blob: 0d94c94d9d810f260e7f33e7d9a2ff73cf057de2 [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>
6
7#ifdef __KERNEL__
8
9#include <linux/config.h>
10#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>
16
17struct mempolicy;
18struct anon_vma;
19
20#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
21extern unsigned long max_mapnr;
22#endif
23
24extern unsigned long num_physpages;
25extern void * high_memory;
26extern unsigned long vmalloc_earlyreserve;
27extern int page_cluster;
28
29#ifdef CONFIG_SYSCTL
30extern int sysctl_legacy_va_layout;
31#else
32#define sysctl_legacy_va_layout 0
33#endif
34
35#include <asm/page.h>
36#include <asm/pgtable.h>
37#include <asm/processor.h>
38#include <asm/atomic.h>
39
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
41
42/*
43 * Linux kernel virtual memory manager primitives.
44 * The idea being to have a "virtual" mm in the same way
45 * we have a virtual fs - giving a cleaner interface to the
46 * mm details, and allowing different kinds of memory mappings
47 * (from shared memory to executable loading to arbitrary
48 * mmap() functions).
49 */
50
51/*
52 * This struct defines a memory VMM memory area. There is one of these
53 * per VM-area/task. A VM area is any part of the process virtual memory
54 * space that has a special rule for the page-fault handlers (ie a shared
55 * library, the executable area etc).
56 */
57struct vm_area_struct {
58 struct mm_struct * vm_mm; /* The address space we belong to. */
59 unsigned long vm_start; /* Our start address within vm_mm. */
60 unsigned long vm_end; /* The first byte after our end address
61 within vm_mm. */
62
63 /* linked list of VM areas per task, sorted by address */
64 struct vm_area_struct *vm_next;
65
66 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
67 unsigned long vm_flags; /* Flags, listed below. */
68
69 struct rb_node vm_rb;
70
71 /*
72 * For areas with an address space and backing store,
73 * linkage into the address_space->i_mmap prio tree, or
74 * linkage to the list of like vmas hanging off its node, or
75 * linkage of vma in the address_space->i_mmap_nonlinear list.
76 */
77 union {
78 struct {
79 struct list_head list;
80 void *parent; /* aligns with prio_tree_node parent */
81 struct vm_area_struct *head;
82 } vm_set;
83
84 struct raw_prio_tree_node prio_tree_node;
85 } shared;
86
87 /*
88 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
89 * list, after a COW of one of the file pages. A MAP_SHARED vma
90 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
91 * or brk vma (with NULL file) can only be in an anon_vma list.
92 */
93 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
94 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
95
96 /* Function pointers to deal with this struct. */
97 struct vm_operations_struct * vm_ops;
98
99 /* Information about our backing store: */
100 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
101 units, *not* PAGE_CACHE_SIZE */
102 struct file * vm_file; /* File we map to (can be NULL). */
103 void * vm_private_data; /* was vm_pte (shared mem) */
104 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
105
106#ifndef CONFIG_MMU
107 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
108#endif
109#ifdef CONFIG_NUMA
110 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
111#endif
112};
113
114/*
115 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
116 * disabled, then there's a single shared list of VMAs maintained by the
117 * system, and mm's subscribe to these individually
118 */
119struct vm_list_struct {
120 struct vm_list_struct *next;
121 struct vm_area_struct *vma;
122};
123
124#ifndef CONFIG_MMU
125extern struct rb_root nommu_vma_tree;
126extern struct rw_semaphore nommu_vma_sem;
127
128extern unsigned int kobjsize(const void *objp);
129#endif
130
131/*
132 * vm_flags..
133 */
134#define VM_READ 0x00000001 /* currently active flags */
135#define VM_WRITE 0x00000002
136#define VM_EXEC 0x00000004
137#define VM_SHARED 0x00000008
138
139#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
140#define VM_MAYWRITE 0x00000020
141#define VM_MAYEXEC 0x00000040
142#define VM_MAYSHARE 0x00000080
143
144#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
145#define VM_GROWSUP 0x00000200
146#define VM_SHM 0x00000400 /* shared memory area, don't swap out */
147#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
148
149#define VM_EXECUTABLE 0x00001000
150#define VM_LOCKED 0x00002000
151#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
152
153 /* Used by sys_madvise() */
154#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
155#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
156
157#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
158#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
159#define VM_RESERVED 0x00080000 /* Don't unmap it from swap_out */
160#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
161#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
162#define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
163#define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
164
165#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
166#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
167#endif
168
169#ifdef CONFIG_STACK_GROWSUP
170#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
171#else
172#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
173#endif
174
175#define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
176#define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
177#define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
178#define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
179#define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
180
181/*
182 * mapping from the currently active vm_flags protection bits (the
183 * low four bits) to a page protection mask..
184 */
185extern pgprot_t protection_map[16];
186
187
188/*
189 * These are the virtual MM functions - opening of an area, closing and
190 * unmapping it (needed to keep files on disk up-to-date etc), pointer
191 * to the functions called when a no-page or a wp-page exception occurs.
192 */
193struct vm_operations_struct {
194 void (*open)(struct vm_area_struct * area);
195 void (*close)(struct vm_area_struct * area);
196 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
197 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
198#ifdef CONFIG_NUMA
199 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
200 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
201 unsigned long addr);
202#endif
203};
204
205struct mmu_gather;
206struct inode;
207
208#ifdef ARCH_HAS_ATOMIC_UNSIGNED
209typedef unsigned page_flags_t;
210#else
211typedef unsigned long page_flags_t;
212#endif
213
214/*
215 * Each physical page in the system has a struct page associated with
216 * it to keep track of whatever it is we are using the page for at the
217 * moment. Note that we have no way to track which tasks are using
218 * a page.
219 */
220struct page {
221 page_flags_t flags; /* Atomic flags, some possibly
222 * updated asynchronously */
223 atomic_t _count; /* Usage count, see below. */
224 atomic_t _mapcount; /* Count of ptes mapped in mms,
225 * to show when page is mapped
226 * & limit reverse map searches.
227 */
228 unsigned long private; /* Mapping-private opaque data:
229 * usually used for buffer_heads
230 * if PagePrivate set; used for
231 * swp_entry_t if PageSwapCache
232 * When page is free, this indicates
233 * order in the buddy system.
234 */
235 struct address_space *mapping; /* If low bit clear, points to
236 * inode address_space, or NULL.
237 * If page mapped as anonymous
238 * memory, low bit is set, and
239 * it points to anon_vma object:
240 * see PAGE_MAPPING_ANON below.
241 */
242 pgoff_t index; /* Our offset within mapping. */
243 struct list_head lru; /* Pageout list, eg. active_list
244 * protected by zone->lru_lock !
245 */
246 /*
247 * On machines where all RAM is mapped into kernel address space,
248 * we can simply calculate the virtual address. On machines with
249 * highmem some memory is mapped into kernel virtual memory
250 * dynamically, so we need a place to store that address.
251 * Note that this field could be 16 bits on x86 ... ;)
252 *
253 * Architectures with slow multiplication can define
254 * WANT_PAGE_VIRTUAL in asm/page.h
255 */
256#if defined(WANT_PAGE_VIRTUAL)
257 void *virtual; /* Kernel virtual address (NULL if
258 not kmapped, ie. highmem) */
259#endif /* WANT_PAGE_VIRTUAL */
260};
261
262/*
263 * FIXME: take this include out, include page-flags.h in
264 * files which need it (119 of them)
265 */
266#include <linux/page-flags.h>
267
268/*
269 * Methods to modify the page usage count.
270 *
271 * What counts for a page usage:
272 * - cache mapping (page->mapping)
273 * - private data (page->private)
274 * - page mapped in a task's page tables, each mapping
275 * is counted separately
276 *
277 * Also, many kernel routines increase the page count before a critical
278 * routine so they can be sure the page doesn't go away from under them.
279 *
280 * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we
281 * can use atomic_add_negative(-1, page->_count) to detect when the page
282 * becomes free and so that we can also use atomic_inc_and_test to atomically
283 * detect when we just tried to grab a ref on a page which some other CPU has
284 * already deemed to be freeable.
285 *
286 * NO code should make assumptions about this internal detail! Use the provided
287 * macros which retain the old rules: page_count(page) == 0 is a free page.
288 */
289
290/*
291 * Drop a ref, return true if the logical refcount fell to zero (the page has
292 * no users)
293 */
294#define put_page_testzero(p) \
295 ({ \
296 BUG_ON(page_count(p) == 0); \
297 atomic_add_negative(-1, &(p)->_count); \
298 })
299
300/*
301 * Grab a ref, return true if the page previously had a logical refcount of
302 * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page
303 */
304#define get_page_testone(p) atomic_inc_and_test(&(p)->_count)
305
306#define set_page_count(p,v) atomic_set(&(p)->_count, v - 1)
307#define __put_page(p) atomic_dec(&(p)->_count)
308
309extern void FASTCALL(__page_cache_release(struct page *));
310
311#ifdef CONFIG_HUGETLB_PAGE
312
313static inline int page_count(struct page *p)
314{
315 if (PageCompound(p))
316 p = (struct page *)p->private;
317 return atomic_read(&(p)->_count) + 1;
318}
319
320static inline void get_page(struct page *page)
321{
322 if (unlikely(PageCompound(page)))
323 page = (struct page *)page->private;
324 atomic_inc(&page->_count);
325}
326
327void put_page(struct page *page);
328
329#else /* CONFIG_HUGETLB_PAGE */
330
331#define page_count(p) (atomic_read(&(p)->_count) + 1)
332
333static inline void get_page(struct page *page)
334{
335 atomic_inc(&page->_count);
336}
337
338static inline void put_page(struct page *page)
339{
340 if (!PageReserved(page) && put_page_testzero(page))
341 __page_cache_release(page);
342}
343
344#endif /* CONFIG_HUGETLB_PAGE */
345
346/*
347 * Multiple processes may "see" the same page. E.g. for untouched
348 * mappings of /dev/null, all processes see the same page full of
349 * zeroes, and text pages of executables and shared libraries have
350 * only one copy in memory, at most, normally.
351 *
352 * For the non-reserved pages, page_count(page) denotes a reference count.
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700353 * page_count() == 0 means the page is free. page->lru is then used for
354 * freelist management in the buddy allocator.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 * page_count() == 1 means the page is used for exactly one purpose
356 * (e.g. a private data page of one process).
357 *
358 * A page may be used for kmalloc() or anyone else who does a
359 * __get_free_page(). In this case the page_count() is at least 1, and
360 * all other fields are unused but should be 0 or NULL. The
361 * management of this page is the responsibility of the one who uses
362 * it.
363 *
364 * The other pages (we may call them "process pages") are completely
365 * managed by the Linux memory manager: I/O, buffers, swapping etc.
366 * The following discussion applies only to them.
367 *
368 * A page may belong to an inode's memory mapping. In this case,
369 * page->mapping is the pointer to the inode, and page->index is the
370 * file offset of the page, in units of PAGE_CACHE_SIZE.
371 *
372 * A page contains an opaque `private' member, which belongs to the
373 * page's address_space. Usually, this is the address of a circular
374 * list of the page's disk buffers.
375 *
376 * For pages belonging to inodes, the page_count() is the number of
377 * attaches, plus 1 if `private' contains something, plus one for
378 * the page cache itself.
379 *
Paolo 'Blaisorblade' Giarrusso7e871b62005-09-21 09:55:38 -0700380 * Instead of keeping dirty/clean pages in per address-space lists, we instead
381 * now tag pages as dirty/under writeback in the radix tree.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 *
383 * There is also a per-mapping radix tree mapping index to the page
384 * in memory if present. The tree is rooted at mapping->root.
385 *
386 * All process pages can do I/O:
387 * - inode pages may need to be read from disk,
388 * - inode pages which have been modified and are MAP_SHARED may need
389 * to be written to disk,
390 * - private pages which have been modified may need to be swapped out
391 * to swap space and (later) to be read back into memory.
392 */
393
394/*
395 * The zone field is never updated after free_area_init_core()
396 * sets it, so none of the operations on it need to be atomic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700398
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700399
400/*
401 * page->flags layout:
402 *
403 * There are three possibilities for how page->flags get
404 * laid out. The first is for the normal case, without
405 * sparsemem. The second is for sparsemem when there is
406 * plenty of space for node and section. The last is when
407 * we have run out of space and have to fall back to an
408 * alternate (slower) way of determining the node.
409 *
410 * No sparsemem: | NODE | ZONE | ... | FLAGS |
411 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
412 * no space for node: | SECTION | ZONE | ... | FLAGS |
413 */
414#ifdef CONFIG_SPARSEMEM
415#define SECTIONS_WIDTH SECTIONS_SHIFT
416#else
417#define SECTIONS_WIDTH 0
418#endif
419
420#define ZONES_WIDTH ZONES_SHIFT
421
422#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
423#define NODES_WIDTH NODES_SHIFT
424#else
425#define NODES_WIDTH 0
426#endif
427
428/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
429#define SECTIONS_PGOFF ((sizeof(page_flags_t)*8) - SECTIONS_WIDTH)
430#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
431#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
432
433/*
434 * We are going to use the flags for the page to node mapping if its in
435 * there. This includes the case where there is no node, so it is implicit.
436 */
437#define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
438
439#ifndef PFN_SECTION_SHIFT
440#define PFN_SECTION_SHIFT 0
441#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700442
443/*
444 * Define the bit shifts to access each section. For non-existant
445 * sections we define the shift as 0; that plus a 0 mask ensures
446 * the compiler will optimise away reference to them.
447 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700448#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
449#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
450#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
Dave Hansen348f8b62005-06-23 00:07:40 -0700451
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700452/* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
453#if FLAGS_HAS_NODE
Dave Hansen348f8b62005-06-23 00:07:40 -0700454#define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700455#else
456#define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
457#endif
Dave Hansen348f8b62005-06-23 00:07:40 -0700458#define ZONETABLE_PGSHIFT ZONES_PGSHIFT
459
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700460#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
461#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
Dave Hansen348f8b62005-06-23 00:07:40 -0700462#endif
463
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700464#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
465#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
466#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
Dave Hansen348f8b62005-06-23 00:07:40 -0700467#define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469static inline unsigned long page_zonenum(struct page *page)
470{
Dave Hansen348f8b62005-06-23 00:07:40 -0700471 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
474struct zone;
475extern struct zone *zone_table[];
476
477static inline struct zone *page_zone(struct page *page)
478{
Dave Hansen348f8b62005-06-23 00:07:40 -0700479 return zone_table[(page->flags >> ZONETABLE_PGSHIFT) &
480 ZONETABLE_MASK];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481}
482
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700483static inline unsigned long page_to_nid(struct page *page)
484{
485 if (FLAGS_HAS_NODE)
486 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
487 else
488 return page_zone(page)->zone_pgdat->node_id;
489}
490static inline unsigned long page_to_section(struct page *page)
491{
492 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
493}
494
Dave Hansen348f8b62005-06-23 00:07:40 -0700495static inline void set_page_zone(struct page *page, unsigned long zone)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
Dave Hansen348f8b62005-06-23 00:07:40 -0700497 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
498 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
499}
500static inline void set_page_node(struct page *page, unsigned long node)
501{
502 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
503 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
504}
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700505static inline void set_page_section(struct page *page, unsigned long section)
506{
507 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
508 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
509}
Dave Hansen348f8b62005-06-23 00:07:40 -0700510
511static inline void set_page_links(struct page *page, unsigned long zone,
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700512 unsigned long node, unsigned long pfn)
Dave Hansen348f8b62005-06-23 00:07:40 -0700513{
514 set_page_zone(page, zone);
515 set_page_node(page, node);
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700516 set_page_section(page, pfn_to_section_nr(pfn));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517}
518
519#ifndef CONFIG_DISCONTIGMEM
520/* The array of struct pages - for discontigmem use pgdat->lmem_map */
521extern struct page *mem_map;
522#endif
523
524static inline void *lowmem_page_address(struct page *page)
525{
526 return __va(page_to_pfn(page) << PAGE_SHIFT);
527}
528
529#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
530#define HASHED_PAGE_VIRTUAL
531#endif
532
533#if defined(WANT_PAGE_VIRTUAL)
534#define page_address(page) ((page)->virtual)
535#define set_page_address(page, address) \
536 do { \
537 (page)->virtual = (address); \
538 } while(0)
539#define page_address_init() do { } while(0)
540#endif
541
542#if defined(HASHED_PAGE_VIRTUAL)
543void *page_address(struct page *page);
544void set_page_address(struct page *page, void *virtual);
545void page_address_init(void);
546#endif
547
548#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
549#define page_address(page) lowmem_page_address(page)
550#define set_page_address(page, address) do { } while(0)
551#define page_address_init() do { } while(0)
552#endif
553
554/*
555 * On an anonymous page mapped into a user virtual memory area,
556 * page->mapping points to its anon_vma, not to a struct address_space;
557 * with the PAGE_MAPPING_ANON bit set to distinguish it.
558 *
559 * Please note that, confusingly, "page_mapping" refers to the inode
560 * address_space which maps the page from disk; whereas "page_mapped"
561 * refers to user virtual address space into which the page is mapped.
562 */
563#define PAGE_MAPPING_ANON 1
564
565extern struct address_space swapper_space;
566static inline struct address_space *page_mapping(struct page *page)
567{
568 struct address_space *mapping = page->mapping;
569
570 if (unlikely(PageSwapCache(page)))
571 mapping = &swapper_space;
572 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
573 mapping = NULL;
574 return mapping;
575}
576
577static inline int PageAnon(struct page *page)
578{
579 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
580}
581
582/*
583 * Return the pagecache index of the passed page. Regular pagecache pages
584 * use ->index whereas swapcache pages use ->private
585 */
586static inline pgoff_t page_index(struct page *page)
587{
588 if (unlikely(PageSwapCache(page)))
589 return page->private;
590 return page->index;
591}
592
593/*
594 * The atomic page->_mapcount, like _count, starts from -1:
595 * so that transitions both from it and to it can be tracked,
596 * using atomic_inc_and_test and atomic_add_negative(-1).
597 */
598static inline void reset_page_mapcount(struct page *page)
599{
600 atomic_set(&(page)->_mapcount, -1);
601}
602
603static inline int page_mapcount(struct page *page)
604{
605 return atomic_read(&(page)->_mapcount) + 1;
606}
607
608/*
609 * Return true if this page is mapped into pagetables.
610 */
611static inline int page_mapped(struct page *page)
612{
613 return atomic_read(&(page)->_mapcount) >= 0;
614}
615
616/*
617 * Error return values for the *_nopage functions
618 */
619#define NOPAGE_SIGBUS (NULL)
620#define NOPAGE_OOM ((struct page *) (-1))
621
622/*
623 * Different kinds of faults, as returned by handle_mm_fault().
624 * Used to decide whether a process gets delivered SIGBUS or
625 * just gets major/minor fault counters bumped up.
626 */
Nick Pigginf33ea7f2005-08-03 20:24:01 +1000627#define VM_FAULT_OOM 0x00
628#define VM_FAULT_SIGBUS 0x01
629#define VM_FAULT_MINOR 0x02
630#define VM_FAULT_MAJOR 0x03
631
632/*
633 * Special case for get_user_pages.
634 * Must be in a distinct bit from the above VM_FAULT_ flags.
635 */
636#define VM_FAULT_WRITE 0x10
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
638#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
639
640extern void show_free_areas(void);
641
642#ifdef CONFIG_SHMEM
643struct page *shmem_nopage(struct vm_area_struct *vma,
644 unsigned long address, int *type);
645int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
646struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
647 unsigned long addr);
648int shmem_lock(struct file *file, int lock, struct user_struct *user);
649#else
650#define shmem_nopage filemap_nopage
651#define shmem_lock(a, b, c) ({0;}) /* always in memory, no need to lock */
652#define shmem_set_policy(a, b) (0)
653#define shmem_get_policy(a, b) (NULL)
654#endif
655struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
656
657int shmem_zero_setup(struct vm_area_struct *);
658
659static inline int can_do_mlock(void)
660{
661 if (capable(CAP_IPC_LOCK))
662 return 1;
663 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
664 return 1;
665 return 0;
666}
667extern int user_shm_lock(size_t, struct user_struct *);
668extern void user_shm_unlock(size_t, struct user_struct *);
669
670/*
671 * Parameter block passed down to zap_pte_range in exceptional cases.
672 */
673struct zap_details {
674 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
675 struct address_space *check_mapping; /* Check page->mapping if set */
676 pgoff_t first_index; /* Lowest page->index to unmap */
677 pgoff_t last_index; /* Highest page->index to unmap */
678 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 unsigned long truncate_count; /* Compare vm_truncate_count */
680};
681
Hugh Dickinsee39b372005-04-19 13:29:15 -0700682unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 unsigned long size, struct zap_details *);
Hugh Dickinsee39b372005-04-19 13:29:15 -0700684unsigned long unmap_vmas(struct mmu_gather **tlb, struct mm_struct *mm,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 struct vm_area_struct *start_vma, unsigned long start_addr,
686 unsigned long end_addr, unsigned long *nr_accounted,
687 struct zap_details *);
Hugh Dickins3bf5ee92005-04-19 13:29:16 -0700688void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
689 unsigned long end, unsigned long floor, unsigned long ceiling);
690void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
Hugh Dickinse0da3822005-04-19 13:29:15 -0700691 unsigned long floor, unsigned long ceiling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
693 struct vm_area_struct *vma);
694int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
695 unsigned long size, pgprot_t prot);
696void unmap_mapping_range(struct address_space *mapping,
697 loff_t const holebegin, loff_t const holelen, int even_cows);
698
699static inline void unmap_shared_mapping_range(struct address_space *mapping,
700 loff_t const holebegin, loff_t const holelen)
701{
702 unmap_mapping_range(mapping, holebegin, holelen, 0);
703}
704
705extern int vmtruncate(struct inode * inode, loff_t offset);
706extern pud_t *FASTCALL(__pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address));
707extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address));
708extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address));
709extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address));
710extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
711extern 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 +1000712extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access);
713
714static inline int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, int write_access)
715{
716 return __handle_mm_fault(mm, vma, address, write_access) & (~VM_FAULT_WRITE);
717}
718
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719extern int make_pages_present(unsigned long addr, unsigned long end);
720extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
721void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
722
723int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
724 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
725
726int __set_page_dirty_buffers(struct page *page);
727int __set_page_dirty_nobuffers(struct page *page);
728int redirty_page_for_writepage(struct writeback_control *wbc,
729 struct page *page);
730int FASTCALL(set_page_dirty(struct page *page));
731int set_page_dirty_lock(struct page *page);
732int clear_page_dirty_for_io(struct page *page);
733
734extern unsigned long do_mremap(unsigned long addr,
735 unsigned long old_len, unsigned long new_len,
736 unsigned long flags, unsigned long new_addr);
737
738/*
739 * Prototype to add a shrinker callback for ageable caches.
740 *
741 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
742 * scan `nr_to_scan' objects, attempting to free them.
743 *
Domen Puncer845d3432005-05-05 16:16:14 -0700744 * The callback must return the number of objects which remain in the cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 *
Domen Puncer845d3432005-05-05 16:16:14 -0700746 * The callback will be passed nr_to_scan == 0 when the VM is querying the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 * cache size, so a fastpath for that case is appropriate.
748 */
749typedef int (*shrinker_t)(int nr_to_scan, unsigned int gfp_mask);
750
751/*
752 * Add an aging callback. The int is the number of 'seeks' it takes
753 * to recreate one of the objects that these functions age.
754 */
755
756#define DEFAULT_SEEKS 2
757struct shrinker;
758extern struct shrinker *set_shrinker(int, shrinker_t);
759extern void remove_shrinker(struct shrinker *shrinker);
760
761/*
762 * On a two-level or three-level page table, this ends up being trivial. Thus
763 * the inlining and the symmetry break with pte_alloc_map() that does all
764 * of this out-of-line.
765 */
766/*
767 * The following ifdef needed to get the 4level-fixup.h header to work.
768 * Remove it when 4level-fixup.h has been removed.
769 */
770#ifdef CONFIG_MMU
771#ifndef __ARCH_HAS_4LEVEL_HACK
772static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
773{
774 if (pgd_none(*pgd))
775 return __pud_alloc(mm, pgd, address);
776 return pud_offset(pgd, address);
777}
778
779static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
780{
781 if (pud_none(*pud))
782 return __pmd_alloc(mm, pud, address);
783 return pmd_offset(pud, address);
784}
785#endif
786#endif /* CONFIG_MMU */
787
788extern void free_area_init(unsigned long * zones_size);
789extern void free_area_init_node(int nid, pg_data_t *pgdat,
790 unsigned long * zones_size, unsigned long zone_start_pfn,
791 unsigned long *zholes_size);
792extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
793extern void mem_init(void);
794extern void show_mem(void);
795extern void si_meminfo(struct sysinfo * val);
796extern void si_meminfo_node(struct sysinfo *val, int nid);
797
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700798#ifdef CONFIG_NUMA
799extern void setup_per_cpu_pageset(void);
800#else
801static inline void setup_per_cpu_pageset(void) {}
802#endif
803
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804/* prio_tree.c */
805void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
806void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
807void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
808struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
809 struct prio_tree_iter *iter);
810
811#define vma_prio_tree_foreach(vma, iter, root, begin, end) \
812 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
813 (vma = vma_prio_tree_next(vma, iter)); )
814
815static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
816 struct list_head *list)
817{
818 vma->shared.vm_set.parent = NULL;
819 list_add_tail(&vma->shared.vm_set.list, list);
820}
821
822/* mmap.c */
823extern int __vm_enough_memory(long pages, int cap_sys_admin);
824extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
825 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
826extern struct vm_area_struct *vma_merge(struct mm_struct *,
827 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
828 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
829 struct mempolicy *);
830extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
831extern int split_vma(struct mm_struct *,
832 struct vm_area_struct *, unsigned long addr, int new_below);
833extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
834extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
835 struct rb_node **, struct rb_node *);
836extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
837 unsigned long addr, unsigned long len, pgoff_t pgoff);
838extern void exit_mmap(struct mm_struct *);
akpm@osdl.org119f6572005-05-01 08:58:35 -0700839extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
841extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
842
843extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
844 unsigned long len, unsigned long prot,
845 unsigned long flag, unsigned long pgoff);
846
847static inline unsigned long do_mmap(struct file *file, unsigned long addr,
848 unsigned long len, unsigned long prot,
849 unsigned long flag, unsigned long offset)
850{
851 unsigned long ret = -EINVAL;
852 if ((offset + PAGE_ALIGN(len)) < offset)
853 goto out;
854 if (!(offset & ~PAGE_MASK))
855 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
856out:
857 return ret;
858}
859
860extern int do_munmap(struct mm_struct *, unsigned long, size_t);
861
862extern unsigned long do_brk(unsigned long, unsigned long);
863
864/* filemap.c */
865extern unsigned long page_unuse(struct page *);
866extern void truncate_inode_pages(struct address_space *, loff_t);
867
868/* generic vm_area_ops exported for stackable file systems */
869extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
870extern int filemap_populate(struct vm_area_struct *, unsigned long,
871 unsigned long, pgprot_t, unsigned long, int);
872
873/* mm/page-writeback.c */
874int write_one_page(struct page *page, int wait);
875
876/* readahead.c */
877#define VM_MAX_READAHEAD 128 /* kbytes */
878#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
879#define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
880 * turning readahead off */
881
882int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
883 unsigned long offset, unsigned long nr_to_read);
884int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
885 unsigned long offset, unsigned long nr_to_read);
886unsigned long page_cache_readahead(struct address_space *mapping,
887 struct file_ra_state *ra,
888 struct file *filp,
889 unsigned long offset,
890 unsigned long size);
891void handle_ra_miss(struct address_space *mapping,
892 struct file_ra_state *ra, pgoff_t offset);
893unsigned long max_sane_readahead(unsigned long nr);
894
895/* Do stack extension */
896extern int expand_stack(struct vm_area_struct * vma, unsigned long address);
897
898/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
899extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
900extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
901 struct vm_area_struct **pprev);
902
903/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
904 NULL if none. Assume start_addr < end_addr. */
905static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
906{
907 struct vm_area_struct * vma = find_vma(mm,start_addr);
908
909 if (vma && end_addr <= vma->vm_start)
910 vma = NULL;
911 return vma;
912}
913
914static inline unsigned long vma_pages(struct vm_area_struct *vma)
915{
916 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
917}
918
919extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr);
920
921extern struct page * vmalloc_to_page(void *addr);
922extern unsigned long vmalloc_to_pfn(void *addr);
923extern struct page * follow_page(struct mm_struct *mm, unsigned long address,
924 int write);
925extern int check_user_page_readable(struct mm_struct *mm, unsigned long address);
926int remap_pfn_range(struct vm_area_struct *, unsigned long,
927 unsigned long, unsigned long, pgprot_t);
928
929#ifdef CONFIG_PROC_FS
930void __vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
931#else
932static inline void __vm_stat_account(struct mm_struct *mm,
933 unsigned long flags, struct file *file, long pages)
934{
935}
936#endif /* CONFIG_PROC_FS */
937
938static inline void vm_stat_account(struct vm_area_struct *vma)
939{
940 __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file,
941 vma_pages(vma));
942}
943
944static inline void vm_stat_unaccount(struct vm_area_struct *vma)
945{
946 __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file,
947 -vma_pages(vma));
948}
949
950/* update per process rss and vm hiwater data */
951extern void update_mem_hiwater(struct task_struct *tsk);
952
953#ifndef CONFIG_DEBUG_PAGEALLOC
954static inline void
955kernel_map_pages(struct page *page, int numpages, int enable)
956{
957}
958#endif
959
960extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
961#ifdef __HAVE_ARCH_GATE_AREA
962int in_gate_area_no_task(unsigned long addr);
963int in_gate_area(struct task_struct *task, unsigned long addr);
964#else
965int in_gate_area_no_task(unsigned long addr);
966#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
967#endif /* __HAVE_ARCH_GATE_AREA */
968
Andrea Arcangeli79befd02005-04-16 15:24:05 -0700969/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
970#define OOM_DISABLE -17
971
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972#endif /* __KERNEL__ */
973#endif /* _LINUX_MM_H */