_vma(mm,start_addr);
1219
1220 if (vma && end_addr <= vma->vm_start)
1221 vma = NULL;
1222 return vma;
1223}
1224
1225static inline unsigned long vma_pages(struct vm_area_struct *vma)
1226{
1227 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1228}
1229
Hugh Dickins804af2c2006-07-26 21:39:49 +01001230pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001231struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001232int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1233 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001234int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001235int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1236 unsigned long pfn);
Nick Piggin423bad602008-04-28 02:13:01 -07001237int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1238 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
Linus Torvalds6aab3412005-11-28 14:34:23 -08001240struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001241 unsigned int foll_flags);
1242#define FOLL_WRITE 0x01 /* check pte is writable */
1243#define FOLL_TOUCH 0x02 /* mark page accessed */
1244#define FOLL_GET 0x04 /* do get_page on page */
1245#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001247typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001248 void *data);
1249extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1250 unsigned long size, pte_fn_t fn, void *data);
1251
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001253void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001255static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 unsigned long flags, struct file *file, long pages)
1257{
1258}
1259#endif /* CONFIG_PROC_FS */
1260
Ingo Molnar12d6f212008-01-30 13:33:58 +01001261#ifdef CONFIG_DEBUG_PAGEALLOC
1262extern int debug_pagealloc_enabled;
1263
1264extern void kernel_map_pages(struct page *page, int numpages, int enable);
1265
1266static inline void enable_debug_pagealloc(void)
1267{
1268 debug_pagealloc_enabled = 1;
1269}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001270#ifdef CONFIG_HIBERNATION
1271extern bool kernel_page_present(struct page *page);
1272#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001273#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001275kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001276static inline void enable_debug_pagealloc(void)
1277{
1278}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001279#ifdef CONFIG_HIBERNATION
1280static inline bool kernel_page_present(struct page *page) { return true; }
1281#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282#endif
1283
1284extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1285#ifdef __HAVE_ARCH_GATE_AREA
1286int in_gate_area_no_task(unsigned long addr);
1287int in_gate_area(struct task_struct *task, unsigned long addr);
1288#else
1289int in_gate_area_no_task(unsigned long addr);
1290#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1291#endif /* __HAVE_ARCH_GATE_AREA */
1292
Andrew Morton9d0243b2006-01-08 01:00:39 -08001293int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1294 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001295unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001296 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001297
Luke Yang7a9166e2006-02-20 18:28:07 -08001298#ifndef CONFIG_MMU
1299#define randomize_va_space 0
1300#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001301extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001302#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001303
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001304const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001305void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001306
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001307struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001308pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1309pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1310pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1311pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001312void *vmemmap_alloc_block(unsigned long size, int node);
1313void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001314int vmemmap_populate_basepages(struct page *start_page,
1315 unsigned long pages, int node);
1316int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001317void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001318
Markus Metzgerc5dee612008-12-19 15:17:02 +01001319extern void *alloc_locked_buffer(size_t size);
1320extern void free_locked_buffer(void *buffer, size_t size);
Markus Metzger9f339e72009-02-11 15:10:27 +01001321extern void release_locked_buffer(void *buffer, size_t size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322#endif /* __KERNEL__ */
1323#endif /* _LINUX_MM_H */