Version:  2.0.40 2.2.26 2.4.37 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 4.3 4.4

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 #include <linux/err.h>
 24 
 25 struct mempolicy;
 26 struct anon_vma;
 27 struct anon_vma_chain;
 28 struct file_ra_state;
 29 struct user_struct;
 30 struct writeback_control;
 31 struct bdi_writeback;
 32 
 33 #ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
 34 extern unsigned long max_mapnr;
 35 
 36 static inline void set_max_mapnr(unsigned long limit)
 37 {
 38         max_mapnr = limit;
 39 }
 40 #else
 41 static inline void set_max_mapnr(unsigned long limit) { }
 42 #endif
 43 
 44 extern unsigned long totalram_pages;
 45 extern void * high_memory;
 46 extern int page_cluster;
 47 
 48 #ifdef CONFIG_SYSCTL
 49 extern int sysctl_legacy_va_layout;
 50 #else
 51 #define sysctl_legacy_va_layout 0
 52 #endif
 53 
 54 #include <asm/page.h>
 55 #include <asm/pgtable.h>
 56 #include <asm/processor.h>
 57 
 58 #ifndef __pa_symbol
 59 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 60 #endif
 61 
 62 /*
 63  * To prevent common memory management code establishing
 64  * a zero page mapping on a read fault.
 65  * This macro should be defined within <asm/pgtable.h>.
 66  * s390 does this to prevent multiplexing of hardware bits
 67  * related to the physical page in case of virtualization.
 68  */
 69 #ifndef mm_forbids_zeropage
 70 #define mm_forbids_zeropage(X)  (0)
 71 #endif
 72 
 73 extern unsigned long sysctl_user_reserve_kbytes;
 74 extern unsigned long sysctl_admin_reserve_kbytes;
 75 
 76 extern int sysctl_overcommit_memory;
 77 extern int sysctl_overcommit_ratio;
 78 extern unsigned long sysctl_overcommit_kbytes;
 79 
 80 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
 81                                     size_t *, loff_t *);
 82 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
 83                                     size_t *, loff_t *);
 84 
 85 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
 86 
 87 /* to align the pointer to the (next) page boundary */
 88 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
 89 
 90 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
 91 #define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
 92 
 93 /*
 94  * Linux kernel virtual memory manager primitives.
 95  * The idea being to have a "virtual" mm in the same way
 96  * we have a virtual fs - giving a cleaner interface to the
 97  * mm details, and allowing different kinds of memory mappings
 98  * (from shared memory to executable loading to arbitrary
 99  * mmap() functions).
100  */
101 
102 extern struct kmem_cache *vm_area_cachep;
103 
104 #ifndef CONFIG_MMU
105 extern struct rb_root nommu_region_tree;
106 extern struct rw_semaphore nommu_region_sem;
107 
108 extern unsigned int kobjsize(const void *objp);
109 #endif
110 
111 /*
112  * vm_flags in vm_area_struct, see mm_types.h.
113  */
114 #define VM_NONE         0x00000000
115 
116 #define VM_READ         0x00000001      /* currently active flags */
117 #define VM_WRITE        0x00000002
118 #define VM_EXEC         0x00000004
119 #define VM_SHARED       0x00000008
120 
121 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
122 #define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
123 #define VM_MAYWRITE     0x00000020
124 #define VM_MAYEXEC      0x00000040
125 #define VM_MAYSHARE     0x00000080
126 
127 #define VM_GROWSDOWN    0x00000100      /* general info on the segment */
128 #define VM_UFFD_MISSING 0x00000200      /* missing pages tracking */
129 #define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
130 #define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
131 #define VM_UFFD_WP      0x00001000      /* wrprotect pages tracking */
132 
133 #define VM_LOCKED       0x00002000
134 #define VM_IO           0x00004000      /* Memory mapped I/O or similar */
135 
136                                         /* Used by sys_madvise() */
137 #define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
138 #define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
139 
140 #define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
141 #define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
142 #define VM_LOCKONFAULT  0x00080000      /* Lock the pages covered when they are faulted in */
143 #define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
144 #define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
145 #define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
146 #define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
147 #define VM_ARCH_2       0x02000000
148 #define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
149 
150 #ifdef CONFIG_MEM_SOFT_DIRTY
151 # define VM_SOFTDIRTY   0x08000000      /* Not soft dirty clean area */
152 #else
153 # define VM_SOFTDIRTY   0
154 #endif
155 
156 #define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
157 #define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
158 #define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
159 #define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
160 
161 #if defined(CONFIG_X86)
162 # define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
163 #elif defined(CONFIG_PPC)
164 # define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
165 #elif defined(CONFIG_PARISC)
166 # define VM_GROWSUP     VM_ARCH_1
167 #elif defined(CONFIG_METAG)
168 # define VM_GROWSUP     VM_ARCH_1
169 #elif defined(CONFIG_IA64)
170 # define VM_GROWSUP     VM_ARCH_1
171 #elif !defined(CONFIG_MMU)
172 # define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
173 #endif
174 
175 #if defined(CONFIG_X86)
176 /* MPX specific bounds table or bounds directory */
177 # define VM_MPX         VM_ARCH_2
178 #endif
179 
180 #ifndef VM_GROWSUP
181 # define VM_GROWSUP     VM_NONE
182 #endif
183 
184 /* Bits set in the VMA until the stack is in its final location */
185 #define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
186 
187 #ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
188 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
189 #endif
190 
191 #ifdef CONFIG_STACK_GROWSUP
192 #define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
193 #else
194 #define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
195 #endif
196 
197 /*
198  * Special vmas that are non-mergable, non-mlock()able.
199  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
200  */
201 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
202 
203 /* This mask defines which mm->def_flags a process can inherit its parent */
204 #define VM_INIT_DEF_MASK        VM_NOHUGEPAGE
205 
206 /* This mask is used to clear all the VMA flags used by mlock */
207 #define VM_LOCKED_CLEAR_MASK    (~(VM_LOCKED | VM_LOCKONFAULT))
208 
209 /*
210  * mapping from the currently active vm_flags protection bits (the
211  * low four bits) to a page protection mask..
212  */
213 extern pgprot_t protection_map[16];
214 
215 #define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
216 #define FAULT_FLAG_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
217 #define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
218 #define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
219 #define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
220 #define FAULT_FLAG_TRIED        0x20    /* Second try */
221 #define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
222 
223 /*
224  * vm_fault is filled by the the pagefault handler and passed to the vma's
225  * ->fault function. The vma's ->fault is responsible for returning a bitmask
226  * of VM_FAULT_xxx flags that give details about how the fault was handled.
227  *
228  * pgoff should be used in favour of virtual_address, if possible.
229  */
230 struct vm_fault {
231         unsigned int flags;             /* FAULT_FLAG_xxx flags */
232         pgoff_t pgoff;                  /* Logical page offset based on vma */
233         void __user *virtual_address;   /* Faulting virtual address */
234 
235         struct page *cow_page;          /* Handler may choose to COW */
236         struct page *page;              /* ->fault handlers should return a
237                                          * page here, unless VM_FAULT_NOPAGE
238                                          * is set (which is also implied by
239                                          * VM_FAULT_ERROR).
240                                          */
241         /* for ->map_pages() only */
242         pgoff_t max_pgoff;              /* map pages for offset from pgoff till
243                                          * max_pgoff inclusive */
244         pte_t *pte;                     /* pte entry associated with ->pgoff */
245 };
246 
247 /*
248  * These are the virtual MM functions - opening of an area, closing and
249  * unmapping it (needed to keep files on disk up-to-date etc), pointer
250  * to the functions called when a no-page or a wp-page exception occurs. 
251  */
252 struct vm_operations_struct {
253         void (*open)(struct vm_area_struct * area);
254         void (*close)(struct vm_area_struct * area);
255         int (*mremap)(struct vm_area_struct * area);
256         int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
257         int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
258                                                 pmd_t *, unsigned int flags);
259         void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
260 
261         /* notification that a previously read-only page is about to become
262          * writable, if an error is returned it will cause a SIGBUS */
263         int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
264 
265         /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
266         int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
267 
268         /* called by access_process_vm when get_user_pages() fails, typically
269          * for use by special VMAs that can switch between memory and hardware
270          */
271         int (*access)(struct vm_area_struct *vma, unsigned long addr,
272                       void *buf, int len, int write);
273 
274         /* Called by the /proc/PID/maps code to ask the vma whether it
275          * has a special name.  Returning non-NULL will also cause this
276          * vma to be dumped unconditionally. */
277         const char *(*name)(struct vm_area_struct *vma);
278 
279 #ifdef CONFIG_NUMA
280         /*
281          * set_policy() op must add a reference to any non-NULL @new mempolicy
282          * to hold the policy upon return.  Caller should pass NULL @new to
283          * remove a policy and fall back to surrounding context--i.e. do not
284          * install a MPOL_DEFAULT policy, nor the task or system default
285          * mempolicy.
286          */
287         int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
288 
289         /*
290          * get_policy() op must add reference [mpol_get()] to any policy at
291          * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
292          * in mm/mempolicy.c will do this automatically.
293          * get_policy() must NOT add a ref if the policy at (vma,addr) is not
294          * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
295          * If no [shared/vma] mempolicy exists at the addr, get_policy() op
296          * must return NULL--i.e., do not "fallback" to task or system default
297          * policy.
298          */
299         struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
300                                         unsigned long addr);
301 #endif
302         /*
303          * Called by vm_normal_page() for special PTEs to find the
304          * page for @addr.  This is useful if the default behavior
305          * (using pte_page()) would not find the correct page.
306          */
307         struct page *(*find_special_page)(struct vm_area_struct *vma,
308                                           unsigned long addr);
309 };
310 
311 struct mmu_gather;
312 struct inode;
313 
314 #define page_private(page)              ((page)->private)
315 #define set_page_private(page, v)       ((page)->private = (v))
316 
317 /*
318  * FIXME: take this include out, include page-flags.h in
319  * files which need it (119 of them)
320  */
321 #include <linux/page-flags.h>
322 #include <linux/huge_mm.h>
323 
324 /*
325  * Methods to modify the page usage count.
326  *
327  * What counts for a page usage:
328  * - cache mapping   (page->mapping)
329  * - private data    (page->private)
330  * - page mapped in a task's page tables, each mapping
331  *   is counted separately
332  *
333  * Also, many kernel routines increase the page count before a critical
334  * routine so they can be sure the page doesn't go away from under them.
335  */
336 
337 /*
338  * Drop a ref, return true if the refcount fell to zero (the page has no users)
339  */
340 static inline int put_page_testzero(struct page *page)
341 {
342         VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
343         return atomic_dec_and_test(&page->_count);
344 }
345 
346 /*
347  * Try to grab a ref unless the page has a refcount of zero, return false if
348  * that is the case.
349  * This can be called when MMU is off so it must not access
350  * any of the virtual mappings.
351  */
352 static inline int get_page_unless_zero(struct page *page)
353 {
354         return atomic_inc_not_zero(&page->_count);
355 }
356 
357 extern int page_is_ram(unsigned long pfn);
358 
359 enum {
360         REGION_INTERSECTS,
361         REGION_DISJOINT,
362         REGION_MIXED,
363 };
364 
365 int region_intersects(resource_size_t offset, size_t size, const char *type);
366 
367 /* Support for virtually mapped pages */
368 struct page *vmalloc_to_page(const void *addr);
369 unsigned long vmalloc_to_pfn(const void *addr);
370 
371 /*
372  * Determine if an address is within the vmalloc range
373  *
374  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
375  * is no special casing required.
376  */
377 static inline int is_vmalloc_addr(const void *x)
378 {
379 #ifdef CONFIG_MMU
380         unsigned long addr = (unsigned long)x;
381 
382         return addr >= VMALLOC_START && addr < VMALLOC_END;
383 #else
384         return 0;
385 #endif
386 }
387 #ifdef CONFIG_MMU
388 extern int is_vmalloc_or_module_addr(const void *x);
389 #else
390 static inline int is_vmalloc_or_module_addr(const void *x)
391 {
392         return 0;
393 }
394 #endif
395 
396 extern void kvfree(const void *addr);
397 
398 static inline void compound_lock(struct page *page)
399 {
400 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
401         VM_BUG_ON_PAGE(PageSlab(page), page);
402         bit_spin_lock(PG_compound_lock, &page->flags);
403 #endif
404 }
405 
406 static inline void compound_unlock(struct page *page)
407 {
408 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
409         VM_BUG_ON_PAGE(PageSlab(page), page);
410         bit_spin_unlock(PG_compound_lock, &page->flags);
411 #endif
412 }
413 
414 static inline unsigned long compound_lock_irqsave(struct page *page)
415 {
416         unsigned long uninitialized_var(flags);
417 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
418         local_irq_save(flags);
419         compound_lock(page);
420 #endif
421         return flags;
422 }
423 
424 static inline void compound_unlock_irqrestore(struct page *page,
425                                               unsigned long flags)
426 {
427 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
428         compound_unlock(page);
429         local_irq_restore(flags);
430 #endif
431 }
432 
433 /*
434  * The atomic page->_mapcount, starts from -1: so that transitions
435  * both from it and to it can be tracked, using atomic_inc_and_test
436  * and atomic_add_negative(-1).
437  */
438 static inline void page_mapcount_reset(struct page *page)
439 {
440         atomic_set(&(page)->_mapcount, -1);
441 }
442 
443 static inline int page_mapcount(struct page *page)
444 {
445         VM_BUG_ON_PAGE(PageSlab(page), page);
446         return atomic_read(&page->_mapcount) + 1;
447 }
448 
449 static inline int page_count(struct page *page)
450 {
451         return atomic_read(&compound_head(page)->_count);
452 }
453 
454 static inline bool __compound_tail_refcounted(struct page *page)
455 {
456         return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
457 }
458 
459 /*
460  * This takes a head page as parameter and tells if the
461  * tail page reference counting can be skipped.
462  *
463  * For this to be safe, PageSlab and PageHeadHuge must remain true on
464  * any given page where they return true here, until all tail pins
465  * have been released.
466  */
467 static inline bool compound_tail_refcounted(struct page *page)
468 {
469         VM_BUG_ON_PAGE(!PageHead(page), page);
470         return __compound_tail_refcounted(page);
471 }
472 
473 static inline void get_huge_page_tail(struct page *page)
474 {
475         /*
476          * __split_huge_page_refcount() cannot run from under us.
477          */
478         VM_BUG_ON_PAGE(!PageTail(page), page);
479         VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
480         VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
481         if (compound_tail_refcounted(compound_head(page)))
482                 atomic_inc(&page->_mapcount);
483 }
484 
485 extern bool __get_page_tail(struct page *page);
486 
487 static inline void get_page(struct page *page)
488 {
489         if (unlikely(PageTail(page)))
490                 if (likely(__get_page_tail(page)))
491                         return;
492         /*
493          * Getting a normal page or the head of a compound page
494          * requires to already have an elevated page->_count.
495          */
496         VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
497         atomic_inc(&page->_count);
498 }
499 
500 static inline struct page *virt_to_head_page(const void *x)
501 {
502         struct page *page = virt_to_page(x);
503 
504         return compound_head(page);
505 }
506 
507 /*
508  * Setup the page count before being freed into the page allocator for
509  * the first time (boot or memory hotplug)
510  */
511 static inline void init_page_count(struct page *page)
512 {
513         atomic_set(&page->_count, 1);
514 }
515 
516 void put_page(struct page *page);
517 void put_pages_list(struct list_head *pages);
518 
519 void split_page(struct page *page, unsigned int order);
520 int split_free_page(struct page *page);
521 
522 /*
523  * Compound pages have a destructor function.  Provide a
524  * prototype for that function and accessor functions.
525  * These are _only_ valid on the head of a compound page.
526  */
527 typedef void compound_page_dtor(struct page *);
528 
529 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
530 enum compound_dtor_id {
531         NULL_COMPOUND_DTOR,
532         COMPOUND_PAGE_DTOR,
533 #ifdef CONFIG_HUGETLB_PAGE
534         HUGETLB_PAGE_DTOR,
535 #endif
536         NR_COMPOUND_DTORS,
537 };
538 extern compound_page_dtor * const compound_page_dtors[];
539 
540 static inline void set_compound_page_dtor(struct page *page,
541                 enum compound_dtor_id compound_dtor)
542 {
543         VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
544         page[1].compound_dtor = compound_dtor;
545 }
546 
547 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
548 {
549         VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
550         return compound_page_dtors[page[1].compound_dtor];
551 }
552 
553 static inline unsigned int compound_order(struct page *page)
554 {
555         if (!PageHead(page))
556                 return 0;
557         return page[1].compound_order;
558 }
559 
560 static inline void set_compound_order(struct page *page, unsigned int order)
561 {
562         page[1].compound_order = order;
563 }
564 
565 #ifdef CONFIG_MMU
566 /*
567  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
568  * servicing faults for write access.  In the normal case, do always want
569  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
570  * that do not have writing enabled, when used by access_process_vm.
571  */
572 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
573 {
574         if (likely(vma->vm_flags & VM_WRITE))
575                 pte = pte_mkwrite(pte);
576         return pte;
577 }
578 
579 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
580                 struct page *page, pte_t *pte, bool write, bool anon);
581 #endif
582 
583 /*
584  * Multiple processes may "see" the same page. E.g. for untouched
585  * mappings of /dev/null, all processes see the same page full of
586  * zeroes, and text pages of executables and shared libraries have
587  * only one copy in memory, at most, normally.
588  *
589  * For the non-reserved pages, page_count(page) denotes a reference count.
590  *   page_count() == 0 means the page is free. page->lru is then used for
591  *   freelist management in the buddy allocator.
592  *   page_count() > 0  means the page has been allocated.
593  *
594  * Pages are allocated by the slab allocator in order to provide memory
595  * to kmalloc and kmem_cache_alloc. In this case, the management of the
596  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
597  * unless a particular usage is carefully commented. (the responsibility of
598  * freeing the kmalloc memory is the caller's, of course).
599  *
600  * A page may be used by anyone else who does a __get_free_page().
601  * In this case, page_count still tracks the references, and should only
602  * be used through the normal accessor functions. The top bits of page->flags
603  * and page->virtual store page management information, but all other fields
604  * are unused and could be used privately, carefully. The management of this
605  * page is the responsibility of the one who allocated it, and those who have
606  * subsequently been given references to it.
607  *
608  * The other pages (we may call them "pagecache pages") are completely
609  * managed by the Linux memory manager: I/O, buffers, swapping etc.
610  * The following discussion applies only to them.
611  *
612  * A pagecache page contains an opaque `private' member, which belongs to the
613  * page's address_space. Usually, this is the address of a circular list of
614  * the page's disk buffers. PG_private must be set to tell the VM to call
615  * into the filesystem to release these pages.
616  *
617  * A page may belong to an inode's memory mapping. In this case, page->mapping
618  * is the pointer to the inode, and page->index is the file offset of the page,
619  * in units of PAGE_CACHE_SIZE.
620  *
621  * If pagecache pages are not associated with an inode, they are said to be
622  * anonymous pages. These may become associated with the swapcache, and in that
623  * case PG_swapcache is set, and page->private is an offset into the swapcache.
624  *
625  * In either case (swapcache or inode backed), the pagecache itself holds one
626  * reference to the page. Setting PG_private should also increment the
627  * refcount. The each user mapping also has a reference to the page.
628  *
629  * The pagecache pages are stored in a per-mapping radix tree, which is
630  * rooted at mapping->page_tree, and indexed by offset.
631  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
632  * lists, we instead now tag pages as dirty/writeback in the radix tree.
633  *
634  * All pagecache pages may be subject to I/O:
635  * - inode pages may need to be read from disk,
636  * - inode pages which have been modified and are MAP_SHARED may need
637  *   to be written back to the inode on disk,
638  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
639  *   modified may need to be swapped out to swap space and (later) to be read
640  *   back into memory.
641  */
642 
643 /*
644  * The zone field is never updated after free_area_init_core()
645  * sets it, so none of the operations on it need to be atomic.
646  */
647 
648 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
649 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
650 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
651 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
652 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
653 
654 /*
655  * Define the bit shifts to access each section.  For non-existent
656  * sections we define the shift as 0; that plus a 0 mask ensures
657  * the compiler will optimise away reference to them.
658  */
659 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
660 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
661 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
662 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
663 
664 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
665 #ifdef NODE_NOT_IN_PAGE_FLAGS
666 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
667 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
668                                                 SECTIONS_PGOFF : ZONES_PGOFF)
669 #else
670 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
671 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
672                                                 NODES_PGOFF : ZONES_PGOFF)
673 #endif
674 
675 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
676 
677 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
678 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
679 #endif
680 
681 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
682 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
683 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
684 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
685 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
686 
687 static inline enum zone_type page_zonenum(const struct page *page)
688 {
689         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
690 }
691 
692 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
693 #define SECTION_IN_PAGE_FLAGS
694 #endif
695 
696 /*
697  * The identification function is mainly used by the buddy allocator for
698  * determining if two pages could be buddies. We are not really identifying
699  * the zone since we could be using the section number id if we do not have
700  * node id available in page flags.
701  * We only guarantee that it will return the same value for two combinable
702  * pages in a zone.
703  */
704 static inline int page_zone_id(struct page *page)
705 {
706         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
707 }
708 
709 static inline int zone_to_nid(struct zone *zone)
710 {
711 #ifdef CONFIG_NUMA
712         return zone->node;
713 #else
714         return 0;
715 #endif
716 }
717 
718 #ifdef NODE_NOT_IN_PAGE_FLAGS
719 extern int page_to_nid(const struct page *page);
720 #else
721 static inline int page_to_nid(const struct page *page)
722 {
723         return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
724 }
725 #endif
726 
727 #ifdef CONFIG_NUMA_BALANCING
728 static inline int cpu_pid_to_cpupid(int cpu, int pid)
729 {
730         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
731 }
732 
733 static inline int cpupid_to_pid(int cpupid)
734 {
735         return cpupid & LAST__PID_MASK;
736 }
737 
738 static inline int cpupid_to_cpu(int cpupid)
739 {
740         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
741 }
742 
743 static inline int cpupid_to_nid(int cpupid)
744 {
745         return cpu_to_node(cpupid_to_cpu(cpupid));
746 }
747 
748 static inline bool cpupid_pid_unset(int cpupid)
749 {
750         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
751 }
752 
753 static inline bool cpupid_cpu_unset(int cpupid)
754 {
755         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
756 }
757 
758 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
759 {
760         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
761 }
762 
763 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
764 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
765 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
766 {
767         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
768 }
769 
770 static inline int page_cpupid_last(struct page *page)
771 {
772         return page->_last_cpupid;
773 }
774 static inline void page_cpupid_reset_last(struct page *page)
775 {
776         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
777 }
778 #else
779 static inline int page_cpupid_last(struct page *page)
780 {
781         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
782 }
783 
784 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
785 
786 static inline void page_cpupid_reset_last(struct page *page)
787 {
788         int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
789 
790         page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
791         page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
792 }
793 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
794 #else /* !CONFIG_NUMA_BALANCING */
795 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
796 {
797         return page_to_nid(page); /* XXX */
798 }
799 
800 static inline int page_cpupid_last(struct page *page)
801 {
802         return page_to_nid(page); /* XXX */
803 }
804 
805 static inline int cpupid_to_nid(int cpupid)
806 {
807         return -1;
808 }
809 
810 static inline int cpupid_to_pid(int cpupid)
811 {
812         return -1;
813 }
814 
815 static inline int cpupid_to_cpu(int cpupid)
816 {
817         return -1;
818 }
819 
820 static inline int cpu_pid_to_cpupid(int nid, int pid)
821 {
822         return -1;
823 }
824 
825 static inline bool cpupid_pid_unset(int cpupid)
826 {
827         return 1;
828 }
829 
830 static inline void page_cpupid_reset_last(struct page *page)
831 {
832 }
833 
834 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
835 {
836         return false;
837 }
838 #endif /* CONFIG_NUMA_BALANCING */
839 
840 static inline struct zone *page_zone(const struct page *page)
841 {
842         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
843 }
844 
845 #ifdef SECTION_IN_PAGE_FLAGS
846 static inline void set_page_section(struct page *page, unsigned long section)
847 {
848         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
849         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
850 }
851 
852 static inline unsigned long page_to_section(const struct page *page)
853 {
854         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
855 }
856 #endif
857 
858 static inline void set_page_zone(struct page *page, enum zone_type zone)
859 {
860         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
861         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
862 }
863 
864 static inline void set_page_node(struct page *page, unsigned long node)
865 {
866         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
867         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
868 }
869 
870 static inline void set_page_links(struct page *page, enum zone_type zone,
871         unsigned long node, unsigned long pfn)
872 {
873         set_page_zone(page, zone);
874         set_page_node(page, node);
875 #ifdef SECTION_IN_PAGE_FLAGS
876         set_page_section(page, pfn_to_section_nr(pfn));
877 #endif
878 }
879 
880 #ifdef CONFIG_MEMCG
881 static inline struct mem_cgroup *page_memcg(struct page *page)
882 {
883         return page->mem_cgroup;
884 }
885 
886 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
887 {
888         page->mem_cgroup = memcg;
889 }
890 #else
891 static inline struct mem_cgroup *page_memcg(struct page *page)
892 {
893         return NULL;
894 }
895 
896 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
897 {
898 }
899 #endif
900 
901 /*
902  * Some inline functions in vmstat.h depend on page_zone()
903  */
904 #include <linux/vmstat.h>
905 
906 static __always_inline void *lowmem_page_address(const struct page *page)
907 {
908         return __va(PFN_PHYS(page_to_pfn(page)));
909 }
910 
911 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
912 #define HASHED_PAGE_VIRTUAL
913 #endif
914 
915 #if defined(WANT_PAGE_VIRTUAL)
916 static inline void *page_address(const struct page *page)
917 {
918         return page->virtual;
919 }
920 static inline void set_page_address(struct page *page, void *address)
921 {
922         page->virtual = address;
923 }
924 #define page_address_init()  do { } while(0)
925 #endif
926 
927 #if defined(HASHED_PAGE_VIRTUAL)
928 void *page_address(const struct page *page);
929 void set_page_address(struct page *page, void *virtual);
930 void page_address_init(void);
931 #endif
932 
933 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
934 #define page_address(page) lowmem_page_address(page)
935 #define set_page_address(page, address)  do { } while(0)
936 #define page_address_init()  do { } while(0)
937 #endif
938 
939 extern void *page_rmapping(struct page *page);
940 extern struct anon_vma *page_anon_vma(struct page *page);
941 extern struct address_space *page_mapping(struct page *page);
942 
943 extern struct address_space *__page_file_mapping(struct page *);
944 
945 static inline
946 struct address_space *page_file_mapping(struct page *page)
947 {
948         if (unlikely(PageSwapCache(page)))
949                 return __page_file_mapping(page);
950 
951         return page->mapping;
952 }
953 
954 /*
955  * Return the pagecache index of the passed page.  Regular pagecache pages
956  * use ->index whereas swapcache pages use ->private
957  */
958 static inline pgoff_t page_index(struct page *page)
959 {
960         if (unlikely(PageSwapCache(page)))
961                 return page_private(page);
962         return page->index;
963 }
964 
965 extern pgoff_t __page_file_index(struct page *page);
966 
967 /*
968  * Return the file index of the page. Regular pagecache pages use ->index
969  * whereas swapcache pages use swp_offset(->private)
970  */
971 static inline pgoff_t page_file_index(struct page *page)
972 {
973         if (unlikely(PageSwapCache(page)))
974                 return __page_file_index(page);
975 
976         return page->index;
977 }
978 
979 /*
980  * Return true if this page is mapped into pagetables.
981  */
982 static inline int page_mapped(struct page *page)
983 {
984         return atomic_read(&(page)->_mapcount) >= 0;
985 }
986 
987 /*
988  * Return true only if the page has been allocated with
989  * ALLOC_NO_WATERMARKS and the low watermark was not
990  * met implying that the system is under some pressure.
991  */
992 static inline bool page_is_pfmemalloc(struct page *page)
993 {
994         /*
995          * Page index cannot be this large so this must be
996          * a pfmemalloc page.
997          */
998         return page->index == -1UL;
999 }
1000 
1001 /*
1002  * Only to be called by the page allocator on a freshly allocated
1003  * page.
1004  */
1005 static inline void set_page_pfmemalloc(struct page *page)
1006 {
1007         page->index = -1UL;
1008 }
1009 
1010 static inline void clear_page_pfmemalloc(struct page *page)
1011 {
1012         page->index = 0;
1013 }
1014 
1015 /*
1016  * Different kinds of faults, as returned by handle_mm_fault().
1017  * Used to decide whether a process gets delivered SIGBUS or
1018  * just gets major/minor fault counters bumped up.
1019  */
1020 
1021 #define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
1022 
1023 #define VM_FAULT_OOM    0x0001
1024 #define VM_FAULT_SIGBUS 0x0002
1025 #define VM_FAULT_MAJOR  0x0004
1026 #define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
1027 #define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
1028 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1029 #define VM_FAULT_SIGSEGV 0x0040
1030 
1031 #define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
1032 #define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
1033 #define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
1034 #define VM_FAULT_FALLBACK 0x0800        /* huge page fault failed, fall back to small */
1035 
1036 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1037 
1038 #define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1039                          VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1040                          VM_FAULT_FALLBACK)
1041 
1042 /* Encode hstate index for a hwpoisoned large page */
1043 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1044 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1045 
1046 /*
1047  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1048  */
1049 extern void pagefault_out_of_memory(void);
1050 
1051 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1052 
1053 /*
1054  * Flags passed to show_mem() and show_free_areas() to suppress output in
1055  * various contexts.
1056  */
1057 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1058 
1059 extern void show_free_areas(unsigned int flags);
1060 extern bool skip_free_areas_node(unsigned int flags, int nid);
1061 
1062 int shmem_zero_setup(struct vm_area_struct *);
1063 #ifdef CONFIG_SHMEM
1064 bool shmem_mapping(struct address_space *mapping);
1065 #else
1066 static inline bool shmem_mapping(struct address_space *mapping)
1067 {
1068         return false;
1069 }
1070 #endif
1071 
1072 extern int can_do_mlock(void);
1073 extern int user_shm_lock(size_t, struct user_struct *);
1074 extern void user_shm_unlock(size_t, struct user_struct *);
1075 
1076 /*
1077  * Parameter block passed down to zap_pte_range in exceptional cases.
1078  */
1079 struct zap_details {
1080         struct address_space *check_mapping;    /* Check page->mapping if set */
1081         pgoff_t first_index;                    /* Lowest page->index to unmap */
1082         pgoff_t last_index;                     /* Highest page->index to unmap */
1083 };
1084 
1085 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1086                 pte_t pte);
1087 
1088 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1089                 unsigned long size);
1090 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1091                 unsigned long size, struct zap_details *);
1092 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1093                 unsigned long start, unsigned long end);
1094 
1095 /**
1096  * mm_walk - callbacks for walk_page_range
1097  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1098  *             this handler is required to be able to handle
1099  *             pmd_trans_huge() pmds.  They may simply choose to
1100  *             split_huge_page() instead of handling it explicitly.
1101  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1102  * @pte_hole: if set, called for each hole at all levels
1103  * @hugetlb_entry: if set, called for each hugetlb entry
1104  * @test_walk: caller specific callback function to determine whether
1105  *             we walk over the current vma or not. A positive returned
1106  *             value means "do page table walk over the current vma,"
1107  *             and a negative one means "abort current page table walk
1108  *             right now." 0 means "skip the current vma."
1109  * @mm:        mm_struct representing the target process of page table walk
1110  * @vma:       vma currently walked (NULL if walking outside vmas)
1111  * @private:   private data for callbacks' usage
1112  *
1113  * (see the comment on walk_page_range() for more details)
1114  */
1115 struct mm_walk {
1116         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1117                          unsigned long next, struct mm_walk *walk);
1118         int (*pte_entry)(pte_t *pte, unsigned long addr,
1119                          unsigned long next, struct mm_walk *walk);
1120         int (*pte_hole)(unsigned long addr, unsigned long next,
1121                         struct mm_walk *walk);
1122         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1123                              unsigned long addr, unsigned long next,
1124                              struct mm_walk *walk);
1125         int (*test_walk)(unsigned long addr, unsigned long next,
1126                         struct mm_walk *walk);
1127         struct mm_struct *mm;
1128         struct vm_area_struct *vma;
1129         void *private;
1130 };
1131 
1132 int walk_page_range(unsigned long addr, unsigned long end,
1133                 struct mm_walk *walk);
1134 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1135 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1136                 unsigned long end, unsigned long floor, unsigned long ceiling);
1137 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1138                         struct vm_area_struct *vma);
1139 void unmap_mapping_range(struct address_space *mapping,
1140                 loff_t const holebegin, loff_t const holelen, int even_cows);
1141 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1142         unsigned long *pfn);
1143 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1144                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1145 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1146                         void *buf, int len, int write);
1147 
1148 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1149                 loff_t const holebegin, loff_t const holelen)
1150 {
1151         unmap_mapping_range(mapping, holebegin, holelen, 0);
1152 }
1153 
1154 extern void truncate_pagecache(struct inode *inode, loff_t new);
1155 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1156 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1157 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1158 int truncate_inode_page(struct address_space *mapping, struct page *page);
1159 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1160 int invalidate_inode_page(struct page *page);
1161 
1162 #ifdef CONFIG_MMU
1163 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1164                         unsigned long address, unsigned int flags);
1165 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1166                             unsigned long address, unsigned int fault_flags);
1167 #else
1168 static inline int handle_mm_fault(struct mm_struct *mm,
1169                         struct vm_area_struct *vma, unsigned long address,
1170                         unsigned int flags)
1171 {
1172         /* should never happen if there's no MMU */
1173         BUG();
1174         return VM_FAULT_SIGBUS;
1175 }
1176 static inline int fixup_user_fault(struct task_struct *tsk,
1177                 struct mm_struct *mm, unsigned long address,
1178                 unsigned int fault_flags)
1179 {
1180         /* should never happen if there's no MMU */
1181         BUG();
1182         return -EFAULT;
1183 }
1184 #endif
1185 
1186 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1187 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1188                 void *buf, int len, int write);
1189 
1190 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1191                       unsigned long start, unsigned long nr_pages,
1192                       unsigned int foll_flags, struct page **pages,
1193                       struct vm_area_struct **vmas, int *nonblocking);
1194 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1195                     unsigned long start, unsigned long nr_pages,
1196                     int write, int force, struct page **pages,
1197                     struct vm_area_struct **vmas);
1198 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1199                     unsigned long start, unsigned long nr_pages,
1200                     int write, int force, struct page **pages,
1201                     int *locked);
1202 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1203                                unsigned long start, unsigned long nr_pages,
1204                                int write, int force, struct page **pages,
1205                                unsigned int gup_flags);
1206 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1207                     unsigned long start, unsigned long nr_pages,
1208                     int write, int force, struct page **pages);
1209 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1210                         struct page **pages);
1211 
1212 /* Container for pinned pfns / pages */
1213 struct frame_vector {
1214         unsigned int nr_allocated;      /* Number of frames we have space for */
1215         unsigned int nr_frames; /* Number of frames stored in ptrs array */
1216         bool got_ref;           /* Did we pin pages by getting page ref? */
1217         bool is_pfns;           /* Does array contain pages or pfns? */
1218         void *ptrs[0];          /* Array of pinned pfns / pages. Use
1219                                  * pfns_vector_pages() or pfns_vector_pfns()
1220                                  * for access */
1221 };
1222 
1223 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1224 void frame_vector_destroy(struct frame_vector *vec);
1225 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1226                      bool write, bool force, struct frame_vector *vec);
1227 void put_vaddr_frames(struct frame_vector *vec);
1228 int frame_vector_to_pages(struct frame_vector *vec);
1229 void frame_vector_to_pfns(struct frame_vector *vec);
1230 
1231 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1232 {
1233         return vec->nr_frames;
1234 }
1235 
1236 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1237 {
1238         if (vec->is_pfns) {
1239                 int err = frame_vector_to_pages(vec);
1240 
1241                 if (err)
1242                         return ERR_PTR(err);
1243         }
1244         return (struct page **)(vec->ptrs);
1245 }
1246 
1247 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1248 {
1249         if (!vec->is_pfns)
1250                 frame_vector_to_pfns(vec);
1251         return (unsigned long *)(vec->ptrs);
1252 }
1253 
1254 struct kvec;
1255 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1256                         struct page **pages);
1257 int get_kernel_page(unsigned long start, int write, struct page **pages);
1258 struct page *get_dump_page(unsigned long addr);
1259 
1260 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1261 extern void do_invalidatepage(struct page *page, unsigned int offset,
1262                               unsigned int length);
1263 
1264 int __set_page_dirty_nobuffers(struct page *page);
1265 int __set_page_dirty_no_writeback(struct page *page);
1266 int redirty_page_for_writepage(struct writeback_control *wbc,
1267                                 struct page *page);
1268 void account_page_dirtied(struct page *page, struct address_space *mapping,
1269                           struct mem_cgroup *memcg);
1270 void account_page_cleaned(struct page *page, struct address_space *mapping,
1271                           struct mem_cgroup *memcg, struct bdi_writeback *wb);
1272 int set_page_dirty(struct page *page);
1273 int set_page_dirty_lock(struct page *page);
1274 void cancel_dirty_page(struct page *page);
1275 int clear_page_dirty_for_io(struct page *page);
1276 
1277 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1278 
1279 /* Is the vma a continuation of the stack vma above it? */
1280 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1281 {
1282         return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1283 }
1284 
1285 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1286 {
1287         return !vma->vm_ops;
1288 }
1289 
1290 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1291                                              unsigned long addr)
1292 {
1293         return (vma->vm_flags & VM_GROWSDOWN) &&
1294                 (vma->vm_start == addr) &&
1295                 !vma_growsdown(vma->vm_prev, addr);
1296 }
1297 
1298 /* Is the vma a continuation of the stack vma below it? */
1299 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1300 {
1301         return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1302 }
1303 
1304 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1305                                            unsigned long addr)
1306 {
1307         return (vma->vm_flags & VM_GROWSUP) &&
1308                 (vma->vm_end == addr) &&
1309                 !vma_growsup(vma->vm_next, addr);
1310 }
1311 
1312 extern struct task_struct *task_of_stack(struct task_struct *task,
1313                                 struct vm_area_struct *vma, bool in_group);
1314 
1315 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1316                 unsigned long old_addr, struct vm_area_struct *new_vma,
1317                 unsigned long new_addr, unsigned long len,
1318                 bool need_rmap_locks);
1319 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1320                               unsigned long end, pgprot_t newprot,
1321                               int dirty_accountable, int prot_numa);
1322 extern int mprotect_fixup(struct vm_area_struct *vma,
1323                           struct vm_area_struct **pprev, unsigned long start,
1324                           unsigned long end, unsigned long newflags);
1325 
1326 /*
1327  * doesn't attempt to fault and will return short.
1328  */
1329 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1330                           struct page **pages);
1331 /*
1332  * per-process(per-mm_struct) statistics.
1333  */
1334 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1335 {
1336         long val = atomic_long_read(&mm->rss_stat.count[member]);
1337 
1338 #ifdef SPLIT_RSS_COUNTING
1339         /*
1340          * counter is updated in asynchronous manner and may go to minus.
1341          * But it's never be expected number for users.
1342          */
1343         if (val < 0)
1344                 val = 0;
1345 #endif
1346         return (unsigned long)val;
1347 }
1348 
1349 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1350 {
1351         atomic_long_add(value, &mm->rss_stat.count[member]);
1352 }
1353 
1354 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1355 {
1356         atomic_long_inc(&mm->rss_stat.count[member]);
1357 }
1358 
1359 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1360 {
1361         atomic_long_dec(&mm->rss_stat.count[member]);
1362 }
1363 
1364 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1365 {
1366         return get_mm_counter(mm, MM_FILEPAGES) +
1367                 get_mm_counter(mm, MM_ANONPAGES);
1368 }
1369 
1370 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1371 {
1372         return max(mm->hiwater_rss, get_mm_rss(mm));
1373 }
1374 
1375 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1376 {
1377         return max(mm->hiwater_vm, mm->total_vm);
1378 }
1379 
1380 static inline void update_hiwater_rss(struct mm_struct *mm)
1381 {
1382         unsigned long _rss = get_mm_rss(mm);
1383 
1384         if ((mm)->hiwater_rss < _rss)
1385                 (mm)->hiwater_rss = _rss;
1386 }
1387 
1388 static inline void update_hiwater_vm(struct mm_struct *mm)
1389 {
1390         if (mm->hiwater_vm < mm->total_vm)
1391                 mm->hiwater_vm = mm->total_vm;
1392 }
1393 
1394 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1395 {
1396         mm->hiwater_rss = get_mm_rss(mm);
1397 }
1398 
1399 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1400                                          struct mm_struct *mm)
1401 {
1402         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1403 
1404         if (*maxrss < hiwater_rss)
1405                 *maxrss = hiwater_rss;
1406 }
1407 
1408 #if defined(SPLIT_RSS_COUNTING)
1409 void sync_mm_rss(struct mm_struct *mm);
1410 #else
1411 static inline void sync_mm_rss(struct mm_struct *mm)
1412 {
1413 }
1414 #endif
1415 
1416 int vma_wants_writenotify(struct vm_area_struct *vma);
1417 
1418 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1419                                spinlock_t **ptl);
1420 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1421                                     spinlock_t **ptl)
1422 {
1423         pte_t *ptep;
1424         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1425         return ptep;
1426 }
1427 
1428 #ifdef __PAGETABLE_PUD_FOLDED
1429 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1430                                                 unsigned long address)
1431 {
1432         return 0;
1433 }
1434 #else
1435 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1436 #endif
1437 
1438 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1439 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1440                                                 unsigned long address)
1441 {
1442         return 0;
1443 }
1444 
1445 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1446 
1447 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1448 {
1449         return 0;
1450 }
1451 
1452 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1453 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1454 
1455 #else
1456 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1457 
1458 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1459 {
1460         atomic_long_set(&mm->nr_pmds, 0);
1461 }
1462 
1463 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1464 {
1465         return atomic_long_read(&mm->nr_pmds);
1466 }
1467 
1468 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1469 {
1470         atomic_long_inc(&mm->nr_pmds);
1471 }
1472 
1473 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1474 {
1475         atomic_long_dec(&mm->nr_pmds);
1476 }
1477 #endif
1478 
1479 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1480                 pmd_t *pmd, unsigned long address);
1481 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1482 
1483 /*
1484  * The following ifdef needed to get the 4level-fixup.h header to work.
1485  * Remove it when 4level-fixup.h has been removed.
1486  */
1487 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1488 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1489 {
1490         return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1491                 NULL: pud_offset(pgd, address);
1492 }
1493 
1494 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1495 {
1496         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1497                 NULL: pmd_offset(pud, address);
1498 }
1499 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1500 
1501 #if USE_SPLIT_PTE_PTLOCKS
1502 #if ALLOC_SPLIT_PTLOCKS
1503 void __init ptlock_cache_init(void);
1504 extern bool ptlock_alloc(struct page *page);
1505 extern void ptlock_free(struct page *page);
1506 
1507 static inline spinlock_t *ptlock_ptr(struct page *page)
1508 {
1509         return page->ptl;
1510 }
1511 #else /* ALLOC_SPLIT_PTLOCKS */
1512 static inline void ptlock_cache_init(void)
1513 {
1514 }
1515 
1516 static inline bool ptlock_alloc(struct page *page)
1517 {
1518         return true;
1519 }
1520 
1521 static inline void ptlock_free(struct page *page)
1522 {
1523 }
1524 
1525 static inline spinlock_t *ptlock_ptr(struct page *page)
1526 {
1527         return &page->ptl;
1528 }
1529 #endif /* ALLOC_SPLIT_PTLOCKS */
1530 
1531 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1532 {
1533         return ptlock_ptr(pmd_page(*pmd));
1534 }
1535 
1536 static inline bool ptlock_init(struct page *page)
1537 {
1538         /*
1539          * prep_new_page() initialize page->private (and therefore page->ptl)
1540          * with 0. Make sure nobody took it in use in between.
1541          *
1542          * It can happen if arch try to use slab for page table allocation:
1543          * slab code uses page->slab_cache, which share storage with page->ptl.
1544          */
1545         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1546         if (!ptlock_alloc(page))
1547                 return false;
1548         spin_lock_init(ptlock_ptr(page));
1549         return true;
1550 }
1551 
1552 /* Reset page->mapping so free_pages_check won't complain. */
1553 static inline void pte_lock_deinit(struct page *page)
1554 {
1555         page->mapping = NULL;
1556         ptlock_free(page);
1557 }
1558 
1559 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1560 /*
1561  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1562  */
1563 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1564 {
1565         return &mm->page_table_lock;
1566 }
1567 static inline void ptlock_cache_init(void) {}
1568 static inline bool ptlock_init(struct page *page) { return true; }
1569 static inline void pte_lock_deinit(struct page *page) {}
1570 #endif /* USE_SPLIT_PTE_PTLOCKS */
1571 
1572 static inline void pgtable_init(void)
1573 {
1574         ptlock_cache_init();
1575         pgtable_cache_init();
1576 }
1577 
1578 static inline bool pgtable_page_ctor(struct page *page)
1579 {
1580         if (!ptlock_init(page))
1581                 return false;
1582         inc_zone_page_state(page, NR_PAGETABLE);
1583         return true;
1584 }
1585 
1586 static inline void pgtable_page_dtor(struct page *page)
1587 {
1588         pte_lock_deinit(page);
1589         dec_zone_page_state(page, NR_PAGETABLE);
1590 }
1591 
1592 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1593 ({                                                      \
1594         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1595         pte_t *__pte = pte_offset_map(pmd, address);    \
1596         *(ptlp) = __ptl;                                \
1597         spin_lock(__ptl);                               \
1598         __pte;                                          \
1599 })
1600 
1601 #define pte_unmap_unlock(pte, ptl)      do {            \
1602         spin_unlock(ptl);                               \
1603         pte_unmap(pte);                                 \
1604 } while (0)
1605 
1606 #define pte_alloc_map(mm, vma, pmd, address)                            \
1607         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1608                                                         pmd, address))? \
1609          NULL: pte_offset_map(pmd, address))
1610 
1611 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1612         ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1613                                                         pmd, address))? \
1614                 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1615 
1616 #define pte_alloc_kernel(pmd, address)                  \
1617         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1618                 NULL: pte_offset_kernel(pmd, address))
1619 
1620 #if USE_SPLIT_PMD_PTLOCKS
1621 
1622 static struct page *pmd_to_page(pmd_t *pmd)
1623 {
1624         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1625         return virt_to_page((void *)((unsigned long) pmd & mask));
1626 }
1627 
1628 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1629 {
1630         return ptlock_ptr(pmd_to_page(pmd));
1631 }
1632 
1633 static inline bool pgtable_pmd_page_ctor(struct page *page)
1634 {
1635 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1636         page->pmd_huge_pte = NULL;
1637 #endif
1638         return ptlock_init(page);
1639 }
1640 
1641 static inline void pgtable_pmd_page_dtor(struct page *page)
1642 {
1643 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1644         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1645 #endif
1646         ptlock_free(page);
1647 }
1648 
1649 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1650 
1651 #else
1652 
1653 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1654 {
1655         return &mm->page_table_lock;
1656 }
1657 
1658 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1659 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1660 
1661 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1662 
1663 #endif
1664 
1665 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1666 {
1667         spinlock_t *ptl = pmd_lockptr(mm, pmd);
1668         spin_lock(ptl);
1669         return ptl;
1670 }
1671 
1672 extern void free_area_init(unsigned long * zones_size);
1673 extern void free_area_init_node(int nid, unsigned long * zones_size,
1674                 unsigned long zone_start_pfn, unsigned long *zholes_size);
1675 extern void free_initmem(void);
1676 
1677 /*
1678  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1679  * into the buddy system. The freed pages will be poisoned with pattern
1680  * "poison" if it's within range [0, UCHAR_MAX].
1681  * Return pages freed into the buddy system.
1682  */
1683 extern unsigned long free_reserved_area(void *start, void *end,
1684                                         int poison, char *s);
1685 
1686 #ifdef  CONFIG_HIGHMEM
1687 /*
1688  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1689  * and totalram_pages.
1690  */
1691 extern void free_highmem_page(struct page *page);
1692 #endif
1693 
1694 extern void adjust_managed_page_count(struct page *page, long count);
1695 extern void mem_init_print_info(const char *str);
1696 
1697 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1698 
1699 /* Free the reserved page into the buddy system, so it gets managed. */
1700 static inline void __free_reserved_page(struct page *page)
1701 {
1702         ClearPageReserved(page);
1703         init_page_count(page);
1704         __free_page(page);
1705 }
1706 
1707 static inline void free_reserved_page(struct page *page)
1708 {
1709         __free_reserved_page(page);
1710         adjust_managed_page_count(page, 1);
1711 }
1712 
1713 static inline void mark_page_reserved(struct page *page)
1714 {
1715         SetPageReserved(page);
1716         adjust_managed_page_count(page, -1);
1717 }
1718 
1719 /*
1720  * Default method to free all the __init memory into the buddy system.
1721  * The freed pages will be poisoned with pattern "poison" if it's within
1722  * range [0, UCHAR_MAX].
1723  * Return pages freed into the buddy system.
1724  */
1725 static inline unsigned long free_initmem_default(int poison)
1726 {
1727         extern char __init_begin[], __init_end[];
1728 
1729         return free_reserved_area(&__init_begin, &__init_end,
1730                                   poison, "unused kernel");
1731 }
1732 
1733 static inline unsigned long get_num_physpages(void)
1734 {
1735         int nid;
1736         unsigned long phys_pages = 0;
1737 
1738         for_each_online_node(nid)
1739                 phys_pages += node_present_pages(nid);
1740 
1741         return phys_pages;
1742 }
1743 
1744 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1745 /*
1746  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1747  * zones, allocate the backing mem_map and account for memory holes in a more
1748  * architecture independent manner. This is a substitute for creating the
1749  * zone_sizes[] and zholes_size[] arrays and passing them to
1750  * free_area_init_node()
1751  *
1752  * An architecture is expected to register range of page frames backed by
1753  * physical memory with memblock_add[_node]() before calling
1754  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1755  * usage, an architecture is expected to do something like
1756  *
1757  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1758  *                                                       max_highmem_pfn};
1759  * for_each_valid_physical_page_range()
1760  *      memblock_add_node(base, size, nid)
1761  * free_area_init_nodes(max_zone_pfns);
1762  *
1763  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1764  * registered physical page range.  Similarly
1765  * sparse_memory_present_with_active_regions() calls memory_present() for
1766  * each range when SPARSEMEM is enabled.
1767  *
1768  * See mm/page_alloc.c for more information on each function exposed by
1769  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1770  */
1771 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1772 unsigned long node_map_pfn_alignment(void);
1773 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1774                                                 unsigned long end_pfn);
1775 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1776                                                 unsigned long end_pfn);
1777 extern void get_pfn_range_for_nid(unsigned int nid,
1778                         unsigned long *start_pfn, unsigned long *end_pfn);
1779 extern unsigned long find_min_pfn_with_active_regions(void);
1780 extern void free_bootmem_with_active_regions(int nid,
1781                                                 unsigned long max_low_pfn);
1782 extern void sparse_memory_present_with_active_regions(int nid);
1783 
1784 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1785 
1786 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1787     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1788 static inline int __early_pfn_to_nid(unsigned long pfn,
1789                                         struct mminit_pfnnid_cache *state)
1790 {
1791         return 0;
1792 }
1793 #else
1794 /* please see mm/page_alloc.c */
1795 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1796 /* there is a per-arch backend function. */
1797 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1798                                         struct mminit_pfnnid_cache *state);
1799 #endif
1800 
1801 extern void set_dma_reserve(unsigned long new_dma_reserve);
1802 extern void memmap_init_zone(unsigned long, int, unsigned long,
1803                                 unsigned long, enum memmap_context);
1804 extern void setup_per_zone_wmarks(void);
1805 extern int __meminit init_per_zone_wmark_min(void);
1806 extern void mem_init(void);
1807 extern void __init mmap_init(void);
1808 extern void show_mem(unsigned int flags);
1809 extern void si_meminfo(struct sysinfo * val);
1810 extern void si_meminfo_node(struct sysinfo *val, int nid);
1811 
1812 extern __printf(3, 4)
1813 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1814                 const char *fmt, ...);
1815 
1816 extern void setup_per_cpu_pageset(void);
1817 
1818 extern void zone_pcp_update(struct zone *zone);
1819 extern void zone_pcp_reset(struct zone *zone);
1820 
1821 /* page_alloc.c */
1822 extern int min_free_kbytes;
1823 
1824 /* nommu.c */
1825 extern atomic_long_t mmap_pages_allocated;
1826 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1827 
1828 /* interval_tree.c */
1829 void vma_interval_tree_insert(struct vm_area_struct *node,
1830                               struct rb_root *root);
1831 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1832                                     struct vm_area_struct *prev,
1833                                     struct rb_root *root);
1834 void vma_interval_tree_remove(struct vm_area_struct *node,
1835                               struct rb_root *root);
1836 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1837                                 unsigned long start, unsigned long last);
1838 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1839                                 unsigned long start, unsigned long last);
1840 
1841 #define vma_interval_tree_foreach(vma, root, start, last)               \
1842         for (vma = vma_interval_tree_iter_first(root, start, last);     \
1843              vma; vma = vma_interval_tree_iter_next(vma, start, last))
1844 
1845 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1846                                    struct rb_root *root);
1847 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1848                                    struct rb_root *root);
1849 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1850         struct rb_root *root, unsigned long start, unsigned long last);
1851 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1852         struct anon_vma_chain *node, unsigned long start, unsigned long last);
1853 #ifdef CONFIG_DEBUG_VM_RB
1854 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1855 #endif
1856 
1857 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1858         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1859              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1860 
1861 /* mmap.c */
1862 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1863 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1864         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1865 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1866         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1867         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1868         struct mempolicy *, struct vm_userfaultfd_ctx);
1869 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1870 extern int split_vma(struct mm_struct *,
1871         struct vm_area_struct *, unsigned long addr, int new_below);
1872 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1873 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1874         struct rb_node **, struct rb_node *);
1875 extern void unlink_file_vma(struct vm_area_struct *);
1876 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1877         unsigned long addr, unsigned long len, pgoff_t pgoff,
1878         bool *need_rmap_locks);
1879 extern void exit_mmap(struct mm_struct *);
1880 
1881 static inline int check_data_rlimit(unsigned long rlim,
1882                                     unsigned long new,
1883                                     unsigned long start,
1884                                     unsigned long end_data,
1885                                     unsigned long start_data)
1886 {
1887         if (rlim < RLIM_INFINITY) {
1888                 if (((new - start) + (end_data - start_data)) > rlim)
1889                         return -ENOSPC;
1890         }
1891 
1892         return 0;
1893 }
1894 
1895 extern int mm_take_all_locks(struct mm_struct *mm);
1896 extern void mm_drop_all_locks(struct mm_struct *mm);
1897 
1898 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1899 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1900 
1901 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1902 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1903                                    unsigned long addr, unsigned long len,
1904                                    unsigned long flags,
1905                                    const struct vm_special_mapping *spec);
1906 /* This is an obsolete alternative to _install_special_mapping. */
1907 extern int install_special_mapping(struct mm_struct *mm,
1908                                    unsigned long addr, unsigned long len,
1909                                    unsigned long flags, struct page **pages);
1910 
1911 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1912 
1913 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1914         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1915 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1916         unsigned long len, unsigned long prot, unsigned long flags,
1917         vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1918 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1919 
1920 static inline unsigned long
1921 do_mmap_pgoff(struct file *file, unsigned long addr,
1922         unsigned long len, unsigned long prot, unsigned long flags,
1923         unsigned long pgoff, unsigned long *populate)
1924 {
1925         return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1926 }
1927 
1928 #ifdef CONFIG_MMU
1929 extern int __mm_populate(unsigned long addr, unsigned long len,
1930                          int ignore_errors);
1931 static inline void mm_populate(unsigned long addr, unsigned long len)
1932 {
1933         /* Ignore errors */
1934         (void) __mm_populate(addr, len, 1);
1935 }
1936 #else
1937 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1938 #endif
1939 
1940 /* These take the mm semaphore themselves */
1941 extern unsigned long vm_brk(unsigned long, unsigned long);
1942 extern int vm_munmap(unsigned long, size_t);
1943 extern unsigned long vm_mmap(struct file *, unsigned long,
1944         unsigned long, unsigned long,
1945         unsigned long, unsigned long);
1946 
1947 struct vm_unmapped_area_info {
1948 #define VM_UNMAPPED_AREA_TOPDOWN 1
1949         unsigned long flags;
1950         unsigned long length;
1951         unsigned long low_limit;
1952         unsigned long high_limit;
1953         unsigned long align_mask;
1954         unsigned long align_offset;
1955 };
1956 
1957 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1958 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1959 
1960 /*
1961  * Search for an unmapped address range.
1962  *
1963  * We are looking for a range that:
1964  * - does not intersect with any VMA;
1965  * - is contained within the [low_limit, high_limit) interval;
1966  * - is at least the desired size.
1967  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1968  */
1969 static inline unsigned long
1970 vm_unmapped_area(struct vm_unmapped_area_info *info)
1971 {
1972         if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1973                 return unmapped_area_topdown(info);
1974         else
1975                 return unmapped_area(info);
1976 }
1977 
1978 /* truncate.c */
1979 extern void truncate_inode_pages(struct address_space *, loff_t);
1980 extern void truncate_inode_pages_range(struct address_space *,
1981                                        loff_t lstart, loff_t lend);
1982 extern void truncate_inode_pages_final(struct address_space *);
1983 
1984 /* generic vm_area_ops exported for stackable file systems */
1985 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1986 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1987 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1988 
1989 /* mm/page-writeback.c */
1990 int write_one_page(struct page *page, int wait);
1991 void task_dirty_inc(struct task_struct *tsk);
1992 
1993 /* readahead.c */
1994 #define VM_MAX_READAHEAD        128     /* kbytes */
1995 #define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1996 
1997 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1998                         pgoff_t offset, unsigned long nr_to_read);
1999 
2000 void page_cache_sync_readahead(struct address_space *mapping,
2001                                struct file_ra_state *ra,
2002                                struct file *filp,
2003                                pgoff_t offset,
2004                                unsigned long size);
2005 
2006 void page_cache_async_readahead(struct address_space *mapping,
2007                                 struct file_ra_state *ra,
2008                                 struct file *filp,
2009                                 struct page *pg,
2010                                 pgoff_t offset,
2011                                 unsigned long size);
2012 
2013 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2014 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2015 
2016 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2017 extern int expand_downwards(struct vm_area_struct *vma,
2018                 unsigned long address);
2019 #if VM_GROWSUP
2020 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2021 #else
2022   #define expand_upwards(vma, address) (0)
2023 #endif
2024 
2025 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2026 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2027 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2028                                              struct vm_area_struct **pprev);
2029 
2030 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2031    NULL if none.  Assume start_addr < end_addr. */
2032 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2033 {
2034         struct vm_area_struct * vma = find_vma(mm,start_addr);
2035 
2036         if (vma && end_addr <= vma->vm_start)
2037                 vma = NULL;
2038         return vma;
2039 }
2040 
2041 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2042 {
2043         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2044 }
2045 
2046 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2047 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2048                                 unsigned long vm_start, unsigned long vm_end)
2049 {
2050         struct vm_area_struct *vma = find_vma(mm, vm_start);
2051 
2052         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2053                 vma = NULL;
2054 
2055         return vma;
2056 }
2057 
2058 #ifdef CONFIG_MMU
2059 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2060 void vma_set_page_prot(struct vm_area_struct *vma);
2061 #else
2062 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2063 {
2064         return __pgprot(0);
2065 }
2066 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2067 {
2068         vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2069 }
2070 #endif
2071 
2072 #ifdef CONFIG_NUMA_BALANCING
2073 unsigned long change_prot_numa(struct vm_area_struct *vma,
2074                         unsigned long start, unsigned long end);
2075 #endif
2076 
2077 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2078 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2079                         unsigned long pfn, unsigned long size, pgprot_t);
2080 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2081 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2082                         unsigned long pfn);
2083 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2084                         unsigned long pfn);
2085 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2086 
2087 
2088 struct page *follow_page_mask(struct vm_area_struct *vma,
2089                               unsigned long address, unsigned int foll_flags,
2090                               unsigned int *page_mask);
2091 
2092 static inline struct page *follow_page(struct vm_area_struct *vma,
2093                 unsigned long address, unsigned int foll_flags)
2094 {
2095         unsigned int unused_page_mask;
2096         return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2097 }
2098 
2099 #define FOLL_WRITE      0x01    /* check pte is writable */
2100 #define FOLL_TOUCH      0x02    /* mark page accessed */
2101 #define FOLL_GET        0x04    /* do get_page on page */
2102 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2103 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2104 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2105                                  * and return without waiting upon it */
2106 #define FOLL_POPULATE   0x40    /* fault in page */
2107 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2108 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2109 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2110 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2111 #define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2112 #define FOLL_MLOCK      0x1000  /* lock present pages */
2113 
2114 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2115                         void *data);
2116 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2117                                unsigned long size, pte_fn_t fn, void *data);
2118 
2119 #ifdef CONFIG_PROC_FS
2120 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2121 #else
2122 static inline void vm_stat_account(struct mm_struct *mm,
2123                         unsigned long flags, struct file *file, long pages)
2124 {
2125         mm->total_vm += pages;
2126 }
2127 #endif /* CONFIG_PROC_FS */
2128 
2129 #ifdef CONFIG_DEBUG_PAGEALLOC
2130 extern bool _debug_pagealloc_enabled;
2131 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2132 
2133 static inline bool debug_pagealloc_enabled(void)
2134 {
2135         return _debug_pagealloc_enabled;
2136 }
2137 
2138 static inline void
2139 kernel_map_pages(struct page *page, int numpages, int enable)
2140 {
2141         if (!debug_pagealloc_enabled())
2142                 return;
2143 
2144         __kernel_map_pages(page, numpages, enable);
2145 }
2146 #ifdef CONFIG_HIBERNATION
2147 extern bool kernel_page_present(struct page *page);
2148 #endif /* CONFIG_HIBERNATION */
2149 #else
2150 static inline void
2151 kernel_map_pages(struct page *page, int numpages, int enable) {}
2152 #ifdef CONFIG_HIBERNATION
2153 static inline bool kernel_page_present(struct page *page) { return true; }
2154 #endif /* CONFIG_HIBERNATION */
2155 #endif
2156 
2157 #ifdef __HAVE_ARCH_GATE_AREA
2158 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2159 extern int in_gate_area_no_mm(unsigned long addr);
2160 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2161 #else
2162 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2163 {
2164         return NULL;
2165 }
2166 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2167 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2168 {
2169         return 0;
2170 }
2171 #endif  /* __HAVE_ARCH_GATE_AREA */
2172 
2173 #ifdef CONFIG_SYSCTL
2174 extern int sysctl_drop_caches;
2175 int drop_caches_sysctl_handler(struct ctl_table *, int,
2176                                         void __user *, size_t *, loff_t *);
2177 #endif
2178 
2179 void drop_slab(void);
2180 void drop_slab_node(int nid);
2181 
2182 #ifndef CONFIG_MMU
2183 #define randomize_va_space 0
2184 #else
2185 extern int randomize_va_space;
2186 #endif
2187 
2188 const char * arch_vma_name(struct vm_area_struct *vma);
2189 void print_vma_addr(char *prefix, unsigned long rip);
2190 
2191 void sparse_mem_maps_populate_node(struct page **map_map,
2192                                    unsigned long pnum_begin,
2193                                    unsigned long pnum_end,
2194                                    unsigned long map_count,
2195                                    int nodeid);
2196 
2197 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2198 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2199 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2200 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2201 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2202 void *vmemmap_alloc_block(unsigned long size, int node);
2203 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2204 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2205 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2206                                int node);
2207 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2208 void vmemmap_populate_print_last(void);
2209 #ifdef CONFIG_MEMORY_HOTPLUG
2210 void vmemmap_free(unsigned long start, unsigned long end);
2211 #endif
2212 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2213                                   unsigned long size);
2214 
2215 enum mf_flags {
2216         MF_COUNT_INCREASED = 1 << 0,
2217         MF_ACTION_REQUIRED = 1 << 1,
2218         MF_MUST_KILL = 1 << 2,
2219         MF_SOFT_OFFLINE = 1 << 3,
2220 };
2221 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2222 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2223 extern int unpoison_memory(unsigned long pfn);
2224 extern int get_hwpoison_page(struct page *page);
2225 extern void put_hwpoison_page(struct page *page);
2226 extern int sysctl_memory_failure_early_kill;
2227 extern int sysctl_memory_failure_recovery;
2228 extern void shake_page(struct page *p, int access);
2229 extern atomic_long_t num_poisoned_pages;
2230 extern int soft_offline_page(struct page *page, int flags);
2231 
2232 
2233 /*
2234  * Error handlers for various types of pages.
2235  */
2236 enum mf_result {
2237         MF_IGNORED,     /* Error: cannot be handled */
2238         MF_FAILED,      /* Error: handling failed */
2239         MF_DELAYED,     /* Will be handled later */
2240         MF_RECOVERED,   /* Successfully recovered */
2241 };
2242 
2243 enum mf_action_page_type {
2244         MF_MSG_KERNEL,
2245         MF_MSG_KERNEL_HIGH_ORDER,
2246         MF_MSG_SLAB,
2247         MF_MSG_DIFFERENT_COMPOUND,
2248         MF_MSG_POISONED_HUGE,
2249         MF_MSG_HUGE,
2250         MF_MSG_FREE_HUGE,
2251         MF_MSG_UNMAP_FAILED,
2252         MF_MSG_DIRTY_SWAPCACHE,
2253         MF_MSG_CLEAN_SWAPCACHE,
2254         MF_MSG_DIRTY_MLOCKED_LRU,
2255         MF_MSG_CLEAN_MLOCKED_LRU,
2256         MF_MSG_DIRTY_UNEVICTABLE_LRU,
2257         MF_MSG_CLEAN_UNEVICTABLE_LRU,
2258         MF_MSG_DIRTY_LRU,
2259         MF_MSG_CLEAN_LRU,
2260         MF_MSG_TRUNCATED_LRU,
2261         MF_MSG_BUDDY,
2262         MF_MSG_BUDDY_2ND,
2263         MF_MSG_UNKNOWN,
2264 };
2265 
2266 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2267 extern void clear_huge_page(struct page *page,
2268                             unsigned long addr,
2269                             unsigned int pages_per_huge_page);
2270 extern void copy_user_huge_page(struct page *dst, struct page *src,
2271                                 unsigned long addr, struct vm_area_struct *vma,
2272                                 unsigned int pages_per_huge_page);
2273 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2274 
2275 extern struct page_ext_operations debug_guardpage_ops;
2276 extern struct page_ext_operations page_poisoning_ops;
2277 
2278 #ifdef CONFIG_DEBUG_PAGEALLOC
2279 extern unsigned int _debug_guardpage_minorder;
2280 extern bool _debug_guardpage_enabled;
2281 
2282 static inline unsigned int debug_guardpage_minorder(void)
2283 {
2284         return _debug_guardpage_minorder;
2285 }
2286 
2287 static inline bool debug_guardpage_enabled(void)
2288 {
2289         return _debug_guardpage_enabled;
2290 }
2291 
2292 static inline bool page_is_guard(struct page *page)
2293 {
2294         struct page_ext *page_ext;
2295 
2296         if (!debug_guardpage_enabled())
2297                 return false;
2298 
2299         page_ext = lookup_page_ext(page);
2300         return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2301 }
2302 #else
2303 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2304 static inline bool debug_guardpage_enabled(void) { return false; }
2305 static inline bool page_is_guard(struct page *page) { return false; }
2306 #endif /* CONFIG_DEBUG_PAGEALLOC */
2307 
2308 #if MAX_NUMNODES > 1
2309 void __init setup_nr_node_ids(void);
2310 #else
2311 static inline void setup_nr_node_ids(void) {}
2312 #endif
2313 
2314 #endif /* __KERNEL__ */
2315 #endif /* _LINUX_MM_H */
2316 

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