~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/linux/mm.h

Version: ~ [ linux-5.15-rc1 ] ~ [ linux-5.14.5 ] ~ [ linux-5.13.18 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.66 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.147 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.206 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.246 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.282 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.283 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 
 22 struct mempolicy;
 23 struct anon_vma;
 24 struct anon_vma_chain;
 25 struct file_ra_state;
 26 struct user_struct;
 27 struct writeback_control;
 28 
 29 #ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
 30 extern unsigned long max_mapnr;
 31 
 32 static inline void set_max_mapnr(unsigned long limit)
 33 {
 34         max_mapnr = limit;
 35 }
 36 #else
 37 static inline void set_max_mapnr(unsigned long limit) { }
 38 #endif
 39 
 40 extern unsigned long totalram_pages;
 41 extern void * high_memory;
 42 extern int page_cluster;
 43 
 44 #ifdef CONFIG_SYSCTL
 45 extern int sysctl_legacy_va_layout;
 46 #else
 47 #define sysctl_legacy_va_layout 0
 48 #endif
 49 
 50 #include <asm/page.h>
 51 #include <asm/pgtable.h>
 52 #include <asm/processor.h>
 53 
 54 #ifndef __pa_symbol
 55 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 56 #endif
 57 
 58 extern unsigned long sysctl_user_reserve_kbytes;
 59 extern unsigned long sysctl_admin_reserve_kbytes;
 60 
 61 extern int sysctl_overcommit_memory;
 62 extern int sysctl_overcommit_ratio;
 63 extern unsigned long sysctl_overcommit_kbytes;
 64 
 65 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 66                                     size_t *, loff_t *);
 67 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 68                                     size_t *, loff_t *);
 69 
 70 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 71 
 72 /* to align the pointer to the (next) page boundary */
 73 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 74 
 75 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 76 #define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
 77 
 78 /*
 79  * Linux kernel virtual memory manager primitives.
 80  * The idea being to have a "virtual" mm in the same way
 81  * we have a virtual fs - giving a cleaner interface to the
 82  * mm details, and allowing different kinds of memory mappings
 83  * (from shared memory to executable loading to arbitrary
 84  * mmap() functions).
 85  */
 86 
 87 extern struct kmem_cache *vm_area_cachep;
 88 
 89 #ifndef CONFIG_MMU
 90 extern struct rb_root nommu_region_tree;
 91 extern struct rw_semaphore nommu_region_sem;
 92 
 93 extern unsigned int kobjsize(const void *objp);
 94 #endif
 95 
 96 /*
 97  * vm_flags in vm_area_struct, see mm_types.h.
 98  */
 99 #define VM_NONE         0x00000000
100 
101 #define VM_READ         0x00000001      /* currently active flags */
102 #define VM_WRITE        0x00000002
103 #define VM_EXEC         0x00000004
104 #define VM_SHARED       0x00000008
105 
106 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
107 #define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
108 #define VM_MAYWRITE     0x00000020
109 #define VM_MAYEXEC      0x00000040
110 #define VM_MAYSHARE     0x00000080
111 
112 #define VM_GROWSDOWN    0x00000100      /* general info on the segment */
113 #define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
114 #define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
115 
116 #define VM_LOCKED       0x00002000
117 #define VM_IO           0x00004000      /* Memory mapped I/O or similar */
118 
119                                         /* Used by sys_madvise() */
120 #define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
121 #define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
122 
123 #define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
124 #define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
125 #define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
126 #define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
127 #define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
128 #define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
129 #define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
130 
131 #ifdef CONFIG_MEM_SOFT_DIRTY
132 # define VM_SOFTDIRTY   0x08000000      /* Not soft dirty clean area */
133 #else
134 # define VM_SOFTDIRTY   0
135 #endif
136 
137 #define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
138 #define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
139 #define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
140 #define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
141 
142 #if defined(CONFIG_X86)
143 # define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
144 #elif defined(CONFIG_PPC)
145 # define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
146 #elif defined(CONFIG_PARISC)
147 # define VM_GROWSUP     VM_ARCH_1
148 #elif defined(CONFIG_METAG)
149 # define VM_GROWSUP     VM_ARCH_1
150 #elif defined(CONFIG_IA64)
151 # define VM_GROWSUP     VM_ARCH_1
152 #elif !defined(CONFIG_MMU)
153 # define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
154 #endif
155 
156 #ifndef VM_GROWSUP
157 # define VM_GROWSUP     VM_NONE
158 #endif
159 
160 /* Bits set in the VMA until the stack is in its final location */
161 #define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
162 
163 #ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
164 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
165 #endif
166 
167 #ifdef CONFIG_STACK_GROWSUP
168 #define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
169 #else
170 #define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
171 #endif
172 
173 /*
174  * Special vmas that are non-mergable, non-mlock()able.
175  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
176  */
177 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
178 
179 /* This mask defines which mm->def_flags a process can inherit its parent */
180 #define VM_INIT_DEF_MASK        VM_NOHUGEPAGE
181 
182 /*
183  * mapping from the currently active vm_flags protection bits (the
184  * low four bits) to a page protection mask..
185  */
186 extern pgprot_t protection_map[16];
187 
188 #define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
189 #define FAULT_FLAG_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
190 #define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
191 #define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
192 #define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
193 #define FAULT_FLAG_TRIED        0x20    /* Second try */
194 #define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
195 
196 /*
197  * vm_fault is filled by the the pagefault handler and passed to the vma's
198  * ->fault function. The vma's ->fault is responsible for returning a bitmask
199  * of VM_FAULT_xxx flags that give details about how the fault was handled.
200  *
201  * pgoff should be used in favour of virtual_address, if possible.
202  */
203 struct vm_fault {
204         unsigned int flags;             /* FAULT_FLAG_xxx flags */
205         pgoff_t pgoff;                  /* Logical page offset based on vma */
206         void __user *virtual_address;   /* Faulting virtual address */
207 
208         struct page *page;              /* ->fault handlers should return a
209                                          * page here, unless VM_FAULT_NOPAGE
210                                          * is set (which is also implied by
211                                          * VM_FAULT_ERROR).
212                                          */
213         /* for ->map_pages() only */
214         pgoff_t max_pgoff;              /* map pages for offset from pgoff till
215                                          * max_pgoff inclusive */
216         pte_t *pte;                     /* pte entry associated with ->pgoff */
217 };
218 
219 /*
220  * These are the virtual MM functions - opening of an area, closing and
221  * unmapping it (needed to keep files on disk up-to-date etc), pointer
222  * to the functions called when a no-page or a wp-page exception occurs. 
223  */
224 struct vm_operations_struct {
225         void (*open)(struct vm_area_struct * area);
226         void (*close)(struct vm_area_struct * area);
227         int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
228         void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
229 
230         /* notification that a previously read-only page is about to become
231          * writable, if an error is returned it will cause a SIGBUS */
232         int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
233 
234         /* called by access_process_vm when get_user_pages() fails, typically
235          * for use by special VMAs that can switch between memory and hardware
236          */
237         int (*access)(struct vm_area_struct *vma, unsigned long addr,
238                       void *buf, int len, int write);
239 
240         /* Called by the /proc/PID/maps code to ask the vma whether it
241          * has a special name.  Returning non-NULL will also cause this
242          * vma to be dumped unconditionally. */
243         const char *(*name)(struct vm_area_struct *vma);
244 
245 #ifdef CONFIG_NUMA
246         /*
247          * set_policy() op must add a reference to any non-NULL @new mempolicy
248          * to hold the policy upon return.  Caller should pass NULL @new to
249          * remove a policy and fall back to surrounding context--i.e. do not
250          * install a MPOL_DEFAULT policy, nor the task or system default
251          * mempolicy.
252          */
253         int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
254 
255         /*
256          * get_policy() op must add reference [mpol_get()] to any policy at
257          * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
258          * in mm/mempolicy.c will do this automatically.
259          * get_policy() must NOT add a ref if the policy at (vma,addr) is not
260          * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
261          * If no [shared/vma] mempolicy exists at the addr, get_policy() op
262          * must return NULL--i.e., do not "fallback" to task or system default
263          * policy.
264          */
265         struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
266                                         unsigned long addr);
267         int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
268                 const nodemask_t *to, unsigned long flags);
269 #endif
270 };
271 
272 struct mmu_gather;
273 struct inode;
274 
275 #define page_private(page)              ((page)->private)
276 #define set_page_private(page, v)       ((page)->private = (v))
277 
278 /* It's valid only if the page is free path or free_list */
279 static inline void set_freepage_migratetype(struct page *page, int migratetype)
280 {
281         page->index = migratetype;
282 }
283 
284 /* It's valid only if the page is free path or free_list */
285 static inline int get_freepage_migratetype(struct page *page)
286 {
287         return page->index;
288 }
289 
290 /*
291  * FIXME: take this include out, include page-flags.h in
292  * files which need it (119 of them)
293  */
294 #include <linux/page-flags.h>
295 #include <linux/huge_mm.h>
296 
297 /*
298  * Methods to modify the page usage count.
299  *
300  * What counts for a page usage:
301  * - cache mapping   (page->mapping)
302  * - private data    (page->private)
303  * - page mapped in a task's page tables, each mapping
304  *   is counted separately
305  *
306  * Also, many kernel routines increase the page count before a critical
307  * routine so they can be sure the page doesn't go away from under them.
308  */
309 
310 /*
311  * Drop a ref, return true if the refcount fell to zero (the page has no users)
312  */
313 static inline int put_page_testzero(struct page *page)
314 {
315         VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
316         return atomic_dec_and_test(&page->_count);
317 }
318 
319 /*
320  * Try to grab a ref unless the page has a refcount of zero, return false if
321  * that is the case.
322  * This can be called when MMU is off so it must not access
323  * any of the virtual mappings.
324  */
325 static inline int get_page_unless_zero(struct page *page)
326 {
327         return atomic_inc_not_zero(&page->_count);
328 }
329 
330 /*
331  * Try to drop a ref unless the page has a refcount of one, return false if
332  * that is the case.
333  * This is to make sure that the refcount won't become zero after this drop.
334  * This can be called when MMU is off so it must not access
335  * any of the virtual mappings.
336  */
337 static inline int put_page_unless_one(struct page *page)
338 {
339         return atomic_add_unless(&page->_count, -1, 1);
340 }
341 
342 extern int page_is_ram(unsigned long pfn);
343 
344 /* Support for virtually mapped pages */
345 struct page *vmalloc_to_page(const void *addr);
346 unsigned long vmalloc_to_pfn(const void *addr);
347 
348 /*
349  * Determine if an address is within the vmalloc range
350  *
351  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
352  * is no special casing required.
353  */
354 static inline int is_vmalloc_addr(const void *x)
355 {
356 #ifdef CONFIG_MMU
357         unsigned long addr = (unsigned long)x;
358 
359         return addr >= VMALLOC_START && addr < VMALLOC_END;
360 #else
361         return 0;
362 #endif
363 }
364 #ifdef CONFIG_MMU
365 extern int is_vmalloc_or_module_addr(const void *x);
366 #else
367 static inline int is_vmalloc_or_module_addr(const void *x)
368 {
369         return 0;
370 }
371 #endif
372 
373 extern void kvfree(const void *addr);
374 
375 static inline void compound_lock(struct page *page)
376 {
377 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
378         VM_BUG_ON_PAGE(PageSlab(page), page);
379         bit_spin_lock(PG_compound_lock, &page->flags);
380 #endif
381 }
382 
383 static inline void compound_unlock(struct page *page)
384 {
385 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
386         VM_BUG_ON_PAGE(PageSlab(page), page);
387         bit_spin_unlock(PG_compound_lock, &page->flags);
388 #endif
389 }
390 
391 static inline unsigned long compound_lock_irqsave(struct page *page)
392 {
393         unsigned long uninitialized_var(flags);
394 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
395         local_irq_save(flags);
396         compound_lock(page);
397 #endif
398         return flags;
399 }
400 
401 static inline void compound_unlock_irqrestore(struct page *page,
402                                               unsigned long flags)
403 {
404 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
405         compound_unlock(page);
406         local_irq_restore(flags);
407 #endif
408 }
409 
410 static inline struct page *compound_head_by_tail(struct page *tail)
411 {
412         struct page *head = tail->first_page;
413 
414         /*
415          * page->first_page may be a dangling pointer to an old
416          * compound page, so recheck that it is still a tail
417          * page before returning.
418          */
419         smp_rmb();
420         if (likely(PageTail(tail)))
421                 return head;
422         return tail;
423 }
424 
425 static inline struct page *compound_head(struct page *page)
426 {
427         if (unlikely(PageTail(page)))
428                 return compound_head_by_tail(page);
429         return page;
430 }
431 
432 /*
433  * The atomic page->_mapcount, starts from -1: so that transitions
434  * both from it and to it can be tracked, using atomic_inc_and_test
435  * and atomic_add_negative(-1).
436  */
437 static inline void page_mapcount_reset(struct page *page)
438 {
439         atomic_set(&(page)->_mapcount, -1);
440 }
441 
442 static inline int page_mapcount(struct page *page)
443 {
444         return atomic_read(&(page)->_mapcount) + 1;
445 }
446 
447 static inline int page_count(struct page *page)
448 {
449         return atomic_read(&compound_head(page)->_count);
450 }
451 
452 #ifdef CONFIG_HUGETLB_PAGE
453 extern int PageHeadHuge(struct page *page_head);
454 #else /* CONFIG_HUGETLB_PAGE */
455 static inline int PageHeadHuge(struct page *page_head)
456 {
457         return 0;
458 }
459 #endif /* CONFIG_HUGETLB_PAGE */
460 
461 static inline bool __compound_tail_refcounted(struct page *page)
462 {
463         return !PageSlab(page) && !PageHeadHuge(page);
464 }
465 
466 /*
467  * This takes a head page as parameter and tells if the
468  * tail page reference counting can be skipped.
469  *
470  * For this to be safe, PageSlab and PageHeadHuge must remain true on
471  * any given page where they return true here, until all tail pins
472  * have been released.
473  */
474 static inline bool compound_tail_refcounted(struct page *page)
475 {
476         VM_BUG_ON_PAGE(!PageHead(page), page);
477         return __compound_tail_refcounted(page);
478 }
479 
480 static inline void get_huge_page_tail(struct page *page)
481 {
482         /*
483          * __split_huge_page_refcount() cannot run from under us.
484          */
485         VM_BUG_ON_PAGE(!PageTail(page), page);
486         VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
487         VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
488         if (compound_tail_refcounted(page->first_page))
489                 atomic_inc(&page->_mapcount);
490 }
491 
492 extern bool __get_page_tail(struct page *page);
493 
494 static inline void get_page(struct page *page)
495 {
496         if (unlikely(PageTail(page)))
497                 if (likely(__get_page_tail(page)))
498                         return;
499         /*
500          * Getting a normal page or the head of a compound page
501          * requires to already have an elevated page->_count.
502          */
503         VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
504         atomic_inc(&page->_count);
505 }
506 
507 static inline struct page *virt_to_head_page(const void *x)
508 {
509         struct page *page = virt_to_page(x);
510         return compound_head(page);
511 }
512 
513 /*
514  * Setup the page count before being freed into the page allocator for
515  * the first time (boot or memory hotplug)
516  */
517 static inline void init_page_count(struct page *page)
518 {
519         atomic_set(&page->_count, 1);
520 }
521 
522 /*
523  * PageBuddy() indicate that the page is free and in the buddy system
524  * (see mm/page_alloc.c).
525  *
526  * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
527  * -2 so that an underflow of the page_mapcount() won't be mistaken
528  * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
529  * efficiently by most CPU architectures.
530  */
531 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
532 
533 static inline int PageBuddy(struct page *page)
534 {
535         return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
536 }
537 
538 static inline void __SetPageBuddy(struct page *page)
539 {
540         VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
541         atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
542 }
543 
544 static inline void __ClearPageBuddy(struct page *page)
545 {
546         VM_BUG_ON_PAGE(!PageBuddy(page), page);
547         atomic_set(&page->_mapcount, -1);
548 }
549 
550 #define PAGE_BALLOON_MAPCOUNT_VALUE (-256)
551 
552 static inline int PageBalloon(struct page *page)
553 {
554         return atomic_read(&page->_mapcount) == PAGE_BALLOON_MAPCOUNT_VALUE;
555 }
556 
557 static inline void __SetPageBalloon(struct page *page)
558 {
559         VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
560         atomic_set(&page->_mapcount, PAGE_BALLOON_MAPCOUNT_VALUE);
561 }
562 
563 static inline void __ClearPageBalloon(struct page *page)
564 {
565         VM_BUG_ON_PAGE(!PageBalloon(page), page);
566         atomic_set(&page->_mapcount, -1);
567 }
568 
569 void put_page(struct page *page);
570 void put_pages_list(struct list_head *pages);
571 
572 void split_page(struct page *page, unsigned int order);
573 int split_free_page(struct page *page);
574 
575 /*
576  * Compound pages have a destructor function.  Provide a
577  * prototype for that function and accessor functions.
578  * These are _only_ valid on the head of a PG_compound page.
579  */
580 typedef void compound_page_dtor(struct page *);
581 
582 static inline void set_compound_page_dtor(struct page *page,
583                                                 compound_page_dtor *dtor)
584 {
585         page[1].lru.next = (void *)dtor;
586 }
587 
588 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
589 {
590         return (compound_page_dtor *)page[1].lru.next;
591 }
592 
593 static inline int compound_order(struct page *page)
594 {
595         if (!PageHead(page))
596                 return 0;
597         return (unsigned long)page[1].lru.prev;
598 }
599 
600 static inline void set_compound_order(struct page *page, unsigned long order)
601 {
602         page[1].lru.prev = (void *)order;
603 }
604 
605 #ifdef CONFIG_MMU
606 /*
607  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
608  * servicing faults for write access.  In the normal case, do always want
609  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
610  * that do not have writing enabled, when used by access_process_vm.
611  */
612 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
613 {
614         if (likely(vma->vm_flags & VM_WRITE))
615                 pte = pte_mkwrite(pte);
616         return pte;
617 }
618 
619 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
620                 struct page *page, pte_t *pte, bool write, bool anon);
621 #endif
622 
623 /*
624  * Multiple processes may "see" the same page. E.g. for untouched
625  * mappings of /dev/null, all processes see the same page full of
626  * zeroes, and text pages of executables and shared libraries have
627  * only one copy in memory, at most, normally.
628  *
629  * For the non-reserved pages, page_count(page) denotes a reference count.
630  *   page_count() == 0 means the page is free. page->lru is then used for
631  *   freelist management in the buddy allocator.
632  *   page_count() > 0  means the page has been allocated.
633  *
634  * Pages are allocated by the slab allocator in order to provide memory
635  * to kmalloc and kmem_cache_alloc. In this case, the management of the
636  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
637  * unless a particular usage is carefully commented. (the responsibility of
638  * freeing the kmalloc memory is the caller's, of course).
639  *
640  * A page may be used by anyone else who does a __get_free_page().
641  * In this case, page_count still tracks the references, and should only
642  * be used through the normal accessor functions. The top bits of page->flags
643  * and page->virtual store page management information, but all other fields
644  * are unused and could be used privately, carefully. The management of this
645  * page is the responsibility of the one who allocated it, and those who have
646  * subsequently been given references to it.
647  *
648  * The other pages (we may call them "pagecache pages") are completely
649  * managed by the Linux memory manager: I/O, buffers, swapping etc.
650  * The following discussion applies only to them.
651  *
652  * A pagecache page contains an opaque `private' member, which belongs to the
653  * page's address_space. Usually, this is the address of a circular list of
654  * the page's disk buffers. PG_private must be set to tell the VM to call
655  * into the filesystem to release these pages.
656  *
657  * A page may belong to an inode's memory mapping. In this case, page->mapping
658  * is the pointer to the inode, and page->index is the file offset of the page,
659  * in units of PAGE_CACHE_SIZE.
660  *
661  * If pagecache pages are not associated with an inode, they are said to be
662  * anonymous pages. These may become associated with the swapcache, and in that
663  * case PG_swapcache is set, and page->private is an offset into the swapcache.
664  *
665  * In either case (swapcache or inode backed), the pagecache itself holds one
666  * reference to the page. Setting PG_private should also increment the
667  * refcount. The each user mapping also has a reference to the page.
668  *
669  * The pagecache pages are stored in a per-mapping radix tree, which is
670  * rooted at mapping->page_tree, and indexed by offset.
671  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
672  * lists, we instead now tag pages as dirty/writeback in the radix tree.
673  *
674  * All pagecache pages may be subject to I/O:
675  * - inode pages may need to be read from disk,
676  * - inode pages which have been modified and are MAP_SHARED may need
677  *   to be written back to the inode on disk,
678  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
679  *   modified may need to be swapped out to swap space and (later) to be read
680  *   back into memory.
681  */
682 
683 /*
684  * The zone field is never updated after free_area_init_core()
685  * sets it, so none of the operations on it need to be atomic.
686  */
687 
688 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
689 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
690 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
691 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
692 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
693 
694 /*
695  * Define the bit shifts to access each section.  For non-existent
696  * sections we define the shift as 0; that plus a 0 mask ensures
697  * the compiler will optimise away reference to them.
698  */
699 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
700 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
701 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
702 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
703 
704 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
705 #ifdef NODE_NOT_IN_PAGE_FLAGS
706 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
707 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
708                                                 SECTIONS_PGOFF : ZONES_PGOFF)
709 #else
710 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
711 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
712                                                 NODES_PGOFF : ZONES_PGOFF)
713 #endif
714 
715 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
716 
717 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
718 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
719 #endif
720 
721 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
722 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
723 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
724 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
725 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
726 
727 static inline enum zone_type page_zonenum(const struct page *page)
728 {
729         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
730 }
731 
732 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
733 #define SECTION_IN_PAGE_FLAGS
734 #endif
735 
736 /*
737  * The identification function is mainly used by the buddy allocator for
738  * determining if two pages could be buddies. We are not really identifying
739  * the zone since we could be using the section number id if we do not have
740  * node id available in page flags.
741  * We only guarantee that it will return the same value for two combinable
742  * pages in a zone.
743  */
744 static inline int page_zone_id(struct page *page)
745 {
746         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
747 }
748 
749 static inline int zone_to_nid(struct zone *zone)
750 {
751 #ifdef CONFIG_NUMA
752         return zone->node;
753 #else
754         return 0;
755 #endif
756 }
757 
758 #ifdef NODE_NOT_IN_PAGE_FLAGS
759 extern int page_to_nid(const struct page *page);
760 #else
761 static inline int page_to_nid(const struct page *page)
762 {
763         return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
764 }
765 #endif
766 
767 #ifdef CONFIG_NUMA_BALANCING
768 static inline int cpu_pid_to_cpupid(int cpu, int pid)
769 {
770         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
771 }
772 
773 static inline int cpupid_to_pid(int cpupid)
774 {
775         return cpupid & LAST__PID_MASK;
776 }
777 
778 static inline int cpupid_to_cpu(int cpupid)
779 {
780         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
781 }
782 
783 static inline int cpupid_to_nid(int cpupid)
784 {
785         return cpu_to_node(cpupid_to_cpu(cpupid));
786 }
787 
788 static inline bool cpupid_pid_unset(int cpupid)
789 {
790         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
791 }
792 
793 static inline bool cpupid_cpu_unset(int cpupid)
794 {
795         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
796 }
797 
798 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
799 {
800         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
801 }
802 
803 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
804 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
805 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
806 {
807         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
808 }
809 
810 static inline int page_cpupid_last(struct page *page)
811 {
812         return page->_last_cpupid;
813 }
814 static inline void page_cpupid_reset_last(struct page *page)
815 {
816         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
817 }
818 #else
819 static inline int page_cpupid_last(struct page *page)
820 {
821         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
822 }
823 
824 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
825 
826 static inline void page_cpupid_reset_last(struct page *page)
827 {
828         int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
829 
830         page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
831         page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
832 }
833 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
834 #else /* !CONFIG_NUMA_BALANCING */
835 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
836 {
837         return page_to_nid(page); /* XXX */
838 }
839 
840 static inline int page_cpupid_last(struct page *page)
841 {
842         return page_to_nid(page); /* XXX */
843 }
844 
845 static inline int cpupid_to_nid(int cpupid)
846 {
847         return -1;
848 }
849 
850 static inline int cpupid_to_pid(int cpupid)
851 {
852         return -1;
853 }
854 
855 static inline int cpupid_to_cpu(int cpupid)
856 {
857         return -1;
858 }
859 
860 static inline int cpu_pid_to_cpupid(int nid, int pid)
861 {
862         return -1;
863 }
864 
865 static inline bool cpupid_pid_unset(int cpupid)
866 {
867         return 1;
868 }
869 
870 static inline void page_cpupid_reset_last(struct page *page)
871 {
872 }
873 
874 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
875 {
876         return false;
877 }
878 #endif /* CONFIG_NUMA_BALANCING */
879 
880 static inline struct zone *page_zone(const struct page *page)
881 {
882         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
883 }
884 
885 #ifdef SECTION_IN_PAGE_FLAGS
886 static inline void set_page_section(struct page *page, unsigned long section)
887 {
888         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
889         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
890 }
891 
892 static inline unsigned long page_to_section(const struct page *page)
893 {
894         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
895 }
896 #endif
897 
898 static inline void set_page_zone(struct page *page, enum zone_type zone)
899 {
900         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
901         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
902 }
903 
904 static inline void set_page_node(struct page *page, unsigned long node)
905 {
906         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
907         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
908 }
909 
910 static inline void set_page_links(struct page *page, enum zone_type zone,
911         unsigned long node, unsigned long pfn)
912 {
913         set_page_zone(page, zone);
914         set_page_node(page, node);
915 #ifdef SECTION_IN_PAGE_FLAGS
916         set_page_section(page, pfn_to_section_nr(pfn));
917 #endif
918 }
919 
920 /*
921  * Some inline functions in vmstat.h depend on page_zone()
922  */
923 #include <linux/vmstat.h>
924 
925 static __always_inline void *lowmem_page_address(const struct page *page)
926 {
927         return __va(PFN_PHYS(page_to_pfn(page)));
928 }
929 
930 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
931 #define HASHED_PAGE_VIRTUAL
932 #endif
933 
934 #if defined(WANT_PAGE_VIRTUAL)
935 static inline void *page_address(const struct page *page)
936 {
937         return page->virtual;
938 }
939 static inline void set_page_address(struct page *page, void *address)
940 {
941         page->virtual = address;
942 }
943 #define page_address_init()  do { } while(0)
944 #endif
945 
946 #if defined(HASHED_PAGE_VIRTUAL)
947 void *page_address(const struct page *page);
948 void set_page_address(struct page *page, void *virtual);
949 void page_address_init(void);
950 #endif
951 
952 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
953 #define page_address(page) lowmem_page_address(page)
954 #define set_page_address(page, address)  do { } while(0)
955 #define page_address_init()  do { } while(0)
956 #endif
957 
958 /*
959  * On an anonymous page mapped into a user virtual memory area,
960  * page->mapping points to its anon_vma, not to a struct address_space;
961  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
962  *
963  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
964  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
965  * and then page->mapping points, not to an anon_vma, but to a private
966  * structure which KSM associates with that merged page.  See ksm.h.
967  *
968  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
969  *
970  * Please note that, confusingly, "page_mapping" refers to the inode
971  * address_space which maps the page from disk; whereas "page_mapped"
972  * refers to user virtual address space into which the page is mapped.
973  */
974 #define PAGE_MAPPING_ANON       1
975 #define PAGE_MAPPING_KSM        2
976 #define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
977 
978 extern struct address_space *page_mapping(struct page *page);
979 
980 /* Neutral page->mapping pointer to address_space or anon_vma or other */
981 static inline void *page_rmapping(struct page *page)
982 {
983         return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
984 }
985 
986 extern struct address_space *__page_file_mapping(struct page *);
987 
988 static inline
989 struct address_space *page_file_mapping(struct page *page)
990 {
991         if (unlikely(PageSwapCache(page)))
992                 return __page_file_mapping(page);
993 
994         return page->mapping;
995 }
996 
997 static inline int PageAnon(struct page *page)
998 {
999         return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
1000 }
1001 
1002 /*
1003  * Return the pagecache index of the passed page.  Regular pagecache pages
1004  * use ->index whereas swapcache pages use ->private
1005  */
1006 static inline pgoff_t page_index(struct page *page)
1007 {
1008         if (unlikely(PageSwapCache(page)))
1009                 return page_private(page);
1010         return page->index;
1011 }
1012 
1013 extern pgoff_t __page_file_index(struct page *page);
1014 
1015 /*
1016  * Return the file index of the page. Regular pagecache pages use ->index
1017  * whereas swapcache pages use swp_offset(->private)
1018  */
1019 static inline pgoff_t page_file_index(struct page *page)
1020 {
1021         if (unlikely(PageSwapCache(page)))
1022                 return __page_file_index(page);
1023 
1024         return page->index;
1025 }
1026 
1027 /*
1028  * Return true if this page is mapped into pagetables.
1029  */
1030 static inline int page_mapped(struct page *page)
1031 {
1032         return atomic_read(&(page)->_mapcount) >= 0;
1033 }
1034 
1035 /*
1036  * Different kinds of faults, as returned by handle_mm_fault().
1037  * Used to decide whether a process gets delivered SIGBUS or
1038  * just gets major/minor fault counters bumped up.
1039  */
1040 
1041 #define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
1042 
1043 #define VM_FAULT_OOM    0x0001
1044 #define VM_FAULT_SIGBUS 0x0002
1045 #define VM_FAULT_MAJOR  0x0004
1046 #define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
1047 #define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
1048 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1049 #define VM_FAULT_SIGSEGV 0x0040
1050 
1051 #define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
1052 #define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
1053 #define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
1054 #define VM_FAULT_FALLBACK 0x0800        /* huge page fault failed, fall back to small */
1055 
1056 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1057 
1058 #define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1059                          VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1060                          VM_FAULT_FALLBACK)
1061 
1062 /* Encode hstate index for a hwpoisoned large page */
1063 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1064 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1065 
1066 /*
1067  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1068  */
1069 extern void pagefault_out_of_memory(void);
1070 
1071 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1072 
1073 /*
1074  * Flags passed to show_mem() and show_free_areas() to suppress output in
1075  * various contexts.
1076  */
1077 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1078 
1079 extern void show_free_areas(unsigned int flags);
1080 extern bool skip_free_areas_node(unsigned int flags, int nid);
1081 
1082 int shmem_zero_setup(struct vm_area_struct *);
1083 #ifdef CONFIG_SHMEM
1084 bool shmem_mapping(struct address_space *mapping);
1085 #else
1086 static inline bool shmem_mapping(struct address_space *mapping)
1087 {
1088         return false;
1089 }
1090 #endif
1091 
1092 extern int can_do_mlock(void);
1093 extern int user_shm_lock(size_t, struct user_struct *);
1094 extern void user_shm_unlock(size_t, struct user_struct *);
1095 
1096 /*
1097  * Parameter block passed down to zap_pte_range in exceptional cases.
1098  */
1099 struct zap_details {
1100         struct address_space *check_mapping;    /* Check page->mapping if set */
1101         pgoff_t first_index;                    /* Lowest page->index to unmap */
1102         pgoff_t last_index;                     /* Highest page->index to unmap */
1103 };
1104 
1105 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1106                 pte_t pte);
1107 
1108 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1109                 unsigned long size);
1110 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1111                 unsigned long size, struct zap_details *);
1112 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1113                 unsigned long start, unsigned long end);
1114 
1115 /**
1116  * mm_walk - callbacks for walk_page_range
1117  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1118  *             this handler is required to be able to handle
1119  *             pmd_trans_huge() pmds.  They may simply choose to
1120  *             split_huge_page() instead of handling it explicitly.
1121  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1122  * @pte_hole: if set, called for each hole at all levels
1123  * @hugetlb_entry: if set, called for each hugetlb entry
1124  * @test_walk: caller specific callback function to determine whether
1125  *             we walk over the current vma or not. A positive returned
1126  *             value means "do page table walk over the current vma,"
1127  *             and a negative one means "abort current page table walk
1128  *             right now." 0 means "skip the current vma."
1129  * @mm:        mm_struct representing the target process of page table walk
1130  * @vma:       vma currently walked (NULL if walking outside vmas)
1131  * @private:   private data for callbacks' usage
1132  *
1133  * (see the comment on walk_page_range() for more details)
1134  */
1135 struct mm_walk {
1136         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1137                          unsigned long next, struct mm_walk *walk);
1138         int (*pte_entry)(pte_t *pte, unsigned long addr,
1139                          unsigned long next, struct mm_walk *walk);
1140         int (*pte_hole)(unsigned long addr, unsigned long next,
1141                         struct mm_walk *walk);
1142         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1143                              unsigned long addr, unsigned long next,
1144                              struct mm_walk *walk);
1145         int (*test_walk)(unsigned long addr, unsigned long next,
1146                         struct mm_walk *walk);
1147         struct mm_struct *mm;
1148         struct vm_area_struct *vma;
1149         void *private;
1150 };
1151 
1152 int walk_page_range(unsigned long addr, unsigned long end,
1153                 struct mm_walk *walk);
1154 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1155                 unsigned long end, unsigned long floor, unsigned long ceiling);
1156 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1157                         struct vm_area_struct *vma);
1158 void unmap_mapping_range(struct address_space *mapping,
1159                 loff_t const holebegin, loff_t const holelen, int even_cows);
1160 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1161         unsigned long *pfn);
1162 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1163                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1164 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1165                         void *buf, int len, int write);
1166 
1167 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1168                 loff_t const holebegin, loff_t const holelen)
1169 {
1170         unmap_mapping_range(mapping, holebegin, holelen, 0);
1171 }
1172 
1173 extern void truncate_pagecache(struct inode *inode, loff_t new);
1174 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1175 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1176 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1177 int truncate_inode_page(struct address_space *mapping, struct page *page);
1178 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1179 int invalidate_inode_page(struct page *page);
1180 
1181 #ifdef CONFIG_MMU
1182 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1183                         unsigned long address, unsigned int flags);
1184 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1185                             unsigned long address, unsigned int fault_flags);
1186 #else
1187 static inline int handle_mm_fault(struct mm_struct *mm,
1188                         struct vm_area_struct *vma, unsigned long address,
1189                         unsigned int flags)
1190 {
1191         /* should never happen if there's no MMU */
1192         BUG();
1193         return VM_FAULT_SIGBUS;
1194 }
1195 static inline int fixup_user_fault(struct task_struct *tsk,
1196                 struct mm_struct *mm, unsigned long address,
1197                 unsigned int fault_flags)
1198 {
1199         /* should never happen if there's no MMU */
1200         BUG();
1201         return -EFAULT;
1202 }
1203 #endif
1204 
1205 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1206 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1207                 void *buf, int len, int write);
1208 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1209                 unsigned long addr, void *buf, int len, int write);
1210 
1211 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1212                       unsigned long start, unsigned long nr_pages,
1213                       unsigned int foll_flags, struct page **pages,
1214                       struct vm_area_struct **vmas, int *nonblocking);
1215 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1216                     unsigned long start, unsigned long nr_pages,
1217                     int write, int force, struct page **pages,
1218                     struct vm_area_struct **vmas);
1219 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1220                         struct page **pages);
1221 struct kvec;
1222 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1223                         struct page **pages);
1224 int get_kernel_page(unsigned long start, int write, struct page **pages);
1225 struct page *get_dump_page(unsigned long addr);
1226 
1227 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1228 extern void do_invalidatepage(struct page *page, unsigned int offset,
1229                               unsigned int length);
1230 
1231 int __set_page_dirty_nobuffers(struct page *page);
1232 int __set_page_dirty_no_writeback(struct page *page);
1233 int redirty_page_for_writepage(struct writeback_control *wbc,
1234                                 struct page *page);
1235 void account_page_dirtied(struct page *page, struct address_space *mapping);
1236 void account_page_writeback(struct page *page);
1237 int set_page_dirty(struct page *page);
1238 int set_page_dirty_lock(struct page *page);
1239 int clear_page_dirty_for_io(struct page *page);
1240 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1241 
1242 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1243 
1244 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1245 {
1246         return !vma->vm_ops;
1247 }
1248 
1249 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1250                 unsigned long old_addr, struct vm_area_struct *new_vma,
1251                 unsigned long new_addr, unsigned long len,
1252                 bool need_rmap_locks);
1253 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1254                               unsigned long end, pgprot_t newprot,
1255                               int dirty_accountable, int prot_numa);
1256 extern int mprotect_fixup(struct vm_area_struct *vma,
1257                           struct vm_area_struct **pprev, unsigned long start,
1258                           unsigned long end, unsigned long newflags);
1259 
1260 /*
1261  * doesn't attempt to fault and will return short.
1262  */
1263 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1264                           struct page **pages);
1265 /*
1266  * per-process(per-mm_struct) statistics.
1267  */
1268 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1269 {
1270         long val = atomic_long_read(&mm->rss_stat.count[member]);
1271 
1272 #ifdef SPLIT_RSS_COUNTING
1273         /*
1274          * counter is updated in asynchronous manner and may go to minus.
1275          * But it's never be expected number for users.
1276          */
1277         if (val < 0)
1278                 val = 0;
1279 #endif
1280         return (unsigned long)val;
1281 }
1282 
1283 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1284 {
1285         atomic_long_add(value, &mm->rss_stat.count[member]);
1286 }
1287 
1288 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1289 {
1290         atomic_long_inc(&mm->rss_stat.count[member]);
1291 }
1292 
1293 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1294 {
1295         atomic_long_dec(&mm->rss_stat.count[member]);
1296 }
1297 
1298 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1299 {
1300         return get_mm_counter(mm, MM_FILEPAGES) +
1301                 get_mm_counter(mm, MM_ANONPAGES);
1302 }
1303 
1304 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1305 {
1306         return max(mm->hiwater_rss, get_mm_rss(mm));
1307 }
1308 
1309 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1310 {
1311         return max(mm->hiwater_vm, mm->total_vm);
1312 }
1313 
1314 static inline void update_hiwater_rss(struct mm_struct *mm)
1315 {
1316         unsigned long _rss = get_mm_rss(mm);
1317 
1318         if ((mm)->hiwater_rss < _rss)
1319                 (mm)->hiwater_rss = _rss;
1320 }
1321 
1322 static inline void update_hiwater_vm(struct mm_struct *mm)
1323 {
1324         if (mm->hiwater_vm < mm->total_vm)
1325                 mm->hiwater_vm = mm->total_vm;
1326 }
1327 
1328 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1329                                          struct mm_struct *mm)
1330 {
1331         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1332 
1333         if (*maxrss < hiwater_rss)
1334                 *maxrss = hiwater_rss;
1335 }
1336 
1337 #if defined(SPLIT_RSS_COUNTING)
1338 void sync_mm_rss(struct mm_struct *mm);
1339 #else
1340 static inline void sync_mm_rss(struct mm_struct *mm)
1341 {
1342 }
1343 #endif
1344 
1345 int vma_wants_writenotify(struct vm_area_struct *vma);
1346 
1347 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1348                                spinlock_t **ptl);
1349 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1350                                     spinlock_t **ptl)
1351 {
1352         pte_t *ptep;
1353         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1354         return ptep;
1355 }
1356 
1357 #ifdef __PAGETABLE_PUD_FOLDED
1358 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1359                                                 unsigned long address)
1360 {
1361         return 0;
1362 }
1363 #else
1364 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1365 #endif
1366 
1367 #ifdef __PAGETABLE_PMD_FOLDED
1368 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1369                                                 unsigned long address)
1370 {
1371         return 0;
1372 }
1373 #else
1374 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1375 #endif
1376 
1377 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1378                 pmd_t *pmd, unsigned long address);
1379 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1380 
1381 /*
1382  * The following ifdef needed to get the 4level-fixup.h header to work.
1383  * Remove it when 4level-fixup.h has been removed.
1384  */
1385 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1386 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1387 {
1388         return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1389                 NULL: pud_offset(pgd, address);
1390 }
1391 
1392 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1393 {
1394         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1395                 NULL: pmd_offset(pud, address);
1396 }
1397 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1398 
1399 #if USE_SPLIT_PTE_PTLOCKS
1400 #if ALLOC_SPLIT_PTLOCKS
1401 void __init ptlock_cache_init(void);
1402 extern bool ptlock_alloc(struct page *page);
1403 extern void ptlock_free(struct page *page);
1404 
1405 static inline spinlock_t *ptlock_ptr(struct page *page)
1406 {
1407         return page->ptl;
1408 }
1409 #else /* ALLOC_SPLIT_PTLOCKS */
1410 static inline void ptlock_cache_init(void)
1411 {
1412 }
1413 
1414 static inline bool ptlock_alloc(struct page *page)
1415 {
1416         return true;
1417 }
1418 
1419 static inline void ptlock_free(struct page *page)
1420 {
1421 }
1422 
1423 static inline spinlock_t *ptlock_ptr(struct page *page)
1424 {
1425         return &page->ptl;
1426 }
1427 #endif /* ALLOC_SPLIT_PTLOCKS */
1428 
1429 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1430 {
1431         return ptlock_ptr(pmd_page(*pmd));
1432 }
1433 
1434 static inline bool ptlock_init(struct page *page)
1435 {
1436         /*
1437          * prep_new_page() initialize page->private (and therefore page->ptl)
1438          * with 0. Make sure nobody took it in use in between.
1439          *
1440          * It can happen if arch try to use slab for page table allocation:
1441          * slab code uses page->slab_cache and page->first_page (for tail
1442          * pages), which share storage with page->ptl.
1443          */
1444         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1445         if (!ptlock_alloc(page))
1446                 return false;
1447         spin_lock_init(ptlock_ptr(page));
1448         return true;
1449 }
1450 
1451 /* Reset page->mapping so free_pages_check won't complain. */
1452 static inline void pte_lock_deinit(struct page *page)
1453 {
1454         page->mapping = NULL;
1455         ptlock_free(page);
1456 }
1457 
1458 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1459 /*
1460  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1461  */
1462 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1463 {
1464         return &mm->page_table_lock;
1465 }
1466 static inline void ptlock_cache_init(void) {}
1467 static inline bool ptlock_init(struct page *page) { return true; }
1468 static inline void pte_lock_deinit(struct page *page) {}
1469 #endif /* USE_SPLIT_PTE_PTLOCKS */
1470 
1471 static inline void pgtable_init(void)
1472 {
1473         ptlock_cache_init();
1474         pgtable_cache_init();
1475 }
1476 
1477 static inline bool pgtable_page_ctor(struct page *page)
1478 {
1479         inc_zone_page_state(page, NR_PAGETABLE);
1480         return ptlock_init(page);
1481 }
1482 
1483 static inline void pgtable_page_dtor(struct page *page)
1484 {
1485         pte_lock_deinit(page);
1486         dec_zone_page_state(page, NR_PAGETABLE);
1487 }
1488 
1489 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1490 ({                                                      \
1491         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1492         pte_t *__pte = pte_offset_map(pmd, address);    \
1493         *(ptlp) = __ptl;                                \
1494         spin_lock(__ptl);                               \
1495         __pte;                                          \
1496 })
1497 
1498 #define pte_unmap_unlock(pte, ptl)      do {            \
1499         spin_unlock(ptl);                               \
1500         pte_unmap(pte);                                 \
1501 } while (0)
1502 
1503 #define pte_alloc_map(mm, vma, pmd, address)                            \
1504         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1505                                                         pmd, address))? \
1506          NULL: pte_offset_map(pmd, address))
1507 
1508 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1509         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1510                                                         pmd, address))? \
1511                 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1512 
1513 #define pte_alloc_kernel(pmd, address)                  \
1514         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1515                 NULL: pte_offset_kernel(pmd, address))
1516 
1517 #if USE_SPLIT_PMD_PTLOCKS
1518 
1519 static struct page *pmd_to_page(pmd_t *pmd)
1520 {
1521         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1522         return virt_to_page((void *)((unsigned long) pmd & mask));
1523 }
1524 
1525 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1526 {
1527         return ptlock_ptr(pmd_to_page(pmd));
1528 }
1529 
1530 static inline bool pgtable_pmd_page_ctor(struct page *page)
1531 {
1532 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1533         page->pmd_huge_pte = NULL;
1534 #endif
1535         return ptlock_init(page);
1536 }
1537 
1538 static inline void pgtable_pmd_page_dtor(struct page *page)
1539 {
1540 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1541         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1542 #endif
1543         ptlock_free(page);
1544 }
1545 
1546 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1547 
1548 #else
1549 
1550 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1551 {
1552         return &mm->page_table_lock;
1553 }
1554 
1555 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1556 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1557 
1558 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1559 
1560 #endif
1561 
1562 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1563 {
1564         spinlock_t *ptl = pmd_lockptr(mm, pmd);
1565         spin_lock(ptl);
1566         return ptl;
1567 }
1568 
1569 extern void free_area_init(unsigned long * zones_size);
1570 extern void free_area_init_node(int nid, unsigned long * zones_size,
1571                 unsigned long zone_start_pfn, unsigned long *zholes_size);
1572 extern void free_initmem(void);
1573 
1574 /*
1575  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1576  * into the buddy system. The freed pages will be poisoned with pattern
1577  * "poison" if it's within range [0, UCHAR_MAX].
1578  * Return pages freed into the buddy system.
1579  */
1580 extern unsigned long free_reserved_area(void *start, void *end,
1581                                         int poison, char *s);
1582 
1583 #ifdef  CONFIG_HIGHMEM
1584 /*
1585  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1586  * and totalram_pages.
1587  */
1588 extern void free_highmem_page(struct page *page);
1589 #endif
1590 
1591 extern void adjust_managed_page_count(struct page *page, long count);
1592 extern void mem_init_print_info(const char *str);
1593 
1594 /* Free the reserved page into the buddy system, so it gets managed. */
1595 static inline void __free_reserved_page(struct page *page)
1596 {
1597         ClearPageReserved(page);
1598         init_page_count(page);
1599         __free_page(page);
1600 }
1601 
1602 static inline void free_reserved_page(struct page *page)
1603 {
1604         __free_reserved_page(page);
1605         adjust_managed_page_count(page, 1);
1606 }
1607 
1608 static inline void mark_page_reserved(struct page *page)
1609 {
1610         SetPageReserved(page);
1611         adjust_managed_page_count(page, -1);
1612 }
1613 
1614 /*
1615  * Default method to free all the __init memory into the buddy system.
1616  * The freed pages will be poisoned with pattern "poison" if it's within
1617  * range [0, UCHAR_MAX].
1618  * Return pages freed into the buddy system.
1619  */
1620 static inline unsigned long free_initmem_default(int poison)
1621 {
1622         extern char __init_begin[], __init_end[];
1623 
1624         return free_reserved_area(&__init_begin, &__init_end,
1625                                   poison, "unused kernel");
1626 }
1627 
1628 static inline unsigned long get_num_physpages(void)
1629 {
1630         int nid;
1631         unsigned long phys_pages = 0;
1632 
1633         for_each_online_node(nid)
1634                 phys_pages += node_present_pages(nid);
1635 
1636         return phys_pages;
1637 }
1638 
1639 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1640 /*
1641  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1642  * zones, allocate the backing mem_map and account for memory holes in a more
1643  * architecture independent manner. This is a substitute for creating the
1644  * zone_sizes[] and zholes_size[] arrays and passing them to
1645  * free_area_init_node()
1646  *
1647  * An architecture is expected to register range of page frames backed by
1648  * physical memory with memblock_add[_node]() before calling
1649  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1650  * usage, an architecture is expected to do something like
1651  *
1652  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1653  *                                                       max_highmem_pfn};
1654  * for_each_valid_physical_page_range()
1655  *      memblock_add_node(base, size, nid)
1656  * free_area_init_nodes(max_zone_pfns);
1657  *
1658  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1659  * registered physical page range.  Similarly
1660  * sparse_memory_present_with_active_regions() calls memory_present() for
1661  * each range when SPARSEMEM is enabled.
1662  *
1663  * See mm/page_alloc.c for more information on each function exposed by
1664  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1665  */
1666 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1667 unsigned long node_map_pfn_alignment(void);
1668 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1669                                                 unsigned long end_pfn);
1670 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1671                                                 unsigned long end_pfn);
1672 extern void get_pfn_range_for_nid(unsigned int nid,
1673                         unsigned long *start_pfn, unsigned long *end_pfn);
1674 extern unsigned long find_min_pfn_with_active_regions(void);
1675 extern void free_bootmem_with_active_regions(int nid,
1676                                                 unsigned long max_low_pfn);
1677 extern void sparse_memory_present_with_active_regions(int nid);
1678 
1679 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1680 
1681 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1682     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1683 static inline int __early_pfn_to_nid(unsigned long pfn)
1684 {
1685         return 0;
1686 }
1687 #else
1688 /* please see mm/page_alloc.c */
1689 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1690 /* there is a per-arch backend function. */
1691 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1692 #endif
1693 
1694 extern void set_dma_reserve(unsigned long new_dma_reserve);
1695 extern void memmap_init_zone(unsigned long, int, unsigned long,
1696                                 unsigned long, enum memmap_context);
1697 extern void setup_per_zone_wmarks(void);
1698 extern int __meminit init_per_zone_wmark_min(void);
1699 extern void mem_init(void);
1700 extern void __init mmap_init(void);
1701 extern void show_mem(unsigned int flags);
1702 extern long si_mem_available(void);
1703 extern void si_meminfo(struct sysinfo * val);
1704 extern void si_meminfo_node(struct sysinfo *val, int nid);
1705 
1706 extern __printf(3, 4)
1707 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1708 
1709 extern void setup_per_cpu_pageset(void);
1710 
1711 extern void zone_pcp_update(struct zone *zone);
1712 extern void zone_pcp_reset(struct zone *zone);
1713 
1714 /* page_alloc.c */
1715 extern int min_free_kbytes;
1716 
1717 /* nommu.c */
1718 extern atomic_long_t mmap_pages_allocated;
1719 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1720 
1721 /* interval_tree.c */
1722 void vma_interval_tree_insert(struct vm_area_struct *node,
1723                               struct rb_root *root);
1724 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1725                                     struct vm_area_struct *prev,
1726                                     struct rb_root *root);
1727 void vma_interval_tree_remove(struct vm_area_struct *node,
1728                               struct rb_root *root);
1729 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1730                                 unsigned long start, unsigned long last);
1731 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1732                                 unsigned long start, unsigned long last);
1733 
1734 #define vma_interval_tree_foreach(vma, root, start, last)               \
1735         for (vma = vma_interval_tree_iter_first(root, start, last);     \
1736              vma; vma = vma_interval_tree_iter_next(vma, start, last))
1737 
1738 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1739                                    struct rb_root *root);
1740 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1741                                    struct rb_root *root);
1742 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1743         struct rb_root *root, unsigned long start, unsigned long last);
1744 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1745         struct anon_vma_chain *node, unsigned long start, unsigned long last);
1746 #ifdef CONFIG_DEBUG_VM_RB
1747 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1748 #endif
1749 
1750 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1751         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1752              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1753 
1754 /* mmap.c */
1755 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1756 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1757         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1758 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1759         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1760         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1761         struct mempolicy *);
1762 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1763 extern int split_vma(struct mm_struct *,
1764         struct vm_area_struct *, unsigned long addr, int new_below);
1765 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1766 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1767         struct rb_node **, struct rb_node *);
1768 extern void unlink_file_vma(struct vm_area_struct *);
1769 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1770         unsigned long addr, unsigned long len, pgoff_t pgoff,
1771         bool *need_rmap_locks);
1772 extern void exit_mmap(struct mm_struct *);
1773 
1774 extern int mm_take_all_locks(struct mm_struct *mm);
1775 extern void mm_drop_all_locks(struct mm_struct *mm);
1776 
1777 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1778 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1779 
1780 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1781 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1782                                    unsigned long addr, unsigned long len,
1783                                    unsigned long flags,
1784                                    const struct vm_special_mapping *spec);
1785 /* This is an obsolete alternative to _install_special_mapping. */
1786 extern int install_special_mapping(struct mm_struct *mm,
1787                                    unsigned long addr, unsigned long len,
1788                                    unsigned long flags, struct page **pages);
1789 
1790 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1791 
1792 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1793         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1794 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1795         unsigned long len, unsigned long prot, unsigned long flags,
1796         unsigned long pgoff, unsigned long *populate);
1797 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1798 
1799 #ifdef CONFIG_MMU
1800 extern int __mm_populate(unsigned long addr, unsigned long len,
1801                          int ignore_errors);
1802 static inline void mm_populate(unsigned long addr, unsigned long len)
1803 {
1804         /* Ignore errors */
1805         (void) __mm_populate(addr, len, 1);
1806 }
1807 #else
1808 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1809 #endif
1810 
1811 /* These take the mm semaphore themselves */
1812 extern unsigned long vm_brk(unsigned long, unsigned long);
1813 extern int vm_munmap(unsigned long, size_t);
1814 extern unsigned long vm_mmap(struct file *, unsigned long,
1815         unsigned long, unsigned long,
1816         unsigned long, unsigned long);
1817 
1818 struct vm_unmapped_area_info {
1819 #define VM_UNMAPPED_AREA_TOPDOWN 1
1820         unsigned long flags;
1821         unsigned long length;
1822         unsigned long low_limit;
1823         unsigned long high_limit;
1824         unsigned long align_mask;
1825         unsigned long align_offset;
1826 };
1827 
1828 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1829 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1830 
1831 /*
1832  * Search for an unmapped address range.
1833  *
1834  * We are looking for a range that:
1835  * - does not intersect with any VMA;
1836  * - is contained within the [low_limit, high_limit) interval;
1837  * - is at least the desired size.
1838  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1839  */
1840 static inline unsigned long
1841 vm_unmapped_area(struct vm_unmapped_area_info *info)
1842 {
1843         if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1844                 return unmapped_area(info);
1845         else
1846                 return unmapped_area_topdown(info);
1847 }
1848 
1849 /* truncate.c */
1850 extern void truncate_inode_pages(struct address_space *, loff_t);
1851 extern void truncate_inode_pages_range(struct address_space *,
1852                                        loff_t lstart, loff_t lend);
1853 extern void truncate_inode_pages_final(struct address_space *);
1854 
1855 /* generic vm_area_ops exported for stackable file systems */
1856 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1857 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1858 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1859 
1860 /* mm/page-writeback.c */
1861 int write_one_page(struct page *page, int wait);
1862 void task_dirty_inc(struct task_struct *tsk);
1863 
1864 /* readahead.c */
1865 #define VM_MAX_READAHEAD        128     /* kbytes */
1866 #define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1867 
1868 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1869                         pgoff_t offset, unsigned long nr_to_read);
1870 
1871 void page_cache_sync_readahead(struct address_space *mapping,
1872                                struct file_ra_state *ra,
1873                                struct file *filp,
1874                                pgoff_t offset,
1875                                unsigned long size);
1876 
1877 void page_cache_async_readahead(struct address_space *mapping,
1878                                 struct file_ra_state *ra,
1879                                 struct file *filp,
1880                                 struct page *pg,
1881                                 pgoff_t offset,
1882                                 unsigned long size);
1883 
1884 unsigned long max_sane_readahead(unsigned long nr);
1885 
1886 extern unsigned long stack_guard_gap;
1887 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1888 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1889 
1890 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1891 extern int expand_downwards(struct vm_area_struct *vma,
1892                 unsigned long address);
1893 #if VM_GROWSUP
1894 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1895 #else
1896   #define expand_upwards(vma, address) (0)
1897 #endif
1898 
1899 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1900 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1901 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1902                                              struct vm_area_struct **pprev);
1903 
1904 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1905    NULL if none.  Assume start_addr < end_addr. */
1906 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1907 {
1908         struct vm_area_struct * vma = find_vma(mm,start_addr);
1909 
1910         if (vma && end_addr <= vma->vm_start)
1911                 vma = NULL;
1912         return vma;
1913 }
1914 
1915 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
1916 {
1917         unsigned long vm_start = vma->vm_start;
1918 
1919         if (vma->vm_flags & VM_GROWSDOWN) {
1920                 vm_start -= stack_guard_gap;
1921                 if (vm_start > vma->vm_start)
1922                         vm_start = 0;
1923         }
1924         return vm_start;
1925 }
1926 
1927 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
1928 {
1929         unsigned long vm_end = vma->vm_end;
1930 
1931         if (vma->vm_flags & VM_GROWSUP) {
1932                 vm_end += stack_guard_gap;
1933                 if (vm_end < vma->vm_end)
1934                         vm_end = -PAGE_SIZE;
1935         }
1936         return vm_end;
1937 }
1938 
1939 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1940 {
1941         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1942 }
1943 
1944 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1945 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1946                                 unsigned long vm_start, unsigned long vm_end)
1947 {
1948         struct vm_area_struct *vma = find_vma(mm, vm_start);
1949 
1950         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1951                 vma = NULL;
1952 
1953         return vma;
1954 }
1955 
1956 static inline bool range_in_vma(struct vm_area_struct *vma,
1957                                 unsigned long start, unsigned long end)
1958 {
1959         return (vma && vma->vm_start <= start && end <= vma->vm_end);
1960 }
1961 
1962 #ifdef CONFIG_MMU
1963 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1964 #else
1965 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1966 {
1967         return __pgprot(0);
1968 }
1969 #endif
1970 
1971 #ifdef CONFIG_NUMA_BALANCING
1972 unsigned long change_prot_numa(struct vm_area_struct *vma,
1973                         unsigned long start, unsigned long end);
1974 #endif
1975 
1976 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1977 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1978                         unsigned long pfn, unsigned long size, pgprot_t);
1979 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1980 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1981                         unsigned long pfn);
1982 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
1983                         unsigned long pfn, pgprot_t pgprot);
1984 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1985                         unsigned long pfn);
1986 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1987 
1988 
1989 struct page *follow_page_mask(struct vm_area_struct *vma,
1990                               unsigned long address, unsigned int foll_flags,
1991                               unsigned int *page_mask);
1992 
1993 static inline struct page *follow_page(struct vm_area_struct *vma,
1994                 unsigned long address, unsigned int foll_flags)
1995 {
1996         unsigned int unused_page_mask;
1997         return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1998 }
1999 
2000 #define FOLL_WRITE      0x01    /* check pte is writable */
2001 #define FOLL_TOUCH      0x02    /* mark page accessed */
2002 #define FOLL_GET        0x04    /* do get_page on page */
2003 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2004 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2005 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2006                                  * and return without waiting upon it */
2007 #define FOLL_MLOCK      0x40    /* mark page as mlocked */
2008 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2009 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2010 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2011 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2012 #define FOLL_COW        0x4000  /* internal GUP flag */
2013 
2014 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2015                         void *data);
2016 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2017                                unsigned long size, pte_fn_t fn, void *data);
2018 
2019 #ifdef CONFIG_PROC_FS
2020 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2021 #else
2022 static inline void vm_stat_account(struct mm_struct *mm,
2023                         unsigned long flags, struct file *file, long pages)
2024 {
2025         mm->total_vm += pages;
2026 }
2027 #endif /* CONFIG_PROC_FS */
2028 
2029 #ifdef CONFIG_DEBUG_PAGEALLOC
2030 extern void kernel_map_pages(struct page *page, int numpages, int enable);
2031 #ifdef CONFIG_HIBERNATION
2032 extern bool kernel_page_present(struct page *page);
2033 #endif /* CONFIG_HIBERNATION */
2034 #else
2035 static inline void
2036 kernel_map_pages(struct page *page, int numpages, int enable) {}
2037 #ifdef CONFIG_HIBERNATION
2038 static inline bool kernel_page_present(struct page *page) { return true; }
2039 #endif /* CONFIG_HIBERNATION */
2040 #endif
2041 
2042 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2043 #ifdef  __HAVE_ARCH_GATE_AREA
2044 int in_gate_area_no_mm(unsigned long addr);
2045 int in_gate_area(struct mm_struct *mm, unsigned long addr);
2046 #else
2047 int in_gate_area_no_mm(unsigned long addr);
2048 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
2049 #endif  /* __HAVE_ARCH_GATE_AREA */
2050 
2051 #ifdef CONFIG_SYSCTL
2052 extern int sysctl_drop_caches;
2053 int drop_caches_sysctl_handler(struct ctl_table *, int,
2054                                         void __user *, size_t *, loff_t *);
2055 #endif
2056 
2057 unsigned long shrink_slab(struct shrink_control *shrink,
2058                           unsigned long nr_pages_scanned,
2059                           unsigned long lru_pages);
2060 
2061 #ifndef CONFIG_MMU
2062 #define randomize_va_space 0
2063 #else
2064 extern int randomize_va_space;
2065 #endif
2066 
2067 const char * arch_vma_name(struct vm_area_struct *vma);
2068 void print_vma_addr(char *prefix, unsigned long rip);
2069 
2070 void sparse_mem_maps_populate_node(struct page **map_map,
2071                                    unsigned long pnum_begin,
2072                                    unsigned long pnum_end,
2073                                    unsigned long map_count,
2074                                    int nodeid);
2075 
2076 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2077 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2078 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2079 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2080 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2081 void *vmemmap_alloc_block(unsigned long size, int node);
2082 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2083 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2084 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2085                                int node);
2086 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2087 void vmemmap_populate_print_last(void);
2088 #ifdef CONFIG_MEMORY_HOTPLUG
2089 void vmemmap_free(unsigned long start, unsigned long end);
2090 #endif
2091 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2092                                   unsigned long size);
2093 
2094 enum mf_flags {
2095         MF_COUNT_INCREASED = 1 << 0,
2096         MF_ACTION_REQUIRED = 1 << 1,
2097         MF_MUST_KILL = 1 << 2,
2098         MF_SOFT_OFFLINE = 1 << 3,
2099 };
2100 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2101 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2102 extern int unpoison_memory(unsigned long pfn);
2103 extern int sysctl_memory_failure_early_kill;
2104 extern int sysctl_memory_failure_recovery;
2105 extern void shake_page(struct page *p, int access);
2106 extern atomic_long_t num_poisoned_pages;
2107 extern int soft_offline_page(struct page *page, int flags);
2108 
2109 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2110 extern void clear_huge_page(struct page *page,
2111                             unsigned long addr,
2112                             unsigned int pages_per_huge_page);
2113 extern void copy_user_huge_page(struct page *dst, struct page *src,
2114                                 unsigned long addr, struct vm_area_struct *vma,
2115                                 unsigned int pages_per_huge_page);
2116 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2117 
2118 #ifdef CONFIG_DEBUG_PAGEALLOC
2119 extern unsigned int _debug_guardpage_minorder;
2120 
2121 static inline unsigned int debug_guardpage_minorder(void)
2122 {
2123         return _debug_guardpage_minorder;
2124 }
2125 
2126 static inline bool page_is_guard(struct page *page)
2127 {
2128         return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2129 }
2130 #else
2131 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2132 static inline bool page_is_guard(struct page *page) { return false; }
2133 #endif /* CONFIG_DEBUG_PAGEALLOC */
2134 
2135 #if MAX_NUMNODES > 1
2136 void __init setup_nr_node_ids(void);
2137 #else
2138 static inline void setup_nr_node_ids(void) {}
2139 #endif
2140 
2141 #endif /* __KERNEL__ */
2142 #endif /* _LINUX_MM_H */
2143 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp