blob: bff1f0d475c7593240d5a464ae8ecc7c03e07299 [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) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700101#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) */
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700107#define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
Nick Piggind00806b2007-07-19 01:46:57 -0700108
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
110#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
111#endif
112
113#ifdef CONFIG_STACK_GROWSUP
114#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
115#else
116#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
117#endif
118
119#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
120#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
121#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
122#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
123#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
124
125/*
Nick Pigginb291f002008-10-18 20:26:44 -0700126 * special vmas that are non-mergable, non-mlock()able
127 */
128#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
129
130/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 * mapping from the currently active vm_flags protection bits (the
132 * low four bits) to a page protection mask..
133 */
134extern pgprot_t protection_map[16];
135
Nick Piggind0217ac2007-07-19 01:47:03 -0700136#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
137#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700138#define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
Nick Piggind0217ac2007-07-19 01:47:03 -0700139
venkatesh.pallipadi@intel.com6bd9cd52008-12-19 13:47:26 -0800140/*
141 * This interface is used by x86 PAT code to identify a pfn mapping that is
142 * linear over entire vma. This is to optimize PAT code that deals with
143 * marking the physical region with a particular prot. This is not for generic
144 * mm use. Note also that this check will not work if the pfn mapping is
145 * linear for a vma starting at physical address 0. In which case PAT code
146 * falls back to slow path of reserving physical range page by page.
147 */
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800148static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
149{
Pallipadi, Venkatesh895791d2009-03-13 16:35:44 -0700150 return (vma->vm_flags & VM_PFN_AT_MMAP);
venkatesh.pallipadi@intel.com3c8bb732008-12-18 11:41:27 -0800151}
152
153static inline int is_pfn_mapping(struct vm_area_struct *vma)
154{
155 return (vma->vm_flags & VM_PFNMAP);
156}
Nick Piggind0217ac2007-07-19 01:47:03 -0700157
Nick Piggin54cb8822007-07-19 01:46:59 -0700158/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700159 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700160 * ->fault function. The vma's ->fault is responsible for returning a bitmask
161 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700162 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700163 * pgoff should be used in favour of virtual_address, if possible. If pgoff
164 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
165 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700166 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700167struct vm_fault {
168 unsigned int flags; /* FAULT_FLAG_xxx flags */
169 pgoff_t pgoff; /* Logical page offset based on vma */
170 void __user *virtual_address; /* Faulting virtual address */
171
172 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700173 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700174 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700175 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700176 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700177};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178
179/*
180 * These are the virtual MM functions - opening of an area, closing and
181 * unmapping it (needed to keep files on disk up-to-date etc), pointer
182 * to the functions called when a no-page or a wp-page exception occurs.
183 */
184struct vm_operations_struct {
185 void (*open)(struct vm_area_struct * area);
186 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700187 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700188
189 /* notification that a previously read-only page is about to become
190 * writable, if an error is returned it will cause a SIGBUS */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700191 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700192
193 /* called by access_process_vm when get_user_pages() fails, typically
194 * for use by special VMAs that can switch between memory and hardware
195 */
196 int (*access)(struct vm_area_struct *vma, unsigned long addr,
197 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700199 /*
200 * set_policy() op must add a reference to any non-NULL @new mempolicy
201 * to hold the policy upon return. Caller should pass NULL @new to
202 * remove a policy and fall back to surrounding context--i.e. do not
203 * install a MPOL_DEFAULT policy, nor the task or system default
204 * mempolicy.
205 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700207
208 /*
209 * get_policy() op must add reference [mpol_get()] to any policy at
210 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
211 * in mm/mempolicy.c will do this automatically.
212 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
213 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
214 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
215 * must return NULL--i.e., do not "fallback" to task or system default
216 * policy.
217 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
219 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700220 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
221 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222#endif
223};
224
225struct mmu_gather;
226struct inode;
227
Andrew Morton349aef02006-01-08 01:04:36 -0800228#define page_private(page) ((page)->private)
229#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700230
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231/*
232 * FIXME: take this include out, include page-flags.h in
233 * files which need it (119 of them)
234 */
235#include <linux/page-flags.h>
236
237/*
238 * Methods to modify the page usage count.
239 *
240 * What counts for a page usage:
241 * - cache mapping (page->mapping)
242 * - private data (page->private)
243 * - page mapped in a task's page tables, each mapping
244 * is counted separately
245 *
246 * Also, many kernel routines increase the page count before a critical
247 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 */
249
250/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700251 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800253static inline int put_page_testzero(struct page *page)
254{
Nick Piggin725d7042006-09-25 23:30:55 -0700255 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800256 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800257}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800260 * Try to grab a ref unless the page has a refcount of zero, return false if
261 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800263static inline int get_page_unless_zero(struct page *page)
264{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800265 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800266}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Christoph Lameter48667e72008-02-04 22:28:31 -0800268/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800269struct page *vmalloc_to_page(const void *addr);
270unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800271
Paul Mundt0738c4b2008-03-12 16:51:31 +0900272/*
273 * Determine if an address is within the vmalloc range
274 *
275 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
276 * is no special casing required.
277 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800278static inline int is_vmalloc_addr(const void *x)
279{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900280#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800281 unsigned long addr = (unsigned long)x;
282
283 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900284#else
285 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800286#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900287}
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800288
Christoph Lameterd85f3382007-05-06 14:49:39 -0700289static inline struct page *compound_head(struct page *page)
290{
Christoph Lameter6d777952007-05-06 14:49:40 -0700291 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700292 return page->first_page;
293 return page;
294}
295
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700296static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700298 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299}
300
301static inline void get_page(struct page *page)
302{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700303 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700304 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 atomic_inc(&page->_count);
306}
307
Christoph Lameterb49af682007-05-06 14:49:41 -0700308static inline struct page *virt_to_head_page(const void *x)
309{
310 struct page *page = virt_to_page(x);
311 return compound_head(page);
312}
313
Nick Piggin7835e982006-03-22 00:08:40 -0800314/*
315 * Setup the page count before being freed into the page allocator for
316 * the first time (boot or memory hotplug)
317 */
318static inline void init_page_count(struct page *page)
319{
320 atomic_set(&page->_count, 1);
321}
322
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700324void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800326void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800327
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800329 * Compound pages have a destructor function. Provide a
330 * prototype for that function and accessor functions.
331 * These are _only_ valid on the head of a PG_compound page.
332 */
333typedef void compound_page_dtor(struct page *);
334
335static inline void set_compound_page_dtor(struct page *page,
336 compound_page_dtor *dtor)
337{
338 page[1].lru.next = (void *)dtor;
339}
340
341static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
342{
343 return (compound_page_dtor *)page[1].lru.next;
344}
345
Christoph Lameterd85f3382007-05-06 14:49:39 -0700346static inline int compound_order(struct page *page)
347{
Christoph Lameter6d777952007-05-06 14:49:40 -0700348 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700349 return 0;
350 return (unsigned long)page[1].lru.prev;
351}
352
353static inline void set_compound_order(struct page *page, unsigned long order)
354{
355 page[1].lru.prev = (void *)order;
356}
357
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800358/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 * Multiple processes may "see" the same page. E.g. for untouched
360 * mappings of /dev/null, all processes see the same page full of
361 * zeroes, and text pages of executables and shared libraries have
362 * only one copy in memory, at most, normally.
363 *
364 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700365 * page_count() == 0 means the page is free. page->lru is then used for
366 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700367 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700369 * Pages are allocated by the slab allocator in order to provide memory
370 * to kmalloc and kmem_cache_alloc. In this case, the management of the
371 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
372 * unless a particular usage is carefully commented. (the responsibility of
373 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700375 * A page may be used by anyone else who does a __get_free_page().
376 * In this case, page_count still tracks the references, and should only
377 * be used through the normal accessor functions. The top bits of page->flags
378 * and page->virtual store page management information, but all other fields
379 * are unused and could be used privately, carefully. The management of this
380 * page is the responsibility of the one who allocated it, and those who have
381 * subsequently been given references to it.
382 *
383 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 * managed by the Linux memory manager: I/O, buffers, swapping etc.
385 * The following discussion applies only to them.
386 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700387 * A pagecache page contains an opaque `private' member, which belongs to the
388 * page's address_space. Usually, this is the address of a circular list of
389 * the page's disk buffers. PG_private must be set to tell the VM to call
390 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700392 * A page may belong to an inode's memory mapping. In this case, page->mapping
393 * is the pointer to the inode, and page->index is the file offset of the page,
394 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700396 * If pagecache pages are not associated with an inode, they are said to be
397 * anonymous pages. These may become associated with the swapcache, and in that
398 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700400 * In either case (swapcache or inode backed), the pagecache itself holds one
401 * reference to the page. Setting PG_private should also increment the
402 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700404 * The pagecache pages are stored in a per-mapping radix tree, which is
405 * rooted at mapping->page_tree, and indexed by offset.
406 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
407 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700409 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 * - inode pages may need to be read from disk,
411 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700412 * to be written back to the inode on disk,
413 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
414 * modified may need to be swapped out to swap space and (later) to be read
415 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 */
417
418/*
419 * The zone field is never updated after free_area_init_core()
420 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700422
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700423
424/*
425 * page->flags layout:
426 *
427 * There are three possibilities for how page->flags get
428 * laid out. The first is for the normal case, without
429 * sparsemem. The second is for sparsemem when there is
430 * plenty of space for node and section. The last is when
431 * we have run out of space and have to fall back to an
432 * alternate (slower) way of determining the node.
433 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700434 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
435 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
436 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700437 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700438#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700439#define SECTIONS_WIDTH SECTIONS_SHIFT
440#else
441#define SECTIONS_WIDTH 0
442#endif
443
444#define ZONES_WIDTH ZONES_SHIFT
445
Christoph Lameter9223b412008-04-28 02:12:48 -0700446#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700447#define NODES_WIDTH NODES_SHIFT
448#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700449#ifdef CONFIG_SPARSEMEM_VMEMMAP
450#error "Vmemmap: No space for nodes field in page flags"
451#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700452#define NODES_WIDTH 0
453#endif
454
455/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100456#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700457#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
458#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
459
460/*
461 * We are going to use the flags for the page to node mapping if its in
462 * there. This includes the case where there is no node, so it is implicit.
463 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800464#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
465#define NODE_NOT_IN_PAGE_FLAGS
466#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700467
468#ifndef PFN_SECTION_SHIFT
469#define PFN_SECTION_SHIFT 0
470#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700471
472/*
473 * Define the bit shifts to access each section. For non-existant
474 * sections we define the shift as 0; that plus a 0 mask ensures
475 * the compiler will optimise away reference to them.
476 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700477#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
478#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
479#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700480
Christoph Lameter89689ae2006-12-06 20:31:45 -0800481/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
482#ifdef NODE_NOT_IN_PAGEFLAGS
483#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800484#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
485 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700486#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800487#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800488#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
489 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700490#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800491
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800492#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700493
Christoph Lameter9223b412008-04-28 02:12:48 -0700494#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
495#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700496#endif
497
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700498#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
499#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
500#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800501#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700502
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700503static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
Dave Hansen348f8b62005-06-23 00:07:40 -0700505 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Christoph Lameter89689ae2006-12-06 20:31:45 -0800508/*
509 * The identification function is only used by the buddy allocator for
510 * determining if two pages could be buddies. We are not really
511 * identifying a zone since we could be using a the section number
512 * id if we have not node id available in page flags.
513 * We guarantee only that it will return the same value for two
514 * combinable pages in a zone.
515 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700516static inline int page_zone_id(struct page *page)
517{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800518 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519}
520
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800521static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700522{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700523#ifdef CONFIG_NUMA
524 return zone->node;
525#else
526 return 0;
527#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700528}
529
Christoph Lameter89689ae2006-12-06 20:31:45 -0800530#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800531extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800532#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800533static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700534{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800535 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700536}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800537#endif
538
539static inline struct zone *page_zone(struct page *page)
540{
541 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
542}
543
Christoph Lameter308c05e2008-04-28 02:12:43 -0700544#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700545static inline unsigned long page_to_section(struct page *page)
546{
547 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
548}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700549#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700550
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700551static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
Dave Hansen348f8b62005-06-23 00:07:40 -0700553 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
554 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
555}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700556
Dave Hansen348f8b62005-06-23 00:07:40 -0700557static inline void set_page_node(struct page *page, unsigned long node)
558{
559 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
560 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
561}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800562
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700563static inline void set_page_section(struct page *page, unsigned long section)
564{
565 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
566 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
567}
Dave Hansen348f8b62005-06-23 00:07:40 -0700568
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700569static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700570 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700571{
572 set_page_zone(page, zone);
573 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700574 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575}
576
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700577/*
Eric Paris7cd94142007-11-26 18:47:40 -0500578 * If a hint addr is less than mmap_min_addr change hint to be as
579 * low as possible but still greater than mmap_min_addr
580 */
581static inline unsigned long round_hint_to_min(unsigned long hint)
582{
583#ifdef CONFIG_SECURITY
584 hint &= PAGE_MASK;
585 if (((void *)hint != NULL) &&
586 (hint < mmap_min_addr))
587 return PAGE_ALIGN(mmap_min_addr);
588#endif
589 return hint;
590}
591
592/*
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700593 * Some inline functions in vmstat.h depend on page_zone()
594 */
595#include <linux/vmstat.h>
596
Ingo Molnar652050a2006-01-14 13:21:30 -0800597static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598{
599 return __va(page_to_pfn(page) << PAGE_SHIFT);
600}
601
602#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
603#define HASHED_PAGE_VIRTUAL
604#endif
605
606#if defined(WANT_PAGE_VIRTUAL)
607#define page_address(page) ((page)->virtual)
608#define set_page_address(page, address) \
609 do { \
610 (page)->virtual = (address); \
611 } while(0)
612#define page_address_init() do { } while(0)
613#endif
614
615#if defined(HASHED_PAGE_VIRTUAL)
616void *page_address(struct page *page);
617void set_page_address(struct page *page, void *virtual);
618void page_address_init(void);
619#endif
620
621#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
622#define page_address(page) lowmem_page_address(page)
623#define set_page_address(page, address) do { } while(0)
624#define page_address_init() do { } while(0)
625#endif
626
627/*
628 * On an anonymous page mapped into a user virtual memory area,
629 * page->mapping points to its anon_vma, not to a struct address_space;
630 * with the PAGE_MAPPING_ANON bit set to distinguish it.
631 *
632 * Please note that, confusingly, "page_mapping" refers to the inode
633 * address_space which maps the page from disk; whereas "page_mapped"
634 * refers to user virtual address space into which the page is mapped.
635 */
636#define PAGE_MAPPING_ANON 1
637
638extern struct address_space swapper_space;
639static inline struct address_space *page_mapping(struct page *page)
640{
641 struct address_space *mapping = page->mapping;
642
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700643 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700644#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 if (unlikely(PageSwapCache(page)))
646 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700647 else
648#endif
649 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 mapping = NULL;
651 return mapping;
652}
653
654static inline int PageAnon(struct page *page)
655{
656 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
657}
658
659/*
660 * Return the pagecache index of the passed page. Regular pagecache pages
661 * use ->index whereas swapcache pages use ->private
662 */
663static inline pgoff_t page_index(struct page *page)
664{
665 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700666 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 return page->index;
668}
669
670/*
671 * The atomic page->_mapcount, like _count, starts from -1:
672 * so that transitions both from it and to it can be tracked,
673 * using atomic_inc_and_test and atomic_add_negative(-1).
674 */
675static inline void reset_page_mapcount(struct page *page)
676{
677 atomic_set(&(page)->_mapcount, -1);
678}
679
680static inline int page_mapcount(struct page *page)
681{
682 return atomic_read(&(page)->_mapcount) + 1;
683}
684
685/*
686 * Return true if this page is mapped into pagetables.
687 */
688static inline int page_mapped(struct page *page)
689{
690 return atomic_read(&(page)->_mapcount) >= 0;
691}
692
693/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 * Different kinds of faults, as returned by handle_mm_fault().
695 * Used to decide whether a process gets delivered SIGBUS or
696 * just gets major/minor fault counters bumped up.
697 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700698
Nick Piggin83c54072007-07-19 01:47:05 -0700699#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700700
Nick Piggin83c54072007-07-19 01:47:05 -0700701#define VM_FAULT_OOM 0x0001
702#define VM_FAULT_SIGBUS 0x0002
703#define VM_FAULT_MAJOR 0x0004
704#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000705
Nick Piggin83c54072007-07-19 01:47:05 -0700706#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
707#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
Nick Piggin83c54072007-07-19 01:47:05 -0700709#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700710
Nick Piggin1c0fe6e2009-01-06 14:38:59 -0800711/*
712 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
713 */
714extern void pagefault_out_of_memory(void);
715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
717
718extern void show_free_areas(void);
719
720#ifdef CONFIG_SHMEM
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700721extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722#else
Russell King03b00eb2006-01-06 00:10:52 -0800723static inline int shmem_lock(struct file *file, int lock,
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700724 struct user_struct *user)
Russell King03b00eb2006-01-06 00:10:52 -0800725{
726 return 0;
727}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728#endif
729struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
730
731int shmem_zero_setup(struct vm_area_struct *);
732
David Howellsb0e15192006-01-06 00:11:42 -0800733#ifndef CONFIG_MMU
734extern unsigned long shmem_get_unmapped_area(struct file *file,
735 unsigned long addr,
736 unsigned long len,
737 unsigned long pgoff,
738 unsigned long flags);
739#endif
740
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400741extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742extern int user_shm_lock(size_t, struct user_struct *);
743extern void user_shm_unlock(size_t, struct user_struct *);
744
745/*
746 * Parameter block passed down to zap_pte_range in exceptional cases.
747 */
748struct zap_details {
749 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
750 struct address_space *check_mapping; /* Check page->mapping if set */
751 pgoff_t first_index; /* Lowest page->index to unmap */
752 pgoff_t last_index; /* Highest page->index to unmap */
753 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 unsigned long truncate_count; /* Compare vm_truncate_count */
755};
756
Nick Piggin7e675132008-04-28 02:13:00 -0700757struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
758 pte_t pte);
759
Jack Steinerc627f9c2008-07-29 22:33:53 -0700760int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
761 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700762unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700764unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 struct vm_area_struct *start_vma, unsigned long start_addr,
766 unsigned long end_addr, unsigned long *nr_accounted,
767 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800768
769/**
770 * mm_walk - callbacks for walk_page_range
771 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
772 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
773 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
774 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
775 * @pte_hole: if set, called for each hole at all levels
776 *
777 * (see walk_page_range for more details)
778 */
779struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700780 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
781 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
782 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
783 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
784 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
785 struct mm_struct *mm;
786 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800787};
788
Dave Hansen21650092008-06-12 15:21:47 -0700789int walk_page_range(unsigned long addr, unsigned long end,
790 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700791void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700792 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
794 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795void unmap_mapping_range(struct address_space *mapping,
796 loff_t const holebegin, loff_t const holelen, int even_cows);
venkatesh.pallipadi@intel.comd87fe662008-12-19 13:47:27 -0800797int follow_phys(struct vm_area_struct *vma, unsigned long address,
798 unsigned int flags, unsigned long *prot, resource_size_t *phys);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700799int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
800 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
802static inline void unmap_shared_mapping_range(struct address_space *mapping,
803 loff_t const holebegin, loff_t const holelen)
804{
805 unmap_mapping_range(mapping, holebegin, holelen, 0);
806}
807
808extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800809extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000810
David Howells7ee1dd32006-01-06 00:11:44 -0800811#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700812extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
David Howells7ee1dd32006-01-06 00:11:44 -0800813 unsigned long address, int write_access);
David Howells7ee1dd32006-01-06 00:11:44 -0800814#else
815static inline int handle_mm_fault(struct mm_struct *mm,
816 struct vm_area_struct *vma, unsigned long address,
817 int write_access)
818{
819 /* should never happen if there's no MMU */
820 BUG();
821 return VM_FAULT_SIGBUS;
822}
823#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825extern int make_pages_present(unsigned long addr, unsigned long end);
826extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
829 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
830
David Howellscf9a2ae2006-08-29 19:05:54 +0100831extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
832extern void do_invalidatepage(struct page *page, unsigned long offset);
833
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800835int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836int redirty_page_for_writepage(struct writeback_control *wbc,
837 struct page *page);
Edward Shishkine3a7cca2009-03-31 15:19:39 -0700838void account_page_dirtied(struct page *page, struct address_space *mapping);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800839int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840int set_page_dirty_lock(struct page *page);
841int clear_page_dirty_for_io(struct page *page);
842
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700843extern unsigned long move_page_tables(struct vm_area_struct *vma,
844 unsigned long old_addr, struct vm_area_struct *new_vma,
845 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846extern unsigned long do_mremap(unsigned long addr,
847 unsigned long old_len, unsigned long new_len,
848 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700849extern int mprotect_fixup(struct vm_area_struct *vma,
850 struct vm_area_struct **pprev, unsigned long start,
851 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
Nick Piggin21cc1992008-07-25 19:45:22 -0700853/*
854 * get_user_pages_fast provides equivalent functionality to get_user_pages,
855 * operating on current and current->mm (force=0 and doesn't return any vmas).
856 *
857 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
858 * can be made about locking. get_user_pages_fast is to be implemented in a
859 * way that is advantageous (vs get_user_pages()) when the user memory area is
860 * already faulted in and present in ptes. However if the pages have to be
861 * faulted in, it may turn out to be slightly slower).
862 */
863int get_user_pages_fast(unsigned long start, int nr_pages, int write,
864 struct page **pages);
865
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700867 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700869 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
870 * look through the least-recently-used 'nr_to_scan' entries and
871 * attempt to free them up. It should return the number of objects
872 * which remain in the cache. If it returns -1, it means it cannot do
873 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700875 * The 'gfpmask' refers to the allocation we are currently trying to
876 * fulfil.
877 *
878 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
879 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700881struct shrinker {
882 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
883 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
Rusty Russell8e1f9362007-07-17 04:03:17 -0700885 /* These are for internal use */
886 struct list_head list;
887 long nr; /* objs pending delete */
888};
889#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
890extern void register_shrinker(struct shrinker *);
891extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400893int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700894
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800895extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800896
Nick Piggin5f22df02007-05-06 14:49:02 -0700897#ifdef __PAGETABLE_PUD_FOLDED
898static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
899 unsigned long address)
900{
901 return 0;
902}
903#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700904int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700905#endif
906
907#ifdef __PAGETABLE_PMD_FOLDED
908static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
909 unsigned long address)
910{
911 return 0;
912}
913#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700914int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700915#endif
916
Hugh Dickins1bb36302005-10-29 18:16:22 -0700917int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
918int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
919
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920/*
921 * The following ifdef needed to get the 4level-fixup.h header to work.
922 * Remove it when 4level-fixup.h has been removed.
923 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700924#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
926{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700927 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
928 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929}
930
931static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
932{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700933 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
934 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700936#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
937
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700938#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700939/*
940 * We tuck a spinlock to guard each pagetable page into its struct page,
941 * at page->private, with BUILD_BUG_ON to make sure that this will not
942 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
943 * When freeing, reset page->mapping so free_pages_check won't complain.
944 */
Andrew Morton349aef02006-01-08 01:04:36 -0800945#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700946#define pte_lock_init(_page) do { \
947 spin_lock_init(__pte_lockptr(_page)); \
948} while (0)
949#define pte_lock_deinit(page) ((page)->mapping = NULL)
950#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700951#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700952/*
953 * We use mm->page_table_lock to guard all pagetable pages of the mm.
954 */
955#define pte_lock_init(page) do {} while (0)
956#define pte_lock_deinit(page) do {} while (0)
957#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700958#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700959
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800960static inline void pgtable_page_ctor(struct page *page)
961{
962 pte_lock_init(page);
963 inc_zone_page_state(page, NR_PAGETABLE);
964}
965
966static inline void pgtable_page_dtor(struct page *page)
967{
968 pte_lock_deinit(page);
969 dec_zone_page_state(page, NR_PAGETABLE);
970}
971
Hugh Dickinsc74df322005-10-29 18:16:23 -0700972#define pte_offset_map_lock(mm, pmd, address, ptlp) \
973({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700974 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700975 pte_t *__pte = pte_offset_map(pmd, address); \
976 *(ptlp) = __ptl; \
977 spin_lock(__ptl); \
978 __pte; \
979})
980
981#define pte_unmap_unlock(pte, ptl) do { \
982 spin_unlock(ptl); \
983 pte_unmap(pte); \
984} while (0)
985
Hugh Dickins1bb36302005-10-29 18:16:22 -0700986#define pte_alloc_map(mm, pmd, address) \
987 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
988 NULL: pte_offset_map(pmd, address))
989
Hugh Dickinsc74df322005-10-29 18:16:23 -0700990#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
991 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
992 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
993
Hugh Dickins1bb36302005-10-29 18:16:22 -0700994#define pte_alloc_kernel(pmd, address) \
995 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
996 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
998extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -0700999extern void free_area_init_node(int nid, unsigned long * zones_size,
1000 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -07001001#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1002/*
1003 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1004 * zones, allocate the backing mem_map and account for memory holes in a more
1005 * architecture independent manner. This is a substitute for creating the
1006 * zone_sizes[] and zholes_size[] arrays and passing them to
1007 * free_area_init_node()
1008 *
1009 * An architecture is expected to register range of page frames backed by
1010 * physical memory with add_active_range() before calling
1011 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1012 * usage, an architecture is expected to do something like
1013 *
1014 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1015 * max_highmem_pfn};
1016 * for_each_valid_physical_page_range()
1017 * add_active_range(node_id, start_pfn, end_pfn)
1018 * free_area_init_nodes(max_zone_pfns);
1019 *
1020 * If the architecture guarantees that there are no holes in the ranges
1021 * registered with add_active_range(), free_bootmem_active_regions()
1022 * will call free_bootmem_node() for each registered physical page range.
1023 * Similarly sparse_memory_present_with_active_regions() calls
1024 * memory_present() for each range when SPARSEMEM is enabled.
1025 *
1026 * See mm/page_alloc.c for more information on each function exposed by
1027 * CONFIG_ARCH_POPULATES_NODE_MAP
1028 */
1029extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1030extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1031 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001032extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1033 unsigned long end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -07001034extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1035 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001036extern void remove_all_active_ranges(void);
1037extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1038 unsigned long end_pfn);
1039extern void get_pfn_range_for_nid(unsigned int nid,
1040 unsigned long *start_pfn, unsigned long *end_pfn);
1041extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001042extern void free_bootmem_with_active_regions(int nid,
1043 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001044typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001045extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001046extern void sparse_memory_present_with_active_regions(int nid);
Mel Gormanc7132162006-09-27 01:49:43 -07001047#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
KAMEZAWA Hiroyukif2dbcfa2009-02-18 14:48:32 -08001048
1049#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1050 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1051static inline int __early_pfn_to_nid(unsigned long pfn)
1052{
1053 return 0;
1054}
1055#else
1056/* please see mm/page_alloc.c */
1057extern int __meminit early_pfn_to_nid(unsigned long pfn);
1058#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1059/* there is a per-arch backend function. */
1060extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1061#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1062#endif
1063
Mel Gorman0e0b8642006-09-27 01:49:56 -07001064extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001065extern void memmap_init_zone(unsigned long, int, unsigned long,
1066 unsigned long, enum memmap_context);
Dave Hansen3947be12005-10-29 18:16:54 -07001067extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068extern void mem_init(void);
David Howells8feae132009-01-08 12:04:47 +00001069extern void __init mmap_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070extern void show_mem(void);
1071extern void si_meminfo(struct sysinfo * val);
1072extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001073extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001075#ifdef CONFIG_NUMA
1076extern void setup_per_cpu_pageset(void);
1077#else
1078static inline void setup_per_cpu_pageset(void) {}
1079#endif
1080
David Howells8feae132009-01-08 12:04:47 +00001081/* nommu.c */
David Howells33e5d7692009-04-02 16:56:32 -07001082extern atomic_long_t mmap_pages_allocated;
David Howells8feae132009-01-08 12:04:47 +00001083
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084/* prio_tree.c */
1085void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1086void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1087void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1088struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1089 struct prio_tree_iter *iter);
1090
1091#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1092 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1093 (vma = vma_prio_tree_next(vma, iter)); )
1094
1095static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1096 struct list_head *list)
1097{
1098 vma->shared.vm_set.parent = NULL;
1099 list_add_tail(&vma->shared.vm_set.list, list);
1100}
1101
1102/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001103extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1105 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1106extern struct vm_area_struct *vma_merge(struct mm_struct *,
1107 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1108 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1109 struct mempolicy *);
1110extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1111extern int split_vma(struct mm_struct *,
1112 struct vm_area_struct *, unsigned long addr, int new_below);
1113extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1114extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1115 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001116extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1118 unsigned long addr, unsigned long len, pgoff_t pgoff);
1119extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001120
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001121extern int mm_take_all_locks(struct mm_struct *mm);
1122extern void mm_drop_all_locks(struct mm_struct *mm);
1123
Matt Helsley925d1c42008-04-29 01:01:36 -07001124#ifdef CONFIG_PROC_FS
1125/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1126extern void added_exe_file_vma(struct mm_struct *mm);
1127extern void removed_exe_file_vma(struct mm_struct *mm);
1128#else
1129static inline void added_exe_file_vma(struct mm_struct *mm)
1130{}
1131
1132static inline void removed_exe_file_vma(struct mm_struct *mm)
1133{}
1134#endif /* CONFIG_PROC_FS */
1135
akpm@osdl.org119f6572005-05-01 08:58:35 -07001136extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001137extern int install_special_mapping(struct mm_struct *mm,
1138 unsigned long addr, unsigned long len,
1139 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
1141extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1142
1143extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1144 unsigned long len, unsigned long prot,
1145 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001146extern unsigned long mmap_region(struct file *file, unsigned long addr,
1147 unsigned long len, unsigned long flags,
Mel Gorman5a6fe122009-02-10 14:02:27 +00001148 unsigned int vm_flags, unsigned long pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
1150static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1151 unsigned long len, unsigned long prot,
1152 unsigned long flag, unsigned long offset)
1153{
1154 unsigned long ret = -EINVAL;
1155 if ((offset + PAGE_ALIGN(len)) < offset)
1156 goto out;
1157 if (!(offset & ~PAGE_MASK))
1158 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1159out:
1160 return ret;
1161}
1162
1163extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1164
1165extern unsigned long do_brk(unsigned long, unsigned long);
1166
1167/* filemap.c */
1168extern unsigned long page_unuse(struct page *);
1169extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001170extern void truncate_inode_pages_range(struct address_space *,
1171 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
1173/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001174extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
1176/* mm/page-writeback.c */
1177int write_one_page(struct page *page, int wait);
Nick Piggin1cf6e7d2009-02-18 14:48:18 -08001178void task_dirty_inc(struct task_struct *tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
1180/* readahead.c */
1181#define VM_MAX_READAHEAD 128 /* kbytes */
1182#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
1184int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001185 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001187 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001188
1189void page_cache_sync_readahead(struct address_space *mapping,
1190 struct file_ra_state *ra,
1191 struct file *filp,
1192 pgoff_t offset,
1193 unsigned long size);
1194
1195void page_cache_async_readahead(struct address_space *mapping,
1196 struct file_ra_state *ra,
1197 struct file *filp,
1198 struct page *pg,
1199 pgoff_t offset,
1200 unsigned long size);
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202unsigned long max_sane_readahead(unsigned long nr);
1203
1204/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001205extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001206#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001207extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001208#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001209extern int expand_stack_downwards(struct vm_area_struct *vma,
1210 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
1212/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1213extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1214extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1215 struct vm_area_struct **pprev);
1216
1217/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1218 NULL if none. Assume start_addr < end_addr. */
1219static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1220{
1221 struct vm_area_struct * vma = find_vma(mm,start_addr);
1222
1223 if (vma && end_addr <= vma->vm_start)
1224 vma = NULL;
1225 return vma;
1226}
1227
1228static inline unsigned long vma_pages(struct vm_area_struct *vma)
1229{
1230 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1231}
1232
Hugh Dickins804af2c2006-07-26 21:39:49 +01001233pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001234struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001235int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1236 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001237int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001238int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1239 unsigned long pfn);
Nick Piggin423bad602008-04-28 02:13:01 -07001240int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1241 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
Linus Torvalds6aab3412005-11-28 14:34:23 -08001243struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001244 unsigned int foll_flags);
1245#define FOLL_WRITE 0x01 /* check pte is writable */
1246#define FOLL_TOUCH 0x02 /* mark page accessed */
1247#define FOLL_GET 0x04 /* do get_page on page */
1248#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001250typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001251 void *data);
1252extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1253 unsigned long size, pte_fn_t fn, void *data);
1254
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001256void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001258static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 unsigned long flags, struct file *file, long pages)
1260{
1261}
1262#endif /* CONFIG_PROC_FS */
1263
Ingo Molnar12d6f212008-01-30 13:33:58 +01001264#ifdef CONFIG_DEBUG_PAGEALLOC
1265extern int debug_pagealloc_enabled;
1266
1267extern void kernel_map_pages(struct page *page, int numpages, int enable);
1268
1269static inline void enable_debug_pagealloc(void)
1270{
1271 debug_pagealloc_enabled = 1;
1272}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001273#ifdef CONFIG_HIBERNATION
1274extern bool kernel_page_present(struct page *page);
1275#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001276#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001278kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001279static inline void enable_debug_pagealloc(void)
1280{
1281}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001282#ifdef CONFIG_HIBERNATION
1283static inline bool kernel_page_present(struct page *page) { return true; }
1284#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285#endif
1286
1287extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1288#ifdef __HAVE_ARCH_GATE_AREA
1289int in_gate_area_no_task(unsigned long addr);
1290int in_gate_area(struct task_struct *task, unsigned long addr);
1291#else
1292int in_gate_area_no_task(unsigned long addr);
1293#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1294#endif /* __HAVE_ARCH_GATE_AREA */
1295
Andrew Morton9d0243b2006-01-08 01:00:39 -08001296int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1297 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001298unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001299 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001300
Luke Yang7a9166e2006-02-20 18:28:07 -08001301#ifndef CONFIG_MMU
1302#define randomize_va_space 0
1303#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001304extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001305#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001306
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001307const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001308void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001309
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001310struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001311pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1312pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1313pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1314pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001315void *vmemmap_alloc_block(unsigned long size, int node);
1316void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001317int vmemmap_populate_basepages(struct page *start_page,
1318 unsigned long pages, int node);
1319int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001320void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001321
Markus Metzgerc5dee612008-12-19 15:17:02 +01001322extern void *alloc_locked_buffer(size_t size);
1323extern void free_locked_buffer(void *buffer, size_t size);
Markus Metzger9f339e72009-02-11 15:10:27 +01001324extern void release_locked_buffer(void *buffer, size_t size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325#endif /* __KERNEL__ */
1326#endif /* _LINUX_MM_H */