blob: d808cf832c4de5d479d2c4a704b291934a4f1a3f [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>
10#include <linux/mmzone.h>
11#include <linux/rbtree.h>
12#include <linux/prio_tree.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070013#include <linux/debug_locks.h>
Heiko Carstens5b99cd02006-09-27 01:50:01 -070014#include <linux/mm_types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015
16struct mempolicy;
17struct anon_vma;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040018struct file_ra_state;
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040019struct user_struct;
Alexey Dobriyan4e950f62007-07-30 02:36:13 +040020struct writeback_control;
Markus Metzger1cb81b12009-04-24 09:51:43 +020021struct rlimit;
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
37#include <asm/page.h>
38#include <asm/pgtable.h>
39#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
42
Andrea Righi27ac7922008-07-23 21:28:13 -070043/* to align the pointer to the (next) page boundary */
44#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
45
Linus Torvalds1da177e2005-04-16 15:20:36 -070046/*
47 * Linux kernel virtual memory manager primitives.
48 * The idea being to have a "virtual" mm in the same way
49 * we have a virtual fs - giving a cleaner interface to the
50 * mm details, and allowing different kinds of memory mappings
51 * (from shared memory to executable loading to arbitrary
52 * mmap() functions).
53 */
54
Christoph Lameterc43692e2006-12-06 20:32:48 -080055extern struct kmem_cache *vm_area_cachep;
56
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#ifndef CONFIG_MMU
David Howells8feae132009-01-08 12:04:47 +000058extern struct rb_root nommu_region_tree;
59extern struct rw_semaphore nommu_region_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
61extern unsigned int kobjsize(const void *objp);
62#endif
63
64/*
Hugh Dickins605d9282008-08-16 11:07:21 +010065 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 */
67#define VM_READ 0x00000001 /* currently active flags */
68#define VM_WRITE 0x00000002
69#define VM_EXEC 0x00000004
70#define VM_SHARED 0x00000008
71
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070072/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
74#define VM_MAYWRITE 0x00000020
75#define VM_MAYEXEC 0x00000040
76#define VM_MAYSHARE 0x00000080
77
78#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
79#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080080#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
82
83#define VM_EXECUTABLE 0x00001000
84#define VM_LOCKED 0x00002000
85#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
86
87 /* Used by sys_madvise() */
88#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
89#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
90
91#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
92#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -080093#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -070095#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
97#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
98#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -070099#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800100#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
Nick Piggind0217ac2007-07-19 01:47:03 -0700102#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700103#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000104#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700105#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700106#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
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 */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700137#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Nick Piggind0217ac2007-07-19 01:47:03 -0700138
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800139/*
140 * This interface is used by x86 PAT code to identify a pfn mapping that is
141 * linear over entire vma. This is to optimize PAT code that deals with
142 * marking the physical region with a particular prot. This is not for generic
143 * mm use. Note also that this check will not work if the pfn mapping is
144 * linear for a vma starting at physical address 0. In which case PAT code
145 * falls back to slow path of reserving physical range page by page.
146 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800147static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
148{
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700149 return (vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800150}
151
152static inline int is_pfn_mapping(struct vm_area_struct *vma)
153{
154 return (vma->vm_flags & VM_PFNMAP);
155}
Nick Piggind0217ac2007-07-19 01:47:03 -0700156
Nick Piggin54cb8822007-07-19 01:46:59 -0700157/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700158 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700159 * ->fault function. The vma's ->fault is responsible for returning a bitmask
160 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700161 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700162 * pgoff should be used in favour of virtual_address, if possible. If pgoff
163 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
164 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700165 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700166struct vm_fault {
167 unsigned int flags; /* FAULT_FLAG_xxx flags */
168 pgoff_t pgoff; /* Logical page offset based on vma */
169 void __user *virtual_address; /* Faulting virtual address */
170
171 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700172 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700173 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700174 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700175 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700176};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178/*
179 * These are the virtual MM functions - opening of an area, closing and
180 * unmapping it (needed to keep files on disk up-to-date etc), pointer
181 * to the functions called when a no-page or a wp-page exception occurs.
182 */
183struct vm_operations_struct {
184 void (*open)(struct vm_area_struct * area);
185 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700186 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700187
188 /* notification that a previously read-only page is about to become
189 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700190 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700191
192 /* called by access_process_vm when get_user_pages() fails, typically
193 * for use by special VMAs that can switch between memory and hardware
194 */
195 int (*access)(struct vm_area_struct *vma, unsigned long addr,
196 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700198 /*
199 * set_policy() op must add a reference to any non-NULL @new mempolicy
200 * to hold the policy upon return. Caller should pass NULL @new to
201 * remove a policy and fall back to surrounding context--i.e. do not
202 * install a MPOL_DEFAULT policy, nor the task or system default
203 * mempolicy.
204 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700206
207 /*
208 * get_policy() op must add reference [mpol_get()] to any policy at
209 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
210 * in mm/mempolicy.c will do this automatically.
211 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
212 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
213 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
214 * must return NULL--i.e., do not "fallback" to task or system default
215 * policy.
216 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
218 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700219 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
220 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221#endif
222};
223
224struct mmu_gather;
225struct inode;
226
Andrew Morton349aef02006-01-08 01:04:36 -0800227#define page_private(page) ((page)->private)
228#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230/*
231 * FIXME: take this include out, include page-flags.h in
232 * files which need it (119 of them)
233 */
234#include <linux/page-flags.h>
235
236/*
237 * Methods to modify the page usage count.
238 *
239 * What counts for a page usage:
240 * - cache mapping (page->mapping)
241 * - private data (page->private)
242 * - page mapped in a task's page tables, each mapping
243 * is counted separately
244 *
245 * Also, many kernel routines increase the page count before a critical
246 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 */
248
249/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700250 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800252static inline int put_page_testzero(struct page *page)
253{
Nick Piggin725d7042006-09-25 23:30:55 -0700254 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800255 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800256}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800259 * Try to grab a ref unless the page has a refcount of zero, return false if
260 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800262static inline int get_page_unless_zero(struct page *page)
263{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800264 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800265}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
Christoph Lameter48667e72008-02-04 22:28:31 -0800267/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800268struct page *vmalloc_to_page(const void *addr);
269unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800270
Paul Mundt0738c4b2008-03-12 16:51:31 +0900271/*
272 * Determine if an address is within the vmalloc range
273 *
274 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
275 * is no special casing required.
276 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800277static inline int is_vmalloc_addr(const void *x)
278{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900279#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800280 unsigned long addr = (unsigned long)x;
281
282 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900283#else
284 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800285#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900286}
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800287
Christoph Lameterd85f3382007-05-06 14:49:39 -0700288static inline struct page *compound_head(struct page *page)
289{
Christoph Lameter6d777952007-05-06 14:49:40 -0700290 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700291 return page->first_page;
292 return page;
293}
294
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700295static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700297 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298}
299
300static inline void get_page(struct page *page)
301{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700302 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700303 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 atomic_inc(&page->_count);
305}
306
Christoph Lameterb49af682007-05-06 14:49:41 -0700307static inline struct page *virt_to_head_page(const void *x)
308{
309 struct page *page = virt_to_page(x);
310 return compound_head(page);
311}
312
Nick Piggin7835e982006-03-22 00:08:40 -0800313/*
314 * Setup the page count before being freed into the page allocator for
315 * the first time (boot or memory hotplug)
316 */
317static inline void init_page_count(struct page *page)
318{
319 atomic_set(&page->_count, 1);
320}
321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700323void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800325void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800326
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800328 * Compound pages have a destructor function. Provide a
329 * prototype for that function and accessor functions.
330 * These are _only_ valid on the head of a PG_compound page.
331 */
332typedef void compound_page_dtor(struct page *);
333
334static inline void set_compound_page_dtor(struct page *page,
335 compound_page_dtor *dtor)
336{
337 page[1].lru.next = (void *)dtor;
338}
339
340static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
341{
342 return (compound_page_dtor *)page[1].lru.next;
343}
344
Christoph Lameterd85f3382007-05-06 14:49:39 -0700345static inline int compound_order(struct page *page)
346{
Christoph Lameter6d777952007-05-06 14:49:40 -0700347 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700348 return 0;
349 return (unsigned long)page[1].lru.prev;
350}
351
352static inline void set_compound_order(struct page *page, unsigned long order)
353{
354 page[1].lru.prev = (void *)order;
355}
356
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800357/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 * Multiple processes may "see" the same page. E.g. for untouched
359 * mappings of /dev/null, all processes see the same page full of
360 * zeroes, and text pages of executables and shared libraries have
361 * only one copy in memory, at most, normally.
362 *
363 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700364 * page_count() == 0 means the page is free. page->lru is then used for
365 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700366 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700368 * Pages are allocated by the slab allocator in order to provide memory
369 * to kmalloc and kmem_cache_alloc. In this case, the management of the
370 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
371 * unless a particular usage is carefully commented. (the responsibility of
372 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700374 * A page may be used by anyone else who does a __get_free_page().
375 * In this case, page_count still tracks the references, and should only
376 * be used through the normal accessor functions. The top bits of page->flags
377 * and page->virtual store page management information, but all other fields
378 * are unused and could be used privately, carefully. The management of this
379 * page is the responsibility of the one who allocated it, and those who have
380 * subsequently been given references to it.
381 *
382 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 * managed by the Linux memory manager: I/O, buffers, swapping etc.
384 * The following discussion applies only to them.
385 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700386 * A pagecache page contains an opaque `private' member, which belongs to the
387 * page's address_space. Usually, this is the address of a circular list of
388 * the page's disk buffers. PG_private must be set to tell the VM to call
389 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700391 * A page may belong to an inode's memory mapping. In this case, page->mapping
392 * is the pointer to the inode, and page->index is the file offset of the page,
393 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700395 * If pagecache pages are not associated with an inode, they are said to be
396 * anonymous pages. These may become associated with the swapcache, and in that
397 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700399 * In either case (swapcache or inode backed), the pagecache itself holds one
400 * reference to the page. Setting PG_private should also increment the
401 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700403 * The pagecache pages are stored in a per-mapping radix tree, which is
404 * rooted at mapping->page_tree, and indexed by offset.
405 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
406 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700408 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 * - inode pages may need to be read from disk,
410 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700411 * to be written back to the inode on disk,
412 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
413 * modified may need to be swapped out to swap space and (later) to be read
414 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 */
416
417/*
418 * The zone field is never updated after free_area_init_core()
419 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700421
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700422
423/*
424 * page->flags layout:
425 *
426 * There are three possibilities for how page->flags get
427 * laid out. The first is for the normal case, without
428 * sparsemem. The second is for sparsemem when there is
429 * plenty of space for node and section. The last is when
430 * we have run out of space and have to fall back to an
431 * alternate (slower) way of determining the node.
432 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700433 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
434 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
435 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700436 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700437#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700438#define SECTIONS_WIDTH SECTIONS_SHIFT
439#else
440#define SECTIONS_WIDTH 0
441#endif
442
443#define ZONES_WIDTH ZONES_SHIFT
444
Christoph Lameter9223b4192008-04-28 02:12:48 -0700445#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700446#define NODES_WIDTH NODES_SHIFT
447#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700448#ifdef CONFIG_SPARSEMEM_VMEMMAP
449#error "Vmemmap: No space for nodes field in page flags"
450#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700451#define NODES_WIDTH 0
452#endif
453
454/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100455#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700456#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
457#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
458
459/*
460 * We are going to use the flags for the page to node mapping if its in
461 * there. This includes the case where there is no node, so it is implicit.
462 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800463#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
464#define NODE_NOT_IN_PAGE_FLAGS
465#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700466
467#ifndef PFN_SECTION_SHIFT
468#define PFN_SECTION_SHIFT 0
469#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700470
471/*
472 * Define the bit shifts to access each section. For non-existant
473 * sections we define the shift as 0; that plus a 0 mask ensures
474 * the compiler will optimise away reference to them.
475 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700476#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
477#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
478#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700479
Christoph Lameter89689ae2006-12-06 20:31:45 -0800480/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
481#ifdef NODE_NOT_IN_PAGEFLAGS
482#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800483#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
484 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700485#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800486#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800487#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
488 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700489#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800490
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800491#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700492
Christoph Lameter9223b4192008-04-28 02:12:48 -0700493#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
494#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700495#endif
496
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700497#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
498#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
499#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800500#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700501
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700502static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503{
Dave Hansen348f8b62005-06-23 00:07:40 -0700504 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
Christoph Lameter89689ae2006-12-06 20:31:45 -0800507/*
508 * The identification function is only used by the buddy allocator for
509 * determining if two pages could be buddies. We are not really
510 * identifying a zone since we could be using a the section number
511 * id if we have not node id available in page flags.
512 * We guarantee only that it will return the same value for two
513 * combinable pages in a zone.
514 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700515static inline int page_zone_id(struct page *page)
516{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800517 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518}
519
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800520static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700521{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700522#ifdef CONFIG_NUMA
523 return zone->node;
524#else
525 return 0;
526#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700527}
528
Christoph Lameter89689ae2006-12-06 20:31:45 -0800529#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800530extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800531#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800532static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700533{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800534 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700535}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800536#endif
537
538static inline struct zone *page_zone(struct page *page)
539{
540 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
541}
542
Christoph Lameter308c05e2008-04-28 02:12:43 -0700543#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700544static inline unsigned long page_to_section(struct page *page)
545{
546 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
547}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700548#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700549
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700550static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551{
Dave Hansen348f8b62005-06-23 00:07:40 -0700552 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
553 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
554}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700555
Dave Hansen348f8b62005-06-23 00:07:40 -0700556static inline void set_page_node(struct page *page, unsigned long node)
557{
558 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
559 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
560}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800561
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700562static inline void set_page_section(struct page *page, unsigned long section)
563{
564 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
565 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
566}
Dave Hansen348f8b62005-06-23 00:07:40 -0700567
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700568static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700569 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700570{
571 set_page_zone(page, zone);
572 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700573 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574}
575
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700576/*
577 * Some inline functions in vmstat.h depend on page_zone()
578 */
579#include <linux/vmstat.h>
580
Ingo Molnar652050a2006-01-14 13:21:30 -0800581static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582{
583 return __va(page_to_pfn(page) << PAGE_SHIFT);
584}
585
586#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
587#define HASHED_PAGE_VIRTUAL
588#endif
589
590#if defined(WANT_PAGE_VIRTUAL)
591#define page_address(page) ((page)->virtual)
592#define set_page_address(page, address) \
593 do { \
594 (page)->virtual = (address); \
595 } while(0)
596#define page_address_init() do { } while(0)
597#endif
598
599#if defined(HASHED_PAGE_VIRTUAL)
600void *page_address(struct page *page);
601void set_page_address(struct page *page, void *virtual);
602void page_address_init(void);
603#endif
604
605#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
606#define page_address(page) lowmem_page_address(page)
607#define set_page_address(page, address) do { } while(0)
608#define page_address_init() do { } while(0)
609#endif
610
611/*
612 * On an anonymous page mapped into a user virtual memory area,
613 * page->mapping points to its anon_vma, not to a struct address_space;
614 * with the PAGE_MAPPING_ANON bit set to distinguish it.
615 *
616 * Please note that, confusingly, "page_mapping" refers to the inode
617 * address_space which maps the page from disk; whereas "page_mapped"
618 * refers to user virtual address space into which the page is mapped.
619 */
620#define PAGE_MAPPING_ANON 1
621
622extern struct address_space swapper_space;
623static inline struct address_space *page_mapping(struct page *page)
624{
625 struct address_space *mapping = page->mapping;
626
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700627 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700628#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 if (unlikely(PageSwapCache(page)))
630 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700631 else
632#endif
633 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 mapping = NULL;
635 return mapping;
636}
637
638static inline int PageAnon(struct page *page)
639{
640 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
641}
642
643/*
644 * Return the pagecache index of the passed page. Regular pagecache pages
645 * use ->index whereas swapcache pages use ->private
646 */
647static inline pgoff_t page_index(struct page *page)
648{
649 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700650 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 return page->index;
652}
653
654/*
655 * The atomic page->_mapcount, like _count, starts from -1:
656 * so that transitions both from it and to it can be tracked,
657 * using atomic_inc_and_test and atomic_add_negative(-1).
658 */
659static inline void reset_page_mapcount(struct page *page)
660{
661 atomic_set(&(page)->_mapcount, -1);
662}
663
664static inline int page_mapcount(struct page *page)
665{
666 return atomic_read(&(page)->_mapcount) + 1;
667}
668
669/*
670 * Return true if this page is mapped into pagetables.
671 */
672static inline int page_mapped(struct page *page)
673{
674 return atomic_read(&(page)->_mapcount) >= 0;
675}
676
677/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 * Different kinds of faults, as returned by handle_mm_fault().
679 * Used to decide whether a process gets delivered SIGBUS or
680 * just gets major/minor fault counters bumped up.
681 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700682
Nick Piggin83c54072007-07-19 01:47:05 -0700683#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700684
Nick Piggin83c54072007-07-19 01:47:05 -0700685#define VM_FAULT_OOM 0x0001
686#define VM_FAULT_SIGBUS 0x0002
687#define VM_FAULT_MAJOR 0x0004
688#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000689
Nick Piggin83c54072007-07-19 01:47:05 -0700690#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
691#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
Nick Piggin83c54072007-07-19 01:47:05 -0700693#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700694
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800695/*
696 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
697 */
698extern void pagefault_out_of_memory(void);
699
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
701
702extern void show_free_areas(void);
703
704#ifdef CONFIG_SHMEM
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700705extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706#else
Russell King03b00eb2006-01-06 00:10:52 -0800707static inline int shmem_lock(struct file *file, int lock,
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700708 struct user_struct *user)
Russell King03b00eb2006-01-06 00:10:52 -0800709{
710 return 0;
711}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712#endif
Sergei Trofimovich168f5ac2009-06-16 15:33:02 -0700713struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714
715int shmem_zero_setup(struct vm_area_struct *);
716
David Howellsb0e15192006-01-06 00:11:42 -0800717#ifndef CONFIG_MMU
718extern unsigned long shmem_get_unmapped_area(struct file *file,
719 unsigned long addr,
720 unsigned long len,
721 unsigned long pgoff,
722 unsigned long flags);
723#endif
724
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400725extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726extern int user_shm_lock(size_t, struct user_struct *);
727extern void user_shm_unlock(size_t, struct user_struct *);
728
729/*
730 * Parameter block passed down to zap_pte_range in exceptional cases.
731 */
732struct zap_details {
733 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
734 struct address_space *check_mapping; /* Check page->mapping if set */
735 pgoff_t first_index; /* Lowest page->index to unmap */
736 pgoff_t last_index; /* Highest page->index to unmap */
737 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 unsigned long truncate_count; /* Compare vm_truncate_count */
739};
740
Nick Piggin7e675132008-04-28 02:13:00 -0700741struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
742 pte_t pte);
743
Jack Steinerc627f9c2008-07-29 22:33:53 -0700744int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
745 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700746unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700748unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 struct vm_area_struct *start_vma, unsigned long start_addr,
750 unsigned long end_addr, unsigned long *nr_accounted,
751 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800752
753/**
754 * mm_walk - callbacks for walk_page_range
755 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
756 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
757 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
758 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
759 * @pte_hole: if set, called for each hole at all levels
760 *
761 * (see walk_page_range for more details)
762 */
763struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700764 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
765 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
766 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
767 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
768 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
769 struct mm_struct *mm;
770 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800771};
772
Dave Hansen21650092008-06-12 15:21:47 -0700773int walk_page_range(unsigned long addr, unsigned long end,
774 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700775void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700776 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
778 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779void unmap_mapping_range(struct address_space *mapping,
780 loff_t const holebegin, loff_t const holelen, int even_cows);
Johannes Weiner3b6748e2009-06-16 15:32:35 -0700781int follow_pfn(struct vm_area_struct *vma, unsigned long address,
782 unsigned long *pfn);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800783int follow_phys(struct vm_area_struct *vma, unsigned long address,
784 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700785int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
786 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788static inline void unmap_shared_mapping_range(struct address_space *mapping,
789 loff_t const holebegin, loff_t const holelen)
790{
791 unmap_mapping_range(mapping, holebegin, holelen, 0);
792}
793
794extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800795extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000796
David Howells7ee1dd32006-01-06 00:11:44 -0800797#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700798extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700799 unsigned long address, unsigned int flags);
David Howells7ee1dd32006-01-06 00:11:44 -0800800#else
801static inline int handle_mm_fault(struct mm_struct *mm,
802 struct vm_area_struct *vma, unsigned long address,
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700803 unsigned int flags)
David Howells7ee1dd32006-01-06 00:11:44 -0800804{
805 /* should never happen if there's no MMU */
806 BUG();
807 return VM_FAULT_SIGBUS;
808}
809#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000810
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811extern int make_pages_present(unsigned long addr, unsigned long end);
812extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Nick Piggind2bf6be2009-06-16 15:31:39 -0700814int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
Peter Zijlstra9d737772009-06-25 11:58:55 +0200815 unsigned long start, int nr_pages, int write, int force,
Nick Piggind2bf6be2009-06-16 15:31:39 -0700816 struct page **pages, struct vm_area_struct **vmas);
817int get_user_pages_fast(unsigned long start, int nr_pages, int write,
818 struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
David Howellscf9a2ae2006-08-29 19:05:54 +0100820extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
821extern void do_invalidatepage(struct page *page, unsigned long offset);
822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800824int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825int redirty_page_for_writepage(struct writeback_control *wbc,
826 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700827void account_page_dirtied(struct page *page, struct address_space *mapping);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800828int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829int set_page_dirty_lock(struct page *page);
830int clear_page_dirty_for_io(struct page *page);
831
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700832extern unsigned long move_page_tables(struct vm_area_struct *vma,
833 unsigned long old_addr, struct vm_area_struct *new_vma,
834 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835extern unsigned long do_mremap(unsigned long addr,
836 unsigned long old_len, unsigned long new_len,
837 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700838extern int mprotect_fixup(struct vm_area_struct *vma,
839 struct vm_area_struct **pprev, unsigned long start,
840 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
Nick Piggin21cc1992008-07-25 19:45:22 -0700842/*
Peter Zijlstra465a4542009-06-15 12:31:37 +0200843 * doesn't attempt to fault and will return short.
844 */
845int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
846 struct page **pages);
847
848/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700849 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700851 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
852 * look through the least-recently-used 'nr_to_scan' entries and
853 * attempt to free them up. It should return the number of objects
854 * which remain in the cache. If it returns -1, it means it cannot do
855 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700857 * The 'gfpmask' refers to the allocation we are currently trying to
858 * fulfil.
859 *
860 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
861 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700863struct shrinker {
864 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
865 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
Rusty Russell8e1f9362007-07-17 04:03:17 -0700867 /* These are for internal use */
868 struct list_head list;
869 long nr; /* objs pending delete */
870};
871#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
872extern void register_shrinker(struct shrinker *);
873extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400875int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700876
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800877extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800878
Nick Piggin5f22df02007-05-06 14:49:02 -0700879#ifdef __PAGETABLE_PUD_FOLDED
880static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
881 unsigned long address)
882{
883 return 0;
884}
885#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700886int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700887#endif
888
889#ifdef __PAGETABLE_PMD_FOLDED
890static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
891 unsigned long address)
892{
893 return 0;
894}
895#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700896int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700897#endif
898
Hugh Dickins1bb36302005-10-29 18:16:22 -0700899int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
900int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902/*
903 * The following ifdef needed to get the 4level-fixup.h header to work.
904 * Remove it when 4level-fixup.h has been removed.
905 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700906#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
908{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700909 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
910 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911}
912
913static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
914{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700915 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
916 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700918#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
919
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700920#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700921/*
922 * We tuck a spinlock to guard each pagetable page into its struct page,
923 * at page->private, with BUILD_BUG_ON to make sure that this will not
924 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
925 * When freeing, reset page->mapping so free_pages_check won't complain.
926 */
Andrew Morton349aef02006-01-08 01:04:36 -0800927#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700928#define pte_lock_init(_page) do { \
929 spin_lock_init(__pte_lockptr(_page)); \
930} while (0)
931#define pte_lock_deinit(page) ((page)->mapping = NULL)
932#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700933#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700934/*
935 * We use mm->page_table_lock to guard all pagetable pages of the mm.
936 */
937#define pte_lock_init(page) do {} while (0)
938#define pte_lock_deinit(page) do {} while (0)
939#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700940#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700941
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800942static inline void pgtable_page_ctor(struct page *page)
943{
944 pte_lock_init(page);
945 inc_zone_page_state(page, NR_PAGETABLE);
946}
947
948static inline void pgtable_page_dtor(struct page *page)
949{
950 pte_lock_deinit(page);
951 dec_zone_page_state(page, NR_PAGETABLE);
952}
953
Hugh Dickinsc74df322005-10-29 18:16:23 -0700954#define pte_offset_map_lock(mm, pmd, address, ptlp) \
955({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700956 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700957 pte_t *__pte = pte_offset_map(pmd, address); \
958 *(ptlp) = __ptl; \
959 spin_lock(__ptl); \
960 __pte; \
961})
962
963#define pte_unmap_unlock(pte, ptl) do { \
964 spin_unlock(ptl); \
965 pte_unmap(pte); \
966} while (0)
967
Hugh Dickins1bb36302005-10-29 18:16:22 -0700968#define pte_alloc_map(mm, pmd, address) \
969 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
970 NULL: pte_offset_map(pmd, address))
971
Hugh Dickinsc74df322005-10-29 18:16:23 -0700972#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
973 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
974 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
975
Hugh Dickins1bb36302005-10-29 18:16:22 -0700976#define pte_alloc_kernel(pmd, address) \
977 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
978 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
980extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -0700981extern void free_area_init_node(int nid, unsigned long * zones_size,
982 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700983#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
984/*
985 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
986 * zones, allocate the backing mem_map and account for memory holes in a more
987 * architecture independent manner. This is a substitute for creating the
988 * zone_sizes[] and zholes_size[] arrays and passing them to
989 * free_area_init_node()
990 *
991 * An architecture is expected to register range of page frames backed by
992 * physical memory with add_active_range() before calling
993 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
994 * usage, an architecture is expected to do something like
995 *
996 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
997 * max_highmem_pfn};
998 * for_each_valid_physical_page_range()
999 * add_active_range(node_id, start_pfn, end_pfn)
1000 * free_area_init_nodes(max_zone_pfns);
1001 *
1002 * If the architecture guarantees that there are no holes in the ranges
1003 * registered with add_active_range(), free_bootmem_active_regions()
1004 * will call free_bootmem_node() for each registered physical page range.
1005 * Similarly sparse_memory_present_with_active_regions() calls
1006 * memory_present() for each range when SPARSEMEM is enabled.
1007 *
1008 * See mm/page_alloc.c for more information on each function exposed by
1009 * CONFIG_ARCH_POPULATES_NODE_MAP
1010 */
1011extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1012extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1013 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001014extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1015 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001016extern void remove_all_active_ranges(void);
1017extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1018 unsigned long end_pfn);
1019extern void get_pfn_range_for_nid(unsigned int nid,
1020 unsigned long *start_pfn, unsigned long *end_pfn);
1021extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001022extern void free_bootmem_with_active_regions(int nid,
1023 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001024typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001025extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001026extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001027#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001028
1029#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1030 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1031static inline int __early_pfn_to_nid(unsigned long pfn)
1032{
1033 return 0;
1034}
1035#else
1036/* please see mm/page_alloc.c */
1037extern int __meminit early_pfn_to_nid(unsigned long pfn);
1038#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1039/* there is a per-arch backend function. */
1040extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1041#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1042#endif
1043
Mel Gorman0e0b8642006-09-27 01:49:56 -07001044extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001045extern void memmap_init_zone(unsigned long, int, unsigned long,
1046 unsigned long, enum memmap_context);
Minchan Kimbc75d332009-06-16 15:32:48 -07001047extern void setup_per_zone_wmarks(void);
Minchan Kim96cb4df2009-06-16 15:32:49 -07001048extern void calculate_zone_inactive_ratio(struct zone *zone);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001050extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051extern void show_mem(void);
1052extern void si_meminfo(struct sysinfo * val);
1053extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001054extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001056#ifdef CONFIG_NUMA
1057extern void setup_per_cpu_pageset(void);
1058#else
1059static inline void setup_per_cpu_pageset(void) {}
1060#endif
1061
Shaohua Li112067f2009-09-21 17:01:16 -07001062extern void zone_pcp_update(struct zone *zone);
1063
David Howells8feae132009-01-08 12:04:47 +00001064/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001065extern atomic_long_t mmap_pages_allocated;
David Howells8feae132009-01-08 12:04:47 +00001066
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067/* prio_tree.c */
1068void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1069void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1070void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1071struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1072 struct prio_tree_iter *iter);
1073
1074#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1075 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1076 (vma = vma_prio_tree_next(vma, iter)); )
1077
1078static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1079 struct list_head *list)
1080{
1081 vma->shared.vm_set.parent = NULL;
1082 list_add_tail(&vma->shared.vm_set.list, list);
1083}
1084
1085/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001086extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1088 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1089extern struct vm_area_struct *vma_merge(struct mm_struct *,
1090 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1091 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1092 struct mempolicy *);
1093extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1094extern int split_vma(struct mm_struct *,
1095 struct vm_area_struct *, unsigned long addr, int new_below);
1096extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1097extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1098 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001099extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1101 unsigned long addr, unsigned long len, pgoff_t pgoff);
1102extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001103
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001104extern int mm_take_all_locks(struct mm_struct *mm);
1105extern void mm_drop_all_locks(struct mm_struct *mm);
1106
Matt Helsley925d1c42008-04-29 01:01:36 -07001107#ifdef CONFIG_PROC_FS
1108/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1109extern void added_exe_file_vma(struct mm_struct *mm);
1110extern void removed_exe_file_vma(struct mm_struct *mm);
1111#else
1112static inline void added_exe_file_vma(struct mm_struct *mm)
1113{}
1114
1115static inline void removed_exe_file_vma(struct mm_struct *mm)
1116{}
1117#endif /* CONFIG_PROC_FS */
1118
akpm@osdl.org119f6572005-05-01 08:58:35 -07001119extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001120extern int install_special_mapping(struct mm_struct *mm,
1121 unsigned long addr, unsigned long len,
1122 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
1124extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1125
1126extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1127 unsigned long len, unsigned long prot,
1128 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001129extern unsigned long mmap_region(struct file *file, unsigned long addr,
1130 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001131 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
1133static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1134 unsigned long len, unsigned long prot,
1135 unsigned long flag, unsigned long offset)
1136{
1137 unsigned long ret = -EINVAL;
1138 if ((offset + PAGE_ALIGN(len)) < offset)
1139 goto out;
1140 if (!(offset & ~PAGE_MASK))
1141 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1142out:
1143 return ret;
1144}
1145
1146extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1147
1148extern unsigned long do_brk(unsigned long, unsigned long);
1149
1150/* filemap.c */
1151extern unsigned long page_unuse(struct page *);
1152extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001153extern void truncate_inode_pages_range(struct address_space *,
1154 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
1156/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001157extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
1159/* mm/page-writeback.c */
1160int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001161void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163/* readahead.c */
1164#define VM_MAX_READAHEAD 128 /* kbytes */
1165#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001168 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001169
1170void page_cache_sync_readahead(struct address_space *mapping,
1171 struct file_ra_state *ra,
1172 struct file *filp,
1173 pgoff_t offset,
1174 unsigned long size);
1175
1176void page_cache_async_readahead(struct address_space *mapping,
1177 struct file_ra_state *ra,
1178 struct file *filp,
1179 struct page *pg,
1180 pgoff_t offset,
1181 unsigned long size);
1182
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183unsigned long max_sane_readahead(unsigned long nr);
Wu Fengguangd30a1102009-06-16 15:31:30 -07001184unsigned long ra_submit(struct file_ra_state *ra,
1185 struct address_space *mapping,
1186 struct file *filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
1188/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001189extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001190#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001191extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001192#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001193extern int expand_stack_downwards(struct vm_area_struct *vma,
1194 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
1196/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1197extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1198extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1199 struct vm_area_struct **pprev);
1200
1201/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1202 NULL if none. Assume start_addr < end_addr. */
1203static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1204{
1205 struct vm_area_struct * vma = find_vma(mm,start_addr);
1206
1207 if (vma && end_addr <= vma->vm_start)
1208 vma = NULL;
1209 return vma;
1210}
1211
1212static inline unsigned long vma_pages(struct vm_area_struct *vma)
1213{
1214 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1215}
1216
Hugh Dickins804af2c2006-07-26 21:39:49 +01001217pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001218struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001219int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1220 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001221int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001222int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1223 unsigned long pfn);
Nick Piggin423bad602008-04-28 02:13:01 -07001224int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1225 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Linus Torvalds6aab3412005-11-28 14:34:23 -08001227struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001228 unsigned int foll_flags);
1229#define FOLL_WRITE 0x01 /* check pte is writable */
1230#define FOLL_TOUCH 0x02 /* mark page accessed */
1231#define FOLL_GET 0x04 /* do get_page on page */
1232#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001234typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001235 void *data);
1236extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1237 unsigned long size, pte_fn_t fn, void *data);
1238
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001240void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001242static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 unsigned long flags, struct file *file, long pages)
1244{
1245}
1246#endif /* CONFIG_PROC_FS */
1247
Ingo Molnar12d6f212008-01-30 13:33:58 +01001248#ifdef CONFIG_DEBUG_PAGEALLOC
1249extern int debug_pagealloc_enabled;
1250
1251extern void kernel_map_pages(struct page *page, int numpages, int enable);
1252
1253static inline void enable_debug_pagealloc(void)
1254{
1255 debug_pagealloc_enabled = 1;
1256}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001257#ifdef CONFIG_HIBERNATION
1258extern bool kernel_page_present(struct page *page);
1259#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001260#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001262kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001263static inline void enable_debug_pagealloc(void)
1264{
1265}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001266#ifdef CONFIG_HIBERNATION
1267static inline bool kernel_page_present(struct page *page) { return true; }
1268#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269#endif
1270
1271extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1272#ifdef __HAVE_ARCH_GATE_AREA
1273int in_gate_area_no_task(unsigned long addr);
1274int in_gate_area(struct task_struct *task, unsigned long addr);
1275#else
1276int in_gate_area_no_task(unsigned long addr);
1277#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1278#endif /* __HAVE_ARCH_GATE_AREA */
1279
Andrew Morton9d0243b2006-01-08 01:00:39 -08001280int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1281 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001282unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001283 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001284
Luke Yang7a9166e2006-02-20 18:28:07 -08001285#ifndef CONFIG_MMU
1286#define randomize_va_space 0
1287#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001288extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001289#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001290
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001291const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001292void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001293
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001294struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001295pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1296pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1297pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1298pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001299void *vmemmap_alloc_block(unsigned long size, int node);
1300void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001301int vmemmap_populate_basepages(struct page *start_page,
1302 unsigned long pages, int node);
1303int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001304void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001305
Markus Metzger1cb81b12009-04-24 09:51:43 +02001306extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1307 size_t size);
1308extern void refund_locked_memory(struct mm_struct *mm, size_t size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309#endif /* __KERNEL__ */
1310#endif /* _LINUX_MM_H */