Version:  2.0.40 2.2.26 2.4.37 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19

Linux/include/linux/mm.h

  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 static inline struct page *compound_head(struct page *page)
450 {
451         if (unlikely(PageTail(page)))
452                 return compound_head_by_tail(page);
453         return page;
454 }
455 
456 /*
457  * The atomic page->_mapcount, starts from -1: so that transitions
458  * both from it and to it can be tracked, using atomic_inc_and_test
459  * and atomic_add_negative(-1).
460  */
461 static inline void page_mapcount_reset(struct page *page)
462 {
463         atomic_set(&(page)->_mapcount, -1);
464 }
465 
466 static inline int page_mapcount(struct page *page)
467 {
468         return atomic_read(&(page)->_mapcount) + 1;
469 }
470 
471 static inline int page_count(struct page *page)
472 {
473         return atomic_read(&compound_head(page)->_count);
474 }
475 
476 #ifdef CONFIG_HUGETLB_PAGE
477 extern int PageHeadHuge(struct page *page_head);
478 #else /* CONFIG_HUGETLB_PAGE */
479 static inline int PageHeadHuge(struct page *page_head)
480 {
481         return 0;
482 }
483 #endif /* CONFIG_HUGETLB_PAGE */
484 
485 static inline bool __compound_tail_refcounted(struct page *page)
486 {
487         return !PageSlab(page) && !PageHeadHuge(page);
488 }
489 
490 /*
491  * This takes a head page as parameter and tells if the
492  * tail page reference counting can be skipped.
493  *
494  * For this to be safe, PageSlab and PageHeadHuge must remain true on
495  * any given page where they return true here, until all tail pins
496  * have been released.
497  */
498 static inline bool compound_tail_refcounted(struct page *page)
499 {
500         VM_BUG_ON_PAGE(!PageHead(page), page);
501         return __compound_tail_refcounted(page);
502 }
503 
504 static inline void get_huge_page_tail(struct page *page)
505 {
506         /*
507          * __split_huge_page_refcount() cannot run from under us.
508          */
509         VM_BUG_ON_PAGE(!PageTail(page), page);
510         VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
511         VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
512         if (compound_tail_refcounted(page->first_page))
513                 atomic_inc(&page->_mapcount);
514 }
515 
516 extern bool __get_page_tail(struct page *page);
517 
518 static inline void get_page(struct page *page)
519 {
520         if (unlikely(PageTail(page)))
521                 if (likely(__get_page_tail(page)))
522                         return;
523         /*
524          * Getting a normal page or the head of a compound page
525          * requires to already have an elevated page->_count.
526          */
527         VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
528         atomic_inc(&page->_count);
529 }
530 
531 static inline struct page *virt_to_head_page(const void *x)
532 {
533         struct page *page = virt_to_page(x);
534         return compound_head(page);
535 }
536 
537 /*
538  * Setup the page count before being freed into the page allocator for
539  * the first time (boot or memory hotplug)
540  */
541 static inline void init_page_count(struct page *page)
542 {
543         atomic_set(&page->_count, 1);
544 }
545 
546 /*
547  * PageBuddy() indicate that the page is free and in the buddy system
548  * (see mm/page_alloc.c).
549  *
550  * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
551  * -2 so that an underflow of the page_mapcount() won't be mistaken
552  * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
553  * efficiently by most CPU architectures.
554  */
555 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
556 
557 static inline int PageBuddy(struct page *page)
558 {
559         return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
560 }
561 
562 static inline void __SetPageBuddy(struct page *page)
563 {
564         VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
565         atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
566 }
567 
568 static inline void __ClearPageBuddy(struct page *page)
569 {
570         VM_BUG_ON_PAGE(!PageBuddy(page), page);
571         atomic_set(&page->_mapcount, -1);
572 }
573 
574 #define PAGE_BALLOON_MAPCOUNT_VALUE (-256)
575 
576 static inline int PageBalloon(struct page *page)
577 {
578         return atomic_read(&page->_mapcount) == PAGE_BALLOON_MAPCOUNT_VALUE;
579 }
580 
581 static inline void __SetPageBalloon(struct page *page)
582 {
583         VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
584         atomic_set(&page->_mapcount, PAGE_BALLOON_MAPCOUNT_VALUE);
585 }
586 
587 static inline void __ClearPageBalloon(struct page *page)
588 {
589         VM_BUG_ON_PAGE(!PageBalloon(page), page);
590         atomic_set(&page->_mapcount, -1);
591 }
592 
593 void put_page(struct page *page);
594 void put_pages_list(struct list_head *pages);
595 
596 void split_page(struct page *page, unsigned int order);
597 int split_free_page(struct page *page);
598 
599 /*
600  * Compound pages have a destructor function.  Provide a
601  * prototype for that function and accessor functions.
602  * These are _only_ valid on the head of a PG_compound page.
603  */
604 typedef void compound_page_dtor(struct page *);
605 
606 static inline void set_compound_page_dtor(struct page *page,
607                                                 compound_page_dtor *dtor)
608 {
609         page[1].lru.next = (void *)dtor;
610 }
611 
612 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
613 {
614         return (compound_page_dtor *)page[1].lru.next;
615 }
616 
617 static inline int compound_order(struct page *page)
618 {
619         if (!PageHead(page))
620                 return 0;
621         return (unsigned long)page[1].lru.prev;
622 }
623 
624 static inline void set_compound_order(struct page *page, unsigned long order)
625 {
626         page[1].lru.prev = (void *)order;
627 }
628 
629 #ifdef CONFIG_MMU
630 /*
631  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
632  * servicing faults for write access.  In the normal case, do always want
633  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
634  * that do not have writing enabled, when used by access_process_vm.
635  */
636 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
637 {
638         if (likely(vma->vm_flags & VM_WRITE))
639                 pte = pte_mkwrite(pte);
640         return pte;
641 }
642 
643 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
644                 struct page *page, pte_t *pte, bool write, bool anon);
645 #endif
646 
647 /*
648  * Multiple processes may "see" the same page. E.g. for untouched
649  * mappings of /dev/null, all processes see the same page full of
650  * zeroes, and text pages of executables and shared libraries have
651  * only one copy in memory, at most, normally.
652  *
653  * For the non-reserved pages, page_count(page) denotes a reference count.
654  *   page_count() == 0 means the page is free. page->lru is then used for
655  *   freelist management in the buddy allocator.
656  *   page_count() > 0  means the page has been allocated.
657  *
658  * Pages are allocated by the slab allocator in order to provide memory
659  * to kmalloc and kmem_cache_alloc. In this case, the management of the
660  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
661  * unless a particular usage is carefully commented. (the responsibility of
662  * freeing the kmalloc memory is the caller's, of course).
663  *
664  * A page may be used by anyone else who does a __get_free_page().
665  * In this case, page_count still tracks the references, and should only
666  * be used through the normal accessor functions. The top bits of page->flags
667  * and page->virtual store page management information, but all other fields
668  * are unused and could be used privately, carefully. The management of this
669  * page is the responsibility of the one who allocated it, and those who have
670  * subsequently been given references to it.
671  *
672  * The other pages (we may call them "pagecache pages") are completely
673  * managed by the Linux memory manager: I/O, buffers, swapping etc.
674  * The following discussion applies only to them.
675  *
676  * A pagecache page contains an opaque `private' member, which belongs to the
677  * page's address_space. Usually, this is the address of a circular list of
678  * the page's disk buffers. PG_private must be set to tell the VM to call
679  * into the filesystem to release these pages.
680  *
681  * A page may belong to an inode's memory mapping. In this case, page->mapping
682  * is the pointer to the inode, and page->index is the file offset of the page,
683  * in units of PAGE_CACHE_SIZE.
684  *
685  * If pagecache pages are not associated with an inode, they are said to be
686  * anonymous pages. These may become associated with the swapcache, and in that
687  * case PG_swapcache is set, and page->private is an offset into the swapcache.
688  *
689  * In either case (swapcache or inode backed), the pagecache itself holds one
690  * reference to the page. Setting PG_private should also increment the
691  * refcount. The each user mapping also has a reference to the page.
692  *
693  * The pagecache pages are stored in a per-mapping radix tree, which is
694  * rooted at mapping->page_tree, and indexed by offset.
695  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
696  * lists, we instead now tag pages as dirty/writeback in the radix tree.
697  *
698  * All pagecache pages may be subject to I/O:
699  * - inode pages may need to be read from disk,
700  * - inode pages which have been modified and are MAP_SHARED may need
701  *   to be written back to the inode on disk,
702  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
703  *   modified may need to be swapped out to swap space and (later) to be read
704  *   back into memory.
705  */
706 
707 /*
708  * The zone field is never updated after free_area_init_core()
709  * sets it, so none of the operations on it need to be atomic.
710  */
711 
712 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
713 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
714 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
715 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
716 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
717 
718 /*
719  * Define the bit shifts to access each section.  For non-existent
720  * sections we define the shift as 0; that plus a 0 mask ensures
721  * the compiler will optimise away reference to them.
722  */
723 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
724 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
725 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
726 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
727 
728 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
729 #ifdef NODE_NOT_IN_PAGE_FLAGS
730 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
731 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
732                                                 SECTIONS_PGOFF : ZONES_PGOFF)
733 #else
734 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
735 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
736                                                 NODES_PGOFF : ZONES_PGOFF)
737 #endif
738 
739 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
740 
741 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
742 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
743 #endif
744 
745 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
746 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
747 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
748 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
749 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
750 
751 static inline enum zone_type page_zonenum(const struct page *page)
752 {
753         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
754 }
755 
756 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
757 #define SECTION_IN_PAGE_FLAGS
758 #endif
759 
760 /*
761  * The identification function is mainly used by the buddy allocator for
762  * determining if two pages could be buddies. We are not really identifying
763  * the zone since we could be using the section number id if we do not have
764  * node id available in page flags.
765  * We only guarantee that it will return the same value for two combinable
766  * pages in a zone.
767  */
768 static inline int page_zone_id(struct page *page)
769 {
770         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
771 }
772 
773 static inline int zone_to_nid(struct zone *zone)
774 {
775 #ifdef CONFIG_NUMA
776         return zone->node;
777 #else
778         return 0;
779 #endif
780 }
781 
782 #ifdef NODE_NOT_IN_PAGE_FLAGS
783 extern int page_to_nid(const struct page *page);
784 #else
785 static inline int page_to_nid(const struct page *page)
786 {
787         return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
788 }
789 #endif
790 
791 #ifdef CONFIG_NUMA_BALANCING
792 static inline int cpu_pid_to_cpupid(int cpu, int pid)
793 {
794         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
795 }
796 
797 static inline int cpupid_to_pid(int cpupid)
798 {
799         return cpupid & LAST__PID_MASK;
800 }
801 
802 static inline int cpupid_to_cpu(int cpupid)
803 {
804         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
805 }
806 
807 static inline int cpupid_to_nid(int cpupid)
808 {
809         return cpu_to_node(cpupid_to_cpu(cpupid));
810 }
811 
812 static inline bool cpupid_pid_unset(int cpupid)
813 {
814         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
815 }
816 
817 static inline bool cpupid_cpu_unset(int cpupid)
818 {
819         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
820 }
821 
822 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
823 {
824         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
825 }
826 
827 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
828 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
829 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
830 {
831         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
832 }
833 
834 static inline int page_cpupid_last(struct page *page)
835 {
836         return page->_last_cpupid;
837 }
838 static inline void page_cpupid_reset_last(struct page *page)
839 {
840         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
841 }
842 #else
843 static inline int page_cpupid_last(struct page *page)
844 {
845         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
846 }
847 
848 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
849 
850 static inline void page_cpupid_reset_last(struct page *page)
851 {
852         int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
853 
854         page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
855         page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
856 }
857 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
858 #else /* !CONFIG_NUMA_BALANCING */
859 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
860 {
861         return page_to_nid(page); /* XXX */
862 }
863 
864 static inline int page_cpupid_last(struct page *page)
865 {
866         return page_to_nid(page); /* XXX */
867 }
868 
869 static inline int cpupid_to_nid(int cpupid)
870 {
871         return -1;
872 }
873 
874 static inline int cpupid_to_pid(int cpupid)
875 {
876         return -1;
877 }
878 
879 static inline int cpupid_to_cpu(int cpupid)
880 {
881         return -1;
882 }
883 
884 static inline int cpu_pid_to_cpupid(int nid, int pid)
885 {
886         return -1;
887 }
888 
889 static inline bool cpupid_pid_unset(int cpupid)
890 {
891         return 1;
892 }
893 
894 static inline void page_cpupid_reset_last(struct page *page)
895 {
896 }
897 
898 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
899 {
900         return false;
901 }
902 #endif /* CONFIG_NUMA_BALANCING */
903 
904 static inline struct zone *page_zone(const struct page *page)
905 {
906         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
907 }
908 
909 #ifdef SECTION_IN_PAGE_FLAGS
910 static inline void set_page_section(struct page *page, unsigned long section)
911 {
912         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
913         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
914 }
915 
916 static inline unsigned long page_to_section(const struct page *page)
917 {
918         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
919 }
920 #endif
921 
922 static inline void set_page_zone(struct page *page, enum zone_type zone)
923 {
924         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
925         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
926 }
927 
928 static inline void set_page_node(struct page *page, unsigned long node)
929 {
930         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
931         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
932 }
933 
934 static inline void set_page_links(struct page *page, enum zone_type zone,
935         unsigned long node, unsigned long pfn)
936 {
937         set_page_zone(page, zone);
938         set_page_node(page, node);
939 #ifdef SECTION_IN_PAGE_FLAGS
940         set_page_section(page, pfn_to_section_nr(pfn));
941 #endif
942 }
943 
944 /*
945  * Some inline functions in vmstat.h depend on page_zone()
946  */
947 #include <linux/vmstat.h>
948 
949 static __always_inline void *lowmem_page_address(const struct page *page)
950 {
951         return __va(PFN_PHYS(page_to_pfn(page)));
952 }
953 
954 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
955 #define HASHED_PAGE_VIRTUAL
956 #endif
957 
958 #if defined(WANT_PAGE_VIRTUAL)
959 static inline void *page_address(const struct page *page)
960 {
961         return page->virtual;
962 }
963 static inline void set_page_address(struct page *page, void *address)
964 {
965         page->virtual = address;
966 }
967 #define page_address_init()  do { } while(0)
968 #endif
969 
970 #if defined(HASHED_PAGE_VIRTUAL)
971 void *page_address(const struct page *page);
972 void set_page_address(struct page *page, void *virtual);
973 void page_address_init(void);
974 #endif
975 
976 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
977 #define page_address(page) lowmem_page_address(page)
978 #define set_page_address(page, address)  do { } while(0)
979 #define page_address_init()  do { } while(0)
980 #endif
981 
982 /*
983  * On an anonymous page mapped into a user virtual memory area,
984  * page->mapping points to its anon_vma, not to a struct address_space;
985  * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
986  *
987  * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
988  * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
989  * and then page->mapping points, not to an anon_vma, but to a private
990  * structure which KSM associates with that merged page.  See ksm.h.
991  *
992  * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
993  *
994  * Please note that, confusingly, "page_mapping" refers to the inode
995  * address_space which maps the page from disk; whereas "page_mapped"
996  * refers to user virtual address space into which the page is mapped.
997  */
998 #define PAGE_MAPPING_ANON       1
999 #define PAGE_MAPPING_KSM        2
1000 #define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
1001 
1002 extern struct address_space *page_mapping(struct page *page);
1003 
1004 /* Neutral page->mapping pointer to address_space or anon_vma or other */
1005 static inline void *page_rmapping(struct page *page)
1006 {
1007         return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
1008 }
1009 
1010 extern struct address_space *__page_file_mapping(struct page *);
1011 
1012 static inline
1013 struct address_space *page_file_mapping(struct page *page)
1014 {
1015         if (unlikely(PageSwapCache(page)))
1016                 return __page_file_mapping(page);
1017 
1018         return page->mapping;
1019 }
1020 
1021 static inline int PageAnon(struct page *page)
1022 {
1023         return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
1024 }
1025 
1026 /*
1027  * Return the pagecache index of the passed page.  Regular pagecache pages
1028  * use ->index whereas swapcache pages use ->private
1029  */
1030 static inline pgoff_t page_index(struct page *page)
1031 {
1032         if (unlikely(PageSwapCache(page)))
1033                 return page_private(page);
1034         return page->index;
1035 }
1036 
1037 extern pgoff_t __page_file_index(struct page *page);
1038 
1039 /*
1040  * Return the file index of the page. Regular pagecache pages use ->index
1041  * whereas swapcache pages use swp_offset(->private)
1042  */
1043 static inline pgoff_t page_file_index(struct page *page)
1044 {
1045         if (unlikely(PageSwapCache(page)))
1046                 return __page_file_index(page);
1047 
1048         return page->index;
1049 }
1050 
1051 /*
1052  * Return true if this page is mapped into pagetables.
1053  */
1054 static inline int page_mapped(struct page *page)
1055 {
1056         return atomic_read(&(page)->_mapcount) >= 0;
1057 }
1058 
1059 /*
1060  * Different kinds of faults, as returned by handle_mm_fault().
1061  * Used to decide whether a process gets delivered SIGBUS or
1062  * just gets major/minor fault counters bumped up.
1063  */
1064 
1065 #define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
1066 
1067 #define VM_FAULT_OOM    0x0001
1068 #define VM_FAULT_SIGBUS 0x0002
1069 #define VM_FAULT_MAJOR  0x0004
1070 #define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
1071 #define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
1072 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1073 #define VM_FAULT_SIGSEGV 0x0040
1074 
1075 #define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
1076 #define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
1077 #define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
1078 #define VM_FAULT_FALLBACK 0x0800        /* huge page fault failed, fall back to small */
1079 
1080 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1081 
1082 #define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1083                          VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1084                          VM_FAULT_FALLBACK)
1085 
1086 /* Encode hstate index for a hwpoisoned large page */
1087 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1088 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1089 
1090 /*
1091  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1092  */
1093 extern void pagefault_out_of_memory(void);
1094 
1095 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1096 
1097 /*
1098  * Flags passed to show_mem() and show_free_areas() to suppress output in
1099  * various contexts.
1100  */
1101 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1102 
1103 extern void show_free_areas(unsigned int flags);
1104 extern bool skip_free_areas_node(unsigned int flags, int nid);
1105 
1106 int shmem_zero_setup(struct vm_area_struct *);
1107 #ifdef CONFIG_SHMEM
1108 bool shmem_mapping(struct address_space *mapping);
1109 #else
1110 static inline bool shmem_mapping(struct address_space *mapping)
1111 {
1112         return false;
1113 }
1114 #endif
1115 
1116 extern int can_do_mlock(void);
1117 extern int user_shm_lock(size_t, struct user_struct *);
1118 extern void user_shm_unlock(size_t, struct user_struct *);
1119 
1120 /*
1121  * Parameter block passed down to zap_pte_range in exceptional cases.
1122  */
1123 struct zap_details {
1124         struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
1125         struct address_space *check_mapping;    /* Check page->mapping if set */
1126         pgoff_t first_index;                    /* Lowest page->index to unmap */
1127         pgoff_t last_index;                     /* Highest page->index to unmap */
1128 };
1129 
1130 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1131                 pte_t pte);
1132 
1133 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1134                 unsigned long size);
1135 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1136                 unsigned long size, struct zap_details *);
1137 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1138                 unsigned long start, unsigned long end);
1139 
1140 /**
1141  * mm_walk - callbacks for walk_page_range
1142  * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1143  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1144  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1145  *             this handler is required to be able to handle
1146  *             pmd_trans_huge() pmds.  They may simply choose to
1147  *             split_huge_page() instead of handling it explicitly.
1148  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1149  * @pte_hole: if set, called for each hole at all levels
1150  * @hugetlb_entry: if set, called for each hugetlb entry
1151  *                 *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1152  *                            is used.
1153  *
1154  * (see walk_page_range for more details)
1155  */
1156 struct mm_walk {
1157         int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1158                          unsigned long next, struct mm_walk *walk);
1159         int (*pud_entry)(pud_t *pud, unsigned long addr,
1160                          unsigned long next, struct mm_walk *walk);
1161         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1162                          unsigned long next, struct mm_walk *walk);
1163         int (*pte_entry)(pte_t *pte, unsigned long addr,
1164                          unsigned long next, struct mm_walk *walk);
1165         int (*pte_hole)(unsigned long addr, unsigned long next,
1166                         struct mm_walk *walk);
1167         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1168                              unsigned long addr, unsigned long next,
1169                              struct mm_walk *walk);
1170         struct mm_struct *mm;
1171         void *private;
1172 };
1173 
1174 int walk_page_range(unsigned long addr, unsigned long end,
1175                 struct mm_walk *walk);
1176 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1177                 unsigned long end, unsigned long floor, unsigned long ceiling);
1178 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1179                         struct vm_area_struct *vma);
1180 void unmap_mapping_range(struct address_space *mapping,
1181                 loff_t const holebegin, loff_t const holelen, int even_cows);
1182 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1183         unsigned long *pfn);
1184 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1185                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1186 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1187                         void *buf, int len, int write);
1188 
1189 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1190                 loff_t const holebegin, loff_t const holelen)
1191 {
1192         unmap_mapping_range(mapping, holebegin, holelen, 0);
1193 }
1194 
1195 extern void truncate_pagecache(struct inode *inode, loff_t new);
1196 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1197 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1198 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1199 int truncate_inode_page(struct address_space *mapping, struct page *page);
1200 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1201 int invalidate_inode_page(struct page *page);
1202 
1203 #ifdef CONFIG_MMU
1204 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1205                         unsigned long address, unsigned int flags);
1206 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1207                             unsigned long address, unsigned int fault_flags);
1208 #else
1209 static inline int handle_mm_fault(struct mm_struct *mm,
1210                         struct vm_area_struct *vma, unsigned long address,
1211                         unsigned int flags)
1212 {
1213         /* should never happen if there's no MMU */
1214         BUG();
1215         return VM_FAULT_SIGBUS;
1216 }
1217 static inline int fixup_user_fault(struct task_struct *tsk,
1218                 struct mm_struct *mm, unsigned long address,
1219                 unsigned int fault_flags)
1220 {
1221         /* should never happen if there's no MMU */
1222         BUG();
1223         return -EFAULT;
1224 }
1225 #endif
1226 
1227 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1228 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1229                 void *buf, int len, int write);
1230 
1231 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1232                       unsigned long start, unsigned long nr_pages,
1233                       unsigned int foll_flags, struct page **pages,
1234                       struct vm_area_struct **vmas, int *nonblocking);
1235 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1236                     unsigned long start, unsigned long nr_pages,
1237                     int write, int force, struct page **pages,
1238                     struct vm_area_struct **vmas);
1239 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1240                         struct page **pages);
1241 struct kvec;
1242 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1243                         struct page **pages);
1244 int get_kernel_page(unsigned long start, int write, struct page **pages);
1245 struct page *get_dump_page(unsigned long addr);
1246 
1247 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1248 extern void do_invalidatepage(struct page *page, unsigned int offset,
1249                               unsigned int length);
1250 
1251 int __set_page_dirty_nobuffers(struct page *page);
1252 int __set_page_dirty_no_writeback(struct page *page);
1253 int redirty_page_for_writepage(struct writeback_control *wbc,
1254                                 struct page *page);
1255 void account_page_dirtied(struct page *page, struct address_space *mapping);
1256 int set_page_dirty(struct page *page);
1257 int set_page_dirty_lock(struct page *page);
1258 int clear_page_dirty_for_io(struct page *page);
1259 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1260 
1261 /* Is the vma a continuation of the stack vma above it? */
1262 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1263 {
1264         return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1265 }
1266 
1267 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1268                                              unsigned long addr)
1269 {
1270         return (vma->vm_flags & VM_GROWSDOWN) &&
1271                 (vma->vm_start == addr) &&
1272                 !vma_growsdown(vma->vm_prev, addr);
1273 }
1274 
1275 /* Is the vma a continuation of the stack vma below it? */
1276 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1277 {
1278         return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1279 }
1280 
1281 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1282                                            unsigned long addr)
1283 {
1284         return (vma->vm_flags & VM_GROWSUP) &&
1285                 (vma->vm_end == addr) &&
1286                 !vma_growsup(vma->vm_next, addr);
1287 }
1288 
1289 extern struct task_struct *task_of_stack(struct task_struct *task,
1290                                 struct vm_area_struct *vma, bool in_group);
1291 
1292 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1293                 unsigned long old_addr, struct vm_area_struct *new_vma,
1294                 unsigned long new_addr, unsigned long len,
1295                 bool need_rmap_locks);
1296 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1297                               unsigned long end, pgprot_t newprot,
1298                               int dirty_accountable, int prot_numa);
1299 extern int mprotect_fixup(struct vm_area_struct *vma,
1300                           struct vm_area_struct **pprev, unsigned long start,
1301                           unsigned long end, unsigned long newflags);
1302 
1303 /*
1304  * doesn't attempt to fault and will return short.
1305  */
1306 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1307                           struct page **pages);
1308 /*
1309  * per-process(per-mm_struct) statistics.
1310  */
1311 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1312 {
1313         long val = atomic_long_read(&mm->rss_stat.count[member]);
1314 
1315 #ifdef SPLIT_RSS_COUNTING
1316         /*
1317          * counter is updated in asynchronous manner and may go to minus.
1318          * But it's never be expected number for users.
1319          */
1320         if (val < 0)
1321                 val = 0;
1322 #endif
1323         return (unsigned long)val;
1324 }
1325 
1326 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1327 {
1328         atomic_long_add(value, &mm->rss_stat.count[member]);
1329 }
1330 
1331 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1332 {
1333         atomic_long_inc(&mm->rss_stat.count[member]);
1334 }
1335 
1336 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1337 {
1338         atomic_long_dec(&mm->rss_stat.count[member]);
1339 }
1340 
1341 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1342 {
1343         return get_mm_counter(mm, MM_FILEPAGES) +
1344                 get_mm_counter(mm, MM_ANONPAGES);
1345 }
1346 
1347 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1348 {
1349         return max(mm->hiwater_rss, get_mm_rss(mm));
1350 }
1351 
1352 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1353 {
1354         return max(mm->hiwater_vm, mm->total_vm);
1355 }
1356 
1357 static inline void update_hiwater_rss(struct mm_struct *mm)
1358 {
1359         unsigned long _rss = get_mm_rss(mm);
1360 
1361         if ((mm)->hiwater_rss < _rss)
1362                 (mm)->hiwater_rss = _rss;
1363 }
1364 
1365 static inline void update_hiwater_vm(struct mm_struct *mm)
1366 {
1367         if (mm->hiwater_vm < mm->total_vm)
1368                 mm->hiwater_vm = mm->total_vm;
1369 }
1370 
1371 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1372                                          struct mm_struct *mm)
1373 {
1374         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1375 
1376         if (*maxrss < hiwater_rss)
1377                 *maxrss = hiwater_rss;
1378 }
1379 
1380 #if defined(SPLIT_RSS_COUNTING)
1381 void sync_mm_rss(struct mm_struct *mm);
1382 #else
1383 static inline void sync_mm_rss(struct mm_struct *mm)
1384 {
1385 }
1386 #endif
1387 
1388 int vma_wants_writenotify(struct vm_area_struct *vma);
1389 
1390 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1391                                spinlock_t **ptl);
1392 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1393                                     spinlock_t **ptl)
1394 {
1395         pte_t *ptep;
1396         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1397         return ptep;
1398 }
1399 
1400 #ifdef __PAGETABLE_PUD_FOLDED
1401 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1402                                                 unsigned long address)
1403 {
1404         return 0;
1405 }
1406 #else
1407 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1408 #endif
1409 
1410 #ifdef __PAGETABLE_PMD_FOLDED
1411 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1412                                                 unsigned long address)
1413 {
1414         return 0;
1415 }
1416 #else
1417 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1418 #endif
1419 
1420 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1421                 pmd_t *pmd, unsigned long address);
1422 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1423 
1424 /*
1425  * The following ifdef needed to get the 4level-fixup.h header to work.
1426  * Remove it when 4level-fixup.h has been removed.
1427  */
1428 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1429 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1430 {
1431         return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1432                 NULL: pud_offset(pgd, address);
1433 }
1434 
1435 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1436 {
1437         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1438                 NULL: pmd_offset(pud, address);
1439 }
1440 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1441 
1442 #if USE_SPLIT_PTE_PTLOCKS
1443 #if ALLOC_SPLIT_PTLOCKS
1444 void __init ptlock_cache_init(void);
1445 extern bool ptlock_alloc(struct page *page);
1446 extern void ptlock_free(struct page *page);
1447 
1448 static inline spinlock_t *ptlock_ptr(struct page *page)
1449 {
1450         return page->ptl;
1451 }
1452 #else /* ALLOC_SPLIT_PTLOCKS */
1453 static inline void ptlock_cache_init(void)
1454 {
1455 }
1456 
1457 static inline bool ptlock_alloc(struct page *page)
1458 {
1459         return true;
1460 }
1461 
1462 static inline void ptlock_free(struct page *page)
1463 {
1464 }
1465 
1466 static inline spinlock_t *ptlock_ptr(struct page *page)
1467 {
1468         return &page->ptl;
1469 }
1470 #endif /* ALLOC_SPLIT_PTLOCKS */
1471 
1472 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1473 {
1474         return ptlock_ptr(pmd_page(*pmd));
1475 }
1476 
1477 static inline bool ptlock_init(struct page *page)
1478 {
1479         /*
1480          * prep_new_page() initialize page->private (and therefore page->ptl)
1481          * with 0. Make sure nobody took it in use in between.
1482          *
1483          * It can happen if arch try to use slab for page table allocation:
1484          * slab code uses page->slab_cache and page->first_page (for tail
1485          * pages), which share storage with page->ptl.
1486          */
1487         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1488         if (!ptlock_alloc(page))
1489                 return false;
1490         spin_lock_init(ptlock_ptr(page));
1491         return true;
1492 }
1493 
1494 /* Reset page->mapping so free_pages_check won't complain. */
1495 static inline void pte_lock_deinit(struct page *page)
1496 {
1497         page->mapping = NULL;
1498         ptlock_free(page);
1499 }
1500 
1501 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1502 /*
1503  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1504  */
1505 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1506 {
1507         return &mm->page_table_lock;
1508 }
1509 static inline void ptlock_cache_init(void) {}
1510 static inline bool ptlock_init(struct page *page) { return true; }
1511 static inline void pte_lock_deinit(struct page *page) {}
1512 #endif /* USE_SPLIT_PTE_PTLOCKS */
1513 
1514 static inline void pgtable_init(void)
1515 {
1516         ptlock_cache_init();
1517         pgtable_cache_init();
1518 }
1519 
1520 static inline bool pgtable_page_ctor(struct page *page)
1521 {
1522         inc_zone_page_state(page, NR_PAGETABLE);
1523         return ptlock_init(page);
1524 }
1525 
1526 static inline void pgtable_page_dtor(struct page *page)
1527 {
1528         pte_lock_deinit(page);
1529         dec_zone_page_state(page, NR_PAGETABLE);
1530 }
1531 
1532 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1533 ({                                                      \
1534         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1535         pte_t *__pte = pte_offset_map(pmd, address);    \
1536         *(ptlp) = __ptl;                                \
1537         spin_lock(__ptl);                               \
1538         __pte;                                          \
1539 })
1540 
1541 #define pte_unmap_unlock(pte, ptl)      do {            \
1542         spin_unlock(ptl);                               \
1543         pte_unmap(pte);                                 \
1544 } while (0)
1545 
1546 #define pte_alloc_map(mm, vma, pmd, address)                            \
1547         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1548                                                         pmd, address))? \
1549          NULL: pte_offset_map(pmd, address))
1550 
1551 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1552         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1553                                                         pmd, address))? \
1554                 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1555 
1556 #define pte_alloc_kernel(pmd, address)                  \
1557         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1558                 NULL: pte_offset_kernel(pmd, address))
1559 
1560 #if USE_SPLIT_PMD_PTLOCKS
1561 
1562 static struct page *pmd_to_page(pmd_t *pmd)
1563 {
1564         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1565         return virt_to_page((void *)((unsigned long) pmd & mask));
1566 }
1567 
1568 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1569 {
1570         return ptlock_ptr(pmd_to_page(pmd));
1571 }
1572 
1573 static inline bool pgtable_pmd_page_ctor(struct page *page)
1574 {
1575 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1576         page->pmd_huge_pte = NULL;
1577 #endif
1578         return ptlock_init(page);
1579 }
1580 
1581 static inline void pgtable_pmd_page_dtor(struct page *page)
1582 {
1583 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1584         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1585 #endif
1586         ptlock_free(page);
1587 }
1588 
1589 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1590 
1591 #else
1592 
1593 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1594 {
1595         return &mm->page_table_lock;
1596 }
1597 
1598 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1599 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1600 
1601 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1602 
1603 #endif
1604 
1605 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1606 {
1607         spinlock_t *ptl = pmd_lockptr(mm, pmd);
1608         spin_lock(ptl);
1609         return ptl;
1610 }
1611 
1612 extern void free_area_init(unsigned long * zones_size);
1613 extern void free_area_init_node(int nid, unsigned long * zones_size,
1614                 unsigned long zone_start_pfn, unsigned long *zholes_size);
1615 extern void free_initmem(void);
1616 
1617 /*
1618  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1619  * into the buddy system. The freed pages will be poisoned with pattern
1620  * "poison" if it's within range [0, UCHAR_MAX].
1621  * Return pages freed into the buddy system.
1622  */
1623 extern unsigned long free_reserved_area(void *start, void *end,
1624                                         int poison, char *s);
1625 
1626 #ifdef  CONFIG_HIGHMEM
1627 /*
1628  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1629  * and totalram_pages.
1630  */
1631 extern void free_highmem_page(struct page *page);
1632 #endif
1633 
1634 extern void adjust_managed_page_count(struct page *page, long count);
1635 extern void mem_init_print_info(const char *str);
1636 
1637 /* Free the reserved page into the buddy system, so it gets managed. */
1638 static inline void __free_reserved_page(struct page *page)
1639 {
1640         ClearPageReserved(page);
1641         init_page_count(page);
1642         __free_page(page);
1643 }
1644 
1645 static inline void free_reserved_page(struct page *page)
1646 {
1647         __free_reserved_page(page);
1648         adjust_managed_page_count(page, 1);
1649 }
1650 
1651 static inline void mark_page_reserved(struct page *page)
1652 {
1653         SetPageReserved(page);
1654         adjust_managed_page_count(page, -1);
1655 }
1656 
1657 /*
1658  * Default method to free all the __init memory into the buddy system.
1659  * The freed pages will be poisoned with pattern "poison" if it's within
1660  * range [0, UCHAR_MAX].
1661  * Return pages freed into the buddy system.
1662  */
1663 static inline unsigned long free_initmem_default(int poison)
1664 {
1665         extern char __init_begin[], __init_end[];
1666 
1667         return free_reserved_area(&__init_begin, &__init_end,
1668                                   poison, "unused kernel");
1669 }
1670 
1671 static inline unsigned long get_num_physpages(void)
1672 {
1673         int nid;
1674         unsigned long phys_pages = 0;
1675 
1676         for_each_online_node(nid)
1677                 phys_pages += node_present_pages(nid);
1678 
1679         return phys_pages;
1680 }
1681 
1682 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1683 /*
1684  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1685  * zones, allocate the backing mem_map and account for memory holes in a more
1686  * architecture independent manner. This is a substitute for creating the
1687  * zone_sizes[] and zholes_size[] arrays and passing them to
1688  * free_area_init_node()
1689  *
1690  * An architecture is expected to register range of page frames backed by
1691  * physical memory with memblock_add[_node]() before calling
1692  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1693  * usage, an architecture is expected to do something like
1694  *
1695  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1696  *                                                       max_highmem_pfn};
1697  * for_each_valid_physical_page_range()
1698  *      memblock_add_node(base, size, nid)
1699  * free_area_init_nodes(max_zone_pfns);
1700  *
1701  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1702  * registered physical page range.  Similarly
1703  * sparse_memory_present_with_active_regions() calls memory_present() for
1704  * each range when SPARSEMEM is enabled.
1705  *
1706  * See mm/page_alloc.c for more information on each function exposed by
1707  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1708  */
1709 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1710 unsigned long node_map_pfn_alignment(void);
1711 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1712                                                 unsigned long end_pfn);
1713 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1714                                                 unsigned long end_pfn);
1715 extern void get_pfn_range_for_nid(unsigned int nid,
1716                         unsigned long *start_pfn, unsigned long *end_pfn);
1717 extern unsigned long find_min_pfn_with_active_regions(void);
1718 extern void free_bootmem_with_active_regions(int nid,
1719                                                 unsigned long max_low_pfn);
1720 extern void sparse_memory_present_with_active_regions(int nid);
1721 
1722 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1723 
1724 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1725     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1726 static inline int __early_pfn_to_nid(unsigned long pfn)
1727 {
1728         return 0;
1729 }
1730 #else
1731 /* please see mm/page_alloc.c */
1732 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1733 /* there is a per-arch backend function. */
1734 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1735 #endif
1736 
1737 extern void set_dma_reserve(unsigned long new_dma_reserve);
1738 extern void memmap_init_zone(unsigned long, int, unsigned long,
1739                                 unsigned long, enum memmap_context);
1740 extern void setup_per_zone_wmarks(void);
1741 extern int __meminit init_per_zone_wmark_min(void);
1742 extern void mem_init(void);
1743 extern void __init mmap_init(void);
1744 extern void show_mem(unsigned int flags);
1745 extern void si_meminfo(struct sysinfo * val);
1746 extern void si_meminfo_node(struct sysinfo *val, int nid);
1747 
1748 extern __printf(3, 4)
1749 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1750 
1751 extern void setup_per_cpu_pageset(void);
1752 
1753 extern void zone_pcp_update(struct zone *zone);
1754 extern void zone_pcp_reset(struct zone *zone);
1755 
1756 /* page_alloc.c */
1757 extern int min_free_kbytes;
1758 
1759 /* nommu.c */
1760 extern atomic_long_t mmap_pages_allocated;
1761 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1762 
1763 /* interval_tree.c */
1764 void vma_interval_tree_insert(struct vm_area_struct *node,
1765                               struct rb_root *root);
1766 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1767                                     struct vm_area_struct *prev,
1768                                     struct rb_root *root);
1769 void vma_interval_tree_remove(struct vm_area_struct *node,
1770                               struct rb_root *root);
1771 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1772                                 unsigned long start, unsigned long last);
1773 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1774                                 unsigned long start, unsigned long last);
1775 
1776 #define vma_interval_tree_foreach(vma, root, start, last)               \
1777         for (vma = vma_interval_tree_iter_first(root, start, last);     \
1778              vma; vma = vma_interval_tree_iter_next(vma, start, last))
1779 
1780 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1781                                         struct list_head *list)
1782 {
1783         list_add_tail(&vma->shared.nonlinear, list);
1784 }
1785 
1786 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1787                                    struct rb_root *root);
1788 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1789                                    struct rb_root *root);
1790 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1791         struct rb_root *root, unsigned long start, unsigned long last);
1792 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1793         struct anon_vma_chain *node, unsigned long start, unsigned long last);
1794 #ifdef CONFIG_DEBUG_VM_RB
1795 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1796 #endif
1797 
1798 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1799         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1800              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1801 
1802 /* mmap.c */
1803 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1804 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1805         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1806 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1807         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1808         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1809         struct mempolicy *);
1810 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1811 extern int split_vma(struct mm_struct *,
1812         struct vm_area_struct *, unsigned long addr, int new_below);
1813 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1814 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1815         struct rb_node **, struct rb_node *);
1816 extern void unlink_file_vma(struct vm_area_struct *);
1817 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1818         unsigned long addr, unsigned long len, pgoff_t pgoff,
1819         bool *need_rmap_locks);
1820 extern void exit_mmap(struct mm_struct *);
1821 
1822 static inline int check_data_rlimit(unsigned long rlim,
1823                                     unsigned long new,
1824                                     unsigned long start,
1825                                     unsigned long end_data,
1826                                     unsigned long start_data)
1827 {
1828         if (rlim < RLIM_INFINITY) {
1829                 if (((new - start) + (end_data - start_data)) > rlim)
1830                         return -ENOSPC;
1831         }
1832 
1833         return 0;
1834 }
1835 
1836 extern int mm_take_all_locks(struct mm_struct *mm);
1837 extern void mm_drop_all_locks(struct mm_struct *mm);
1838 
1839 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1840 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1841 
1842 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1843 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1844                                    unsigned long addr, unsigned long len,
1845                                    unsigned long flags,
1846                                    const struct vm_special_mapping *spec);
1847 /* This is an obsolete alternative to _install_special_mapping. */
1848 extern int install_special_mapping(struct mm_struct *mm,
1849                                    unsigned long addr, unsigned long len,
1850                                    unsigned long flags, struct page **pages);
1851 
1852 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1853 
1854 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1855         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1856 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1857         unsigned long len, unsigned long prot, unsigned long flags,
1858         unsigned long pgoff, unsigned long *populate);
1859 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1860 
1861 #ifdef CONFIG_MMU
1862 extern int __mm_populate(unsigned long addr, unsigned long len,
1863                          int ignore_errors);
1864 static inline void mm_populate(unsigned long addr, unsigned long len)
1865 {
1866         /* Ignore errors */
1867         (void) __mm_populate(addr, len, 1);
1868 }
1869 #else
1870 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1871 #endif
1872 
1873 /* These take the mm semaphore themselves */
1874 extern unsigned long vm_brk(unsigned long, unsigned long);
1875 extern int vm_munmap(unsigned long, size_t);
1876 extern unsigned long vm_mmap(struct file *, unsigned long,
1877         unsigned long, unsigned long,
1878         unsigned long, unsigned long);
1879 
1880 struct vm_unmapped_area_info {
1881 #define VM_UNMAPPED_AREA_TOPDOWN 1
1882         unsigned long flags;
1883         unsigned long length;
1884         unsigned long low_limit;
1885         unsigned long high_limit;
1886         unsigned long align_mask;
1887         unsigned long align_offset;
1888 };
1889 
1890 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1891 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1892 
1893 /*
1894  * Search for an unmapped address range.
1895  *
1896  * We are looking for a range that:
1897  * - does not intersect with any VMA;
1898  * - is contained within the [low_limit, high_limit) interval;
1899  * - is at least the desired size.
1900  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1901  */
1902 static inline unsigned long
1903 vm_unmapped_area(struct vm_unmapped_area_info *info)
1904 {
1905         if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1906                 return unmapped_area(info);
1907         else
1908                 return unmapped_area_topdown(info);
1909 }
1910 
1911 /* truncate.c */
1912 extern void truncate_inode_pages(struct address_space *, loff_t);
1913 extern void truncate_inode_pages_range(struct address_space *,
1914                                        loff_t lstart, loff_t lend);
1915 extern void truncate_inode_pages_final(struct address_space *);
1916 
1917 /* generic vm_area_ops exported for stackable file systems */
1918 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1919 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1920 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1921 
1922 /* mm/page-writeback.c */
1923 int write_one_page(struct page *page, int wait);
1924 void task_dirty_inc(struct task_struct *tsk);
1925 
1926 /* readahead.c */
1927 #define VM_MAX_READAHEAD        128     /* kbytes */
1928 #define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1929 
1930 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1931                         pgoff_t offset, unsigned long nr_to_read);
1932 
1933 void page_cache_sync_readahead(struct address_space *mapping,
1934                                struct file_ra_state *ra,
1935                                struct file *filp,
1936                                pgoff_t offset,
1937                                unsigned long size);
1938 
1939 void page_cache_async_readahead(struct address_space *mapping,
1940                                 struct file_ra_state *ra,
1941                                 struct file *filp,
1942                                 struct page *pg,
1943                                 pgoff_t offset,
1944                                 unsigned long size);
1945 
1946 unsigned long max_sane_readahead(unsigned long nr);
1947 
1948 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1949 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1950 
1951 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1952 extern int expand_downwards(struct vm_area_struct *vma,
1953                 unsigned long address);
1954 #if VM_GROWSUP
1955 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1956 #else
1957   #define expand_upwards(vma, address) (0)
1958 #endif
1959 
1960 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1961 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1962 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1963                                              struct vm_area_struct **pprev);
1964 
1965 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1966    NULL if none.  Assume start_addr < end_addr. */
1967 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1968 {
1969         struct vm_area_struct * vma = find_vma(mm,start_addr);
1970 
1971         if (vma && end_addr <= vma->vm_start)
1972                 vma = NULL;
1973         return vma;
1974 }
1975 
1976 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1977 {
1978         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1979 }
1980 
1981 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1982 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1983                                 unsigned long vm_start, unsigned long vm_end)
1984 {
1985         struct vm_area_struct *vma = find_vma(mm, vm_start);
1986 
1987         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1988                 vma = NULL;
1989 
1990         return vma;
1991 }
1992 
1993 #ifdef CONFIG_MMU
1994 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1995 void vma_set_page_prot(struct vm_area_struct *vma);
1996 #else
1997 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1998 {
1999         return __pgprot(0);
2000 }
2001 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2002 {
2003         vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2004 }
2005 #endif
2006 
2007 #ifdef CONFIG_NUMA_BALANCING
2008 unsigned long change_prot_numa(struct vm_area_struct *vma,
2009                         unsigned long start, unsigned long end);
2010 #endif
2011 
2012 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2013 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2014                         unsigned long pfn, unsigned long size, pgprot_t);
2015 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2016 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2017                         unsigned long pfn);
2018 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2019                         unsigned long pfn);
2020 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2021 
2022 
2023 struct page *follow_page_mask(struct vm_area_struct *vma,
2024                               unsigned long address, unsigned int foll_flags,
2025                               unsigned int *page_mask);
2026 
2027 static inline struct page *follow_page(struct vm_area_struct *vma,
2028                 unsigned long address, unsigned int foll_flags)
2029 {
2030         unsigned int unused_page_mask;
2031         return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2032 }
2033 
2034 #define FOLL_WRITE      0x01    /* check pte is writable */
2035 #define FOLL_TOUCH      0x02    /* mark page accessed */
2036 #define FOLL_GET        0x04    /* do get_page on page */
2037 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2038 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2039 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2040                                  * and return without waiting upon it */
2041 #define FOLL_MLOCK      0x40    /* mark page as mlocked */
2042 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2043 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2044 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2045 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2046 #define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2047 
2048 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2049                         void *data);
2050 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2051                                unsigned long size, pte_fn_t fn, void *data);
2052 
2053 #ifdef CONFIG_PROC_FS
2054 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2055 #else
2056 static inline void vm_stat_account(struct mm_struct *mm,
2057                         unsigned long flags, struct file *file, long pages)
2058 {
2059         mm->total_vm += pages;
2060 }
2061 #endif /* CONFIG_PROC_FS */
2062 
2063 #ifdef CONFIG_DEBUG_PAGEALLOC
2064 extern bool _debug_pagealloc_enabled;
2065 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2066 
2067 static inline bool debug_pagealloc_enabled(void)
2068 {
2069         return _debug_pagealloc_enabled;
2070 }
2071 
2072 static inline void
2073 kernel_map_pages(struct page *page, int numpages, int enable)
2074 {
2075         if (!debug_pagealloc_enabled())
2076                 return;
2077 
2078         __kernel_map_pages(page, numpages, enable);
2079 }
2080 #ifdef CONFIG_HIBERNATION
2081 extern bool kernel_page_present(struct page *page);
2082 #endif /* CONFIG_HIBERNATION */
2083 #else
2084 static inline void
2085 kernel_map_pages(struct page *page, int numpages, int enable) {}
2086 #ifdef CONFIG_HIBERNATION
2087 static inline bool kernel_page_present(struct page *page) { return true; }
2088 #endif /* CONFIG_HIBERNATION */
2089 #endif
2090 
2091 #ifdef __HAVE_ARCH_GATE_AREA
2092 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2093 extern int in_gate_area_no_mm(unsigned long addr);
2094 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2095 #else
2096 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2097 {
2098         return NULL;
2099 }
2100 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2101 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2102 {
2103         return 0;
2104 }
2105 #endif  /* __HAVE_ARCH_GATE_AREA */
2106 
2107 #ifdef CONFIG_SYSCTL
2108 extern int sysctl_drop_caches;
2109 int drop_caches_sysctl_handler(struct ctl_table *, int,
2110                                         void __user *, size_t *, loff_t *);
2111 #endif
2112 
2113 unsigned long shrink_node_slabs(gfp_t gfp_mask, int nid,
2114                                 unsigned long nr_scanned,
2115                                 unsigned long nr_eligible);
2116 
2117 #ifndef CONFIG_MMU
2118 #define randomize_va_space 0
2119 #else
2120 extern int randomize_va_space;
2121 #endif
2122 
2123 const char * arch_vma_name(struct vm_area_struct *vma);
2124 void print_vma_addr(char *prefix, unsigned long rip);
2125 
2126 void sparse_mem_maps_populate_node(struct page **map_map,
2127                                    unsigned long pnum_begin,
2128                                    unsigned long pnum_end,
2129                                    unsigned long map_count,
2130                                    int nodeid);
2131 
2132 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2133 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2134 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2135 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2136 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2137 void *vmemmap_alloc_block(unsigned long size, int node);
2138 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2139 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2140 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2141                                int node);
2142 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2143 void vmemmap_populate_print_last(void);
2144 #ifdef CONFIG_MEMORY_HOTPLUG
2145 void vmemmap_free(unsigned long start, unsigned long end);
2146 #endif
2147 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2148                                   unsigned long size);
2149 
2150 enum mf_flags {
2151         MF_COUNT_INCREASED = 1 << 0,
2152         MF_ACTION_REQUIRED = 1 << 1,
2153         MF_MUST_KILL = 1 << 2,
2154         MF_SOFT_OFFLINE = 1 << 3,
2155 };
2156 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2157 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2158 extern int unpoison_memory(unsigned long pfn);
2159 extern int sysctl_memory_failure_early_kill;
2160 extern int sysctl_memory_failure_recovery;
2161 extern void shake_page(struct page *p, int access);
2162 extern atomic_long_t num_poisoned_pages;
2163 extern int soft_offline_page(struct page *page, int flags);
2164 
2165 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2166 extern void clear_huge_page(struct page *page,
2167                             unsigned long addr,
2168                             unsigned int pages_per_huge_page);
2169 extern void copy_user_huge_page(struct page *dst, struct page *src,
2170                                 unsigned long addr, struct vm_area_struct *vma,
2171                                 unsigned int pages_per_huge_page);
2172 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2173 
2174 extern struct page_ext_operations debug_guardpage_ops;
2175 extern struct page_ext_operations page_poisoning_ops;
2176 
2177 #ifdef CONFIG_DEBUG_PAGEALLOC
2178 extern unsigned int _debug_guardpage_minorder;
2179 extern bool _debug_guardpage_enabled;
2180 
2181 static inline unsigned int debug_guardpage_minorder(void)
2182 {
2183         return _debug_guardpage_minorder;
2184 }
2185 
2186 static inline bool debug_guardpage_enabled(void)
2187 {
2188         return _debug_guardpage_enabled;
2189 }
2190 
2191 static inline bool page_is_guard(struct page *page)
2192 {
2193         struct page_ext *page_ext;
2194 
2195         if (!debug_guardpage_enabled())
2196                 return false;
2197 
2198         page_ext = lookup_page_ext(page);
2199         return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2200 }
2201 #else
2202 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2203 static inline bool debug_guardpage_enabled(void) { return false; }
2204 static inline bool page_is_guard(struct page *page) { return false; }
2205 #endif /* CONFIG_DEBUG_PAGEALLOC */
2206 
2207 #if MAX_NUMNODES > 1
2208 void __init setup_nr_node_ids(void);
2209 #else
2210 static inline void setup_nr_node_ids(void) {}
2211 #endif
2212 
2213 #endif /* __KERNEL__ */
2214 #endif /* _LINUX_MM_H */
2215 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us