Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #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 | ||
17 | struct mempolicy; | |
18 | struct anon_vma; | |
19 | ||
20 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ | |
21 | extern unsigned long max_mapnr; | |
22 | #endif | |
23 | ||
24 | extern unsigned long num_physpages; | |
25 | extern void * high_memory; | |
26 | extern unsigned long vmalloc_earlyreserve; | |
27 | extern int page_cluster; | |
28 | ||
29 | #ifdef CONFIG_SYSCTL | |
30 | extern 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 | ||
1da177e4 LT |
40 | #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 | */ | |
57 | struct 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 | */ | |
119 | struct vm_list_struct { | |
120 | struct vm_list_struct *next; | |
121 | struct vm_area_struct *vma; | |
122 | }; | |
123 | ||
124 | #ifndef CONFIG_MMU | |
125 | extern struct rb_root nommu_vma_tree; | |
126 | extern struct rw_semaphore nommu_vma_sem; | |
127 | ||
128 | extern 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 | */ | |
185 | extern 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 | */ | |
193 | struct 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 | ||
205 | struct mmu_gather; | |
206 | struct inode; | |
207 | ||
208 | #ifdef ARCH_HAS_ATOMIC_UNSIGNED | |
209 | typedef unsigned page_flags_t; | |
210 | #else | |
211 | typedef 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 | */ | |
220 | struct 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 | ||
309 | extern void FASTCALL(__page_cache_release(struct page *)); | |
310 | ||
311 | #ifdef CONFIG_HUGETLB_PAGE | |
312 | ||
313 | static 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 | ||
320 | static 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 | ||
327 | void put_page(struct page *page); | |
328 | ||
329 | #else /* CONFIG_HUGETLB_PAGE */ | |
330 | ||
331 | #define page_count(p) (atomic_read(&(p)->_count) + 1) | |
332 | ||
333 | static inline void get_page(struct page *page) | |
334 | { | |
335 | atomic_inc(&page->_count); | |
336 | } | |
337 | ||
338 | static 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. | |
353 | * page_count() == 0 means the page is free. | |
354 | * page_count() == 1 means the page is used for exactly one purpose | |
355 | * (e.g. a private data page of one process). | |
356 | * | |
357 | * A page may be used for kmalloc() or anyone else who does a | |
358 | * __get_free_page(). In this case the page_count() is at least 1, and | |
359 | * all other fields are unused but should be 0 or NULL. The | |
360 | * management of this page is the responsibility of the one who uses | |
361 | * it. | |
362 | * | |
363 | * The other pages (we may call them "process pages") are completely | |
364 | * managed by the Linux memory manager: I/O, buffers, swapping etc. | |
365 | * The following discussion applies only to them. | |
366 | * | |
367 | * A page may belong to an inode's memory mapping. In this case, | |
368 | * page->mapping is the pointer to the inode, and page->index is the | |
369 | * file offset of the page, in units of PAGE_CACHE_SIZE. | |
370 | * | |
371 | * A page contains an opaque `private' member, which belongs to the | |
372 | * page's address_space. Usually, this is the address of a circular | |
373 | * list of the page's disk buffers. | |
374 | * | |
375 | * For pages belonging to inodes, the page_count() is the number of | |
376 | * attaches, plus 1 if `private' contains something, plus one for | |
377 | * the page cache itself. | |
378 | * | |
379 | * All pages belonging to an inode are in these doubly linked lists: | |
380 | * mapping->clean_pages, mapping->dirty_pages and mapping->locked_pages; | |
381 | * using the page->list list_head. These fields are also used for | |
382 | * freelist managemet (when page_count()==0). | |
383 | * | |
384 | * There is also a per-mapping radix tree mapping index to the page | |
385 | * in memory if present. The tree is rooted at mapping->root. | |
386 | * | |
387 | * All process pages can do I/O: | |
388 | * - inode pages may need to be read from disk, | |
389 | * - inode pages which have been modified and are MAP_SHARED may need | |
390 | * to be written to disk, | |
391 | * - private pages which have been modified may need to be swapped out | |
392 | * to swap space and (later) to be read back into memory. | |
393 | */ | |
394 | ||
395 | /* | |
396 | * The zone field is never updated after free_area_init_core() | |
397 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 398 | */ |
348f8b6c | 399 | |
d41dee36 AW |
400 | |
401 | /* | |
402 | * page->flags layout: | |
403 | * | |
404 | * There are three possibilities for how page->flags get | |
405 | * laid out. The first is for the normal case, without | |
406 | * sparsemem. The second is for sparsemem when there is | |
407 | * plenty of space for node and section. The last is when | |
408 | * we have run out of space and have to fall back to an | |
409 | * alternate (slower) way of determining the node. | |
410 | * | |
411 | * No sparsemem: | NODE | ZONE | ... | FLAGS | | |
412 | * with space for node: | SECTION | NODE | ZONE | ... | FLAGS | | |
413 | * no space for node: | SECTION | ZONE | ... | FLAGS | | |
414 | */ | |
415 | #ifdef CONFIG_SPARSEMEM | |
416 | #define SECTIONS_WIDTH SECTIONS_SHIFT | |
417 | #else | |
418 | #define SECTIONS_WIDTH 0 | |
419 | #endif | |
420 | ||
421 | #define ZONES_WIDTH ZONES_SHIFT | |
422 | ||
423 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED | |
424 | #define NODES_WIDTH NODES_SHIFT | |
425 | #else | |
426 | #define NODES_WIDTH 0 | |
427 | #endif | |
428 | ||
429 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ | |
430 | #define SECTIONS_PGOFF ((sizeof(page_flags_t)*8) - SECTIONS_WIDTH) | |
431 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) | |
432 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
433 | ||
434 | /* | |
435 | * We are going to use the flags for the page to node mapping if its in | |
436 | * there. This includes the case where there is no node, so it is implicit. | |
437 | */ | |
438 | #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0) | |
439 | ||
440 | #ifndef PFN_SECTION_SHIFT | |
441 | #define PFN_SECTION_SHIFT 0 | |
442 | #endif | |
348f8b6c DH |
443 | |
444 | /* | |
445 | * Define the bit shifts to access each section. For non-existant | |
446 | * sections we define the shift as 0; that plus a 0 mask ensures | |
447 | * the compiler will optimise away reference to them. | |
448 | */ | |
d41dee36 AW |
449 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
450 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
451 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
348f8b6c | 452 | |
d41dee36 AW |
453 | /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */ |
454 | #if FLAGS_HAS_NODE | |
348f8b6c | 455 | #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
d41dee36 AW |
456 | #else |
457 | #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
458 | #endif | |
348f8b6c DH |
459 | #define ZONETABLE_PGSHIFT ZONES_PGSHIFT |
460 | ||
d41dee36 AW |
461 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED |
462 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED | |
348f8b6c DH |
463 | #endif |
464 | ||
d41dee36 AW |
465 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
466 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
467 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
348f8b6c DH |
468 | #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1) |
469 | ||
1da177e4 LT |
470 | static inline unsigned long page_zonenum(struct page *page) |
471 | { | |
348f8b6c | 472 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 473 | } |
1da177e4 LT |
474 | |
475 | struct zone; | |
476 | extern struct zone *zone_table[]; | |
477 | ||
478 | static inline struct zone *page_zone(struct page *page) | |
479 | { | |
348f8b6c DH |
480 | return zone_table[(page->flags >> ZONETABLE_PGSHIFT) & |
481 | ZONETABLE_MASK]; | |
482 | } | |
483 | ||
d41dee36 AW |
484 | static inline unsigned long page_to_nid(struct page *page) |
485 | { | |
486 | if (FLAGS_HAS_NODE) | |
487 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; | |
488 | else | |
489 | return page_zone(page)->zone_pgdat->node_id; | |
490 | } | |
491 | static inline unsigned long page_to_section(struct page *page) | |
492 | { | |
493 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
494 | } | |
495 | ||
348f8b6c DH |
496 | static inline void set_page_zone(struct page *page, unsigned long zone) |
497 | { | |
498 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
499 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
500 | } | |
501 | static inline void set_page_node(struct page *page, unsigned long node) | |
502 | { | |
503 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
504 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 505 | } |
d41dee36 AW |
506 | static inline void set_page_section(struct page *page, unsigned long section) |
507 | { | |
508 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
509 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
510 | } | |
1da177e4 | 511 | |
348f8b6c | 512 | static inline void set_page_links(struct page *page, unsigned long zone, |
d41dee36 | 513 | unsigned long node, unsigned long pfn) |
1da177e4 | 514 | { |
348f8b6c DH |
515 | set_page_zone(page, zone); |
516 | set_page_node(page, node); | |
d41dee36 | 517 | set_page_section(page, pfn_to_section_nr(pfn)); |
1da177e4 LT |
518 | } |
519 | ||
520 | #ifndef CONFIG_DISCONTIGMEM | |
521 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ | |
522 | extern struct page *mem_map; | |
523 | #endif | |
524 | ||
525 | static inline void *lowmem_page_address(struct page *page) | |
526 | { | |
527 | return __va(page_to_pfn(page) << PAGE_SHIFT); | |
528 | } | |
529 | ||
530 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
531 | #define HASHED_PAGE_VIRTUAL | |
532 | #endif | |
533 | ||
534 | #if defined(WANT_PAGE_VIRTUAL) | |
535 | #define page_address(page) ((page)->virtual) | |
536 | #define set_page_address(page, address) \ | |
537 | do { \ | |
538 | (page)->virtual = (address); \ | |
539 | } while(0) | |
540 | #define page_address_init() do { } while(0) | |
541 | #endif | |
542 | ||
543 | #if defined(HASHED_PAGE_VIRTUAL) | |
544 | void *page_address(struct page *page); | |
545 | void set_page_address(struct page *page, void *virtual); | |
546 | void page_address_init(void); | |
547 | #endif | |
548 | ||
549 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
550 | #define page_address(page) lowmem_page_address(page) | |
551 | #define set_page_address(page, address) do { } while(0) | |
552 | #define page_address_init() do { } while(0) | |
553 | #endif | |
554 | ||
555 | /* | |
556 | * On an anonymous page mapped into a user virtual memory area, | |
557 | * page->mapping points to its anon_vma, not to a struct address_space; | |
558 | * with the PAGE_MAPPING_ANON bit set to distinguish it. | |
559 | * | |
560 | * Please note that, confusingly, "page_mapping" refers to the inode | |
561 | * address_space which maps the page from disk; whereas "page_mapped" | |
562 | * refers to user virtual address space into which the page is mapped. | |
563 | */ | |
564 | #define PAGE_MAPPING_ANON 1 | |
565 | ||
566 | extern struct address_space swapper_space; | |
567 | static inline struct address_space *page_mapping(struct page *page) | |
568 | { | |
569 | struct address_space *mapping = page->mapping; | |
570 | ||
571 | if (unlikely(PageSwapCache(page))) | |
572 | mapping = &swapper_space; | |
573 | else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) | |
574 | mapping = NULL; | |
575 | return mapping; | |
576 | } | |
577 | ||
578 | static inline int PageAnon(struct page *page) | |
579 | { | |
580 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
581 | } | |
582 | ||
583 | /* | |
584 | * Return the pagecache index of the passed page. Regular pagecache pages | |
585 | * use ->index whereas swapcache pages use ->private | |
586 | */ | |
587 | static inline pgoff_t page_index(struct page *page) | |
588 | { | |
589 | if (unlikely(PageSwapCache(page))) | |
590 | return page->private; | |
591 | return page->index; | |
592 | } | |
593 | ||
594 | /* | |
595 | * The atomic page->_mapcount, like _count, starts from -1: | |
596 | * so that transitions both from it and to it can be tracked, | |
597 | * using atomic_inc_and_test and atomic_add_negative(-1). | |
598 | */ | |
599 | static inline void reset_page_mapcount(struct page *page) | |
600 | { | |
601 | atomic_set(&(page)->_mapcount, -1); | |
602 | } | |
603 | ||
604 | static inline int page_mapcount(struct page *page) | |
605 | { | |
606 | return atomic_read(&(page)->_mapcount) + 1; | |
607 | } | |
608 | ||
609 | /* | |
610 | * Return true if this page is mapped into pagetables. | |
611 | */ | |
612 | static inline int page_mapped(struct page *page) | |
613 | { | |
614 | return atomic_read(&(page)->_mapcount) >= 0; | |
615 | } | |
616 | ||
617 | /* | |
618 | * Error return values for the *_nopage functions | |
619 | */ | |
620 | #define NOPAGE_SIGBUS (NULL) | |
621 | #define NOPAGE_OOM ((struct page *) (-1)) | |
622 | ||
623 | /* | |
624 | * Different kinds of faults, as returned by handle_mm_fault(). | |
625 | * Used to decide whether a process gets delivered SIGBUS or | |
626 | * just gets major/minor fault counters bumped up. | |
627 | */ | |
628 | #define VM_FAULT_OOM (-1) | |
629 | #define VM_FAULT_SIGBUS 0 | |
630 | #define VM_FAULT_MINOR 1 | |
631 | #define VM_FAULT_MAJOR 2 | |
632 | ||
633 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | |
634 | ||
635 | extern void show_free_areas(void); | |
636 | ||
637 | #ifdef CONFIG_SHMEM | |
638 | struct page *shmem_nopage(struct vm_area_struct *vma, | |
639 | unsigned long address, int *type); | |
640 | int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new); | |
641 | struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, | |
642 | unsigned long addr); | |
643 | int shmem_lock(struct file *file, int lock, struct user_struct *user); | |
644 | #else | |
645 | #define shmem_nopage filemap_nopage | |
646 | #define shmem_lock(a, b, c) ({0;}) /* always in memory, no need to lock */ | |
647 | #define shmem_set_policy(a, b) (0) | |
648 | #define shmem_get_policy(a, b) (NULL) | |
649 | #endif | |
650 | struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); | |
651 | ||
652 | int shmem_zero_setup(struct vm_area_struct *); | |
653 | ||
654 | static inline int can_do_mlock(void) | |
655 | { | |
656 | if (capable(CAP_IPC_LOCK)) | |
657 | return 1; | |
658 | if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0) | |
659 | return 1; | |
660 | return 0; | |
661 | } | |
662 | extern int user_shm_lock(size_t, struct user_struct *); | |
663 | extern void user_shm_unlock(size_t, struct user_struct *); | |
664 | ||
665 | /* | |
666 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
667 | */ | |
668 | struct zap_details { | |
669 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | |
670 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
671 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
672 | pgoff_t last_index; /* Highest page->index to unmap */ | |
673 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ | |
1da177e4 LT |
674 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
675 | }; | |
676 | ||
ee39b37b | 677 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
1da177e4 | 678 | unsigned long size, struct zap_details *); |
ee39b37b | 679 | unsigned long unmap_vmas(struct mmu_gather **tlb, struct mm_struct *mm, |
1da177e4 LT |
680 | struct vm_area_struct *start_vma, unsigned long start_addr, |
681 | unsigned long end_addr, unsigned long *nr_accounted, | |
682 | struct zap_details *); | |
3bf5ee95 HD |
683 | void free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
684 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
685 | void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma, | |
e0da382c | 686 | unsigned long floor, unsigned long ceiling); |
1da177e4 LT |
687 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
688 | struct vm_area_struct *vma); | |
689 | int zeromap_page_range(struct vm_area_struct *vma, unsigned long from, | |
690 | unsigned long size, pgprot_t prot); | |
691 | void unmap_mapping_range(struct address_space *mapping, | |
692 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
693 | ||
694 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
695 | loff_t const holebegin, loff_t const holelen) | |
696 | { | |
697 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
698 | } | |
699 | ||
700 | extern int vmtruncate(struct inode * inode, loff_t offset); | |
701 | extern pud_t *FASTCALL(__pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)); | |
702 | extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)); | |
703 | extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); | |
704 | extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); | |
705 | extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); | |
706 | extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot); | |
707 | extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access); | |
708 | extern int make_pages_present(unsigned long addr, unsigned long end); | |
709 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
710 | void install_arg_page(struct vm_area_struct *, struct page *, unsigned long); | |
711 | ||
712 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, | |
713 | int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); | |
714 | ||
715 | int __set_page_dirty_buffers(struct page *page); | |
716 | int __set_page_dirty_nobuffers(struct page *page); | |
717 | int redirty_page_for_writepage(struct writeback_control *wbc, | |
718 | struct page *page); | |
719 | int FASTCALL(set_page_dirty(struct page *page)); | |
720 | int set_page_dirty_lock(struct page *page); | |
721 | int clear_page_dirty_for_io(struct page *page); | |
722 | ||
723 | extern unsigned long do_mremap(unsigned long addr, | |
724 | unsigned long old_len, unsigned long new_len, | |
725 | unsigned long flags, unsigned long new_addr); | |
726 | ||
727 | /* | |
728 | * Prototype to add a shrinker callback for ageable caches. | |
729 | * | |
730 | * These functions are passed a count `nr_to_scan' and a gfpmask. They should | |
731 | * scan `nr_to_scan' objects, attempting to free them. | |
732 | * | |
845d3431 | 733 | * The callback must return the number of objects which remain in the cache. |
1da177e4 | 734 | * |
845d3431 | 735 | * The callback will be passed nr_to_scan == 0 when the VM is querying the |
1da177e4 LT |
736 | * cache size, so a fastpath for that case is appropriate. |
737 | */ | |
738 | typedef int (*shrinker_t)(int nr_to_scan, unsigned int gfp_mask); | |
739 | ||
740 | /* | |
741 | * Add an aging callback. The int is the number of 'seeks' it takes | |
742 | * to recreate one of the objects that these functions age. | |
743 | */ | |
744 | ||
745 | #define DEFAULT_SEEKS 2 | |
746 | struct shrinker; | |
747 | extern struct shrinker *set_shrinker(int, shrinker_t); | |
748 | extern void remove_shrinker(struct shrinker *shrinker); | |
749 | ||
750 | /* | |
751 | * On a two-level or three-level page table, this ends up being trivial. Thus | |
752 | * the inlining and the symmetry break with pte_alloc_map() that does all | |
753 | * of this out-of-line. | |
754 | */ | |
755 | /* | |
756 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
757 | * Remove it when 4level-fixup.h has been removed. | |
758 | */ | |
759 | #ifdef CONFIG_MMU | |
760 | #ifndef __ARCH_HAS_4LEVEL_HACK | |
761 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) | |
762 | { | |
763 | if (pgd_none(*pgd)) | |
764 | return __pud_alloc(mm, pgd, address); | |
765 | return pud_offset(pgd, address); | |
766 | } | |
767 | ||
768 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
769 | { | |
770 | if (pud_none(*pud)) | |
771 | return __pmd_alloc(mm, pud, address); | |
772 | return pmd_offset(pud, address); | |
773 | } | |
774 | #endif | |
775 | #endif /* CONFIG_MMU */ | |
776 | ||
777 | extern void free_area_init(unsigned long * zones_size); | |
778 | extern void free_area_init_node(int nid, pg_data_t *pgdat, | |
779 | unsigned long * zones_size, unsigned long zone_start_pfn, | |
780 | unsigned long *zholes_size); | |
781 | extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long); | |
782 | extern void mem_init(void); | |
783 | extern void show_mem(void); | |
784 | extern void si_meminfo(struct sysinfo * val); | |
785 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
786 | ||
e7c8d5c9 CL |
787 | #ifdef CONFIG_NUMA |
788 | extern void setup_per_cpu_pageset(void); | |
789 | #else | |
790 | static inline void setup_per_cpu_pageset(void) {} | |
791 | #endif | |
792 | ||
1da177e4 LT |
793 | /* prio_tree.c */ |
794 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); | |
795 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); | |
796 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); | |
797 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, | |
798 | struct prio_tree_iter *iter); | |
799 | ||
800 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ | |
801 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ | |
802 | (vma = vma_prio_tree_next(vma, iter)); ) | |
803 | ||
804 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
805 | struct list_head *list) | |
806 | { | |
807 | vma->shared.vm_set.parent = NULL; | |
808 | list_add_tail(&vma->shared.vm_set.list, list); | |
809 | } | |
810 | ||
811 | /* mmap.c */ | |
812 | extern int __vm_enough_memory(long pages, int cap_sys_admin); | |
813 | extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
814 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | |
815 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
816 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
817 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
818 | struct mempolicy *); | |
819 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
820 | extern int split_vma(struct mm_struct *, | |
821 | struct vm_area_struct *, unsigned long addr, int new_below); | |
822 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
823 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
824 | struct rb_node **, struct rb_node *); | |
825 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, | |
826 | unsigned long addr, unsigned long len, pgoff_t pgoff); | |
827 | extern void exit_mmap(struct mm_struct *); | |
119f657c | 828 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
1da177e4 LT |
829 | |
830 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
831 | ||
832 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
833 | unsigned long len, unsigned long prot, | |
834 | unsigned long flag, unsigned long pgoff); | |
835 | ||
836 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, | |
837 | unsigned long len, unsigned long prot, | |
838 | unsigned long flag, unsigned long offset) | |
839 | { | |
840 | unsigned long ret = -EINVAL; | |
841 | if ((offset + PAGE_ALIGN(len)) < offset) | |
842 | goto out; | |
843 | if (!(offset & ~PAGE_MASK)) | |
844 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); | |
845 | out: | |
846 | return ret; | |
847 | } | |
848 | ||
849 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
850 | ||
851 | extern unsigned long do_brk(unsigned long, unsigned long); | |
852 | ||
853 | /* filemap.c */ | |
854 | extern unsigned long page_unuse(struct page *); | |
855 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
856 | ||
857 | /* generic vm_area_ops exported for stackable file systems */ | |
858 | extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *); | |
859 | extern int filemap_populate(struct vm_area_struct *, unsigned long, | |
860 | unsigned long, pgprot_t, unsigned long, int); | |
861 | ||
862 | /* mm/page-writeback.c */ | |
863 | int write_one_page(struct page *page, int wait); | |
864 | ||
865 | /* readahead.c */ | |
866 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
867 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
868 | #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before | |
869 | * turning readahead off */ | |
870 | ||
871 | int do_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
872 | unsigned long offset, unsigned long nr_to_read); | |
873 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
874 | unsigned long offset, unsigned long nr_to_read); | |
875 | unsigned long page_cache_readahead(struct address_space *mapping, | |
876 | struct file_ra_state *ra, | |
877 | struct file *filp, | |
878 | unsigned long offset, | |
879 | unsigned long size); | |
880 | void handle_ra_miss(struct address_space *mapping, | |
881 | struct file_ra_state *ra, pgoff_t offset); | |
882 | unsigned long max_sane_readahead(unsigned long nr); | |
883 | ||
884 | /* Do stack extension */ | |
885 | extern int expand_stack(struct vm_area_struct * vma, unsigned long address); | |
886 | ||
887 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
888 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
889 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
890 | struct vm_area_struct **pprev); | |
891 | ||
892 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
893 | NULL if none. Assume start_addr < end_addr. */ | |
894 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
895 | { | |
896 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
897 | ||
898 | if (vma && end_addr <= vma->vm_start) | |
899 | vma = NULL; | |
900 | return vma; | |
901 | } | |
902 | ||
903 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
904 | { | |
905 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
906 | } | |
907 | ||
908 | extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr); | |
909 | ||
910 | extern struct page * vmalloc_to_page(void *addr); | |
911 | extern unsigned long vmalloc_to_pfn(void *addr); | |
912 | extern struct page * follow_page(struct mm_struct *mm, unsigned long address, | |
913 | int write); | |
914 | extern int check_user_page_readable(struct mm_struct *mm, unsigned long address); | |
915 | int remap_pfn_range(struct vm_area_struct *, unsigned long, | |
916 | unsigned long, unsigned long, pgprot_t); | |
917 | ||
918 | #ifdef CONFIG_PROC_FS | |
919 | void __vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); | |
920 | #else | |
921 | static inline void __vm_stat_account(struct mm_struct *mm, | |
922 | unsigned long flags, struct file *file, long pages) | |
923 | { | |
924 | } | |
925 | #endif /* CONFIG_PROC_FS */ | |
926 | ||
927 | static inline void vm_stat_account(struct vm_area_struct *vma) | |
928 | { | |
929 | __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file, | |
930 | vma_pages(vma)); | |
931 | } | |
932 | ||
933 | static inline void vm_stat_unaccount(struct vm_area_struct *vma) | |
934 | { | |
935 | __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file, | |
936 | -vma_pages(vma)); | |
937 | } | |
938 | ||
939 | /* update per process rss and vm hiwater data */ | |
940 | extern void update_mem_hiwater(struct task_struct *tsk); | |
941 | ||
942 | #ifndef CONFIG_DEBUG_PAGEALLOC | |
943 | static inline void | |
944 | kernel_map_pages(struct page *page, int numpages, int enable) | |
945 | { | |
946 | } | |
947 | #endif | |
948 | ||
949 | extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); | |
950 | #ifdef __HAVE_ARCH_GATE_AREA | |
951 | int in_gate_area_no_task(unsigned long addr); | |
952 | int in_gate_area(struct task_struct *task, unsigned long addr); | |
953 | #else | |
954 | int in_gate_area_no_task(unsigned long addr); | |
955 | #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) | |
956 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
957 | ||
79befd0c AA |
958 | /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */ |
959 | #define OOM_DISABLE -17 | |
960 | ||
1da177e4 LT |
961 | #endif /* __KERNEL__ */ |
962 | #endif /* _LINUX_MM_H */ |