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 | 170 | |
d0217ac0 | 171 | #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ |
1da177e4 LT |
172 | |
173 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | |
174 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
175 | #endif | |
176 | ||
177 | #ifdef CONFIG_STACK_GROWSUP | |
178 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
179 | #else | |
180 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
181 | #endif | |
182 | ||
183 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) | |
184 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK | |
185 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) | |
186 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) | |
187 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) | |
188 | ||
189 | /* | |
190 | * mapping from the currently active vm_flags protection bits (the | |
191 | * low four bits) to a page protection mask.. | |
192 | */ | |
193 | extern pgprot_t protection_map[16]; | |
194 | ||
d0217ac0 NP |
195 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
196 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ | |
197 | ||
198 | ||
54cb8821 | 199 | /* |
d0217ac0 | 200 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
83c54070 NP |
201 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
202 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
54cb8821 | 203 | * |
d0217ac0 NP |
204 | * pgoff should be used in favour of virtual_address, if possible. If pgoff |
205 | * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear | |
206 | * mapping support. | |
54cb8821 | 207 | */ |
d0217ac0 NP |
208 | struct vm_fault { |
209 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | |
210 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
211 | void __user *virtual_address; /* Faulting virtual address */ | |
212 | ||
213 | struct page *page; /* ->fault handlers should return a | |
83c54070 | 214 | * page here, unless VM_FAULT_NOPAGE |
d0217ac0 | 215 | * is set (which is also implied by |
83c54070 | 216 | * VM_FAULT_ERROR). |
d0217ac0 | 217 | */ |
54cb8821 | 218 | }; |
1da177e4 LT |
219 | |
220 | /* | |
221 | * These are the virtual MM functions - opening of an area, closing and | |
222 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
223 | * to the functions called when a no-page or a wp-page exception occurs. | |
224 | */ | |
225 | struct vm_operations_struct { | |
226 | void (*open)(struct vm_area_struct * area); | |
227 | void (*close)(struct vm_area_struct * area); | |
d0217ac0 | 228 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); |
54cb8821 NP |
229 | struct page *(*nopage)(struct vm_area_struct *area, |
230 | unsigned long address, int *type); | |
231 | unsigned long (*nopfn)(struct vm_area_struct *area, | |
232 | unsigned long address); | |
9637a5ef DH |
233 | |
234 | /* notification that a previously read-only page is about to become | |
235 | * writable, if an error is returned it will cause a SIGBUS */ | |
236 | int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page); | |
1da177e4 LT |
237 | #ifdef CONFIG_NUMA |
238 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | |
239 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | |
240 | unsigned long addr); | |
7b2259b3 CL |
241 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
242 | const nodemask_t *to, unsigned long flags); | |
1da177e4 LT |
243 | #endif |
244 | }; | |
245 | ||
246 | struct mmu_gather; | |
247 | struct inode; | |
248 | ||
349aef0b AM |
249 | #define page_private(page) ((page)->private) |
250 | #define set_page_private(page, v) ((page)->private = (v)) | |
4c21e2f2 | 251 | |
1da177e4 LT |
252 | /* |
253 | * FIXME: take this include out, include page-flags.h in | |
254 | * files which need it (119 of them) | |
255 | */ | |
256 | #include <linux/page-flags.h> | |
257 | ||
725d704e NP |
258 | #ifdef CONFIG_DEBUG_VM |
259 | #define VM_BUG_ON(cond) BUG_ON(cond) | |
260 | #else | |
261 | #define VM_BUG_ON(condition) do { } while(0) | |
262 | #endif | |
263 | ||
1da177e4 LT |
264 | /* |
265 | * Methods to modify the page usage count. | |
266 | * | |
267 | * What counts for a page usage: | |
268 | * - cache mapping (page->mapping) | |
269 | * - private data (page->private) | |
270 | * - page mapped in a task's page tables, each mapping | |
271 | * is counted separately | |
272 | * | |
273 | * Also, many kernel routines increase the page count before a critical | |
274 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
275 | */ |
276 | ||
277 | /* | |
da6052f7 | 278 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 279 | */ |
7c8ee9a8 NP |
280 | static inline int put_page_testzero(struct page *page) |
281 | { | |
725d704e | 282 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
8dc04efb | 283 | return atomic_dec_and_test(&page->_count); |
7c8ee9a8 | 284 | } |
1da177e4 LT |
285 | |
286 | /* | |
7c8ee9a8 NP |
287 | * Try to grab a ref unless the page has a refcount of zero, return false if |
288 | * that is the case. | |
1da177e4 | 289 | */ |
7c8ee9a8 NP |
290 | static inline int get_page_unless_zero(struct page *page) |
291 | { | |
725d704e | 292 | VM_BUG_ON(PageCompound(page)); |
8dc04efb | 293 | return atomic_inc_not_zero(&page->_count); |
7c8ee9a8 | 294 | } |
1da177e4 | 295 | |
d85f3385 CL |
296 | static inline struct page *compound_head(struct page *page) |
297 | { | |
6d777953 | 298 | if (unlikely(PageTail(page))) |
d85f3385 CL |
299 | return page->first_page; |
300 | return page; | |
301 | } | |
302 | ||
4c21e2f2 | 303 | static inline int page_count(struct page *page) |
1da177e4 | 304 | { |
d85f3385 | 305 | return atomic_read(&compound_head(page)->_count); |
1da177e4 LT |
306 | } |
307 | ||
308 | static inline void get_page(struct page *page) | |
309 | { | |
d85f3385 | 310 | page = compound_head(page); |
725d704e | 311 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
1da177e4 LT |
312 | atomic_inc(&page->_count); |
313 | } | |
314 | ||
b49af68f CL |
315 | static inline struct page *virt_to_head_page(const void *x) |
316 | { | |
317 | struct page *page = virt_to_page(x); | |
318 | return compound_head(page); | |
319 | } | |
320 | ||
7835e98b NP |
321 | /* |
322 | * Setup the page count before being freed into the page allocator for | |
323 | * the first time (boot or memory hotplug) | |
324 | */ | |
325 | static inline void init_page_count(struct page *page) | |
326 | { | |
327 | atomic_set(&page->_count, 1); | |
328 | } | |
329 | ||
1da177e4 | 330 | void put_page(struct page *page); |
1d7ea732 | 331 | void put_pages_list(struct list_head *pages); |
1da177e4 | 332 | |
8dfcc9ba | 333 | void split_page(struct page *page, unsigned int order); |
8dfcc9ba | 334 | |
33f2ef89 AW |
335 | /* |
336 | * Compound pages have a destructor function. Provide a | |
337 | * prototype for that function and accessor functions. | |
338 | * These are _only_ valid on the head of a PG_compound page. | |
339 | */ | |
340 | typedef void compound_page_dtor(struct page *); | |
341 | ||
342 | static inline void set_compound_page_dtor(struct page *page, | |
343 | compound_page_dtor *dtor) | |
344 | { | |
345 | page[1].lru.next = (void *)dtor; | |
346 | } | |
347 | ||
348 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
349 | { | |
350 | return (compound_page_dtor *)page[1].lru.next; | |
351 | } | |
352 | ||
d85f3385 CL |
353 | static inline int compound_order(struct page *page) |
354 | { | |
6d777953 | 355 | if (!PageHead(page)) |
d85f3385 CL |
356 | return 0; |
357 | return (unsigned long)page[1].lru.prev; | |
358 | } | |
359 | ||
360 | static inline void set_compound_order(struct page *page, unsigned long order) | |
361 | { | |
362 | page[1].lru.prev = (void *)order; | |
363 | } | |
364 | ||
1da177e4 LT |
365 | /* |
366 | * Multiple processes may "see" the same page. E.g. for untouched | |
367 | * mappings of /dev/null, all processes see the same page full of | |
368 | * zeroes, and text pages of executables and shared libraries have | |
369 | * only one copy in memory, at most, normally. | |
370 | * | |
371 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
372 | * page_count() == 0 means the page is free. page->lru is then used for |
373 | * freelist management in the buddy allocator. | |
da6052f7 | 374 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 375 | * |
da6052f7 NP |
376 | * Pages are allocated by the slab allocator in order to provide memory |
377 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
378 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
379 | * unless a particular usage is carefully commented. (the responsibility of | |
380 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 381 | * |
da6052f7 NP |
382 | * A page may be used by anyone else who does a __get_free_page(). |
383 | * In this case, page_count still tracks the references, and should only | |
384 | * be used through the normal accessor functions. The top bits of page->flags | |
385 | * and page->virtual store page management information, but all other fields | |
386 | * are unused and could be used privately, carefully. The management of this | |
387 | * page is the responsibility of the one who allocated it, and those who have | |
388 | * subsequently been given references to it. | |
389 | * | |
390 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
391 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
392 | * The following discussion applies only to them. | |
393 | * | |
da6052f7 NP |
394 | * A pagecache page contains an opaque `private' member, which belongs to the |
395 | * page's address_space. Usually, this is the address of a circular list of | |
396 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
397 | * into the filesystem to release these pages. | |
1da177e4 | 398 | * |
da6052f7 NP |
399 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
400 | * is the pointer to the inode, and page->index is the file offset of the page, | |
401 | * in units of PAGE_CACHE_SIZE. | |
1da177e4 | 402 | * |
da6052f7 NP |
403 | * If pagecache pages are not associated with an inode, they are said to be |
404 | * anonymous pages. These may become associated with the swapcache, and in that | |
405 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 406 | * |
da6052f7 NP |
407 | * In either case (swapcache or inode backed), the pagecache itself holds one |
408 | * reference to the page. Setting PG_private should also increment the | |
409 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 410 | * |
da6052f7 NP |
411 | * The pagecache pages are stored in a per-mapping radix tree, which is |
412 | * rooted at mapping->page_tree, and indexed by offset. | |
413 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
414 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 415 | * |
da6052f7 | 416 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
417 | * - inode pages may need to be read from disk, |
418 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
419 | * to be written back to the inode on disk, |
420 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
421 | * modified may need to be swapped out to swap space and (later) to be read | |
422 | * back into memory. | |
1da177e4 LT |
423 | */ |
424 | ||
425 | /* | |
426 | * The zone field is never updated after free_area_init_core() | |
427 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 428 | */ |
348f8b6c | 429 | |
d41dee36 AW |
430 | |
431 | /* | |
432 | * page->flags layout: | |
433 | * | |
434 | * There are three possibilities for how page->flags get | |
435 | * laid out. The first is for the normal case, without | |
436 | * sparsemem. The second is for sparsemem when there is | |
437 | * plenty of space for node and section. The last is when | |
438 | * we have run out of space and have to fall back to an | |
439 | * alternate (slower) way of determining the node. | |
440 | * | |
441 | * No sparsemem: | NODE | ZONE | ... | FLAGS | | |
442 | * with space for node: | SECTION | NODE | ZONE | ... | FLAGS | | |
443 | * no space for node: | SECTION | ZONE | ... | FLAGS | | |
444 | */ | |
445 | #ifdef CONFIG_SPARSEMEM | |
446 | #define SECTIONS_WIDTH SECTIONS_SHIFT | |
447 | #else | |
448 | #define SECTIONS_WIDTH 0 | |
449 | #endif | |
450 | ||
451 | #define ZONES_WIDTH ZONES_SHIFT | |
452 | ||
453 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED | |
454 | #define NODES_WIDTH NODES_SHIFT | |
455 | #else | |
456 | #define NODES_WIDTH 0 | |
457 | #endif | |
458 | ||
459 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ | |
07808b74 | 460 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
d41dee36 AW |
461 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
462 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
463 | ||
464 | /* | |
465 | * We are going to use the flags for the page to node mapping if its in | |
466 | * there. This includes the case where there is no node, so it is implicit. | |
467 | */ | |
89689ae7 CL |
468 | #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) |
469 | #define NODE_NOT_IN_PAGE_FLAGS | |
470 | #endif | |
d41dee36 AW |
471 | |
472 | #ifndef PFN_SECTION_SHIFT | |
473 | #define PFN_SECTION_SHIFT 0 | |
474 | #endif | |
348f8b6c DH |
475 | |
476 | /* | |
477 | * Define the bit shifts to access each section. For non-existant | |
478 | * sections we define the shift as 0; that plus a 0 mask ensures | |
479 | * the compiler will optimise away reference to them. | |
480 | */ | |
d41dee36 AW |
481 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
482 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
483 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
348f8b6c | 484 | |
89689ae7 CL |
485 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ |
486 | #ifdef NODE_NOT_IN_PAGEFLAGS | |
487 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
bd8029b6 AW |
488 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
489 | SECTIONS_PGOFF : ZONES_PGOFF) | |
d41dee36 | 490 | #else |
89689ae7 | 491 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
492 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
493 | NODES_PGOFF : ZONES_PGOFF) | |
89689ae7 CL |
494 | #endif |
495 | ||
bd8029b6 | 496 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
348f8b6c | 497 | |
d41dee36 AW |
498 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED |
499 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED | |
348f8b6c DH |
500 | #endif |
501 | ||
d41dee36 AW |
502 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
503 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
504 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
89689ae7 | 505 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
348f8b6c | 506 | |
2f1b6248 | 507 | static inline enum zone_type page_zonenum(struct page *page) |
1da177e4 | 508 | { |
348f8b6c | 509 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 510 | } |
1da177e4 | 511 | |
89689ae7 CL |
512 | /* |
513 | * The identification function is only used by the buddy allocator for | |
514 | * determining if two pages could be buddies. We are not really | |
515 | * identifying a zone since we could be using a the section number | |
516 | * id if we have not node id available in page flags. | |
517 | * We guarantee only that it will return the same value for two | |
518 | * combinable pages in a zone. | |
519 | */ | |
cb2b95e1 AW |
520 | static inline int page_zone_id(struct page *page) |
521 | { | |
89689ae7 | 522 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
523 | } |
524 | ||
25ba77c1 | 525 | static inline int zone_to_nid(struct zone *zone) |
89fa3024 | 526 | { |
d5f541ed CL |
527 | #ifdef CONFIG_NUMA |
528 | return zone->node; | |
529 | #else | |
530 | return 0; | |
531 | #endif | |
89fa3024 CL |
532 | } |
533 | ||
89689ae7 | 534 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
25ba77c1 | 535 | extern int page_to_nid(struct page *page); |
89689ae7 | 536 | #else |
25ba77c1 | 537 | static inline int page_to_nid(struct page *page) |
d41dee36 | 538 | { |
89689ae7 | 539 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
d41dee36 | 540 | } |
89689ae7 CL |
541 | #endif |
542 | ||
543 | static inline struct zone *page_zone(struct page *page) | |
544 | { | |
545 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
546 | } | |
547 | ||
d41dee36 AW |
548 | static inline unsigned long page_to_section(struct page *page) |
549 | { | |
550 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
551 | } | |
552 | ||
2f1b6248 | 553 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
554 | { |
555 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
556 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
557 | } | |
2f1b6248 | 558 | |
348f8b6c DH |
559 | static inline void set_page_node(struct page *page, unsigned long node) |
560 | { | |
561 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
562 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 563 | } |
89689ae7 | 564 | |
d41dee36 AW |
565 | static inline void set_page_section(struct page *page, unsigned long section) |
566 | { | |
567 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
568 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
569 | } | |
1da177e4 | 570 | |
2f1b6248 | 571 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 572 | unsigned long node, unsigned long pfn) |
1da177e4 | 573 | { |
348f8b6c DH |
574 | set_page_zone(page, zone); |
575 | set_page_node(page, node); | |
d41dee36 | 576 | set_page_section(page, pfn_to_section_nr(pfn)); |
1da177e4 LT |
577 | } |
578 | ||
f6ac2354 CL |
579 | /* |
580 | * Some inline functions in vmstat.h depend on page_zone() | |
581 | */ | |
582 | #include <linux/vmstat.h> | |
583 | ||
652050ae | 584 | static __always_inline void *lowmem_page_address(struct page *page) |
1da177e4 LT |
585 | { |
586 | return __va(page_to_pfn(page) << PAGE_SHIFT); | |
587 | } | |
588 | ||
589 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
590 | #define HASHED_PAGE_VIRTUAL | |
591 | #endif | |
592 | ||
593 | #if defined(WANT_PAGE_VIRTUAL) | |
594 | #define page_address(page) ((page)->virtual) | |
595 | #define set_page_address(page, address) \ | |
596 | do { \ | |
597 | (page)->virtual = (address); \ | |
598 | } while(0) | |
599 | #define page_address_init() do { } while(0) | |
600 | #endif | |
601 | ||
602 | #if defined(HASHED_PAGE_VIRTUAL) | |
603 | void *page_address(struct page *page); | |
604 | void set_page_address(struct page *page, void *virtual); | |
605 | void page_address_init(void); | |
606 | #endif | |
607 | ||
608 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
609 | #define page_address(page) lowmem_page_address(page) | |
610 | #define set_page_address(page, address) do { } while(0) | |
611 | #define page_address_init() do { } while(0) | |
612 | #endif | |
613 | ||
614 | /* | |
615 | * On an anonymous page mapped into a user virtual memory area, | |
616 | * page->mapping points to its anon_vma, not to a struct address_space; | |
617 | * with the PAGE_MAPPING_ANON bit set to distinguish it. | |
618 | * | |
619 | * Please note that, confusingly, "page_mapping" refers to the inode | |
620 | * address_space which maps the page from disk; whereas "page_mapped" | |
621 | * refers to user virtual address space into which the page is mapped. | |
622 | */ | |
623 | #define PAGE_MAPPING_ANON 1 | |
624 | ||
625 | extern struct address_space swapper_space; | |
626 | static inline struct address_space *page_mapping(struct page *page) | |
627 | { | |
628 | struct address_space *mapping = page->mapping; | |
629 | ||
b5fab14e | 630 | VM_BUG_ON(PageSlab(page)); |
1da177e4 LT |
631 | if (unlikely(PageSwapCache(page))) |
632 | mapping = &swapper_space; | |
b9bae340 HD |
633 | #ifdef CONFIG_SLUB |
634 | else if (unlikely(PageSlab(page))) | |
635 | mapping = NULL; | |
636 | #endif | |
1da177e4 LT |
637 | else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) |
638 | mapping = NULL; | |
639 | return mapping; | |
640 | } | |
641 | ||
642 | static inline int PageAnon(struct page *page) | |
643 | { | |
644 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
645 | } | |
646 | ||
647 | /* | |
648 | * Return the pagecache index of the passed page. Regular pagecache pages | |
649 | * use ->index whereas swapcache pages use ->private | |
650 | */ | |
651 | static inline pgoff_t page_index(struct page *page) | |
652 | { | |
653 | if (unlikely(PageSwapCache(page))) | |
4c21e2f2 | 654 | return page_private(page); |
1da177e4 LT |
655 | return page->index; |
656 | } | |
657 | ||
658 | /* | |
659 | * The atomic page->_mapcount, like _count, starts from -1: | |
660 | * so that transitions both from it and to it can be tracked, | |
661 | * using atomic_inc_and_test and atomic_add_negative(-1). | |
662 | */ | |
663 | static inline void reset_page_mapcount(struct page *page) | |
664 | { | |
665 | atomic_set(&(page)->_mapcount, -1); | |
666 | } | |
667 | ||
668 | static inline int page_mapcount(struct page *page) | |
669 | { | |
670 | return atomic_read(&(page)->_mapcount) + 1; | |
671 | } | |
672 | ||
673 | /* | |
674 | * Return true if this page is mapped into pagetables. | |
675 | */ | |
676 | static inline int page_mapped(struct page *page) | |
677 | { | |
678 | return atomic_read(&(page)->_mapcount) >= 0; | |
679 | } | |
680 | ||
681 | /* | |
682 | * Error return values for the *_nopage functions | |
683 | */ | |
684 | #define NOPAGE_SIGBUS (NULL) | |
685 | #define NOPAGE_OOM ((struct page *) (-1)) | |
686 | ||
f4b81804 JS |
687 | /* |
688 | * Error return values for the *_nopfn functions | |
689 | */ | |
690 | #define NOPFN_SIGBUS ((unsigned long) -1) | |
691 | #define NOPFN_OOM ((unsigned long) -2) | |
22cd25ed | 692 | #define NOPFN_REFAULT ((unsigned long) -3) |
f4b81804 | 693 | |
1da177e4 LT |
694 | /* |
695 | * Different kinds of faults, as returned by handle_mm_fault(). | |
696 | * Used to decide whether a process gets delivered SIGBUS or | |
697 | * just gets major/minor fault counters bumped up. | |
698 | */ | |
d0217ac0 | 699 | |
83c54070 | 700 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ |
d0217ac0 | 701 | |
83c54070 NP |
702 | #define VM_FAULT_OOM 0x0001 |
703 | #define VM_FAULT_SIGBUS 0x0002 | |
704 | #define VM_FAULT_MAJOR 0x0004 | |
705 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | |
f33ea7f4 | 706 | |
83c54070 NP |
707 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ |
708 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | |
1da177e4 | 709 | |
83c54070 | 710 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS) |
d0217ac0 | 711 | |
1da177e4 LT |
712 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
713 | ||
714 | extern void show_free_areas(void); | |
715 | ||
716 | #ifdef CONFIG_SHMEM | |
1da177e4 LT |
717 | int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new); |
718 | struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, | |
719 | unsigned long addr); | |
720 | int shmem_lock(struct file *file, int lock, struct user_struct *user); | |
721 | #else | |
03b00ebc RK |
722 | static inline int shmem_lock(struct file *file, int lock, |
723 | struct user_struct *user) | |
724 | { | |
725 | return 0; | |
726 | } | |
727 | ||
728 | static inline int shmem_set_policy(struct vm_area_struct *vma, | |
729 | struct mempolicy *new) | |
730 | { | |
731 | return 0; | |
732 | } | |
733 | ||
734 | static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, | |
735 | unsigned long addr) | |
736 | { | |
737 | return NULL; | |
738 | } | |
1da177e4 LT |
739 | #endif |
740 | struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); | |
741 | ||
742 | int shmem_zero_setup(struct vm_area_struct *); | |
743 | ||
b0e15190 DH |
744 | #ifndef CONFIG_MMU |
745 | extern unsigned long shmem_get_unmapped_area(struct file *file, | |
746 | unsigned long addr, | |
747 | unsigned long len, | |
748 | unsigned long pgoff, | |
749 | unsigned long flags); | |
750 | #endif | |
751 | ||
e8edc6e0 | 752 | extern int can_do_mlock(void); |
1da177e4 LT |
753 | extern int user_shm_lock(size_t, struct user_struct *); |
754 | extern void user_shm_unlock(size_t, struct user_struct *); | |
755 | ||
756 | /* | |
757 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
758 | */ | |
759 | struct zap_details { | |
760 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | |
761 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
762 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
763 | pgoff_t last_index; /* Highest page->index to unmap */ | |
764 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ | |
1da177e4 LT |
765 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
766 | }; | |
767 | ||
6aab341e | 768 | struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t); |
ee39b37b | 769 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
1da177e4 | 770 | unsigned long size, struct zap_details *); |
508034a3 | 771 | unsigned long unmap_vmas(struct mmu_gather **tlb, |
1da177e4 LT |
772 | struct vm_area_struct *start_vma, unsigned long start_addr, |
773 | unsigned long end_addr, unsigned long *nr_accounted, | |
774 | struct zap_details *); | |
3bf5ee95 HD |
775 | void free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
776 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
777 | void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma, | |
e0da382c | 778 | unsigned long floor, unsigned long ceiling); |
1da177e4 LT |
779 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
780 | struct vm_area_struct *vma); | |
781 | int zeromap_page_range(struct vm_area_struct *vma, unsigned long from, | |
782 | unsigned long size, pgprot_t prot); | |
783 | void unmap_mapping_range(struct address_space *mapping, | |
784 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
785 | ||
786 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
787 | loff_t const holebegin, loff_t const holelen) | |
788 | { | |
789 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
790 | } | |
791 | ||
792 | extern int vmtruncate(struct inode * inode, loff_t offset); | |
f6b3ec23 | 793 | extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end); |
f33ea7f4 | 794 | |
7ee1dd3f | 795 | #ifdef CONFIG_MMU |
83c54070 | 796 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
7ee1dd3f | 797 | unsigned long address, int write_access); |
7ee1dd3f DH |
798 | #else |
799 | static inline int handle_mm_fault(struct mm_struct *mm, | |
800 | struct vm_area_struct *vma, unsigned long address, | |
801 | int write_access) | |
802 | { | |
803 | /* should never happen if there's no MMU */ | |
804 | BUG(); | |
805 | return VM_FAULT_SIGBUS; | |
806 | } | |
807 | #endif | |
f33ea7f4 | 808 | |
1da177e4 LT |
809 | extern int make_pages_present(unsigned long addr, unsigned long end); |
810 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
1da177e4 LT |
811 | |
812 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, | |
813 | int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); | |
b5810039 | 814 | void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long); |
1da177e4 | 815 | |
cf9a2ae8 DH |
816 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
817 | extern void do_invalidatepage(struct page *page, unsigned long offset); | |
818 | ||
1da177e4 | 819 | int __set_page_dirty_nobuffers(struct page *page); |
76719325 | 820 | int __set_page_dirty_no_writeback(struct page *page); |
1da177e4 LT |
821 | int redirty_page_for_writepage(struct writeback_control *wbc, |
822 | struct page *page); | |
823 | int FASTCALL(set_page_dirty(struct page *page)); | |
824 | int set_page_dirty_lock(struct page *page); | |
825 | int clear_page_dirty_for_io(struct page *page); | |
826 | ||
b6a2fea3 OW |
827 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
828 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
829 | unsigned long new_addr, unsigned long len); | |
1da177e4 LT |
830 | extern unsigned long do_mremap(unsigned long addr, |
831 | unsigned long old_len, unsigned long new_len, | |
832 | unsigned long flags, unsigned long new_addr); | |
b6a2fea3 OW |
833 | extern int mprotect_fixup(struct vm_area_struct *vma, |
834 | struct vm_area_struct **pprev, unsigned long start, | |
835 | unsigned long end, unsigned long newflags); | |
1da177e4 LT |
836 | |
837 | /* | |
8e1f936b | 838 | * A callback you can register to apply pressure to ageable caches. |
1da177e4 | 839 | * |
8e1f936b RR |
840 | * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should |
841 | * look through the least-recently-used 'nr_to_scan' entries and | |
842 | * attempt to free them up. It should return the number of objects | |
843 | * which remain in the cache. If it returns -1, it means it cannot do | |
844 | * any scanning at this time (eg. there is a risk of deadlock). | |
1da177e4 | 845 | * |
8e1f936b RR |
846 | * The 'gfpmask' refers to the allocation we are currently trying to |
847 | * fulfil. | |
848 | * | |
849 | * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is | |
850 | * querying the cache size, so a fastpath for that case is appropriate. | |
1da177e4 | 851 | */ |
8e1f936b RR |
852 | struct shrinker { |
853 | int (*shrink)(int nr_to_scan, gfp_t gfp_mask); | |
854 | int seeks; /* seeks to recreate an obj */ | |
1da177e4 | 855 | |
8e1f936b RR |
856 | /* These are for internal use */ |
857 | struct list_head list; | |
858 | long nr; /* objs pending delete */ | |
859 | }; | |
860 | #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ | |
861 | extern void register_shrinker(struct shrinker *); | |
862 | extern void unregister_shrinker(struct shrinker *); | |
1da177e4 | 863 | |
d08b3851 PZ |
864 | /* |
865 | * Some shared mappigns will want the pages marked read-only | |
866 | * to track write events. If so, we'll downgrade vm_page_prot | |
867 | * to the private version (using protection_map[] without the | |
868 | * VM_SHARED bit). | |
869 | */ | |
870 | static inline int vma_wants_writenotify(struct vm_area_struct *vma) | |
871 | { | |
872 | unsigned int vm_flags = vma->vm_flags; | |
873 | ||
874 | /* If it was private or non-writable, the write bit is already clear */ | |
875 | if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED))) | |
876 | return 0; | |
877 | ||
878 | /* The backer wishes to know when pages are first written to? */ | |
879 | if (vma->vm_ops && vma->vm_ops->page_mkwrite) | |
880 | return 1; | |
881 | ||
882 | /* The open routine did something to the protections already? */ | |
883 | if (pgprot_val(vma->vm_page_prot) != | |
884 | pgprot_val(protection_map[vm_flags & | |
885 | (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)])) | |
886 | return 0; | |
887 | ||
888 | /* Specialty mapping? */ | |
889 | if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE)) | |
890 | return 0; | |
891 | ||
892 | /* Can the mapping track the dirty pages? */ | |
893 | return vma->vm_file && vma->vm_file->f_mapping && | |
894 | mapping_cap_account_dirty(vma->vm_file->f_mapping); | |
895 | } | |
896 | ||
c9cfcddf LT |
897 | extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl)); |
898 | ||
5f22df00 NP |
899 | #ifdef __PAGETABLE_PUD_FOLDED |
900 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
901 | unsigned long address) | |
902 | { | |
903 | return 0; | |
904 | } | |
905 | #else | |
1bb3630e | 906 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
5f22df00 NP |
907 | #endif |
908 | ||
909 | #ifdef __PAGETABLE_PMD_FOLDED | |
910 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
911 | unsigned long address) | |
912 | { | |
913 | return 0; | |
914 | } | |
915 | #else | |
1bb3630e | 916 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
5f22df00 NP |
917 | #endif |
918 | ||
1bb3630e HD |
919 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); |
920 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | |
921 | ||
1da177e4 LT |
922 | /* |
923 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
924 | * Remove it when 4level-fixup.h has been removed. | |
925 | */ | |
1bb3630e | 926 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
1da177e4 LT |
927 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
928 | { | |
1bb3630e HD |
929 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
930 | NULL: pud_offset(pgd, address); | |
1da177e4 LT |
931 | } |
932 | ||
933 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
934 | { | |
1bb3630e HD |
935 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
936 | NULL: pmd_offset(pud, address); | |
1da177e4 | 937 | } |
1bb3630e HD |
938 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
939 | ||
4c21e2f2 HD |
940 | #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS |
941 | /* | |
942 | * We tuck a spinlock to guard each pagetable page into its struct page, | |
943 | * at page->private, with BUILD_BUG_ON to make sure that this will not | |
944 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). | |
945 | * When freeing, reset page->mapping so free_pages_check won't complain. | |
946 | */ | |
349aef0b | 947 | #define __pte_lockptr(page) &((page)->ptl) |
4c21e2f2 HD |
948 | #define pte_lock_init(_page) do { \ |
949 | spin_lock_init(__pte_lockptr(_page)); \ | |
950 | } while (0) | |
951 | #define pte_lock_deinit(page) ((page)->mapping = NULL) | |
952 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) | |
953 | #else | |
954 | /* | |
955 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
956 | */ | |
957 | #define pte_lock_init(page) do {} while (0) | |
958 | #define pte_lock_deinit(page) do {} while (0) | |
959 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) | |
960 | #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */ | |
961 | ||
c74df32c HD |
962 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
963 | ({ \ | |
4c21e2f2 | 964 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
965 | pte_t *__pte = pte_offset_map(pmd, address); \ |
966 | *(ptlp) = __ptl; \ | |
967 | spin_lock(__ptl); \ | |
968 | __pte; \ | |
969 | }) | |
970 | ||
971 | #define pte_unmap_unlock(pte, ptl) do { \ | |
972 | spin_unlock(ptl); \ | |
973 | pte_unmap(pte); \ | |
974 | } while (0) | |
975 | ||
1bb3630e HD |
976 | #define pte_alloc_map(mm, pmd, address) \ |
977 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ | |
978 | NULL: pte_offset_map(pmd, address)) | |
979 | ||
c74df32c HD |
980 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
981 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ | |
982 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) | |
983 | ||
1bb3630e HD |
984 | #define pte_alloc_kernel(pmd, address) \ |
985 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | |
986 | NULL: pte_offset_kernel(pmd, address)) | |
1da177e4 LT |
987 | |
988 | extern void free_area_init(unsigned long * zones_size); | |
989 | extern void free_area_init_node(int nid, pg_data_t *pgdat, | |
990 | unsigned long * zones_size, unsigned long zone_start_pfn, | |
991 | unsigned long *zholes_size); | |
c713216d MG |
992 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
993 | /* | |
994 | * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its | |
995 | * zones, allocate the backing mem_map and account for memory holes in a more | |
996 | * architecture independent manner. This is a substitute for creating the | |
997 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
998 | * free_area_init_node() | |
999 | * | |
1000 | * An architecture is expected to register range of page frames backed by | |
1001 | * physical memory with add_active_range() before calling | |
1002 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
1003 | * usage, an architecture is expected to do something like | |
1004 | * | |
1005 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
1006 | * max_highmem_pfn}; | |
1007 | * for_each_valid_physical_page_range() | |
1008 | * add_active_range(node_id, start_pfn, end_pfn) | |
1009 | * free_area_init_nodes(max_zone_pfns); | |
1010 | * | |
1011 | * If the architecture guarantees that there are no holes in the ranges | |
1012 | * registered with add_active_range(), free_bootmem_active_regions() | |
1013 | * will call free_bootmem_node() for each registered physical page range. | |
1014 | * Similarly sparse_memory_present_with_active_regions() calls | |
1015 | * memory_present() for each range when SPARSEMEM is enabled. | |
1016 | * | |
1017 | * See mm/page_alloc.c for more information on each function exposed by | |
1018 | * CONFIG_ARCH_POPULATES_NODE_MAP | |
1019 | */ | |
1020 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1021 | extern void add_active_range(unsigned int nid, unsigned long start_pfn, | |
1022 | unsigned long end_pfn); | |
1023 | extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn, | |
1024 | unsigned long new_end_pfn); | |
fb01439c MG |
1025 | extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn, |
1026 | unsigned long end_pfn); | |
c713216d MG |
1027 | extern void remove_all_active_ranges(void); |
1028 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | |
1029 | unsigned long end_pfn); | |
1030 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1031 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1032 | extern unsigned long find_min_pfn_with_active_regions(void); | |
1033 | extern unsigned long find_max_pfn_with_active_regions(void); | |
1034 | extern void free_bootmem_with_active_regions(int nid, | |
1035 | unsigned long max_low_pfn); | |
1036 | extern void sparse_memory_present_with_active_regions(int nid); | |
1037 | #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID | |
1038 | extern int early_pfn_to_nid(unsigned long pfn); | |
1039 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ | |
1040 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ | |
0e0b864e | 1041 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
a2f3aa02 DH |
1042 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
1043 | unsigned long, enum memmap_context); | |
3947be19 | 1044 | extern void setup_per_zone_pages_min(void); |
1da177e4 LT |
1045 | extern void mem_init(void); |
1046 | extern void show_mem(void); | |
1047 | extern void si_meminfo(struct sysinfo * val); | |
1048 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
1049 | ||
e7c8d5c9 CL |
1050 | #ifdef CONFIG_NUMA |
1051 | extern void setup_per_cpu_pageset(void); | |
1052 | #else | |
1053 | static inline void setup_per_cpu_pageset(void) {} | |
1054 | #endif | |
1055 | ||
1da177e4 LT |
1056 | /* prio_tree.c */ |
1057 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); | |
1058 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); | |
1059 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); | |
1060 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, | |
1061 | struct prio_tree_iter *iter); | |
1062 | ||
1063 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ | |
1064 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ | |
1065 | (vma = vma_prio_tree_next(vma, iter)); ) | |
1066 | ||
1067 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
1068 | struct list_head *list) | |
1069 | { | |
1070 | vma->shared.vm_set.parent = NULL; | |
1071 | list_add_tail(&vma->shared.vm_set.list, list); | |
1072 | } | |
1073 | ||
1074 | /* mmap.c */ | |
1075 | extern int __vm_enough_memory(long pages, int cap_sys_admin); | |
1076 | extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1077 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | |
1078 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1079 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1080 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1081 | struct mempolicy *); | |
1082 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1083 | extern int split_vma(struct mm_struct *, | |
1084 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1085 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1086 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1087 | struct rb_node **, struct rb_node *); | |
a8fb5618 | 1088 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 LT |
1089 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
1090 | unsigned long addr, unsigned long len, pgoff_t pgoff); | |
1091 | extern void exit_mmap(struct mm_struct *); | |
119f657c | 1092 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
fa5dc22f RM |
1093 | extern int install_special_mapping(struct mm_struct *mm, |
1094 | unsigned long addr, unsigned long len, | |
1095 | unsigned long flags, struct page **pages); | |
1da177e4 LT |
1096 | |
1097 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
1098 | ||
1099 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
1100 | unsigned long len, unsigned long prot, | |
1101 | unsigned long flag, unsigned long pgoff); | |
0165ab44 MS |
1102 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
1103 | unsigned long len, unsigned long flags, | |
1104 | unsigned int vm_flags, unsigned long pgoff, | |
1105 | int accountable); | |
1da177e4 LT |
1106 | |
1107 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, | |
1108 | unsigned long len, unsigned long prot, | |
1109 | unsigned long flag, unsigned long offset) | |
1110 | { | |
1111 | unsigned long ret = -EINVAL; | |
1112 | if ((offset + PAGE_ALIGN(len)) < offset) | |
1113 | goto out; | |
1114 | if (!(offset & ~PAGE_MASK)) | |
1115 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); | |
1116 | out: | |
1117 | return ret; | |
1118 | } | |
1119 | ||
1120 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
1121 | ||
1122 | extern unsigned long do_brk(unsigned long, unsigned long); | |
1123 | ||
1124 | /* filemap.c */ | |
1125 | extern unsigned long page_unuse(struct page *); | |
1126 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
d7339071 HR |
1127 | extern void truncate_inode_pages_range(struct address_space *, |
1128 | loff_t lstart, loff_t lend); | |
1da177e4 LT |
1129 | |
1130 | /* generic vm_area_ops exported for stackable file systems */ | |
d0217ac0 | 1131 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); |
1da177e4 LT |
1132 | |
1133 | /* mm/page-writeback.c */ | |
1134 | int write_one_page(struct page *page, int wait); | |
1135 | ||
1136 | /* readahead.c */ | |
1137 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
1138 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
1139 | #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before | |
1140 | * turning readahead off */ | |
1141 | ||
1142 | int do_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
7361f4d8 | 1143 | pgoff_t offset, unsigned long nr_to_read); |
1da177e4 | 1144 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8 | 1145 | pgoff_t offset, unsigned long nr_to_read); |
cf914a7d RR |
1146 | |
1147 | void page_cache_sync_readahead(struct address_space *mapping, | |
1148 | struct file_ra_state *ra, | |
1149 | struct file *filp, | |
1150 | pgoff_t offset, | |
1151 | unsigned long size); | |
1152 | ||
1153 | void page_cache_async_readahead(struct address_space *mapping, | |
1154 | struct file_ra_state *ra, | |
1155 | struct file *filp, | |
1156 | struct page *pg, | |
1157 | pgoff_t offset, | |
1158 | unsigned long size); | |
1159 | ||
1da177e4 LT |
1160 | unsigned long max_sane_readahead(unsigned long nr); |
1161 | ||
1162 | /* Do stack extension */ | |
46dea3d0 | 1163 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
9ab88515 | 1164 | #ifdef CONFIG_IA64 |
46dea3d0 | 1165 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
9ab88515 | 1166 | #endif |
b6a2fea3 OW |
1167 | extern int expand_stack_downwards(struct vm_area_struct *vma, |
1168 | unsigned long address); | |
1da177e4 LT |
1169 | |
1170 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
1171 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
1172 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
1173 | struct vm_area_struct **pprev); | |
1174 | ||
1175 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
1176 | NULL if none. Assume start_addr < end_addr. */ | |
1177 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
1178 | { | |
1179 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
1180 | ||
1181 | if (vma && end_addr <= vma->vm_start) | |
1182 | vma = NULL; | |
1183 | return vma; | |
1184 | } | |
1185 | ||
1186 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
1187 | { | |
1188 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
1189 | } | |
1190 | ||
804af2cf | 1191 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
deceb6cd HD |
1192 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
1193 | struct page *vmalloc_to_page(void *addr); | |
1194 | unsigned long vmalloc_to_pfn(void *addr); | |
1195 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | |
1196 | unsigned long pfn, unsigned long size, pgprot_t); | |
a145dd41 | 1197 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
e0dc0d8f NP |
1198 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
1199 | unsigned long pfn); | |
deceb6cd | 1200 | |
6aab341e | 1201 | struct page *follow_page(struct vm_area_struct *, unsigned long address, |
deceb6cd HD |
1202 | unsigned int foll_flags); |
1203 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
1204 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
1205 | #define FOLL_GET 0x04 /* do get_page on page */ | |
1206 | #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */ | |
1da177e4 | 1207 | |
aee16b3c JF |
1208 | typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr, |
1209 | void *data); | |
1210 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
1211 | unsigned long size, pte_fn_t fn, void *data); | |
1212 | ||
1da177e4 | 1213 | #ifdef CONFIG_PROC_FS |
ab50b8ed | 1214 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
1da177e4 | 1215 | #else |
ab50b8ed | 1216 | static inline void vm_stat_account(struct mm_struct *mm, |
1da177e4 LT |
1217 | unsigned long flags, struct file *file, long pages) |
1218 | { | |
1219 | } | |
1220 | #endif /* CONFIG_PROC_FS */ | |
1221 | ||
1da177e4 LT |
1222 | #ifndef CONFIG_DEBUG_PAGEALLOC |
1223 | static inline void | |
9858db50 | 1224 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
1da177e4 LT |
1225 | #endif |
1226 | ||
1227 | extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); | |
1228 | #ifdef __HAVE_ARCH_GATE_AREA | |
1229 | int in_gate_area_no_task(unsigned long addr); | |
1230 | int in_gate_area(struct task_struct *task, unsigned long addr); | |
1231 | #else | |
1232 | int in_gate_area_no_task(unsigned long addr); | |
1233 | #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) | |
1234 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
1235 | ||
9d0243bc AM |
1236 | int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *, |
1237 | void __user *, size_t *, loff_t *); | |
69e05944 | 1238 | unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, |
9d0243bc AM |
1239 | unsigned long lru_pages); |
1240 | void drop_pagecache(void); | |
1241 | void drop_slab(void); | |
1242 | ||
7a9166e3 LY |
1243 | #ifndef CONFIG_MMU |
1244 | #define randomize_va_space 0 | |
1245 | #else | |
a62eaf15 | 1246 | extern int randomize_va_space; |
7a9166e3 | 1247 | #endif |
a62eaf15 | 1248 | |
045e72ac | 1249 | const char * arch_vma_name(struct vm_area_struct *vma); |
e6e5494c | 1250 | |
1da177e4 LT |
1251 | #endif /* __KERNEL__ */ |
1252 | #endif /* _LINUX_MM_H */ |