]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _LINUX_MM_H |
2 | #define _LINUX_MM_H | |
3 | ||
1da177e4 LT |
4 | #include <linux/errno.h> |
5 | ||
6 | #ifdef __KERNEL__ | |
7 | ||
1da177e4 LT |
8 | #include <linux/gfp.h> |
9 | #include <linux/list.h> | |
10 | #include <linux/mmzone.h> | |
11 | #include <linux/rbtree.h> | |
12 | #include <linux/prio_tree.h> | |
13 | #include <linux/fs.h> | |
de5097c2 | 14 | #include <linux/mutex.h> |
9a11b49a | 15 | #include <linux/debug_locks.h> |
d08b3851 | 16 | #include <linux/backing-dev.h> |
5b99cd0e | 17 | #include <linux/mm_types.h> |
1da177e4 LT |
18 | |
19 | struct mempolicy; | |
20 | struct anon_vma; | |
e8edc6e0 | 21 | struct user_struct; |
1da177e4 LT |
22 | |
23 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ | |
24 | extern unsigned long max_mapnr; | |
25 | #endif | |
26 | ||
27 | extern unsigned long num_physpages; | |
28 | extern void * high_memory; | |
1da177e4 LT |
29 | extern int page_cluster; |
30 | ||
31 | #ifdef CONFIG_SYSCTL | |
32 | extern 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> | |
1da177e4 | 40 | |
1da177e4 LT |
41 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
42 | ||
43 | /* | |
44 | * Linux kernel virtual memory manager primitives. | |
45 | * The idea being to have a "virtual" mm in the same way | |
46 | * we have a virtual fs - giving a cleaner interface to the | |
47 | * mm details, and allowing different kinds of memory mappings | |
48 | * (from shared memory to executable loading to arbitrary | |
49 | * mmap() functions). | |
50 | */ | |
51 | ||
52 | /* | |
53 | * This struct defines a memory VMM memory area. There is one of these | |
54 | * per VM-area/task. A VM area is any part of the process virtual memory | |
55 | * space that has a special rule for the page-fault handlers (ie a shared | |
56 | * library, the executable area etc). | |
57 | */ | |
58 | struct vm_area_struct { | |
59 | struct mm_struct * vm_mm; /* The address space we belong to. */ | |
60 | unsigned long vm_start; /* Our start address within vm_mm. */ | |
61 | unsigned long vm_end; /* The first byte after our end address | |
62 | within vm_mm. */ | |
63 | ||
64 | /* linked list of VM areas per task, sorted by address */ | |
65 | struct vm_area_struct *vm_next; | |
66 | ||
67 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ | |
68 | unsigned long vm_flags; /* Flags, listed below. */ | |
69 | ||
70 | struct rb_node vm_rb; | |
71 | ||
72 | /* | |
73 | * For areas with an address space and backing store, | |
74 | * linkage into the address_space->i_mmap prio tree, or | |
75 | * linkage to the list of like vmas hanging off its node, or | |
76 | * linkage of vma in the address_space->i_mmap_nonlinear list. | |
77 | */ | |
78 | union { | |
79 | struct { | |
80 | struct list_head list; | |
81 | void *parent; /* aligns with prio_tree_node parent */ | |
82 | struct vm_area_struct *head; | |
83 | } vm_set; | |
84 | ||
85 | struct raw_prio_tree_node prio_tree_node; | |
86 | } shared; | |
87 | ||
88 | /* | |
89 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
90 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
91 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
92 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
93 | */ | |
94 | struct list_head anon_vma_node; /* Serialized by anon_vma->lock */ | |
95 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ | |
96 | ||
97 | /* Function pointers to deal with this struct. */ | |
98 | struct vm_operations_struct * vm_ops; | |
99 | ||
100 | /* Information about our backing store: */ | |
101 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
102 | units, *not* PAGE_CACHE_SIZE */ | |
103 | struct file * vm_file; /* File we map to (can be NULL). */ | |
104 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
105 | unsigned long vm_truncate_count;/* truncate_count or restart_addr */ | |
106 | ||
107 | #ifndef CONFIG_MMU | |
108 | atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */ | |
109 | #endif | |
110 | #ifdef CONFIG_NUMA | |
111 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
112 | #endif | |
113 | }; | |
114 | ||
c43692e8 CL |
115 | extern struct kmem_cache *vm_area_cachep; |
116 | ||
1da177e4 LT |
117 | /* |
118 | * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is | |
119 | * disabled, then there's a single shared list of VMAs maintained by the | |
120 | * system, and mm's subscribe to these individually | |
121 | */ | |
122 | struct vm_list_struct { | |
123 | struct vm_list_struct *next; | |
124 | struct vm_area_struct *vma; | |
125 | }; | |
126 | ||
127 | #ifndef CONFIG_MMU | |
128 | extern struct rb_root nommu_vma_tree; | |
129 | extern struct rw_semaphore nommu_vma_sem; | |
130 | ||
131 | extern unsigned int kobjsize(const void *objp); | |
132 | #endif | |
133 | ||
134 | /* | |
135 | * vm_flags.. | |
136 | */ | |
137 | #define VM_READ 0x00000001 /* currently active flags */ | |
138 | #define VM_WRITE 0x00000002 | |
139 | #define VM_EXEC 0x00000004 | |
140 | #define VM_SHARED 0x00000008 | |
141 | ||
7e2cff42 | 142 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
1da177e4 LT |
143 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
144 | #define VM_MAYWRITE 0x00000020 | |
145 | #define VM_MAYEXEC 0x00000040 | |
146 | #define VM_MAYSHARE 0x00000080 | |
147 | ||
148 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
149 | #define VM_GROWSUP 0x00000200 | |
6aab341e | 150 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
1da177e4 LT |
151 | #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() */ | |
0b14c179 | 163 | #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ |
1da177e4 LT |
164 | #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) */ | |
4d7672b4 | 168 | #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ |
e5b97dde | 169 | #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ |
d00806b1 NP |
170 | |
171 | #define VM_CAN_INVALIDATE 0x08000000 /* The mapping may be invalidated, | |
172 | * eg. truncate or invalidate_inode_*. | |
173 | * In this case, do_no_page must | |
174 | * return with the page locked. | |
175 | */ | |
1da177e4 LT |
176 | |
177 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | |
178 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
179 | #endif | |
180 | ||
181 | #ifdef CONFIG_STACK_GROWSUP | |
182 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
183 | #else | |
184 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
185 | #endif | |
186 | ||
187 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) | |
188 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK | |
189 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) | |
190 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) | |
191 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) | |
192 | ||
193 | /* | |
194 | * mapping from the currently active vm_flags protection bits (the | |
195 | * low four bits) to a page protection mask.. | |
196 | */ | |
197 | extern pgprot_t protection_map[16]; | |
198 | ||
199 | ||
200 | /* | |
201 | * These are the virtual MM functions - opening of an area, closing and | |
202 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
203 | * to the functions called when a no-page or a wp-page exception occurs. | |
204 | */ | |
205 | struct vm_operations_struct { | |
206 | void (*open)(struct vm_area_struct * area); | |
207 | void (*close)(struct vm_area_struct * area); | |
208 | struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type); | |
f4b81804 | 209 | unsigned long (*nopfn)(struct vm_area_struct * area, unsigned long address); |
1da177e4 | 210 | int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock); |
9637a5ef DH |
211 | |
212 | /* notification that a previously read-only page is about to become | |
213 | * writable, if an error is returned it will cause a SIGBUS */ | |
214 | int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page); | |
1da177e4 LT |
215 | #ifdef CONFIG_NUMA |
216 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | |
217 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | |
218 | unsigned long addr); | |
7b2259b3 CL |
219 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
220 | const nodemask_t *to, unsigned long flags); | |
1da177e4 LT |
221 | #endif |
222 | }; | |
223 | ||
224 | struct mmu_gather; | |
225 | struct inode; | |
226 | ||
349aef0b AM |
227 | #define page_private(page) ((page)->private) |
228 | #define set_page_private(page, v) ((page)->private = (v)) | |
4c21e2f2 | 229 | |
1da177e4 LT |
230 | /* |
231 | * FIXME: take this include out, include page-flags.h in | |
232 | * files which need it (119 of them) | |
233 | */ | |
234 | #include <linux/page-flags.h> | |
235 | ||
725d704e NP |
236 | #ifdef CONFIG_DEBUG_VM |
237 | #define VM_BUG_ON(cond) BUG_ON(cond) | |
238 | #else | |
239 | #define VM_BUG_ON(condition) do { } while(0) | |
240 | #endif | |
241 | ||
1da177e4 LT |
242 | /* |
243 | * Methods to modify the page usage count. | |
244 | * | |
245 | * What counts for a page usage: | |
246 | * - cache mapping (page->mapping) | |
247 | * - private data (page->private) | |
248 | * - page mapped in a task's page tables, each mapping | |
249 | * is counted separately | |
250 | * | |
251 | * Also, many kernel routines increase the page count before a critical | |
252 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
253 | */ |
254 | ||
255 | /* | |
da6052f7 | 256 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 257 | */ |
7c8ee9a8 NP |
258 | static inline int put_page_testzero(struct page *page) |
259 | { | |
725d704e | 260 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
8dc04efb | 261 | return atomic_dec_and_test(&page->_count); |
7c8ee9a8 | 262 | } |
1da177e4 LT |
263 | |
264 | /* | |
7c8ee9a8 NP |
265 | * Try to grab a ref unless the page has a refcount of zero, return false if |
266 | * that is the case. | |
1da177e4 | 267 | */ |
7c8ee9a8 NP |
268 | static inline int get_page_unless_zero(struct page *page) |
269 | { | |
725d704e | 270 | VM_BUG_ON(PageCompound(page)); |
8dc04efb | 271 | return atomic_inc_not_zero(&page->_count); |
7c8ee9a8 | 272 | } |
1da177e4 | 273 | |
d85f3385 CL |
274 | static inline struct page *compound_head(struct page *page) |
275 | { | |
6d777953 | 276 | if (unlikely(PageTail(page))) |
d85f3385 CL |
277 | return page->first_page; |
278 | return page; | |
279 | } | |
280 | ||
4c21e2f2 | 281 | static inline int page_count(struct page *page) |
1da177e4 | 282 | { |
d85f3385 | 283 | return atomic_read(&compound_head(page)->_count); |
1da177e4 LT |
284 | } |
285 | ||
286 | static inline void get_page(struct page *page) | |
287 | { | |
d85f3385 | 288 | page = compound_head(page); |
725d704e | 289 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
1da177e4 LT |
290 | atomic_inc(&page->_count); |
291 | } | |
292 | ||
b49af68f CL |
293 | static inline struct page *virt_to_head_page(const void *x) |
294 | { | |
295 | struct page *page = virt_to_page(x); | |
296 | return compound_head(page); | |
297 | } | |
298 | ||
7835e98b NP |
299 | /* |
300 | * Setup the page count before being freed into the page allocator for | |
301 | * the first time (boot or memory hotplug) | |
302 | */ | |
303 | static inline void init_page_count(struct page *page) | |
304 | { | |
305 | atomic_set(&page->_count, 1); | |
306 | } | |
307 | ||
1da177e4 | 308 | void put_page(struct page *page); |
1d7ea732 | 309 | void put_pages_list(struct list_head *pages); |
1da177e4 | 310 | |
8dfcc9ba | 311 | void split_page(struct page *page, unsigned int order); |
8dfcc9ba | 312 | |
33f2ef89 AW |
313 | /* |
314 | * Compound pages have a destructor function. Provide a | |
315 | * prototype for that function and accessor functions. | |
316 | * These are _only_ valid on the head of a PG_compound page. | |
317 | */ | |
318 | typedef void compound_page_dtor(struct page *); | |
319 | ||
320 | static inline void set_compound_page_dtor(struct page *page, | |
321 | compound_page_dtor *dtor) | |
322 | { | |
323 | page[1].lru.next = (void *)dtor; | |
324 | } | |
325 | ||
326 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
327 | { | |
328 | return (compound_page_dtor *)page[1].lru.next; | |
329 | } | |
330 | ||
d85f3385 CL |
331 | static inline int compound_order(struct page *page) |
332 | { | |
6d777953 | 333 | if (!PageHead(page)) |
d85f3385 CL |
334 | return 0; |
335 | return (unsigned long)page[1].lru.prev; | |
336 | } | |
337 | ||
338 | static inline void set_compound_order(struct page *page, unsigned long order) | |
339 | { | |
340 | page[1].lru.prev = (void *)order; | |
341 | } | |
342 | ||
1da177e4 LT |
343 | /* |
344 | * Multiple processes may "see" the same page. E.g. for untouched | |
345 | * mappings of /dev/null, all processes see the same page full of | |
346 | * zeroes, and text pages of executables and shared libraries have | |
347 | * only one copy in memory, at most, normally. | |
348 | * | |
349 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
350 | * page_count() == 0 means the page is free. page->lru is then used for |
351 | * freelist management in the buddy allocator. | |
da6052f7 | 352 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 353 | * |
da6052f7 NP |
354 | * Pages are allocated by the slab allocator in order to provide memory |
355 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
356 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
357 | * unless a particular usage is carefully commented. (the responsibility of | |
358 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 359 | * |
da6052f7 NP |
360 | * A page may be used by anyone else who does a __get_free_page(). |
361 | * In this case, page_count still tracks the references, and should only | |
362 | * be used through the normal accessor functions. The top bits of page->flags | |
363 | * and page->virtual store page management information, but all other fields | |
364 | * are unused and could be used privately, carefully. The management of this | |
365 | * page is the responsibility of the one who allocated it, and those who have | |
366 | * subsequently been given references to it. | |
367 | * | |
368 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
369 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
370 | * The following discussion applies only to them. | |
371 | * | |
da6052f7 NP |
372 | * A pagecache page contains an opaque `private' member, which belongs to the |
373 | * page's address_space. Usually, this is the address of a circular list of | |
374 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
375 | * into the filesystem to release these pages. | |
1da177e4 | 376 | * |
da6052f7 NP |
377 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
378 | * is the pointer to the inode, and page->index is the file offset of the page, | |
379 | * in units of PAGE_CACHE_SIZE. | |
1da177e4 | 380 | * |
da6052f7 NP |
381 | * If pagecache pages are not associated with an inode, they are said to be |
382 | * anonymous pages. These may become associated with the swapcache, and in that | |
383 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 384 | * |
da6052f7 NP |
385 | * In either case (swapcache or inode backed), the pagecache itself holds one |
386 | * reference to the page. Setting PG_private should also increment the | |
387 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 388 | * |
da6052f7 NP |
389 | * The pagecache pages are stored in a per-mapping radix tree, which is |
390 | * rooted at mapping->page_tree, and indexed by offset. | |
391 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
392 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 393 | * |
da6052f7 | 394 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
395 | * - inode pages may need to be read from disk, |
396 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
397 | * to be written back to the inode on disk, |
398 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
399 | * modified may need to be swapped out to swap space and (later) to be read | |
400 | * back into memory. | |
1da177e4 LT |
401 | */ |
402 | ||
403 | /* | |
404 | * The zone field is never updated after free_area_init_core() | |
405 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 406 | */ |
348f8b6c | 407 | |
d41dee36 AW |
408 | |
409 | /* | |
410 | * page->flags layout: | |
411 | * | |
412 | * There are three possibilities for how page->flags get | |
413 | * laid out. The first is for the normal case, without | |
414 | * sparsemem. The second is for sparsemem when there is | |
415 | * plenty of space for node and section. The last is when | |
416 | * we have run out of space and have to fall back to an | |
417 | * alternate (slower) way of determining the node. | |
418 | * | |
419 | * No sparsemem: | NODE | ZONE | ... | FLAGS | | |
420 | * with space for node: | SECTION | NODE | ZONE | ... | FLAGS | | |
421 | * no space for node: | SECTION | ZONE | ... | FLAGS | | |
422 | */ | |
423 | #ifdef CONFIG_SPARSEMEM | |
424 | #define SECTIONS_WIDTH SECTIONS_SHIFT | |
425 | #else | |
426 | #define SECTIONS_WIDTH 0 | |
427 | #endif | |
428 | ||
429 | #define ZONES_WIDTH ZONES_SHIFT | |
430 | ||
431 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED | |
432 | #define NODES_WIDTH NODES_SHIFT | |
433 | #else | |
434 | #define NODES_WIDTH 0 | |
435 | #endif | |
436 | ||
437 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ | |
07808b74 | 438 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
d41dee36 AW |
439 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
440 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
441 | ||
442 | /* | |
443 | * We are going to use the flags for the page to node mapping if its in | |
444 | * there. This includes the case where there is no node, so it is implicit. | |
445 | */ | |
89689ae7 CL |
446 | #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) |
447 | #define NODE_NOT_IN_PAGE_FLAGS | |
448 | #endif | |
d41dee36 AW |
449 | |
450 | #ifndef PFN_SECTION_SHIFT | |
451 | #define PFN_SECTION_SHIFT 0 | |
452 | #endif | |
348f8b6c DH |
453 | |
454 | /* | |
455 | * Define the bit shifts to access each section. For non-existant | |
456 | * sections we define the shift as 0; that plus a 0 mask ensures | |
457 | * the compiler will optimise away reference to them. | |
458 | */ | |
d41dee36 AW |
459 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
460 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
461 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
348f8b6c | 462 | |
89689ae7 CL |
463 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ |
464 | #ifdef NODE_NOT_IN_PAGEFLAGS | |
465 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
bd8029b6 AW |
466 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
467 | SECTIONS_PGOFF : ZONES_PGOFF) | |
d41dee36 | 468 | #else |
89689ae7 | 469 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
470 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
471 | NODES_PGOFF : ZONES_PGOFF) | |
89689ae7 CL |
472 | #endif |
473 | ||
bd8029b6 | 474 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
348f8b6c | 475 | |
d41dee36 AW |
476 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED |
477 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED | |
348f8b6c DH |
478 | #endif |
479 | ||
d41dee36 AW |
480 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
481 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
482 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
89689ae7 | 483 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
348f8b6c | 484 | |
2f1b6248 | 485 | static inline enum zone_type page_zonenum(struct page *page) |
1da177e4 | 486 | { |
348f8b6c | 487 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 488 | } |
1da177e4 | 489 | |
89689ae7 CL |
490 | /* |
491 | * The identification function is only used by the buddy allocator for | |
492 | * determining if two pages could be buddies. We are not really | |
493 | * identifying a zone since we could be using a the section number | |
494 | * id if we have not node id available in page flags. | |
495 | * We guarantee only that it will return the same value for two | |
496 | * combinable pages in a zone. | |
497 | */ | |
cb2b95e1 AW |
498 | static inline int page_zone_id(struct page *page) |
499 | { | |
89689ae7 | 500 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
501 | } |
502 | ||
25ba77c1 | 503 | static inline int zone_to_nid(struct zone *zone) |
89fa3024 | 504 | { |
d5f541ed CL |
505 | #ifdef CONFIG_NUMA |
506 | return zone->node; | |
507 | #else | |
508 | return 0; | |
509 | #endif | |
89fa3024 CL |
510 | } |
511 | ||
89689ae7 | 512 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
25ba77c1 | 513 | extern int page_to_nid(struct page *page); |
89689ae7 | 514 | #else |
25ba77c1 | 515 | static inline int page_to_nid(struct page *page) |
d41dee36 | 516 | { |
89689ae7 | 517 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
d41dee36 | 518 | } |
89689ae7 CL |
519 | #endif |
520 | ||
521 | static inline struct zone *page_zone(struct page *page) | |
522 | { | |
523 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
524 | } | |
525 | ||
d41dee36 AW |
526 | static inline unsigned long page_to_section(struct page *page) |
527 | { | |
528 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
529 | } | |
530 | ||
2f1b6248 | 531 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
532 | { |
533 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
534 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
535 | } | |
2f1b6248 | 536 | |
348f8b6c DH |
537 | static inline void set_page_node(struct page *page, unsigned long node) |
538 | { | |
539 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
540 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 541 | } |
89689ae7 | 542 | |
d41dee36 AW |
543 | static inline void set_page_section(struct page *page, unsigned long section) |
544 | { | |
545 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
546 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
547 | } | |
1da177e4 | 548 | |
2f1b6248 | 549 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 550 | unsigned long node, unsigned long pfn) |
1da177e4 | 551 | { |
348f8b6c DH |
552 | set_page_zone(page, zone); |
553 | set_page_node(page, node); | |
d41dee36 | 554 | set_page_section(page, pfn_to_section_nr(pfn)); |
1da177e4 LT |
555 | } |
556 | ||
f6ac2354 CL |
557 | /* |
558 | * Some inline functions in vmstat.h depend on page_zone() | |
559 | */ | |
560 | #include <linux/vmstat.h> | |
561 | ||
652050ae | 562 | static __always_inline void *lowmem_page_address(struct page *page) |
1da177e4 LT |
563 | { |
564 | return __va(page_to_pfn(page) << PAGE_SHIFT); | |
565 | } | |
566 | ||
567 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
568 | #define HASHED_PAGE_VIRTUAL | |
569 | #endif | |
570 | ||
571 | #if defined(WANT_PAGE_VIRTUAL) | |
572 | #define page_address(page) ((page)->virtual) | |
573 | #define set_page_address(page, address) \ | |
574 | do { \ | |
575 | (page)->virtual = (address); \ | |
576 | } while(0) | |
577 | #define page_address_init() do { } while(0) | |
578 | #endif | |
579 | ||
580 | #if defined(HASHED_PAGE_VIRTUAL) | |
581 | void *page_address(struct page *page); | |
582 | void set_page_address(struct page *page, void *virtual); | |
583 | void page_address_init(void); | |
584 | #endif | |
585 | ||
586 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
587 | #define page_address(page) lowmem_page_address(page) | |
588 | #define set_page_address(page, address) do { } while(0) | |
589 | #define page_address_init() do { } while(0) | |
590 | #endif | |
591 | ||
592 | /* | |
593 | * On an anonymous page mapped into a user virtual memory area, | |
594 | * page->mapping points to its anon_vma, not to a struct address_space; | |
595 | * with the PAGE_MAPPING_ANON bit set to distinguish it. | |
596 | * | |
597 | * Please note that, confusingly, "page_mapping" refers to the inode | |
598 | * address_space which maps the page from disk; whereas "page_mapped" | |
599 | * refers to user virtual address space into which the page is mapped. | |
600 | */ | |
601 | #define PAGE_MAPPING_ANON 1 | |
602 | ||
603 | extern struct address_space swapper_space; | |
604 | static inline struct address_space *page_mapping(struct page *page) | |
605 | { | |
606 | struct address_space *mapping = page->mapping; | |
607 | ||
b5fab14e | 608 | VM_BUG_ON(PageSlab(page)); |
1da177e4 LT |
609 | if (unlikely(PageSwapCache(page))) |
610 | mapping = &swapper_space; | |
b9bae340 HD |
611 | #ifdef CONFIG_SLUB |
612 | else if (unlikely(PageSlab(page))) | |
613 | mapping = NULL; | |
614 | #endif | |
1da177e4 LT |
615 | else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) |
616 | mapping = NULL; | |
617 | return mapping; | |
618 | } | |
619 | ||
620 | static inline int PageAnon(struct page *page) | |
621 | { | |
622 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
623 | } | |
624 | ||
625 | /* | |
626 | * Return the pagecache index of the passed page. Regular pagecache pages | |
627 | * use ->index whereas swapcache pages use ->private | |
628 | */ | |
629 | static inline pgoff_t page_index(struct page *page) | |
630 | { | |
631 | if (unlikely(PageSwapCache(page))) | |
4c21e2f2 | 632 | return page_private(page); |
1da177e4 LT |
633 | return page->index; |
634 | } | |
635 | ||
636 | /* | |
637 | * The atomic page->_mapcount, like _count, starts from -1: | |
638 | * so that transitions both from it and to it can be tracked, | |
639 | * using atomic_inc_and_test and atomic_add_negative(-1). | |
640 | */ | |
641 | static inline void reset_page_mapcount(struct page *page) | |
642 | { | |
643 | atomic_set(&(page)->_mapcount, -1); | |
644 | } | |
645 | ||
646 | static inline int page_mapcount(struct page *page) | |
647 | { | |
648 | return atomic_read(&(page)->_mapcount) + 1; | |
649 | } | |
650 | ||
651 | /* | |
652 | * Return true if this page is mapped into pagetables. | |
653 | */ | |
654 | static inline int page_mapped(struct page *page) | |
655 | { | |
656 | return atomic_read(&(page)->_mapcount) >= 0; | |
657 | } | |
658 | ||
659 | /* | |
660 | * Error return values for the *_nopage functions | |
661 | */ | |
662 | #define NOPAGE_SIGBUS (NULL) | |
663 | #define NOPAGE_OOM ((struct page *) (-1)) | |
7f7bbbe5 | 664 | #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */ |
1da177e4 | 665 | |
f4b81804 JS |
666 | /* |
667 | * Error return values for the *_nopfn functions | |
668 | */ | |
669 | #define NOPFN_SIGBUS ((unsigned long) -1) | |
670 | #define NOPFN_OOM ((unsigned long) -2) | |
22cd25ed | 671 | #define NOPFN_REFAULT ((unsigned long) -3) |
f4b81804 | 672 | |
1da177e4 LT |
673 | /* |
674 | * Different kinds of faults, as returned by handle_mm_fault(). | |
675 | * Used to decide whether a process gets delivered SIGBUS or | |
676 | * just gets major/minor fault counters bumped up. | |
677 | */ | |
f33ea7f4 NP |
678 | #define VM_FAULT_OOM 0x00 |
679 | #define VM_FAULT_SIGBUS 0x01 | |
680 | #define VM_FAULT_MINOR 0x02 | |
681 | #define VM_FAULT_MAJOR 0x03 | |
682 | ||
683 | /* | |
684 | * Special case for get_user_pages. | |
685 | * Must be in a distinct bit from the above VM_FAULT_ flags. | |
686 | */ | |
687 | #define VM_FAULT_WRITE 0x10 | |
1da177e4 LT |
688 | |
689 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | |
690 | ||
691 | extern void show_free_areas(void); | |
692 | ||
693 | #ifdef CONFIG_SHMEM | |
1da177e4 LT |
694 | int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new); |
695 | struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, | |
696 | unsigned long addr); | |
697 | int shmem_lock(struct file *file, int lock, struct user_struct *user); | |
698 | #else | |
03b00ebc RK |
699 | static inline int shmem_lock(struct file *file, int lock, |
700 | struct user_struct *user) | |
701 | { | |
702 | return 0; | |
703 | } | |
704 | ||
705 | static inline int shmem_set_policy(struct vm_area_struct *vma, | |
706 | struct mempolicy *new) | |
707 | { | |
708 | return 0; | |
709 | } | |
710 | ||
711 | static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, | |
712 | unsigned long addr) | |
713 | { | |
714 | return NULL; | |
715 | } | |
1da177e4 LT |
716 | #endif |
717 | struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); | |
718 | ||
719 | int shmem_zero_setup(struct vm_area_struct *); | |
720 | ||
b0e15190 DH |
721 | #ifndef CONFIG_MMU |
722 | extern unsigned long shmem_get_unmapped_area(struct file *file, | |
723 | unsigned long addr, | |
724 | unsigned long len, | |
725 | unsigned long pgoff, | |
726 | unsigned long flags); | |
727 | #endif | |
728 | ||
e8edc6e0 | 729 | extern int can_do_mlock(void); |
1da177e4 LT |
730 | extern int user_shm_lock(size_t, struct user_struct *); |
731 | extern void user_shm_unlock(size_t, struct user_struct *); | |
732 | ||
733 | /* | |
734 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
735 | */ | |
736 | struct zap_details { | |
737 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | |
738 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
739 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
740 | pgoff_t last_index; /* Highest page->index to unmap */ | |
741 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ | |
1da177e4 LT |
742 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
743 | }; | |
744 | ||
6aab341e | 745 | struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t); |
ee39b37b | 746 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
1da177e4 | 747 | unsigned long size, struct zap_details *); |
508034a3 | 748 | unsigned long unmap_vmas(struct mmu_gather **tlb, |
1da177e4 LT |
749 | struct vm_area_struct *start_vma, unsigned long start_addr, |
750 | unsigned long end_addr, unsigned long *nr_accounted, | |
751 | struct zap_details *); | |
3bf5ee95 HD |
752 | void free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
753 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
754 | void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma, | |
e0da382c | 755 | unsigned long floor, unsigned long ceiling); |
1da177e4 LT |
756 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
757 | struct vm_area_struct *vma); | |
758 | int zeromap_page_range(struct vm_area_struct *vma, unsigned long from, | |
759 | unsigned long size, pgprot_t prot); | |
760 | void unmap_mapping_range(struct address_space *mapping, | |
761 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
762 | ||
763 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
764 | loff_t const holebegin, loff_t const holelen) | |
765 | { | |
766 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
767 | } | |
768 | ||
769 | extern int vmtruncate(struct inode * inode, loff_t offset); | |
f6b3ec23 | 770 | extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end); |
1da177e4 LT |
771 | extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); |
772 | extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot); | |
f33ea7f4 | 773 | |
7ee1dd3f DH |
774 | #ifdef CONFIG_MMU |
775 | extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, | |
776 | unsigned long address, int write_access); | |
777 | ||
778 | static inline int handle_mm_fault(struct mm_struct *mm, | |
779 | struct vm_area_struct *vma, unsigned long address, | |
780 | int write_access) | |
f33ea7f4 | 781 | { |
7ee1dd3f DH |
782 | return __handle_mm_fault(mm, vma, address, write_access) & |
783 | (~VM_FAULT_WRITE); | |
f33ea7f4 | 784 | } |
7ee1dd3f DH |
785 | #else |
786 | static inline int handle_mm_fault(struct mm_struct *mm, | |
787 | struct vm_area_struct *vma, unsigned long address, | |
788 | int write_access) | |
789 | { | |
790 | /* should never happen if there's no MMU */ | |
791 | BUG(); | |
792 | return VM_FAULT_SIGBUS; | |
793 | } | |
794 | #endif | |
f33ea7f4 | 795 | |
1da177e4 LT |
796 | extern int make_pages_present(unsigned long addr, unsigned long end); |
797 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
798 | void install_arg_page(struct vm_area_struct *, struct page *, unsigned long); | |
799 | ||
800 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, | |
801 | int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); | |
b5810039 | 802 | void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long); |
1da177e4 | 803 | |
cf9a2ae8 DH |
804 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
805 | extern void do_invalidatepage(struct page *page, unsigned long offset); | |
806 | ||
1da177e4 | 807 | int __set_page_dirty_nobuffers(struct page *page); |
76719325 | 808 | int __set_page_dirty_no_writeback(struct page *page); |
1da177e4 LT |
809 | int redirty_page_for_writepage(struct writeback_control *wbc, |
810 | struct page *page); | |
811 | int FASTCALL(set_page_dirty(struct page *page)); | |
812 | int set_page_dirty_lock(struct page *page); | |
813 | int clear_page_dirty_for_io(struct page *page); | |
814 | ||
815 | extern unsigned long do_mremap(unsigned long addr, | |
816 | unsigned long old_len, unsigned long new_len, | |
817 | unsigned long flags, unsigned long new_addr); | |
818 | ||
819 | /* | |
8e1f936b | 820 | * A callback you can register to apply pressure to ageable caches. |
1da177e4 | 821 | * |
8e1f936b RR |
822 | * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should |
823 | * look through the least-recently-used 'nr_to_scan' entries and | |
824 | * attempt to free them up. It should return the number of objects | |
825 | * which remain in the cache. If it returns -1, it means it cannot do | |
826 | * any scanning at this time (eg. there is a risk of deadlock). | |
1da177e4 | 827 | * |
8e1f936b RR |
828 | * The 'gfpmask' refers to the allocation we are currently trying to |
829 | * fulfil. | |
830 | * | |
831 | * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is | |
832 | * querying the cache size, so a fastpath for that case is appropriate. | |
1da177e4 | 833 | */ |
8e1f936b RR |
834 | struct shrinker { |
835 | int (*shrink)(int nr_to_scan, gfp_t gfp_mask); | |
836 | int seeks; /* seeks to recreate an obj */ | |
1da177e4 | 837 | |
8e1f936b RR |
838 | /* These are for internal use */ |
839 | struct list_head list; | |
840 | long nr; /* objs pending delete */ | |
841 | }; | |
842 | #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ | |
843 | extern void register_shrinker(struct shrinker *); | |
844 | extern void unregister_shrinker(struct shrinker *); | |
1da177e4 | 845 | |
d08b3851 PZ |
846 | /* |
847 | * Some shared mappigns will want the pages marked read-only | |
848 | * to track write events. If so, we'll downgrade vm_page_prot | |
849 | * to the private version (using protection_map[] without the | |
850 | * VM_SHARED bit). | |
851 | */ | |
852 | static inline int vma_wants_writenotify(struct vm_area_struct *vma) | |
853 | { | |
854 | unsigned int vm_flags = vma->vm_flags; | |
855 | ||
856 | /* If it was private or non-writable, the write bit is already clear */ | |
857 | if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED))) | |
858 | return 0; | |
859 | ||
860 | /* The backer wishes to know when pages are first written to? */ | |
861 | if (vma->vm_ops && vma->vm_ops->page_mkwrite) | |
862 | return 1; | |
863 | ||
864 | /* The open routine did something to the protections already? */ | |
865 | if (pgprot_val(vma->vm_page_prot) != | |
866 | pgprot_val(protection_map[vm_flags & | |
867 | (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)])) | |
868 | return 0; | |
869 | ||
870 | /* Specialty mapping? */ | |
871 | if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE)) | |
872 | return 0; | |
873 | ||
874 | /* Can the mapping track the dirty pages? */ | |
875 | return vma->vm_file && vma->vm_file->f_mapping && | |
876 | mapping_cap_account_dirty(vma->vm_file->f_mapping); | |
877 | } | |
878 | ||
c9cfcddf LT |
879 | extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl)); |
880 | ||
5f22df00 NP |
881 | #ifdef __PAGETABLE_PUD_FOLDED |
882 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
883 | unsigned long address) | |
884 | { | |
885 | return 0; | |
886 | } | |
887 | #else | |
1bb3630e | 888 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
5f22df00 NP |
889 | #endif |
890 | ||
891 | #ifdef __PAGETABLE_PMD_FOLDED | |
892 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
893 | unsigned long address) | |
894 | { | |
895 | return 0; | |
896 | } | |
897 | #else | |
1bb3630e | 898 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
5f22df00 NP |
899 | #endif |
900 | ||
1bb3630e HD |
901 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); |
902 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | |
903 | ||
1da177e4 LT |
904 | /* |
905 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
906 | * Remove it when 4level-fixup.h has been removed. | |
907 | */ | |
1bb3630e | 908 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
1da177e4 LT |
909 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
910 | { | |
1bb3630e HD |
911 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
912 | NULL: pud_offset(pgd, address); | |
1da177e4 LT |
913 | } |
914 | ||
915 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
916 | { | |
1bb3630e HD |
917 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
918 | NULL: pmd_offset(pud, address); | |
1da177e4 | 919 | } |
1bb3630e HD |
920 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
921 | ||
4c21e2f2 HD |
922 | #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS |
923 | /* | |
924 | * We tuck a spinlock to guard each pagetable page into its struct page, | |
925 | * at page->private, with BUILD_BUG_ON to make sure that this will not | |
926 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). | |
927 | * When freeing, reset page->mapping so free_pages_check won't complain. | |
928 | */ | |
349aef0b | 929 | #define __pte_lockptr(page) &((page)->ptl) |
4c21e2f2 HD |
930 | #define pte_lock_init(_page) do { \ |
931 | spin_lock_init(__pte_lockptr(_page)); \ | |
932 | } while (0) | |
933 | #define pte_lock_deinit(page) ((page)->mapping = NULL) | |
934 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) | |
935 | #else | |
936 | /* | |
937 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
938 | */ | |
939 | #define pte_lock_init(page) do {} while (0) | |
940 | #define pte_lock_deinit(page) do {} while (0) | |
941 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) | |
942 | #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */ | |
943 | ||
c74df32c HD |
944 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
945 | ({ \ | |
4c21e2f2 | 946 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
947 | pte_t *__pte = pte_offset_map(pmd, address); \ |
948 | *(ptlp) = __ptl; \ | |
949 | spin_lock(__ptl); \ | |
950 | __pte; \ | |
951 | }) | |
952 | ||
953 | #define pte_unmap_unlock(pte, ptl) do { \ | |
954 | spin_unlock(ptl); \ | |
955 | pte_unmap(pte); \ | |
956 | } while (0) | |
957 | ||
1bb3630e HD |
958 | #define pte_alloc_map(mm, pmd, address) \ |
959 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ | |
960 | NULL: pte_offset_map(pmd, address)) | |
961 | ||
c74df32c HD |
962 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
963 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ | |
964 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) | |
965 | ||
1bb3630e HD |
966 | #define pte_alloc_kernel(pmd, address) \ |
967 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | |
968 | NULL: pte_offset_kernel(pmd, address)) | |
1da177e4 LT |
969 | |
970 | extern void free_area_init(unsigned long * zones_size); | |
971 | extern void free_area_init_node(int nid, pg_data_t *pgdat, | |
972 | unsigned long * zones_size, unsigned long zone_start_pfn, | |
973 | unsigned long *zholes_size); | |
c713216d MG |
974 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
975 | /* | |
976 | * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its | |
977 | * zones, allocate the backing mem_map and account for memory holes in a more | |
978 | * architecture independent manner. This is a substitute for creating the | |
979 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
980 | * free_area_init_node() | |
981 | * | |
982 | * An architecture is expected to register range of page frames backed by | |
983 | * physical memory with add_active_range() before calling | |
984 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
985 | * usage, an architecture is expected to do something like | |
986 | * | |
987 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
988 | * max_highmem_pfn}; | |
989 | * for_each_valid_physical_page_range() | |
990 | * add_active_range(node_id, start_pfn, end_pfn) | |
991 | * free_area_init_nodes(max_zone_pfns); | |
992 | * | |
993 | * If the architecture guarantees that there are no holes in the ranges | |
994 | * registered with add_active_range(), free_bootmem_active_regions() | |
995 | * will call free_bootmem_node() for each registered physical page range. | |
996 | * Similarly sparse_memory_present_with_active_regions() calls | |
997 | * memory_present() for each range when SPARSEMEM is enabled. | |
998 | * | |
999 | * See mm/page_alloc.c for more information on each function exposed by | |
1000 | * CONFIG_ARCH_POPULATES_NODE_MAP | |
1001 | */ | |
1002 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1003 | extern void add_active_range(unsigned int nid, unsigned long start_pfn, | |
1004 | unsigned long end_pfn); | |
1005 | extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn, | |
1006 | unsigned long new_end_pfn); | |
fb01439c MG |
1007 | extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn, |
1008 | unsigned long end_pfn); | |
c713216d MG |
1009 | extern void remove_all_active_ranges(void); |
1010 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | |
1011 | unsigned long end_pfn); | |
1012 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1013 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1014 | extern unsigned long find_min_pfn_with_active_regions(void); | |
1015 | extern unsigned long find_max_pfn_with_active_regions(void); | |
1016 | extern void free_bootmem_with_active_regions(int nid, | |
1017 | unsigned long max_low_pfn); | |
1018 | extern void sparse_memory_present_with_active_regions(int nid); | |
1019 | #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID | |
1020 | extern int early_pfn_to_nid(unsigned long pfn); | |
1021 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ | |
1022 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ | |
0e0b864e | 1023 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
a2f3aa02 DH |
1024 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
1025 | unsigned long, enum memmap_context); | |
3947be19 | 1026 | extern void setup_per_zone_pages_min(void); |
1da177e4 LT |
1027 | extern void mem_init(void); |
1028 | extern void show_mem(void); | |
1029 | extern void si_meminfo(struct sysinfo * val); | |
1030 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
1031 | ||
e7c8d5c9 CL |
1032 | #ifdef CONFIG_NUMA |
1033 | extern void setup_per_cpu_pageset(void); | |
1034 | #else | |
1035 | static inline void setup_per_cpu_pageset(void) {} | |
1036 | #endif | |
1037 | ||
1da177e4 LT |
1038 | /* prio_tree.c */ |
1039 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); | |
1040 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); | |
1041 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); | |
1042 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, | |
1043 | struct prio_tree_iter *iter); | |
1044 | ||
1045 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ | |
1046 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ | |
1047 | (vma = vma_prio_tree_next(vma, iter)); ) | |
1048 | ||
1049 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
1050 | struct list_head *list) | |
1051 | { | |
1052 | vma->shared.vm_set.parent = NULL; | |
1053 | list_add_tail(&vma->shared.vm_set.list, list); | |
1054 | } | |
1055 | ||
1056 | /* mmap.c */ | |
1057 | extern int __vm_enough_memory(long pages, int cap_sys_admin); | |
1058 | extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1059 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | |
1060 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1061 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1062 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1063 | struct mempolicy *); | |
1064 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1065 | extern int split_vma(struct mm_struct *, | |
1066 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1067 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1068 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1069 | struct rb_node **, struct rb_node *); | |
a8fb5618 | 1070 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 LT |
1071 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
1072 | unsigned long addr, unsigned long len, pgoff_t pgoff); | |
1073 | extern void exit_mmap(struct mm_struct *); | |
119f657c | 1074 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
fa5dc22f RM |
1075 | extern int install_special_mapping(struct mm_struct *mm, |
1076 | unsigned long addr, unsigned long len, | |
1077 | unsigned long flags, struct page **pages); | |
1da177e4 LT |
1078 | |
1079 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
1080 | ||
1081 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
1082 | unsigned long len, unsigned long prot, | |
1083 | unsigned long flag, unsigned long pgoff); | |
0165ab44 MS |
1084 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
1085 | unsigned long len, unsigned long flags, | |
1086 | unsigned int vm_flags, unsigned long pgoff, | |
1087 | int accountable); | |
1da177e4 LT |
1088 | |
1089 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, | |
1090 | unsigned long len, unsigned long prot, | |
1091 | unsigned long flag, unsigned long offset) | |
1092 | { | |
1093 | unsigned long ret = -EINVAL; | |
1094 | if ((offset + PAGE_ALIGN(len)) < offset) | |
1095 | goto out; | |
1096 | if (!(offset & ~PAGE_MASK)) | |
1097 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); | |
1098 | out: | |
1099 | return ret; | |
1100 | } | |
1101 | ||
1102 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
1103 | ||
1104 | extern unsigned long do_brk(unsigned long, unsigned long); | |
1105 | ||
1106 | /* filemap.c */ | |
1107 | extern unsigned long page_unuse(struct page *); | |
1108 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
d7339071 HR |
1109 | extern void truncate_inode_pages_range(struct address_space *, |
1110 | loff_t lstart, loff_t lend); | |
1da177e4 LT |
1111 | |
1112 | /* generic vm_area_ops exported for stackable file systems */ | |
1113 | extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *); | |
1114 | extern int filemap_populate(struct vm_area_struct *, unsigned long, | |
1115 | unsigned long, pgprot_t, unsigned long, int); | |
1116 | ||
1117 | /* mm/page-writeback.c */ | |
1118 | int write_one_page(struct page *page, int wait); | |
1119 | ||
1120 | /* readahead.c */ | |
1121 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
1122 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
1123 | #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before | |
1124 | * turning readahead off */ | |
1125 | ||
1126 | int do_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
7361f4d8 | 1127 | pgoff_t offset, unsigned long nr_to_read); |
1da177e4 | 1128 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8 AM |
1129 | pgoff_t offset, unsigned long nr_to_read); |
1130 | unsigned long page_cache_readahead(struct address_space *mapping, | |
1da177e4 LT |
1131 | struct file_ra_state *ra, |
1132 | struct file *filp, | |
7361f4d8 | 1133 | pgoff_t offset, |
1da177e4 LT |
1134 | unsigned long size); |
1135 | void handle_ra_miss(struct address_space *mapping, | |
1136 | struct file_ra_state *ra, pgoff_t offset); | |
1137 | unsigned long max_sane_readahead(unsigned long nr); | |
1138 | ||
1139 | /* Do stack extension */ | |
46dea3d0 | 1140 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
9ab88515 | 1141 | #ifdef CONFIG_IA64 |
46dea3d0 | 1142 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
9ab88515 | 1143 | #endif |
1da177e4 LT |
1144 | |
1145 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
1146 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
1147 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
1148 | struct vm_area_struct **pprev); | |
1149 | ||
1150 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
1151 | NULL if none. Assume start_addr < end_addr. */ | |
1152 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
1153 | { | |
1154 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
1155 | ||
1156 | if (vma && end_addr <= vma->vm_start) | |
1157 | vma = NULL; | |
1158 | return vma; | |
1159 | } | |
1160 | ||
1161 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
1162 | { | |
1163 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
1164 | } | |
1165 | ||
804af2cf | 1166 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
deceb6cd HD |
1167 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
1168 | struct page *vmalloc_to_page(void *addr); | |
1169 | unsigned long vmalloc_to_pfn(void *addr); | |
1170 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | |
1171 | unsigned long pfn, unsigned long size, pgprot_t); | |
a145dd41 | 1172 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
e0dc0d8f NP |
1173 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
1174 | unsigned long pfn); | |
deceb6cd | 1175 | |
6aab341e | 1176 | struct page *follow_page(struct vm_area_struct *, unsigned long address, |
deceb6cd HD |
1177 | unsigned int foll_flags); |
1178 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
1179 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
1180 | #define FOLL_GET 0x04 /* do get_page on page */ | |
1181 | #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */ | |
1da177e4 | 1182 | |
aee16b3c JF |
1183 | typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr, |
1184 | void *data); | |
1185 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
1186 | unsigned long size, pte_fn_t fn, void *data); | |
1187 | ||
1da177e4 | 1188 | #ifdef CONFIG_PROC_FS |
ab50b8ed | 1189 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
1da177e4 | 1190 | #else |
ab50b8ed | 1191 | static inline void vm_stat_account(struct mm_struct *mm, |
1da177e4 LT |
1192 | unsigned long flags, struct file *file, long pages) |
1193 | { | |
1194 | } | |
1195 | #endif /* CONFIG_PROC_FS */ | |
1196 | ||
1da177e4 LT |
1197 | #ifndef CONFIG_DEBUG_PAGEALLOC |
1198 | static inline void | |
9858db50 | 1199 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
1da177e4 LT |
1200 | #endif |
1201 | ||
1202 | extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); | |
1203 | #ifdef __HAVE_ARCH_GATE_AREA | |
1204 | int in_gate_area_no_task(unsigned long addr); | |
1205 | int in_gate_area(struct task_struct *task, unsigned long addr); | |
1206 | #else | |
1207 | int in_gate_area_no_task(unsigned long addr); | |
1208 | #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) | |
1209 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
1210 | ||
9d0243bc AM |
1211 | int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *, |
1212 | void __user *, size_t *, loff_t *); | |
69e05944 | 1213 | unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, |
9d0243bc AM |
1214 | unsigned long lru_pages); |
1215 | void drop_pagecache(void); | |
1216 | void drop_slab(void); | |
1217 | ||
7a9166e3 LY |
1218 | #ifndef CONFIG_MMU |
1219 | #define randomize_va_space 0 | |
1220 | #else | |
a62eaf15 | 1221 | extern int randomize_va_space; |
7a9166e3 | 1222 | #endif |
a62eaf15 | 1223 | |
f269fdd1 | 1224 | __attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma); |
e6e5494c | 1225 | |
1da177e4 LT |
1226 | #endif /* __KERNEL__ */ |
1227 | #endif /* _LINUX_MM_H */ |