blob: 85854b867463484cf1cc1c7fe76ae5c4f5ca4201 [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
10#include <linux/config.h>
11#include <linux/gfp.h>
12#include <linux/list.h>
13#include <linux/mmzone.h>
14#include <linux/rbtree.h>
15#include <linux/prio_tree.h>
16#include <linux/fs.h>
Ingo Molnarde5097c2006-01-09 15:59:21 -080017#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
19struct mempolicy;
20struct anon_vma;
21
22#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23extern unsigned long max_mapnr;
24#endif
25
26extern unsigned long num_physpages;
27extern void * high_memory;
28extern unsigned long vmalloc_earlyreserve;
29extern int page_cluster;
30
31#ifdef CONFIG_SYSCTL
32extern int sysctl_legacy_va_layout;
33#else
34#define sysctl_legacy_va_layout 0
35#endif
36
37#include <asm/page.h>
38#include <asm/pgtable.h>
39#include <asm/processor.h>
40#include <asm/atomic.h>
41
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
43
44/*
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
50 * mmap() functions).
51 */
52
53/*
54 * This struct defines a memory VMM memory area. There is one of these
55 * per VM-area/task. A VM area is any part of the process virtual memory
56 * space that has a special rule for the page-fault handlers (ie a shared
57 * library, the executable area etc).
58 */
59struct vm_area_struct {
60 struct mm_struct * vm_mm; /* The address space we belong to. */
61 unsigned long vm_start; /* Our start address within vm_mm. */
62 unsigned long vm_end; /* The first byte after our end address
63 within vm_mm. */
64
65 /* linked list of VM areas per task, sorted by address */
66 struct vm_area_struct *vm_next;
67
68 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
69 unsigned long vm_flags; /* Flags, listed below. */
70
71 struct rb_node vm_rb;
72
73 /*
74 * For areas with an address space and backing store,
75 * linkage into the address_space->i_mmap prio tree, or
76 * linkage to the list of like vmas hanging off its node, or
77 * linkage of vma in the address_space->i_mmap_nonlinear list.
78 */
79 union {
80 struct {
81 struct list_head list;
82 void *parent; /* aligns with prio_tree_node parent */
83 struct vm_area_struct *head;
84 } vm_set;
85
86 struct raw_prio_tree_node prio_tree_node;
87 } shared;
88
89 /*
90 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
91 * list, after a COW of one of the file pages. A MAP_SHARED vma
92 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
93 * or brk vma (with NULL file) can only be in an anon_vma list.
94 */
95 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
96 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
97
98 /* Function pointers to deal with this struct. */
99 struct vm_operations_struct * vm_ops;
100
101 /* Information about our backing store: */
102 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
103 units, *not* PAGE_CACHE_SIZE */
104 struct file * vm_file; /* File we map to (can be NULL). */
105 void * vm_private_data; /* was vm_pte (shared mem) */
106 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
107
108#ifndef CONFIG_MMU
109 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
110#endif
111#ifdef CONFIG_NUMA
112 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
113#endif
114};
115
116/*
117 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
118 * disabled, then there's a single shared list of VMAs maintained by the
119 * system, and mm's subscribe to these individually
120 */
121struct vm_list_struct {
122 struct vm_list_struct *next;
123 struct vm_area_struct *vma;
124};
125
126#ifndef CONFIG_MMU
127extern struct rb_root nommu_vma_tree;
128extern struct rw_semaphore nommu_vma_sem;
129
130extern unsigned int kobjsize(const void *objp);
131#endif
132
133/*
134 * vm_flags..
135 */
136#define VM_READ 0x00000001 /* currently active flags */
137#define VM_WRITE 0x00000002
138#define VM_EXEC 0x00000004
139#define VM_SHARED 0x00000008
140
Paolo 'Blaisorblade' Giarrusso7e2cff42005-09-21 09:55:39 -0700141/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
143#define VM_MAYWRITE 0x00000020
144#define VM_MAYEXEC 0x00000040
145#define VM_MAYSHARE 0x00000080
146
147#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
148#define VM_GROWSUP 0x00000200
Hugh Dickins0b14c172005-11-21 21:32:15 -0800149#define VM_SHM 0x00000000 /* Means nothing: delete it later */
Linus Torvalds6aab3412005-11-28 14:34:23 -0800150#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
152
153#define VM_EXECUTABLE 0x00001000
154#define VM_LOCKED 0x00002000
155#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
156
157 /* Used by sys_madvise() */
158#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
160
161#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
Hugh Dickins0b14c172005-11-21 21:32:15 -0800163#define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
Linus Torvalds4d7672b2005-12-16 10:21:23 -0800168#define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
170#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
171#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
172#endif
173
174#ifdef CONFIG_STACK_GROWSUP
175#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176#else
177#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
178#endif
179
180#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
181#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
182#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
183#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
184#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
185
186/*
187 * mapping from the currently active vm_flags protection bits (the
188 * low four bits) to a page protection mask..
189 */
190extern pgprot_t protection_map[16];
191
192
193/*
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
197 */
198struct vm_operations_struct {
199 void (*open)(struct vm_area_struct * area);
200 void (*close)(struct vm_area_struct * area);
201 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
202 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
203#ifdef CONFIG_NUMA
204 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
205 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
206 unsigned long addr);
207#endif
208};
209
210struct mmu_gather;
211struct inode;
212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213/*
214 * Each physical page in the system has a struct page associated with
215 * it to keep track of whatever it is we are using the page for at the
216 * moment. Note that we have no way to track which tasks are using
217 * a page.
218 */
219struct page {
Andi Kleen07808b72005-11-05 17:25:53 +0100220 unsigned long flags; /* Atomic flags, some possibly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 * updated asynchronously */
222 atomic_t _count; /* Usage count, see below. */
223 atomic_t _mapcount; /* Count of ptes mapped in mms,
224 * to show when page is mapped
225 * & limit reverse map searches.
226 */
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700227 union {
Andrew Morton349aef02006-01-08 01:04:36 -0800228 struct {
229 unsigned long private; /* Mapping-private opaque data:
230 * usually used for buffer_heads
231 * if PagePrivate set; used for
232 * swp_entry_t if PageSwapCache.
233 * When page is free, this
234 * indicates order in the buddy
235 * system.
236 */
237 struct address_space *mapping; /* If low bit clear, points to
238 * inode address_space, or NULL.
239 * If page mapped as anonymous
240 * memory, low bit is set, and
241 * it points to anon_vma object:
242 * see PAGE_MAPPING_ANON below.
243 */
244 };
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700245#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
Andrew Morton349aef02006-01-08 01:04:36 -0800246 spinlock_t ptl;
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700247#endif
Andrew Morton349aef02006-01-08 01:04:36 -0800248 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 pgoff_t index; /* Our offset within mapping. */
250 struct list_head lru; /* Pageout list, eg. active_list
251 * protected by zone->lru_lock !
252 */
253 /*
254 * On machines where all RAM is mapped into kernel address space,
255 * we can simply calculate the virtual address. On machines with
256 * highmem some memory is mapped into kernel virtual memory
257 * dynamically, so we need a place to store that address.
258 * Note that this field could be 16 bits on x86 ... ;)
259 *
260 * Architectures with slow multiplication can define
261 * WANT_PAGE_VIRTUAL in asm/page.h
262 */
263#if defined(WANT_PAGE_VIRTUAL)
264 void *virtual; /* Kernel virtual address (NULL if
265 not kmapped, ie. highmem) */
266#endif /* WANT_PAGE_VIRTUAL */
267};
268
Andrew Morton349aef02006-01-08 01:04:36 -0800269#define page_private(page) ((page)->private)
270#define set_page_private(page, v) ((page)->private = (v))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272/*
273 * FIXME: take this include out, include page-flags.h in
274 * files which need it (119 of them)
275 */
276#include <linux/page-flags.h>
277
278/*
279 * Methods to modify the page usage count.
280 *
281 * What counts for a page usage:
282 * - cache mapping (page->mapping)
283 * - private data (page->private)
284 * - page mapped in a task's page tables, each mapping
285 * is counted separately
286 *
287 * Also, many kernel routines increase the page count before a critical
288 * routine so they can be sure the page doesn't go away from under them.
289 *
290 * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we
291 * can use atomic_add_negative(-1, page->_count) to detect when the page
292 * becomes free and so that we can also use atomic_inc_and_test to atomically
293 * detect when we just tried to grab a ref on a page which some other CPU has
294 * already deemed to be freeable.
295 *
296 * NO code should make assumptions about this internal detail! Use the provided
297 * macros which retain the old rules: page_count(page) == 0 is a free page.
298 */
299
300/*
301 * Drop a ref, return true if the logical refcount fell to zero (the page has
302 * no users)
303 */
304#define put_page_testzero(p) \
305 ({ \
306 BUG_ON(page_count(p) == 0); \
307 atomic_add_negative(-1, &(p)->_count); \
308 })
309
310/*
311 * Grab a ref, return true if the page previously had a logical refcount of
312 * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page
313 */
314#define get_page_testone(p) atomic_inc_and_test(&(p)->_count)
315
Avishay Traeger152194a2006-01-08 01:00:58 -0800316#define set_page_count(p,v) atomic_set(&(p)->_count, (v) - 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317#define __put_page(p) atomic_dec(&(p)->_count)
318
319extern void FASTCALL(__page_cache_release(struct page *));
320
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700321static inline int page_count(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322{
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700323 if (PageCompound(page))
324 page = (struct page *)page_private(page);
325 return atomic_read(&page->_count) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326}
327
328static inline void get_page(struct page *page)
329{
330 if (unlikely(PageCompound(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700331 page = (struct page *)page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 atomic_inc(&page->_count);
333}
334
335void put_page(struct page *page);
336
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337/*
338 * Multiple processes may "see" the same page. E.g. for untouched
339 * mappings of /dev/null, all processes see the same page full of
340 * zeroes, and text pages of executables and shared libraries have
341 * only one copy in memory, at most, normally.
342 *
343 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700344 * page_count() == 0 means the page is free. page->lru is then used for
345 * freelist management in the buddy allocator.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * page_count() == 1 means the page is used for exactly one purpose
347 * (e.g. a private data page of one process).
348 *
349 * A page may be used for kmalloc() or anyone else who does a
350 * __get_free_page(). In this case the page_count() is at least 1, and
351 * all other fields are unused but should be 0 or NULL. The
352 * management of this page is the responsibility of the one who uses
353 * it.
354 *
355 * The other pages (we may call them "process pages") are completely
356 * managed by the Linux memory manager: I/O, buffers, swapping etc.
357 * The following discussion applies only to them.
358 *
359 * A page may belong to an inode's memory mapping. In this case,
360 * page->mapping is the pointer to the inode, and page->index is the
361 * file offset of the page, in units of PAGE_CACHE_SIZE.
362 *
363 * A page contains an opaque `private' member, which belongs to the
364 * page's address_space. Usually, this is the address of a circular
365 * list of the page's disk buffers.
366 *
367 * For pages belonging to inodes, the page_count() is the number of
368 * attaches, plus 1 if `private' contains something, plus one for
369 * the page cache itself.
370 *
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700371 * Instead of keeping dirty/clean pages in per address-space lists, we instead
372 * now tag pages as dirty/under writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 *
374 * There is also a per-mapping radix tree mapping index to the page
375 * in memory if present. The tree is rooted at mapping->root.
376 *
377 * All process pages can do I/O:
378 * - inode pages may need to be read from disk,
379 * - inode pages which have been modified and are MAP_SHARED may need
380 * to be written to disk,
381 * - private pages which have been modified may need to be swapped out
382 * to swap space and (later) to be read back into memory.
383 */
384
385/*
386 * The zone field is never updated after free_area_init_core()
387 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700389
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700390
391/*
392 * page->flags layout:
393 *
394 * There are three possibilities for how page->flags get
395 * laid out. The first is for the normal case, without
396 * sparsemem. The second is for sparsemem when there is
397 * plenty of space for node and section. The last is when
398 * we have run out of space and have to fall back to an
399 * alternate (slower) way of determining the node.
400 *
401 * No sparsemem: | NODE | ZONE | ... | FLAGS |
402 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
403 * no space for node: | SECTION | ZONE | ... | FLAGS |
404 */
405#ifdef CONFIG_SPARSEMEM
406#define SECTIONS_WIDTH SECTIONS_SHIFT
407#else
408#define SECTIONS_WIDTH 0
409#endif
410
411#define ZONES_WIDTH ZONES_SHIFT
412
413#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
414#define NODES_WIDTH NODES_SHIFT
415#else
416#define NODES_WIDTH 0
417#endif
418
419/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
Andi Kleen07808b72005-11-05 17:25:53 +0100420#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700421#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
422#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
423
424/*
425 * We are going to use the flags for the page to node mapping if its in
426 * there. This includes the case where there is no node, so it is implicit.
427 */
428#define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
429
430#ifndef PFN_SECTION_SHIFT
431#define PFN_SECTION_SHIFT 0
432#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700433
434/*
435 * Define the bit shifts to access each section. For non-existant
436 * sections we define the shift as 0; that plus a 0 mask ensures
437 * the compiler will optimise away reference to them.
438 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700439#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
440#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
441#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700442
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700443/* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
444#if FLAGS_HAS_NODE
Dave Hansen348f8b62005-06-23 00:07:40 -0700445#define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700446#else
447#define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
448#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700449#define ZONETABLE_PGSHIFT ZONES_PGSHIFT
450
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700451#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
452#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700453#endif
454
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700455#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
456#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
457#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700458#define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460static inline unsigned long page_zonenum(struct page *page)
461{
Dave Hansen348f8b62005-06-23 00:07:40 -0700462 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465struct zone;
466extern struct zone *zone_table[];
467
468static inline struct zone *page_zone(struct page *page)
469{
Dave Hansen348f8b62005-06-23 00:07:40 -0700470 return zone_table[(page->flags >> ZONETABLE_PGSHIFT) &
471 ZONETABLE_MASK];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472}
473
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700474static inline unsigned long page_to_nid(struct page *page)
475{
476 if (FLAGS_HAS_NODE)
477 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
478 else
479 return page_zone(page)->zone_pgdat->node_id;
480}
481static inline unsigned long page_to_section(struct page *page)
482{
483 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
484}
485
Dave Hansen348f8b62005-06-23 00:07:40 -0700486static inline void set_page_zone(struct page *page, unsigned long zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Dave Hansen348f8b62005-06-23 00:07:40 -0700488 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
489 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
490}
491static inline void set_page_node(struct page *page, unsigned long node)
492{
493 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
494 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
495}
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700496static inline void set_page_section(struct page *page, unsigned long section)
497{
498 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
499 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
500}
Dave Hansen348f8b62005-06-23 00:07:40 -0700501
502static inline void set_page_links(struct page *page, unsigned long zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700503 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700504{
505 set_page_zone(page, zone);
506 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700507 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508}
509
510#ifndef CONFIG_DISCONTIGMEM
511/* The array of struct pages - for discontigmem use pgdat->lmem_map */
512extern struct page *mem_map;
513#endif
514
Ingo Molnar652050a2006-01-14 13:21:30 -0800515static __always_inline void *lowmem_page_address(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
517 return __va(page_to_pfn(page) << PAGE_SHIFT);
518}
519
520#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
521#define HASHED_PAGE_VIRTUAL
522#endif
523
524#if defined(WANT_PAGE_VIRTUAL)
525#define page_address(page) ((page)->virtual)
526#define set_page_address(page, address) \
527 do { \
528 (page)->virtual = (address); \
529 } while(0)
530#define page_address_init() do { } while(0)
531#endif
532
533#if defined(HASHED_PAGE_VIRTUAL)
534void *page_address(struct page *page);
535void set_page_address(struct page *page, void *virtual);
536void page_address_init(void);
537#endif
538
539#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
540#define page_address(page) lowmem_page_address(page)
541#define set_page_address(page, address) do { } while(0)
542#define page_address_init() do { } while(0)
543#endif
544
545/*
546 * On an anonymous page mapped into a user virtual memory area,
547 * page->mapping points to its anon_vma, not to a struct address_space;
548 * with the PAGE_MAPPING_ANON bit set to distinguish it.
549 *
550 * Please note that, confusingly, "page_mapping" refers to the inode
551 * address_space which maps the page from disk; whereas "page_mapped"
552 * refers to user virtual address space into which the page is mapped.
553 */
554#define PAGE_MAPPING_ANON 1
555
556extern struct address_space swapper_space;
557static inline struct address_space *page_mapping(struct page *page)
558{
559 struct address_space *mapping = page->mapping;
560
561 if (unlikely(PageSwapCache(page)))
562 mapping = &swapper_space;
563 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
564 mapping = NULL;
565 return mapping;
566}
567
568static inline int PageAnon(struct page *page)
569{
570 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
571}
572
573/*
574 * Return the pagecache index of the passed page. Regular pagecache pages
575 * use ->index whereas swapcache pages use ->private
576 */
577static inline pgoff_t page_index(struct page *page)
578{
579 if (unlikely(PageSwapCache(page)))
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700580 return page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 return page->index;
582}
583
584/*
585 * The atomic page->_mapcount, like _count, starts from -1:
586 * so that transitions both from it and to it can be tracked,
587 * using atomic_inc_and_test and atomic_add_negative(-1).
588 */
589static inline void reset_page_mapcount(struct page *page)
590{
591 atomic_set(&(page)->_mapcount, -1);
592}
593
594static inline int page_mapcount(struct page *page)
595{
596 return atomic_read(&(page)->_mapcount) + 1;
597}
598
599/*
600 * Return true if this page is mapped into pagetables.
601 */
602static inline int page_mapped(struct page *page)
603{
604 return atomic_read(&(page)->_mapcount) >= 0;
605}
606
607/*
608 * Error return values for the *_nopage functions
609 */
610#define NOPAGE_SIGBUS (NULL)
611#define NOPAGE_OOM ((struct page *) (-1))
612
613/*
614 * Different kinds of faults, as returned by handle_mm_fault().
615 * Used to decide whether a process gets delivered SIGBUS or
616 * just gets major/minor fault counters bumped up.
617 */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000618#define VM_FAULT_OOM 0x00
619#define VM_FAULT_SIGBUS 0x01
620#define VM_FAULT_MINOR 0x02
621#define VM_FAULT_MAJOR 0x03
622
623/*
624 * Special case for get_user_pages.
625 * Must be in a distinct bit from the above VM_FAULT_ flags.
626 */
627#define VM_FAULT_WRITE 0x10
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
630
631extern void show_free_areas(void);
632
633#ifdef CONFIG_SHMEM
634struct page *shmem_nopage(struct vm_area_struct *vma,
635 unsigned long address, int *type);
636int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
637struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
638 unsigned long addr);
639int shmem_lock(struct file *file, int lock, struct user_struct *user);
640#else
641#define shmem_nopage filemap_nopage
Russell King03b00eb2006-01-06 00:10:52 -0800642
643static inline int shmem_lock(struct file *file, int lock,
644 struct user_struct *user)
645{
646 return 0;
647}
648
649static inline int shmem_set_policy(struct vm_area_struct *vma,
650 struct mempolicy *new)
651{
652 return 0;
653}
654
655static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
656 unsigned long addr)
657{
658 return NULL;
659}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660#endif
661struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
David Howellsb0e15192006-01-06 00:11:42 -0800662extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
664int shmem_zero_setup(struct vm_area_struct *);
665
David Howellsb0e15192006-01-06 00:11:42 -0800666#ifndef CONFIG_MMU
667extern unsigned long shmem_get_unmapped_area(struct file *file,
668 unsigned long addr,
669 unsigned long len,
670 unsigned long pgoff,
671 unsigned long flags);
672#endif
673
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674static inline int can_do_mlock(void)
675{
676 if (capable(CAP_IPC_LOCK))
677 return 1;
678 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
679 return 1;
680 return 0;
681}
682extern int user_shm_lock(size_t, struct user_struct *);
683extern void user_shm_unlock(size_t, struct user_struct *);
684
685/*
686 * Parameter block passed down to zap_pte_range in exceptional cases.
687 */
688struct zap_details {
689 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
690 struct address_space *check_mapping; /* Check page->mapping if set */
691 pgoff_t first_index; /* Lowest page->index to unmap */
692 pgoff_t last_index; /* Highest page->index to unmap */
693 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 unsigned long truncate_count; /* Compare vm_truncate_count */
695};
696
Linus Torvalds6aab3412005-11-28 14:34:23 -0800697struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700698unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 unsigned long size, struct zap_details *);
Hugh Dickins508034a2005-10-29 18:16:30 -0700700unsigned long unmap_vmas(struct mmu_gather **tlb,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 struct vm_area_struct *start_vma, unsigned long start_addr,
702 unsigned long end_addr, unsigned long *nr_accounted,
703 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700704void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
705 unsigned long end, unsigned long floor, unsigned long ceiling);
706void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700707 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
709 struct vm_area_struct *vma);
710int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
711 unsigned long size, pgprot_t prot);
712void unmap_mapping_range(struct address_space *mapping,
713 loff_t const holebegin, loff_t const holelen, int even_cows);
714
715static inline void unmap_shared_mapping_range(struct address_space *mapping,
716 loff_t const holebegin, loff_t const holelen)
717{
718 unmap_mapping_range(mapping, holebegin, holelen, 0);
719}
720
721extern int vmtruncate(struct inode * inode, loff_t offset);
Badari Pulavartyf6b3ec22006-01-06 00:10:38 -0800722extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
724extern 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 +1000725
David Howells7ee1dd32006-01-06 00:11:44 -0800726#ifdef CONFIG_MMU
727extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
728 unsigned long address, int write_access);
729
730static inline int handle_mm_fault(struct mm_struct *mm,
731 struct vm_area_struct *vma, unsigned long address,
732 int write_access)
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000733{
David Howells7ee1dd32006-01-06 00:11:44 -0800734 return __handle_mm_fault(mm, vma, address, write_access) &
735 (~VM_FAULT_WRITE);
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000736}
David Howells7ee1dd32006-01-06 00:11:44 -0800737#else
738static inline int handle_mm_fault(struct mm_struct *mm,
739 struct vm_area_struct *vma, unsigned long address,
740 int write_access)
741{
742 /* should never happen if there's no MMU */
743 BUG();
744 return VM_FAULT_SIGBUS;
745}
746#endif
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748extern int make_pages_present(unsigned long addr, unsigned long end);
749extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
750void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
751
752int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
753 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
Nick Pigginb5810032005-10-29 18:16:12 -0700754void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755
756int __set_page_dirty_buffers(struct page *page);
757int __set_page_dirty_nobuffers(struct page *page);
758int redirty_page_for_writepage(struct writeback_control *wbc,
759 struct page *page);
760int FASTCALL(set_page_dirty(struct page *page));
761int set_page_dirty_lock(struct page *page);
762int clear_page_dirty_for_io(struct page *page);
763
764extern unsigned long do_mremap(unsigned long addr,
765 unsigned long old_len, unsigned long new_len,
766 unsigned long flags, unsigned long new_addr);
767
768/*
769 * Prototype to add a shrinker callback for ageable caches.
770 *
771 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
772 * scan `nr_to_scan' objects, attempting to free them.
773 *
Domen Puncer845d3432005-05-05 16:16:14 -0700774 * The callback must return the number of objects which remain in the cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 *
Domen Puncer845d3432005-05-05 16:16:14 -0700776 * The callback will be passed nr_to_scan == 0 when the VM is querying the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 * cache size, so a fastpath for that case is appropriate.
778 */
Al Viro6daa0e22005-10-21 03:18:50 -0400779typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
781/*
782 * Add an aging callback. The int is the number of 'seeks' it takes
783 * to recreate one of the objects that these functions age.
784 */
785
786#define DEFAULT_SEEKS 2
787struct shrinker;
788extern struct shrinker *set_shrinker(int, shrinker_t);
789extern void remove_shrinker(struct shrinker *shrinker);
790
Linus Torvaldsc9cfcdd2005-11-29 14:03:14 -0800791extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
792
Hugh Dickins1bb36302005-10-29 18:16:22 -0700793int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
794int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
795int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
796int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
797
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798/*
799 * The following ifdef needed to get the 4level-fixup.h header to work.
800 * Remove it when 4level-fixup.h has been removed.
801 */
Hugh Dickins1bb36302005-10-29 18:16:22 -0700802#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
804{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700805 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
806 NULL: pud_offset(pgd, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807}
808
809static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
810{
Hugh Dickins1bb36302005-10-29 18:16:22 -0700811 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
812 NULL: pmd_offset(pud, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813}
Hugh Dickins1bb36302005-10-29 18:16:22 -0700814#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
815
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700816#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
817/*
818 * We tuck a spinlock to guard each pagetable page into its struct page,
819 * at page->private, with BUILD_BUG_ON to make sure that this will not
820 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
821 * When freeing, reset page->mapping so free_pages_check won't complain.
822 */
Andrew Morton349aef02006-01-08 01:04:36 -0800823#define __pte_lockptr(page) &((page)->ptl)
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700824#define pte_lock_init(_page) do { \
825 spin_lock_init(__pte_lockptr(_page)); \
826} while (0)
827#define pte_lock_deinit(page) ((page)->mapping = NULL)
828#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
829#else
830/*
831 * We use mm->page_table_lock to guard all pagetable pages of the mm.
832 */
833#define pte_lock_init(page) do {} while (0)
834#define pte_lock_deinit(page) do {} while (0)
835#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
836#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
837
Hugh Dickinsc74df322005-10-29 18:16:23 -0700838#define pte_offset_map_lock(mm, pmd, address, ptlp) \
839({ \
Hugh Dickins4c21e2f2005-10-29 18:16:40 -0700840 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
Hugh Dickinsc74df322005-10-29 18:16:23 -0700841 pte_t *__pte = pte_offset_map(pmd, address); \
842 *(ptlp) = __ptl; \
843 spin_lock(__ptl); \
844 __pte; \
845})
846
847#define pte_unmap_unlock(pte, ptl) do { \
848 spin_unlock(ptl); \
849 pte_unmap(pte); \
850} while (0)
851
Hugh Dickins1bb36302005-10-29 18:16:22 -0700852#define pte_alloc_map(mm, pmd, address) \
853 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
854 NULL: pte_offset_map(pmd, address))
855
Hugh Dickinsc74df322005-10-29 18:16:23 -0700856#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
857 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
858 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
859
Hugh Dickins1bb36302005-10-29 18:16:22 -0700860#define pte_alloc_kernel(pmd, address) \
861 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
862 NULL: pte_offset_kernel(pmd, address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864extern void free_area_init(unsigned long * zones_size);
865extern void free_area_init_node(int nid, pg_data_t *pgdat,
866 unsigned long * zones_size, unsigned long zone_start_pfn,
867 unsigned long *zholes_size);
868extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
Dave Hansen3947be12005-10-29 18:16:54 -0700869extern void setup_per_zone_pages_min(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870extern void mem_init(void);
871extern void show_mem(void);
872extern void si_meminfo(struct sysinfo * val);
873extern void si_meminfo_node(struct sysinfo *val, int nid);
874
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700875#ifdef CONFIG_NUMA
876extern void setup_per_cpu_pageset(void);
877#else
878static inline void setup_per_cpu_pageset(void) {}
879#endif
880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881/* prio_tree.c */
882void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
883void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
884void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
885struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
886 struct prio_tree_iter *iter);
887
888#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
889 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
890 (vma = vma_prio_tree_next(vma, iter)); )
891
892static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
893 struct list_head *list)
894{
895 vma->shared.vm_set.parent = NULL;
896 list_add_tail(&vma->shared.vm_set.list, list);
897}
898
899/* mmap.c */
900extern int __vm_enough_memory(long pages, int cap_sys_admin);
901extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
902 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
903extern struct vm_area_struct *vma_merge(struct mm_struct *,
904 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
905 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
906 struct mempolicy *);
907extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
908extern int split_vma(struct mm_struct *,
909 struct vm_area_struct *, unsigned long addr, int new_below);
910extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
911extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
912 struct rb_node **, struct rb_node *);
Hugh Dickinsa8fb5612005-10-29 18:15:57 -0700913extern void unlink_file_vma(struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
915 unsigned long addr, unsigned long len, pgoff_t pgoff);
916extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -0700917extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
919extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
920
921extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
922 unsigned long len, unsigned long prot,
923 unsigned long flag, unsigned long pgoff);
924
925static inline unsigned long do_mmap(struct file *file, unsigned long addr,
926 unsigned long len, unsigned long prot,
927 unsigned long flag, unsigned long offset)
928{
929 unsigned long ret = -EINVAL;
930 if ((offset + PAGE_ALIGN(len)) < offset)
931 goto out;
932 if (!(offset & ~PAGE_MASK))
933 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
934out:
935 return ret;
936}
937
938extern int do_munmap(struct mm_struct *, unsigned long, size_t);
939
940extern unsigned long do_brk(unsigned long, unsigned long);
941
942/* filemap.c */
943extern unsigned long page_unuse(struct page *);
944extern void truncate_inode_pages(struct address_space *, loff_t);
Hans Reiserd7339072006-01-06 00:10:36 -0800945extern void truncate_inode_pages_range(struct address_space *,
946 loff_t lstart, loff_t lend);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947
948/* generic vm_area_ops exported for stackable file systems */
949extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
950extern int filemap_populate(struct vm_area_struct *, unsigned long,
951 unsigned long, pgprot_t, unsigned long, int);
952
953/* mm/page-writeback.c */
954int write_one_page(struct page *page, int wait);
955
956/* readahead.c */
957#define VM_MAX_READAHEAD 128 /* kbytes */
958#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
959#define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
960 * turning readahead off */
961
962int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800963 pgoff_t offset, unsigned long nr_to_read);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800965 pgoff_t offset, unsigned long nr_to_read);
966unsigned long page_cache_readahead(struct address_space *mapping,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 struct file_ra_state *ra,
968 struct file *filp,
Andrew Morton7361f4d2005-11-07 00:59:28 -0800969 pgoff_t offset,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 unsigned long size);
971void handle_ra_miss(struct address_space *mapping,
972 struct file_ra_state *ra, pgoff_t offset);
973unsigned long max_sane_readahead(unsigned long nr);
974
975/* Do stack extension */
Hugh Dickins46dea3d2005-10-29 18:16:20 -0700976extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -0500977#ifdef CONFIG_IA64
Hugh Dickins46dea3d2005-10-29 18:16:20 -0700978extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
Matthew Wilcox9ab88512005-11-18 16:16:42 -0500979#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
981/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
982extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
983extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
984 struct vm_area_struct **pprev);
985
986/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
987 NULL if none. Assume start_addr < end_addr. */
988static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
989{
990 struct vm_area_struct * vma = find_vma(mm,start_addr);
991
992 if (vma && end_addr <= vma->vm_start)
993 vma = NULL;
994 return vma;
995}
996
997static inline unsigned long vma_pages(struct vm_area_struct *vma)
998{
999 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1000}
1001
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001002struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1003struct page *vmalloc_to_page(void *addr);
1004unsigned long vmalloc_to_pfn(void *addr);
1005int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1006 unsigned long pfn, unsigned long size, pgprot_t);
Linus Torvaldsa145dd42005-11-30 09:35:19 -08001007int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008
Linus Torvalds6aab3412005-11-28 14:34:23 -08001009struct page *follow_page(struct vm_area_struct *, unsigned long address,
Hugh Dickinsdeceb6c2005-10-29 18:16:33 -07001010 unsigned int foll_flags);
1011#define FOLL_WRITE 0x01 /* check pte is writable */
1012#define FOLL_TOUCH 0x02 /* mark page accessed */
1013#define FOLL_GET 0x04 /* do get_page on page */
1014#define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
1016#ifdef CONFIG_PROC_FS
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001017void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018#else
Hugh Dickinsab50b8e2005-10-29 18:15:56 -07001019static inline void vm_stat_account(struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 unsigned long flags, struct file *file, long pages)
1021{
1022}
1023#endif /* CONFIG_PROC_FS */
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025#ifndef CONFIG_DEBUG_PAGEALLOC
1026static inline void
1027kernel_map_pages(struct page *page, int numpages, int enable)
1028{
Ingo Molnarde5097c2006-01-09 15:59:21 -08001029 if (!PageHighMem(page) && !enable)
1030 mutex_debug_check_no_locks_freed(page_address(page),
David Woodhousea4fc7ab2006-01-11 14:41:26 +00001031 numpages * PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032}
1033#endif
1034
1035extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1036#ifdef __HAVE_ARCH_GATE_AREA
1037int in_gate_area_no_task(unsigned long addr);
1038int in_gate_area(struct task_struct *task, unsigned long addr);
1039#else
1040int in_gate_area_no_task(unsigned long addr);
1041#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1042#endif /* __HAVE_ARCH_GATE_AREA */
1043
Andrea Arcangeli79befd02005-04-16 15:24:05 -07001044/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1045#define OOM_DISABLE -17
1046
Andrew Morton9d0243b2006-01-08 01:00:39 -08001047int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1048 void __user *, size_t *, loff_t *);
1049int shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1050 unsigned long lru_pages);
1051void drop_pagecache(void);
1052void drop_slab(void);
1053
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054#endif /* __KERNEL__ */
1055#endif /* _LINUX_MM_H */