blob: a929ea197e4844da5d11fc9f43da506138b276f7 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
4#include <linux/sched.h>
5#include <linux/errno.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -08006#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007
8#ifdef __KERNEL__
9
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/gfp.h>
11#include <linux/list.h>
12#include <linux/mmzone.h>
13#include <linux/rbtree.h>
14#include <linux/prio_tree.h>
15#include <linux/fs.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080016#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017
18struct mempolicy;
19struct anon_vma;
20
21#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
22extern unsigned long max_mapnr;
23#endif
24
25extern unsigned long num_physpages;
26extern void * high_memory;
27extern unsigned long vmalloc_earlyreserve;
28extern int page_cluster;
29
30#ifdef CONFIG_SYSCTL
31extern int sysctl_legacy_va_layout;
32#else
33#define sysctl_legacy_va_layout 0
34#endif
35
36#include <asm/page.h>
37#include <asm/pgtable.h>
38#include <asm/processor.h>
39#include <asm/atomic.h>
40
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
42
43/*
44 * Linux kernel virtual memory manager primitives.
45 * The idea being to have a "virtual" mm in the same way
46 * we have a virtual fs - giving a cleaner interface to the
47 * mm details, and allowing different kinds of memory mappings
48 * (from shared memory to executable loading to arbitrary
49 * mmap() functions).
50 */
51
52/*
53 * This struct defines a memory VMM memory area. There is one of these
54 * per VM-area/task. A VM area is any part of the process virtual memory
55 * space that has a special rule for the page-fault handlers (ie a shared
56 * library, the executable area etc).
57 */
58struct vm_area_struct {
59 struct mm_struct * vm_mm; /* The address space we belong to. */
60 unsigned long vm_start; /* Our start address within vm_mm. */
61 unsigned long vm_end; /* The first byte after our end address
62 within vm_mm. */
63
64 /* linked list of VM areas per task, sorted by address */
65 struct vm_area_struct *vm_next;
66
67 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
68 unsigned long vm_flags; /* Flags, listed below. */
69
70 struct rb_node vm_rb;
71
72 /*
73 * For areas with an address space and backing store,
74 * linkage into the address_space->i_mmap prio tree, or
75 * linkage to the list of like vmas hanging off its node, or
76 * linkage of vma in the address_space->i_mmap_nonlinear list.
77 */
78 union {
79 struct {
80 struct list_head list;
81 void *parent; /* aligns with prio_tree_node parent */
82 struct vm_area_struct *head;
83 } vm_set;
84
85 struct raw_prio_tree_node prio_tree_node;
86 } shared;
87
88 /*
89 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
90 * list, after a COW of one of the file pages. A MAP_SHARED vma
91 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
92 * or brk vma (with NULL file) can only be in an anon_vma list.
93 */
94 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
95 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
96
97 /* Function pointers to deal with this struct. */
98 struct vm_operations_struct * vm_ops;
99
100 /* Information about our backing store: */
101 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
102 units, *not* PAGE_CACHE_SIZE */
103 struct file * vm_file; /* File we map to (can be NULL). */
104 void * vm_private_data; /* was vm_pte (shared mem) */
105 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
106
107#ifndef CONFIG_MMU
108 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
109#endif
110#ifdef CONFIG_NUMA
111 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
112#endif
113};
114
115/*
116 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
117 * disabled, then there's a single shared list of VMAs maintained by the
118 * system, and mm's subscribe to these individually
119 */
120struct vm_list_struct {
121 struct vm_list_struct *next;
122 struct vm_area_struct *vma;
123};
124
125#ifndef CONFIG_MMU
126extern struct rb_root nommu_vma_tree;
127extern struct rw_semaphore nommu_vma_sem;
128
129extern unsigned int kobjsize(const void *objp);
130#endif
131
132/*
133 * vm_flags..
134 */
135#define VM_READ 0x00000001 /* currently active flags */
136#define VM_WRITE 0x00000002
137#define VM_EXEC 0x00000004
138#define VM_SHARED 0x00000008
139
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700140/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
142#define VM_MAYWRITE 0x00000020
143#define VM_MAYEXEC 0x00000040
144#define VM_MAYSHARE 0x00000080
145
146#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
147#define VM_GROWSUP 0x00000200
Linus Torvalds6aab3412005-11-28 14:34:23 -0800148#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
150
151#define VM_EXECUTABLE 0x00001000
152#define VM_LOCKED 0x00002000
153#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
154
155 /* Used by sys_madvise() */
156#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
157#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
158
159#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
160#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800161#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
163#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
164#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
165#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800166#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
168#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
169#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
170#endif
171
172#ifdef CONFIG_STACK_GROWSUP
173#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
174#else
175#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176#endif
177
178#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
179#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
180#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
181#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
182#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
183
184/*
185 * mapping from the currently active vm_flags protection bits (the
186 * low four bits) to a page protection mask..
187 */
188extern pgprot_t protection_map[16];
189
190
191/*
192 * These are the virtual MM functions - opening of an area, closing and
193 * unmapping it (needed to keep files on disk up-to-date etc), pointer
194 * to the functions called when a no-page or a wp-page exception occurs.
195 */
196struct vm_operations_struct {
197 void (*open)(struct vm_area_struct * area);
198 void (*close)(struct vm_area_struct * area);
199 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
200 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
David Howells9637a5e2006-06-23 02:03:43 -0700201
202 /* notification that a previously read-only page is about to become
203 * writable, if an error is returned it will cause a SIGBUS */
204 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205#ifdef CONFIG_NUMA
206 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
207 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
208 unsigned long addr);
Christoph Lameter7b2259b2006-06-25 05:46:48 -0700209 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
210 const nodemask_t *to, unsigned long flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211#endif
212};
213
214struct mmu_gather;
215struct inode;
216
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217/*
218 * Each physical page in the system has a struct page associated with
219 * it to keep track of whatever it is we are using the page for at the
220 * moment. Note that we have no way to track which tasks are using
221 * a page.
222 */
223struct page {
Andi Kleen07808b72005-11-05 17:25:53 +0100224 unsigned long flags; /* Atomic flags, some possibly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 * updated asynchronously */
226 atomic_t _count; /* Usage count, see below. */
227 atomic_t _mapcount; /* Count of ptes mapped in mms,
228 * to show when page is mapped
229 * & limit reverse map searches.
230 */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700231 union {
Andrew Morton349aef02006-01-08 01:04:36 -0800232 struct {
233 unsigned long private; /* Mapping-private opaque data:
234 * usually used for buffer_heads
235 * if PagePrivate set; used for
Nick Piggin676165a2006-04-10 11:21:48 +1000236 * swp_entry_t if PageSwapCache;
Andrew Morton349aef02006-01-08 01:04:36 -0800237 * indicates order in the buddy
Nick Piggin676165a2006-04-10 11:21:48 +1000238 * system if PG_buddy is set.
Andrew Morton349aef02006-01-08 01:04:36 -0800239 */
240 struct address_space *mapping; /* If low bit clear, points to
241 * inode address_space, or NULL.
242 * If page mapped as anonymous
243 * memory, low bit is set, and
244 * it points to anon_vma object:
245 * see PAGE_MAPPING_ANON below.
246 */
247 };
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700248#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
Andrew Morton349aef02006-01-08 01:04:36 -0800249 spinlock_t ptl;
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700250#endif
Andrew Morton349aef02006-01-08 01:04:36 -0800251 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 pgoff_t index; /* Our offset within mapping. */
253 struct list_head lru; /* Pageout list, eg. active_list
254 * protected by zone->lru_lock !
255 */
256 /*
257 * On machines where all RAM is mapped into kernel address space,
258 * we can simply calculate the virtual address. On machines with
259 * highmem some memory is mapped into kernel virtual memory
260 * dynamically, so we need a place to store that address.
261 * Note that this field could be 16 bits on x86 ... ;)
262 *
263 * Architectures with slow multiplication can define
264 * WANT_PAGE_VIRTUAL in asm/page.h
265 */
266#if defined(WANT_PAGE_VIRTUAL)
267 void *virtual; /* Kernel virtual address (NULL if
268 not kmapped, ie. highmem) */
269#endif /* WANT_PAGE_VIRTUAL */
270};
271
Andrew Morton349aef02006-01-08 01:04:36 -0800272#define page_private(page) ((page)->private)
273#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275/*
276 * FIXME: take this include out, include page-flags.h in
277 * files which need it (119 of them)
278 */
279#include <linux/page-flags.h>
280
281/*
282 * Methods to modify the page usage count.
283 *
284 * What counts for a page usage:
285 * - cache mapping (page->mapping)
286 * - private data (page->private)
287 * - page mapped in a task's page tables, each mapping
288 * is counted separately
289 *
290 * Also, many kernel routines increase the page count before a critical
291 * routine so they can be sure the page doesn't go away from under them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 */
293
294/*
295 * Drop a ref, return true if the logical refcount fell to zero (the page has
296 * no users)
297 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800298static inline int put_page_testzero(struct page *page)
299{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800300 BUG_ON(atomic_read(&page->_count) == 0);
301 return atomic_dec_and_test(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800302}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
304/*
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800305 * Try to grab a ref unless the page has a refcount of zero, return false if
306 * that is the case.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 */
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800308static inline int get_page_unless_zero(struct page *page)
309{
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800310 return atomic_inc_not_zero(&page->_count);
Nick Piggin7c8ee9a2006-03-22 00:08:03 -0800311}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313extern void FASTCALL(__page_cache_release(struct page *));
314
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700315static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316{
Nick Piggin617d2212006-03-22 00:08:43 -0800317 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700318 page = (struct page *)page_private(page);
Nick Piggin8dc04ef2006-03-22 00:08:03 -0800319 return atomic_read(&page->_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320}
321
322static inline void get_page(struct page *page)
323{
324 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700325 page = (struct page *)page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 atomic_inc(&page->_count);
327}
328
Nick Piggin7835e982006-03-22 00:08:40 -0800329/*
330 * Setup the page count before being freed into the page allocator for
331 * the first time (boot or memory hotplug)
332 */
333static inline void init_page_count(struct page *page)
334{
335 atomic_set(&page->_count, 1);
336}
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338void put_page(struct page *page);
339
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800340void split_page(struct page *page, unsigned int order);
Nick Piggin8dfcc9b2006-03-22 00:08:05 -0800341
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342/*
343 * Multiple processes may "see" the same page. E.g. for untouched
344 * mappings of /dev/null, all processes see the same page full of
345 * zeroes, and text pages of executables and shared libraries have
346 * only one copy in memory, at most, normally.
347 *
348 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700349 * page_count() == 0 means the page is free. page->lru is then used for
350 * freelist management in the buddy allocator.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 * page_count() == 1 means the page is used for exactly one purpose
352 * (e.g. a private data page of one process).
353 *
354 * A page may be used for kmalloc() or anyone else who does a
355 * __get_free_page(). In this case the page_count() is at least 1, and
356 * all other fields are unused but should be 0 or NULL. The
357 * management of this page is the responsibility of the one who uses
358 * it.
359 *
360 * The other pages (we may call them "process pages") are completely
361 * managed by the Linux memory manager: I/O, buffers, swapping etc.
362 * The following discussion applies only to them.
363 *
364 * A page may belong to an inode's memory mapping. In this case,
365 * page->mapping is the pointer to the inode, and page->index is the
366 * file offset of the page, in units of PAGE_CACHE_SIZE.
367 *
368 * A page contains an opaque `private' member, which belongs to the
369 * page's address_space. Usually, this is the address of a circular
370 * list of the page's disk buffers.
371 *
372 * For pages belonging to inodes, the page_count() is the number of
373 * attaches, plus 1 if `private' contains something, plus one for
374 * the page cache itself.
375 *
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700376 * Instead of keeping dirty/clean pages in per address-space lists, we instead
377 * now tag pages as dirty/under writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 *
379 * There is also a per-mapping radix tree mapping index to the page
380 * in memory if present. The tree is rooted at mapping->root.
381 *
382 * All process pages can do I/O:
383 * - inode pages may need to be read from disk,
384 * - inode pages which have been modified and are MAP_SHARED may need
385 * to be written to disk,
386 * - private pages which have been modified may need to be swapped out
387 * to swap space and (later) to be read back into memory.
388 */
389
390/*
391 * The zone field is never updated after free_area_init_core()
392 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700394
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700395
396/*
397 * page->flags layout:
398 *
399 * There are three possibilities for how page->flags get
400 * laid out. The first is for the normal case, without
401 * sparsemem. The second is for sparsemem when there is
402 * plenty of space for node and section. The last is when
403 * we have run out of space and have to fall back to an
404 * alternate (slower) way of determining the node.
405 *
406 * No sparsemem: | NODE | ZONE | ... | FLAGS |
407 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
408 * no space for node: | SECTION | ZONE | ... | FLAGS |
409 */
410#ifdef CONFIG_SPARSEMEM
411#define SECTIONS_WIDTH SECTIONS_SHIFT
412#else
413#define SECTIONS_WIDTH 0
414#endif
415
416#define ZONES_WIDTH ZONES_SHIFT
417
418#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
419#define NODES_WIDTH NODES_SHIFT
420#else
421#define NODES_WIDTH 0
422#endif
423
424/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100425#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700426#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
427#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
428
429/*
430 * We are going to use the flags for the page to node mapping if its in
431 * there. This includes the case where there is no node, so it is implicit.
432 */
433#define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
434
435#ifndef PFN_SECTION_SHIFT
436#define PFN_SECTION_SHIFT 0
437#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700438
439/*
440 * Define the bit shifts to access each section. For non-existant
441 * sections we define the shift as 0; that plus a 0 mask ensures
442 * the compiler will optimise away reference to them.
443 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700444#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
445#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
446#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700447
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700448/* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
449#if FLAGS_HAS_NODE
Dave Hansen348f8b62005-06-23 00:07:40 -0700450#define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700451#else
452#define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
453#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700454#define ZONETABLE_PGSHIFT ZONES_PGSHIFT
455
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700456#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
457#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700458#endif
459
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700460#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
461#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
462#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700463#define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
464
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465static inline unsigned long page_zonenum(struct page *page)
466{
Dave Hansen348f8b62005-06-23 00:07:40 -0700467 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470struct zone;
471extern struct zone *zone_table[];
472
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700473static inline int page_zone_id(struct page *page)
474{
475 return (page->flags >> ZONETABLE_PGSHIFT) & ZONETABLE_MASK;
476}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477static inline struct zone *page_zone(struct page *page)
478{
Andy Whitcroftcb2b95e2006-06-23 02:03:01 -0700479 return zone_table[page_zone_id(page)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480}
481
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700482static inline unsigned long page_to_nid(struct page *page)
483{
484 if (FLAGS_HAS_NODE)
485 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
486 else
487 return page_zone(page)->zone_pgdat->node_id;
488}
489static inline unsigned long page_to_section(struct page *page)
490{
491 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
492}
493
Dave Hansen348f8b62005-06-23 00:07:40 -0700494static inline void set_page_zone(struct page *page, unsigned long zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495{
Dave Hansen348f8b62005-06-23 00:07:40 -0700496 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
497 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
498}
499static inline void set_page_node(struct page *page, unsigned long node)
500{
501 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
502 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
503}
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700504static inline void set_page_section(struct page *page, unsigned long section)
505{
506 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
507 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
508}
Dave Hansen348f8b62005-06-23 00:07:40 -0700509
510static inline void set_page_links(struct page *page, unsigned long zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700511 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700512{
513 set_page_zone(page, zone);
514 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700515 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516}
517
518#ifndef CONFIG_DISCONTIGMEM
519/* The array of struct pages - for discontigmem use pgdat->lmem_map */
520extern struct page *mem_map;
521#endif
522
Ingo Molnar652050a2006-01-14 13:21:30 -0800523static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524{
525 return __va(page_to_pfn(page) << PAGE_SHIFT);
526}
527
528#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
529#define HASHED_PAGE_VIRTUAL
530#endif
531
532#if defined(WANT_PAGE_VIRTUAL)
533#define page_address(page) ((page)->virtual)
534#define set_page_address(page, address) \
535 do { \
536 (page)->virtual = (address); \
537 } while(0)
538#define page_address_init() do { } while(0)
539#endif
540
541#if defined(HASHED_PAGE_VIRTUAL)
542void *page_address(struct page *page);
543void set_page_address(struct page *page, void *virtual);
544void page_address_init(void);
545#endif
546
547#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
548#define page_address(page) lowmem_page_address(page)
549#define set_page_address(page, address) do { } while(0)
550#define page_address_init() do { } while(0)
551#endif
552
553/*
554 * On an anonymous page mapped into a user virtual memory area,
555 * page->mapping points to its anon_vma, not to a struct address_space;
556 * with the PAGE_MAPPING_ANON bit set to distinguish it.
557 *
558 * Please note that, confusingly, "page_mapping" refers to the inode
559 * address_space which maps the page from disk; whereas "page_mapped"
560 * refers to user virtual address space into which the page is mapped.
561 */
562#define PAGE_MAPPING_ANON 1
563
564extern struct address_space swapper_space;
565static inline struct address_space *page_mapping(struct page *page)
566{
567 struct address_space *mapping = page->mapping;
568
569 if (unlikely(PageSwapCache(page)))
570 mapping = &swapper_space;
571 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
572 mapping = NULL;
573 return mapping;
574}
575
576static inline int PageAnon(struct page *page)
577{
578 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
579}
580
581/*
582 * Return the pagecache index of the passed page. Regular pagecache pages
583 * use ->index whereas swapcache pages use ->private
584 */
585static inline pgoff_t page_index(struct page *page)
586{
587 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700588 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 return page->index;
590}
591
592/*
593 * The atomic page->_mapcount, like _count, starts from -1:
594 * so that transitions both from it and to it can be tracked,
595 * using atomic_inc_and_test and atomic_add_negative(-1).
596 */
597static inline void reset_page_mapcount(struct page *page)
598{
599 atomic_set(&(page)->_mapcount, -1);
600}
601
602static inline int page_mapcount(struct page *page)
603{
604 return atomic_read(&(page)->_mapcount) + 1;
605}
606
607/*
608 * Return true if this page is mapped into pagetables.
609 */
610static inline int page_mapped(struct page *page)
611{
612 return atomic_read(&(page)->_mapcount) >= 0;
613}
614
615/*
616 * Error return values for the *_nopage functions
617 */
618#define NOPAGE_SIGBUS (NULL)
619#define NOPAGE_OOM ((struct page *) (-1))
620
621/*
622 * Different kinds of faults, as returned by handle_mm_fault().
623 * Used to decide whether a process gets delivered SIGBUS or
624 * just gets major/minor fault counters bumped up.
625 */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000626#define VM_FAULT_OOM 0x00
627#define VM_FAULT_SIGBUS 0x01
628#define VM_FAULT_MINOR 0x02
629#define VM_FAULT_MAJOR 0x03
630
631/*
632 * Special case for get_user_pages.
633 * Must be in a distinct bit from the above VM_FAULT_ flags.
634 */
635#define VM_FAULT_WRITE 0x10
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
638
639extern void show_free_areas(void);
640
641#ifdef CONFIG_SHMEM
642struct page *shmem_nopage(struct vm_area_struct *vma,
643 unsigned long address, int *type);
644int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
645struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
646 unsigned long addr);
647int shmem_lock(struct file *file, int lock, struct user_struct *user);
648#else
649#define shmem_nopage filemap_nopage
Russell King03b00eb2006-01-06 00:10:52 -0800650
651static inline int shmem_lock(struct file *file, int lock,
652 struct user_struct *user)
653{
654 return 0;
655}
656
657static inline int shmem_set_policy(struct vm_area_struct *vma,
658 struct mempolicy *new)
659{
660 return 0;
661}
662
663static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
664 unsigned long addr)
665{
666 return NULL;
667}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668#endif
669struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
David Howellsb0e15192006-01-06 00:11:42 -0800670extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672int shmem_zero_setup(struct vm_area_struct *);
673
David Howellsb0e15192006-01-06 00:11:42 -0800674#ifndef CONFIG_MMU
675extern unsigned long shmem_get_unmapped_area(struct file *file,
676 unsigned long addr,
677 unsigned long len,
678 unsigned long pgoff,
679 unsigned long flags);
680#endif
681
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682static inline int can_do_mlock(void)
683{
684 if (capable(CAP_IPC_LOCK))
685 return 1;
686 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
687 return 1;
688 return 0;
689}
690extern int user_shm_lock(size_t, struct user_struct *);
691extern void user_shm_unlock(size_t, struct user_struct *);
692
693/*
694 * Parameter block passed down to zap_pte_range in exceptional cases.
695 */
696struct zap_details {
697 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
698 struct address_space *check_mapping; /* Check page->mapping if set */
699 pgoff_t first_index; /* Lowest page->index to unmap */
700 pgoff_t last_index; /* Highest page->index to unmap */
701 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 unsigned long truncate_count; /* Compare vm_truncate_count */
703};
704
Linus Torvalds6aab3412005-11-28 14:34:23 -0800705struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700706unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700708unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 struct vm_area_struct *start_vma, unsigned long start_addr,
710 unsigned long end_addr, unsigned long *nr_accounted,
711 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700712void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
713 unsigned long end, unsigned long floor, unsigned long ceiling);
714void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700715 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
717 struct vm_area_struct *vma);
718int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
719 unsigned long size, pgprot_t prot);
720void unmap_mapping_range(struct address_space *mapping,
721 loff_t const holebegin, loff_t const holelen, int even_cows);
722
723static inline void unmap_shared_mapping_range(struct address_space *mapping,
724 loff_t const holebegin, loff_t const holelen)
725{
726 unmap_mapping_range(mapping, holebegin, holelen, 0);
727}
728
729extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800730extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
732extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000733
David Howells7ee1dd32006-01-06 00:11:44 -0800734#ifdef CONFIG_MMU
735extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
736 unsigned long address, int write_access);
737
738static inline int handle_mm_fault(struct mm_struct *mm,
739 struct vm_area_struct *vma, unsigned long address,
740 int write_access)
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000741{
David Howells7ee1dd32006-01-06 00:11:44 -0800742 return __handle_mm_fault(mm, vma, address, write_access) &
743 (~VM_FAULT_WRITE);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000744}
David Howells7ee1dd32006-01-06 00:11:44 -0800745#else
746static inline int handle_mm_fault(struct mm_struct *mm,
747 struct vm_area_struct *vma, unsigned long address,
748 int write_access)
749{
750 /* should never happen if there's no MMU */
751 BUG();
752 return VM_FAULT_SIGBUS;
753}
754#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000755
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756extern int make_pages_present(unsigned long addr, unsigned long end);
757extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
758void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
759
760int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
761 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
Nick Pigginb5810032005-10-29 18:16:12 -0700762void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
764int __set_page_dirty_buffers(struct page *page);
765int __set_page_dirty_nobuffers(struct page *page);
766int redirty_page_for_writepage(struct writeback_control *wbc,
767 struct page *page);
768int FASTCALL(set_page_dirty(struct page *page));
769int set_page_dirty_lock(struct page *page);
770int clear_page_dirty_for_io(struct page *page);
771
772extern unsigned long do_mremap(unsigned long addr,
773 unsigned long old_len, unsigned long new_len,
774 unsigned long flags, unsigned long new_addr);
775
776/*
777 * Prototype to add a shrinker callback for ageable caches.
778 *
779 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
780 * scan `nr_to_scan' objects, attempting to free them.
781 *
Domen Puncer845d3432005-05-05 16:16:14 -0700782 * The callback must return the number of objects which remain in the cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 *
Domen Puncer845d3432005-05-05 16:16:14 -0700784 * The callback will be passed nr_to_scan == 0 when the VM is querying the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 * cache size, so a fastpath for that case is appropriate.
786 */
Al Viro6daa0e22005-10-21 03:18:50 -0400787typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789/*
790 * Add an aging callback. The int is the number of 'seeks' it takes
791 * to recreate one of the objects that these functions age.
792 */
793
794#define DEFAULT_SEEKS 2
795struct shrinker;
796extern struct shrinker *set_shrinker(int, shrinker_t);
797extern void remove_shrinker(struct shrinker *shrinker);
798
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800799extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
800
Hugh Dickins1bb36302005-10-29 18:16:22 -0700801int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
802int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
803int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
804int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
805
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806/*
807 * The following ifdef needed to get the 4level-fixup.h header to work.
808 * Remove it when 4level-fixup.h has been removed.
809 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700810#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
812{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700813 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
814 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815}
816
817static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
818{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700819 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
820 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700822#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
823
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700824#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
825/*
826 * We tuck a spinlock to guard each pagetable page into its struct page,
827 * at page->private, with BUILD_BUG_ON to make sure that this will not
828 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
829 * When freeing, reset page->mapping so free_pages_check won't complain.
830 */
Andrew Morton349aef02006-01-08 01:04:36 -0800831#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700832#define pte_lock_init(_page) do { \
833 spin_lock_init(__pte_lockptr(_page)); \
834} while (0)
835#define pte_lock_deinit(page) ((page)->mapping = NULL)
836#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
837#else
838/*
839 * We use mm->page_table_lock to guard all pagetable pages of the mm.
840 */
841#define pte_lock_init(page) do {} while (0)
842#define pte_lock_deinit(page) do {} while (0)
843#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
844#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
845
Hugh Dickinsc74df322005-10-29 18:16:23 -0700846#define pte_offset_map_lock(mm, pmd, address, ptlp) \
847({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700848 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700849 pte_t *__pte = pte_offset_map(pmd, address); \
850 *(ptlp) = __ptl; \
851 spin_lock(__ptl); \
852 __pte; \
853})
854
855#define pte_unmap_unlock(pte, ptl) do { \
856 spin_unlock(ptl); \
857 pte_unmap(pte); \
858} while (0)
859
Hugh Dickins1bb36302005-10-29 18:16:22 -0700860#define pte_alloc_map(mm, pmd, address) \
861 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
862 NULL: pte_offset_map(pmd, address))
863
Hugh Dickinsc74df322005-10-29 18:16:23 -0700864#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
865 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
866 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
867
Hugh Dickins1bb36302005-10-29 18:16:22 -0700868#define pte_alloc_kernel(pmd, address) \
869 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
870 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
872extern void free_area_init(unsigned long * zones_size);
873extern void free_area_init_node(int nid, pg_data_t *pgdat,
874 unsigned long * zones_size, unsigned long zone_start_pfn,
875 unsigned long *zholes_size);
876extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
Dave Hansen3947be12005-10-29 18:16:54 -0700877extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878extern void mem_init(void);
879extern void show_mem(void);
880extern void si_meminfo(struct sysinfo * val);
881extern void si_meminfo_node(struct sysinfo *val, int nid);
882
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700883#ifdef CONFIG_NUMA
884extern void setup_per_cpu_pageset(void);
885#else
886static inline void setup_per_cpu_pageset(void) {}
887#endif
888
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889/* prio_tree.c */
890void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
891void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
892void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
893struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
894 struct prio_tree_iter *iter);
895
896#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
897 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
898 (vma = vma_prio_tree_next(vma, iter)); )
899
900static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
901 struct list_head *list)
902{
903 vma->shared.vm_set.parent = NULL;
904 list_add_tail(&vma->shared.vm_set.list, list);
905}
906
907/* mmap.c */
908extern int __vm_enough_memory(long pages, int cap_sys_admin);
909extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
910 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
911extern struct vm_area_struct *vma_merge(struct mm_struct *,
912 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
913 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
914 struct mempolicy *);
915extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
916extern int split_vma(struct mm_struct *,
917 struct vm_area_struct *, unsigned long addr, int new_below);
918extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
919extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
920 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -0700921extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
923 unsigned long addr, unsigned long len, pgoff_t pgoff);
924extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -0700925extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926
927extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
928
929extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
930 unsigned long len, unsigned long prot,
931 unsigned long flag, unsigned long pgoff);
932
933static inline unsigned long do_mmap(struct file *file, unsigned long addr,
934 unsigned long len, unsigned long prot,
935 unsigned long flag, unsigned long offset)
936{
937 unsigned long ret = -EINVAL;
938 if ((offset + PAGE_ALIGN(len)) < offset)
939 goto out;
940 if (!(offset & ~PAGE_MASK))
941 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
942out:
943 return ret;
944}
945
946extern int do_munmap(struct mm_struct *, unsigned long, size_t);
947
948extern unsigned long do_brk(unsigned long, unsigned long);
949
950/* filemap.c */
951extern unsigned long page_unuse(struct page *);
952extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -0800953extern void truncate_inode_pages_range(struct address_space *,
954 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
956/* generic vm_area_ops exported for stackable file systems */
957extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
958extern int filemap_populate(struct vm_area_struct *, unsigned long,
959 unsigned long, pgprot_t, unsigned long, int);
960
961/* mm/page-writeback.c */
962int write_one_page(struct page *page, int wait);
963
964/* readahead.c */
965#define VM_MAX_READAHEAD 128 /* kbytes */
966#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
967#define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
968 * turning readahead off */
969
970int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800971 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800973 pgoff_t offset, unsigned long nr_to_read);
974unsigned long page_cache_readahead(struct address_space *mapping,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 struct file_ra_state *ra,
976 struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800977 pgoff_t offset,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 unsigned long size);
979void handle_ra_miss(struct address_space *mapping,
980 struct file_ra_state *ra, pgoff_t offset);
981unsigned long max_sane_readahead(unsigned long nr);
982
983/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -0700984extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -0500985#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -0700986extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -0500987#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
989/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
990extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
991extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
992 struct vm_area_struct **pprev);
993
994/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
995 NULL if none. Assume start_addr < end_addr. */
996static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
997{
998 struct vm_area_struct * vma = find_vma(mm,start_addr);
999
1000 if (vma && end_addr <= vma->vm_start)
1001 vma = NULL;
1002 return vma;
1003}
1004
1005static inline unsigned long vma_pages(struct vm_area_struct *vma)
1006{
1007 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1008}
1009
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001010struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1011struct page *vmalloc_to_page(void *addr);
1012unsigned long vmalloc_to_pfn(void *addr);
1013int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1014 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001015int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
Linus Torvalds6aab3412005-11-28 14:34:23 -08001017struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001018 unsigned int foll_flags);
1019#define FOLL_WRITE 0x01 /* check pte is writable */
1020#define FOLL_TOUCH 0x02 /* mark page accessed */
1021#define FOLL_GET 0x04 /* do get_page on page */
1022#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023
1024#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001025void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001027static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 unsigned long flags, struct file *file, long pages)
1029{
1030}
1031#endif /* CONFIG_PROC_FS */
1032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033#ifndef CONFIG_DEBUG_PAGEALLOC
1034static inline void
1035kernel_map_pages(struct page *page, int numpages, int enable)
1036{
Ingo Molnarde5097c2006-01-09 15:59:21 -08001037 if (!PageHighMem(page) && !enable)
1038 mutex_debug_check_no_locks_freed(page_address(page),
David Woodhousea4fc7ab2006-01-11 14:41:26 +00001039 numpages * PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040}
1041#endif
1042
1043extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1044#ifdef __HAVE_ARCH_GATE_AREA
1045int in_gate_area_no_task(unsigned long addr);
1046int in_gate_area(struct task_struct *task, unsigned long addr);
1047#else
1048int in_gate_area_no_task(unsigned long addr);
1049#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1050#endif /* __HAVE_ARCH_GATE_AREA */
1051
Andrea Arcangeli79befd02005-04-16 15:24:05 -07001052/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1053#define OOM_DISABLE -17
1054
Andrew Morton9d0243b2006-01-08 01:00:39 -08001055int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1056 void __user *, size_t *, loff_t *);
Andrew Morton69e05942006-03-22 00:08:19 -08001057unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
Andrew Morton9d0243b2006-01-08 01:00:39 -08001058 unsigned long lru_pages);
1059void drop_pagecache(void);
1060void drop_slab(void);
1061
Luke Yang7a9166e2006-02-20 18:28:07 -08001062#ifndef CONFIG_MMU
1063#define randomize_va_space 0
1064#else
Andi Kleena62eaf12006-02-16 23:41:58 +01001065extern int randomize_va_space;
Luke Yang7a9166e2006-02-20 18:28:07 -08001066#endif
Andi Kleena62eaf12006-02-16 23:41:58 +01001067
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068#endif /* __KERNEL__ */
1069#endif /* _LINUX_MM_H */