blob: 40236290e2ae5f9469445b4079e47524cd56b91c [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/*
60 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
61 * disabled, then there's a single shared list of VMAs maintained by the
62 * system, and mm's subscribe to these individually
63 */
64struct vm_list_struct {
65 struct vm_list_struct *next;
66 struct vm_area_struct *vma;
67};
68
69#ifndef CONFIG_MMU
70extern struct rb_root nommu_vma_tree;
71extern struct rw_semaphore nommu_vma_sem;
72
73extern unsigned int kobjsize(const void *objp);
74#endif
75
76/*
Hugh Dickins605d9282008-08-16 11:07:21 +010077 * vm_flags in vm_area_struct, see mm_types.h.
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 */
79#define VM_READ 0x00000001 /* currently active flags */
80#define VM_WRITE 0x00000002
81#define VM_EXEC 0x00000004
82#define VM_SHARED 0x00000008
83
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -070084/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
86#define VM_MAYWRITE 0x00000020
87#define VM_MAYEXEC 0x00000040
88#define VM_MAYSHARE 0x00000080
89
90#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
91#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -080092#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
94
95#define VM_EXECUTABLE 0x00001000
96#define VM_LOCKED 0x00002000
97#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
98
99 /* Used by sys_madvise() */
100#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
101#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
102
103#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
104#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800105#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
Andy Whitcroftcdfd4322008-07-23 21:27:28 -0700107#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
109#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
110#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800111#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Roland McGrathe5b97dd2007-01-26 00:56:48 -0800112#define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Nick Piggind0217ac2007-07-19 01:47:03 -0700114#define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
Jared Hulbertb379d792008-04-28 02:12:58 -0700115#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
Dave Kleikampaba46c52008-07-08 00:28:52 +1000116#define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
Nick Piggind00806b2007-07-19 01:46:57 -0700117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
119#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
120#endif
121
122#ifdef CONFIG_STACK_GROWSUP
123#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
124#else
125#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
126#endif
127
128#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
129#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
130#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
131#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
132#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
133
134/*
135 * mapping from the currently active vm_flags protection bits (the
136 * low four bits) to a page protection mask..
137 */
138extern pgprot_t protection_map[16];
139
Nick Piggind0217ac2007-07-19 01:47:03 -0700140#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
141#define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
142
143
Nick Piggin54cb8822007-07-19 01:46:59 -0700144/*
Nick Piggind0217ac2007-07-19 01:47:03 -0700145 * vm_fault is filled by the the pagefault handler and passed to the vma's
Nick Piggin83c54072007-07-19 01:47:05 -0700146 * ->fault function. The vma's ->fault is responsible for returning a bitmask
147 * of VM_FAULT_xxx flags that give details about how the fault was handled.
Nick Piggin54cb8822007-07-19 01:46:59 -0700148 *
Nick Piggind0217ac2007-07-19 01:47:03 -0700149 * pgoff should be used in favour of virtual_address, if possible. If pgoff
150 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
151 * mapping support.
Nick Piggin54cb8822007-07-19 01:46:59 -0700152 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700153struct vm_fault {
154 unsigned int flags; /* FAULT_FLAG_xxx flags */
155 pgoff_t pgoff; /* Logical page offset based on vma */
156 void __user *virtual_address; /* Faulting virtual address */
157
158 struct page *page; /* ->fault handlers should return a
Nick Piggin83c54072007-07-19 01:47:05 -0700159 * page here, unless VM_FAULT_NOPAGE
Nick Piggind0217ac2007-07-19 01:47:03 -0700160 * is set (which is also implied by
Nick Piggin83c54072007-07-19 01:47:05 -0700161 * VM_FAULT_ERROR).
Nick Piggind0217ac2007-07-19 01:47:03 -0700162 */
Nick Piggin54cb8822007-07-19 01:46:59 -0700163};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
165/*
166 * These are the virtual MM functions - opening of an area, closing and
167 * unmapping it (needed to keep files on disk up-to-date etc), pointer
168 * to the functions called when a no-page or a wp-page exception occurs.
169 */
170struct vm_operations_struct {
171 void (*open)(struct vm_area_struct * area);
172 void (*close)(struct vm_area_struct * area);
Nick Piggind0217ac2007-07-19 01:47:03 -0700173 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
David Howells9637a5e2006-06-23 02:03:43 -0700174
175 /* notification that a previously read-only page is about to become
176 * writable, if an error is returned it will cause a SIGBUS */
177 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700178
179 /* called by access_process_vm when get_user_pages() fails, typically
180 * for use by special VMAs that can switch between memory and hardware
181 */
182 int (*access)(struct vm_area_struct *vma, unsigned long addr,
183 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184#ifdef CONFIG_NUMA
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700185 /*
186 * set_policy() op must add a reference to any non-NULL @new mempolicy
187 * to hold the policy upon return. Caller should pass NULL @new to
188 * remove a policy and fall back to surrounding context--i.e. do not
189 * install a MPOL_DEFAULT policy, nor the task or system default
190 * mempolicy.
191 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
Lee Schermerhorna6020ed2008-04-28 02:13:14 -0700193
194 /*
195 * get_policy() op must add reference [mpol_get()] to any policy at
196 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
197 * in mm/mempolicy.c will do this automatically.
198 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
199 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
200 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
201 * must return NULL--i.e., do not "fallback" to task or system default
202 * policy.
203 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
205 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700206 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
207 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208#endif
209};
210
211struct mmu_gather;
212struct inode;
213
Andrew Morton349aef02006-01-08 01:04:36 -0800214#define page_private(page) ((page)->private)
215#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700216
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217/*
218 * FIXME: take this include out, include page-flags.h in
219 * files which need it (119 of them)
220 */
221#include <linux/page-flags.h>
222
223/*
224 * Methods to modify the page usage count.
225 *
226 * What counts for a page usage:
227 * - cache mapping (page->mapping)
228 * - private data (page->private)
229 * - page mapped in a task's page tables, each mapping
230 * is counted separately
231 *
232 * Also, many kernel routines increase the page count before a critical
233 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235
236/*
Nick Pigginda6052f2006-09-25 23:31:35 -0700237 * Drop a ref, return true if the refcount fell to zero (the page has no users)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800239static inline int put_page_testzero(struct page *page)
240{
Nick Piggin725d7042006-09-25 23:30:55 -0700241 VM_BUG_ON(atomic_read(&page->_count) == 0);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800242 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800243}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800246 * Try to grab a ref unless the page has a refcount of zero, return false if
247 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800249static inline int get_page_unless_zero(struct page *page)
250{
Christoph Lameteraec2c3e2008-02-04 22:28:35 -0800251 VM_BUG_ON(PageTail(page));
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800252 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800253}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
Christoph Lameter48667e72008-02-04 22:28:31 -0800255/* Support for virtually mapped pages */
Christoph Lameterb3bdda02008-02-04 22:28:32 -0800256struct page *vmalloc_to_page(const void *addr);
257unsigned long vmalloc_to_pfn(const void *addr);
Christoph Lameter48667e72008-02-04 22:28:31 -0800258
Paul Mundt0738c4b2008-03-12 16:51:31 +0900259/*
260 * Determine if an address is within the vmalloc range
261 *
262 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
263 * is no special casing required.
264 */
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800265static inline int is_vmalloc_addr(const void *x)
266{
Paul Mundt0738c4b2008-03-12 16:51:31 +0900267#ifdef CONFIG_MMU
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800268 unsigned long addr = (unsigned long)x;
269
270 return addr >= VMALLOC_START && addr < VMALLOC_END;
Paul Mundt0738c4b2008-03-12 16:51:31 +0900271#else
272 return 0;
David Howells8ca3ed82008-02-23 15:23:37 -0800273#endif
Paul Mundt0738c4b2008-03-12 16:51:31 +0900274}
Christoph Lameter9e2779f2008-02-04 22:28:34 -0800275
Christoph Lameterd85f3382007-05-06 14:49:39 -0700276static inline struct page *compound_head(struct page *page)
277{
Christoph Lameter6d777952007-05-06 14:49:40 -0700278 if (unlikely(PageTail(page)))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700279 return page->first_page;
280 return page;
281}
282
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700283static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700285 return atomic_read(&compound_head(page)->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286}
287
288static inline void get_page(struct page *page)
289{
Christoph Lameterd85f3382007-05-06 14:49:39 -0700290 page = compound_head(page);
Nick Piggin725d7042006-09-25 23:30:55 -0700291 VM_BUG_ON(atomic_read(&page->_count) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 atomic_inc(&page->_count);
293}
294
Christoph Lameterb49af682007-05-06 14:49:41 -0700295static inline struct page *virt_to_head_page(const void *x)
296{
297 struct page *page = virt_to_page(x);
298 return compound_head(page);
299}
300
Nick Piggin7835e982006-03-22 00:08:40 -0800301/*
302 * Setup the page count before being freed into the page allocator for
303 * the first time (boot or memory hotplug)
304 */
305static inline void init_page_count(struct page *page)
306{
307 atomic_set(&page->_count, 1);
308}
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310void put_page(struct page *page);
Alexander Zarochentsev1d7ea732006-08-13 23:24:27 -0700311void put_pages_list(struct list_head *pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800313void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800314
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315/*
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800316 * Compound pages have a destructor function. Provide a
317 * prototype for that function and accessor functions.
318 * These are _only_ valid on the head of a PG_compound page.
319 */
320typedef void compound_page_dtor(struct page *);
321
322static inline void set_compound_page_dtor(struct page *page,
323 compound_page_dtor *dtor)
324{
325 page[1].lru.next = (void *)dtor;
326}
327
328static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
329{
330 return (compound_page_dtor *)page[1].lru.next;
331}
332
Christoph Lameterd85f3382007-05-06 14:49:39 -0700333static inline int compound_order(struct page *page)
334{
Christoph Lameter6d777952007-05-06 14:49:40 -0700335 if (!PageHead(page))
Christoph Lameterd85f3382007-05-06 14:49:39 -0700336 return 0;
337 return (unsigned long)page[1].lru.prev;
338}
339
340static inline void set_compound_order(struct page *page, unsigned long order)
341{
342 page[1].lru.prev = (void *)order;
343}
344
Andy Whitcroft33f2ef82006-12-06 20:33:32 -0800345/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * Multiple processes may "see" the same page. E.g. for untouched
347 * mappings of /dev/null, all processes see the same page full of
348 * zeroes, and text pages of executables and shared libraries have
349 * only one copy in memory, at most, normally.
350 *
351 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700352 * page_count() == 0 means the page is free. page->lru is then used for
353 * freelist management in the buddy allocator.
Nick Pigginda6052f2006-09-25 23:31:35 -0700354 * page_count() > 0 means the page has been allocated.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700356 * Pages are allocated by the slab allocator in order to provide memory
357 * to kmalloc and kmem_cache_alloc. In this case, the management of the
358 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
359 * unless a particular usage is carefully commented. (the responsibility of
360 * freeing the kmalloc memory is the caller's, of course).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700362 * A page may be used by anyone else who does a __get_free_page().
363 * In this case, page_count still tracks the references, and should only
364 * be used through the normal accessor functions. The top bits of page->flags
365 * and page->virtual store page management information, but all other fields
366 * are unused and could be used privately, carefully. The management of this
367 * page is the responsibility of the one who allocated it, and those who have
368 * subsequently been given references to it.
369 *
370 * The other pages (we may call them "pagecache pages") are completely
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 * managed by the Linux memory manager: I/O, buffers, swapping etc.
372 * The following discussion applies only to them.
373 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700374 * A pagecache page contains an opaque `private' member, which belongs to the
375 * page's address_space. Usually, this is the address of a circular list of
376 * the page's disk buffers. PG_private must be set to tell the VM to call
377 * into the filesystem to release these pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700379 * A page may belong to an inode's memory mapping. In this case, page->mapping
380 * is the pointer to the inode, and page->index is the file offset of the page,
381 * in units of PAGE_CACHE_SIZE.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700383 * If pagecache pages are not associated with an inode, they are said to be
384 * anonymous pages. These may become associated with the swapcache, and in that
385 * case PG_swapcache is set, and page->private is an offset into the swapcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700387 * In either case (swapcache or inode backed), the pagecache itself holds one
388 * reference to the page. Setting PG_private should also increment the
389 * refcount. The each user mapping also has a reference to the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700391 * The pagecache pages are stored in a per-mapping radix tree, which is
392 * rooted at mapping->page_tree, and indexed by offset.
393 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
394 * lists, we instead now tag pages as dirty/writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 *
Nick Pigginda6052f2006-09-25 23:31:35 -0700396 * All pagecache pages may be subject to I/O:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 * - inode pages may need to be read from disk,
398 * - inode pages which have been modified and are MAP_SHARED may need
Nick Pigginda6052f2006-09-25 23:31:35 -0700399 * to be written back to the inode on disk,
400 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
401 * modified may need to be swapped out to swap space and (later) to be read
402 * back into memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 */
404
405/*
406 * The zone field is never updated after free_area_init_core()
407 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700409
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700410
411/*
412 * page->flags layout:
413 *
414 * There are three possibilities for how page->flags get
415 * laid out. The first is for the normal case, without
416 * sparsemem. The second is for sparsemem when there is
417 * plenty of space for node and section. The last is when
418 * we have run out of space and have to fall back to an
419 * alternate (slower) way of determining the node.
420 *
Christoph Lameter308c05e2008-04-28 02:12:43 -0700421 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
422 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
423 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700424 */
Christoph Lameter308c05e2008-04-28 02:12:43 -0700425#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700426#define SECTIONS_WIDTH SECTIONS_SHIFT
427#else
428#define SECTIONS_WIDTH 0
429#endif
430
431#define ZONES_WIDTH ZONES_SHIFT
432
Christoph Lameter9223b412008-04-28 02:12:48 -0700433#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700434#define NODES_WIDTH NODES_SHIFT
435#else
Christoph Lameter308c05e2008-04-28 02:12:43 -0700436#ifdef CONFIG_SPARSEMEM_VMEMMAP
437#error "Vmemmap: No space for nodes field in page flags"
438#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700439#define NODES_WIDTH 0
440#endif
441
442/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100443#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700444#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
445#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
446
447/*
448 * We are going to use the flags for the page to node mapping if its in
449 * there. This includes the case where there is no node, so it is implicit.
450 */
Christoph Lameter89689ae2006-12-06 20:31:45 -0800451#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
452#define NODE_NOT_IN_PAGE_FLAGS
453#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700454
455#ifndef PFN_SECTION_SHIFT
456#define PFN_SECTION_SHIFT 0
457#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700458
459/*
460 * Define the bit shifts to access each section. For non-existant
461 * sections we define the shift as 0; that plus a 0 mask ensures
462 * the compiler will optimise away reference to them.
463 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700464#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
465#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
466#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700467
Christoph Lameter89689ae2006-12-06 20:31:45 -0800468/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
469#ifdef NODE_NOT_IN_PAGEFLAGS
470#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800471#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
472 SECTIONS_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700473#else
Christoph Lameter89689ae2006-12-06 20:31:45 -0800474#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800475#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
476 NODES_PGOFF : ZONES_PGOFF)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700477#endif
Christoph Lameter89689ae2006-12-06 20:31:45 -0800478
Andy Whitcroftbd8029b2007-02-10 01:43:14 -0800479#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700480
Christoph Lameter9223b412008-04-28 02:12:48 -0700481#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
482#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
Dave Hansen348f8b62005-06-23 00:07:40 -0700483#endif
484
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700485#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
486#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
487#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Christoph Lameter89689ae2006-12-06 20:31:45 -0800488#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700489
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700490static inline enum zone_type page_zonenum(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491{
Dave Hansen348f8b62005-06-23 00:07:40 -0700492 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494
Christoph Lameter89689ae2006-12-06 20:31:45 -0800495/*
496 * The identification function is only used by the buddy allocator for
497 * determining if two pages could be buddies. We are not really
498 * identifying a zone since we could be using a the section number
499 * id if we have not node id available in page flags.
500 * We guarantee only that it will return the same value for two
501 * combinable pages in a zone.
502 */
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700503static inline int page_zone_id(struct page *page)
504{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800505 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506}
507
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800508static inline int zone_to_nid(struct zone *zone)
Christoph Lameter89fa3022006-09-25 23:31:55 -0700509{
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700510#ifdef CONFIG_NUMA
511 return zone->node;
512#else
513 return 0;
514#endif
Christoph Lameter89fa3022006-09-25 23:31:55 -0700515}
516
Christoph Lameter89689ae2006-12-06 20:31:45 -0800517#ifdef NODE_NOT_IN_PAGE_FLAGS
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800518extern int page_to_nid(struct page *page);
Christoph Lameter89689ae2006-12-06 20:31:45 -0800519#else
Andy Whitcroft25ba77c2006-12-06 20:33:03 -0800520static inline int page_to_nid(struct page *page)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700521{
Christoph Lameter89689ae2006-12-06 20:31:45 -0800522 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700523}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800524#endif
525
526static inline struct zone *page_zone(struct page *page)
527{
528 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
529}
530
Christoph Lameter308c05e2008-04-28 02:12:43 -0700531#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700532static inline unsigned long page_to_section(struct page *page)
533{
534 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
535}
Christoph Lameter308c05e2008-04-28 02:12:43 -0700536#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700537
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700538static inline void set_page_zone(struct page *page, enum zone_type zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539{
Dave Hansen348f8b62005-06-23 00:07:40 -0700540 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
541 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
542}
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700543
Dave Hansen348f8b62005-06-23 00:07:40 -0700544static inline void set_page_node(struct page *page, unsigned long node)
545{
546 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
547 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
548}
Christoph Lameter89689ae2006-12-06 20:31:45 -0800549
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700550static inline void set_page_section(struct page *page, unsigned long section)
551{
552 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
553 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
554}
Dave Hansen348f8b62005-06-23 00:07:40 -0700555
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700556static inline void set_page_links(struct page *page, enum zone_type zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700557 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700558{
559 set_page_zone(page, zone);
560 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700561 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562}
563
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700564/*
Eric Paris7cd94142007-11-26 18:47:40 -0500565 * If a hint addr is less than mmap_min_addr change hint to be as
566 * low as possible but still greater than mmap_min_addr
567 */
568static inline unsigned long round_hint_to_min(unsigned long hint)
569{
570#ifdef CONFIG_SECURITY
571 hint &= PAGE_MASK;
572 if (((void *)hint != NULL) &&
573 (hint < mmap_min_addr))
574 return PAGE_ALIGN(mmap_min_addr);
575#endif
576 return hint;
577}
578
579/*
Christoph Lameterf6ac2352006-06-30 01:55:32 -0700580 * Some inline functions in vmstat.h depend on page_zone()
581 */
582#include <linux/vmstat.h>
583
Ingo Molnar652050a2006-01-14 13:21:30 -0800584static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585{
586 return __va(page_to_pfn(page) << PAGE_SHIFT);
587}
588
589#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
590#define HASHED_PAGE_VIRTUAL
591#endif
592
593#if defined(WANT_PAGE_VIRTUAL)
594#define page_address(page) ((page)->virtual)
595#define set_page_address(page, address) \
596 do { \
597 (page)->virtual = (address); \
598 } while(0)
599#define page_address_init() do { } while(0)
600#endif
601
602#if defined(HASHED_PAGE_VIRTUAL)
603void *page_address(struct page *page);
604void set_page_address(struct page *page, void *virtual);
605void page_address_init(void);
606#endif
607
608#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
609#define page_address(page) lowmem_page_address(page)
610#define set_page_address(page, address) do { } while(0)
611#define page_address_init() do { } while(0)
612#endif
613
614/*
615 * On an anonymous page mapped into a user virtual memory area,
616 * page->mapping points to its anon_vma, not to a struct address_space;
617 * with the PAGE_MAPPING_ANON bit set to distinguish it.
618 *
619 * Please note that, confusingly, "page_mapping" refers to the inode
620 * address_space which maps the page from disk; whereas "page_mapped"
621 * refers to user virtual address space into which the page is mapped.
622 */
623#define PAGE_MAPPING_ANON 1
624
625extern struct address_space swapper_space;
626static inline struct address_space *page_mapping(struct page *page)
627{
628 struct address_space *mapping = page->mapping;
629
Christoph Lameterb5fab14e2007-07-17 04:03:33 -0700630 VM_BUG_ON(PageSlab(page));
Andrew Morton726b8012008-04-28 02:12:44 -0700631#ifdef CONFIG_SWAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 if (unlikely(PageSwapCache(page)))
633 mapping = &swapper_space;
Andrew Morton726b8012008-04-28 02:12:44 -0700634 else
635#endif
636 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 mapping = NULL;
638 return mapping;
639}
640
641static inline int PageAnon(struct page *page)
642{
643 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
644}
645
646/*
647 * Return the pagecache index of the passed page. Regular pagecache pages
648 * use ->index whereas swapcache pages use ->private
649 */
650static inline pgoff_t page_index(struct page *page)
651{
652 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700653 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 return page->index;
655}
656
657/*
658 * The atomic page->_mapcount, like _count, starts from -1:
659 * so that transitions both from it and to it can be tracked,
660 * using atomic_inc_and_test and atomic_add_negative(-1).
661 */
662static inline void reset_page_mapcount(struct page *page)
663{
664 atomic_set(&(page)->_mapcount, -1);
665}
666
667static inline int page_mapcount(struct page *page)
668{
669 return atomic_read(&(page)->_mapcount) + 1;
670}
671
672/*
673 * Return true if this page is mapped into pagetables.
674 */
675static inline int page_mapped(struct page *page)
676{
677 return atomic_read(&(page)->_mapcount) >= 0;
678}
679
680/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 * Different kinds of faults, as returned by handle_mm_fault().
682 * Used to decide whether a process gets delivered SIGBUS or
683 * just gets major/minor fault counters bumped up.
684 */
Nick Piggind0217ac2007-07-19 01:47:03 -0700685
Nick Piggin83c54072007-07-19 01:47:05 -0700686#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
Nick Piggind0217ac2007-07-19 01:47:03 -0700687
Nick Piggin83c54072007-07-19 01:47:05 -0700688#define VM_FAULT_OOM 0x0001
689#define VM_FAULT_SIGBUS 0x0002
690#define VM_FAULT_MAJOR 0x0004
691#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000692
Nick Piggin83c54072007-07-19 01:47:05 -0700693#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
694#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
Nick Piggin83c54072007-07-19 01:47:05 -0700696#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
Nick Piggind0217ac2007-07-19 01:47:03 -0700697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
699
700extern void show_free_areas(void);
701
702#ifdef CONFIG_SHMEM
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700703extern int shmem_lock(struct file *file, int lock, struct user_struct *user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704#else
Russell King03b00eb2006-01-06 00:10:52 -0800705static inline int shmem_lock(struct file *file, int lock,
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700706 struct user_struct *user)
Russell King03b00eb2006-01-06 00:10:52 -0800707{
708 return 0;
709}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710#endif
711struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
712
713int shmem_zero_setup(struct vm_area_struct *);
714
David Howellsb0e15192006-01-06 00:11:42 -0800715#ifndef CONFIG_MMU
716extern unsigned long shmem_get_unmapped_area(struct file *file,
717 unsigned long addr,
718 unsigned long len,
719 unsigned long pgoff,
720 unsigned long flags);
721#endif
722
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +0400723extern int can_do_mlock(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724extern int user_shm_lock(size_t, struct user_struct *);
725extern void user_shm_unlock(size_t, struct user_struct *);
726
727/*
728 * Parameter block passed down to zap_pte_range in exceptional cases.
729 */
730struct zap_details {
731 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
732 struct address_space *check_mapping; /* Check page->mapping if set */
733 pgoff_t first_index; /* Lowest page->index to unmap */
734 pgoff_t last_index; /* Highest page->index to unmap */
735 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 unsigned long truncate_count; /* Compare vm_truncate_count */
737};
738
Nick Piggin7e675132008-04-28 02:13:00 -0700739struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
740 pte_t pte);
741
Jack Steinerc627f9c2008-07-29 22:33:53 -0700742int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
743 unsigned long size);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700744unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700746unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 struct vm_area_struct *start_vma, unsigned long start_addr,
748 unsigned long end_addr, unsigned long *nr_accounted,
749 struct zap_details *);
Matt Mackalle6473092008-02-04 22:29:01 -0800750
751/**
752 * mm_walk - callbacks for walk_page_range
753 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
754 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
755 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
756 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
757 * @pte_hole: if set, called for each hole at all levels
758 *
759 * (see walk_page_range for more details)
760 */
761struct mm_walk {
Dave Hansen21650092008-06-12 15:21:47 -0700762 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
763 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
764 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
765 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
766 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
767 struct mm_struct *mm;
768 void *private;
Matt Mackalle6473092008-02-04 22:29:01 -0800769};
770
Dave Hansen21650092008-06-12 15:21:47 -0700771int walk_page_range(unsigned long addr, unsigned long end,
772 struct mm_walk *walk);
Jan Beulich42b77722008-07-23 21:27:10 -0700773void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700774 unsigned long end, unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
776 struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777void unmap_mapping_range(struct address_space *mapping,
778 loff_t const holebegin, loff_t const holelen, int even_cows);
Rik van Riel28b2ee22008-07-23 21:27:05 -0700779int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
780 void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782static inline void unmap_shared_mapping_range(struct address_space *mapping,
783 loff_t const holebegin, loff_t const holelen)
784{
785 unmap_mapping_range(mapping, holebegin, holelen, 0);
786}
787
788extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800789extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000790
David Howells7ee1dd32006-01-06 00:11:44 -0800791#ifdef CONFIG_MMU
Nick Piggin83c54072007-07-19 01:47:05 -0700792extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
David Howells7ee1dd32006-01-06 00:11:44 -0800793 unsigned long address, int write_access);
David Howells7ee1dd32006-01-06 00:11:44 -0800794#else
795static inline int handle_mm_fault(struct mm_struct *mm,
796 struct vm_area_struct *vma, unsigned long address,
797 int write_access)
798{
799 /* should never happen if there's no MMU */
800 BUG();
801 return VM_FAULT_SIGBUS;
802}
803#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000804
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805extern int make_pages_present(unsigned long addr, unsigned long end);
806extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
808int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
809 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
810
David Howellscf9a2ae2006-08-29 19:05:54 +0100811extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
812extern void do_invalidatepage(struct page *page, unsigned long offset);
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814int __set_page_dirty_nobuffers(struct page *page);
Ken Chen76719322007-02-10 01:43:15 -0800815int __set_page_dirty_no_writeback(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816int redirty_page_for_writepage(struct writeback_control *wbc,
817 struct page *page);
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800818int set_page_dirty(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819int set_page_dirty_lock(struct page *page);
820int clear_page_dirty_for_io(struct page *page);
821
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700822extern unsigned long move_page_tables(struct vm_area_struct *vma,
823 unsigned long old_addr, struct vm_area_struct *new_vma,
824 unsigned long new_addr, unsigned long len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825extern unsigned long do_mremap(unsigned long addr,
826 unsigned long old_len, unsigned long new_len,
827 unsigned long flags, unsigned long new_addr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700828extern int mprotect_fixup(struct vm_area_struct *vma,
829 struct vm_area_struct **pprev, unsigned long start,
830 unsigned long end, unsigned long newflags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
Nick Piggin21cc1992008-07-25 19:45:22 -0700832/*
833 * get_user_pages_fast provides equivalent functionality to get_user_pages,
834 * operating on current and current->mm (force=0 and doesn't return any vmas).
835 *
836 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
837 * can be made about locking. get_user_pages_fast is to be implemented in a
838 * way that is advantageous (vs get_user_pages()) when the user memory area is
839 * already faulted in and present in ptes. However if the pages have to be
840 * faulted in, it may turn out to be slightly slower).
841 */
842int get_user_pages_fast(unsigned long start, int nr_pages, int write,
843 struct page **pages);
844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845/*
Rusty Russell8e1f9362007-07-17 04:03:17 -0700846 * A callback you can register to apply pressure to ageable caches.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700848 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
849 * look through the least-recently-used 'nr_to_scan' entries and
850 * attempt to free them up. It should return the number of objects
851 * which remain in the cache. If it returns -1, it means it cannot do
852 * any scanning at this time (eg. there is a risk of deadlock).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 *
Rusty Russell8e1f9362007-07-17 04:03:17 -0700854 * The 'gfpmask' refers to the allocation we are currently trying to
855 * fulfil.
856 *
857 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
858 * querying the cache size, so a fastpath for that case is appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 */
Rusty Russell8e1f9362007-07-17 04:03:17 -0700860struct shrinker {
861 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
862 int seeks; /* seeks to recreate an obj */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
Rusty Russell8e1f9362007-07-17 04:03:17 -0700864 /* These are for internal use */
865 struct list_head list;
866 long nr; /* objs pending delete */
867};
868#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
869extern void register_shrinker(struct shrinker *);
870extern void unregister_shrinker(struct shrinker *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
Alexey Dobriyan4e950f62007-07-30 02:36:13 +0400872int vma_wants_writenotify(struct vm_area_struct *vma);
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700873
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800874extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800875
Nick Piggin5f22df02007-05-06 14:49:02 -0700876#ifdef __PAGETABLE_PUD_FOLDED
877static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
878 unsigned long address)
879{
880 return 0;
881}
882#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700883int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700884#endif
885
886#ifdef __PAGETABLE_PMD_FOLDED
887static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
888 unsigned long address)
889{
890 return 0;
891}
892#else
Hugh Dickins1bb36302005-10-29 18:16:22 -0700893int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
Nick Piggin5f22df02007-05-06 14:49:02 -0700894#endif
895
Hugh Dickins1bb36302005-10-29 18:16:22 -0700896int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
897int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
898
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899/*
900 * The following ifdef needed to get the 4level-fixup.h header to work.
901 * Remove it when 4level-fixup.h has been removed.
902 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700903#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
905{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700906 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
907 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908}
909
910static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
911{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700912 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
913 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700915#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
916
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700917#if USE_SPLIT_PTLOCKS
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700918/*
919 * We tuck a spinlock to guard each pagetable page into its struct page,
920 * at page->private, with BUILD_BUG_ON to make sure that this will not
921 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
922 * When freeing, reset page->mapping so free_pages_check won't complain.
923 */
Andrew Morton349aef02006-01-08 01:04:36 -0800924#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700925#define pte_lock_init(_page) do { \
926 spin_lock_init(__pte_lockptr(_page)); \
927} while (0)
928#define pte_lock_deinit(page) ((page)->mapping = NULL)
929#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700930#else /* !USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700931/*
932 * We use mm->page_table_lock to guard all pagetable pages of the mm.
933 */
934#define pte_lock_init(page) do {} while (0)
935#define pte_lock_deinit(page) do {} while (0)
936#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
Jeremy Fitzhardingef7d0b922008-09-09 15:43:22 -0700937#endif /* USE_SPLIT_PTLOCKS */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700938
Martin Schwidefsky2f569af2008-02-08 04:22:04 -0800939static inline void pgtable_page_ctor(struct page *page)
940{
941 pte_lock_init(page);
942 inc_zone_page_state(page, NR_PAGETABLE);
943}
944
945static inline void pgtable_page_dtor(struct page *page)
946{
947 pte_lock_deinit(page);
948 dec_zone_page_state(page, NR_PAGETABLE);
949}
950
Hugh Dickinsc74df322005-10-29 18:16:23 -0700951#define pte_offset_map_lock(mm, pmd, address, ptlp) \
952({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700953 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700954 pte_t *__pte = pte_offset_map(pmd, address); \
955 *(ptlp) = __ptl; \
956 spin_lock(__ptl); \
957 __pte; \
958})
959
960#define pte_unmap_unlock(pte, ptl) do { \
961 spin_unlock(ptl); \
962 pte_unmap(pte); \
963} while (0)
964
Hugh Dickins1bb36302005-10-29 18:16:22 -0700965#define pte_alloc_map(mm, pmd, address) \
966 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
967 NULL: pte_offset_map(pmd, address))
968
Hugh Dickinsc74df322005-10-29 18:16:23 -0700969#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
970 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
971 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
972
Hugh Dickins1bb36302005-10-29 18:16:22 -0700973#define pte_alloc_kernel(pmd, address) \
974 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
975 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
977extern void free_area_init(unsigned long * zones_size);
Johannes Weiner9109fb72008-07-23 21:27:20 -0700978extern void free_area_init_node(int nid, unsigned long * zones_size,
979 unsigned long zone_start_pfn, unsigned long *zholes_size);
Mel Gormanc7132162006-09-27 01:49:43 -0700980#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
981/*
982 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
983 * zones, allocate the backing mem_map and account for memory holes in a more
984 * architecture independent manner. This is a substitute for creating the
985 * zone_sizes[] and zholes_size[] arrays and passing them to
986 * free_area_init_node()
987 *
988 * An architecture is expected to register range of page frames backed by
989 * physical memory with add_active_range() before calling
990 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
991 * usage, an architecture is expected to do something like
992 *
993 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
994 * max_highmem_pfn};
995 * for_each_valid_physical_page_range()
996 * add_active_range(node_id, start_pfn, end_pfn)
997 * free_area_init_nodes(max_zone_pfns);
998 *
999 * If the architecture guarantees that there are no holes in the ranges
1000 * registered with add_active_range(), free_bootmem_active_regions()
1001 * will call free_bootmem_node() for each registered physical page range.
1002 * Similarly sparse_memory_present_with_active_regions() calls
1003 * memory_present() for each range when SPARSEMEM is enabled.
1004 *
1005 * See mm/page_alloc.c for more information on each function exposed by
1006 * CONFIG_ARCH_POPULATES_NODE_MAP
1007 */
1008extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1009extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1010 unsigned long end_pfn);
Yinghai Lucc1050b2008-06-13 19:08:52 -07001011extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1012 unsigned long end_pfn);
Mel Gormanfb014392006-09-27 01:49:59 -07001013extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1014 unsigned long end_pfn);
Mel Gormanc7132162006-09-27 01:49:43 -07001015extern void remove_all_active_ranges(void);
1016extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1017 unsigned long end_pfn);
1018extern void get_pfn_range_for_nid(unsigned int nid,
1019 unsigned long *start_pfn, unsigned long *end_pfn);
1020extern unsigned long find_min_pfn_with_active_regions(void);
Mel Gormanc7132162006-09-27 01:49:43 -07001021extern void free_bootmem_with_active_regions(int nid,
1022 unsigned long max_low_pfn);
Yinghai Lud52d53b2008-06-16 20:10:55 -07001023typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
Yinghai Lub5bc6c02008-06-14 18:32:52 -07001024extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
Mel Gormanc7132162006-09-27 01:49:43 -07001025extern void sparse_memory_present_with_active_regions(int nid);
1026#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1027extern int early_pfn_to_nid(unsigned long pfn);
1028#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1029#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
Mel Gorman0e0b8642006-09-27 01:49:56 -07001030extern void set_dma_reserve(unsigned long new_dma_reserve);
Dave Hansena2f3aa022007-01-10 23:15:30 -08001031extern void memmap_init_zone(unsigned long, int, unsigned long,
1032 unsigned long, enum memmap_context);
Dave Hansen3947be12005-10-29 18:16:54 -07001033extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034extern void mem_init(void);
1035extern void show_mem(void);
1036extern void si_meminfo(struct sysinfo * val);
1037extern void si_meminfo_node(struct sysinfo *val, int nid);
Mike Travis3461b0a2008-05-12 21:21:13 +02001038extern int after_bootmem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
Christoph Lametere7c8d5c2005-06-21 17:14:47 -07001040#ifdef CONFIG_NUMA
1041extern void setup_per_cpu_pageset(void);
1042#else
1043static inline void setup_per_cpu_pageset(void) {}
1044#endif
1045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046/* prio_tree.c */
1047void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1048void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1049void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1050struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1051 struct prio_tree_iter *iter);
1052
1053#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1054 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1055 (vma = vma_prio_tree_next(vma, iter)); )
1056
1057static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1058 struct list_head *list)
1059{
1060 vma->shared.vm_set.parent = NULL;
1061 list_add_tail(&vma->shared.vm_set.list, list);
1062}
1063
1064/* mmap.c */
Alan Cox34b4e4a2007-08-22 14:01:28 -07001065extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1067 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1068extern struct vm_area_struct *vma_merge(struct mm_struct *,
1069 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1070 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1071 struct mempolicy *);
1072extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1073extern int split_vma(struct mm_struct *,
1074 struct vm_area_struct *, unsigned long addr, int new_below);
1075extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1076extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1077 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -07001078extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1080 unsigned long addr, unsigned long len, pgoff_t pgoff);
1081extern void exit_mmap(struct mm_struct *);
Matt Helsley925d1c42008-04-29 01:01:36 -07001082
Andrea Arcangeli7906d002008-07-28 15:46:26 -07001083extern int mm_take_all_locks(struct mm_struct *mm);
1084extern void mm_drop_all_locks(struct mm_struct *mm);
1085
Matt Helsley925d1c42008-04-29 01:01:36 -07001086#ifdef CONFIG_PROC_FS
1087/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1088extern void added_exe_file_vma(struct mm_struct *mm);
1089extern void removed_exe_file_vma(struct mm_struct *mm);
1090#else
1091static inline void added_exe_file_vma(struct mm_struct *mm)
1092{}
1093
1094static inline void removed_exe_file_vma(struct mm_struct *mm)
1095{}
1096#endif /* CONFIG_PROC_FS */
1097
akpm@osdl.org119f6572005-05-01 08:58:35 -07001098extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Roland McGrathfa5dc222007-02-08 14:20:41 -08001099extern int install_special_mapping(struct mm_struct *mm,
1100 unsigned long addr, unsigned long len,
1101 unsigned long flags, struct page **pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
1103extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1104
1105extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1106 unsigned long len, unsigned long prot,
1107 unsigned long flag, unsigned long pgoff);
Miklos Szeredi0165ab42007-07-15 23:38:26 -07001108extern unsigned long mmap_region(struct file *file, unsigned long addr,
1109 unsigned long len, unsigned long flags,
1110 unsigned int vm_flags, unsigned long pgoff,
1111 int accountable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1114 unsigned long len, unsigned long prot,
1115 unsigned long flag, unsigned long offset)
1116{
1117 unsigned long ret = -EINVAL;
1118 if ((offset + PAGE_ALIGN(len)) < offset)
1119 goto out;
1120 if (!(offset & ~PAGE_MASK))
1121 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1122out:
1123 return ret;
1124}
1125
1126extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1127
1128extern unsigned long do_brk(unsigned long, unsigned long);
1129
1130/* filemap.c */
1131extern unsigned long page_unuse(struct page *);
1132extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -08001133extern void truncate_inode_pages_range(struct address_space *,
1134 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136/* generic vm_area_ops exported for stackable file systems */
Nick Piggind0217ac2007-07-19 01:47:03 -07001137extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
1139/* mm/page-writeback.c */
1140int write_one_page(struct page *page, int wait);
1141
1142/* readahead.c */
1143#define VM_MAX_READAHEAD 128 /* kbytes */
1144#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001147 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -08001149 pgoff_t offset, unsigned long nr_to_read);
Rusty Russellcf914a72007-07-19 01:48:08 -07001150
1151void page_cache_sync_readahead(struct address_space *mapping,
1152 struct file_ra_state *ra,
1153 struct file *filp,
1154 pgoff_t offset,
1155 unsigned long size);
1156
1157void page_cache_async_readahead(struct address_space *mapping,
1158 struct file_ra_state *ra,
1159 struct file *filp,
1160 struct page *pg,
1161 pgoff_t offset,
1162 unsigned long size);
1163
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164unsigned long max_sane_readahead(unsigned long nr);
1165
1166/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001167extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001168#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -07001169extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -05001170#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001171extern int expand_stack_downwards(struct vm_area_struct *vma,
1172 unsigned long address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173
1174/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1175extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1176extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1177 struct vm_area_struct **pprev);
1178
1179/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1180 NULL if none. Assume start_addr < end_addr. */
1181static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1182{
1183 struct vm_area_struct * vma = find_vma(mm,start_addr);
1184
1185 if (vma && end_addr <= vma->vm_start)
1186 vma = NULL;
1187 return vma;
1188}
1189
1190static inline unsigned long vma_pages(struct vm_area_struct *vma)
1191{
1192 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1193}
1194
Hugh Dickins804af2c2006-07-26 21:39:49 +01001195pgprot_t vm_get_page_prot(unsigned long vm_flags);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001196struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001197int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1198 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001199int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Nick Piggine0dc0d82007-02-12 00:51:36 -08001200int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1201 unsigned long pfn);
Nick Piggin423bad62008-04-28 02:13:01 -07001202int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1203 unsigned long pfn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
Linus Torvalds6aab3412005-11-28 14:34:23 -08001205struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001206 unsigned int foll_flags);
1207#define FOLL_WRITE 0x01 /* check pte is writable */
1208#define FOLL_TOUCH 0x02 /* mark page accessed */
1209#define FOLL_GET 0x04 /* do get_page on page */
1210#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
Martin Schwidefsky2f569af2008-02-08 04:22:04 -08001212typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
Jeremy Fitzhardingeaee16b32007-05-06 14:48:54 -07001213 void *data);
1214extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1215 unsigned long size, pte_fn_t fn, void *data);
1216
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001218void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001220static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 unsigned long flags, struct file *file, long pages)
1222{
1223}
1224#endif /* CONFIG_PROC_FS */
1225
Ingo Molnar12d6f212008-01-30 13:33:58 +01001226#ifdef CONFIG_DEBUG_PAGEALLOC
1227extern int debug_pagealloc_enabled;
1228
1229extern void kernel_map_pages(struct page *page, int numpages, int enable);
1230
1231static inline void enable_debug_pagealloc(void)
1232{
1233 debug_pagealloc_enabled = 1;
1234}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001235#ifdef CONFIG_HIBERNATION
1236extern bool kernel_page_present(struct page *page);
1237#endif /* CONFIG_HIBERNATION */
Ingo Molnar12d6f212008-01-30 13:33:58 +01001238#else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239static inline void
Nick Piggin9858db52006-10-11 01:21:30 -07001240kernel_map_pages(struct page *page, int numpages, int enable) {}
Ingo Molnar12d6f212008-01-30 13:33:58 +01001241static inline void enable_debug_pagealloc(void)
1242{
1243}
Rafael J. Wysocki8a235ef2008-02-20 01:47:44 +01001244#ifdef CONFIG_HIBERNATION
1245static inline bool kernel_page_present(struct page *page) { return true; }
1246#endif /* CONFIG_HIBERNATION */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247#endif
1248
1249extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1250#ifdef __HAVE_ARCH_GATE_AREA
1251int in_gate_area_no_task(unsigned long addr);
1252int in_gate_area(struct task_struct *task, unsigned long addr);
1253#else
1254int in_gate_area_no_task(unsigned long addr);
1255#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1256#endif /* __HAVE_ARCH_GATE_AREA */
1257
Andrew Morton9d0243b2006-01-08 01:00:39 -08001258int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1259 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001260unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001261 unsigned long lru_pages);
Andrew Morton9d0243b2006-01-08 01:00:39 -08001262
Luke Yang7a9166e2006-02-20 18:28:07 -08001263#ifndef CONFIG_MMU
1264#define randomize_va_space 0
1265#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001266extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001267#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001268
Sam Ravnborg045e72a2007-07-26 10:41:13 -07001269const char * arch_vma_name(struct vm_area_struct *vma);
Andi Kleen03252912008-01-30 13:33:18 +01001270void print_vma_addr(char *prefix, unsigned long rip);
Ingo Molnare6e54942006-06-27 02:53:50 -07001271
Yasunori Goto98f3cfc2007-10-16 01:26:14 -07001272struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001273pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1274pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1275pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1276pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001277void *vmemmap_alloc_block(unsigned long size, int node);
1278void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
Andy Whitcroft29c71112007-10-16 01:24:14 -07001279int vmemmap_populate_basepages(struct page *start_page,
1280 unsigned long pages, int node);
1281int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
Yinghai Luc2b91e22008-04-12 01:19:24 -07001282void vmemmap_populate_print_last(void);
Christoph Lameter8f6aac42007-10-16 01:24:13 -07001283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284#endif /* __KERNEL__ */
1285#endif /* _LINUX_MM_H */