Version:  2.0.40 2.2.26 2.4.37 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 4.0 4.1 4.2

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 struct bdi_writeback;
 31 
 32 #ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
 33 extern unsigned long max_mapnr;
 34 
 35 static inline void set_max_mapnr(unsigned long limit)
 36 {
 37         max_mapnr = limit;
 38 }
 39 #else
 40 static inline void set_max_mapnr(unsigned long limit) { }
 41 #endif
 42 
 43 extern unsigned long totalram_pages;
 44 extern void * high_memory;
 45 extern int page_cluster;
 46 
 47 #ifdef CONFIG_SYSCTL
 48 extern int sysctl_legacy_va_layout;
 49 #else
 50 #define sysctl_legacy_va_layout 0
 51 #endif
 52 
 53 #include <asm/page.h>
 54 #include <asm/pgtable.h>
 55 #include <asm/processor.h>
 56 
 57 #ifndef __pa_symbol
 58 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 59 #endif
 60 
 61 /*
 62  * To prevent common memory management code establishing
 63  * a zero page mapping on a read fault.
 64  * This macro should be defined within <asm/pgtable.h>.
 65  * s390 does this to prevent multiplexing of hardware bits
 66  * related to the physical page in case of virtualization.
 67  */
 68 #ifndef mm_forbids_zeropage
 69 #define mm_forbids_zeropage(X)  (0)
 70 #endif
 71 
 72 extern unsigned long sysctl_user_reserve_kbytes;
 73 extern unsigned long sysctl_admin_reserve_kbytes;
 74 
 75 extern int sysctl_overcommit_memory;
 76 extern int sysctl_overcommit_ratio;
 77 extern unsigned long sysctl_overcommit_kbytes;
 78 
 79 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 80                                     size_t *, loff_t *);
 81 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 82                                     size_t *, loff_t *);
 83 
 84 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 85 
 86 /* to align the pointer to the (next) page boundary */
 87 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 88 
 89 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 90 #define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
 91 
 92 /*
 93  * Linux kernel virtual memory manager primitives.
 94  * The idea being to have a "virtual" mm in the same way
 95  * we have a virtual fs - giving a cleaner interface to the
 96  * mm details, and allowing different kinds of memory mappings
 97  * (from shared memory to executable loading to arbitrary
 98  * mmap() functions).
 99  */
100 
101 extern struct kmem_cache *vm_area_cachep;
102 
103 #ifndef CONFIG_MMU
104 extern struct rb_root nommu_region_tree;
105 extern struct rw_semaphore nommu_region_sem;
106 
107 extern unsigned int kobjsize(const void *objp);
108 #endif
109 
110 /*
111  * vm_flags in vm_area_struct, see mm_types.h.
112  */
113 #define VM_NONE         0x00000000
114 
115 #define VM_READ         0x00000001      /* currently active flags */
116 #define VM_WRITE        0x00000002
117 #define VM_EXEC         0x00000004
118 #define VM_SHARED       0x00000008
119 
120 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
121 #define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
122 #define VM_MAYWRITE     0x00000020
123 #define VM_MAYEXEC      0x00000040
124 #define VM_MAYSHARE     0x00000080
125 
126 #define VM_GROWSDOWN    0x00000100      /* general info on the segment */
127 #define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
128 #define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
129 
130 #define VM_LOCKED       0x00002000
131 #define VM_IO           0x00004000      /* Memory mapped I/O or similar */
132 
133                                         /* Used by sys_madvise() */
134 #define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
135 #define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
136 
137 #define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
138 #define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
139 #define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
140 #define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
141 #define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
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_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
210 #define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
211 #define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
212 #define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
213 #define FAULT_FLAG_TRIED        0x20    /* Second try */
214 #define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
215 
216 /*
217  * vm_fault is filled by the the pagefault handler and passed to the vma's
218  * ->fault function. The vma's ->fault is responsible for returning a bitmask
219  * of VM_FAULT_xxx flags that give details about how the fault was handled.
220  *
221  * pgoff should be used in favour of virtual_address, if possible.
222  */
223 struct vm_fault {
224         unsigned int flags;             /* FAULT_FLAG_xxx flags */
225         pgoff_t pgoff;                  /* Logical page offset based on vma */
226         void __user *virtual_address;   /* Faulting virtual address */
227 
228         struct page *cow_page;          /* Handler may choose to COW */
229         struct page *page;              /* ->fault handlers should return a
230                                          * page here, unless VM_FAULT_NOPAGE
231                                          * is set (which is also implied by
232                                          * VM_FAULT_ERROR).
233                                          */
234         /* for ->map_pages() only */
235         pgoff_t max_pgoff;              /* map pages for offset from pgoff till
236                                          * max_pgoff inclusive */
237         pte_t *pte;                     /* pte entry associated with ->pgoff */
238 };
239 
240 /*
241  * These are the virtual MM functions - opening of an area, closing and
242  * unmapping it (needed to keep files on disk up-to-date etc), pointer
243  * to the functions called when a no-page or a wp-page exception occurs. 
244  */
245 struct vm_operations_struct {
246         void (*open)(struct vm_area_struct * area);
247         void (*close)(struct vm_area_struct * area);
248         int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
249         void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
250 
251         /* notification that a previously read-only page is about to become
252          * writable, if an error is returned it will cause a SIGBUS */
253         int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
254 
255         /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
256         int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
257 
258         /* called by access_process_vm when get_user_pages() fails, typically
259          * for use by special VMAs that can switch between memory and hardware
260          */
261         int (*access)(struct vm_area_struct *vma, unsigned long addr,
262                       void *buf, int len, int write);
263 
264         /* Called by the /proc/PID/maps code to ask the vma whether it
265          * has a special name.  Returning non-NULL will also cause this
266          * vma to be dumped unconditionally. */
267         const char *(*name)(struct vm_area_struct *vma);
268 
269 #ifdef CONFIG_NUMA
270         /*
271          * set_policy() op must add a reference to any non-NULL @new mempolicy
272          * to hold the policy upon return.  Caller should pass NULL @new to
273          * remove a policy and fall back to surrounding context--i.e. do not
274          * install a MPOL_DEFAULT policy, nor the task or system default
275          * mempolicy.
276          */
277         int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
278 
279         /*
280          * get_policy() op must add reference [mpol_get()] to any policy at
281          * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
282          * in mm/mempolicy.c will do this automatically.
283          * get_policy() must NOT add a ref if the policy at (vma,addr) is not
284          * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
285          * If no [shared/vma] mempolicy exists at the addr, get_policy() op
286          * must return NULL--i.e., do not "fallback" to task or system default
287          * policy.
288          */
289         struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
290                                         unsigned long addr);
291 #endif
292         /*
293          * Called by vm_normal_page() for special PTEs to find the
294          * page for @addr.  This is useful if the default behavior
295          * (using pte_page()) would not find the correct page.
296          */
297         struct page *(*find_special_page)(struct vm_area_struct *vma,
298                                           unsigned long addr);
299 };
300 
301 struct mmu_gather;
302 struct inode;
303 
304 #define page_private(page)              ((page)->private)
305 #define set_page_private(page, v)       ((page)->private = (v))
306 
307 /* It's valid only if the page is free path or free_list */
308 static inline void set_freepage_migratetype(struct page *page, int migratetype)
309 {
310         page->index = migratetype;
311 }
312 
313 /* It's valid only if the page is free path or free_list */
314 static inline int get_freepage_migratetype(struct page *page)
315 {
316         return page->index;
317 }
318 
319 /*
320  * FIXME: take this include out, include page-flags.h in
321  * files which need it (119 of them)
322  */
323 #include <linux/page-flags.h>
324 #include <linux/huge_mm.h>
325 
326 /*
327  * Methods to modify the page usage count.
328  *
329  * What counts for a page usage:
330  * - cache mapping   (page->mapping)
331  * - private data    (page->private)
332  * - page mapped in a task's page tables, each mapping
333  *   is counted separately
334  *
335  * Also, many kernel routines increase the page count before a critical
336  * routine so they can be sure the page doesn't go away from under them.
337  */
338 
339 /*
340  * Drop a ref, return true if the refcount fell to zero (the page has no users)
341  */
342 static inline int put_page_testzero(struct page *page)
343 {
344         VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
345         return atomic_dec_and_test(&page->_count);
346 }
347 
348 /*
349  * Try to grab a ref unless the page has a refcount of zero, return false if
350  * that is the case.
351  * This can be called when MMU is off so it must not access
352  * any of the virtual mappings.
353  */
354 static inline int get_page_unless_zero(struct page *page)
355 {
356         return atomic_inc_not_zero(&page->_count);
357 }
358 
359 /*
360  * Try to drop a ref unless the page has a refcount of one, return false if
361  * that is the case.
362  * This is to make sure that the refcount won't become zero after this drop.
363  * This can be called when MMU is off so it must not access
364  * any of the virtual mappings.
365  */
366 static inline int put_page_unless_one(struct page *page)
367 {
368         return atomic_add_unless(&page->_count, -1, 1);
369 }
370 
371 extern int page_is_ram(unsigned long pfn);
372 extern int region_is_ram(resource_size_t phys_addr, unsigned long size);
373 
374 /* Support for virtually mapped pages */
375 struct page *vmalloc_to_page(const void *addr);
376 unsigned long vmalloc_to_pfn(const void *addr);
377 
378 /*
379  * Determine if an address is within the vmalloc range
380  *
381  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
382  * is no special casing required.
383  */
384 static inline int is_vmalloc_addr(const void *x)
385 {
386 #ifdef CONFIG_MMU
387         unsigned long addr = (unsigned long)x;
388 
389         return addr >= VMALLOC_START && addr < VMALLOC_END;
390 #else
391         return 0;
392 #endif
393 }
394 #ifdef CONFIG_MMU
395 extern int is_vmalloc_or_module_addr(const void *x);
396 #else
397 static inline int is_vmalloc_or_module_addr(const void *x)
398 {
399         return 0;
400 }
401 #endif
402 
403 extern void kvfree(const void *addr);
404 
405 static inline void compound_lock(struct page *page)
406 {
407 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
408         VM_BUG_ON_PAGE(PageSlab(page), page);
409         bit_spin_lock(PG_compound_lock, &page->flags);
410 #endif
411 }
412 
413 static inline void compound_unlock(struct page *page)
414 {
415 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
416         VM_BUG_ON_PAGE(PageSlab(page), page);
417         bit_spin_unlock(PG_compound_lock, &page->flags);
418 #endif
419 }
420 
421 static inline unsigned long compound_lock_irqsave(struct page *page)
422 {
423         unsigned long uninitialized_var(flags);
424 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
425         local_irq_save(flags);
426         compound_lock(page);
427 #endif
428         return flags;
429 }
430 
431 static inline void compound_unlock_irqrestore(struct page *page,
432                                               unsigned long flags)
433 {
434 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
435         compound_unlock(page);
436         local_irq_restore(flags);
437 #endif
438 }
439 
440 static inline struct page *compound_head_by_tail(struct page *tail)
441 {
442         struct page *head = tail->first_page;
443 
444         /*
445          * page->first_page may be a dangling pointer to an old
446          * compound page, so recheck that it is still a tail
447          * page before returning.
448          */
449         smp_rmb();
450         if (likely(PageTail(tail)))
451                 return head;
452         return tail;
453 }
454 
455 /*
456  * Since either compound page could be dismantled asynchronously in THP
457  * or we access asynchronously arbitrary positioned struct page, there
458  * would be tail flag race. To handle this race, we should call
459  * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
460  */
461 static inline struct page *compound_head(struct page *page)
462 {
463         if (unlikely(PageTail(page)))
464                 return compound_head_by_tail(page);
465         return page;
466 }
467 
468 /*
469  * If we access compound page synchronously such as access to
470  * allocated page, there is no need to handle tail flag race, so we can
471  * check tail flag directly without any synchronization primitive.
472  */
473 static inline struct page *compound_head_fast(struct page *page)
474 {
475         if (unlikely(PageTail(page)))
476                 return page->first_page;
477         return page;
478 }
479 
480 /*
481  * The atomic page->_mapcount, starts from -1: so that transitions
482  * both from it and to it can be tracked, using atomic_inc_and_test
483  * and atomic_add_negative(-1).
484  */
485 static inline void page_mapcount_reset(struct page *page)
486 {
487         atomic_set(&(page)->_mapcount, -1);
488 }
489 
490 static inline int page_mapcount(struct page *page)
491 {
492         VM_BUG_ON_PAGE(PageSlab(page), page);
493         return atomic_read(&page->_mapcount) + 1;
494 }
495 
496 static inline int page_count(struct page *page)
497 {
498         return atomic_read(&compound_head(page)->_count);
499 }
500 
501 static inline bool __compound_tail_refcounted(struct page *page)
502 {
503         return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
504 }
505 
506 /*
507  * This takes a head page as parameter and tells if the
508  * tail page reference counting can be skipped.
509  *
510  * For this to be safe, PageSlab and PageHeadHuge must remain true on
511  * any given page where they return true here, until all tail pins
512  * have been released.
513  */
514 static inline bool compound_tail_refcounted(struct page *page)
515 {
516         VM_BUG_ON_PAGE(!PageHead(page), page);
517         return __compound_tail_refcounted(page);
518 }
519 
520 static inline void get_huge_page_tail(struct page *page)
521 {
522         /*
523          * __split_huge_page_refcount() cannot run from under us.
524          */
525         VM_BUG_ON_PAGE(!PageTail(page), page);
526         VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
527         VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
528         if (compound_tail_refcounted(page->first_page))
529                 atomic_inc(&page->_mapcount);
530 }
531 
532 extern bool __get_page_tail(struct page *page);
533 
534 static inline void get_page(struct page *page)
535 {
536         if (unlikely(PageTail(page)))
537                 if (likely(__get_page_tail(page)))
538                         return;
539         /*
540          * Getting a normal page or the head of a compound page
541          * requires to already have an elevated page->_count.
542          */
543         VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
544         atomic_inc(&page->_count);
545 }
546 
547 static inline struct page *virt_to_head_page(const void *x)
548 {
549         struct page *page = virt_to_page(x);
550 
551         /*
552          * We don't need to worry about synchronization of tail flag
553          * when we call virt_to_head_page() since it is only called for
554          * already allocated page and this page won't be freed until
555          * this virt_to_head_page() is finished. So use _fast variant.
556          */
557         return compound_head_fast(page);
558 }
559 
560 /*
561  * Setup the page count before being freed into the page allocator for
562  * the first time (boot or memory hotplug)
563  */
564 static inline void init_page_count(struct page *page)
565 {
566         atomic_set(&page->_count, 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 
581 static inline void set_compound_page_dtor(struct page *page,
582                                                 compound_page_dtor *dtor)
583 {
584         page[1].compound_dtor = dtor;
585 }
586 
587 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
588 {
589         return page[1].compound_dtor;
590 }
591 
592 static inline int compound_order(struct page *page)
593 {
594         if (!PageHead(page))
595                 return 0;
596         return page[1].compound_order;
597 }
598 
599 static inline void set_compound_order(struct page *page, unsigned long order)
600 {
601         page[1].compound_order = order;
602 }
603 
604 #ifdef CONFIG_MMU
605 /*
606  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
607  * servicing faults for write access.  In the normal case, do always want
608  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
609  * that do not have writing enabled, when used by access_process_vm.
610  */
611 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
612 {
613         if (likely(vma->vm_flags & VM_WRITE))
614                 pte = pte_mkwrite(pte);
615         return pte;
616 }
617 
618 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
619                 struct page *page, pte_t *pte, bool write, bool anon);
620 #endif
621 
622 /*
623  * Multiple processes may "see" the same page. E.g. for untouched
624  * mappings of /dev/null, all processes see the same page full of
625  * zeroes, and text pages of executables and shared libraries have
626  * only one copy in memory, at most, normally.
627  *
628  * For the non-reserved pages, page_count(page) denotes a reference count.
629  *   page_count() == 0 means the page is free. page->lru is then used for
630  *   freelist management in the buddy allocator.
631  *   page_count() > 0  means the page has been allocated.
632  *
633  * Pages are allocated by the slab allocator in order to provide memory
634  * to kmalloc and kmem_cache_alloc. In this case, the management of the
635  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
636  * unless a particular usage is carefully commented. (the responsibility of
637  * freeing the kmalloc memory is the caller's, of course).
638  *
639  * A page may be used by anyone else who does a __get_free_page().
640  * In this case, page_count still tracks the references, and should only
641  * be used through the normal accessor functions. The top bits of page->flags
642  * and page->virtual store page management information, but all other fields
643  * are unused and could be used privately, carefully. The management of this
644  * page is the responsibility of the one who allocated it, and those who have
645  * subsequently been given references to it.
646  *
647  * The other pages (we may call them "pagecache pages") are completely
648  * managed by the Linux memory manager: I/O, buffers, swapping etc.
649  * The following discussion applies only to them.
650  *
651  * A pagecache page contains an opaque `private' member, which belongs to the
652  * page's address_space. Usually, this is the address of a circular list of
653  * the page's disk buffers. PG_private must be set to tell the VM to call
654  * into the filesystem to release these pages.
655  *
656  * A page may belong to an inode's memory mapping. In this case, page->mapping
657  * is the pointer to the inode, and page->index is the file offset of the page,
658  * in units of PAGE_CACHE_SIZE.
659  *
660  * If pagecache pages are not associated with an inode, they are said to be
661  * anonymous pages. These may become associated with the swapcache, and in that
662  * case PG_swapcache is set, and page->private is an offset into the swapcache.
663  *
664  * In either case (swapcache or inode backed), the pagecache itself holds one
665  * reference to the page. Setting PG_private should also increment the
666  * refcount. The each user mapping also has a reference to the page.
667  *
668  * The pagecache pages are stored in a per-mapping radix tree, which is
669  * rooted at mapping->page_tree, and indexed by offset.
670  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
671  * lists, we instead now tag pages as dirty/writeback in the radix tree.
672  *
673  * All pagecache pages may be subject to I/O:
674  * - inode pages may need to be read from disk,
675  * - inode pages which have been modified and are MAP_SHARED may need
676  *   to be written back to the inode on disk,
677  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
678  *   modified may need to be swapped out to swap space and (later) to be read
679  *   back into memory.
680  */
681 
682 /*
683  * The zone field is never updated after free_area_init_core()
684  * sets it, so none of the operations on it need to be atomic.
685  */
686 
687 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
688 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
689 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
690 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
691 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
692 
693 /*
694  * Define the bit shifts to access each section.  For non-existent
695  * sections we define the shift as 0; that plus a 0 mask ensures
696  * the compiler will optimise away reference to them.
697  */
698 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
699 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
700 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
701 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
702 
703 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
704 #ifdef NODE_NOT_IN_PAGE_FLAGS
705 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
706 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
707                                                 SECTIONS_PGOFF : ZONES_PGOFF)
708 #else
709 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
710 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
711                                                 NODES_PGOFF : ZONES_PGOFF)
712 #endif
713 
714 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
715 
716 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
717 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
718 #endif
719 
720 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
721 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
722 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
723 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
724 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
725 
726 static inline enum zone_type page_zonenum(const struct page *page)
727 {
728         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
729 }
730 
731 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
732 #define SECTION_IN_PAGE_FLAGS
733 #endif
734 
735 /*
736  * The identification function is mainly used by the buddy allocator for
737  * determining if two pages could be buddies. We are not really identifying
738  * the zone since we could be using the section number id if we do not have
739  * node id available in page flags.
740  * We only guarantee that it will return the same value for two combinable
741  * pages in a zone.
742  */
743 static inline int page_zone_id(struct page *page)
744 {
745         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
746 }
747 
748 static inline int zone_to_nid(struct zone *zone)
749 {
750 #ifdef CONFIG_NUMA
751         return zone->node;
752 #else
753         return 0;
754 #endif
755 }
756 
757 #ifdef NODE_NOT_IN_PAGE_FLAGS
758 extern int page_to_nid(const struct page *page);
759 #else
760 static inline int page_to_nid(const struct page *page)
761 {
762         return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
763 }
764 #endif
765 
766 #ifdef CONFIG_NUMA_BALANCING
767 static inline int cpu_pid_to_cpupid(int cpu, int pid)
768 {
769         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
770 }
771 
772 static inline int cpupid_to_pid(int cpupid)
773 {
774         return cpupid & LAST__PID_MASK;
775 }
776 
777 static inline int cpupid_to_cpu(int cpupid)
778 {
779         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
780 }
781 
782 static inline int cpupid_to_nid(int cpupid)
783 {
784         return cpu_to_node(cpupid_to_cpu(cpupid));
785 }
786 
787 static inline bool cpupid_pid_unset(int cpupid)
788 {
789         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
790 }
791 
792 static inline bool cpupid_cpu_unset(int cpupid)
793 {
794         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
795 }
796 
797 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
798 {
799         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
800 }
801 
802 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
803 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
804 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
805 {
806         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
807 }
808 
809 static inline int page_cpupid_last(struct page *page)
810 {
811         return page->_last_cpupid;
812 }
813 static inline void page_cpupid_reset_last(struct page *page)
814 {
815         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
816 }
817 #else
818 static inline int page_cpupid_last(struct page *page)
819 {
820         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
821 }
822 
823 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
824 
825 static inline void page_cpupid_reset_last(struct page *page)
826 {
827         int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
828 
829         page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
830         page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
831 }
832 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
833 #else /* !CONFIG_NUMA_BALANCING */
834 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
835 {
836         return page_to_nid(page); /* XXX */
837 }
838 
839 static inline int page_cpupid_last(struct page *page)
840 {
841         return page_to_nid(page); /* XXX */
842 }
843 
844 static inline int cpupid_to_nid(int cpupid)
845 {
846         return -1;
847 }
848 
849 static inline int cpupid_to_pid(int cpupid)
850 {
851         return -1;
852 }
853 
854 static inline int cpupid_to_cpu(int cpupid)
855 {
856         return -1;
857 }
858 
859 static inline int cpu_pid_to_cpupid(int nid, int pid)
860 {
861         return -1;
862 }
863 
864 static inline bool cpupid_pid_unset(int cpupid)
865 {
866         return 1;
867 }
868 
869 static inline void page_cpupid_reset_last(struct page *page)
870 {
871 }
872 
873 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
874 {
875         return false;
876 }
877 #endif /* CONFIG_NUMA_BALANCING */
878 
879 static inline struct zone *page_zone(const struct page *page)
880 {
881         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
882 }
883 
884 #ifdef SECTION_IN_PAGE_FLAGS
885 static inline void set_page_section(struct page *page, unsigned long section)
886 {
887         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
888         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
889 }
890 
891 static inline unsigned long page_to_section(const struct page *page)
892 {
893         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
894 }
895 #endif
896 
897 static inline void set_page_zone(struct page *page, enum zone_type zone)
898 {
899         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
900         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
901 }
902 
903 static inline void set_page_node(struct page *page, unsigned long node)
904 {
905         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
906         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
907 }
908 
909 static inline void set_page_links(struct page *page, enum zone_type zone,
910         unsigned long node, unsigned long pfn)
911 {
912         set_page_zone(page, zone);
913         set_page_node(page, node);
914 #ifdef SECTION_IN_PAGE_FLAGS
915         set_page_section(page, pfn_to_section_nr(pfn));
916 #endif
917 }
918 
919 /*
920  * Some inline functions in vmstat.h depend on page_zone()
921  */
922 #include <linux/vmstat.h>
923 
924 static __always_inline void *lowmem_page_address(const struct page *page)
925 {
926         return __va(PFN_PHYS(page_to_pfn(page)));
927 }
928 
929 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
930 #define HASHED_PAGE_VIRTUAL
931 #endif
932 
933 #if defined(WANT_PAGE_VIRTUAL)
934 static inline void *page_address(const struct page *page)
935 {
936         return page->virtual;
937 }
938 static inline void set_page_address(struct page *page, void *address)
939 {
940         page->virtual = address;
941 }
942 #define page_address_init()  do { } while(0)
943 #endif
944 
945 #if defined(HASHED_PAGE_VIRTUAL)
946 void *page_address(const struct page *page);
947 void set_page_address(struct page *page, void *virtual);
948 void page_address_init(void);
949 #endif
950 
951 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
952 #define page_address(page) lowmem_page_address(page)
953 #define set_page_address(page, address)  do { } while(0)
954 #define page_address_init()  do { } while(0)
955 #endif
956 
957 extern void *page_rmapping(struct page *page);
958 extern struct anon_vma *page_anon_vma(struct page *page);
959 extern struct address_space *page_mapping(struct page *page);
960 
961 extern struct address_space *__page_file_mapping(struct page *);
962 
963 static inline
964 struct address_space *page_file_mapping(struct page *page)
965 {
966         if (unlikely(PageSwapCache(page)))
967                 return __page_file_mapping(page);
968 
969         return page->mapping;
970 }
971 
972 /*
973  * Return the pagecache index of the passed page.  Regular pagecache pages
974  * use ->index whereas swapcache pages use ->private
975  */
976 static inline pgoff_t page_index(struct page *page)
977 {
978         if (unlikely(PageSwapCache(page)))
979                 return page_private(page);
980         return page->index;
981 }
982 
983 extern pgoff_t __page_file_index(struct page *page);
984 
985 /*
986  * Return the file index of the page. Regular pagecache pages use ->index
987  * whereas swapcache pages use swp_offset(->private)
988  */
989 static inline pgoff_t page_file_index(struct page *page)
990 {
991         if (unlikely(PageSwapCache(page)))
992                 return __page_file_index(page);
993 
994         return page->index;
995 }
996 
997 /*
998  * Return true if this page is mapped into pagetables.
999  */
1000 static inline int page_mapped(struct page *page)
1001 {
1002         return atomic_read(&(page)->_mapcount) >= 0;
1003 }
1004 
1005 /*
1006  * Return true only if the page has been allocated with
1007  * ALLOC_NO_WATERMARKS and the low watermark was not
1008  * met implying that the system is under some pressure.
1009  */
1010 static inline bool page_is_pfmemalloc(struct page *page)
1011 {
1012         /*
1013          * Page index cannot be this large so this must be
1014          * a pfmemalloc page.
1015          */
1016         return page->index == -1UL;
1017 }
1018 
1019 /*
1020  * Only to be called by the page allocator on a freshly allocated
1021  * page.
1022  */
1023 static inline void set_page_pfmemalloc(struct page *page)
1024 {
1025         page->index = -1UL;
1026 }
1027 
1028 static inline void clear_page_pfmemalloc(struct page *page)
1029 {
1030         page->index = 0;
1031 }
1032 
1033 /*
1034  * Different kinds of faults, as returned by handle_mm_fault().
1035  * Used to decide whether a process gets delivered SIGBUS or
1036  * just gets major/minor fault counters bumped up.
1037  */
1038 
1039 #define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
1040 
1041 #define VM_FAULT_OOM    0x0001
1042 #define VM_FAULT_SIGBUS 0x0002
1043 #define VM_FAULT_MAJOR  0x0004
1044 #define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
1045 #define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
1046 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1047 #define VM_FAULT_SIGSEGV 0x0040
1048 
1049 #define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
1050 #define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
1051 #define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
1052 #define VM_FAULT_FALLBACK 0x0800        /* huge page fault failed, fall back to small */
1053 
1054 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1055 
1056 #define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1057                          VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1058                          VM_FAULT_FALLBACK)
1059 
1060 /* Encode hstate index for a hwpoisoned large page */
1061 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1062 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1063 
1064 /*
1065  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1066  */
1067 extern void pagefault_out_of_memory(void);
1068 
1069 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1070 
1071 /*
1072  * Flags passed to show_mem() and show_free_areas() to suppress output in
1073  * various contexts.
1074  */
1075 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1076 
1077 extern void show_free_areas(unsigned int flags);
1078 extern bool skip_free_areas_node(unsigned int flags, int nid);
1079 
1080 int shmem_zero_setup(struct vm_area_struct *);
1081 #ifdef CONFIG_SHMEM
1082 bool shmem_mapping(struct address_space *mapping);
1083 #else
1084 static inline bool shmem_mapping(struct address_space *mapping)
1085 {
1086         return false;
1087 }
1088 #endif
1089 
1090 extern int can_do_mlock(void);
1091 extern int user_shm_lock(size_t, struct user_struct *);
1092 extern void user_shm_unlock(size_t, struct user_struct *);
1093 
1094 /*
1095  * Parameter block passed down to zap_pte_range in exceptional cases.
1096  */
1097 struct zap_details {
1098         struct address_space *check_mapping;    /* Check page->mapping if set */
1099         pgoff_t first_index;                    /* Lowest page->index to unmap */
1100         pgoff_t last_index;                     /* Highest page->index to unmap */
1101 };
1102 
1103 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1104                 pte_t pte);
1105 
1106 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1107                 unsigned long size);
1108 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1109                 unsigned long size, struct zap_details *);
1110 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1111                 unsigned long start, unsigned long end);
1112 
1113 /**
1114  * mm_walk - callbacks for walk_page_range
1115  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1116  *             this handler is required to be able to handle
1117  *             pmd_trans_huge() pmds.  They may simply choose to
1118  *             split_huge_page() instead of handling it explicitly.
1119  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1120  * @pte_hole: if set, called for each hole at all levels
1121  * @hugetlb_entry: if set, called for each hugetlb entry
1122  * @test_walk: caller specific callback function to determine whether
1123  *             we walk over the current vma or not. A positive returned
1124  *             value means "do page table walk over the current vma,"
1125  *             and a negative one means "abort current page table walk
1126  *             right now." 0 means "skip the current vma."
1127  * @mm:        mm_struct representing the target process of page table walk
1128  * @vma:       vma currently walked (NULL if walking outside vmas)
1129  * @private:   private data for callbacks' usage
1130  *
1131  * (see the comment on walk_page_range() for more details)
1132  */
1133 struct mm_walk {
1134         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1135                          unsigned long next, struct mm_walk *walk);
1136         int (*pte_entry)(pte_t *pte, unsigned long addr,
1137                          unsigned long next, struct mm_walk *walk);
1138         int (*pte_hole)(unsigned long addr, unsigned long next,
1139                         struct mm_walk *walk);
1140         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1141                              unsigned long addr, unsigned long next,
1142                              struct mm_walk *walk);
1143         int (*test_walk)(unsigned long addr, unsigned long next,
1144                         struct mm_walk *walk);
1145         struct mm_struct *mm;
1146         struct vm_area_struct *vma;
1147         void *private;
1148 };
1149 
1150 int walk_page_range(unsigned long addr, unsigned long end,
1151                 struct mm_walk *walk);
1152 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1153 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1154                 unsigned long end, unsigned long floor, unsigned long ceiling);
1155 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1156                         struct vm_area_struct *vma);
1157 void unmap_mapping_range(struct address_space *mapping,
1158                 loff_t const holebegin, loff_t const holelen, int even_cows);
1159 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1160         unsigned long *pfn);
1161 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1162                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1163 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1164                         void *buf, int len, int write);
1165 
1166 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1167                 loff_t const holebegin, loff_t const holelen)
1168 {
1169         unmap_mapping_range(mapping, holebegin, holelen, 0);
1170 }
1171 
1172 extern void truncate_pagecache(struct inode *inode, loff_t new);
1173 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1174 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1175 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1176 int truncate_inode_page(struct address_space *mapping, struct page *page);
1177 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1178 int invalidate_inode_page(struct page *page);
1179 
1180 #ifdef CONFIG_MMU
1181 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1182                         unsigned long address, unsigned int flags);
1183 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1184                             unsigned long address, unsigned int fault_flags);
1185 #else
1186 static inline int handle_mm_fault(struct mm_struct *mm,
1187                         struct vm_area_struct *vma, unsigned long address,
1188                         unsigned int flags)
1189 {
1190         /* should never happen if there's no MMU */
1191         BUG();
1192         return VM_FAULT_SIGBUS;
1193 }
1194 static inline int fixup_user_fault(struct task_struct *tsk,
1195                 struct mm_struct *mm, unsigned long address,
1196                 unsigned int fault_flags)
1197 {
1198         /* should never happen if there's no MMU */
1199         BUG();
1200         return -EFAULT;
1201 }
1202 #endif
1203 
1204 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1205 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1206                 void *buf, int len, int write);
1207 
1208 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1209                       unsigned long start, unsigned long nr_pages,
1210                       unsigned int foll_flags, struct page **pages,
1211                       struct vm_area_struct **vmas, int *nonblocking);
1212 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1213                     unsigned long start, unsigned long nr_pages,
1214                     int write, int force, struct page **pages,
1215                     struct vm_area_struct **vmas);
1216 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1217                     unsigned long start, unsigned long nr_pages,
1218                     int write, int force, struct page **pages,
1219                     int *locked);
1220 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1221                                unsigned long start, unsigned long nr_pages,
1222                                int write, int force, struct page **pages,
1223                                unsigned int gup_flags);
1224 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1225                     unsigned long start, unsigned long nr_pages,
1226                     int write, int force, struct page **pages);
1227 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1228                         struct page **pages);
1229 struct kvec;
1230 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1231                         struct page **pages);
1232 int get_kernel_page(unsigned long start, int write, struct page **pages);
1233 struct page *get_dump_page(unsigned long addr);
1234 
1235 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1236 extern void do_invalidatepage(struct page *page, unsigned int offset,
1237                               unsigned int length);
1238 
1239 int __set_page_dirty_nobuffers(struct page *page);
1240 int __set_page_dirty_no_writeback(struct page *page);
1241 int redirty_page_for_writepage(struct writeback_control *wbc,
1242                                 struct page *page);
1243 void account_page_dirtied(struct page *page, struct address_space *mapping,
1244                           struct mem_cgroup *memcg);
1245 void account_page_cleaned(struct page *page, struct address_space *mapping,
1246                           struct mem_cgroup *memcg, struct bdi_writeback *wb);
1247 int set_page_dirty(struct page *page);
1248 int set_page_dirty_lock(struct page *page);
1249 void cancel_dirty_page(struct page *page);
1250 int clear_page_dirty_for_io(struct page *page);
1251 
1252 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1253 
1254 /* Is the vma a continuation of the stack vma above it? */
1255 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1256 {
1257         return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1258 }
1259 
1260 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1261                                              unsigned long addr)
1262 {
1263         return (vma->vm_flags & VM_GROWSDOWN) &&
1264                 (vma->vm_start == addr) &&
1265                 !vma_growsdown(vma->vm_prev, addr);
1266 }
1267 
1268 /* Is the vma a continuation of the stack vma below it? */
1269 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1270 {
1271         return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1272 }
1273 
1274 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1275                                            unsigned long addr)
1276 {
1277         return (vma->vm_flags & VM_GROWSUP) &&
1278                 (vma->vm_end == addr) &&
1279                 !vma_growsup(vma->vm_next, addr);
1280 }
1281 
1282 extern struct task_struct *task_of_stack(struct task_struct *task,
1283                                 struct vm_area_struct *vma, bool in_group);
1284 
1285 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1286                 unsigned long old_addr, struct vm_area_struct *new_vma,
1287                 unsigned long new_addr, unsigned long len,
1288                 bool need_rmap_locks);
1289 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1290                               unsigned long end, pgprot_t newprot,
1291                               int dirty_accountable, int prot_numa);
1292 extern int mprotect_fixup(struct vm_area_struct *vma,
1293                           struct vm_area_struct **pprev, unsigned long start,
1294                           unsigned long end, unsigned long newflags);
1295 
1296 /*
1297  * doesn't attempt to fault and will return short.
1298  */
1299 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1300                           struct page **pages);
1301 /*
1302  * per-process(per-mm_struct) statistics.
1303  */
1304 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1305 {
1306         long val = atomic_long_read(&mm->rss_stat.count[member]);
1307 
1308 #ifdef SPLIT_RSS_COUNTING
1309         /*
1310          * counter is updated in asynchronous manner and may go to minus.
1311          * But it's never be expected number for users.
1312          */
1313         if (val < 0)
1314                 val = 0;
1315 #endif
1316         return (unsigned long)val;
1317 }
1318 
1319 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1320 {
1321         atomic_long_add(value, &mm->rss_stat.count[member]);
1322 }
1323 
1324 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1325 {
1326         atomic_long_inc(&mm->rss_stat.count[member]);
1327 }
1328 
1329 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1330 {
1331         atomic_long_dec(&mm->rss_stat.count[member]);
1332 }
1333 
1334 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1335 {
1336         return get_mm_counter(mm, MM_FILEPAGES) +
1337                 get_mm_counter(mm, MM_ANONPAGES);
1338 }
1339 
1340 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1341 {
1342         return max(mm->hiwater_rss, get_mm_rss(mm));
1343 }
1344 
1345 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1346 {
1347         return max(mm->hiwater_vm, mm->total_vm);
1348 }
1349 
1350 static inline void update_hiwater_rss(struct mm_struct *mm)
1351 {
1352         unsigned long _rss = get_mm_rss(mm);
1353 
1354         if ((mm)->hiwater_rss < _rss)
1355                 (mm)->hiwater_rss = _rss;
1356 }
1357 
1358 static inline void update_hiwater_vm(struct mm_struct *mm)
1359 {
1360         if (mm->hiwater_vm < mm->total_vm)
1361                 mm->hiwater_vm = mm->total_vm;
1362 }
1363 
1364 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1365 {
1366         mm->hiwater_rss = get_mm_rss(mm);
1367 }
1368 
1369 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1370                                          struct mm_struct *mm)
1371 {
1372         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1373 
1374         if (*maxrss < hiwater_rss)
1375                 *maxrss = hiwater_rss;
1376 }
1377 
1378 #if defined(SPLIT_RSS_COUNTING)
1379 void sync_mm_rss(struct mm_struct *mm);
1380 #else
1381 static inline void sync_mm_rss(struct mm_struct *mm)
1382 {
1383 }
1384 #endif
1385 
1386 int vma_wants_writenotify(struct vm_area_struct *vma);
1387 
1388 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1389                                spinlock_t **ptl);
1390 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1391                                     spinlock_t **ptl)
1392 {
1393         pte_t *ptep;
1394         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1395         return ptep;
1396 }
1397 
1398 #ifdef __PAGETABLE_PUD_FOLDED
1399 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1400                                                 unsigned long address)
1401 {
1402         return 0;
1403 }
1404 #else
1405 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1406 #endif
1407 
1408 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1409 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1410                                                 unsigned long address)
1411 {
1412         return 0;
1413 }
1414 
1415 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1416 
1417 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1418 {
1419         return 0;
1420 }
1421 
1422 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1423 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1424 
1425 #else
1426 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1427 
1428 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1429 {
1430         atomic_long_set(&mm->nr_pmds, 0);
1431 }
1432 
1433 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1434 {
1435         return atomic_long_read(&mm->nr_pmds);
1436 }
1437 
1438 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1439 {
1440         atomic_long_inc(&mm->nr_pmds);
1441 }
1442 
1443 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1444 {
1445         atomic_long_dec(&mm->nr_pmds);
1446 }
1447 #endif
1448 
1449 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1450                 pmd_t *pmd, unsigned long address);
1451 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1452 
1453 /*
1454  * The following ifdef needed to get the 4level-fixup.h header to work.
1455  * Remove it when 4level-fixup.h has been removed.
1456  */
1457 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1458 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1459 {
1460         return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1461                 NULL: pud_offset(pgd, address);
1462 }
1463 
1464 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1465 {
1466         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1467                 NULL: pmd_offset(pud, address);
1468 }
1469 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1470 
1471 #if USE_SPLIT_PTE_PTLOCKS
1472 #if ALLOC_SPLIT_PTLOCKS
1473 void __init ptlock_cache_init(void);
1474 extern bool ptlock_alloc(struct page *page);
1475 extern void ptlock_free(struct page *page);
1476 
1477 static inline spinlock_t *ptlock_ptr(struct page *page)
1478 {
1479         return page->ptl;
1480 }
1481 #else /* ALLOC_SPLIT_PTLOCKS */
1482 static inline void ptlock_cache_init(void)
1483 {
1484 }
1485 
1486 static inline bool ptlock_alloc(struct page *page)
1487 {
1488         return true;
1489 }
1490 
1491 static inline void ptlock_free(struct page *page)
1492 {
1493 }
1494 
1495 static inline spinlock_t *ptlock_ptr(struct page *page)
1496 {
1497         return &page->ptl;
1498 }
1499 #endif /* ALLOC_SPLIT_PTLOCKS */
1500 
1501 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1502 {
1503         return ptlock_ptr(pmd_page(*pmd));
1504 }
1505 
1506 static inline bool ptlock_init(struct page *page)
1507 {
1508         /*
1509          * prep_new_page() initialize page->private (and therefore page->ptl)
1510          * with 0. Make sure nobody took it in use in between.
1511          *
1512          * It can happen if arch try to use slab for page table allocation:
1513          * slab code uses page->slab_cache and page->first_page (for tail
1514          * pages), which share storage with page->ptl.
1515          */
1516         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1517         if (!ptlock_alloc(page))
1518                 return false;
1519         spin_lock_init(ptlock_ptr(page));
1520         return true;
1521 }
1522 
1523 /* Reset page->mapping so free_pages_check won't complain. */
1524 static inline void pte_lock_deinit(struct page *page)
1525 {
1526         page->mapping = NULL;
1527         ptlock_free(page);
1528 }
1529 
1530 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1531 /*
1532  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1533  */
1534 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1535 {
1536         return &mm->page_table_lock;
1537 }
1538 static inline void ptlock_cache_init(void) {}
1539 static inline bool ptlock_init(struct page *page) { return true; }
1540 static inline void pte_lock_deinit(struct page *page) {}
1541 #endif /* USE_SPLIT_PTE_PTLOCKS */
1542 
1543 static inline void pgtable_init(void)
1544 {
1545         ptlock_cache_init();
1546         pgtable_cache_init();
1547 }
1548 
1549 static inline bool pgtable_page_ctor(struct page *page)
1550 {
1551         inc_zone_page_state(page, NR_PAGETABLE);
1552         return ptlock_init(page);
1553 }
1554 
1555 static inline void pgtable_page_dtor(struct page *page)
1556 {
1557         pte_lock_deinit(page);
1558         dec_zone_page_state(page, NR_PAGETABLE);
1559 }
1560 
1561 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1562 ({                                                      \
1563         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1564         pte_t *__pte = pte_offset_map(pmd, address);    \
1565         *(ptlp) = __ptl;                                \
1566         spin_lock(__ptl);                               \
1567         __pte;                                          \
1568 })
1569 
1570 #define pte_unmap_unlock(pte, ptl)      do {            \
1571         spin_unlock(ptl);                               \
1572         pte_unmap(pte);                                 \
1573 } while (0)
1574 
1575 #define pte_alloc_map(mm, vma, pmd, address)                            \
1576         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1577                                                         pmd, address))? \
1578          NULL: pte_offset_map(pmd, address))
1579 
1580 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1581         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1582                                                         pmd, address))? \
1583                 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1584 
1585 #define pte_alloc_kernel(pmd, address)                  \
1586         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1587                 NULL: pte_offset_kernel(pmd, address))
1588 
1589 #if USE_SPLIT_PMD_PTLOCKS
1590 
1591 static struct page *pmd_to_page(pmd_t *pmd)
1592 {
1593         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1594         return virt_to_page((void *)((unsigned long) pmd & mask));
1595 }
1596 
1597 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1598 {
1599         return ptlock_ptr(pmd_to_page(pmd));
1600 }
1601 
1602 static inline bool pgtable_pmd_page_ctor(struct page *page)
1603 {
1604 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1605         page->pmd_huge_pte = NULL;
1606 #endif
1607         return ptlock_init(page);
1608 }
1609 
1610 static inline void pgtable_pmd_page_dtor(struct page *page)
1611 {
1612 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1613         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1614 #endif
1615         ptlock_free(page);
1616 }
1617 
1618 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1619 
1620 #else
1621 
1622 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1623 {
1624         return &mm->page_table_lock;
1625 }
1626 
1627 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1628 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1629 
1630 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1631 
1632 #endif
1633 
1634 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1635 {
1636         spinlock_t *ptl = pmd_lockptr(mm, pmd);
1637         spin_lock(ptl);
1638         return ptl;
1639 }
1640 
1641 extern void free_area_init(unsigned long * zones_size);
1642 extern void free_area_init_node(int nid, unsigned long * zones_size,
1643                 unsigned long zone_start_pfn, unsigned long *zholes_size);
1644 extern void free_initmem(void);
1645 
1646 /*
1647  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1648  * into the buddy system. The freed pages will be poisoned with pattern
1649  * "poison" if it's within range [0, UCHAR_MAX].
1650  * Return pages freed into the buddy system.
1651  */
1652 extern unsigned long free_reserved_area(void *start, void *end,
1653                                         int poison, char *s);
1654 
1655 #ifdef  CONFIG_HIGHMEM
1656 /*
1657  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1658  * and totalram_pages.
1659  */
1660 extern void free_highmem_page(struct page *page);
1661 #endif
1662 
1663 extern void adjust_managed_page_count(struct page *page, long count);
1664 extern void mem_init_print_info(const char *str);
1665 
1666 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1667 
1668 /* Free the reserved page into the buddy system, so it gets managed. */
1669 static inline void __free_reserved_page(struct page *page)
1670 {
1671         ClearPageReserved(page);
1672         init_page_count(page);
1673         __free_page(page);
1674 }
1675 
1676 static inline void free_reserved_page(struct page *page)
1677 {
1678         __free_reserved_page(page);
1679         adjust_managed_page_count(page, 1);
1680 }
1681 
1682 static inline void mark_page_reserved(struct page *page)
1683 {
1684         SetPageReserved(page);
1685         adjust_managed_page_count(page, -1);
1686 }
1687 
1688 /*
1689  * Default method to free all the __init memory into the buddy system.
1690  * The freed pages will be poisoned with pattern "poison" if it's within
1691  * range [0, UCHAR_MAX].
1692  * Return pages freed into the buddy system.
1693  */
1694 static inline unsigned long free_initmem_default(int poison)
1695 {
1696         extern char __init_begin[], __init_end[];
1697 
1698         return free_reserved_area(&__init_begin, &__init_end,
1699                                   poison, "unused kernel");
1700 }
1701 
1702 static inline unsigned long get_num_physpages(void)
1703 {
1704         int nid;
1705         unsigned long phys_pages = 0;
1706 
1707         for_each_online_node(nid)
1708                 phys_pages += node_present_pages(nid);
1709 
1710         return phys_pages;
1711 }
1712 
1713 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1714 /*
1715  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1716  * zones, allocate the backing mem_map and account for memory holes in a more
1717  * architecture independent manner. This is a substitute for creating the
1718  * zone_sizes[] and zholes_size[] arrays and passing them to
1719  * free_area_init_node()
1720  *
1721  * An architecture is expected to register range of page frames backed by
1722  * physical memory with memblock_add[_node]() before calling
1723  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1724  * usage, an architecture is expected to do something like
1725  *
1726  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1727  *                                                       max_highmem_pfn};
1728  * for_each_valid_physical_page_range()
1729  *      memblock_add_node(base, size, nid)
1730  * free_area_init_nodes(max_zone_pfns);
1731  *
1732  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1733  * registered physical page range.  Similarly
1734  * sparse_memory_present_with_active_regions() calls memory_present() for
1735  * each range when SPARSEMEM is enabled.
1736  *
1737  * See mm/page_alloc.c for more information on each function exposed by
1738  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1739  */
1740 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1741 unsigned long node_map_pfn_alignment(void);
1742 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1743                                                 unsigned long end_pfn);
1744 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1745                                                 unsigned long end_pfn);
1746 extern void get_pfn_range_for_nid(unsigned int nid,
1747                         unsigned long *start_pfn, unsigned long *end_pfn);
1748 extern unsigned long find_min_pfn_with_active_regions(void);
1749 extern void free_bootmem_with_active_regions(int nid,
1750                                                 unsigned long max_low_pfn);
1751 extern void sparse_memory_present_with_active_regions(int nid);
1752 
1753 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1754 
1755 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1756     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1757 static inline int __early_pfn_to_nid(unsigned long pfn,
1758                                         struct mminit_pfnnid_cache *state)
1759 {
1760         return 0;
1761 }
1762 #else
1763 /* please see mm/page_alloc.c */
1764 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1765 /* there is a per-arch backend function. */
1766 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1767                                         struct mminit_pfnnid_cache *state);
1768 #endif
1769 
1770 extern void set_dma_reserve(unsigned long new_dma_reserve);
1771 extern void memmap_init_zone(unsigned long, int, unsigned long,
1772                                 unsigned long, enum memmap_context);
1773 extern void setup_per_zone_wmarks(void);
1774 extern int __meminit init_per_zone_wmark_min(void);
1775 extern void mem_init(void);
1776 extern void __init mmap_init(void);
1777 extern void show_mem(unsigned int flags);
1778 extern void si_meminfo(struct sysinfo * val);
1779 extern void si_meminfo_node(struct sysinfo *val, int nid);
1780 
1781 extern __printf(3, 4)
1782 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1783 
1784 extern void setup_per_cpu_pageset(void);
1785 
1786 extern void zone_pcp_update(struct zone *zone);
1787 extern void zone_pcp_reset(struct zone *zone);
1788 
1789 /* page_alloc.c */
1790 extern int min_free_kbytes;
1791 
1792 /* nommu.c */
1793 extern atomic_long_t mmap_pages_allocated;
1794 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1795 
1796 /* interval_tree.c */
1797 void vma_interval_tree_insert(struct vm_area_struct *node,
1798                               struct rb_root *root);
1799 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1800                                     struct vm_area_struct *prev,
1801                                     struct rb_root *root);
1802 void vma_interval_tree_remove(struct vm_area_struct *node,
1803                               struct rb_root *root);
1804 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1805                                 unsigned long start, unsigned long last);
1806 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1807                                 unsigned long start, unsigned long last);
1808 
1809 #define vma_interval_tree_foreach(vma, root, start, last)               \
1810         for (vma = vma_interval_tree_iter_first(root, start, last);     \
1811              vma; vma = vma_interval_tree_iter_next(vma, start, last))
1812 
1813 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1814                                    struct rb_root *root);
1815 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1816                                    struct rb_root *root);
1817 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1818         struct rb_root *root, unsigned long start, unsigned long last);
1819 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1820         struct anon_vma_chain *node, unsigned long start, unsigned long last);
1821 #ifdef CONFIG_DEBUG_VM_RB
1822 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1823 #endif
1824 
1825 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1826         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1827              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1828 
1829 /* mmap.c */
1830 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1831 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1832         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1833 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1834         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1835         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1836         struct mempolicy *);
1837 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1838 extern int split_vma(struct mm_struct *,
1839         struct vm_area_struct *, unsigned long addr, int new_below);
1840 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1841 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1842         struct rb_node **, struct rb_node *);
1843 extern void unlink_file_vma(struct vm_area_struct *);
1844 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1845         unsigned long addr, unsigned long len, pgoff_t pgoff,
1846         bool *need_rmap_locks);
1847 extern void exit_mmap(struct mm_struct *);
1848 
1849 static inline int check_data_rlimit(unsigned long rlim,
1850                                     unsigned long new,
1851                                     unsigned long start,
1852                                     unsigned long end_data,
1853                                     unsigned long start_data)
1854 {
1855         if (rlim < RLIM_INFINITY) {
1856                 if (((new - start) + (end_data - start_data)) > rlim)
1857                         return -ENOSPC;
1858         }
1859 
1860         return 0;
1861 }
1862 
1863 extern int mm_take_all_locks(struct mm_struct *mm);
1864 extern void mm_drop_all_locks(struct mm_struct *mm);
1865 
1866 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1867 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1868 
1869 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1870 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1871                                    unsigned long addr, unsigned long len,
1872                                    unsigned long flags,
1873                                    const struct vm_special_mapping *spec);
1874 /* This is an obsolete alternative to _install_special_mapping. */
1875 extern int install_special_mapping(struct mm_struct *mm,
1876                                    unsigned long addr, unsigned long len,
1877                                    unsigned long flags, struct page **pages);
1878 
1879 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1880 
1881 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1882         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1883 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1884         unsigned long len, unsigned long prot, unsigned long flags,
1885         unsigned long pgoff, unsigned long *populate);
1886 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1887 
1888 #ifdef CONFIG_MMU
1889 extern int __mm_populate(unsigned long addr, unsigned long len,
1890                          int ignore_errors);
1891 static inline void mm_populate(unsigned long addr, unsigned long len)
1892 {
1893         /* Ignore errors */
1894         (void) __mm_populate(addr, len, 1);
1895 }
1896 #else
1897 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1898 #endif
1899 
1900 /* These take the mm semaphore themselves */
1901 extern unsigned long vm_brk(unsigned long, unsigned long);
1902 extern int vm_munmap(unsigned long, size_t);
1903 extern unsigned long vm_mmap(struct file *, unsigned long,
1904         unsigned long, unsigned long,
1905         unsigned long, unsigned long);
1906 
1907 struct vm_unmapped_area_info {
1908 #define VM_UNMAPPED_AREA_TOPDOWN 1
1909         unsigned long flags;
1910         unsigned long length;
1911         unsigned long low_limit;
1912         unsigned long high_limit;
1913         unsigned long align_mask;
1914         unsigned long align_offset;
1915 };
1916 
1917 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1918 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1919 
1920 /*
1921  * Search for an unmapped address range.
1922  *
1923  * We are looking for a range that:
1924  * - does not intersect with any VMA;
1925  * - is contained within the [low_limit, high_limit) interval;
1926  * - is at least the desired size.
1927  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1928  */
1929 static inline unsigned long
1930 vm_unmapped_area(struct vm_unmapped_area_info *info)
1931 {
1932         if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1933                 return unmapped_area_topdown(info);
1934         else
1935                 return unmapped_area(info);
1936 }
1937 
1938 /* truncate.c */
1939 extern void truncate_inode_pages(struct address_space *, loff_t);
1940 extern void truncate_inode_pages_range(struct address_space *,
1941                                        loff_t lstart, loff_t lend);
1942 extern void truncate_inode_pages_final(struct address_space *);
1943 
1944 /* generic vm_area_ops exported for stackable file systems */
1945 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1946 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1947 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1948 
1949 /* mm/page-writeback.c */
1950 int write_one_page(struct page *page, int wait);
1951 void task_dirty_inc(struct task_struct *tsk);
1952 
1953 /* readahead.c */
1954 #define VM_MAX_READAHEAD        128     /* kbytes */
1955 #define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1956 
1957 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1958                         pgoff_t offset, unsigned long nr_to_read);
1959 
1960 void page_cache_sync_readahead(struct address_space *mapping,
1961                                struct file_ra_state *ra,
1962                                struct file *filp,
1963                                pgoff_t offset,
1964                                unsigned long size);
1965 
1966 void page_cache_async_readahead(struct address_space *mapping,
1967                                 struct file_ra_state *ra,
1968                                 struct file *filp,
1969                                 struct page *pg,
1970                                 pgoff_t offset,
1971                                 unsigned long size);
1972 
1973 unsigned long max_sane_readahead(unsigned long nr);
1974 
1975 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1976 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1977 
1978 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1979 extern int expand_downwards(struct vm_area_struct *vma,
1980                 unsigned long address);
1981 #if VM_GROWSUP
1982 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1983 #else
1984   #define expand_upwards(vma, address) (0)
1985 #endif
1986 
1987 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1988 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1989 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1990                                              struct vm_area_struct **pprev);
1991 
1992 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1993    NULL if none.  Assume start_addr < end_addr. */
1994 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1995 {
1996         struct vm_area_struct * vma = find_vma(mm,start_addr);
1997 
1998         if (vma && end_addr <= vma->vm_start)
1999                 vma = NULL;
2000         return vma;
2001 }
2002 
2003 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2004 {
2005         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2006 }
2007 
2008 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2009 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2010                                 unsigned long vm_start, unsigned long vm_end)
2011 {
2012         struct vm_area_struct *vma = find_vma(mm, vm_start);
2013 
2014         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2015                 vma = NULL;
2016 
2017         return vma;
2018 }
2019 
2020 #ifdef CONFIG_MMU
2021 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2022 void vma_set_page_prot(struct vm_area_struct *vma);
2023 #else
2024 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2025 {
2026         return __pgprot(0);
2027 }
2028 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2029 {
2030         vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2031 }
2032 #endif
2033 
2034 #ifdef CONFIG_NUMA_BALANCING
2035 unsigned long change_prot_numa(struct vm_area_struct *vma,
2036                         unsigned long start, unsigned long end);
2037 #endif
2038 
2039 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2040 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2041                         unsigned long pfn, unsigned long size, pgprot_t);
2042 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2043 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2044                         unsigned long pfn);
2045 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2046                         unsigned long pfn);
2047 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2048 
2049 
2050 struct page *follow_page_mask(struct vm_area_struct *vma,
2051                               unsigned long address, unsigned int foll_flags,
2052                               unsigned int *page_mask);
2053 
2054 static inline struct page *follow_page(struct vm_area_struct *vma,
2055                 unsigned long address, unsigned int foll_flags)
2056 {
2057         unsigned int unused_page_mask;
2058         return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2059 }
2060 
2061 #define FOLL_WRITE      0x01    /* check pte is writable */
2062 #define FOLL_TOUCH      0x02    /* mark page accessed */
2063 #define FOLL_GET        0x04    /* do get_page on page */
2064 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2065 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2066 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2067                                  * and return without waiting upon it */
2068 #define FOLL_POPULATE   0x40    /* fault in page */
2069 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2070 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2071 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2072 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2073 #define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2074 
2075 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2076                         void *data);
2077 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2078                                unsigned long size, pte_fn_t fn, void *data);
2079 
2080 #ifdef CONFIG_PROC_FS
2081 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2082 #else
2083 static inline void vm_stat_account(struct mm_struct *mm,
2084                         unsigned long flags, struct file *file, long pages)
2085 {
2086         mm->total_vm += pages;
2087 }
2088 #endif /* CONFIG_PROC_FS */
2089 
2090 #ifdef CONFIG_DEBUG_PAGEALLOC
2091 extern bool _debug_pagealloc_enabled;
2092 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2093 
2094 static inline bool debug_pagealloc_enabled(void)
2095 {
2096         return _debug_pagealloc_enabled;
2097 }
2098 
2099 static inline void
2100 kernel_map_pages(struct page *page, int numpages, int enable)
2101 {
2102         if (!debug_pagealloc_enabled())
2103                 return;
2104 
2105         __kernel_map_pages(page, numpages, enable);
2106 }
2107 #ifdef CONFIG_HIBERNATION
2108 extern bool kernel_page_present(struct page *page);
2109 #endif /* CONFIG_HIBERNATION */
2110 #else
2111 static inline void
2112 kernel_map_pages(struct page *page, int numpages, int enable) {}
2113 #ifdef CONFIG_HIBERNATION
2114 static inline bool kernel_page_present(struct page *page) { return true; }
2115 #endif /* CONFIG_HIBERNATION */
2116 #endif
2117 
2118 #ifdef __HAVE_ARCH_GATE_AREA
2119 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2120 extern int in_gate_area_no_mm(unsigned long addr);
2121 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2122 #else
2123 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2124 {
2125         return NULL;
2126 }
2127 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2128 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2129 {
2130         return 0;
2131 }
2132 #endif  /* __HAVE_ARCH_GATE_AREA */
2133 
2134 #ifdef CONFIG_SYSCTL
2135 extern int sysctl_drop_caches;
2136 int drop_caches_sysctl_handler(struct ctl_table *, int,
2137                                         void __user *, size_t *, loff_t *);
2138 #endif
2139 
2140 void drop_slab(void);
2141 void drop_slab_node(int nid);
2142 
2143 #ifndef CONFIG_MMU
2144 #define randomize_va_space 0
2145 #else
2146 extern int randomize_va_space;
2147 #endif
2148 
2149 const char * arch_vma_name(struct vm_area_struct *vma);
2150 void print_vma_addr(char *prefix, unsigned long rip);
2151 
2152 void sparse_mem_maps_populate_node(struct page **map_map,
2153                                    unsigned long pnum_begin,
2154                                    unsigned long pnum_end,
2155                                    unsigned long map_count,
2156                                    int nodeid);
2157 
2158 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2159 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2160 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2161 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2162 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2163 void *vmemmap_alloc_block(unsigned long size, int node);
2164 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2165 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2166 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2167                                int node);
2168 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2169 void vmemmap_populate_print_last(void);
2170 #ifdef CONFIG_MEMORY_HOTPLUG
2171 void vmemmap_free(unsigned long start, unsigned long end);
2172 #endif
2173 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2174                                   unsigned long size);
2175 
2176 enum mf_flags {
2177         MF_COUNT_INCREASED = 1 << 0,
2178         MF_ACTION_REQUIRED = 1 << 1,
2179         MF_MUST_KILL = 1 << 2,
2180         MF_SOFT_OFFLINE = 1 << 3,
2181 };
2182 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2183 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2184 extern int unpoison_memory(unsigned long pfn);
2185 extern int get_hwpoison_page(struct page *page);
2186 extern int sysctl_memory_failure_early_kill;
2187 extern int sysctl_memory_failure_recovery;
2188 extern void shake_page(struct page *p, int access);
2189 extern atomic_long_t num_poisoned_pages;
2190 extern int soft_offline_page(struct page *page, int flags);
2191 
2192 
2193 /*
2194  * Error handlers for various types of pages.
2195  */
2196 enum mf_result {
2197         MF_IGNORED,     /* Error: cannot be handled */
2198         MF_FAILED,      /* Error: handling failed */
2199         MF_DELAYED,     /* Will be handled later */
2200         MF_RECOVERED,   /* Successfully recovered */
2201 };
2202 
2203 enum mf_action_page_type {
2204         MF_MSG_KERNEL,
2205         MF_MSG_KERNEL_HIGH_ORDER,
2206         MF_MSG_SLAB,
2207         MF_MSG_DIFFERENT_COMPOUND,
2208         MF_MSG_POISONED_HUGE,
2209         MF_MSG_HUGE,
2210         MF_MSG_FREE_HUGE,
2211         MF_MSG_UNMAP_FAILED,
2212         MF_MSG_DIRTY_SWAPCACHE,
2213         MF_MSG_CLEAN_SWAPCACHE,
2214         MF_MSG_DIRTY_MLOCKED_LRU,
2215         MF_MSG_CLEAN_MLOCKED_LRU,
2216         MF_MSG_DIRTY_UNEVICTABLE_LRU,
2217         MF_MSG_CLEAN_UNEVICTABLE_LRU,
2218         MF_MSG_DIRTY_LRU,
2219         MF_MSG_CLEAN_LRU,
2220         MF_MSG_TRUNCATED_LRU,
2221         MF_MSG_BUDDY,
2222         MF_MSG_BUDDY_2ND,
2223         MF_MSG_UNKNOWN,
2224 };
2225 
2226 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2227 extern void clear_huge_page(struct page *page,
2228                             unsigned long addr,
2229                             unsigned int pages_per_huge_page);
2230 extern void copy_user_huge_page(struct page *dst, struct page *src,
2231                                 unsigned long addr, struct vm_area_struct *vma,
2232                                 unsigned int pages_per_huge_page);
2233 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2234 
2235 extern struct page_ext_operations debug_guardpage_ops;
2236 extern struct page_ext_operations page_poisoning_ops;
2237 
2238 #ifdef CONFIG_DEBUG_PAGEALLOC
2239 extern unsigned int _debug_guardpage_minorder;
2240 extern bool _debug_guardpage_enabled;
2241 
2242 static inline unsigned int debug_guardpage_minorder(void)
2243 {
2244         return _debug_guardpage_minorder;
2245 }
2246 
2247 static inline bool debug_guardpage_enabled(void)
2248 {
2249         return _debug_guardpage_enabled;
2250 }
2251 
2252 static inline bool page_is_guard(struct page *page)
2253 {
2254         struct page_ext *page_ext;
2255 
2256         if (!debug_guardpage_enabled())
2257                 return false;
2258 
2259         page_ext = lookup_page_ext(page);
2260         return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2261 }
2262 #else
2263 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2264 static inline bool debug_guardpage_enabled(void) { return false; }
2265 static inline bool page_is_guard(struct page *page) { return false; }
2266 #endif /* CONFIG_DEBUG_PAGEALLOC */
2267 
2268 #if MAX_NUMNODES > 1
2269 void __init setup_nr_node_ids(void);
2270 #else
2271 static inline void setup_nr_node_ids(void) {}
2272 #endif
2273 
2274 #endif /* __KERNEL__ */
2275 #endif /* _LINUX_MM_H */
2276 

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