]>
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> | |
9a11b49a | 13 | #include <linux/debug_locks.h> |
5b99cd0e | 14 | #include <linux/mm_types.h> |
08677214 | 15 | #include <linux/range.h> |
c6f6b596 | 16 | #include <linux/pfn.h> |
e9da73d6 | 17 | #include <linux/bit_spinlock.h> |
1da177e4 LT |
18 | |
19 | struct mempolicy; | |
20 | struct anon_vma; | |
4e950f6f | 21 | struct file_ra_state; |
e8edc6e0 | 22 | struct user_struct; |
4e950f6f | 23 | struct writeback_control; |
1da177e4 LT |
24 | |
25 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ | |
26 | extern unsigned long max_mapnr; | |
27 | #endif | |
28 | ||
29 | extern unsigned long num_physpages; | |
4481374c | 30 | extern unsigned long totalram_pages; |
1da177e4 | 31 | extern void * high_memory; |
1da177e4 LT |
32 | extern int page_cluster; |
33 | ||
34 | #ifdef CONFIG_SYSCTL | |
35 | extern int sysctl_legacy_va_layout; | |
36 | #else | |
37 | #define sysctl_legacy_va_layout 0 | |
38 | #endif | |
39 | ||
40 | #include <asm/page.h> | |
41 | #include <asm/pgtable.h> | |
42 | #include <asm/processor.h> | |
1da177e4 | 43 | |
1da177e4 LT |
44 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
45 | ||
27ac792c AR |
46 | /* to align the pointer to the (next) page boundary */ |
47 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
48 | ||
1da177e4 LT |
49 | /* |
50 | * Linux kernel virtual memory manager primitives. | |
51 | * The idea being to have a "virtual" mm in the same way | |
52 | * we have a virtual fs - giving a cleaner interface to the | |
53 | * mm details, and allowing different kinds of memory mappings | |
54 | * (from shared memory to executable loading to arbitrary | |
55 | * mmap() functions). | |
56 | */ | |
57 | ||
c43692e8 CL |
58 | extern struct kmem_cache *vm_area_cachep; |
59 | ||
1da177e4 | 60 | #ifndef CONFIG_MMU |
8feae131 DH |
61 | extern struct rb_root nommu_region_tree; |
62 | extern struct rw_semaphore nommu_region_sem; | |
1da177e4 LT |
63 | |
64 | extern unsigned int kobjsize(const void *objp); | |
65 | #endif | |
66 | ||
67 | /* | |
605d9288 | 68 | * vm_flags in vm_area_struct, see mm_types.h. |
1da177e4 LT |
69 | */ |
70 | #define VM_READ 0x00000001 /* currently active flags */ | |
71 | #define VM_WRITE 0x00000002 | |
72 | #define VM_EXEC 0x00000004 | |
73 | #define VM_SHARED 0x00000008 | |
74 | ||
7e2cff42 | 75 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
1da177e4 LT |
76 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
77 | #define VM_MAYWRITE 0x00000020 | |
78 | #define VM_MAYEXEC 0x00000040 | |
79 | #define VM_MAYSHARE 0x00000080 | |
80 | ||
81 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
8ca3eb08 | 82 | #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64) |
1da177e4 | 83 | #define VM_GROWSUP 0x00000200 |
8ca3eb08 TL |
84 | #else |
85 | #define VM_GROWSUP 0x00000000 | |
a664b2d8 | 86 | #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */ |
8ca3eb08 | 87 | #endif |
6aab341e | 88 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
1da177e4 LT |
89 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
90 | ||
91 | #define VM_EXECUTABLE 0x00001000 | |
92 | #define VM_LOCKED 0x00002000 | |
93 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
94 | ||
95 | /* Used by sys_madvise() */ | |
96 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
97 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
98 | ||
99 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
100 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
0b14c179 | 101 | #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ |
1da177e4 | 102 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
cdfd4325 | 103 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ |
1da177e4 LT |
104 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
105 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ | |
f2d6bfe9 | 106 | #ifndef CONFIG_TRANSPARENT_HUGEPAGE |
1da177e4 | 107 | #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ |
f2d6bfe9 JW |
108 | #else |
109 | #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */ | |
110 | #endif | |
895791da | 111 | #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ |
e5b97dde | 112 | #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ |
d00806b1 | 113 | |
d0217ac0 | 114 | #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ |
b379d790 | 115 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ |
aba46c50 | 116 | #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ |
895791da | 117 | #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ |
f8af4da3 | 118 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ |
1da177e4 | 119 | |
a8bef8ff MG |
120 | /* Bits set in the VMA until the stack is in its final location */ |
121 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
122 | ||
1da177e4 LT |
123 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
124 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
125 | #endif | |
126 | ||
127 | #ifdef CONFIG_STACK_GROWSUP | |
128 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
129 | #else | |
130 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
131 | #endif | |
132 | ||
133 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) | |
134 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK | |
135 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) | |
136 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) | |
137 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) | |
138 | ||
b291f000 | 139 | /* |
78f11a25 AA |
140 | * Special vmas that are non-mergable, non-mlock()able. |
141 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | |
b291f000 NP |
142 | */ |
143 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) | |
144 | ||
1da177e4 LT |
145 | /* |
146 | * mapping from the currently active vm_flags protection bits (the | |
147 | * low four bits) to a page protection mask.. | |
148 | */ | |
149 | extern pgprot_t protection_map[16]; | |
150 | ||
d0217ac0 NP |
151 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ |
152 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ | |
c2ec175c | 153 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ |
d065bd81 | 154 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ |
318b275f | 155 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ |
37b23e05 | 156 | #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ |
d0217ac0 | 157 | |
6bd9cd50 | 158 | /* |
159 | * This interface is used by x86 PAT code to identify a pfn mapping that is | |
160 | * linear over entire vma. This is to optimize PAT code that deals with | |
161 | * marking the physical region with a particular prot. This is not for generic | |
162 | * mm use. Note also that this check will not work if the pfn mapping is | |
163 | * linear for a vma starting at physical address 0. In which case PAT code | |
164 | * falls back to slow path of reserving physical range page by page. | |
165 | */ | |
3c8bb73a | 166 | static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) |
167 | { | |
895791da | 168 | return (vma->vm_flags & VM_PFN_AT_MMAP); |
3c8bb73a | 169 | } |
170 | ||
171 | static inline int is_pfn_mapping(struct vm_area_struct *vma) | |
172 | { | |
173 | return (vma->vm_flags & VM_PFNMAP); | |
174 | } | |
d0217ac0 | 175 | |
54cb8821 | 176 | /* |
d0217ac0 | 177 | * vm_fault is filled by the the pagefault handler and passed to the vma's |
83c54070 NP |
178 | * ->fault function. The vma's ->fault is responsible for returning a bitmask |
179 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
54cb8821 | 180 | * |
d0217ac0 NP |
181 | * pgoff should be used in favour of virtual_address, if possible. If pgoff |
182 | * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear | |
183 | * mapping support. | |
54cb8821 | 184 | */ |
d0217ac0 NP |
185 | struct vm_fault { |
186 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | |
187 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
188 | void __user *virtual_address; /* Faulting virtual address */ | |
189 | ||
190 | struct page *page; /* ->fault handlers should return a | |
83c54070 | 191 | * page here, unless VM_FAULT_NOPAGE |
d0217ac0 | 192 | * is set (which is also implied by |
83c54070 | 193 | * VM_FAULT_ERROR). |
d0217ac0 | 194 | */ |
54cb8821 | 195 | }; |
1da177e4 LT |
196 | |
197 | /* | |
198 | * These are the virtual MM functions - opening of an area, closing and | |
199 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
200 | * to the functions called when a no-page or a wp-page exception occurs. | |
201 | */ | |
202 | struct vm_operations_struct { | |
203 | void (*open)(struct vm_area_struct * area); | |
204 | void (*close)(struct vm_area_struct * area); | |
d0217ac0 | 205 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); |
9637a5ef DH |
206 | |
207 | /* notification that a previously read-only page is about to become | |
208 | * writable, if an error is returned it will cause a SIGBUS */ | |
c2ec175c | 209 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); |
28b2ee20 RR |
210 | |
211 | /* called by access_process_vm when get_user_pages() fails, typically | |
212 | * for use by special VMAs that can switch between memory and hardware | |
213 | */ | |
214 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
215 | void *buf, int len, int write); | |
1da177e4 | 216 | #ifdef CONFIG_NUMA |
a6020ed7 LS |
217 | /* |
218 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
219 | * to hold the policy upon return. Caller should pass NULL @new to | |
220 | * remove a policy and fall back to surrounding context--i.e. do not | |
221 | * install a MPOL_DEFAULT policy, nor the task or system default | |
222 | * mempolicy. | |
223 | */ | |
1da177e4 | 224 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
a6020ed7 LS |
225 | |
226 | /* | |
227 | * get_policy() op must add reference [mpol_get()] to any policy at | |
228 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
229 | * in mm/mempolicy.c will do this automatically. | |
230 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
231 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | |
232 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | |
233 | * must return NULL--i.e., do not "fallback" to task or system default | |
234 | * policy. | |
235 | */ | |
1da177e4 LT |
236 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
237 | unsigned long addr); | |
7b2259b3 CL |
238 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
239 | const nodemask_t *to, unsigned long flags); | |
1da177e4 LT |
240 | #endif |
241 | }; | |
242 | ||
243 | struct mmu_gather; | |
244 | struct inode; | |
245 | ||
349aef0b AM |
246 | #define page_private(page) ((page)->private) |
247 | #define set_page_private(page, v) ((page)->private = (v)) | |
4c21e2f2 | 248 | |
1da177e4 LT |
249 | /* |
250 | * FIXME: take this include out, include page-flags.h in | |
251 | * files which need it (119 of them) | |
252 | */ | |
253 | #include <linux/page-flags.h> | |
71e3aac0 | 254 | #include <linux/huge_mm.h> |
1da177e4 LT |
255 | |
256 | /* | |
257 | * Methods to modify the page usage count. | |
258 | * | |
259 | * What counts for a page usage: | |
260 | * - cache mapping (page->mapping) | |
261 | * - private data (page->private) | |
262 | * - page mapped in a task's page tables, each mapping | |
263 | * is counted separately | |
264 | * | |
265 | * Also, many kernel routines increase the page count before a critical | |
266 | * routine so they can be sure the page doesn't go away from under them. | |
1da177e4 LT |
267 | */ |
268 | ||
269 | /* | |
da6052f7 | 270 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
1da177e4 | 271 | */ |
7c8ee9a8 NP |
272 | static inline int put_page_testzero(struct page *page) |
273 | { | |
725d704e | 274 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
8dc04efb | 275 | return atomic_dec_and_test(&page->_count); |
7c8ee9a8 | 276 | } |
1da177e4 LT |
277 | |
278 | /* | |
7c8ee9a8 NP |
279 | * Try to grab a ref unless the page has a refcount of zero, return false if |
280 | * that is the case. | |
1da177e4 | 281 | */ |
7c8ee9a8 NP |
282 | static inline int get_page_unless_zero(struct page *page) |
283 | { | |
8dc04efb | 284 | return atomic_inc_not_zero(&page->_count); |
7c8ee9a8 | 285 | } |
1da177e4 | 286 | |
53df8fdc WF |
287 | extern int page_is_ram(unsigned long pfn); |
288 | ||
48667e7a | 289 | /* Support for virtually mapped pages */ |
b3bdda02 CL |
290 | struct page *vmalloc_to_page(const void *addr); |
291 | unsigned long vmalloc_to_pfn(const void *addr); | |
48667e7a | 292 | |
0738c4bb PM |
293 | /* |
294 | * Determine if an address is within the vmalloc range | |
295 | * | |
296 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
297 | * is no special casing required. | |
298 | */ | |
9e2779fa CL |
299 | static inline int is_vmalloc_addr(const void *x) |
300 | { | |
0738c4bb | 301 | #ifdef CONFIG_MMU |
9e2779fa CL |
302 | unsigned long addr = (unsigned long)x; |
303 | ||
304 | return addr >= VMALLOC_START && addr < VMALLOC_END; | |
0738c4bb PM |
305 | #else |
306 | return 0; | |
8ca3ed87 | 307 | #endif |
0738c4bb | 308 | } |
81ac3ad9 KH |
309 | #ifdef CONFIG_MMU |
310 | extern int is_vmalloc_or_module_addr(const void *x); | |
311 | #else | |
934831d0 | 312 | static inline int is_vmalloc_or_module_addr(const void *x) |
81ac3ad9 KH |
313 | { |
314 | return 0; | |
315 | } | |
316 | #endif | |
9e2779fa | 317 | |
e9da73d6 AA |
318 | static inline void compound_lock(struct page *page) |
319 | { | |
320 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
321 | bit_spin_lock(PG_compound_lock, &page->flags); | |
322 | #endif | |
323 | } | |
324 | ||
325 | static inline void compound_unlock(struct page *page) | |
326 | { | |
327 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
328 | bit_spin_unlock(PG_compound_lock, &page->flags); | |
329 | #endif | |
330 | } | |
331 | ||
332 | static inline unsigned long compound_lock_irqsave(struct page *page) | |
333 | { | |
334 | unsigned long uninitialized_var(flags); | |
335 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
336 | local_irq_save(flags); | |
337 | compound_lock(page); | |
338 | #endif | |
339 | return flags; | |
340 | } | |
341 | ||
342 | static inline void compound_unlock_irqrestore(struct page *page, | |
343 | unsigned long flags) | |
344 | { | |
345 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
346 | compound_unlock(page); | |
347 | local_irq_restore(flags); | |
348 | #endif | |
349 | } | |
350 | ||
d85f3385 CL |
351 | static inline struct page *compound_head(struct page *page) |
352 | { | |
6d777953 | 353 | if (unlikely(PageTail(page))) |
d85f3385 CL |
354 | return page->first_page; |
355 | return page; | |
356 | } | |
357 | ||
4c21e2f2 | 358 | static inline int page_count(struct page *page) |
1da177e4 | 359 | { |
d85f3385 | 360 | return atomic_read(&compound_head(page)->_count); |
1da177e4 LT |
361 | } |
362 | ||
363 | static inline void get_page(struct page *page) | |
364 | { | |
91807063 AA |
365 | /* |
366 | * Getting a normal page or the head of a compound page | |
367 | * requires to already have an elevated page->_count. Only if | |
368 | * we're getting a tail page, the elevated page->_count is | |
369 | * required only in the head page, so for tail pages the | |
370 | * bugcheck only verifies that the page->_count isn't | |
371 | * negative. | |
372 | */ | |
373 | VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page)); | |
1da177e4 | 374 | atomic_inc(&page->_count); |
91807063 AA |
375 | /* |
376 | * Getting a tail page will elevate both the head and tail | |
377 | * page->_count(s). | |
378 | */ | |
379 | if (unlikely(PageTail(page))) { | |
380 | /* | |
381 | * This is safe only because | |
382 | * __split_huge_page_refcount can't run under | |
383 | * get_page(). | |
384 | */ | |
385 | VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0); | |
386 | atomic_inc(&page->first_page->_count); | |
387 | } | |
1da177e4 LT |
388 | } |
389 | ||
b49af68f CL |
390 | static inline struct page *virt_to_head_page(const void *x) |
391 | { | |
392 | struct page *page = virt_to_page(x); | |
393 | return compound_head(page); | |
394 | } | |
395 | ||
7835e98b NP |
396 | /* |
397 | * Setup the page count before being freed into the page allocator for | |
398 | * the first time (boot or memory hotplug) | |
399 | */ | |
400 | static inline void init_page_count(struct page *page) | |
401 | { | |
402 | atomic_set(&page->_count, 1); | |
403 | } | |
404 | ||
5f24ce5f AA |
405 | /* |
406 | * PageBuddy() indicate that the page is free and in the buddy system | |
407 | * (see mm/page_alloc.c). | |
ef2b4b95 AA |
408 | * |
409 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to | |
410 | * -2 so that an underflow of the page_mapcount() won't be mistaken | |
411 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very | |
412 | * efficiently by most CPU architectures. | |
5f24ce5f | 413 | */ |
ef2b4b95 AA |
414 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) |
415 | ||
5f24ce5f AA |
416 | static inline int PageBuddy(struct page *page) |
417 | { | |
ef2b4b95 | 418 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; |
5f24ce5f AA |
419 | } |
420 | ||
421 | static inline void __SetPageBuddy(struct page *page) | |
422 | { | |
423 | VM_BUG_ON(atomic_read(&page->_mapcount) != -1); | |
ef2b4b95 | 424 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); |
5f24ce5f AA |
425 | } |
426 | ||
427 | static inline void __ClearPageBuddy(struct page *page) | |
428 | { | |
429 | VM_BUG_ON(!PageBuddy(page)); | |
430 | atomic_set(&page->_mapcount, -1); | |
431 | } | |
432 | ||
1da177e4 | 433 | void put_page(struct page *page); |
1d7ea732 | 434 | void put_pages_list(struct list_head *pages); |
1da177e4 | 435 | |
8dfcc9ba | 436 | void split_page(struct page *page, unsigned int order); |
748446bb | 437 | int split_free_page(struct page *page); |
8dfcc9ba | 438 | |
33f2ef89 AW |
439 | /* |
440 | * Compound pages have a destructor function. Provide a | |
441 | * prototype for that function and accessor functions. | |
442 | * These are _only_ valid on the head of a PG_compound page. | |
443 | */ | |
444 | typedef void compound_page_dtor(struct page *); | |
445 | ||
446 | static inline void set_compound_page_dtor(struct page *page, | |
447 | compound_page_dtor *dtor) | |
448 | { | |
449 | page[1].lru.next = (void *)dtor; | |
450 | } | |
451 | ||
452 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
453 | { | |
454 | return (compound_page_dtor *)page[1].lru.next; | |
455 | } | |
456 | ||
d85f3385 CL |
457 | static inline int compound_order(struct page *page) |
458 | { | |
6d777953 | 459 | if (!PageHead(page)) |
d85f3385 CL |
460 | return 0; |
461 | return (unsigned long)page[1].lru.prev; | |
462 | } | |
463 | ||
37c2ac78 AA |
464 | static inline int compound_trans_order(struct page *page) |
465 | { | |
466 | int order; | |
467 | unsigned long flags; | |
468 | ||
469 | if (!PageHead(page)) | |
470 | return 0; | |
471 | ||
472 | flags = compound_lock_irqsave(page); | |
473 | order = compound_order(page); | |
474 | compound_unlock_irqrestore(page, flags); | |
475 | return order; | |
476 | } | |
477 | ||
d85f3385 CL |
478 | static inline void set_compound_order(struct page *page, unsigned long order) |
479 | { | |
480 | page[1].lru.prev = (void *)order; | |
481 | } | |
482 | ||
3dece370 | 483 | #ifdef CONFIG_MMU |
14fd403f AA |
484 | /* |
485 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
486 | * servicing faults for write access. In the normal case, do always want | |
487 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
488 | * that do not have writing enabled, when used by access_process_vm. | |
489 | */ | |
490 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
491 | { | |
492 | if (likely(vma->vm_flags & VM_WRITE)) | |
493 | pte = pte_mkwrite(pte); | |
494 | return pte; | |
495 | } | |
3dece370 | 496 | #endif |
14fd403f | 497 | |
1da177e4 LT |
498 | /* |
499 | * Multiple processes may "see" the same page. E.g. for untouched | |
500 | * mappings of /dev/null, all processes see the same page full of | |
501 | * zeroes, and text pages of executables and shared libraries have | |
502 | * only one copy in memory, at most, normally. | |
503 | * | |
504 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
7e871b6c PBG |
505 | * page_count() == 0 means the page is free. page->lru is then used for |
506 | * freelist management in the buddy allocator. | |
da6052f7 | 507 | * page_count() > 0 means the page has been allocated. |
1da177e4 | 508 | * |
da6052f7 NP |
509 | * Pages are allocated by the slab allocator in order to provide memory |
510 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
511 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
512 | * unless a particular usage is carefully commented. (the responsibility of | |
513 | * freeing the kmalloc memory is the caller's, of course). | |
1da177e4 | 514 | * |
da6052f7 NP |
515 | * A page may be used by anyone else who does a __get_free_page(). |
516 | * In this case, page_count still tracks the references, and should only | |
517 | * be used through the normal accessor functions. The top bits of page->flags | |
518 | * and page->virtual store page management information, but all other fields | |
519 | * are unused and could be used privately, carefully. The management of this | |
520 | * page is the responsibility of the one who allocated it, and those who have | |
521 | * subsequently been given references to it. | |
522 | * | |
523 | * The other pages (we may call them "pagecache pages") are completely | |
1da177e4 LT |
524 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
525 | * The following discussion applies only to them. | |
526 | * | |
da6052f7 NP |
527 | * A pagecache page contains an opaque `private' member, which belongs to the |
528 | * page's address_space. Usually, this is the address of a circular list of | |
529 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
530 | * into the filesystem to release these pages. | |
1da177e4 | 531 | * |
da6052f7 NP |
532 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
533 | * is the pointer to the inode, and page->index is the file offset of the page, | |
534 | * in units of PAGE_CACHE_SIZE. | |
1da177e4 | 535 | * |
da6052f7 NP |
536 | * If pagecache pages are not associated with an inode, they are said to be |
537 | * anonymous pages. These may become associated with the swapcache, and in that | |
538 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
1da177e4 | 539 | * |
da6052f7 NP |
540 | * In either case (swapcache or inode backed), the pagecache itself holds one |
541 | * reference to the page. Setting PG_private should also increment the | |
542 | * refcount. The each user mapping also has a reference to the page. | |
1da177e4 | 543 | * |
da6052f7 NP |
544 | * The pagecache pages are stored in a per-mapping radix tree, which is |
545 | * rooted at mapping->page_tree, and indexed by offset. | |
546 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
547 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
1da177e4 | 548 | * |
da6052f7 | 549 | * All pagecache pages may be subject to I/O: |
1da177e4 LT |
550 | * - inode pages may need to be read from disk, |
551 | * - inode pages which have been modified and are MAP_SHARED may need | |
da6052f7 NP |
552 | * to be written back to the inode on disk, |
553 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
554 | * modified may need to be swapped out to swap space and (later) to be read | |
555 | * back into memory. | |
1da177e4 LT |
556 | */ |
557 | ||
558 | /* | |
559 | * The zone field is never updated after free_area_init_core() | |
560 | * sets it, so none of the operations on it need to be atomic. | |
1da177e4 | 561 | */ |
348f8b6c | 562 | |
d41dee36 AW |
563 | |
564 | /* | |
565 | * page->flags layout: | |
566 | * | |
567 | * There are three possibilities for how page->flags get | |
568 | * laid out. The first is for the normal case, without | |
569 | * sparsemem. The second is for sparsemem when there is | |
570 | * plenty of space for node and section. The last is when | |
571 | * we have run out of space and have to fall back to an | |
572 | * alternate (slower) way of determining the node. | |
573 | * | |
308c05e3 CL |
574 | * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | |
575 | * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | | |
576 | * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | | |
d41dee36 | 577 | */ |
308c05e3 | 578 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
d41dee36 AW |
579 | #define SECTIONS_WIDTH SECTIONS_SHIFT |
580 | #else | |
581 | #define SECTIONS_WIDTH 0 | |
582 | #endif | |
583 | ||
584 | #define ZONES_WIDTH ZONES_SHIFT | |
585 | ||
9223b419 | 586 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS |
d41dee36 AW |
587 | #define NODES_WIDTH NODES_SHIFT |
588 | #else | |
308c05e3 CL |
589 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
590 | #error "Vmemmap: No space for nodes field in page flags" | |
591 | #endif | |
d41dee36 AW |
592 | #define NODES_WIDTH 0 |
593 | #endif | |
594 | ||
595 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ | |
07808b74 | 596 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
d41dee36 AW |
597 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
598 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
599 | ||
600 | /* | |
601 | * We are going to use the flags for the page to node mapping if its in | |
602 | * there. This includes the case where there is no node, so it is implicit. | |
603 | */ | |
89689ae7 CL |
604 | #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) |
605 | #define NODE_NOT_IN_PAGE_FLAGS | |
606 | #endif | |
d41dee36 AW |
607 | |
608 | #ifndef PFN_SECTION_SHIFT | |
609 | #define PFN_SECTION_SHIFT 0 | |
610 | #endif | |
348f8b6c DH |
611 | |
612 | /* | |
25985edc | 613 | * Define the bit shifts to access each section. For non-existent |
348f8b6c DH |
614 | * sections we define the shift as 0; that plus a 0 mask ensures |
615 | * the compiler will optimise away reference to them. | |
616 | */ | |
d41dee36 AW |
617 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
618 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
619 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
348f8b6c | 620 | |
bce54bbf WD |
621 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ |
622 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
89689ae7 | 623 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
624 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
625 | SECTIONS_PGOFF : ZONES_PGOFF) | |
d41dee36 | 626 | #else |
89689ae7 | 627 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
bd8029b6 AW |
628 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
629 | NODES_PGOFF : ZONES_PGOFF) | |
89689ae7 CL |
630 | #endif |
631 | ||
bd8029b6 | 632 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
348f8b6c | 633 | |
9223b419 CL |
634 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS |
635 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
348f8b6c DH |
636 | #endif |
637 | ||
d41dee36 AW |
638 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
639 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
640 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
89689ae7 | 641 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
348f8b6c | 642 | |
2f1b6248 | 643 | static inline enum zone_type page_zonenum(struct page *page) |
1da177e4 | 644 | { |
348f8b6c | 645 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
1da177e4 | 646 | } |
1da177e4 | 647 | |
89689ae7 CL |
648 | /* |
649 | * The identification function is only used by the buddy allocator for | |
650 | * determining if two pages could be buddies. We are not really | |
651 | * identifying a zone since we could be using a the section number | |
652 | * id if we have not node id available in page flags. | |
653 | * We guarantee only that it will return the same value for two | |
654 | * combinable pages in a zone. | |
655 | */ | |
cb2b95e1 AW |
656 | static inline int page_zone_id(struct page *page) |
657 | { | |
89689ae7 | 658 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
348f8b6c DH |
659 | } |
660 | ||
25ba77c1 | 661 | static inline int zone_to_nid(struct zone *zone) |
89fa3024 | 662 | { |
d5f541ed CL |
663 | #ifdef CONFIG_NUMA |
664 | return zone->node; | |
665 | #else | |
666 | return 0; | |
667 | #endif | |
89fa3024 CL |
668 | } |
669 | ||
89689ae7 | 670 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
25ba77c1 | 671 | extern int page_to_nid(struct page *page); |
89689ae7 | 672 | #else |
25ba77c1 | 673 | static inline int page_to_nid(struct page *page) |
d41dee36 | 674 | { |
89689ae7 | 675 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
d41dee36 | 676 | } |
89689ae7 CL |
677 | #endif |
678 | ||
679 | static inline struct zone *page_zone(struct page *page) | |
680 | { | |
681 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
682 | } | |
683 | ||
308c05e3 | 684 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) |
d41dee36 AW |
685 | static inline unsigned long page_to_section(struct page *page) |
686 | { | |
687 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
688 | } | |
308c05e3 | 689 | #endif |
d41dee36 | 690 | |
2f1b6248 | 691 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
348f8b6c DH |
692 | { |
693 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
694 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
695 | } | |
2f1b6248 | 696 | |
348f8b6c DH |
697 | static inline void set_page_node(struct page *page, unsigned long node) |
698 | { | |
699 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
700 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
1da177e4 | 701 | } |
89689ae7 | 702 | |
d41dee36 AW |
703 | static inline void set_page_section(struct page *page, unsigned long section) |
704 | { | |
705 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
706 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
707 | } | |
1da177e4 | 708 | |
2f1b6248 | 709 | static inline void set_page_links(struct page *page, enum zone_type zone, |
d41dee36 | 710 | unsigned long node, unsigned long pfn) |
1da177e4 | 711 | { |
348f8b6c DH |
712 | set_page_zone(page, zone); |
713 | set_page_node(page, node); | |
d41dee36 | 714 | set_page_section(page, pfn_to_section_nr(pfn)); |
1da177e4 LT |
715 | } |
716 | ||
f6ac2354 CL |
717 | /* |
718 | * Some inline functions in vmstat.h depend on page_zone() | |
719 | */ | |
720 | #include <linux/vmstat.h> | |
721 | ||
652050ae | 722 | static __always_inline void *lowmem_page_address(struct page *page) |
1da177e4 | 723 | { |
c6f6b596 | 724 | return __va(PFN_PHYS(page_to_pfn(page))); |
1da177e4 LT |
725 | } |
726 | ||
727 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
728 | #define HASHED_PAGE_VIRTUAL | |
729 | #endif | |
730 | ||
731 | #if defined(WANT_PAGE_VIRTUAL) | |
732 | #define page_address(page) ((page)->virtual) | |
733 | #define set_page_address(page, address) \ | |
734 | do { \ | |
735 | (page)->virtual = (address); \ | |
736 | } while(0) | |
737 | #define page_address_init() do { } while(0) | |
738 | #endif | |
739 | ||
740 | #if defined(HASHED_PAGE_VIRTUAL) | |
741 | void *page_address(struct page *page); | |
742 | void set_page_address(struct page *page, void *virtual); | |
743 | void page_address_init(void); | |
744 | #endif | |
745 | ||
746 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
747 | #define page_address(page) lowmem_page_address(page) | |
748 | #define set_page_address(page, address) do { } while(0) | |
749 | #define page_address_init() do { } while(0) | |
750 | #endif | |
751 | ||
752 | /* | |
753 | * On an anonymous page mapped into a user virtual memory area, | |
754 | * page->mapping points to its anon_vma, not to a struct address_space; | |
3ca7b3c5 HD |
755 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. |
756 | * | |
757 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, | |
758 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; | |
759 | * and then page->mapping points, not to an anon_vma, but to a private | |
760 | * structure which KSM associates with that merged page. See ksm.h. | |
761 | * | |
762 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. | |
1da177e4 LT |
763 | * |
764 | * Please note that, confusingly, "page_mapping" refers to the inode | |
765 | * address_space which maps the page from disk; whereas "page_mapped" | |
766 | * refers to user virtual address space into which the page is mapped. | |
767 | */ | |
768 | #define PAGE_MAPPING_ANON 1 | |
3ca7b3c5 HD |
769 | #define PAGE_MAPPING_KSM 2 |
770 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) | |
1da177e4 LT |
771 | |
772 | extern struct address_space swapper_space; | |
773 | static inline struct address_space *page_mapping(struct page *page) | |
774 | { | |
775 | struct address_space *mapping = page->mapping; | |
776 | ||
b5fab14e | 777 | VM_BUG_ON(PageSlab(page)); |
1da177e4 LT |
778 | if (unlikely(PageSwapCache(page))) |
779 | mapping = &swapper_space; | |
e20e8779 | 780 | else if ((unsigned long)mapping & PAGE_MAPPING_ANON) |
1da177e4 LT |
781 | mapping = NULL; |
782 | return mapping; | |
783 | } | |
784 | ||
3ca7b3c5 HD |
785 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ |
786 | static inline void *page_rmapping(struct page *page) | |
787 | { | |
788 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); | |
789 | } | |
790 | ||
1da177e4 LT |
791 | static inline int PageAnon(struct page *page) |
792 | { | |
793 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
794 | } | |
795 | ||
796 | /* | |
797 | * Return the pagecache index of the passed page. Regular pagecache pages | |
798 | * use ->index whereas swapcache pages use ->private | |
799 | */ | |
800 | static inline pgoff_t page_index(struct page *page) | |
801 | { | |
802 | if (unlikely(PageSwapCache(page))) | |
4c21e2f2 | 803 | return page_private(page); |
1da177e4 LT |
804 | return page->index; |
805 | } | |
806 | ||
807 | /* | |
808 | * The atomic page->_mapcount, like _count, starts from -1: | |
809 | * so that transitions both from it and to it can be tracked, | |
810 | * using atomic_inc_and_test and atomic_add_negative(-1). | |
811 | */ | |
812 | static inline void reset_page_mapcount(struct page *page) | |
813 | { | |
814 | atomic_set(&(page)->_mapcount, -1); | |
815 | } | |
816 | ||
817 | static inline int page_mapcount(struct page *page) | |
818 | { | |
819 | return atomic_read(&(page)->_mapcount) + 1; | |
820 | } | |
821 | ||
822 | /* | |
823 | * Return true if this page is mapped into pagetables. | |
824 | */ | |
825 | static inline int page_mapped(struct page *page) | |
826 | { | |
827 | return atomic_read(&(page)->_mapcount) >= 0; | |
828 | } | |
829 | ||
1da177e4 LT |
830 | /* |
831 | * Different kinds of faults, as returned by handle_mm_fault(). | |
832 | * Used to decide whether a process gets delivered SIGBUS or | |
833 | * just gets major/minor fault counters bumped up. | |
834 | */ | |
d0217ac0 | 835 | |
83c54070 | 836 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ |
d0217ac0 | 837 | |
83c54070 NP |
838 | #define VM_FAULT_OOM 0x0001 |
839 | #define VM_FAULT_SIGBUS 0x0002 | |
840 | #define VM_FAULT_MAJOR 0x0004 | |
841 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | |
aa50d3a7 AK |
842 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ |
843 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ | |
f33ea7f4 | 844 | |
83c54070 NP |
845 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ |
846 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | |
d065bd81 | 847 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ |
1da177e4 | 848 | |
aa50d3a7 AK |
849 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ |
850 | ||
851 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ | |
852 | VM_FAULT_HWPOISON_LARGE) | |
853 | ||
854 | /* Encode hstate index for a hwpoisoned large page */ | |
855 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) | |
856 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) | |
d0217ac0 | 857 | |
1c0fe6e3 NP |
858 | /* |
859 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
860 | */ | |
861 | extern void pagefault_out_of_memory(void); | |
862 | ||
1da177e4 LT |
863 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
864 | ||
ddd588b5 | 865 | /* |
7bf02ea2 | 866 | * Flags passed to show_mem() and show_free_areas() to suppress output in |
ddd588b5 DR |
867 | * various contexts. |
868 | */ | |
869 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */ | |
870 | ||
7bf02ea2 DR |
871 | extern void show_free_areas(unsigned int flags); |
872 | extern bool skip_free_areas_node(unsigned int flags, int nid); | |
1da177e4 | 873 | |
3f96b79a | 874 | int shmem_lock(struct file *file, int lock, struct user_struct *user); |
168f5ac6 | 875 | struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); |
1da177e4 LT |
876 | int shmem_zero_setup(struct vm_area_struct *); |
877 | ||
b0e15190 DH |
878 | #ifndef CONFIG_MMU |
879 | extern unsigned long shmem_get_unmapped_area(struct file *file, | |
880 | unsigned long addr, | |
881 | unsigned long len, | |
882 | unsigned long pgoff, | |
883 | unsigned long flags); | |
884 | #endif | |
885 | ||
e8edc6e0 | 886 | extern int can_do_mlock(void); |
1da177e4 LT |
887 | extern int user_shm_lock(size_t, struct user_struct *); |
888 | extern void user_shm_unlock(size_t, struct user_struct *); | |
889 | ||
890 | /* | |
891 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
892 | */ | |
893 | struct zap_details { | |
894 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ | |
895 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
896 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
897 | pgoff_t last_index; /* Highest page->index to unmap */ | |
1da177e4 LT |
898 | }; |
899 | ||
7e675137 NP |
900 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, |
901 | pte_t pte); | |
902 | ||
c627f9cc JS |
903 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, |
904 | unsigned long size); | |
ee39b37b | 905 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
1da177e4 | 906 | unsigned long size, struct zap_details *); |
d16dfc55 | 907 | unsigned long unmap_vmas(struct mmu_gather *tlb, |
1da177e4 LT |
908 | struct vm_area_struct *start_vma, unsigned long start_addr, |
909 | unsigned long end_addr, unsigned long *nr_accounted, | |
910 | struct zap_details *); | |
e6473092 MM |
911 | |
912 | /** | |
913 | * mm_walk - callbacks for walk_page_range | |
914 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry | |
915 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry | |
916 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry | |
03319327 DH |
917 | * this handler is required to be able to handle |
918 | * pmd_trans_huge() pmds. They may simply choose to | |
919 | * split_huge_page() instead of handling it explicitly. | |
e6473092 MM |
920 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry |
921 | * @pte_hole: if set, called for each hole at all levels | |
5dc37642 | 922 | * @hugetlb_entry: if set, called for each hugetlb entry |
e6473092 MM |
923 | * |
924 | * (see walk_page_range for more details) | |
925 | */ | |
926 | struct mm_walk { | |
2165009b DH |
927 | int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); |
928 | int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); | |
929 | int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); | |
930 | int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); | |
931 | int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); | |
116354d1 NH |
932 | int (*hugetlb_entry)(pte_t *, unsigned long, |
933 | unsigned long, unsigned long, struct mm_walk *); | |
2165009b DH |
934 | struct mm_struct *mm; |
935 | void *private; | |
e6473092 MM |
936 | }; |
937 | ||
2165009b DH |
938 | int walk_page_range(unsigned long addr, unsigned long end, |
939 | struct mm_walk *walk); | |
42b77728 | 940 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, |
3bf5ee95 | 941 | unsigned long end, unsigned long floor, unsigned long ceiling); |
1da177e4 LT |
942 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
943 | struct vm_area_struct *vma); | |
1da177e4 LT |
944 | void unmap_mapping_range(struct address_space *mapping, |
945 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
3b6748e2 JW |
946 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, |
947 | unsigned long *pfn); | |
d87fe660 | 948 | int follow_phys(struct vm_area_struct *vma, unsigned long address, |
949 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
28b2ee20 RR |
950 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, |
951 | void *buf, int len, int write); | |
1da177e4 LT |
952 | |
953 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
954 | loff_t const holebegin, loff_t const holelen) | |
955 | { | |
956 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
957 | } | |
958 | ||
25d9e2d1 | 959 | extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); |
2c27c65e | 960 | extern void truncate_setsize(struct inode *inode, loff_t newsize); |
25d9e2d1 | 961 | extern int vmtruncate(struct inode *inode, loff_t offset); |
962 | extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); | |
f33ea7f4 | 963 | |
750b4987 | 964 | int truncate_inode_page(struct address_space *mapping, struct page *page); |
25718736 | 965 | int generic_error_remove_page(struct address_space *mapping, struct page *page); |
750b4987 | 966 | |
83f78668 WF |
967 | int invalidate_inode_page(struct page *page); |
968 | ||
7ee1dd3f | 969 | #ifdef CONFIG_MMU |
83c54070 | 970 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
d06063cc | 971 | unsigned long address, unsigned int flags); |
7ee1dd3f DH |
972 | #else |
973 | static inline int handle_mm_fault(struct mm_struct *mm, | |
974 | struct vm_area_struct *vma, unsigned long address, | |
d06063cc | 975 | unsigned int flags) |
7ee1dd3f DH |
976 | { |
977 | /* should never happen if there's no MMU */ | |
978 | BUG(); | |
979 | return VM_FAULT_SIGBUS; | |
980 | } | |
981 | #endif | |
f33ea7f4 | 982 | |
1da177e4 LT |
983 | extern int make_pages_present(unsigned long addr, unsigned long end); |
984 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
5ddd36b9 SW |
985 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, |
986 | void *buf, int len, int write); | |
1da177e4 | 987 | |
0014bd99 HY |
988 | int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
989 | unsigned long start, int len, unsigned int foll_flags, | |
990 | struct page **pages, struct vm_area_struct **vmas, | |
991 | int *nonblocking); | |
d2bf6be8 | 992 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
9d73777e | 993 | unsigned long start, int nr_pages, int write, int force, |
d2bf6be8 NP |
994 | struct page **pages, struct vm_area_struct **vmas); |
995 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
996 | struct page **pages); | |
f3e8fccd | 997 | struct page *get_dump_page(unsigned long addr); |
1da177e4 | 998 | |
cf9a2ae8 DH |
999 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
1000 | extern void do_invalidatepage(struct page *page, unsigned long offset); | |
1001 | ||
1da177e4 | 1002 | int __set_page_dirty_nobuffers(struct page *page); |
76719325 | 1003 | int __set_page_dirty_no_writeback(struct page *page); |
1da177e4 LT |
1004 | int redirty_page_for_writepage(struct writeback_control *wbc, |
1005 | struct page *page); | |
e3a7cca1 | 1006 | void account_page_dirtied(struct page *page, struct address_space *mapping); |
f629d1c9 | 1007 | void account_page_writeback(struct page *page); |
b3c97528 | 1008 | int set_page_dirty(struct page *page); |
1da177e4 LT |
1009 | int set_page_dirty_lock(struct page *page); |
1010 | int clear_page_dirty_for_io(struct page *page); | |
1011 | ||
39aa3cb3 | 1012 | /* Is the vma a continuation of the stack vma above it? */ |
a09a79f6 | 1013 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) |
39aa3cb3 SB |
1014 | { |
1015 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); | |
1016 | } | |
1017 | ||
a09a79f6 MP |
1018 | static inline int stack_guard_page_start(struct vm_area_struct *vma, |
1019 | unsigned long addr) | |
1020 | { | |
1021 | return (vma->vm_flags & VM_GROWSDOWN) && | |
1022 | (vma->vm_start == addr) && | |
1023 | !vma_growsdown(vma->vm_prev, addr); | |
1024 | } | |
1025 | ||
1026 | /* Is the vma a continuation of the stack vma below it? */ | |
1027 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) | |
1028 | { | |
1029 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); | |
1030 | } | |
1031 | ||
1032 | static inline int stack_guard_page_end(struct vm_area_struct *vma, | |
1033 | unsigned long addr) | |
1034 | { | |
1035 | return (vma->vm_flags & VM_GROWSUP) && | |
1036 | (vma->vm_end == addr) && | |
1037 | !vma_growsup(vma->vm_next, addr); | |
1038 | } | |
1039 | ||
b6a2fea3 OW |
1040 | extern unsigned long move_page_tables(struct vm_area_struct *vma, |
1041 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
1042 | unsigned long new_addr, unsigned long len); | |
1da177e4 LT |
1043 | extern unsigned long do_mremap(unsigned long addr, |
1044 | unsigned long old_len, unsigned long new_len, | |
1045 | unsigned long flags, unsigned long new_addr); | |
b6a2fea3 OW |
1046 | extern int mprotect_fixup(struct vm_area_struct *vma, |
1047 | struct vm_area_struct **pprev, unsigned long start, | |
1048 | unsigned long end, unsigned long newflags); | |
1da177e4 | 1049 | |
465a454f PZ |
1050 | /* |
1051 | * doesn't attempt to fault and will return short. | |
1052 | */ | |
1053 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1054 | struct page **pages); | |
d559db08 KH |
1055 | /* |
1056 | * per-process(per-mm_struct) statistics. | |
1057 | */ | |
34e55232 | 1058 | #if defined(SPLIT_RSS_COUNTING) |
d559db08 KH |
1059 | /* |
1060 | * The mm counters are not protected by its page_table_lock, | |
1061 | * so must be incremented atomically. | |
1062 | */ | |
1063 | static inline void set_mm_counter(struct mm_struct *mm, int member, long value) | |
1064 | { | |
1065 | atomic_long_set(&mm->rss_stat.count[member], value); | |
1066 | } | |
1067 | ||
34e55232 | 1068 | unsigned long get_mm_counter(struct mm_struct *mm, int member); |
d559db08 KH |
1069 | |
1070 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1071 | { | |
1072 | atomic_long_add(value, &mm->rss_stat.count[member]); | |
1073 | } | |
1074 | ||
1075 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1076 | { | |
1077 | atomic_long_inc(&mm->rss_stat.count[member]); | |
1078 | } | |
1079 | ||
1080 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1081 | { | |
1082 | atomic_long_dec(&mm->rss_stat.count[member]); | |
1083 | } | |
1084 | ||
1085 | #else /* !USE_SPLIT_PTLOCKS */ | |
1086 | /* | |
1087 | * The mm counters are protected by its page_table_lock, | |
1088 | * so can be incremented directly. | |
1089 | */ | |
1090 | static inline void set_mm_counter(struct mm_struct *mm, int member, long value) | |
1091 | { | |
1092 | mm->rss_stat.count[member] = value; | |
1093 | } | |
1094 | ||
1095 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) | |
1096 | { | |
1097 | return mm->rss_stat.count[member]; | |
1098 | } | |
1099 | ||
1100 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1101 | { | |
1102 | mm->rss_stat.count[member] += value; | |
1103 | } | |
1104 | ||
1105 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1106 | { | |
1107 | mm->rss_stat.count[member]++; | |
1108 | } | |
1109 | ||
1110 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1111 | { | |
1112 | mm->rss_stat.count[member]--; | |
1113 | } | |
1114 | ||
1115 | #endif /* !USE_SPLIT_PTLOCKS */ | |
1116 | ||
1117 | static inline unsigned long get_mm_rss(struct mm_struct *mm) | |
1118 | { | |
1119 | return get_mm_counter(mm, MM_FILEPAGES) + | |
1120 | get_mm_counter(mm, MM_ANONPAGES); | |
1121 | } | |
1122 | ||
1123 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1124 | { | |
1125 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1126 | } | |
1127 | ||
1128 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1129 | { | |
1130 | return max(mm->hiwater_vm, mm->total_vm); | |
1131 | } | |
1132 | ||
1133 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1134 | { | |
1135 | unsigned long _rss = get_mm_rss(mm); | |
1136 | ||
1137 | if ((mm)->hiwater_rss < _rss) | |
1138 | (mm)->hiwater_rss = _rss; | |
1139 | } | |
1140 | ||
1141 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1142 | { | |
1143 | if (mm->hiwater_vm < mm->total_vm) | |
1144 | mm->hiwater_vm = mm->total_vm; | |
1145 | } | |
1146 | ||
1147 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, | |
1148 | struct mm_struct *mm) | |
1149 | { | |
1150 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
1151 | ||
1152 | if (*maxrss < hiwater_rss) | |
1153 | *maxrss = hiwater_rss; | |
1154 | } | |
1155 | ||
53bddb4e | 1156 | #if defined(SPLIT_RSS_COUNTING) |
34e55232 | 1157 | void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); |
53bddb4e KH |
1158 | #else |
1159 | static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) | |
1160 | { | |
1161 | } | |
1162 | #endif | |
465a454f | 1163 | |
1da177e4 | 1164 | /* |
8e1f936b | 1165 | * A callback you can register to apply pressure to ageable caches. |
1da177e4 | 1166 | * |
8e1f936b RR |
1167 | * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should |
1168 | * look through the least-recently-used 'nr_to_scan' entries and | |
1169 | * attempt to free them up. It should return the number of objects | |
1170 | * which remain in the cache. If it returns -1, it means it cannot do | |
1171 | * any scanning at this time (eg. there is a risk of deadlock). | |
1da177e4 | 1172 | * |
8e1f936b RR |
1173 | * The 'gfpmask' refers to the allocation we are currently trying to |
1174 | * fulfil. | |
1175 | * | |
1176 | * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is | |
1177 | * querying the cache size, so a fastpath for that case is appropriate. | |
1da177e4 | 1178 | */ |
8e1f936b | 1179 | struct shrinker { |
7f8275d0 | 1180 | int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask); |
8e1f936b | 1181 | int seeks; /* seeks to recreate an obj */ |
1da177e4 | 1182 | |
8e1f936b RR |
1183 | /* These are for internal use */ |
1184 | struct list_head list; | |
1185 | long nr; /* objs pending delete */ | |
1186 | }; | |
1187 | #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ | |
1188 | extern void register_shrinker(struct shrinker *); | |
1189 | extern void unregister_shrinker(struct shrinker *); | |
1da177e4 | 1190 | |
4e950f6f | 1191 | int vma_wants_writenotify(struct vm_area_struct *vma); |
d08b3851 | 1192 | |
25ca1d6c NK |
1193 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, |
1194 | spinlock_t **ptl); | |
1195 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1196 | spinlock_t **ptl) | |
1197 | { | |
1198 | pte_t *ptep; | |
1199 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
1200 | return ptep; | |
1201 | } | |
c9cfcddf | 1202 | |
5f22df00 NP |
1203 | #ifdef __PAGETABLE_PUD_FOLDED |
1204 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
1205 | unsigned long address) | |
1206 | { | |
1207 | return 0; | |
1208 | } | |
1209 | #else | |
1bb3630e | 1210 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
5f22df00 NP |
1211 | #endif |
1212 | ||
1213 | #ifdef __PAGETABLE_PMD_FOLDED | |
1214 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
1215 | unsigned long address) | |
1216 | { | |
1217 | return 0; | |
1218 | } | |
1219 | #else | |
1bb3630e | 1220 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
5f22df00 NP |
1221 | #endif |
1222 | ||
8ac1f832 AA |
1223 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, |
1224 | pmd_t *pmd, unsigned long address); | |
1bb3630e HD |
1225 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); |
1226 | ||
1da177e4 LT |
1227 | /* |
1228 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
1229 | * Remove it when 4level-fixup.h has been removed. | |
1230 | */ | |
1bb3630e | 1231 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
1da177e4 LT |
1232 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
1233 | { | |
1bb3630e HD |
1234 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
1235 | NULL: pud_offset(pgd, address); | |
1da177e4 LT |
1236 | } |
1237 | ||
1238 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
1239 | { | |
1bb3630e HD |
1240 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
1241 | NULL: pmd_offset(pud, address); | |
1da177e4 | 1242 | } |
1bb3630e HD |
1243 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
1244 | ||
f7d0b926 | 1245 | #if USE_SPLIT_PTLOCKS |
4c21e2f2 HD |
1246 | /* |
1247 | * We tuck a spinlock to guard each pagetable page into its struct page, | |
1248 | * at page->private, with BUILD_BUG_ON to make sure that this will not | |
1249 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). | |
1250 | * When freeing, reset page->mapping so free_pages_check won't complain. | |
1251 | */ | |
349aef0b | 1252 | #define __pte_lockptr(page) &((page)->ptl) |
4c21e2f2 HD |
1253 | #define pte_lock_init(_page) do { \ |
1254 | spin_lock_init(__pte_lockptr(_page)); \ | |
1255 | } while (0) | |
1256 | #define pte_lock_deinit(page) ((page)->mapping = NULL) | |
1257 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) | |
f7d0b926 | 1258 | #else /* !USE_SPLIT_PTLOCKS */ |
4c21e2f2 HD |
1259 | /* |
1260 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
1261 | */ | |
1262 | #define pte_lock_init(page) do {} while (0) | |
1263 | #define pte_lock_deinit(page) do {} while (0) | |
1264 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) | |
f7d0b926 | 1265 | #endif /* USE_SPLIT_PTLOCKS */ |
4c21e2f2 | 1266 | |
2f569afd MS |
1267 | static inline void pgtable_page_ctor(struct page *page) |
1268 | { | |
1269 | pte_lock_init(page); | |
1270 | inc_zone_page_state(page, NR_PAGETABLE); | |
1271 | } | |
1272 | ||
1273 | static inline void pgtable_page_dtor(struct page *page) | |
1274 | { | |
1275 | pte_lock_deinit(page); | |
1276 | dec_zone_page_state(page, NR_PAGETABLE); | |
1277 | } | |
1278 | ||
c74df32c HD |
1279 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
1280 | ({ \ | |
4c21e2f2 | 1281 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
c74df32c HD |
1282 | pte_t *__pte = pte_offset_map(pmd, address); \ |
1283 | *(ptlp) = __ptl; \ | |
1284 | spin_lock(__ptl); \ | |
1285 | __pte; \ | |
1286 | }) | |
1287 | ||
1288 | #define pte_unmap_unlock(pte, ptl) do { \ | |
1289 | spin_unlock(ptl); \ | |
1290 | pte_unmap(pte); \ | |
1291 | } while (0) | |
1292 | ||
8ac1f832 AA |
1293 | #define pte_alloc_map(mm, vma, pmd, address) \ |
1294 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ | |
1295 | pmd, address))? \ | |
1296 | NULL: pte_offset_map(pmd, address)) | |
1bb3630e | 1297 | |
c74df32c | 1298 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
8ac1f832 AA |
1299 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ |
1300 | pmd, address))? \ | |
c74df32c HD |
1301 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) |
1302 | ||
1bb3630e | 1303 | #define pte_alloc_kernel(pmd, address) \ |
8ac1f832 | 1304 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ |
1bb3630e | 1305 | NULL: pte_offset_kernel(pmd, address)) |
1da177e4 LT |
1306 | |
1307 | extern void free_area_init(unsigned long * zones_size); | |
9109fb7b JW |
1308 | extern void free_area_init_node(int nid, unsigned long * zones_size, |
1309 | unsigned long zone_start_pfn, unsigned long *zholes_size); | |
c713216d MG |
1310 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
1311 | /* | |
1312 | * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its | |
1313 | * zones, allocate the backing mem_map and account for memory holes in a more | |
1314 | * architecture independent manner. This is a substitute for creating the | |
1315 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
1316 | * free_area_init_node() | |
1317 | * | |
1318 | * An architecture is expected to register range of page frames backed by | |
1319 | * physical memory with add_active_range() before calling | |
1320 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
1321 | * usage, an architecture is expected to do something like | |
1322 | * | |
1323 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
1324 | * max_highmem_pfn}; | |
1325 | * for_each_valid_physical_page_range() | |
1326 | * add_active_range(node_id, start_pfn, end_pfn) | |
1327 | * free_area_init_nodes(max_zone_pfns); | |
1328 | * | |
1329 | * If the architecture guarantees that there are no holes in the ranges | |
1330 | * registered with add_active_range(), free_bootmem_active_regions() | |
1331 | * will call free_bootmem_node() for each registered physical page range. | |
1332 | * Similarly sparse_memory_present_with_active_regions() calls | |
1333 | * memory_present() for each range when SPARSEMEM is enabled. | |
1334 | * | |
1335 | * See mm/page_alloc.c for more information on each function exposed by | |
1336 | * CONFIG_ARCH_POPULATES_NODE_MAP | |
1337 | */ | |
1338 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1339 | extern void add_active_range(unsigned int nid, unsigned long start_pfn, | |
1340 | unsigned long end_pfn); | |
cc1050ba YL |
1341 | extern void remove_active_range(unsigned int nid, unsigned long start_pfn, |
1342 | unsigned long end_pfn); | |
c713216d | 1343 | extern void remove_all_active_ranges(void); |
32996250 YL |
1344 | void sort_node_map(void); |
1345 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, | |
1346 | unsigned long end_pfn); | |
c713216d MG |
1347 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
1348 | unsigned long end_pfn); | |
1349 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1350 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1351 | extern unsigned long find_min_pfn_with_active_regions(void); | |
c713216d MG |
1352 | extern void free_bootmem_with_active_regions(int nid, |
1353 | unsigned long max_low_pfn); | |
08677214 YL |
1354 | int add_from_early_node_map(struct range *range, int az, |
1355 | int nr_range, int nid); | |
edbe7d23 YL |
1356 | u64 __init find_memory_core_early(int nid, u64 size, u64 align, |
1357 | u64 goal, u64 limit); | |
d52d53b8 | 1358 | typedef int (*work_fn_t)(unsigned long, unsigned long, void *); |
b5bc6c0e | 1359 | extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); |
c713216d | 1360 | extern void sparse_memory_present_with_active_regions(int nid); |
c713216d | 1361 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ |
f2dbcfa7 KH |
1362 | |
1363 | #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ | |
1364 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) | |
1365 | static inline int __early_pfn_to_nid(unsigned long pfn) | |
1366 | { | |
1367 | return 0; | |
1368 | } | |
1369 | #else | |
1370 | /* please see mm/page_alloc.c */ | |
1371 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
1372 | #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID | |
1373 | /* there is a per-arch backend function. */ | |
1374 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); | |
1375 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ | |
1376 | #endif | |
1377 | ||
0e0b864e | 1378 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
a2f3aa02 DH |
1379 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
1380 | unsigned long, enum memmap_context); | |
bc75d33f | 1381 | extern void setup_per_zone_wmarks(void); |
1b79acc9 | 1382 | extern int __meminit init_per_zone_wmark_min(void); |
1da177e4 | 1383 | extern void mem_init(void); |
8feae131 | 1384 | extern void __init mmap_init(void); |
b2b755b5 | 1385 | extern void show_mem(unsigned int flags); |
1da177e4 LT |
1386 | extern void si_meminfo(struct sysinfo * val); |
1387 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
3461b0af | 1388 | extern int after_bootmem; |
1da177e4 | 1389 | |
a238ab5b DH |
1390 | extern void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); |
1391 | ||
e7c8d5c9 | 1392 | extern void setup_per_cpu_pageset(void); |
e7c8d5c9 | 1393 | |
112067f0 SL |
1394 | extern void zone_pcp_update(struct zone *zone); |
1395 | ||
8feae131 | 1396 | /* nommu.c */ |
33e5d769 | 1397 | extern atomic_long_t mmap_pages_allocated; |
7e660872 | 1398 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); |
8feae131 | 1399 | |
1da177e4 LT |
1400 | /* prio_tree.c */ |
1401 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); | |
1402 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); | |
1403 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); | |
1404 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, | |
1405 | struct prio_tree_iter *iter); | |
1406 | ||
1407 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ | |
1408 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ | |
1409 | (vma = vma_prio_tree_next(vma, iter)); ) | |
1410 | ||
1411 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
1412 | struct list_head *list) | |
1413 | { | |
1414 | vma->shared.vm_set.parent = NULL; | |
1415 | list_add_tail(&vma->shared.vm_set.list, list); | |
1416 | } | |
1417 | ||
1418 | /* mmap.c */ | |
34b4e4aa | 1419 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); |
5beb4930 | 1420 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, |
1da177e4 LT |
1421 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
1422 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1423 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1424 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1425 | struct mempolicy *); | |
1426 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1427 | extern int split_vma(struct mm_struct *, | |
1428 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1429 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1430 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1431 | struct rb_node **, struct rb_node *); | |
a8fb5618 | 1432 | extern void unlink_file_vma(struct vm_area_struct *); |
1da177e4 LT |
1433 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
1434 | unsigned long addr, unsigned long len, pgoff_t pgoff); | |
1435 | extern void exit_mmap(struct mm_struct *); | |
925d1c40 | 1436 | |
7906d00c AA |
1437 | extern int mm_take_all_locks(struct mm_struct *mm); |
1438 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
1439 | ||
925d1c40 MH |
1440 | #ifdef CONFIG_PROC_FS |
1441 | /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ | |
1442 | extern void added_exe_file_vma(struct mm_struct *mm); | |
1443 | extern void removed_exe_file_vma(struct mm_struct *mm); | |
1444 | #else | |
1445 | static inline void added_exe_file_vma(struct mm_struct *mm) | |
1446 | {} | |
1447 | ||
1448 | static inline void removed_exe_file_vma(struct mm_struct *mm) | |
1449 | {} | |
1450 | #endif /* CONFIG_PROC_FS */ | |
1451 | ||
119f657c | 1452 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
fa5dc22f RM |
1453 | extern int install_special_mapping(struct mm_struct *mm, |
1454 | unsigned long addr, unsigned long len, | |
1455 | unsigned long flags, struct page **pages); | |
1da177e4 LT |
1456 | |
1457 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
1458 | ||
1459 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
1460 | unsigned long len, unsigned long prot, | |
1461 | unsigned long flag, unsigned long pgoff); | |
0165ab44 MS |
1462 | extern unsigned long mmap_region(struct file *file, unsigned long addr, |
1463 | unsigned long len, unsigned long flags, | |
5a6fe125 | 1464 | unsigned int vm_flags, unsigned long pgoff); |
1da177e4 LT |
1465 | |
1466 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, | |
1467 | unsigned long len, unsigned long prot, | |
1468 | unsigned long flag, unsigned long offset) | |
1469 | { | |
1470 | unsigned long ret = -EINVAL; | |
1471 | if ((offset + PAGE_ALIGN(len)) < offset) | |
1472 | goto out; | |
1473 | if (!(offset & ~PAGE_MASK)) | |
1474 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); | |
1475 | out: | |
1476 | return ret; | |
1477 | } | |
1478 | ||
1479 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
1480 | ||
1481 | extern unsigned long do_brk(unsigned long, unsigned long); | |
1482 | ||
1483 | /* filemap.c */ | |
1484 | extern unsigned long page_unuse(struct page *); | |
1485 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
d7339071 HR |
1486 | extern void truncate_inode_pages_range(struct address_space *, |
1487 | loff_t lstart, loff_t lend); | |
1da177e4 LT |
1488 | |
1489 | /* generic vm_area_ops exported for stackable file systems */ | |
d0217ac0 | 1490 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); |
1da177e4 LT |
1491 | |
1492 | /* mm/page-writeback.c */ | |
1493 | int write_one_page(struct page *page, int wait); | |
1cf6e7d8 | 1494 | void task_dirty_inc(struct task_struct *tsk); |
1da177e4 LT |
1495 | |
1496 | /* readahead.c */ | |
1497 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
1498 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
1da177e4 | 1499 | |
1da177e4 | 1500 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
7361f4d8 | 1501 | pgoff_t offset, unsigned long nr_to_read); |
cf914a7d RR |
1502 | |
1503 | void page_cache_sync_readahead(struct address_space *mapping, | |
1504 | struct file_ra_state *ra, | |
1505 | struct file *filp, | |
1506 | pgoff_t offset, | |
1507 | unsigned long size); | |
1508 | ||
1509 | void page_cache_async_readahead(struct address_space *mapping, | |
1510 | struct file_ra_state *ra, | |
1511 | struct file *filp, | |
1512 | struct page *pg, | |
1513 | pgoff_t offset, | |
1514 | unsigned long size); | |
1515 | ||
1da177e4 | 1516 | unsigned long max_sane_readahead(unsigned long nr); |
d30a1100 WF |
1517 | unsigned long ra_submit(struct file_ra_state *ra, |
1518 | struct address_space *mapping, | |
1519 | struct file *filp); | |
1da177e4 | 1520 | |
d05f3169 | 1521 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ |
46dea3d0 | 1522 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
d05f3169 MH |
1523 | |
1524 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | |
1525 | extern int expand_downwards(struct vm_area_struct *vma, | |
1526 | unsigned long address); | |
8ca3eb08 | 1527 | #if VM_GROWSUP |
46dea3d0 | 1528 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
8ca3eb08 TL |
1529 | #else |
1530 | #define expand_upwards(vma, address) do { } while (0) | |
9ab88515 | 1531 | #endif |
1da177e4 LT |
1532 | |
1533 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
1534 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
1535 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
1536 | struct vm_area_struct **pprev); | |
1537 | ||
1538 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
1539 | NULL if none. Assume start_addr < end_addr. */ | |
1540 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
1541 | { | |
1542 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
1543 | ||
1544 | if (vma && end_addr <= vma->vm_start) | |
1545 | vma = NULL; | |
1546 | return vma; | |
1547 | } | |
1548 | ||
1549 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
1550 | { | |
1551 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
1552 | } | |
1553 | ||
bad849b3 | 1554 | #ifdef CONFIG_MMU |
804af2cf | 1555 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
bad849b3 DH |
1556 | #else |
1557 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
1558 | { | |
1559 | return __pgprot(0); | |
1560 | } | |
1561 | #endif | |
1562 | ||
deceb6cd | 1563 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
deceb6cd HD |
1564 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
1565 | unsigned long pfn, unsigned long size, pgprot_t); | |
a145dd41 | 1566 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
e0dc0d8f NP |
1567 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
1568 | unsigned long pfn); | |
423bad60 NP |
1569 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, |
1570 | unsigned long pfn); | |
deceb6cd | 1571 | |
6aab341e | 1572 | struct page *follow_page(struct vm_area_struct *, unsigned long address, |
deceb6cd HD |
1573 | unsigned int foll_flags); |
1574 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
1575 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
1576 | #define FOLL_GET 0x04 /* do get_page on page */ | |
8e4b9a60 | 1577 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ |
58fa879e | 1578 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ |
318b275f GN |
1579 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO |
1580 | * and return without waiting upon it */ | |
110d74a9 | 1581 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ |
500d65d4 | 1582 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ |
69ebb83e | 1583 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ |
1da177e4 | 1584 | |
2f569afd | 1585 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, |
aee16b3c JF |
1586 | void *data); |
1587 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
1588 | unsigned long size, pte_fn_t fn, void *data); | |
1589 | ||
1da177e4 | 1590 | #ifdef CONFIG_PROC_FS |
ab50b8ed | 1591 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
1da177e4 | 1592 | #else |
ab50b8ed | 1593 | static inline void vm_stat_account(struct mm_struct *mm, |
1da177e4 LT |
1594 | unsigned long flags, struct file *file, long pages) |
1595 | { | |
1596 | } | |
1597 | #endif /* CONFIG_PROC_FS */ | |
1598 | ||
12d6f21e IM |
1599 | #ifdef CONFIG_DEBUG_PAGEALLOC |
1600 | extern int debug_pagealloc_enabled; | |
1601 | ||
1602 | extern void kernel_map_pages(struct page *page, int numpages, int enable); | |
1603 | ||
1604 | static inline void enable_debug_pagealloc(void) | |
1605 | { | |
1606 | debug_pagealloc_enabled = 1; | |
1607 | } | |
8a235efa RW |
1608 | #ifdef CONFIG_HIBERNATION |
1609 | extern bool kernel_page_present(struct page *page); | |
1610 | #endif /* CONFIG_HIBERNATION */ | |
12d6f21e | 1611 | #else |
1da177e4 | 1612 | static inline void |
9858db50 | 1613 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
12d6f21e IM |
1614 | static inline void enable_debug_pagealloc(void) |
1615 | { | |
1616 | } | |
8a235efa RW |
1617 | #ifdef CONFIG_HIBERNATION |
1618 | static inline bool kernel_page_present(struct page *page) { return true; } | |
1619 | #endif /* CONFIG_HIBERNATION */ | |
1da177e4 LT |
1620 | #endif |
1621 | ||
31db58b3 | 1622 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); |
1da177e4 | 1623 | #ifdef __HAVE_ARCH_GATE_AREA |
cae5d390 | 1624 | int in_gate_area_no_mm(unsigned long addr); |
83b964bb | 1625 | int in_gate_area(struct mm_struct *mm, unsigned long addr); |
1da177e4 | 1626 | #else |
cae5d390 SW |
1627 | int in_gate_area_no_mm(unsigned long addr); |
1628 | #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) | |
1da177e4 LT |
1629 | #endif /* __HAVE_ARCH_GATE_AREA */ |
1630 | ||
8d65af78 | 1631 | int drop_caches_sysctl_handler(struct ctl_table *, int, |
9d0243bc | 1632 | void __user *, size_t *, loff_t *); |
69e05944 | 1633 | unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, |
9d0243bc | 1634 | unsigned long lru_pages); |
9d0243bc | 1635 | |
7a9166e3 LY |
1636 | #ifndef CONFIG_MMU |
1637 | #define randomize_va_space 0 | |
1638 | #else | |
a62eaf15 | 1639 | extern int randomize_va_space; |
7a9166e3 | 1640 | #endif |
a62eaf15 | 1641 | |
045e72ac | 1642 | const char * arch_vma_name(struct vm_area_struct *vma); |
03252919 | 1643 | void print_vma_addr(char *prefix, unsigned long rip); |
e6e5494c | 1644 | |
9bdac914 YL |
1645 | void sparse_mem_maps_populate_node(struct page **map_map, |
1646 | unsigned long pnum_begin, | |
1647 | unsigned long pnum_end, | |
1648 | unsigned long map_count, | |
1649 | int nodeid); | |
1650 | ||
98f3cfc1 | 1651 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); |
29c71111 AW |
1652 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); |
1653 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); | |
1654 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); | |
1655 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | |
8f6aac41 | 1656 | void *vmemmap_alloc_block(unsigned long size, int node); |
9bdac914 | 1657 | void *vmemmap_alloc_block_buf(unsigned long size, int node); |
8f6aac41 | 1658 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); |
29c71111 AW |
1659 | int vmemmap_populate_basepages(struct page *start_page, |
1660 | unsigned long pages, int node); | |
1661 | int vmemmap_populate(struct page *start_page, unsigned long pages, int node); | |
c2b91e2e | 1662 | void vmemmap_populate_print_last(void); |
8f6aac41 | 1663 | |
6a46079c | 1664 | |
82ba011b AK |
1665 | enum mf_flags { |
1666 | MF_COUNT_INCREASED = 1 << 0, | |
1667 | }; | |
6a46079c | 1668 | extern void memory_failure(unsigned long pfn, int trapno); |
82ba011b | 1669 | extern int __memory_failure(unsigned long pfn, int trapno, int flags); |
847ce401 | 1670 | extern int unpoison_memory(unsigned long pfn); |
6a46079c AK |
1671 | extern int sysctl_memory_failure_early_kill; |
1672 | extern int sysctl_memory_failure_recovery; | |
facb6011 | 1673 | extern void shake_page(struct page *p, int access); |
6a46079c | 1674 | extern atomic_long_t mce_bad_pages; |
facb6011 | 1675 | extern int soft_offline_page(struct page *page, int flags); |
6a46079c | 1676 | |
718a3821 WF |
1677 | extern void dump_page(struct page *page); |
1678 | ||
47ad8475 AA |
1679 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) |
1680 | extern void clear_huge_page(struct page *page, | |
1681 | unsigned long addr, | |
1682 | unsigned int pages_per_huge_page); | |
1683 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
1684 | unsigned long addr, struct vm_area_struct *vma, | |
1685 | unsigned int pages_per_huge_page); | |
1686 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ | |
1687 | ||
1da177e4 LT |
1688 | #endif /* __KERNEL__ */ |
1689 | #endif /* _LINUX_MM_H */ |