]>
Commit | Line | Data |
---|---|---|
1 | #ifndef _LINUX_MM_H | |
2 | #define _LINUX_MM_H | |
3 | ||
4 | #include <linux/errno.h> | |
5 | ||
6 | #ifdef __KERNEL__ | |
7 | ||
8 | #include <linux/gfp.h> | |
9 | #include <linux/bug.h> | |
10 | #include <linux/list.h> | |
11 | #include <linux/mmzone.h> | |
12 | #include <linux/rbtree.h> | |
13 | #include <linux/atomic.h> | |
14 | #include <linux/debug_locks.h> | |
15 | #include <linux/mm_types.h> | |
16 | #include <linux/range.h> | |
17 | #include <linux/pfn.h> | |
18 | #include <linux/bit_spinlock.h> | |
19 | #include <linux/shrinker.h> | |
20 | ||
21 | struct mempolicy; | |
22 | struct anon_vma; | |
23 | struct anon_vma_chain; | |
24 | struct file_ra_state; | |
25 | struct user_struct; | |
26 | struct writeback_control; | |
27 | ||
28 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ | |
29 | extern unsigned long max_mapnr; | |
30 | #endif | |
31 | ||
32 | extern unsigned long num_physpages; | |
33 | extern unsigned long totalram_pages; | |
34 | extern void * high_memory; | |
35 | extern int page_cluster; | |
36 | ||
37 | #ifdef CONFIG_SYSCTL | |
38 | extern int sysctl_legacy_va_layout; | |
39 | #else | |
40 | #define sysctl_legacy_va_layout 0 | |
41 | #endif | |
42 | ||
43 | #include <asm/page.h> | |
44 | #include <asm/pgtable.h> | |
45 | #include <asm/processor.h> | |
46 | ||
47 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) | |
48 | ||
49 | /* to align the pointer to the (next) page boundary */ | |
50 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
51 | ||
52 | /* | |
53 | * Linux kernel virtual memory manager primitives. | |
54 | * The idea being to have a "virtual" mm in the same way | |
55 | * we have a virtual fs - giving a cleaner interface to the | |
56 | * mm details, and allowing different kinds of memory mappings | |
57 | * (from shared memory to executable loading to arbitrary | |
58 | * mmap() functions). | |
59 | */ | |
60 | ||
61 | extern struct kmem_cache *vm_area_cachep; | |
62 | ||
63 | #ifndef CONFIG_MMU | |
64 | extern struct rb_root nommu_region_tree; | |
65 | extern struct rw_semaphore nommu_region_sem; | |
66 | ||
67 | extern unsigned int kobjsize(const void *objp); | |
68 | #endif | |
69 | ||
70 | /* | |
71 | * vm_flags in vm_area_struct, see mm_types.h. | |
72 | */ | |
73 | #define VM_NONE 0x00000000 | |
74 | ||
75 | #define VM_READ 0x00000001 /* currently active flags */ | |
76 | #define VM_WRITE 0x00000002 | |
77 | #define VM_EXEC 0x00000004 | |
78 | #define VM_SHARED 0x00000008 | |
79 | ||
80 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ | |
81 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ | |
82 | #define VM_MAYWRITE 0x00000020 | |
83 | #define VM_MAYEXEC 0x00000040 | |
84 | #define VM_MAYSHARE 0x00000080 | |
85 | ||
86 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
87 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ | |
88 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ | |
89 | ||
90 | #define VM_LOCKED 0x00002000 | |
91 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
92 | ||
93 | /* Used by sys_madvise() */ | |
94 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
95 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
96 | ||
97 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
98 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
99 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ | |
100 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ | |
101 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ | |
102 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ | |
103 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ | |
104 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ | |
105 | ||
106 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ | |
107 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ | |
108 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
109 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ | |
110 | ||
111 | #if defined(CONFIG_X86) | |
112 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
113 | #elif defined(CONFIG_PPC) | |
114 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
115 | #elif defined(CONFIG_PARISC) | |
116 | # define VM_GROWSUP VM_ARCH_1 | |
117 | #elif defined(CONFIG_METAG) | |
118 | # define VM_GROWSUP VM_ARCH_1 | |
119 | #elif defined(CONFIG_IA64) | |
120 | # define VM_GROWSUP VM_ARCH_1 | |
121 | #elif !defined(CONFIG_MMU) | |
122 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
123 | #endif | |
124 | ||
125 | #ifndef VM_GROWSUP | |
126 | # define VM_GROWSUP VM_NONE | |
127 | #endif | |
128 | ||
129 | /* Bits set in the VMA until the stack is in its final location */ | |
130 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
131 | ||
132 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | |
133 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
134 | #endif | |
135 | ||
136 | #ifdef CONFIG_STACK_GROWSUP | |
137 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
138 | #else | |
139 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
140 | #endif | |
141 | ||
142 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) | |
143 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK | |
144 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) | |
145 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) | |
146 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) | |
147 | ||
148 | /* | |
149 | * Special vmas that are non-mergable, non-mlock()able. | |
150 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | |
151 | */ | |
152 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP) | |
153 | ||
154 | /* | |
155 | * mapping from the currently active vm_flags protection bits (the | |
156 | * low four bits) to a page protection mask.. | |
157 | */ | |
158 | extern pgprot_t protection_map[16]; | |
159 | ||
160 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ | |
161 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ | |
162 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ | |
163 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ | |
164 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ | |
165 | #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ | |
166 | #define FAULT_FLAG_TRIED 0x40 /* second try */ | |
167 | ||
168 | /* | |
169 | * vm_fault is filled by the the pagefault handler and passed to the vma's | |
170 | * ->fault function. The vma's ->fault is responsible for returning a bitmask | |
171 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
172 | * | |
173 | * pgoff should be used in favour of virtual_address, if possible. If pgoff | |
174 | * is used, one may implement ->remap_pages to get nonlinear mapping support. | |
175 | */ | |
176 | struct vm_fault { | |
177 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | |
178 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
179 | void __user *virtual_address; /* Faulting virtual address */ | |
180 | ||
181 | struct page *page; /* ->fault handlers should return a | |
182 | * page here, unless VM_FAULT_NOPAGE | |
183 | * is set (which is also implied by | |
184 | * VM_FAULT_ERROR). | |
185 | */ | |
186 | }; | |
187 | ||
188 | /* | |
189 | * These are the virtual MM functions - opening of an area, closing and | |
190 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
191 | * to the functions called when a no-page or a wp-page exception occurs. | |
192 | */ | |
193 | struct vm_operations_struct { | |
194 | void (*open)(struct vm_area_struct * area); | |
195 | void (*close)(struct vm_area_struct * area); | |
196 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
197 | ||
198 | /* notification that a previously read-only page is about to become | |
199 | * writable, if an error is returned it will cause a SIGBUS */ | |
200 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
201 | ||
202 | /* called by access_process_vm when get_user_pages() fails, typically | |
203 | * for use by special VMAs that can switch between memory and hardware | |
204 | */ | |
205 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
206 | void *buf, int len, int write); | |
207 | #ifdef CONFIG_NUMA | |
208 | /* | |
209 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
210 | * to hold the policy upon return. Caller should pass NULL @new to | |
211 | * remove a policy and fall back to surrounding context--i.e. do not | |
212 | * install a MPOL_DEFAULT policy, nor the task or system default | |
213 | * mempolicy. | |
214 | */ | |
215 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | |
216 | ||
217 | /* | |
218 | * get_policy() op must add reference [mpol_get()] to any policy at | |
219 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
220 | * in mm/mempolicy.c will do this automatically. | |
221 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
222 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | |
223 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | |
224 | * must return NULL--i.e., do not "fallback" to task or system default | |
225 | * policy. | |
226 | */ | |
227 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | |
228 | unsigned long addr); | |
229 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, | |
230 | const nodemask_t *to, unsigned long flags); | |
231 | #endif | |
232 | /* called by sys_remap_file_pages() to populate non-linear mapping */ | |
233 | int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr, | |
234 | unsigned long size, pgoff_t pgoff); | |
235 | }; | |
236 | ||
237 | struct mmu_gather; | |
238 | struct inode; | |
239 | ||
240 | #define page_private(page) ((page)->private) | |
241 | #define set_page_private(page, v) ((page)->private = (v)) | |
242 | ||
243 | /* It's valid only if the page is free path or free_list */ | |
244 | static inline void set_freepage_migratetype(struct page *page, int migratetype) | |
245 | { | |
246 | page->index = migratetype; | |
247 | } | |
248 | ||
249 | /* It's valid only if the page is free path or free_list */ | |
250 | static inline int get_freepage_migratetype(struct page *page) | |
251 | { | |
252 | return page->index; | |
253 | } | |
254 | ||
255 | /* | |
256 | * FIXME: take this include out, include page-flags.h in | |
257 | * files which need it (119 of them) | |
258 | */ | |
259 | #include <linux/page-flags.h> | |
260 | #include <linux/huge_mm.h> | |
261 | ||
262 | /* | |
263 | * Methods to modify the page usage count. | |
264 | * | |
265 | * What counts for a page usage: | |
266 | * - cache mapping (page->mapping) | |
267 | * - private data (page->private) | |
268 | * - page mapped in a task's page tables, each mapping | |
269 | * is counted separately | |
270 | * | |
271 | * Also, many kernel routines increase the page count before a critical | |
272 | * routine so they can be sure the page doesn't go away from under them. | |
273 | */ | |
274 | ||
275 | /* | |
276 | * Drop a ref, return true if the refcount fell to zero (the page has no users) | |
277 | */ | |
278 | static inline int put_page_testzero(struct page *page) | |
279 | { | |
280 | VM_BUG_ON(atomic_read(&page->_count) == 0); | |
281 | return atomic_dec_and_test(&page->_count); | |
282 | } | |
283 | ||
284 | /* | |
285 | * Try to grab a ref unless the page has a refcount of zero, return false if | |
286 | * that is the case. | |
287 | */ | |
288 | static inline int get_page_unless_zero(struct page *page) | |
289 | { | |
290 | return atomic_inc_not_zero(&page->_count); | |
291 | } | |
292 | ||
293 | extern int page_is_ram(unsigned long pfn); | |
294 | ||
295 | /* Support for virtually mapped pages */ | |
296 | struct page *vmalloc_to_page(const void *addr); | |
297 | unsigned long vmalloc_to_pfn(const void *addr); | |
298 | ||
299 | /* | |
300 | * Determine if an address is within the vmalloc range | |
301 | * | |
302 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
303 | * is no special casing required. | |
304 | */ | |
305 | static inline int is_vmalloc_addr(const void *x) | |
306 | { | |
307 | #ifdef CONFIG_MMU | |
308 | unsigned long addr = (unsigned long)x; | |
309 | ||
310 | return addr >= VMALLOC_START && addr < VMALLOC_END; | |
311 | #else | |
312 | return 0; | |
313 | #endif | |
314 | } | |
315 | #ifdef CONFIG_MMU | |
316 | extern int is_vmalloc_or_module_addr(const void *x); | |
317 | #else | |
318 | static inline int is_vmalloc_or_module_addr(const void *x) | |
319 | { | |
320 | return 0; | |
321 | } | |
322 | #endif | |
323 | ||
324 | static inline void compound_lock(struct page *page) | |
325 | { | |
326 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
327 | VM_BUG_ON(PageSlab(page)); | |
328 | bit_spin_lock(PG_compound_lock, &page->flags); | |
329 | #endif | |
330 | } | |
331 | ||
332 | static inline void compound_unlock(struct page *page) | |
333 | { | |
334 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
335 | VM_BUG_ON(PageSlab(page)); | |
336 | bit_spin_unlock(PG_compound_lock, &page->flags); | |
337 | #endif | |
338 | } | |
339 | ||
340 | static inline unsigned long compound_lock_irqsave(struct page *page) | |
341 | { | |
342 | unsigned long uninitialized_var(flags); | |
343 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
344 | local_irq_save(flags); | |
345 | compound_lock(page); | |
346 | #endif | |
347 | return flags; | |
348 | } | |
349 | ||
350 | static inline void compound_unlock_irqrestore(struct page *page, | |
351 | unsigned long flags) | |
352 | { | |
353 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
354 | compound_unlock(page); | |
355 | local_irq_restore(flags); | |
356 | #endif | |
357 | } | |
358 | ||
359 | static inline struct page *compound_head(struct page *page) | |
360 | { | |
361 | if (unlikely(PageTail(page))) | |
362 | return page->first_page; | |
363 | return page; | |
364 | } | |
365 | ||
366 | /* | |
367 | * The atomic page->_mapcount, starts from -1: so that transitions | |
368 | * both from it and to it can be tracked, using atomic_inc_and_test | |
369 | * and atomic_add_negative(-1). | |
370 | */ | |
371 | static inline void page_mapcount_reset(struct page *page) | |
372 | { | |
373 | atomic_set(&(page)->_mapcount, -1); | |
374 | } | |
375 | ||
376 | static inline int page_mapcount(struct page *page) | |
377 | { | |
378 | return atomic_read(&(page)->_mapcount) + 1; | |
379 | } | |
380 | ||
381 | static inline int page_count(struct page *page) | |
382 | { | |
383 | return atomic_read(&compound_head(page)->_count); | |
384 | } | |
385 | ||
386 | static inline void get_huge_page_tail(struct page *page) | |
387 | { | |
388 | /* | |
389 | * __split_huge_page_refcount() cannot run | |
390 | * from under us. | |
391 | */ | |
392 | VM_BUG_ON(page_mapcount(page) < 0); | |
393 | VM_BUG_ON(atomic_read(&page->_count) != 0); | |
394 | atomic_inc(&page->_mapcount); | |
395 | } | |
396 | ||
397 | extern bool __get_page_tail(struct page *page); | |
398 | ||
399 | static inline void get_page(struct page *page) | |
400 | { | |
401 | if (unlikely(PageTail(page))) | |
402 | if (likely(__get_page_tail(page))) | |
403 | return; | |
404 | /* | |
405 | * Getting a normal page or the head of a compound page | |
406 | * requires to already have an elevated page->_count. | |
407 | */ | |
408 | VM_BUG_ON(atomic_read(&page->_count) <= 0); | |
409 | atomic_inc(&page->_count); | |
410 | } | |
411 | ||
412 | static inline struct page *virt_to_head_page(const void *x) | |
413 | { | |
414 | struct page *page = virt_to_page(x); | |
415 | return compound_head(page); | |
416 | } | |
417 | ||
418 | /* | |
419 | * Setup the page count before being freed into the page allocator for | |
420 | * the first time (boot or memory hotplug) | |
421 | */ | |
422 | static inline void init_page_count(struct page *page) | |
423 | { | |
424 | atomic_set(&page->_count, 1); | |
425 | } | |
426 | ||
427 | /* | |
428 | * PageBuddy() indicate that the page is free and in the buddy system | |
429 | * (see mm/page_alloc.c). | |
430 | * | |
431 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to | |
432 | * -2 so that an underflow of the page_mapcount() won't be mistaken | |
433 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very | |
434 | * efficiently by most CPU architectures. | |
435 | */ | |
436 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) | |
437 | ||
438 | static inline int PageBuddy(struct page *page) | |
439 | { | |
440 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; | |
441 | } | |
442 | ||
443 | static inline void __SetPageBuddy(struct page *page) | |
444 | { | |
445 | VM_BUG_ON(atomic_read(&page->_mapcount) != -1); | |
446 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); | |
447 | } | |
448 | ||
449 | static inline void __ClearPageBuddy(struct page *page) | |
450 | { | |
451 | VM_BUG_ON(!PageBuddy(page)); | |
452 | atomic_set(&page->_mapcount, -1); | |
453 | } | |
454 | ||
455 | void put_page(struct page *page); | |
456 | void put_pages_list(struct list_head *pages); | |
457 | ||
458 | void split_page(struct page *page, unsigned int order); | |
459 | int split_free_page(struct page *page); | |
460 | ||
461 | /* | |
462 | * Compound pages have a destructor function. Provide a | |
463 | * prototype for that function and accessor functions. | |
464 | * These are _only_ valid on the head of a PG_compound page. | |
465 | */ | |
466 | typedef void compound_page_dtor(struct page *); | |
467 | ||
468 | static inline void set_compound_page_dtor(struct page *page, | |
469 | compound_page_dtor *dtor) | |
470 | { | |
471 | page[1].lru.next = (void *)dtor; | |
472 | } | |
473 | ||
474 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
475 | { | |
476 | return (compound_page_dtor *)page[1].lru.next; | |
477 | } | |
478 | ||
479 | static inline int compound_order(struct page *page) | |
480 | { | |
481 | if (!PageHead(page)) | |
482 | return 0; | |
483 | return (unsigned long)page[1].lru.prev; | |
484 | } | |
485 | ||
486 | static inline int compound_trans_order(struct page *page) | |
487 | { | |
488 | int order; | |
489 | unsigned long flags; | |
490 | ||
491 | if (!PageHead(page)) | |
492 | return 0; | |
493 | ||
494 | flags = compound_lock_irqsave(page); | |
495 | order = compound_order(page); | |
496 | compound_unlock_irqrestore(page, flags); | |
497 | return order; | |
498 | } | |
499 | ||
500 | static inline void set_compound_order(struct page *page, unsigned long order) | |
501 | { | |
502 | page[1].lru.prev = (void *)order; | |
503 | } | |
504 | ||
505 | #ifdef CONFIG_MMU | |
506 | /* | |
507 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
508 | * servicing faults for write access. In the normal case, do always want | |
509 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
510 | * that do not have writing enabled, when used by access_process_vm. | |
511 | */ | |
512 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
513 | { | |
514 | if (likely(vma->vm_flags & VM_WRITE)) | |
515 | pte = pte_mkwrite(pte); | |
516 | return pte; | |
517 | } | |
518 | #endif | |
519 | ||
520 | /* | |
521 | * Multiple processes may "see" the same page. E.g. for untouched | |
522 | * mappings of /dev/null, all processes see the same page full of | |
523 | * zeroes, and text pages of executables and shared libraries have | |
524 | * only one copy in memory, at most, normally. | |
525 | * | |
526 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
527 | * page_count() == 0 means the page is free. page->lru is then used for | |
528 | * freelist management in the buddy allocator. | |
529 | * page_count() > 0 means the page has been allocated. | |
530 | * | |
531 | * Pages are allocated by the slab allocator in order to provide memory | |
532 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
533 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
534 | * unless a particular usage is carefully commented. (the responsibility of | |
535 | * freeing the kmalloc memory is the caller's, of course). | |
536 | * | |
537 | * A page may be used by anyone else who does a __get_free_page(). | |
538 | * In this case, page_count still tracks the references, and should only | |
539 | * be used through the normal accessor functions. The top bits of page->flags | |
540 | * and page->virtual store page management information, but all other fields | |
541 | * are unused and could be used privately, carefully. The management of this | |
542 | * page is the responsibility of the one who allocated it, and those who have | |
543 | * subsequently been given references to it. | |
544 | * | |
545 | * The other pages (we may call them "pagecache pages") are completely | |
546 | * managed by the Linux memory manager: I/O, buffers, swapping etc. | |
547 | * The following discussion applies only to them. | |
548 | * | |
549 | * A pagecache page contains an opaque `private' member, which belongs to the | |
550 | * page's address_space. Usually, this is the address of a circular list of | |
551 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
552 | * into the filesystem to release these pages. | |
553 | * | |
554 | * A page may belong to an inode's memory mapping. In this case, page->mapping | |
555 | * is the pointer to the inode, and page->index is the file offset of the page, | |
556 | * in units of PAGE_CACHE_SIZE. | |
557 | * | |
558 | * If pagecache pages are not associated with an inode, they are said to be | |
559 | * anonymous pages. These may become associated with the swapcache, and in that | |
560 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
561 | * | |
562 | * In either case (swapcache or inode backed), the pagecache itself holds one | |
563 | * reference to the page. Setting PG_private should also increment the | |
564 | * refcount. The each user mapping also has a reference to the page. | |
565 | * | |
566 | * The pagecache pages are stored in a per-mapping radix tree, which is | |
567 | * rooted at mapping->page_tree, and indexed by offset. | |
568 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
569 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
570 | * | |
571 | * All pagecache pages may be subject to I/O: | |
572 | * - inode pages may need to be read from disk, | |
573 | * - inode pages which have been modified and are MAP_SHARED may need | |
574 | * to be written back to the inode on disk, | |
575 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
576 | * modified may need to be swapped out to swap space and (later) to be read | |
577 | * back into memory. | |
578 | */ | |
579 | ||
580 | /* | |
581 | * The zone field is never updated after free_area_init_core() | |
582 | * sets it, so none of the operations on it need to be atomic. | |
583 | */ | |
584 | ||
585 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */ | |
586 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) | |
587 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) | |
588 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
589 | #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH) | |
590 | ||
591 | /* | |
592 | * Define the bit shifts to access each section. For non-existent | |
593 | * sections we define the shift as 0; that plus a 0 mask ensures | |
594 | * the compiler will optimise away reference to them. | |
595 | */ | |
596 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) | |
597 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
598 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
599 | #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0)) | |
600 | ||
601 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ | |
602 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
603 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
604 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ | |
605 | SECTIONS_PGOFF : ZONES_PGOFF) | |
606 | #else | |
607 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) | |
608 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ | |
609 | NODES_PGOFF : ZONES_PGOFF) | |
610 | #endif | |
611 | ||
612 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) | |
613 | ||
614 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
615 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
616 | #endif | |
617 | ||
618 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) | |
619 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
620 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
621 | #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1) | |
622 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) | |
623 | ||
624 | static inline enum zone_type page_zonenum(const struct page *page) | |
625 | { | |
626 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; | |
627 | } | |
628 | ||
629 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) | |
630 | #define SECTION_IN_PAGE_FLAGS | |
631 | #endif | |
632 | ||
633 | /* | |
634 | * The identification function is only used by the buddy allocator for | |
635 | * determining if two pages could be buddies. We are not really | |
636 | * identifying a zone since we could be using a the section number | |
637 | * id if we have not node id available in page flags. | |
638 | * We guarantee only that it will return the same value for two | |
639 | * combinable pages in a zone. | |
640 | */ | |
641 | static inline int page_zone_id(struct page *page) | |
642 | { | |
643 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; | |
644 | } | |
645 | ||
646 | static inline int zone_to_nid(struct zone *zone) | |
647 | { | |
648 | #ifdef CONFIG_NUMA | |
649 | return zone->node; | |
650 | #else | |
651 | return 0; | |
652 | #endif | |
653 | } | |
654 | ||
655 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
656 | extern int page_to_nid(const struct page *page); | |
657 | #else | |
658 | static inline int page_to_nid(const struct page *page) | |
659 | { | |
660 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; | |
661 | } | |
662 | #endif | |
663 | ||
664 | #ifdef CONFIG_NUMA_BALANCING | |
665 | #ifdef LAST_NID_NOT_IN_PAGE_FLAGS | |
666 | static inline int page_nid_xchg_last(struct page *page, int nid) | |
667 | { | |
668 | return xchg(&page->_last_nid, nid); | |
669 | } | |
670 | ||
671 | static inline int page_nid_last(struct page *page) | |
672 | { | |
673 | return page->_last_nid; | |
674 | } | |
675 | static inline void page_nid_reset_last(struct page *page) | |
676 | { | |
677 | page->_last_nid = -1; | |
678 | } | |
679 | #else | |
680 | static inline int page_nid_last(struct page *page) | |
681 | { | |
682 | return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK; | |
683 | } | |
684 | ||
685 | extern int page_nid_xchg_last(struct page *page, int nid); | |
686 | ||
687 | static inline void page_nid_reset_last(struct page *page) | |
688 | { | |
689 | int nid = (1 << LAST_NID_SHIFT) - 1; | |
690 | ||
691 | page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT); | |
692 | page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT; | |
693 | } | |
694 | #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */ | |
695 | #else | |
696 | static inline int page_nid_xchg_last(struct page *page, int nid) | |
697 | { | |
698 | return page_to_nid(page); | |
699 | } | |
700 | ||
701 | static inline int page_nid_last(struct page *page) | |
702 | { | |
703 | return page_to_nid(page); | |
704 | } | |
705 | ||
706 | static inline void page_nid_reset_last(struct page *page) | |
707 | { | |
708 | } | |
709 | #endif | |
710 | ||
711 | static inline struct zone *page_zone(const struct page *page) | |
712 | { | |
713 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
714 | } | |
715 | ||
716 | #ifdef SECTION_IN_PAGE_FLAGS | |
717 | static inline void set_page_section(struct page *page, unsigned long section) | |
718 | { | |
719 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
720 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
721 | } | |
722 | ||
723 | static inline unsigned long page_to_section(const struct page *page) | |
724 | { | |
725 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
726 | } | |
727 | #endif | |
728 | ||
729 | static inline void set_page_zone(struct page *page, enum zone_type zone) | |
730 | { | |
731 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
732 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
733 | } | |
734 | ||
735 | static inline void set_page_node(struct page *page, unsigned long node) | |
736 | { | |
737 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
738 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
739 | } | |
740 | ||
741 | static inline void set_page_links(struct page *page, enum zone_type zone, | |
742 | unsigned long node, unsigned long pfn) | |
743 | { | |
744 | set_page_zone(page, zone); | |
745 | set_page_node(page, node); | |
746 | #ifdef SECTION_IN_PAGE_FLAGS | |
747 | set_page_section(page, pfn_to_section_nr(pfn)); | |
748 | #endif | |
749 | } | |
750 | ||
751 | /* | |
752 | * Some inline functions in vmstat.h depend on page_zone() | |
753 | */ | |
754 | #include <linux/vmstat.h> | |
755 | ||
756 | static __always_inline void *lowmem_page_address(const struct page *page) | |
757 | { | |
758 | return __va(PFN_PHYS(page_to_pfn(page))); | |
759 | } | |
760 | ||
761 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
762 | #define HASHED_PAGE_VIRTUAL | |
763 | #endif | |
764 | ||
765 | #if defined(WANT_PAGE_VIRTUAL) | |
766 | #define page_address(page) ((page)->virtual) | |
767 | #define set_page_address(page, address) \ | |
768 | do { \ | |
769 | (page)->virtual = (address); \ | |
770 | } while(0) | |
771 | #define page_address_init() do { } while(0) | |
772 | #endif | |
773 | ||
774 | #if defined(HASHED_PAGE_VIRTUAL) | |
775 | void *page_address(const struct page *page); | |
776 | void set_page_address(struct page *page, void *virtual); | |
777 | void page_address_init(void); | |
778 | #endif | |
779 | ||
780 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
781 | #define page_address(page) lowmem_page_address(page) | |
782 | #define set_page_address(page, address) do { } while(0) | |
783 | #define page_address_init() do { } while(0) | |
784 | #endif | |
785 | ||
786 | /* | |
787 | * On an anonymous page mapped into a user virtual memory area, | |
788 | * page->mapping points to its anon_vma, not to a struct address_space; | |
789 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. | |
790 | * | |
791 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, | |
792 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; | |
793 | * and then page->mapping points, not to an anon_vma, but to a private | |
794 | * structure which KSM associates with that merged page. See ksm.h. | |
795 | * | |
796 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. | |
797 | * | |
798 | * Please note that, confusingly, "page_mapping" refers to the inode | |
799 | * address_space which maps the page from disk; whereas "page_mapped" | |
800 | * refers to user virtual address space into which the page is mapped. | |
801 | */ | |
802 | #define PAGE_MAPPING_ANON 1 | |
803 | #define PAGE_MAPPING_KSM 2 | |
804 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) | |
805 | ||
806 | extern struct address_space *page_mapping(struct page *page); | |
807 | ||
808 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ | |
809 | static inline void *page_rmapping(struct page *page) | |
810 | { | |
811 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); | |
812 | } | |
813 | ||
814 | extern struct address_space *__page_file_mapping(struct page *); | |
815 | ||
816 | static inline | |
817 | struct address_space *page_file_mapping(struct page *page) | |
818 | { | |
819 | if (unlikely(PageSwapCache(page))) | |
820 | return __page_file_mapping(page); | |
821 | ||
822 | return page->mapping; | |
823 | } | |
824 | ||
825 | static inline int PageAnon(struct page *page) | |
826 | { | |
827 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
828 | } | |
829 | ||
830 | /* | |
831 | * Return the pagecache index of the passed page. Regular pagecache pages | |
832 | * use ->index whereas swapcache pages use ->private | |
833 | */ | |
834 | static inline pgoff_t page_index(struct page *page) | |
835 | { | |
836 | if (unlikely(PageSwapCache(page))) | |
837 | return page_private(page); | |
838 | return page->index; | |
839 | } | |
840 | ||
841 | extern pgoff_t __page_file_index(struct page *page); | |
842 | ||
843 | /* | |
844 | * Return the file index of the page. Regular pagecache pages use ->index | |
845 | * whereas swapcache pages use swp_offset(->private) | |
846 | */ | |
847 | static inline pgoff_t page_file_index(struct page *page) | |
848 | { | |
849 | if (unlikely(PageSwapCache(page))) | |
850 | return __page_file_index(page); | |
851 | ||
852 | return page->index; | |
853 | } | |
854 | ||
855 | /* | |
856 | * Return true if this page is mapped into pagetables. | |
857 | */ | |
858 | static inline int page_mapped(struct page *page) | |
859 | { | |
860 | return atomic_read(&(page)->_mapcount) >= 0; | |
861 | } | |
862 | ||
863 | /* | |
864 | * Different kinds of faults, as returned by handle_mm_fault(). | |
865 | * Used to decide whether a process gets delivered SIGBUS or | |
866 | * just gets major/minor fault counters bumped up. | |
867 | */ | |
868 | ||
869 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ | |
870 | ||
871 | #define VM_FAULT_OOM 0x0001 | |
872 | #define VM_FAULT_SIGBUS 0x0002 | |
873 | #define VM_FAULT_MAJOR 0x0004 | |
874 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | |
875 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ | |
876 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ | |
877 | ||
878 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ | |
879 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | |
880 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ | |
881 | ||
882 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ | |
883 | ||
884 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ | |
885 | VM_FAULT_HWPOISON_LARGE) | |
886 | ||
887 | /* Encode hstate index for a hwpoisoned large page */ | |
888 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) | |
889 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) | |
890 | ||
891 | /* | |
892 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
893 | */ | |
894 | extern void pagefault_out_of_memory(void); | |
895 | ||
896 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | |
897 | ||
898 | /* | |
899 | * Flags passed to show_mem() and show_free_areas() to suppress output in | |
900 | * various contexts. | |
901 | */ | |
902 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ | |
903 | #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */ | |
904 | ||
905 | extern void show_free_areas(unsigned int flags); | |
906 | extern bool skip_free_areas_node(unsigned int flags, int nid); | |
907 | ||
908 | int shmem_zero_setup(struct vm_area_struct *); | |
909 | ||
910 | extern int can_do_mlock(void); | |
911 | extern int user_shm_lock(size_t, struct user_struct *); | |
912 | extern void user_shm_unlock(size_t, struct user_struct *); | |
913 | ||
914 | /* | |
915 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
916 | */ | |
917 | struct zap_details { | |
918 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | |
919 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
920 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
921 | pgoff_t last_index; /* Highest page->index to unmap */ | |
922 | }; | |
923 | ||
924 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, | |
925 | pte_t pte); | |
926 | ||
927 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, | |
928 | unsigned long size); | |
929 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, | |
930 | unsigned long size, struct zap_details *); | |
931 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, | |
932 | unsigned long start, unsigned long end); | |
933 | ||
934 | /** | |
935 | * mm_walk - callbacks for walk_page_range | |
936 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry | |
937 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry | |
938 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry | |
939 | * this handler is required to be able to handle | |
940 | * pmd_trans_huge() pmds. They may simply choose to | |
941 | * split_huge_page() instead of handling it explicitly. | |
942 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry | |
943 | * @pte_hole: if set, called for each hole at all levels | |
944 | * @hugetlb_entry: if set, called for each hugetlb entry | |
945 | * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry | |
946 | * is used. | |
947 | * | |
948 | * (see walk_page_range for more details) | |
949 | */ | |
950 | struct mm_walk { | |
951 | int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); | |
952 | int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); | |
953 | int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); | |
954 | int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); | |
955 | int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); | |
956 | int (*hugetlb_entry)(pte_t *, unsigned long, | |
957 | unsigned long, unsigned long, struct mm_walk *); | |
958 | struct mm_struct *mm; | |
959 | void *private; | |
960 | }; | |
961 | ||
962 | int walk_page_range(unsigned long addr, unsigned long end, | |
963 | struct mm_walk *walk); | |
964 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, | |
965 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
966 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, | |
967 | struct vm_area_struct *vma); | |
968 | void unmap_mapping_range(struct address_space *mapping, | |
969 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
970 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, | |
971 | unsigned long *pfn); | |
972 | int follow_phys(struct vm_area_struct *vma, unsigned long address, | |
973 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
974 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, | |
975 | void *buf, int len, int write); | |
976 | ||
977 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
978 | loff_t const holebegin, loff_t const holelen) | |
979 | { | |
980 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
981 | } | |
982 | ||
983 | extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); | |
984 | extern void truncate_setsize(struct inode *inode, loff_t newsize); | |
985 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); | |
986 | int truncate_inode_page(struct address_space *mapping, struct page *page); | |
987 | int generic_error_remove_page(struct address_space *mapping, struct page *page); | |
988 | int invalidate_inode_page(struct page *page); | |
989 | ||
990 | #ifdef CONFIG_MMU | |
991 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, | |
992 | unsigned long address, unsigned int flags); | |
993 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, | |
994 | unsigned long address, unsigned int fault_flags); | |
995 | #else | |
996 | static inline int handle_mm_fault(struct mm_struct *mm, | |
997 | struct vm_area_struct *vma, unsigned long address, | |
998 | unsigned int flags) | |
999 | { | |
1000 | /* should never happen if there's no MMU */ | |
1001 | BUG(); | |
1002 | return VM_FAULT_SIGBUS; | |
1003 | } | |
1004 | static inline int fixup_user_fault(struct task_struct *tsk, | |
1005 | struct mm_struct *mm, unsigned long address, | |
1006 | unsigned int fault_flags) | |
1007 | { | |
1008 | /* should never happen if there's no MMU */ | |
1009 | BUG(); | |
1010 | return -EFAULT; | |
1011 | } | |
1012 | #endif | |
1013 | ||
1014 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
1015 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, | |
1016 | void *buf, int len, int write); | |
1017 | ||
1018 | long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |
1019 | unsigned long start, unsigned long nr_pages, | |
1020 | unsigned int foll_flags, struct page **pages, | |
1021 | struct vm_area_struct **vmas, int *nonblocking); | |
1022 | long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |
1023 | unsigned long start, unsigned long nr_pages, | |
1024 | int write, int force, struct page **pages, | |
1025 | struct vm_area_struct **vmas); | |
1026 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1027 | struct page **pages); | |
1028 | struct kvec; | |
1029 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | |
1030 | struct page **pages); | |
1031 | int get_kernel_page(unsigned long start, int write, struct page **pages); | |
1032 | struct page *get_dump_page(unsigned long addr); | |
1033 | ||
1034 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); | |
1035 | extern void do_invalidatepage(struct page *page, unsigned long offset); | |
1036 | ||
1037 | int __set_page_dirty_nobuffers(struct page *page); | |
1038 | int __set_page_dirty_no_writeback(struct page *page); | |
1039 | int redirty_page_for_writepage(struct writeback_control *wbc, | |
1040 | struct page *page); | |
1041 | void account_page_dirtied(struct page *page, struct address_space *mapping); | |
1042 | void account_page_writeback(struct page *page); | |
1043 | int set_page_dirty(struct page *page); | |
1044 | int set_page_dirty_lock(struct page *page); | |
1045 | int clear_page_dirty_for_io(struct page *page); | |
1046 | ||
1047 | /* Is the vma a continuation of the stack vma above it? */ | |
1048 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) | |
1049 | { | |
1050 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); | |
1051 | } | |
1052 | ||
1053 | static inline int stack_guard_page_start(struct vm_area_struct *vma, | |
1054 | unsigned long addr) | |
1055 | { | |
1056 | return (vma->vm_flags & VM_GROWSDOWN) && | |
1057 | (vma->vm_start == addr) && | |
1058 | !vma_growsdown(vma->vm_prev, addr); | |
1059 | } | |
1060 | ||
1061 | /* Is the vma a continuation of the stack vma below it? */ | |
1062 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) | |
1063 | { | |
1064 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); | |
1065 | } | |
1066 | ||
1067 | static inline int stack_guard_page_end(struct vm_area_struct *vma, | |
1068 | unsigned long addr) | |
1069 | { | |
1070 | return (vma->vm_flags & VM_GROWSUP) && | |
1071 | (vma->vm_end == addr) && | |
1072 | !vma_growsup(vma->vm_next, addr); | |
1073 | } | |
1074 | ||
1075 | extern pid_t | |
1076 | vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group); | |
1077 | ||
1078 | extern unsigned long move_page_tables(struct vm_area_struct *vma, | |
1079 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
1080 | unsigned long new_addr, unsigned long len, | |
1081 | bool need_rmap_locks); | |
1082 | extern unsigned long do_mremap(unsigned long addr, | |
1083 | unsigned long old_len, unsigned long new_len, | |
1084 | unsigned long flags, unsigned long new_addr); | |
1085 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, | |
1086 | unsigned long end, pgprot_t newprot, | |
1087 | int dirty_accountable, int prot_numa); | |
1088 | extern int mprotect_fixup(struct vm_area_struct *vma, | |
1089 | struct vm_area_struct **pprev, unsigned long start, | |
1090 | unsigned long end, unsigned long newflags); | |
1091 | ||
1092 | /* | |
1093 | * doesn't attempt to fault and will return short. | |
1094 | */ | |
1095 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1096 | struct page **pages); | |
1097 | /* | |
1098 | * per-process(per-mm_struct) statistics. | |
1099 | */ | |
1100 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) | |
1101 | { | |
1102 | long val = atomic_long_read(&mm->rss_stat.count[member]); | |
1103 | ||
1104 | #ifdef SPLIT_RSS_COUNTING | |
1105 | /* | |
1106 | * counter is updated in asynchronous manner and may go to minus. | |
1107 | * But it's never be expected number for users. | |
1108 | */ | |
1109 | if (val < 0) | |
1110 | val = 0; | |
1111 | #endif | |
1112 | return (unsigned long)val; | |
1113 | } | |
1114 | ||
1115 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1116 | { | |
1117 | atomic_long_add(value, &mm->rss_stat.count[member]); | |
1118 | } | |
1119 | ||
1120 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1121 | { | |
1122 | atomic_long_inc(&mm->rss_stat.count[member]); | |
1123 | } | |
1124 | ||
1125 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1126 | { | |
1127 | atomic_long_dec(&mm->rss_stat.count[member]); | |
1128 | } | |
1129 | ||
1130 | static inline unsigned long get_mm_rss(struct mm_struct *mm) | |
1131 | { | |
1132 | return get_mm_counter(mm, MM_FILEPAGES) + | |
1133 | get_mm_counter(mm, MM_ANONPAGES); | |
1134 | } | |
1135 | ||
1136 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1137 | { | |
1138 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1139 | } | |
1140 | ||
1141 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1142 | { | |
1143 | return max(mm->hiwater_vm, mm->total_vm); | |
1144 | } | |
1145 | ||
1146 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1147 | { | |
1148 | unsigned long _rss = get_mm_rss(mm); | |
1149 | ||
1150 | if ((mm)->hiwater_rss < _rss) | |
1151 | (mm)->hiwater_rss = _rss; | |
1152 | } | |
1153 | ||
1154 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1155 | { | |
1156 | if (mm->hiwater_vm < mm->total_vm) | |
1157 | mm->hiwater_vm = mm->total_vm; | |
1158 | } | |
1159 | ||
1160 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, | |
1161 | struct mm_struct *mm) | |
1162 | { | |
1163 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
1164 | ||
1165 | if (*maxrss < hiwater_rss) | |
1166 | *maxrss = hiwater_rss; | |
1167 | } | |
1168 | ||
1169 | #if defined(SPLIT_RSS_COUNTING) | |
1170 | void sync_mm_rss(struct mm_struct *mm); | |
1171 | #else | |
1172 | static inline void sync_mm_rss(struct mm_struct *mm) | |
1173 | { | |
1174 | } | |
1175 | #endif | |
1176 | ||
1177 | int vma_wants_writenotify(struct vm_area_struct *vma); | |
1178 | ||
1179 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1180 | spinlock_t **ptl); | |
1181 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1182 | spinlock_t **ptl) | |
1183 | { | |
1184 | pte_t *ptep; | |
1185 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
1186 | return ptep; | |
1187 | } | |
1188 | ||
1189 | #ifdef __PAGETABLE_PUD_FOLDED | |
1190 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
1191 | unsigned long address) | |
1192 | { | |
1193 | return 0; | |
1194 | } | |
1195 | #else | |
1196 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
1197 | #endif | |
1198 | ||
1199 | #ifdef __PAGETABLE_PMD_FOLDED | |
1200 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
1201 | unsigned long address) | |
1202 | { | |
1203 | return 0; | |
1204 | } | |
1205 | #else | |
1206 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); | |
1207 | #endif | |
1208 | ||
1209 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, | |
1210 | pmd_t *pmd, unsigned long address); | |
1211 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | |
1212 | ||
1213 | /* | |
1214 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
1215 | * Remove it when 4level-fixup.h has been removed. | |
1216 | */ | |
1217 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) | |
1218 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) | |
1219 | { | |
1220 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? | |
1221 | NULL: pud_offset(pgd, address); | |
1222 | } | |
1223 | ||
1224 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
1225 | { | |
1226 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? | |
1227 | NULL: pmd_offset(pud, address); | |
1228 | } | |
1229 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ | |
1230 | ||
1231 | #if USE_SPLIT_PTLOCKS | |
1232 | /* | |
1233 | * We tuck a spinlock to guard each pagetable page into its struct page, | |
1234 | * at page->private, with BUILD_BUG_ON to make sure that this will not | |
1235 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). | |
1236 | * When freeing, reset page->mapping so free_pages_check won't complain. | |
1237 | */ | |
1238 | #define __pte_lockptr(page) &((page)->ptl) | |
1239 | #define pte_lock_init(_page) do { \ | |
1240 | spin_lock_init(__pte_lockptr(_page)); \ | |
1241 | } while (0) | |
1242 | #define pte_lock_deinit(page) ((page)->mapping = NULL) | |
1243 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) | |
1244 | #else /* !USE_SPLIT_PTLOCKS */ | |
1245 | /* | |
1246 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
1247 | */ | |
1248 | #define pte_lock_init(page) do {} while (0) | |
1249 | #define pte_lock_deinit(page) do {} while (0) | |
1250 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) | |
1251 | #endif /* USE_SPLIT_PTLOCKS */ | |
1252 | ||
1253 | static inline void pgtable_page_ctor(struct page *page) | |
1254 | { | |
1255 | pte_lock_init(page); | |
1256 | inc_zone_page_state(page, NR_PAGETABLE); | |
1257 | } | |
1258 | ||
1259 | static inline void pgtable_page_dtor(struct page *page) | |
1260 | { | |
1261 | pte_lock_deinit(page); | |
1262 | dec_zone_page_state(page, NR_PAGETABLE); | |
1263 | } | |
1264 | ||
1265 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ | |
1266 | ({ \ | |
1267 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ | |
1268 | pte_t *__pte = pte_offset_map(pmd, address); \ | |
1269 | *(ptlp) = __ptl; \ | |
1270 | spin_lock(__ptl); \ | |
1271 | __pte; \ | |
1272 | }) | |
1273 | ||
1274 | #define pte_unmap_unlock(pte, ptl) do { \ | |
1275 | spin_unlock(ptl); \ | |
1276 | pte_unmap(pte); \ | |
1277 | } while (0) | |
1278 | ||
1279 | #define pte_alloc_map(mm, vma, pmd, address) \ | |
1280 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ | |
1281 | pmd, address))? \ | |
1282 | NULL: pte_offset_map(pmd, address)) | |
1283 | ||
1284 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ | |
1285 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ | |
1286 | pmd, address))? \ | |
1287 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) | |
1288 | ||
1289 | #define pte_alloc_kernel(pmd, address) \ | |
1290 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | |
1291 | NULL: pte_offset_kernel(pmd, address)) | |
1292 | ||
1293 | extern void free_area_init(unsigned long * zones_size); | |
1294 | extern void free_area_init_node(int nid, unsigned long * zones_size, | |
1295 | unsigned long zone_start_pfn, unsigned long *zholes_size); | |
1296 | extern void free_initmem(void); | |
1297 | ||
1298 | /* | |
1299 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
1300 | * into the buddy system. The freed pages will be poisoned with pattern | |
1301 | * "poison" if it's non-zero. | |
1302 | * Return pages freed into the buddy system. | |
1303 | */ | |
1304 | extern unsigned long free_reserved_area(unsigned long start, unsigned long end, | |
1305 | int poison, char *s); | |
1306 | ||
1307 | static inline void adjust_managed_page_count(struct page *page, long count) | |
1308 | { | |
1309 | totalram_pages += count; | |
1310 | } | |
1311 | ||
1312 | /* Free the reserved page into the buddy system, so it gets managed. */ | |
1313 | static inline void __free_reserved_page(struct page *page) | |
1314 | { | |
1315 | ClearPageReserved(page); | |
1316 | init_page_count(page); | |
1317 | __free_page(page); | |
1318 | } | |
1319 | ||
1320 | static inline void free_reserved_page(struct page *page) | |
1321 | { | |
1322 | __free_reserved_page(page); | |
1323 | adjust_managed_page_count(page, 1); | |
1324 | } | |
1325 | ||
1326 | static inline void mark_page_reserved(struct page *page) | |
1327 | { | |
1328 | SetPageReserved(page); | |
1329 | adjust_managed_page_count(page, -1); | |
1330 | } | |
1331 | ||
1332 | /* | |
1333 | * Default method to free all the __init memory into the buddy system. | |
1334 | * The freed pages will be poisoned with pattern "poison" if it is | |
1335 | * non-zero. Return pages freed into the buddy system. | |
1336 | */ | |
1337 | static inline unsigned long free_initmem_default(int poison) | |
1338 | { | |
1339 | extern char __init_begin[], __init_end[]; | |
1340 | ||
1341 | return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin) , | |
1342 | ((unsigned long)&__init_end) & PAGE_MASK, | |
1343 | poison, "unused kernel"); | |
1344 | } | |
1345 | ||
1346 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP | |
1347 | /* | |
1348 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its | |
1349 | * zones, allocate the backing mem_map and account for memory holes in a more | |
1350 | * architecture independent manner. This is a substitute for creating the | |
1351 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
1352 | * free_area_init_node() | |
1353 | * | |
1354 | * An architecture is expected to register range of page frames backed by | |
1355 | * physical memory with memblock_add[_node]() before calling | |
1356 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
1357 | * usage, an architecture is expected to do something like | |
1358 | * | |
1359 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
1360 | * max_highmem_pfn}; | |
1361 | * for_each_valid_physical_page_range() | |
1362 | * memblock_add_node(base, size, nid) | |
1363 | * free_area_init_nodes(max_zone_pfns); | |
1364 | * | |
1365 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each | |
1366 | * registered physical page range. Similarly | |
1367 | * sparse_memory_present_with_active_regions() calls memory_present() for | |
1368 | * each range when SPARSEMEM is enabled. | |
1369 | * | |
1370 | * See mm/page_alloc.c for more information on each function exposed by | |
1371 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. | |
1372 | */ | |
1373 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1374 | unsigned long node_map_pfn_alignment(void); | |
1375 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, | |
1376 | unsigned long end_pfn); | |
1377 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | |
1378 | unsigned long end_pfn); | |
1379 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1380 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1381 | extern unsigned long find_min_pfn_with_active_regions(void); | |
1382 | extern void free_bootmem_with_active_regions(int nid, | |
1383 | unsigned long max_low_pfn); | |
1384 | extern void sparse_memory_present_with_active_regions(int nid); | |
1385 | ||
1386 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
1387 | ||
1388 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ | |
1389 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) | |
1390 | static inline int __early_pfn_to_nid(unsigned long pfn) | |
1391 | { | |
1392 | return 0; | |
1393 | } | |
1394 | #else | |
1395 | /* please see mm/page_alloc.c */ | |
1396 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
1397 | #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID | |
1398 | /* there is a per-arch backend function. */ | |
1399 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); | |
1400 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ | |
1401 | #endif | |
1402 | ||
1403 | extern void set_dma_reserve(unsigned long new_dma_reserve); | |
1404 | extern void memmap_init_zone(unsigned long, int, unsigned long, | |
1405 | unsigned long, enum memmap_context); | |
1406 | extern void setup_per_zone_wmarks(void); | |
1407 | extern int __meminit init_per_zone_wmark_min(void); | |
1408 | extern void mem_init(void); | |
1409 | extern void __init mmap_init(void); | |
1410 | extern void show_mem(unsigned int flags); | |
1411 | extern void si_meminfo(struct sysinfo * val); | |
1412 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
1413 | ||
1414 | extern __printf(3, 4) | |
1415 | void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); | |
1416 | ||
1417 | extern void setup_per_cpu_pageset(void); | |
1418 | ||
1419 | extern void zone_pcp_update(struct zone *zone); | |
1420 | extern void zone_pcp_reset(struct zone *zone); | |
1421 | ||
1422 | /* page_alloc.c */ | |
1423 | extern int min_free_kbytes; | |
1424 | ||
1425 | /* nommu.c */ | |
1426 | extern atomic_long_t mmap_pages_allocated; | |
1427 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); | |
1428 | ||
1429 | /* interval_tree.c */ | |
1430 | void vma_interval_tree_insert(struct vm_area_struct *node, | |
1431 | struct rb_root *root); | |
1432 | void vma_interval_tree_insert_after(struct vm_area_struct *node, | |
1433 | struct vm_area_struct *prev, | |
1434 | struct rb_root *root); | |
1435 | void vma_interval_tree_remove(struct vm_area_struct *node, | |
1436 | struct rb_root *root); | |
1437 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, | |
1438 | unsigned long start, unsigned long last); | |
1439 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
1440 | unsigned long start, unsigned long last); | |
1441 | ||
1442 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
1443 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
1444 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1445 | ||
1446 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
1447 | struct list_head *list) | |
1448 | { | |
1449 | list_add_tail(&vma->shared.nonlinear, list); | |
1450 | } | |
1451 | ||
1452 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, | |
1453 | struct rb_root *root); | |
1454 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, | |
1455 | struct rb_root *root); | |
1456 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( | |
1457 | struct rb_root *root, unsigned long start, unsigned long last); | |
1458 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( | |
1459 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
1460 | #ifdef CONFIG_DEBUG_VM_RB | |
1461 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
1462 | #endif | |
1463 | ||
1464 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
1465 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
1466 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
1467 | ||
1468 | /* mmap.c */ | |
1469 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); | |
1470 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1471 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | |
1472 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1473 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1474 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1475 | struct mempolicy *); | |
1476 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1477 | extern int split_vma(struct mm_struct *, | |
1478 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1479 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1480 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1481 | struct rb_node **, struct rb_node *); | |
1482 | extern void unlink_file_vma(struct vm_area_struct *); | |
1483 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, | |
1484 | unsigned long addr, unsigned long len, pgoff_t pgoff, | |
1485 | bool *need_rmap_locks); | |
1486 | extern void exit_mmap(struct mm_struct *); | |
1487 | ||
1488 | extern int mm_take_all_locks(struct mm_struct *mm); | |
1489 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
1490 | ||
1491 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); | |
1492 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | |
1493 | ||
1494 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); | |
1495 | extern int install_special_mapping(struct mm_struct *mm, | |
1496 | unsigned long addr, unsigned long len, | |
1497 | unsigned long flags, struct page **pages); | |
1498 | ||
1499 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
1500 | ||
1501 | extern unsigned long mmap_region(struct file *file, unsigned long addr, | |
1502 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); | |
1503 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
1504 | unsigned long len, unsigned long prot, unsigned long flags, | |
1505 | unsigned long pgoff, unsigned long *populate); | |
1506 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
1507 | ||
1508 | #ifdef CONFIG_MMU | |
1509 | extern int __mm_populate(unsigned long addr, unsigned long len, | |
1510 | int ignore_errors); | |
1511 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
1512 | { | |
1513 | /* Ignore errors */ | |
1514 | (void) __mm_populate(addr, len, 1); | |
1515 | } | |
1516 | #else | |
1517 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
1518 | #endif | |
1519 | ||
1520 | /* These take the mm semaphore themselves */ | |
1521 | extern unsigned long vm_brk(unsigned long, unsigned long); | |
1522 | extern int vm_munmap(unsigned long, size_t); | |
1523 | extern unsigned long vm_mmap(struct file *, unsigned long, | |
1524 | unsigned long, unsigned long, | |
1525 | unsigned long, unsigned long); | |
1526 | ||
1527 | struct vm_unmapped_area_info { | |
1528 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
1529 | unsigned long flags; | |
1530 | unsigned long length; | |
1531 | unsigned long low_limit; | |
1532 | unsigned long high_limit; | |
1533 | unsigned long align_mask; | |
1534 | unsigned long align_offset; | |
1535 | }; | |
1536 | ||
1537 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); | |
1538 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); | |
1539 | ||
1540 | /* | |
1541 | * Search for an unmapped address range. | |
1542 | * | |
1543 | * We are looking for a range that: | |
1544 | * - does not intersect with any VMA; | |
1545 | * - is contained within the [low_limit, high_limit) interval; | |
1546 | * - is at least the desired size. | |
1547 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) | |
1548 | */ | |
1549 | static inline unsigned long | |
1550 | vm_unmapped_area(struct vm_unmapped_area_info *info) | |
1551 | { | |
1552 | if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN)) | |
1553 | return unmapped_area(info); | |
1554 | else | |
1555 | return unmapped_area_topdown(info); | |
1556 | } | |
1557 | ||
1558 | /* truncate.c */ | |
1559 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
1560 | extern void truncate_inode_pages_range(struct address_space *, | |
1561 | loff_t lstart, loff_t lend); | |
1562 | ||
1563 | /* generic vm_area_ops exported for stackable file systems */ | |
1564 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); | |
1565 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); | |
1566 | ||
1567 | /* mm/page-writeback.c */ | |
1568 | int write_one_page(struct page *page, int wait); | |
1569 | void task_dirty_inc(struct task_struct *tsk); | |
1570 | ||
1571 | /* readahead.c */ | |
1572 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
1573 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
1574 | ||
1575 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
1576 | pgoff_t offset, unsigned long nr_to_read); | |
1577 | ||
1578 | void page_cache_sync_readahead(struct address_space *mapping, | |
1579 | struct file_ra_state *ra, | |
1580 | struct file *filp, | |
1581 | pgoff_t offset, | |
1582 | unsigned long size); | |
1583 | ||
1584 | void page_cache_async_readahead(struct address_space *mapping, | |
1585 | struct file_ra_state *ra, | |
1586 | struct file *filp, | |
1587 | struct page *pg, | |
1588 | pgoff_t offset, | |
1589 | unsigned long size); | |
1590 | ||
1591 | unsigned long max_sane_readahead(unsigned long nr); | |
1592 | unsigned long ra_submit(struct file_ra_state *ra, | |
1593 | struct address_space *mapping, | |
1594 | struct file *filp); | |
1595 | ||
1596 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ | |
1597 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); | |
1598 | ||
1599 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | |
1600 | extern int expand_downwards(struct vm_area_struct *vma, | |
1601 | unsigned long address); | |
1602 | #if VM_GROWSUP | |
1603 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); | |
1604 | #else | |
1605 | #define expand_upwards(vma, address) do { } while (0) | |
1606 | #endif | |
1607 | ||
1608 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
1609 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
1610 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
1611 | struct vm_area_struct **pprev); | |
1612 | ||
1613 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
1614 | NULL if none. Assume start_addr < end_addr. */ | |
1615 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
1616 | { | |
1617 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
1618 | ||
1619 | if (vma && end_addr <= vma->vm_start) | |
1620 | vma = NULL; | |
1621 | return vma; | |
1622 | } | |
1623 | ||
1624 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
1625 | { | |
1626 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
1627 | } | |
1628 | ||
1629 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ | |
1630 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
1631 | unsigned long vm_start, unsigned long vm_end) | |
1632 | { | |
1633 | struct vm_area_struct *vma = find_vma(mm, vm_start); | |
1634 | ||
1635 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
1636 | vma = NULL; | |
1637 | ||
1638 | return vma; | |
1639 | } | |
1640 | ||
1641 | #ifdef CONFIG_MMU | |
1642 | pgprot_t vm_get_page_prot(unsigned long vm_flags); | |
1643 | #else | |
1644 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
1645 | { | |
1646 | return __pgprot(0); | |
1647 | } | |
1648 | #endif | |
1649 | ||
1650 | #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE | |
1651 | unsigned long change_prot_numa(struct vm_area_struct *vma, | |
1652 | unsigned long start, unsigned long end); | |
1653 | #endif | |
1654 | ||
1655 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); | |
1656 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | |
1657 | unsigned long pfn, unsigned long size, pgprot_t); | |
1658 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); | |
1659 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, | |
1660 | unsigned long pfn); | |
1661 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, | |
1662 | unsigned long pfn); | |
1663 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); | |
1664 | ||
1665 | ||
1666 | struct page *follow_page_mask(struct vm_area_struct *vma, | |
1667 | unsigned long address, unsigned int foll_flags, | |
1668 | unsigned int *page_mask); | |
1669 | ||
1670 | static inline struct page *follow_page(struct vm_area_struct *vma, | |
1671 | unsigned long address, unsigned int foll_flags) | |
1672 | { | |
1673 | unsigned int unused_page_mask; | |
1674 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); | |
1675 | } | |
1676 | ||
1677 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
1678 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
1679 | #define FOLL_GET 0x04 /* do get_page on page */ | |
1680 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ | |
1681 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ | |
1682 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO | |
1683 | * and return without waiting upon it */ | |
1684 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ | |
1685 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ | |
1686 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ | |
1687 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ | |
1688 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ | |
1689 | ||
1690 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, | |
1691 | void *data); | |
1692 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
1693 | unsigned long size, pte_fn_t fn, void *data); | |
1694 | ||
1695 | #ifdef CONFIG_PROC_FS | |
1696 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); | |
1697 | #else | |
1698 | static inline void vm_stat_account(struct mm_struct *mm, | |
1699 | unsigned long flags, struct file *file, long pages) | |
1700 | { | |
1701 | mm->total_vm += pages; | |
1702 | } | |
1703 | #endif /* CONFIG_PROC_FS */ | |
1704 | ||
1705 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
1706 | extern void kernel_map_pages(struct page *page, int numpages, int enable); | |
1707 | #ifdef CONFIG_HIBERNATION | |
1708 | extern bool kernel_page_present(struct page *page); | |
1709 | #endif /* CONFIG_HIBERNATION */ | |
1710 | #else | |
1711 | static inline void | |
1712 | kernel_map_pages(struct page *page, int numpages, int enable) {} | |
1713 | #ifdef CONFIG_HIBERNATION | |
1714 | static inline bool kernel_page_present(struct page *page) { return true; } | |
1715 | #endif /* CONFIG_HIBERNATION */ | |
1716 | #endif | |
1717 | ||
1718 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); | |
1719 | #ifdef __HAVE_ARCH_GATE_AREA | |
1720 | int in_gate_area_no_mm(unsigned long addr); | |
1721 | int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
1722 | #else | |
1723 | int in_gate_area_no_mm(unsigned long addr); | |
1724 | #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) | |
1725 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
1726 | ||
1727 | int drop_caches_sysctl_handler(struct ctl_table *, int, | |
1728 | void __user *, size_t *, loff_t *); | |
1729 | unsigned long shrink_slab(struct shrink_control *shrink, | |
1730 | unsigned long nr_pages_scanned, | |
1731 | unsigned long lru_pages); | |
1732 | ||
1733 | #ifndef CONFIG_MMU | |
1734 | #define randomize_va_space 0 | |
1735 | #else | |
1736 | extern int randomize_va_space; | |
1737 | #endif | |
1738 | ||
1739 | const char * arch_vma_name(struct vm_area_struct *vma); | |
1740 | void print_vma_addr(char *prefix, unsigned long rip); | |
1741 | ||
1742 | void sparse_mem_maps_populate_node(struct page **map_map, | |
1743 | unsigned long pnum_begin, | |
1744 | unsigned long pnum_end, | |
1745 | unsigned long map_count, | |
1746 | int nodeid); | |
1747 | ||
1748 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); | |
1749 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); | |
1750 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); | |
1751 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); | |
1752 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | |
1753 | void *vmemmap_alloc_block(unsigned long size, int node); | |
1754 | void *vmemmap_alloc_block_buf(unsigned long size, int node); | |
1755 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); | |
1756 | int vmemmap_populate_basepages(struct page *start_page, | |
1757 | unsigned long pages, int node); | |
1758 | int vmemmap_populate(struct page *start_page, unsigned long pages, int node); | |
1759 | void vmemmap_populate_print_last(void); | |
1760 | #ifdef CONFIG_MEMORY_HOTPLUG | |
1761 | void vmemmap_free(struct page *memmap, unsigned long nr_pages); | |
1762 | #endif | |
1763 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, | |
1764 | unsigned long size); | |
1765 | ||
1766 | enum mf_flags { | |
1767 | MF_COUNT_INCREASED = 1 << 0, | |
1768 | MF_ACTION_REQUIRED = 1 << 1, | |
1769 | MF_MUST_KILL = 1 << 2, | |
1770 | }; | |
1771 | extern int memory_failure(unsigned long pfn, int trapno, int flags); | |
1772 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); | |
1773 | extern int unpoison_memory(unsigned long pfn); | |
1774 | extern int sysctl_memory_failure_early_kill; | |
1775 | extern int sysctl_memory_failure_recovery; | |
1776 | extern void shake_page(struct page *p, int access); | |
1777 | extern atomic_long_t num_poisoned_pages; | |
1778 | extern int soft_offline_page(struct page *page, int flags); | |
1779 | ||
1780 | extern void dump_page(struct page *page); | |
1781 | ||
1782 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) | |
1783 | extern void clear_huge_page(struct page *page, | |
1784 | unsigned long addr, | |
1785 | unsigned int pages_per_huge_page); | |
1786 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
1787 | unsigned long addr, struct vm_area_struct *vma, | |
1788 | unsigned int pages_per_huge_page); | |
1789 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ | |
1790 | ||
1791 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
1792 | extern unsigned int _debug_guardpage_minorder; | |
1793 | ||
1794 | static inline unsigned int debug_guardpage_minorder(void) | |
1795 | { | |
1796 | return _debug_guardpage_minorder; | |
1797 | } | |
1798 | ||
1799 | static inline bool page_is_guard(struct page *page) | |
1800 | { | |
1801 | return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags); | |
1802 | } | |
1803 | #else | |
1804 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
1805 | static inline bool page_is_guard(struct page *page) { return false; } | |
1806 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
1807 | ||
1808 | #endif /* __KERNEL__ */ | |
1809 | #endif /* _LINUX_MM_H */ |