]>
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/mmdebug.h> | |
9 | #include <linux/gfp.h> | |
10 | #include <linux/bug.h> | |
11 | #include <linux/list.h> | |
12 | #include <linux/mmzone.h> | |
13 | #include <linux/rbtree.h> | |
14 | #include <linux/atomic.h> | |
15 | #include <linux/debug_locks.h> | |
16 | #include <linux/mm_types.h> | |
17 | #include <linux/range.h> | |
18 | #include <linux/pfn.h> | |
19 | #include <linux/bit_spinlock.h> | |
20 | #include <linux/shrinker.h> | |
21 | #include <linux/resource.h> | |
22 | #include <linux/page_ext.h> | |
23 | ||
24 | struct mempolicy; | |
25 | struct anon_vma; | |
26 | struct anon_vma_chain; | |
27 | struct file_ra_state; | |
28 | struct user_struct; | |
29 | struct writeback_control; | |
30 | ||
31 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ | |
32 | extern unsigned long max_mapnr; | |
33 | ||
34 | static inline void set_max_mapnr(unsigned long limit) | |
35 | { | |
36 | max_mapnr = limit; | |
37 | } | |
38 | #else | |
39 | static inline void set_max_mapnr(unsigned long limit) { } | |
40 | #endif | |
41 | ||
42 | extern unsigned long totalram_pages; | |
43 | extern void * high_memory; | |
44 | extern int page_cluster; | |
45 | ||
46 | #ifdef CONFIG_SYSCTL | |
47 | extern int sysctl_legacy_va_layout; | |
48 | #else | |
49 | #define sysctl_legacy_va_layout 0 | |
50 | #endif | |
51 | ||
52 | #include <asm/page.h> | |
53 | #include <asm/pgtable.h> | |
54 | #include <asm/processor.h> | |
55 | ||
56 | #ifndef __pa_symbol | |
57 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | |
58 | #endif | |
59 | ||
60 | /* | |
61 | * To prevent common memory management code establishing | |
62 | * a zero page mapping on a read fault. | |
63 | * This macro should be defined within <asm/pgtable.h>. | |
64 | * s390 does this to prevent multiplexing of hardware bits | |
65 | * related to the physical page in case of virtualization. | |
66 | */ | |
67 | #ifndef mm_forbids_zeropage | |
68 | #define mm_forbids_zeropage(X) (0) | |
69 | #endif | |
70 | ||
71 | extern unsigned long sysctl_user_reserve_kbytes; | |
72 | extern unsigned long sysctl_admin_reserve_kbytes; | |
73 | ||
74 | extern int sysctl_overcommit_memory; | |
75 | extern int sysctl_overcommit_ratio; | |
76 | extern unsigned long sysctl_overcommit_kbytes; | |
77 | ||
78 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, | |
79 | size_t *, loff_t *); | |
80 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, | |
81 | size_t *, loff_t *); | |
82 | ||
83 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) | |
84 | ||
85 | /* to align the pointer to the (next) page boundary */ | |
86 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
87 | ||
88 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ | |
89 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE) | |
90 | ||
91 | /* | |
92 | * Linux kernel virtual memory manager primitives. | |
93 | * The idea being to have a "virtual" mm in the same way | |
94 | * we have a virtual fs - giving a cleaner interface to the | |
95 | * mm details, and allowing different kinds of memory mappings | |
96 | * (from shared memory to executable loading to arbitrary | |
97 | * mmap() functions). | |
98 | */ | |
99 | ||
100 | extern struct kmem_cache *vm_area_cachep; | |
101 | ||
102 | #ifndef CONFIG_MMU | |
103 | extern struct rb_root nommu_region_tree; | |
104 | extern struct rw_semaphore nommu_region_sem; | |
105 | ||
106 | extern unsigned int kobjsize(const void *objp); | |
107 | #endif | |
108 | ||
109 | /* | |
110 | * vm_flags in vm_area_struct, see mm_types.h. | |
111 | */ | |
112 | #define VM_NONE 0x00000000 | |
113 | ||
114 | #define VM_READ 0x00000001 /* currently active flags */ | |
115 | #define VM_WRITE 0x00000002 | |
116 | #define VM_EXEC 0x00000004 | |
117 | #define VM_SHARED 0x00000008 | |
118 | ||
119 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ | |
120 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ | |
121 | #define VM_MAYWRITE 0x00000020 | |
122 | #define VM_MAYEXEC 0x00000040 | |
123 | #define VM_MAYSHARE 0x00000080 | |
124 | ||
125 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
126 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ | |
127 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ | |
128 | ||
129 | #define VM_LOCKED 0x00002000 | |
130 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
131 | ||
132 | /* Used by sys_madvise() */ | |
133 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
134 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
135 | ||
136 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
137 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
138 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ | |
139 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ | |
140 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ | |
141 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ | |
142 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ | |
143 | #define VM_ARCH_2 0x02000000 | |
144 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ | |
145 | ||
146 | #ifdef CONFIG_MEM_SOFT_DIRTY | |
147 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | |
148 | #else | |
149 | # define VM_SOFTDIRTY 0 | |
150 | #endif | |
151 | ||
152 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ | |
153 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ | |
154 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
155 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ | |
156 | ||
157 | #if defined(CONFIG_X86) | |
158 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
159 | #elif defined(CONFIG_PPC) | |
160 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
161 | #elif defined(CONFIG_PARISC) | |
162 | # define VM_GROWSUP VM_ARCH_1 | |
163 | #elif defined(CONFIG_METAG) | |
164 | # define VM_GROWSUP VM_ARCH_1 | |
165 | #elif defined(CONFIG_IA64) | |
166 | # define VM_GROWSUP VM_ARCH_1 | |
167 | #elif !defined(CONFIG_MMU) | |
168 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
169 | #endif | |
170 | ||
171 | #if defined(CONFIG_X86) | |
172 | /* MPX specific bounds table or bounds directory */ | |
173 | # define VM_MPX VM_ARCH_2 | |
174 | #endif | |
175 | ||
176 | #ifndef VM_GROWSUP | |
177 | # define VM_GROWSUP VM_NONE | |
178 | #endif | |
179 | ||
180 | /* Bits set in the VMA until the stack is in its final location */ | |
181 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
182 | ||
183 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | |
184 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
185 | #endif | |
186 | ||
187 | #ifdef CONFIG_STACK_GROWSUP | |
188 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
189 | #else | |
190 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
191 | #endif | |
192 | ||
193 | /* | |
194 | * Special vmas that are non-mergable, non-mlock()able. | |
195 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | |
196 | */ | |
197 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) | |
198 | ||
199 | /* This mask defines which mm->def_flags a process can inherit its parent */ | |
200 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | |
201 | ||
202 | /* | |
203 | * mapping from the currently active vm_flags protection bits (the | |
204 | * low four bits) to a page protection mask.. | |
205 | */ | |
206 | extern pgprot_t protection_map[16]; | |
207 | ||
208 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ | |
209 | #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ | |
210 | #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ | |
211 | #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ | |
212 | #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ | |
213 | #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ | |
214 | #define FAULT_FLAG_TRIED 0x40 /* second try */ | |
215 | #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */ | |
216 | ||
217 | /* | |
218 | * vm_fault is filled by the the pagefault handler and passed to the vma's | |
219 | * ->fault function. The vma's ->fault is responsible for returning a bitmask | |
220 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
221 | * | |
222 | * pgoff should be used in favour of virtual_address, if possible. If pgoff | |
223 | * is used, one may implement ->remap_pages to get nonlinear mapping support. | |
224 | */ | |
225 | struct vm_fault { | |
226 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | |
227 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
228 | void __user *virtual_address; /* Faulting virtual address */ | |
229 | ||
230 | struct page *page; /* ->fault handlers should return a | |
231 | * page here, unless VM_FAULT_NOPAGE | |
232 | * is set (which is also implied by | |
233 | * VM_FAULT_ERROR). | |
234 | */ | |
235 | /* for ->map_pages() only */ | |
236 | pgoff_t max_pgoff; /* map pages for offset from pgoff till | |
237 | * max_pgoff inclusive */ | |
238 | pte_t *pte; /* pte entry associated with ->pgoff */ | |
239 | }; | |
240 | ||
241 | /* | |
242 | * These are the virtual MM functions - opening of an area, closing and | |
243 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
244 | * to the functions called when a no-page or a wp-page exception occurs. | |
245 | */ | |
246 | struct vm_operations_struct { | |
247 | void (*open)(struct vm_area_struct * area); | |
248 | void (*close)(struct vm_area_struct * area); | |
249 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
250 | void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
251 | ||
252 | /* notification that a previously read-only page is about to become | |
253 | * writable, if an error is returned it will cause a SIGBUS */ | |
254 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
255 | ||
256 | /* called by access_process_vm when get_user_pages() fails, typically | |
257 | * for use by special VMAs that can switch between memory and hardware | |
258 | */ | |
259 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
260 | void *buf, int len, int write); | |
261 | ||
262 | /* Called by the /proc/PID/maps code to ask the vma whether it | |
263 | * has a special name. Returning non-NULL will also cause this | |
264 | * vma to be dumped unconditionally. */ | |
265 | const char *(*name)(struct vm_area_struct *vma); | |
266 | ||
267 | #ifdef CONFIG_NUMA | |
268 | /* | |
269 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
270 | * to hold the policy upon return. Caller should pass NULL @new to | |
271 | * remove a policy and fall back to surrounding context--i.e. do not | |
272 | * install a MPOL_DEFAULT policy, nor the task or system default | |
273 | * mempolicy. | |
274 | */ | |
275 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | |
276 | ||
277 | /* | |
278 | * get_policy() op must add reference [mpol_get()] to any policy at | |
279 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
280 | * in mm/mempolicy.c will do this automatically. | |
281 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
282 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | |
283 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | |
284 | * must return NULL--i.e., do not "fallback" to task or system default | |
285 | * policy. | |
286 | */ | |
287 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | |
288 | unsigned long addr); | |
289 | #endif | |
290 | /* called by sys_remap_file_pages() to populate non-linear mapping */ | |
291 | int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr, | |
292 | unsigned long size, pgoff_t pgoff); | |
293 | }; | |
294 | ||
295 | struct mmu_gather; | |
296 | struct inode; | |
297 | ||
298 | #define page_private(page) ((page)->private) | |
299 | #define set_page_private(page, v) ((page)->private = (v)) | |
300 | ||
301 | /* It's valid only if the page is free path or free_list */ | |
302 | static inline void set_freepage_migratetype(struct page *page, int migratetype) | |
303 | { | |
304 | page->index = migratetype; | |
305 | } | |
306 | ||
307 | /* It's valid only if the page is free path or free_list */ | |
308 | static inline int get_freepage_migratetype(struct page *page) | |
309 | { | |
310 | return page->index; | |
311 | } | |
312 | ||
313 | /* | |
314 | * FIXME: take this include out, include page-flags.h in | |
315 | * files which need it (119 of them) | |
316 | */ | |
317 | #include <linux/page-flags.h> | |
318 | #include <linux/huge_mm.h> | |
319 | ||
320 | /* | |
321 | * Methods to modify the page usage count. | |
322 | * | |
323 | * What counts for a page usage: | |
324 | * - cache mapping (page->mapping) | |
325 | * - private data (page->private) | |
326 | * - page mapped in a task's page tables, each mapping | |
327 | * is counted separately | |
328 | * | |
329 | * Also, many kernel routines increase the page count before a critical | |
330 | * routine so they can be sure the page doesn't go away from under them. | |
331 | */ | |
332 | ||
333 | /* | |
334 | * Drop a ref, return true if the refcount fell to zero (the page has no users) | |
335 | */ | |
336 | static inline int put_page_testzero(struct page *page) | |
337 | { | |
338 | VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page); | |
339 | return atomic_dec_and_test(&page->_count); | |
340 | } | |
341 | ||
342 | /* | |
343 | * Try to grab a ref unless the page has a refcount of zero, return false if | |
344 | * that is the case. | |
345 | * This can be called when MMU is off so it must not access | |
346 | * any of the virtual mappings. | |
347 | */ | |
348 | static inline int get_page_unless_zero(struct page *page) | |
349 | { | |
350 | return atomic_inc_not_zero(&page->_count); | |
351 | } | |
352 | ||
353 | /* | |
354 | * Try to drop a ref unless the page has a refcount of one, return false if | |
355 | * that is the case. | |
356 | * This is to make sure that the refcount won't become zero after this drop. | |
357 | * This can be called when MMU is off so it must not access | |
358 | * any of the virtual mappings. | |
359 | */ | |
360 | static inline int put_page_unless_one(struct page *page) | |
361 | { | |
362 | return atomic_add_unless(&page->_count, -1, 1); | |
363 | } | |
364 | ||
365 | extern int page_is_ram(unsigned long pfn); | |
366 | extern int region_is_ram(resource_size_t phys_addr, unsigned long size); | |
367 | ||
368 | /* Support for virtually mapped pages */ | |
369 | struct page *vmalloc_to_page(const void *addr); | |
370 | unsigned long vmalloc_to_pfn(const void *addr); | |
371 | ||
372 | /* | |
373 | * Determine if an address is within the vmalloc range | |
374 | * | |
375 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
376 | * is no special casing required. | |
377 | */ | |
378 | static inline int is_vmalloc_addr(const void *x) | |
379 | { | |
380 | #ifdef CONFIG_MMU | |
381 | unsigned long addr = (unsigned long)x; | |
382 | ||
383 | return addr >= VMALLOC_START && addr < VMALLOC_END; | |
384 | #else | |
385 | return 0; | |
386 | #endif | |
387 | } | |
388 | #ifdef CONFIG_MMU | |
389 | extern int is_vmalloc_or_module_addr(const void *x); | |
390 | #else | |
391 | static inline int is_vmalloc_or_module_addr(const void *x) | |
392 | { | |
393 | return 0; | |
394 | } | |
395 | #endif | |
396 | ||
397 | extern void kvfree(const void *addr); | |
398 | ||
399 | static inline void compound_lock(struct page *page) | |
400 | { | |
401 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
402 | VM_BUG_ON_PAGE(PageSlab(page), page); | |
403 | bit_spin_lock(PG_compound_lock, &page->flags); | |
404 | #endif | |
405 | } | |
406 | ||
407 | static inline void compound_unlock(struct page *page) | |
408 | { | |
409 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
410 | VM_BUG_ON_PAGE(PageSlab(page), page); | |
411 | bit_spin_unlock(PG_compound_lock, &page->flags); | |
412 | #endif | |
413 | } | |
414 | ||
415 | static inline unsigned long compound_lock_irqsave(struct page *page) | |
416 | { | |
417 | unsigned long uninitialized_var(flags); | |
418 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
419 | local_irq_save(flags); | |
420 | compound_lock(page); | |
421 | #endif | |
422 | return flags; | |
423 | } | |
424 | ||
425 | static inline void compound_unlock_irqrestore(struct page *page, | |
426 | unsigned long flags) | |
427 | { | |
428 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
429 | compound_unlock(page); | |
430 | local_irq_restore(flags); | |
431 | #endif | |
432 | } | |
433 | ||
434 | static inline struct page *compound_head_by_tail(struct page *tail) | |
435 | { | |
436 | struct page *head = tail->first_page; | |
437 | ||
438 | /* | |
439 | * page->first_page may be a dangling pointer to an old | |
440 | * compound page, so recheck that it is still a tail | |
441 | * page before returning. | |
442 | */ | |
443 | smp_rmb(); | |
444 | if (likely(PageTail(tail))) | |
445 | return head; | |
446 | return tail; | |
447 | } | |
448 | ||
449 | /* | |
450 | * Since either compound page could be dismantled asynchronously in THP | |
451 | * or we access asynchronously arbitrary positioned struct page, there | |
452 | * would be tail flag race. To handle this race, we should call | |
453 | * smp_rmb() before checking tail flag. compound_head_by_tail() did it. | |
454 | */ | |
455 | static inline struct page *compound_head(struct page *page) | |
456 | { | |
457 | if (unlikely(PageTail(page))) | |
458 | return compound_head_by_tail(page); | |
459 | return page; | |
460 | } | |
461 | ||
462 | /* | |
463 | * If we access compound page synchronously such as access to | |
464 | * allocated page, there is no need to handle tail flag race, so we can | |
465 | * check tail flag directly without any synchronization primitive. | |
466 | */ | |
467 | static inline struct page *compound_head_fast(struct page *page) | |
468 | { | |
469 | if (unlikely(PageTail(page))) | |
470 | return page->first_page; | |
471 | return page; | |
472 | } | |
473 | ||
474 | /* | |
475 | * The atomic page->_mapcount, starts from -1: so that transitions | |
476 | * both from it and to it can be tracked, using atomic_inc_and_test | |
477 | * and atomic_add_negative(-1). | |
478 | */ | |
479 | static inline void page_mapcount_reset(struct page *page) | |
480 | { | |
481 | atomic_set(&(page)->_mapcount, -1); | |
482 | } | |
483 | ||
484 | static inline int page_mapcount(struct page *page) | |
485 | { | |
486 | return atomic_read(&(page)->_mapcount) + 1; | |
487 | } | |
488 | ||
489 | static inline int page_count(struct page *page) | |
490 | { | |
491 | return atomic_read(&compound_head(page)->_count); | |
492 | } | |
493 | ||
494 | #ifdef CONFIG_HUGETLB_PAGE | |
495 | extern int PageHeadHuge(struct page *page_head); | |
496 | #else /* CONFIG_HUGETLB_PAGE */ | |
497 | static inline int PageHeadHuge(struct page *page_head) | |
498 | { | |
499 | return 0; | |
500 | } | |
501 | #endif /* CONFIG_HUGETLB_PAGE */ | |
502 | ||
503 | static inline bool __compound_tail_refcounted(struct page *page) | |
504 | { | |
505 | return !PageSlab(page) && !PageHeadHuge(page); | |
506 | } | |
507 | ||
508 | /* | |
509 | * This takes a head page as parameter and tells if the | |
510 | * tail page reference counting can be skipped. | |
511 | * | |
512 | * For this to be safe, PageSlab and PageHeadHuge must remain true on | |
513 | * any given page where they return true here, until all tail pins | |
514 | * have been released. | |
515 | */ | |
516 | static inline bool compound_tail_refcounted(struct page *page) | |
517 | { | |
518 | VM_BUG_ON_PAGE(!PageHead(page), page); | |
519 | return __compound_tail_refcounted(page); | |
520 | } | |
521 | ||
522 | static inline void get_huge_page_tail(struct page *page) | |
523 | { | |
524 | /* | |
525 | * __split_huge_page_refcount() cannot run from under us. | |
526 | */ | |
527 | VM_BUG_ON_PAGE(!PageTail(page), page); | |
528 | VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); | |
529 | VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page); | |
530 | if (compound_tail_refcounted(page->first_page)) | |
531 | atomic_inc(&page->_mapcount); | |
532 | } | |
533 | ||
534 | extern bool __get_page_tail(struct page *page); | |
535 | ||
536 | static inline void get_page(struct page *page) | |
537 | { | |
538 | if (unlikely(PageTail(page))) | |
539 | if (likely(__get_page_tail(page))) | |
540 | return; | |
541 | /* | |
542 | * Getting a normal page or the head of a compound page | |
543 | * requires to already have an elevated page->_count. | |
544 | */ | |
545 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); | |
546 | atomic_inc(&page->_count); | |
547 | } | |
548 | ||
549 | static inline struct page *virt_to_head_page(const void *x) | |
550 | { | |
551 | struct page *page = virt_to_page(x); | |
552 | ||
553 | /* | |
554 | * We don't need to worry about synchronization of tail flag | |
555 | * when we call virt_to_head_page() since it is only called for | |
556 | * already allocated page and this page won't be freed until | |
557 | * this virt_to_head_page() is finished. So use _fast variant. | |
558 | */ | |
559 | return compound_head_fast(page); | |
560 | } | |
561 | ||
562 | /* | |
563 | * Setup the page count before being freed into the page allocator for | |
564 | * the first time (boot or memory hotplug) | |
565 | */ | |
566 | static inline void init_page_count(struct page *page) | |
567 | { | |
568 | atomic_set(&page->_count, 1); | |
569 | } | |
570 | ||
571 | /* | |
572 | * PageBuddy() indicate that the page is free and in the buddy system | |
573 | * (see mm/page_alloc.c). | |
574 | * | |
575 | * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to | |
576 | * -2 so that an underflow of the page_mapcount() won't be mistaken | |
577 | * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very | |
578 | * efficiently by most CPU architectures. | |
579 | */ | |
580 | #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) | |
581 | ||
582 | static inline int PageBuddy(struct page *page) | |
583 | { | |
584 | return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; | |
585 | } | |
586 | ||
587 | static inline void __SetPageBuddy(struct page *page) | |
588 | { | |
589 | VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page); | |
590 | atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); | |
591 | } | |
592 | ||
593 | static inline void __ClearPageBuddy(struct page *page) | |
594 | { | |
595 | VM_BUG_ON_PAGE(!PageBuddy(page), page); | |
596 | atomic_set(&page->_mapcount, -1); | |
597 | } | |
598 | ||
599 | #define PAGE_BALLOON_MAPCOUNT_VALUE (-256) | |
600 | ||
601 | static inline int PageBalloon(struct page *page) | |
602 | { | |
603 | return atomic_read(&page->_mapcount) == PAGE_BALLOON_MAPCOUNT_VALUE; | |
604 | } | |
605 | ||
606 | static inline void __SetPageBalloon(struct page *page) | |
607 | { | |
608 | VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page); | |
609 | atomic_set(&page->_mapcount, PAGE_BALLOON_MAPCOUNT_VALUE); | |
610 | } | |
611 | ||
612 | static inline void __ClearPageBalloon(struct page *page) | |
613 | { | |
614 | VM_BUG_ON_PAGE(!PageBalloon(page), page); | |
615 | atomic_set(&page->_mapcount, -1); | |
616 | } | |
617 | ||
618 | void put_page(struct page *page); | |
619 | void put_pages_list(struct list_head *pages); | |
620 | ||
621 | void split_page(struct page *page, unsigned int order); | |
622 | int split_free_page(struct page *page); | |
623 | ||
624 | /* | |
625 | * Compound pages have a destructor function. Provide a | |
626 | * prototype for that function and accessor functions. | |
627 | * These are _only_ valid on the head of a PG_compound page. | |
628 | */ | |
629 | typedef void compound_page_dtor(struct page *); | |
630 | ||
631 | static inline void set_compound_page_dtor(struct page *page, | |
632 | compound_page_dtor *dtor) | |
633 | { | |
634 | page[1].lru.next = (void *)dtor; | |
635 | } | |
636 | ||
637 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
638 | { | |
639 | return (compound_page_dtor *)page[1].lru.next; | |
640 | } | |
641 | ||
642 | static inline int compound_order(struct page *page) | |
643 | { | |
644 | if (!PageHead(page)) | |
645 | return 0; | |
646 | return (unsigned long)page[1].lru.prev; | |
647 | } | |
648 | ||
649 | static inline void set_compound_order(struct page *page, unsigned long order) | |
650 | { | |
651 | page[1].lru.prev = (void *)order; | |
652 | } | |
653 | ||
654 | #ifdef CONFIG_MMU | |
655 | /* | |
656 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
657 | * servicing faults for write access. In the normal case, do always want | |
658 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
659 | * that do not have writing enabled, when used by access_process_vm. | |
660 | */ | |
661 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
662 | { | |
663 | if (likely(vma->vm_flags & VM_WRITE)) | |
664 | pte = pte_mkwrite(pte); | |
665 | return pte; | |
666 | } | |
667 | ||
668 | void do_set_pte(struct vm_area_struct *vma, unsigned long address, | |
669 | struct page *page, pte_t *pte, bool write, bool anon); | |
670 | #endif | |
671 | ||
672 | /* | |
673 | * Multiple processes may "see" the same page. E.g. for untouched | |
674 | * mappings of /dev/null, all processes see the same page full of | |
675 | * zeroes, and text pages of executables and shared libraries have | |
676 | * only one copy in memory, at most, normally. | |
677 | * | |
678 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
679 | * page_count() == 0 means the page is free. page->lru is then used for | |
680 | * freelist management in the buddy allocator. | |
681 | * page_count() > 0 means the page has been allocated. | |
682 | * | |
683 | * Pages are allocated by the slab allocator in order to provide memory | |
684 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
685 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
686 | * unless a particular usage is carefully commented. (the responsibility of | |
687 | * freeing the kmalloc memory is the caller's, of course). | |
688 | * | |
689 | * A page may be used by anyone else who does a __get_free_page(). | |
690 | * In this case, page_count still tracks the references, and should only | |
691 | * be used through the normal accessor functions. The top bits of page->flags | |
692 | * and page->virtual store page management information, but all other fields | |
693 | * are unused and could be used privately, carefully. The management of this | |
694 | * page is the responsibility of the one who allocated it, and those who have | |
695 | * subsequently been given references to it. | |
696 | * | |
697 | * The other pages (we may call them "pagecache pages") are completely | |
698 | * managed by the Linux memory manager: I/O, buffers, swapping etc. | |
699 | * The following discussion applies only to them. | |
700 | * | |
701 | * A pagecache page contains an opaque `private' member, which belongs to the | |
702 | * page's address_space. Usually, this is the address of a circular list of | |
703 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
704 | * into the filesystem to release these pages. | |
705 | * | |
706 | * A page may belong to an inode's memory mapping. In this case, page->mapping | |
707 | * is the pointer to the inode, and page->index is the file offset of the page, | |
708 | * in units of PAGE_CACHE_SIZE. | |
709 | * | |
710 | * If pagecache pages are not associated with an inode, they are said to be | |
711 | * anonymous pages. These may become associated with the swapcache, and in that | |
712 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
713 | * | |
714 | * In either case (swapcache or inode backed), the pagecache itself holds one | |
715 | * reference to the page. Setting PG_private should also increment the | |
716 | * refcount. The each user mapping also has a reference to the page. | |
717 | * | |
718 | * The pagecache pages are stored in a per-mapping radix tree, which is | |
719 | * rooted at mapping->page_tree, and indexed by offset. | |
720 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
721 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
722 | * | |
723 | * All pagecache pages may be subject to I/O: | |
724 | * - inode pages may need to be read from disk, | |
725 | * - inode pages which have been modified and are MAP_SHARED may need | |
726 | * to be written back to the inode on disk, | |
727 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
728 | * modified may need to be swapped out to swap space and (later) to be read | |
729 | * back into memory. | |
730 | */ | |
731 | ||
732 | /* | |
733 | * The zone field is never updated after free_area_init_core() | |
734 | * sets it, so none of the operations on it need to be atomic. | |
735 | */ | |
736 | ||
737 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ | |
738 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) | |
739 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) | |
740 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
741 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) | |
742 | ||
743 | /* | |
744 | * Define the bit shifts to access each section. For non-existent | |
745 | * sections we define the shift as 0; that plus a 0 mask ensures | |
746 | * the compiler will optimise away reference to them. | |
747 | */ | |
748 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) | |
749 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
750 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
751 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) | |
752 | ||
753 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ | |
754 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
755 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
756 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ | |
757 | SECTIONS_PGOFF : ZONES_PGOFF) | |
758 | #else | |
759 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) | |
760 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ | |
761 | NODES_PGOFF : ZONES_PGOFF) | |
762 | #endif | |
763 | ||
764 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) | |
765 | ||
766 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
767 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
768 | #endif | |
769 | ||
770 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) | |
771 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
772 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
773 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) | |
774 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) | |
775 | ||
776 | static inline enum zone_type page_zonenum(const struct page *page) | |
777 | { | |
778 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; | |
779 | } | |
780 | ||
781 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) | |
782 | #define SECTION_IN_PAGE_FLAGS | |
783 | #endif | |
784 | ||
785 | /* | |
786 | * The identification function is mainly used by the buddy allocator for | |
787 | * determining if two pages could be buddies. We are not really identifying | |
788 | * the zone since we could be using the section number id if we do not have | |
789 | * node id available in page flags. | |
790 | * We only guarantee that it will return the same value for two combinable | |
791 | * pages in a zone. | |
792 | */ | |
793 | static inline int page_zone_id(struct page *page) | |
794 | { | |
795 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; | |
796 | } | |
797 | ||
798 | static inline int zone_to_nid(struct zone *zone) | |
799 | { | |
800 | #ifdef CONFIG_NUMA | |
801 | return zone->node; | |
802 | #else | |
803 | return 0; | |
804 | #endif | |
805 | } | |
806 | ||
807 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
808 | extern int page_to_nid(const struct page *page); | |
809 | #else | |
810 | static inline int page_to_nid(const struct page *page) | |
811 | { | |
812 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; | |
813 | } | |
814 | #endif | |
815 | ||
816 | #ifdef CONFIG_NUMA_BALANCING | |
817 | static inline int cpu_pid_to_cpupid(int cpu, int pid) | |
818 | { | |
819 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); | |
820 | } | |
821 | ||
822 | static inline int cpupid_to_pid(int cpupid) | |
823 | { | |
824 | return cpupid & LAST__PID_MASK; | |
825 | } | |
826 | ||
827 | static inline int cpupid_to_cpu(int cpupid) | |
828 | { | |
829 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; | |
830 | } | |
831 | ||
832 | static inline int cpupid_to_nid(int cpupid) | |
833 | { | |
834 | return cpu_to_node(cpupid_to_cpu(cpupid)); | |
835 | } | |
836 | ||
837 | static inline bool cpupid_pid_unset(int cpupid) | |
838 | { | |
839 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); | |
840 | } | |
841 | ||
842 | static inline bool cpupid_cpu_unset(int cpupid) | |
843 | { | |
844 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); | |
845 | } | |
846 | ||
847 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) | |
848 | { | |
849 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | |
850 | } | |
851 | ||
852 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | |
853 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS | |
854 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
855 | { | |
856 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); | |
857 | } | |
858 | ||
859 | static inline int page_cpupid_last(struct page *page) | |
860 | { | |
861 | return page->_last_cpupid; | |
862 | } | |
863 | static inline void page_cpupid_reset_last(struct page *page) | |
864 | { | |
865 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; | |
866 | } | |
867 | #else | |
868 | static inline int page_cpupid_last(struct page *page) | |
869 | { | |
870 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; | |
871 | } | |
872 | ||
873 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); | |
874 | ||
875 | static inline void page_cpupid_reset_last(struct page *page) | |
876 | { | |
877 | int cpupid = (1 << LAST_CPUPID_SHIFT) - 1; | |
878 | ||
879 | page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT); | |
880 | page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT; | |
881 | } | |
882 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ | |
883 | #else /* !CONFIG_NUMA_BALANCING */ | |
884 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
885 | { | |
886 | return page_to_nid(page); /* XXX */ | |
887 | } | |
888 | ||
889 | static inline int page_cpupid_last(struct page *page) | |
890 | { | |
891 | return page_to_nid(page); /* XXX */ | |
892 | } | |
893 | ||
894 | static inline int cpupid_to_nid(int cpupid) | |
895 | { | |
896 | return -1; | |
897 | } | |
898 | ||
899 | static inline int cpupid_to_pid(int cpupid) | |
900 | { | |
901 | return -1; | |
902 | } | |
903 | ||
904 | static inline int cpupid_to_cpu(int cpupid) | |
905 | { | |
906 | return -1; | |
907 | } | |
908 | ||
909 | static inline int cpu_pid_to_cpupid(int nid, int pid) | |
910 | { | |
911 | return -1; | |
912 | } | |
913 | ||
914 | static inline bool cpupid_pid_unset(int cpupid) | |
915 | { | |
916 | return 1; | |
917 | } | |
918 | ||
919 | static inline void page_cpupid_reset_last(struct page *page) | |
920 | { | |
921 | } | |
922 | ||
923 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | |
924 | { | |
925 | return false; | |
926 | } | |
927 | #endif /* CONFIG_NUMA_BALANCING */ | |
928 | ||
929 | static inline struct zone *page_zone(const struct page *page) | |
930 | { | |
931 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
932 | } | |
933 | ||
934 | #ifdef SECTION_IN_PAGE_FLAGS | |
935 | static inline void set_page_section(struct page *page, unsigned long section) | |
936 | { | |
937 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
938 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
939 | } | |
940 | ||
941 | static inline unsigned long page_to_section(const struct page *page) | |
942 | { | |
943 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
944 | } | |
945 | #endif | |
946 | ||
947 | static inline void set_page_zone(struct page *page, enum zone_type zone) | |
948 | { | |
949 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
950 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
951 | } | |
952 | ||
953 | static inline void set_page_node(struct page *page, unsigned long node) | |
954 | { | |
955 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
956 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
957 | } | |
958 | ||
959 | static inline void set_page_links(struct page *page, enum zone_type zone, | |
960 | unsigned long node, unsigned long pfn) | |
961 | { | |
962 | set_page_zone(page, zone); | |
963 | set_page_node(page, node); | |
964 | #ifdef SECTION_IN_PAGE_FLAGS | |
965 | set_page_section(page, pfn_to_section_nr(pfn)); | |
966 | #endif | |
967 | } | |
968 | ||
969 | /* | |
970 | * Some inline functions in vmstat.h depend on page_zone() | |
971 | */ | |
972 | #include <linux/vmstat.h> | |
973 | ||
974 | static __always_inline void *lowmem_page_address(const struct page *page) | |
975 | { | |
976 | return __va(PFN_PHYS(page_to_pfn(page))); | |
977 | } | |
978 | ||
979 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
980 | #define HASHED_PAGE_VIRTUAL | |
981 | #endif | |
982 | ||
983 | #if defined(WANT_PAGE_VIRTUAL) | |
984 | static inline void *page_address(const struct page *page) | |
985 | { | |
986 | return page->virtual; | |
987 | } | |
988 | static inline void set_page_address(struct page *page, void *address) | |
989 | { | |
990 | page->virtual = address; | |
991 | } | |
992 | #define page_address_init() do { } while(0) | |
993 | #endif | |
994 | ||
995 | #if defined(HASHED_PAGE_VIRTUAL) | |
996 | void *page_address(const struct page *page); | |
997 | void set_page_address(struct page *page, void *virtual); | |
998 | void page_address_init(void); | |
999 | #endif | |
1000 | ||
1001 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
1002 | #define page_address(page) lowmem_page_address(page) | |
1003 | #define set_page_address(page, address) do { } while(0) | |
1004 | #define page_address_init() do { } while(0) | |
1005 | #endif | |
1006 | ||
1007 | /* | |
1008 | * On an anonymous page mapped into a user virtual memory area, | |
1009 | * page->mapping points to its anon_vma, not to a struct address_space; | |
1010 | * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. | |
1011 | * | |
1012 | * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, | |
1013 | * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; | |
1014 | * and then page->mapping points, not to an anon_vma, but to a private | |
1015 | * structure which KSM associates with that merged page. See ksm.h. | |
1016 | * | |
1017 | * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. | |
1018 | * | |
1019 | * Please note that, confusingly, "page_mapping" refers to the inode | |
1020 | * address_space which maps the page from disk; whereas "page_mapped" | |
1021 | * refers to user virtual address space into which the page is mapped. | |
1022 | */ | |
1023 | #define PAGE_MAPPING_ANON 1 | |
1024 | #define PAGE_MAPPING_KSM 2 | |
1025 | #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) | |
1026 | ||
1027 | extern struct address_space *page_mapping(struct page *page); | |
1028 | ||
1029 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ | |
1030 | static inline void *page_rmapping(struct page *page) | |
1031 | { | |
1032 | return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); | |
1033 | } | |
1034 | ||
1035 | extern struct address_space *__page_file_mapping(struct page *); | |
1036 | ||
1037 | static inline | |
1038 | struct address_space *page_file_mapping(struct page *page) | |
1039 | { | |
1040 | if (unlikely(PageSwapCache(page))) | |
1041 | return __page_file_mapping(page); | |
1042 | ||
1043 | return page->mapping; | |
1044 | } | |
1045 | ||
1046 | static inline int PageAnon(struct page *page) | |
1047 | { | |
1048 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; | |
1049 | } | |
1050 | ||
1051 | /* | |
1052 | * Return the pagecache index of the passed page. Regular pagecache pages | |
1053 | * use ->index whereas swapcache pages use ->private | |
1054 | */ | |
1055 | static inline pgoff_t page_index(struct page *page) | |
1056 | { | |
1057 | if (unlikely(PageSwapCache(page))) | |
1058 | return page_private(page); | |
1059 | return page->index; | |
1060 | } | |
1061 | ||
1062 | extern pgoff_t __page_file_index(struct page *page); | |
1063 | ||
1064 | /* | |
1065 | * Return the file index of the page. Regular pagecache pages use ->index | |
1066 | * whereas swapcache pages use swp_offset(->private) | |
1067 | */ | |
1068 | static inline pgoff_t page_file_index(struct page *page) | |
1069 | { | |
1070 | if (unlikely(PageSwapCache(page))) | |
1071 | return __page_file_index(page); | |
1072 | ||
1073 | return page->index; | |
1074 | } | |
1075 | ||
1076 | /* | |
1077 | * Return true if this page is mapped into pagetables. | |
1078 | */ | |
1079 | static inline int page_mapped(struct page *page) | |
1080 | { | |
1081 | return atomic_read(&(page)->_mapcount) >= 0; | |
1082 | } | |
1083 | ||
1084 | /* | |
1085 | * Different kinds of faults, as returned by handle_mm_fault(). | |
1086 | * Used to decide whether a process gets delivered SIGBUS or | |
1087 | * just gets major/minor fault counters bumped up. | |
1088 | */ | |
1089 | ||
1090 | #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ | |
1091 | ||
1092 | #define VM_FAULT_OOM 0x0001 | |
1093 | #define VM_FAULT_SIGBUS 0x0002 | |
1094 | #define VM_FAULT_MAJOR 0x0004 | |
1095 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | |
1096 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ | |
1097 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ | |
1098 | #define VM_FAULT_SIGSEGV 0x0040 | |
1099 | ||
1100 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ | |
1101 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | |
1102 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ | |
1103 | #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ | |
1104 | ||
1105 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ | |
1106 | ||
1107 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \ | |
1108 | VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \ | |
1109 | VM_FAULT_FALLBACK) | |
1110 | ||
1111 | /* Encode hstate index for a hwpoisoned large page */ | |
1112 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) | |
1113 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) | |
1114 | ||
1115 | /* | |
1116 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
1117 | */ | |
1118 | extern void pagefault_out_of_memory(void); | |
1119 | ||
1120 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | |
1121 | ||
1122 | /* | |
1123 | * Flags passed to show_mem() and show_free_areas() to suppress output in | |
1124 | * various contexts. | |
1125 | */ | |
1126 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ | |
1127 | ||
1128 | extern void show_free_areas(unsigned int flags); | |
1129 | extern bool skip_free_areas_node(unsigned int flags, int nid); | |
1130 | ||
1131 | int shmem_zero_setup(struct vm_area_struct *); | |
1132 | #ifdef CONFIG_SHMEM | |
1133 | bool shmem_mapping(struct address_space *mapping); | |
1134 | #else | |
1135 | static inline bool shmem_mapping(struct address_space *mapping) | |
1136 | { | |
1137 | return false; | |
1138 | } | |
1139 | #endif | |
1140 | ||
1141 | extern int can_do_mlock(void); | |
1142 | extern int user_shm_lock(size_t, struct user_struct *); | |
1143 | extern void user_shm_unlock(size_t, struct user_struct *); | |
1144 | ||
1145 | /* | |
1146 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
1147 | */ | |
1148 | struct zap_details { | |
1149 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
1150 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
1151 | pgoff_t last_index; /* Highest page->index to unmap */ | |
1152 | }; | |
1153 | ||
1154 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, | |
1155 | pte_t pte); | |
1156 | ||
1157 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, | |
1158 | unsigned long size); | |
1159 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, | |
1160 | unsigned long size, struct zap_details *); | |
1161 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, | |
1162 | unsigned long start, unsigned long end); | |
1163 | ||
1164 | /** | |
1165 | * mm_walk - callbacks for walk_page_range | |
1166 | * @pgd_entry: if set, called for each non-empty PGD (top-level) entry | |
1167 | * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry | |
1168 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry | |
1169 | * this handler is required to be able to handle | |
1170 | * pmd_trans_huge() pmds. They may simply choose to | |
1171 | * split_huge_page() instead of handling it explicitly. | |
1172 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry | |
1173 | * @pte_hole: if set, called for each hole at all levels | |
1174 | * @hugetlb_entry: if set, called for each hugetlb entry | |
1175 | * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry | |
1176 | * is used. | |
1177 | * | |
1178 | * (see walk_page_range for more details) | |
1179 | */ | |
1180 | struct mm_walk { | |
1181 | int (*pgd_entry)(pgd_t *pgd, unsigned long addr, | |
1182 | unsigned long next, struct mm_walk *walk); | |
1183 | int (*pud_entry)(pud_t *pud, unsigned long addr, | |
1184 | unsigned long next, struct mm_walk *walk); | |
1185 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, | |
1186 | unsigned long next, struct mm_walk *walk); | |
1187 | int (*pte_entry)(pte_t *pte, unsigned long addr, | |
1188 | unsigned long next, struct mm_walk *walk); | |
1189 | int (*pte_hole)(unsigned long addr, unsigned long next, | |
1190 | struct mm_walk *walk); | |
1191 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, | |
1192 | unsigned long addr, unsigned long next, | |
1193 | struct mm_walk *walk); | |
1194 | struct mm_struct *mm; | |
1195 | void *private; | |
1196 | }; | |
1197 | ||
1198 | int walk_page_range(unsigned long addr, unsigned long end, | |
1199 | struct mm_walk *walk); | |
1200 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, | |
1201 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
1202 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, | |
1203 | struct vm_area_struct *vma); | |
1204 | void unmap_mapping_range(struct address_space *mapping, | |
1205 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
1206 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, | |
1207 | unsigned long *pfn); | |
1208 | int follow_phys(struct vm_area_struct *vma, unsigned long address, | |
1209 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
1210 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, | |
1211 | void *buf, int len, int write); | |
1212 | ||
1213 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
1214 | loff_t const holebegin, loff_t const holelen) | |
1215 | { | |
1216 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
1217 | } | |
1218 | ||
1219 | extern void truncate_pagecache(struct inode *inode, loff_t new); | |
1220 | extern void truncate_setsize(struct inode *inode, loff_t newsize); | |
1221 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); | |
1222 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); | |
1223 | int truncate_inode_page(struct address_space *mapping, struct page *page); | |
1224 | int generic_error_remove_page(struct address_space *mapping, struct page *page); | |
1225 | int invalidate_inode_page(struct page *page); | |
1226 | ||
1227 | #ifdef CONFIG_MMU | |
1228 | extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, | |
1229 | unsigned long address, unsigned int flags); | |
1230 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, | |
1231 | unsigned long address, unsigned int fault_flags); | |
1232 | #else | |
1233 | static inline int handle_mm_fault(struct mm_struct *mm, | |
1234 | struct vm_area_struct *vma, unsigned long address, | |
1235 | unsigned int flags) | |
1236 | { | |
1237 | /* should never happen if there's no MMU */ | |
1238 | BUG(); | |
1239 | return VM_FAULT_SIGBUS; | |
1240 | } | |
1241 | static inline int fixup_user_fault(struct task_struct *tsk, | |
1242 | struct mm_struct *mm, unsigned long address, | |
1243 | unsigned int fault_flags) | |
1244 | { | |
1245 | /* should never happen if there's no MMU */ | |
1246 | BUG(); | |
1247 | return -EFAULT; | |
1248 | } | |
1249 | #endif | |
1250 | ||
1251 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); | |
1252 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, | |
1253 | void *buf, int len, int write); | |
1254 | ||
1255 | long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |
1256 | unsigned long start, unsigned long nr_pages, | |
1257 | unsigned int foll_flags, struct page **pages, | |
1258 | struct vm_area_struct **vmas, int *nonblocking); | |
1259 | long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |
1260 | unsigned long start, unsigned long nr_pages, | |
1261 | int write, int force, struct page **pages, | |
1262 | struct vm_area_struct **vmas); | |
1263 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1264 | struct page **pages); | |
1265 | struct kvec; | |
1266 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | |
1267 | struct page **pages); | |
1268 | int get_kernel_page(unsigned long start, int write, struct page **pages); | |
1269 | struct page *get_dump_page(unsigned long addr); | |
1270 | ||
1271 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); | |
1272 | extern void do_invalidatepage(struct page *page, unsigned int offset, | |
1273 | unsigned int length); | |
1274 | ||
1275 | int __set_page_dirty_nobuffers(struct page *page); | |
1276 | int __set_page_dirty_no_writeback(struct page *page); | |
1277 | int redirty_page_for_writepage(struct writeback_control *wbc, | |
1278 | struct page *page); | |
1279 | void account_page_dirtied(struct page *page, struct address_space *mapping); | |
1280 | int set_page_dirty(struct page *page); | |
1281 | int set_page_dirty_lock(struct page *page); | |
1282 | int clear_page_dirty_for_io(struct page *page); | |
1283 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); | |
1284 | ||
1285 | /* Is the vma a continuation of the stack vma above it? */ | |
1286 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) | |
1287 | { | |
1288 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); | |
1289 | } | |
1290 | ||
1291 | static inline int stack_guard_page_start(struct vm_area_struct *vma, | |
1292 | unsigned long addr) | |
1293 | { | |
1294 | return (vma->vm_flags & VM_GROWSDOWN) && | |
1295 | (vma->vm_start == addr) && | |
1296 | !vma_growsdown(vma->vm_prev, addr); | |
1297 | } | |
1298 | ||
1299 | /* Is the vma a continuation of the stack vma below it? */ | |
1300 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) | |
1301 | { | |
1302 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); | |
1303 | } | |
1304 | ||
1305 | static inline int stack_guard_page_end(struct vm_area_struct *vma, | |
1306 | unsigned long addr) | |
1307 | { | |
1308 | return (vma->vm_flags & VM_GROWSUP) && | |
1309 | (vma->vm_end == addr) && | |
1310 | !vma_growsup(vma->vm_next, addr); | |
1311 | } | |
1312 | ||
1313 | extern struct task_struct *task_of_stack(struct task_struct *task, | |
1314 | struct vm_area_struct *vma, bool in_group); | |
1315 | ||
1316 | extern unsigned long move_page_tables(struct vm_area_struct *vma, | |
1317 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
1318 | unsigned long new_addr, unsigned long len, | |
1319 | bool need_rmap_locks); | |
1320 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, | |
1321 | unsigned long end, pgprot_t newprot, | |
1322 | int dirty_accountable, int prot_numa); | |
1323 | extern int mprotect_fixup(struct vm_area_struct *vma, | |
1324 | struct vm_area_struct **pprev, unsigned long start, | |
1325 | unsigned long end, unsigned long newflags); | |
1326 | ||
1327 | /* | |
1328 | * doesn't attempt to fault and will return short. | |
1329 | */ | |
1330 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1331 | struct page **pages); | |
1332 | /* | |
1333 | * per-process(per-mm_struct) statistics. | |
1334 | */ | |
1335 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) | |
1336 | { | |
1337 | long val = atomic_long_read(&mm->rss_stat.count[member]); | |
1338 | ||
1339 | #ifdef SPLIT_RSS_COUNTING | |
1340 | /* | |
1341 | * counter is updated in asynchronous manner and may go to minus. | |
1342 | * But it's never be expected number for users. | |
1343 | */ | |
1344 | if (val < 0) | |
1345 | val = 0; | |
1346 | #endif | |
1347 | return (unsigned long)val; | |
1348 | } | |
1349 | ||
1350 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1351 | { | |
1352 | atomic_long_add(value, &mm->rss_stat.count[member]); | |
1353 | } | |
1354 | ||
1355 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1356 | { | |
1357 | atomic_long_inc(&mm->rss_stat.count[member]); | |
1358 | } | |
1359 | ||
1360 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1361 | { | |
1362 | atomic_long_dec(&mm->rss_stat.count[member]); | |
1363 | } | |
1364 | ||
1365 | static inline unsigned long get_mm_rss(struct mm_struct *mm) | |
1366 | { | |
1367 | return get_mm_counter(mm, MM_FILEPAGES) + | |
1368 | get_mm_counter(mm, MM_ANONPAGES); | |
1369 | } | |
1370 | ||
1371 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1372 | { | |
1373 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1374 | } | |
1375 | ||
1376 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1377 | { | |
1378 | return max(mm->hiwater_vm, mm->total_vm); | |
1379 | } | |
1380 | ||
1381 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1382 | { | |
1383 | unsigned long _rss = get_mm_rss(mm); | |
1384 | ||
1385 | if ((mm)->hiwater_rss < _rss) | |
1386 | (mm)->hiwater_rss = _rss; | |
1387 | } | |
1388 | ||
1389 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1390 | { | |
1391 | if (mm->hiwater_vm < mm->total_vm) | |
1392 | mm->hiwater_vm = mm->total_vm; | |
1393 | } | |
1394 | ||
1395 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, | |
1396 | struct mm_struct *mm) | |
1397 | { | |
1398 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
1399 | ||
1400 | if (*maxrss < hiwater_rss) | |
1401 | *maxrss = hiwater_rss; | |
1402 | } | |
1403 | ||
1404 | #if defined(SPLIT_RSS_COUNTING) | |
1405 | void sync_mm_rss(struct mm_struct *mm); | |
1406 | #else | |
1407 | static inline void sync_mm_rss(struct mm_struct *mm) | |
1408 | { | |
1409 | } | |
1410 | #endif | |
1411 | ||
1412 | int vma_wants_writenotify(struct vm_area_struct *vma); | |
1413 | ||
1414 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1415 | spinlock_t **ptl); | |
1416 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1417 | spinlock_t **ptl) | |
1418 | { | |
1419 | pte_t *ptep; | |
1420 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
1421 | return ptep; | |
1422 | } | |
1423 | ||
1424 | #ifdef __PAGETABLE_PUD_FOLDED | |
1425 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
1426 | unsigned long address) | |
1427 | { | |
1428 | return 0; | |
1429 | } | |
1430 | #else | |
1431 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
1432 | #endif | |
1433 | ||
1434 | #ifdef __PAGETABLE_PMD_FOLDED | |
1435 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
1436 | unsigned long address) | |
1437 | { | |
1438 | return 0; | |
1439 | } | |
1440 | #else | |
1441 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); | |
1442 | #endif | |
1443 | ||
1444 | int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, | |
1445 | pmd_t *pmd, unsigned long address); | |
1446 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | |
1447 | ||
1448 | /* | |
1449 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
1450 | * Remove it when 4level-fixup.h has been removed. | |
1451 | */ | |
1452 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) | |
1453 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) | |
1454 | { | |
1455 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? | |
1456 | NULL: pud_offset(pgd, address); | |
1457 | } | |
1458 | ||
1459 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
1460 | { | |
1461 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? | |
1462 | NULL: pmd_offset(pud, address); | |
1463 | } | |
1464 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ | |
1465 | ||
1466 | #if USE_SPLIT_PTE_PTLOCKS | |
1467 | #if ALLOC_SPLIT_PTLOCKS | |
1468 | void __init ptlock_cache_init(void); | |
1469 | extern bool ptlock_alloc(struct page *page); | |
1470 | extern void ptlock_free(struct page *page); | |
1471 | ||
1472 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1473 | { | |
1474 | return page->ptl; | |
1475 | } | |
1476 | #else /* ALLOC_SPLIT_PTLOCKS */ | |
1477 | static inline void ptlock_cache_init(void) | |
1478 | { | |
1479 | } | |
1480 | ||
1481 | static inline bool ptlock_alloc(struct page *page) | |
1482 | { | |
1483 | return true; | |
1484 | } | |
1485 | ||
1486 | static inline void ptlock_free(struct page *page) | |
1487 | { | |
1488 | } | |
1489 | ||
1490 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1491 | { | |
1492 | return &page->ptl; | |
1493 | } | |
1494 | #endif /* ALLOC_SPLIT_PTLOCKS */ | |
1495 | ||
1496 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1497 | { | |
1498 | return ptlock_ptr(pmd_page(*pmd)); | |
1499 | } | |
1500 | ||
1501 | static inline bool ptlock_init(struct page *page) | |
1502 | { | |
1503 | /* | |
1504 | * prep_new_page() initialize page->private (and therefore page->ptl) | |
1505 | * with 0. Make sure nobody took it in use in between. | |
1506 | * | |
1507 | * It can happen if arch try to use slab for page table allocation: | |
1508 | * slab code uses page->slab_cache and page->first_page (for tail | |
1509 | * pages), which share storage with page->ptl. | |
1510 | */ | |
1511 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); | |
1512 | if (!ptlock_alloc(page)) | |
1513 | return false; | |
1514 | spin_lock_init(ptlock_ptr(page)); | |
1515 | return true; | |
1516 | } | |
1517 | ||
1518 | /* Reset page->mapping so free_pages_check won't complain. */ | |
1519 | static inline void pte_lock_deinit(struct page *page) | |
1520 | { | |
1521 | page->mapping = NULL; | |
1522 | ptlock_free(page); | |
1523 | } | |
1524 | ||
1525 | #else /* !USE_SPLIT_PTE_PTLOCKS */ | |
1526 | /* | |
1527 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
1528 | */ | |
1529 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1530 | { | |
1531 | return &mm->page_table_lock; | |
1532 | } | |
1533 | static inline void ptlock_cache_init(void) {} | |
1534 | static inline bool ptlock_init(struct page *page) { return true; } | |
1535 | static inline void pte_lock_deinit(struct page *page) {} | |
1536 | #endif /* USE_SPLIT_PTE_PTLOCKS */ | |
1537 | ||
1538 | static inline void pgtable_init(void) | |
1539 | { | |
1540 | ptlock_cache_init(); | |
1541 | pgtable_cache_init(); | |
1542 | } | |
1543 | ||
1544 | static inline bool pgtable_page_ctor(struct page *page) | |
1545 | { | |
1546 | inc_zone_page_state(page, NR_PAGETABLE); | |
1547 | return ptlock_init(page); | |
1548 | } | |
1549 | ||
1550 | static inline void pgtable_page_dtor(struct page *page) | |
1551 | { | |
1552 | pte_lock_deinit(page); | |
1553 | dec_zone_page_state(page, NR_PAGETABLE); | |
1554 | } | |
1555 | ||
1556 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ | |
1557 | ({ \ | |
1558 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ | |
1559 | pte_t *__pte = pte_offset_map(pmd, address); \ | |
1560 | *(ptlp) = __ptl; \ | |
1561 | spin_lock(__ptl); \ | |
1562 | __pte; \ | |
1563 | }) | |
1564 | ||
1565 | #define pte_unmap_unlock(pte, ptl) do { \ | |
1566 | spin_unlock(ptl); \ | |
1567 | pte_unmap(pte); \ | |
1568 | } while (0) | |
1569 | ||
1570 | #define pte_alloc_map(mm, vma, pmd, address) \ | |
1571 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ | |
1572 | pmd, address))? \ | |
1573 | NULL: pte_offset_map(pmd, address)) | |
1574 | ||
1575 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ | |
1576 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ | |
1577 | pmd, address))? \ | |
1578 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) | |
1579 | ||
1580 | #define pte_alloc_kernel(pmd, address) \ | |
1581 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | |
1582 | NULL: pte_offset_kernel(pmd, address)) | |
1583 | ||
1584 | #if USE_SPLIT_PMD_PTLOCKS | |
1585 | ||
1586 | static struct page *pmd_to_page(pmd_t *pmd) | |
1587 | { | |
1588 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | |
1589 | return virt_to_page((void *)((unsigned long) pmd & mask)); | |
1590 | } | |
1591 | ||
1592 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1593 | { | |
1594 | return ptlock_ptr(pmd_to_page(pmd)); | |
1595 | } | |
1596 | ||
1597 | static inline bool pgtable_pmd_page_ctor(struct page *page) | |
1598 | { | |
1599 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
1600 | page->pmd_huge_pte = NULL; | |
1601 | #endif | |
1602 | return ptlock_init(page); | |
1603 | } | |
1604 | ||
1605 | static inline void pgtable_pmd_page_dtor(struct page *page) | |
1606 | { | |
1607 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
1608 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); | |
1609 | #endif | |
1610 | ptlock_free(page); | |
1611 | } | |
1612 | ||
1613 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) | |
1614 | ||
1615 | #else | |
1616 | ||
1617 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1618 | { | |
1619 | return &mm->page_table_lock; | |
1620 | } | |
1621 | ||
1622 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } | |
1623 | static inline void pgtable_pmd_page_dtor(struct page *page) {} | |
1624 | ||
1625 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) | |
1626 | ||
1627 | #endif | |
1628 | ||
1629 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) | |
1630 | { | |
1631 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | |
1632 | spin_lock(ptl); | |
1633 | return ptl; | |
1634 | } | |
1635 | ||
1636 | extern void free_area_init(unsigned long * zones_size); | |
1637 | extern void free_area_init_node(int nid, unsigned long * zones_size, | |
1638 | unsigned long zone_start_pfn, unsigned long *zholes_size); | |
1639 | extern void free_initmem(void); | |
1640 | ||
1641 | /* | |
1642 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
1643 | * into the buddy system. The freed pages will be poisoned with pattern | |
1644 | * "poison" if it's within range [0, UCHAR_MAX]. | |
1645 | * Return pages freed into the buddy system. | |
1646 | */ | |
1647 | extern unsigned long free_reserved_area(void *start, void *end, | |
1648 | int poison, char *s); | |
1649 | ||
1650 | #ifdef CONFIG_HIGHMEM | |
1651 | /* | |
1652 | * Free a highmem page into the buddy system, adjusting totalhigh_pages | |
1653 | * and totalram_pages. | |
1654 | */ | |
1655 | extern void free_highmem_page(struct page *page); | |
1656 | #endif | |
1657 | ||
1658 | extern void adjust_managed_page_count(struct page *page, long count); | |
1659 | extern void mem_init_print_info(const char *str); | |
1660 | ||
1661 | /* Free the reserved page into the buddy system, so it gets managed. */ | |
1662 | static inline void __free_reserved_page(struct page *page) | |
1663 | { | |
1664 | ClearPageReserved(page); | |
1665 | init_page_count(page); | |
1666 | __free_page(page); | |
1667 | } | |
1668 | ||
1669 | static inline void free_reserved_page(struct page *page) | |
1670 | { | |
1671 | __free_reserved_page(page); | |
1672 | adjust_managed_page_count(page, 1); | |
1673 | } | |
1674 | ||
1675 | static inline void mark_page_reserved(struct page *page) | |
1676 | { | |
1677 | SetPageReserved(page); | |
1678 | adjust_managed_page_count(page, -1); | |
1679 | } | |
1680 | ||
1681 | /* | |
1682 | * Default method to free all the __init memory into the buddy system. | |
1683 | * The freed pages will be poisoned with pattern "poison" if it's within | |
1684 | * range [0, UCHAR_MAX]. | |
1685 | * Return pages freed into the buddy system. | |
1686 | */ | |
1687 | static inline unsigned long free_initmem_default(int poison) | |
1688 | { | |
1689 | extern char __init_begin[], __init_end[]; | |
1690 | ||
1691 | return free_reserved_area(&__init_begin, &__init_end, | |
1692 | poison, "unused kernel"); | |
1693 | } | |
1694 | ||
1695 | static inline unsigned long get_num_physpages(void) | |
1696 | { | |
1697 | int nid; | |
1698 | unsigned long phys_pages = 0; | |
1699 | ||
1700 | for_each_online_node(nid) | |
1701 | phys_pages += node_present_pages(nid); | |
1702 | ||
1703 | return phys_pages; | |
1704 | } | |
1705 | ||
1706 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP | |
1707 | /* | |
1708 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its | |
1709 | * zones, allocate the backing mem_map and account for memory holes in a more | |
1710 | * architecture independent manner. This is a substitute for creating the | |
1711 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
1712 | * free_area_init_node() | |
1713 | * | |
1714 | * An architecture is expected to register range of page frames backed by | |
1715 | * physical memory with memblock_add[_node]() before calling | |
1716 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
1717 | * usage, an architecture is expected to do something like | |
1718 | * | |
1719 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
1720 | * max_highmem_pfn}; | |
1721 | * for_each_valid_physical_page_range() | |
1722 | * memblock_add_node(base, size, nid) | |
1723 | * free_area_init_nodes(max_zone_pfns); | |
1724 | * | |
1725 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each | |
1726 | * registered physical page range. Similarly | |
1727 | * sparse_memory_present_with_active_regions() calls memory_present() for | |
1728 | * each range when SPARSEMEM is enabled. | |
1729 | * | |
1730 | * See mm/page_alloc.c for more information on each function exposed by | |
1731 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. | |
1732 | */ | |
1733 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1734 | unsigned long node_map_pfn_alignment(void); | |
1735 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, | |
1736 | unsigned long end_pfn); | |
1737 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | |
1738 | unsigned long end_pfn); | |
1739 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1740 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1741 | extern unsigned long find_min_pfn_with_active_regions(void); | |
1742 | extern void free_bootmem_with_active_regions(int nid, | |
1743 | unsigned long max_low_pfn); | |
1744 | extern void sparse_memory_present_with_active_regions(int nid); | |
1745 | ||
1746 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
1747 | ||
1748 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ | |
1749 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) | |
1750 | static inline int __early_pfn_to_nid(unsigned long pfn) | |
1751 | { | |
1752 | return 0; | |
1753 | } | |
1754 | #else | |
1755 | /* please see mm/page_alloc.c */ | |
1756 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
1757 | /* there is a per-arch backend function. */ | |
1758 | extern int __meminit __early_pfn_to_nid(unsigned long pfn); | |
1759 | #endif | |
1760 | ||
1761 | extern void set_dma_reserve(unsigned long new_dma_reserve); | |
1762 | extern void memmap_init_zone(unsigned long, int, unsigned long, | |
1763 | unsigned long, enum memmap_context); | |
1764 | extern void setup_per_zone_wmarks(void); | |
1765 | extern int __meminit init_per_zone_wmark_min(void); | |
1766 | extern void mem_init(void); | |
1767 | extern void __init mmap_init(void); | |
1768 | extern void show_mem(unsigned int flags); | |
1769 | extern void si_meminfo(struct sysinfo * val); | |
1770 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
1771 | ||
1772 | extern __printf(3, 4) | |
1773 | void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); | |
1774 | ||
1775 | extern void setup_per_cpu_pageset(void); | |
1776 | ||
1777 | extern void zone_pcp_update(struct zone *zone); | |
1778 | extern void zone_pcp_reset(struct zone *zone); | |
1779 | ||
1780 | /* page_alloc.c */ | |
1781 | extern int min_free_kbytes; | |
1782 | ||
1783 | /* nommu.c */ | |
1784 | extern atomic_long_t mmap_pages_allocated; | |
1785 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); | |
1786 | ||
1787 | /* interval_tree.c */ | |
1788 | void vma_interval_tree_insert(struct vm_area_struct *node, | |
1789 | struct rb_root *root); | |
1790 | void vma_interval_tree_insert_after(struct vm_area_struct *node, | |
1791 | struct vm_area_struct *prev, | |
1792 | struct rb_root *root); | |
1793 | void vma_interval_tree_remove(struct vm_area_struct *node, | |
1794 | struct rb_root *root); | |
1795 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, | |
1796 | unsigned long start, unsigned long last); | |
1797 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
1798 | unsigned long start, unsigned long last); | |
1799 | ||
1800 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
1801 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
1802 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1803 | ||
1804 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, | |
1805 | struct list_head *list) | |
1806 | { | |
1807 | list_add_tail(&vma->shared.nonlinear, list); | |
1808 | } | |
1809 | ||
1810 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, | |
1811 | struct rb_root *root); | |
1812 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, | |
1813 | struct rb_root *root); | |
1814 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( | |
1815 | struct rb_root *root, unsigned long start, unsigned long last); | |
1816 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( | |
1817 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
1818 | #ifdef CONFIG_DEBUG_VM_RB | |
1819 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
1820 | #endif | |
1821 | ||
1822 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
1823 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
1824 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
1825 | ||
1826 | /* mmap.c */ | |
1827 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); | |
1828 | extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1829 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); | |
1830 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1831 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1832 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1833 | struct mempolicy *); | |
1834 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1835 | extern int split_vma(struct mm_struct *, | |
1836 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1837 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1838 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1839 | struct rb_node **, struct rb_node *); | |
1840 | extern void unlink_file_vma(struct vm_area_struct *); | |
1841 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, | |
1842 | unsigned long addr, unsigned long len, pgoff_t pgoff, | |
1843 | bool *need_rmap_locks); | |
1844 | extern void exit_mmap(struct mm_struct *); | |
1845 | ||
1846 | static inline int check_data_rlimit(unsigned long rlim, | |
1847 | unsigned long new, | |
1848 | unsigned long start, | |
1849 | unsigned long end_data, | |
1850 | unsigned long start_data) | |
1851 | { | |
1852 | if (rlim < RLIM_INFINITY) { | |
1853 | if (((new - start) + (end_data - start_data)) > rlim) | |
1854 | return -ENOSPC; | |
1855 | } | |
1856 | ||
1857 | return 0; | |
1858 | } | |
1859 | ||
1860 | extern int mm_take_all_locks(struct mm_struct *mm); | |
1861 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
1862 | ||
1863 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); | |
1864 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | |
1865 | ||
1866 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); | |
1867 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, | |
1868 | unsigned long addr, unsigned long len, | |
1869 | unsigned long flags, | |
1870 | const struct vm_special_mapping *spec); | |
1871 | /* This is an obsolete alternative to _install_special_mapping. */ | |
1872 | extern int install_special_mapping(struct mm_struct *mm, | |
1873 | unsigned long addr, unsigned long len, | |
1874 | unsigned long flags, struct page **pages); | |
1875 | ||
1876 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
1877 | ||
1878 | extern unsigned long mmap_region(struct file *file, unsigned long addr, | |
1879 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); | |
1880 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, | |
1881 | unsigned long len, unsigned long prot, unsigned long flags, | |
1882 | unsigned long pgoff, unsigned long *populate); | |
1883 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
1884 | ||
1885 | #ifdef CONFIG_MMU | |
1886 | extern int __mm_populate(unsigned long addr, unsigned long len, | |
1887 | int ignore_errors); | |
1888 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
1889 | { | |
1890 | /* Ignore errors */ | |
1891 | (void) __mm_populate(addr, len, 1); | |
1892 | } | |
1893 | #else | |
1894 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
1895 | #endif | |
1896 | ||
1897 | /* These take the mm semaphore themselves */ | |
1898 | extern unsigned long vm_brk(unsigned long, unsigned long); | |
1899 | extern int vm_munmap(unsigned long, size_t); | |
1900 | extern unsigned long vm_mmap(struct file *, unsigned long, | |
1901 | unsigned long, unsigned long, | |
1902 | unsigned long, unsigned long); | |
1903 | ||
1904 | struct vm_unmapped_area_info { | |
1905 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
1906 | unsigned long flags; | |
1907 | unsigned long length; | |
1908 | unsigned long low_limit; | |
1909 | unsigned long high_limit; | |
1910 | unsigned long align_mask; | |
1911 | unsigned long align_offset; | |
1912 | }; | |
1913 | ||
1914 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); | |
1915 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); | |
1916 | ||
1917 | /* | |
1918 | * Search for an unmapped address range. | |
1919 | * | |
1920 | * We are looking for a range that: | |
1921 | * - does not intersect with any VMA; | |
1922 | * - is contained within the [low_limit, high_limit) interval; | |
1923 | * - is at least the desired size. | |
1924 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) | |
1925 | */ | |
1926 | static inline unsigned long | |
1927 | vm_unmapped_area(struct vm_unmapped_area_info *info) | |
1928 | { | |
1929 | if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN)) | |
1930 | return unmapped_area(info); | |
1931 | else | |
1932 | return unmapped_area_topdown(info); | |
1933 | } | |
1934 | ||
1935 | /* truncate.c */ | |
1936 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
1937 | extern void truncate_inode_pages_range(struct address_space *, | |
1938 | loff_t lstart, loff_t lend); | |
1939 | extern void truncate_inode_pages_final(struct address_space *); | |
1940 | ||
1941 | /* generic vm_area_ops exported for stackable file systems */ | |
1942 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); | |
1943 | extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf); | |
1944 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); | |
1945 | ||
1946 | /* mm/page-writeback.c */ | |
1947 | int write_one_page(struct page *page, int wait); | |
1948 | void task_dirty_inc(struct task_struct *tsk); | |
1949 | ||
1950 | /* readahead.c */ | |
1951 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
1952 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
1953 | ||
1954 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
1955 | pgoff_t offset, unsigned long nr_to_read); | |
1956 | ||
1957 | void page_cache_sync_readahead(struct address_space *mapping, | |
1958 | struct file_ra_state *ra, | |
1959 | struct file *filp, | |
1960 | pgoff_t offset, | |
1961 | unsigned long size); | |
1962 | ||
1963 | void page_cache_async_readahead(struct address_space *mapping, | |
1964 | struct file_ra_state *ra, | |
1965 | struct file *filp, | |
1966 | struct page *pg, | |
1967 | pgoff_t offset, | |
1968 | unsigned long size); | |
1969 | ||
1970 | unsigned long max_sane_readahead(unsigned long nr); | |
1971 | ||
1972 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ | |
1973 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); | |
1974 | ||
1975 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | |
1976 | extern int expand_downwards(struct vm_area_struct *vma, | |
1977 | unsigned long address); | |
1978 | #if VM_GROWSUP | |
1979 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); | |
1980 | #else | |
1981 | #define expand_upwards(vma, address) (0) | |
1982 | #endif | |
1983 | ||
1984 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
1985 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
1986 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
1987 | struct vm_area_struct **pprev); | |
1988 | ||
1989 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
1990 | NULL if none. Assume start_addr < end_addr. */ | |
1991 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
1992 | { | |
1993 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
1994 | ||
1995 | if (vma && end_addr <= vma->vm_start) | |
1996 | vma = NULL; | |
1997 | return vma; | |
1998 | } | |
1999 | ||
2000 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
2001 | { | |
2002 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
2003 | } | |
2004 | ||
2005 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ | |
2006 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
2007 | unsigned long vm_start, unsigned long vm_end) | |
2008 | { | |
2009 | struct vm_area_struct *vma = find_vma(mm, vm_start); | |
2010 | ||
2011 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
2012 | vma = NULL; | |
2013 | ||
2014 | return vma; | |
2015 | } | |
2016 | ||
2017 | #ifdef CONFIG_MMU | |
2018 | pgprot_t vm_get_page_prot(unsigned long vm_flags); | |
2019 | void vma_set_page_prot(struct vm_area_struct *vma); | |
2020 | #else | |
2021 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
2022 | { | |
2023 | return __pgprot(0); | |
2024 | } | |
2025 | static inline void vma_set_page_prot(struct vm_area_struct *vma) | |
2026 | { | |
2027 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); | |
2028 | } | |
2029 | #endif | |
2030 | ||
2031 | #ifdef CONFIG_NUMA_BALANCING | |
2032 | unsigned long change_prot_numa(struct vm_area_struct *vma, | |
2033 | unsigned long start, unsigned long end); | |
2034 | #endif | |
2035 | ||
2036 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); | |
2037 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | |
2038 | unsigned long pfn, unsigned long size, pgprot_t); | |
2039 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); | |
2040 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, | |
2041 | unsigned long pfn); | |
2042 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, | |
2043 | unsigned long pfn); | |
2044 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); | |
2045 | ||
2046 | ||
2047 | struct page *follow_page_mask(struct vm_area_struct *vma, | |
2048 | unsigned long address, unsigned int foll_flags, | |
2049 | unsigned int *page_mask); | |
2050 | ||
2051 | static inline struct page *follow_page(struct vm_area_struct *vma, | |
2052 | unsigned long address, unsigned int foll_flags) | |
2053 | { | |
2054 | unsigned int unused_page_mask; | |
2055 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); | |
2056 | } | |
2057 | ||
2058 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
2059 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
2060 | #define FOLL_GET 0x04 /* do get_page on page */ | |
2061 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ | |
2062 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ | |
2063 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO | |
2064 | * and return without waiting upon it */ | |
2065 | #define FOLL_MLOCK 0x40 /* mark page as mlocked */ | |
2066 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ | |
2067 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ | |
2068 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ | |
2069 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ | |
2070 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ | |
2071 | ||
2072 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, | |
2073 | void *data); | |
2074 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
2075 | unsigned long size, pte_fn_t fn, void *data); | |
2076 | ||
2077 | #ifdef CONFIG_PROC_FS | |
2078 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); | |
2079 | #else | |
2080 | static inline void vm_stat_account(struct mm_struct *mm, | |
2081 | unsigned long flags, struct file *file, long pages) | |
2082 | { | |
2083 | mm->total_vm += pages; | |
2084 | } | |
2085 | #endif /* CONFIG_PROC_FS */ | |
2086 | ||
2087 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
2088 | extern bool _debug_pagealloc_enabled; | |
2089 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); | |
2090 | ||
2091 | static inline bool debug_pagealloc_enabled(void) | |
2092 | { | |
2093 | return _debug_pagealloc_enabled; | |
2094 | } | |
2095 | ||
2096 | static inline void | |
2097 | kernel_map_pages(struct page *page, int numpages, int enable) | |
2098 | { | |
2099 | if (!debug_pagealloc_enabled()) | |
2100 | return; | |
2101 | ||
2102 | __kernel_map_pages(page, numpages, enable); | |
2103 | } | |
2104 | #ifdef CONFIG_HIBERNATION | |
2105 | extern bool kernel_page_present(struct page *page); | |
2106 | #endif /* CONFIG_HIBERNATION */ | |
2107 | #else | |
2108 | static inline void | |
2109 | kernel_map_pages(struct page *page, int numpages, int enable) {} | |
2110 | #ifdef CONFIG_HIBERNATION | |
2111 | static inline bool kernel_page_present(struct page *page) { return true; } | |
2112 | #endif /* CONFIG_HIBERNATION */ | |
2113 | #endif | |
2114 | ||
2115 | #ifdef __HAVE_ARCH_GATE_AREA | |
2116 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); | |
2117 | extern int in_gate_area_no_mm(unsigned long addr); | |
2118 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
2119 | #else | |
2120 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) | |
2121 | { | |
2122 | return NULL; | |
2123 | } | |
2124 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } | |
2125 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
2126 | { | |
2127 | return 0; | |
2128 | } | |
2129 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
2130 | ||
2131 | #ifdef CONFIG_SYSCTL | |
2132 | extern int sysctl_drop_caches; | |
2133 | int drop_caches_sysctl_handler(struct ctl_table *, int, | |
2134 | void __user *, size_t *, loff_t *); | |
2135 | #endif | |
2136 | ||
2137 | unsigned long shrink_node_slabs(gfp_t gfp_mask, int nid, | |
2138 | unsigned long nr_scanned, | |
2139 | unsigned long nr_eligible); | |
2140 | ||
2141 | #ifndef CONFIG_MMU | |
2142 | #define randomize_va_space 0 | |
2143 | #else | |
2144 | extern int randomize_va_space; | |
2145 | #endif | |
2146 | ||
2147 | const char * arch_vma_name(struct vm_area_struct *vma); | |
2148 | void print_vma_addr(char *prefix, unsigned long rip); | |
2149 | ||
2150 | void sparse_mem_maps_populate_node(struct page **map_map, | |
2151 | unsigned long pnum_begin, | |
2152 | unsigned long pnum_end, | |
2153 | unsigned long map_count, | |
2154 | int nodeid); | |
2155 | ||
2156 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); | |
2157 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); | |
2158 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); | |
2159 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); | |
2160 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | |
2161 | void *vmemmap_alloc_block(unsigned long size, int node); | |
2162 | void *vmemmap_alloc_block_buf(unsigned long size, int node); | |
2163 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); | |
2164 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, | |
2165 | int node); | |
2166 | int vmemmap_populate(unsigned long start, unsigned long end, int node); | |
2167 | void vmemmap_populate_print_last(void); | |
2168 | #ifdef CONFIG_MEMORY_HOTPLUG | |
2169 | void vmemmap_free(unsigned long start, unsigned long end); | |
2170 | #endif | |
2171 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, | |
2172 | unsigned long size); | |
2173 | ||
2174 | enum mf_flags { | |
2175 | MF_COUNT_INCREASED = 1 << 0, | |
2176 | MF_ACTION_REQUIRED = 1 << 1, | |
2177 | MF_MUST_KILL = 1 << 2, | |
2178 | MF_SOFT_OFFLINE = 1 << 3, | |
2179 | }; | |
2180 | extern int memory_failure(unsigned long pfn, int trapno, int flags); | |
2181 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); | |
2182 | extern int unpoison_memory(unsigned long pfn); | |
2183 | extern int sysctl_memory_failure_early_kill; | |
2184 | extern int sysctl_memory_failure_recovery; | |
2185 | extern void shake_page(struct page *p, int access); | |
2186 | extern atomic_long_t num_poisoned_pages; | |
2187 | extern int soft_offline_page(struct page *page, int flags); | |
2188 | ||
2189 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) | |
2190 | extern void clear_huge_page(struct page *page, | |
2191 | unsigned long addr, | |
2192 | unsigned int pages_per_huge_page); | |
2193 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
2194 | unsigned long addr, struct vm_area_struct *vma, | |
2195 | unsigned int pages_per_huge_page); | |
2196 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ | |
2197 | ||
2198 | extern struct page_ext_operations debug_guardpage_ops; | |
2199 | extern struct page_ext_operations page_poisoning_ops; | |
2200 | ||
2201 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
2202 | extern unsigned int _debug_guardpage_minorder; | |
2203 | extern bool _debug_guardpage_enabled; | |
2204 | ||
2205 | static inline unsigned int debug_guardpage_minorder(void) | |
2206 | { | |
2207 | return _debug_guardpage_minorder; | |
2208 | } | |
2209 | ||
2210 | static inline bool debug_guardpage_enabled(void) | |
2211 | { | |
2212 | return _debug_guardpage_enabled; | |
2213 | } | |
2214 | ||
2215 | static inline bool page_is_guard(struct page *page) | |
2216 | { | |
2217 | struct page_ext *page_ext; | |
2218 | ||
2219 | if (!debug_guardpage_enabled()) | |
2220 | return false; | |
2221 | ||
2222 | page_ext = lookup_page_ext(page); | |
2223 | return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags); | |
2224 | } | |
2225 | #else | |
2226 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
2227 | static inline bool debug_guardpage_enabled(void) { return false; } | |
2228 | static inline bool page_is_guard(struct page *page) { return false; } | |
2229 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
2230 | ||
2231 | #if MAX_NUMNODES > 1 | |
2232 | void __init setup_nr_node_ids(void); | |
2233 | #else | |
2234 | static inline void setup_nr_node_ids(void) {} | |
2235 | #endif | |
2236 | ||
2237 | #endif /* __KERNEL__ */ | |
2238 | #endif /* _LINUX_MM_H */ |