Version:  2.0.40 2.2.26 2.4.37 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 4.5 4.6 4.7 4.8

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/percpu-refcount.h>
 20 #include <linux/bit_spinlock.h>
 21 #include <linux/shrinker.h>
 22 #include <linux/resource.h>
 23 #include <linux/page_ext.h>
 24 #include <linux/err.h>
 25 #include <linux/page_ref.h>
 26 
 27 struct mempolicy;
 28 struct anon_vma;
 29 struct anon_vma_chain;
 30 struct file_ra_state;
 31 struct user_struct;
 32 struct writeback_control;
 33 struct bdi_writeback;
 34 
 35 #ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
 36 extern unsigned long max_mapnr;
 37 
 38 static inline void set_max_mapnr(unsigned long limit)
 39 {
 40         max_mapnr = limit;
 41 }
 42 #else
 43 static inline void set_max_mapnr(unsigned long limit) { }
 44 #endif
 45 
 46 extern unsigned long totalram_pages;
 47 extern void * high_memory;
 48 extern int page_cluster;
 49 
 50 #ifdef CONFIG_SYSCTL
 51 extern int sysctl_legacy_va_layout;
 52 #else
 53 #define sysctl_legacy_va_layout 0
 54 #endif
 55 
 56 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
 57 extern const int mmap_rnd_bits_min;
 58 extern const int mmap_rnd_bits_max;
 59 extern int mmap_rnd_bits __read_mostly;
 60 #endif
 61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
 62 extern const int mmap_rnd_compat_bits_min;
 63 extern const int mmap_rnd_compat_bits_max;
 64 extern int mmap_rnd_compat_bits __read_mostly;
 65 #endif
 66 
 67 #include <asm/page.h>
 68 #include <asm/pgtable.h>
 69 #include <asm/processor.h>
 70 
 71 #ifndef __pa_symbol
 72 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 73 #endif
 74 
 75 #ifndef page_to_virt
 76 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
 77 #endif
 78 
 79 /*
 80  * To prevent common memory management code establishing
 81  * a zero page mapping on a read fault.
 82  * This macro should be defined within <asm/pgtable.h>.
 83  * s390 does this to prevent multiplexing of hardware bits
 84  * related to the physical page in case of virtualization.
 85  */
 86 #ifndef mm_forbids_zeropage
 87 #define mm_forbids_zeropage(X)  (0)
 88 #endif
 89 
 90 /*
 91  * Default maximum number of active map areas, this limits the number of vmas
 92  * per mm struct. Users can overwrite this number by sysctl but there is a
 93  * problem.
 94  *
 95  * When a program's coredump is generated as ELF format, a section is created
 96  * per a vma. In ELF, the number of sections is represented in unsigned short.
 97  * This means the number of sections should be smaller than 65535 at coredump.
 98  * Because the kernel adds some informative sections to a image of program at
 99  * generating coredump, we need some margin. The number of extra sections is
100  * 1-3 now and depends on arch. We use "5" as safe margin, here.
101  *
102  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
103  * not a hard limit any more. Although some userspace tools can be surprised by
104  * that.
105  */
106 #define MAPCOUNT_ELF_CORE_MARGIN        (5)
107 #define DEFAULT_MAX_MAP_COUNT   (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
108 
109 extern int sysctl_max_map_count;
110 
111 extern unsigned long sysctl_user_reserve_kbytes;
112 extern unsigned long sysctl_admin_reserve_kbytes;
113 
114 extern int sysctl_overcommit_memory;
115 extern int sysctl_overcommit_ratio;
116 extern unsigned long sysctl_overcommit_kbytes;
117 
118 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
119                                     size_t *, loff_t *);
120 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
121                                     size_t *, loff_t *);
122 
123 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
124 
125 /* to align the pointer to the (next) page boundary */
126 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
127 
128 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
129 #define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
130 
131 /*
132  * Linux kernel virtual memory manager primitives.
133  * The idea being to have a "virtual" mm in the same way
134  * we have a virtual fs - giving a cleaner interface to the
135  * mm details, and allowing different kinds of memory mappings
136  * (from shared memory to executable loading to arbitrary
137  * mmap() functions).
138  */
139 
140 extern struct kmem_cache *vm_area_cachep;
141 
142 #ifndef CONFIG_MMU
143 extern struct rb_root nommu_region_tree;
144 extern struct rw_semaphore nommu_region_sem;
145 
146 extern unsigned int kobjsize(const void *objp);
147 #endif
148 
149 /*
150  * vm_flags in vm_area_struct, see mm_types.h.
151  * When changing, update also include/trace/events/mmflags.h
152  */
153 #define VM_NONE         0x00000000
154 
155 #define VM_READ         0x00000001      /* currently active flags */
156 #define VM_WRITE        0x00000002
157 #define VM_EXEC         0x00000004
158 #define VM_SHARED       0x00000008
159 
160 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
161 #define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
162 #define VM_MAYWRITE     0x00000020
163 #define VM_MAYEXEC      0x00000040
164 #define VM_MAYSHARE     0x00000080
165 
166 #define VM_GROWSDOWN    0x00000100      /* general info on the segment */
167 #define VM_UFFD_MISSING 0x00000200      /* missing pages tracking */
168 #define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
169 #define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
170 #define VM_UFFD_WP      0x00001000      /* wrprotect pages tracking */
171 
172 #define VM_LOCKED       0x00002000
173 #define VM_IO           0x00004000      /* Memory mapped I/O or similar */
174 
175                                         /* Used by sys_madvise() */
176 #define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
177 #define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
178 
179 #define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
180 #define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
181 #define VM_LOCKONFAULT  0x00080000      /* Lock the pages covered when they are faulted in */
182 #define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
183 #define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
184 #define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
185 #define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
186 #define VM_ARCH_2       0x02000000
187 #define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
188 
189 #ifdef CONFIG_MEM_SOFT_DIRTY
190 # define VM_SOFTDIRTY   0x08000000      /* Not soft dirty clean area */
191 #else
192 # define VM_SOFTDIRTY   0
193 #endif
194 
195 #define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
196 #define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
197 #define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
198 #define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
199 
200 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
201 #define VM_HIGH_ARCH_BIT_0      32      /* bit only usable on 64-bit architectures */
202 #define VM_HIGH_ARCH_BIT_1      33      /* bit only usable on 64-bit architectures */
203 #define VM_HIGH_ARCH_BIT_2      34      /* bit only usable on 64-bit architectures */
204 #define VM_HIGH_ARCH_BIT_3      35      /* bit only usable on 64-bit architectures */
205 #define VM_HIGH_ARCH_0  BIT(VM_HIGH_ARCH_BIT_0)
206 #define VM_HIGH_ARCH_1  BIT(VM_HIGH_ARCH_BIT_1)
207 #define VM_HIGH_ARCH_2  BIT(VM_HIGH_ARCH_BIT_2)
208 #define VM_HIGH_ARCH_3  BIT(VM_HIGH_ARCH_BIT_3)
209 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
210 
211 #if defined(CONFIG_X86)
212 # define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
213 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
214 # define VM_PKEY_SHIFT  VM_HIGH_ARCH_BIT_0
215 # define VM_PKEY_BIT0   VM_HIGH_ARCH_0  /* A protection key is a 4-bit value */
216 # define VM_PKEY_BIT1   VM_HIGH_ARCH_1
217 # define VM_PKEY_BIT2   VM_HIGH_ARCH_2
218 # define VM_PKEY_BIT3   VM_HIGH_ARCH_3
219 #endif
220 #elif defined(CONFIG_PPC)
221 # define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
222 #elif defined(CONFIG_PARISC)
223 # define VM_GROWSUP     VM_ARCH_1
224 #elif defined(CONFIG_METAG)
225 # define VM_GROWSUP     VM_ARCH_1
226 #elif defined(CONFIG_IA64)
227 # define VM_GROWSUP     VM_ARCH_1
228 #elif !defined(CONFIG_MMU)
229 # define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
230 #endif
231 
232 #if defined(CONFIG_X86)
233 /* MPX specific bounds table or bounds directory */
234 # define VM_MPX         VM_ARCH_2
235 #endif
236 
237 #ifndef VM_GROWSUP
238 # define VM_GROWSUP     VM_NONE
239 #endif
240 
241 /* Bits set in the VMA until the stack is in its final location */
242 #define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
243 
244 #ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
245 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
246 #endif
247 
248 #ifdef CONFIG_STACK_GROWSUP
249 #define VM_STACK        VM_GROWSUP
250 #else
251 #define VM_STACK        VM_GROWSDOWN
252 #endif
253 
254 #define VM_STACK_FLAGS  (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
255 
256 /*
257  * Special vmas that are non-mergable, non-mlock()able.
258  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
259  */
260 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
261 
262 /* This mask defines which mm->def_flags a process can inherit its parent */
263 #define VM_INIT_DEF_MASK        VM_NOHUGEPAGE
264 
265 /* This mask is used to clear all the VMA flags used by mlock */
266 #define VM_LOCKED_CLEAR_MASK    (~(VM_LOCKED | VM_LOCKONFAULT))
267 
268 /*
269  * mapping from the currently active vm_flags protection bits (the
270  * low four bits) to a page protection mask..
271  */
272 extern pgprot_t protection_map[16];
273 
274 #define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
275 #define FAULT_FLAG_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
276 #define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
277 #define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
278 #define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
279 #define FAULT_FLAG_TRIED        0x20    /* Second try */
280 #define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
281 #define FAULT_FLAG_REMOTE       0x80    /* faulting for non current tsk/mm */
282 #define FAULT_FLAG_INSTRUCTION  0x100   /* The fault was during an instruction fetch */
283 
284 /*
285  * vm_fault is filled by the the pagefault handler and passed to the vma's
286  * ->fault function. The vma's ->fault is responsible for returning a bitmask
287  * of VM_FAULT_xxx flags that give details about how the fault was handled.
288  *
289  * MM layer fills up gfp_mask for page allocations but fault handler might
290  * alter it if its implementation requires a different allocation context.
291  *
292  * pgoff should be used in favour of virtual_address, if possible.
293  */
294 struct vm_fault {
295         unsigned int flags;             /* FAULT_FLAG_xxx flags */
296         gfp_t gfp_mask;                 /* gfp mask to be used for allocations */
297         pgoff_t pgoff;                  /* Logical page offset based on vma */
298         void __user *virtual_address;   /* Faulting virtual address */
299 
300         struct page *cow_page;          /* Handler may choose to COW */
301         struct page *page;              /* ->fault handlers should return a
302                                          * page here, unless VM_FAULT_NOPAGE
303                                          * is set (which is also implied by
304                                          * VM_FAULT_ERROR).
305                                          */
306         void *entry;                    /* ->fault handler can alternatively
307                                          * return locked DAX entry. In that
308                                          * case handler should return
309                                          * VM_FAULT_DAX_LOCKED and fill in
310                                          * entry here.
311                                          */
312 };
313 
314 /*
315  * Page fault context: passes though page fault handler instead of endless list
316  * of function arguments.
317  */
318 struct fault_env {
319         struct vm_area_struct *vma;     /* Target VMA */
320         unsigned long address;          /* Faulting virtual address */
321         unsigned int flags;             /* FAULT_FLAG_xxx flags */
322         pmd_t *pmd;                     /* Pointer to pmd entry matching
323                                          * the 'address'
324                                          */
325         pte_t *pte;                     /* Pointer to pte entry matching
326                                          * the 'address'. NULL if the page
327                                          * table hasn't been allocated.
328                                          */
329         spinlock_t *ptl;                /* Page table lock.
330                                          * Protects pte page table if 'pte'
331                                          * is not NULL, otherwise pmd.
332                                          */
333         pgtable_t prealloc_pte;         /* Pre-allocated pte page table.
334                                          * vm_ops->map_pages() calls
335                                          * alloc_set_pte() from atomic context.
336                                          * do_fault_around() pre-allocates
337                                          * page table to avoid allocation from
338                                          * atomic context.
339                                          */
340 };
341 
342 /*
343  * These are the virtual MM functions - opening of an area, closing and
344  * unmapping it (needed to keep files on disk up-to-date etc), pointer
345  * to the functions called when a no-page or a wp-page exception occurs. 
346  */
347 struct vm_operations_struct {
348         void (*open)(struct vm_area_struct * area);
349         void (*close)(struct vm_area_struct * area);
350         int (*mremap)(struct vm_area_struct * area);
351         int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
352         int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
353                                                 pmd_t *, unsigned int flags);
354         void (*map_pages)(struct fault_env *fe,
355                         pgoff_t start_pgoff, pgoff_t end_pgoff);
356 
357         /* notification that a previously read-only page is about to become
358          * writable, if an error is returned it will cause a SIGBUS */
359         int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
360 
361         /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
362         int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
363 
364         /* called by access_process_vm when get_user_pages() fails, typically
365          * for use by special VMAs that can switch between memory and hardware
366          */
367         int (*access)(struct vm_area_struct *vma, unsigned long addr,
368                       void *buf, int len, int write);
369 
370         /* Called by the /proc/PID/maps code to ask the vma whether it
371          * has a special name.  Returning non-NULL will also cause this
372          * vma to be dumped unconditionally. */
373         const char *(*name)(struct vm_area_struct *vma);
374 
375 #ifdef CONFIG_NUMA
376         /*
377          * set_policy() op must add a reference to any non-NULL @new mempolicy
378          * to hold the policy upon return.  Caller should pass NULL @new to
379          * remove a policy and fall back to surrounding context--i.e. do not
380          * install a MPOL_DEFAULT policy, nor the task or system default
381          * mempolicy.
382          */
383         int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
384 
385         /*
386          * get_policy() op must add reference [mpol_get()] to any policy at
387          * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
388          * in mm/mempolicy.c will do this automatically.
389          * get_policy() must NOT add a ref if the policy at (vma,addr) is not
390          * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
391          * If no [shared/vma] mempolicy exists at the addr, get_policy() op
392          * must return NULL--i.e., do not "fallback" to task or system default
393          * policy.
394          */
395         struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
396                                         unsigned long addr);
397 #endif
398         /*
399          * Called by vm_normal_page() for special PTEs to find the
400          * page for @addr.  This is useful if the default behavior
401          * (using pte_page()) would not find the correct page.
402          */
403         struct page *(*find_special_page)(struct vm_area_struct *vma,
404                                           unsigned long addr);
405 };
406 
407 struct mmu_gather;
408 struct inode;
409 
410 #define page_private(page)              ((page)->private)
411 #define set_page_private(page, v)       ((page)->private = (v))
412 
413 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
414 static inline int pmd_devmap(pmd_t pmd)
415 {
416         return 0;
417 }
418 #endif
419 
420 /*
421  * FIXME: take this include out, include page-flags.h in
422  * files which need it (119 of them)
423  */
424 #include <linux/page-flags.h>
425 #include <linux/huge_mm.h>
426 
427 /*
428  * Methods to modify the page usage count.
429  *
430  * What counts for a page usage:
431  * - cache mapping   (page->mapping)
432  * - private data    (page->private)
433  * - page mapped in a task's page tables, each mapping
434  *   is counted separately
435  *
436  * Also, many kernel routines increase the page count before a critical
437  * routine so they can be sure the page doesn't go away from under them.
438  */
439 
440 /*
441  * Drop a ref, return true if the refcount fell to zero (the page has no users)
442  */
443 static inline int put_page_testzero(struct page *page)
444 {
445         VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
446         return page_ref_dec_and_test(page);
447 }
448 
449 /*
450  * Try to grab a ref unless the page has a refcount of zero, return false if
451  * that is the case.
452  * This can be called when MMU is off so it must not access
453  * any of the virtual mappings.
454  */
455 static inline int get_page_unless_zero(struct page *page)
456 {
457         return page_ref_add_unless(page, 1, 0);
458 }
459 
460 extern int page_is_ram(unsigned long pfn);
461 
462 enum {
463         REGION_INTERSECTS,
464         REGION_DISJOINT,
465         REGION_MIXED,
466 };
467 
468 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
469                       unsigned long desc);
470 
471 /* Support for virtually mapped pages */
472 struct page *vmalloc_to_page(const void *addr);
473 unsigned long vmalloc_to_pfn(const void *addr);
474 
475 /*
476  * Determine if an address is within the vmalloc range
477  *
478  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
479  * is no special casing required.
480  */
481 static inline bool is_vmalloc_addr(const void *x)
482 {
483 #ifdef CONFIG_MMU
484         unsigned long addr = (unsigned long)x;
485 
486         return addr >= VMALLOC_START && addr < VMALLOC_END;
487 #else
488         return false;
489 #endif
490 }
491 #ifdef CONFIG_MMU
492 extern int is_vmalloc_or_module_addr(const void *x);
493 #else
494 static inline int is_vmalloc_or_module_addr(const void *x)
495 {
496         return 0;
497 }
498 #endif
499 
500 extern void kvfree(const void *addr);
501 
502 static inline atomic_t *compound_mapcount_ptr(struct page *page)
503 {
504         return &page[1].compound_mapcount;
505 }
506 
507 static inline int compound_mapcount(struct page *page)
508 {
509         VM_BUG_ON_PAGE(!PageCompound(page), page);
510         page = compound_head(page);
511         return atomic_read(compound_mapcount_ptr(page)) + 1;
512 }
513 
514 /*
515  * The atomic page->_mapcount, starts from -1: so that transitions
516  * both from it and to it can be tracked, using atomic_inc_and_test
517  * and atomic_add_negative(-1).
518  */
519 static inline void page_mapcount_reset(struct page *page)
520 {
521         atomic_set(&(page)->_mapcount, -1);
522 }
523 
524 int __page_mapcount(struct page *page);
525 
526 static inline int page_mapcount(struct page *page)
527 {
528         VM_BUG_ON_PAGE(PageSlab(page), page);
529 
530         if (unlikely(PageCompound(page)))
531                 return __page_mapcount(page);
532         return atomic_read(&page->_mapcount) + 1;
533 }
534 
535 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
536 int total_mapcount(struct page *page);
537 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
538 #else
539 static inline int total_mapcount(struct page *page)
540 {
541         return page_mapcount(page);
542 }
543 static inline int page_trans_huge_mapcount(struct page *page,
544                                            int *total_mapcount)
545 {
546         int mapcount = page_mapcount(page);
547         if (total_mapcount)
548                 *total_mapcount = mapcount;
549         return mapcount;
550 }
551 #endif
552 
553 static inline struct page *virt_to_head_page(const void *x)
554 {
555         struct page *page = virt_to_page(x);
556 
557         return compound_head(page);
558 }
559 
560 void __put_page(struct page *page);
561 
562 void put_pages_list(struct list_head *pages);
563 
564 void split_page(struct page *page, unsigned int order);
565 
566 /*
567  * Compound pages have a destructor function.  Provide a
568  * prototype for that function and accessor functions.
569  * These are _only_ valid on the head of a compound page.
570  */
571 typedef void compound_page_dtor(struct page *);
572 
573 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
574 enum compound_dtor_id {
575         NULL_COMPOUND_DTOR,
576         COMPOUND_PAGE_DTOR,
577 #ifdef CONFIG_HUGETLB_PAGE
578         HUGETLB_PAGE_DTOR,
579 #endif
580 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
581         TRANSHUGE_PAGE_DTOR,
582 #endif
583         NR_COMPOUND_DTORS,
584 };
585 extern compound_page_dtor * const compound_page_dtors[];
586 
587 static inline void set_compound_page_dtor(struct page *page,
588                 enum compound_dtor_id compound_dtor)
589 {
590         VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
591         page[1].compound_dtor = compound_dtor;
592 }
593 
594 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
595 {
596         VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
597         return compound_page_dtors[page[1].compound_dtor];
598 }
599 
600 static inline unsigned int compound_order(struct page *page)
601 {
602         if (!PageHead(page))
603                 return 0;
604         return page[1].compound_order;
605 }
606 
607 static inline void set_compound_order(struct page *page, unsigned int order)
608 {
609         page[1].compound_order = order;
610 }
611 
612 void free_compound_page(struct page *page);
613 
614 #ifdef CONFIG_MMU
615 /*
616  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
617  * servicing faults for write access.  In the normal case, do always want
618  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
619  * that do not have writing enabled, when used by access_process_vm.
620  */
621 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
622 {
623         if (likely(vma->vm_flags & VM_WRITE))
624                 pte = pte_mkwrite(pte);
625         return pte;
626 }
627 
628 int alloc_set_pte(struct fault_env *fe, struct mem_cgroup *memcg,
629                 struct page *page);
630 #endif
631 
632 /*
633  * Multiple processes may "see" the same page. E.g. for untouched
634  * mappings of /dev/null, all processes see the same page full of
635  * zeroes, and text pages of executables and shared libraries have
636  * only one copy in memory, at most, normally.
637  *
638  * For the non-reserved pages, page_count(page) denotes a reference count.
639  *   page_count() == 0 means the page is free. page->lru is then used for
640  *   freelist management in the buddy allocator.
641  *   page_count() > 0  means the page has been allocated.
642  *
643  * Pages are allocated by the slab allocator in order to provide memory
644  * to kmalloc and kmem_cache_alloc. In this case, the management of the
645  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
646  * unless a particular usage is carefully commented. (the responsibility of
647  * freeing the kmalloc memory is the caller's, of course).
648  *
649  * A page may be used by anyone else who does a __get_free_page().
650  * In this case, page_count still tracks the references, and should only
651  * be used through the normal accessor functions. The top bits of page->flags
652  * and page->virtual store page management information, but all other fields
653  * are unused and could be used privately, carefully. The management of this
654  * page is the responsibility of the one who allocated it, and those who have
655  * subsequently been given references to it.
656  *
657  * The other pages (we may call them "pagecache pages") are completely
658  * managed by the Linux memory manager: I/O, buffers, swapping etc.
659  * The following discussion applies only to them.
660  *
661  * A pagecache page contains an opaque `private' member, which belongs to the
662  * page's address_space. Usually, this is the address of a circular list of
663  * the page's disk buffers. PG_private must be set to tell the VM to call
664  * into the filesystem to release these pages.
665  *
666  * A page may belong to an inode's memory mapping. In this case, page->mapping
667  * is the pointer to the inode, and page->index is the file offset of the page,
668  * in units of PAGE_SIZE.
669  *
670  * If pagecache pages are not associated with an inode, they are said to be
671  * anonymous pages. These may become associated with the swapcache, and in that
672  * case PG_swapcache is set, and page->private is an offset into the swapcache.
673  *
674  * In either case (swapcache or inode backed), the pagecache itself holds one
675  * reference to the page. Setting PG_private should also increment the
676  * refcount. The each user mapping also has a reference to the page.
677  *
678  * The pagecache pages are stored in a per-mapping radix tree, which is
679  * rooted at mapping->page_tree, and indexed by offset.
680  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
681  * lists, we instead now tag pages as dirty/writeback in the radix tree.
682  *
683  * All pagecache pages may be subject to I/O:
684  * - inode pages may need to be read from disk,
685  * - inode pages which have been modified and are MAP_SHARED may need
686  *   to be written back to the inode on disk,
687  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
688  *   modified may need to be swapped out to swap space and (later) to be read
689  *   back into memory.
690  */
691 
692 /*
693  * The zone field is never updated after free_area_init_core()
694  * sets it, so none of the operations on it need to be atomic.
695  */
696 
697 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
698 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
699 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
700 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
701 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
702 
703 /*
704  * Define the bit shifts to access each section.  For non-existent
705  * sections we define the shift as 0; that plus a 0 mask ensures
706  * the compiler will optimise away reference to them.
707  */
708 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
709 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
710 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
711 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
712 
713 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
714 #ifdef NODE_NOT_IN_PAGE_FLAGS
715 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
716 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
717                                                 SECTIONS_PGOFF : ZONES_PGOFF)
718 #else
719 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
720 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
721                                                 NODES_PGOFF : ZONES_PGOFF)
722 #endif
723 
724 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
725 
726 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
727 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
728 #endif
729 
730 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
731 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
732 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
733 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
734 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
735 
736 static inline enum zone_type page_zonenum(const struct page *page)
737 {
738         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
739 }
740 
741 #ifdef CONFIG_ZONE_DEVICE
742 void get_zone_device_page(struct page *page);
743 void put_zone_device_page(struct page *page);
744 static inline bool is_zone_device_page(const struct page *page)
745 {
746         return page_zonenum(page) == ZONE_DEVICE;
747 }
748 #else
749 static inline void get_zone_device_page(struct page *page)
750 {
751 }
752 static inline void put_zone_device_page(struct page *page)
753 {
754 }
755 static inline bool is_zone_device_page(const struct page *page)
756 {
757         return false;
758 }
759 #endif
760 
761 static inline void get_page(struct page *page)
762 {
763         page = compound_head(page);
764         /*
765          * Getting a normal page or the head of a compound page
766          * requires to already have an elevated page->_refcount.
767          */
768         VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
769         page_ref_inc(page);
770 
771         if (unlikely(is_zone_device_page(page)))
772                 get_zone_device_page(page);
773 }
774 
775 static inline void put_page(struct page *page)
776 {
777         page = compound_head(page);
778 
779         if (put_page_testzero(page))
780                 __put_page(page);
781 
782         if (unlikely(is_zone_device_page(page)))
783                 put_zone_device_page(page);
784 }
785 
786 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
787 #define SECTION_IN_PAGE_FLAGS
788 #endif
789 
790 /*
791  * The identification function is mainly used by the buddy allocator for
792  * determining if two pages could be buddies. We are not really identifying
793  * the zone since we could be using the section number id if we do not have
794  * node id available in page flags.
795  * We only guarantee that it will return the same value for two combinable
796  * pages in a zone.
797  */
798 static inline int page_zone_id(struct page *page)
799 {
800         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
801 }
802 
803 static inline int zone_to_nid(struct zone *zone)
804 {
805 #ifdef CONFIG_NUMA
806         return zone->node;
807 #else
808         return 0;
809 #endif
810 }
811 
812 #ifdef NODE_NOT_IN_PAGE_FLAGS
813 extern int page_to_nid(const struct page *page);
814 #else
815 static inline int page_to_nid(const struct page *page)
816 {
817         return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
818 }
819 #endif
820 
821 #ifdef CONFIG_NUMA_BALANCING
822 static inline int cpu_pid_to_cpupid(int cpu, int pid)
823 {
824         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
825 }
826 
827 static inline int cpupid_to_pid(int cpupid)
828 {
829         return cpupid & LAST__PID_MASK;
830 }
831 
832 static inline int cpupid_to_cpu(int cpupid)
833 {
834         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
835 }
836 
837 static inline int cpupid_to_nid(int cpupid)
838 {
839         return cpu_to_node(cpupid_to_cpu(cpupid));
840 }
841 
842 static inline bool cpupid_pid_unset(int cpupid)
843 {
844         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
845 }
846 
847 static inline bool cpupid_cpu_unset(int cpupid)
848 {
849         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
850 }
851 
852 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
853 {
854         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
855 }
856 
857 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
858 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
859 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
860 {
861         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
862 }
863 
864 static inline int page_cpupid_last(struct page *page)
865 {
866         return page->_last_cpupid;
867 }
868 static inline void page_cpupid_reset_last(struct page *page)
869 {
870         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
871 }
872 #else
873 static inline int page_cpupid_last(struct page *page)
874 {
875         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
876 }
877 
878 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
879 
880 static inline void page_cpupid_reset_last(struct page *page)
881 {
882         page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
883 }
884 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
885 #else /* !CONFIG_NUMA_BALANCING */
886 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
887 {
888         return page_to_nid(page); /* XXX */
889 }
890 
891 static inline int page_cpupid_last(struct page *page)
892 {
893         return page_to_nid(page); /* XXX */
894 }
895 
896 static inline int cpupid_to_nid(int cpupid)
897 {
898         return -1;
899 }
900 
901 static inline int cpupid_to_pid(int cpupid)
902 {
903         return -1;
904 }
905 
906 static inline int cpupid_to_cpu(int cpupid)
907 {
908         return -1;
909 }
910 
911 static inline int cpu_pid_to_cpupid(int nid, int pid)
912 {
913         return -1;
914 }
915 
916 static inline bool cpupid_pid_unset(int cpupid)
917 {
918         return 1;
919 }
920 
921 static inline void page_cpupid_reset_last(struct page *page)
922 {
923 }
924 
925 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
926 {
927         return false;
928 }
929 #endif /* CONFIG_NUMA_BALANCING */
930 
931 static inline struct zone *page_zone(const struct page *page)
932 {
933         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
934 }
935 
936 static inline pg_data_t *page_pgdat(const struct page *page)
937 {
938         return NODE_DATA(page_to_nid(page));
939 }
940 
941 #ifdef SECTION_IN_PAGE_FLAGS
942 static inline void set_page_section(struct page *page, unsigned long section)
943 {
944         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
945         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
946 }
947 
948 static inline unsigned long page_to_section(const struct page *page)
949 {
950         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
951 }
952 #endif
953 
954 static inline void set_page_zone(struct page *page, enum zone_type zone)
955 {
956         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
957         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
958 }
959 
960 static inline void set_page_node(struct page *page, unsigned long node)
961 {
962         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
963         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
964 }
965 
966 static inline void set_page_links(struct page *page, enum zone_type zone,
967         unsigned long node, unsigned long pfn)
968 {
969         set_page_zone(page, zone);
970         set_page_node(page, node);
971 #ifdef SECTION_IN_PAGE_FLAGS
972         set_page_section(page, pfn_to_section_nr(pfn));
973 #endif
974 }
975 
976 #ifdef CONFIG_MEMCG
977 static inline struct mem_cgroup *page_memcg(struct page *page)
978 {
979         return page->mem_cgroup;
980 }
981 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
982 {
983         WARN_ON_ONCE(!rcu_read_lock_held());
984         return READ_ONCE(page->mem_cgroup);
985 }
986 #else
987 static inline struct mem_cgroup *page_memcg(struct page *page)
988 {
989         return NULL;
990 }
991 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
992 {
993         WARN_ON_ONCE(!rcu_read_lock_held());
994         return NULL;
995 }
996 #endif
997 
998 /*
999  * Some inline functions in vmstat.h depend on page_zone()
1000  */
1001 #include <linux/vmstat.h>
1002 
1003 static __always_inline void *lowmem_page_address(const struct page *page)
1004 {
1005         return page_to_virt(page);
1006 }
1007 
1008 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1009 #define HASHED_PAGE_VIRTUAL
1010 #endif
1011 
1012 #if defined(WANT_PAGE_VIRTUAL)
1013 static inline void *page_address(const struct page *page)
1014 {
1015         return page->virtual;
1016 }
1017 static inline void set_page_address(struct page *page, void *address)
1018 {
1019         page->virtual = address;
1020 }
1021 #define page_address_init()  do { } while(0)
1022 #endif
1023 
1024 #if defined(HASHED_PAGE_VIRTUAL)
1025 void *page_address(const struct page *page);
1026 void set_page_address(struct page *page, void *virtual);
1027 void page_address_init(void);
1028 #endif
1029 
1030 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1031 #define page_address(page) lowmem_page_address(page)
1032 #define set_page_address(page, address)  do { } while(0)
1033 #define page_address_init()  do { } while(0)
1034 #endif
1035 
1036 extern void *page_rmapping(struct page *page);
1037 extern struct anon_vma *page_anon_vma(struct page *page);
1038 extern struct address_space *page_mapping(struct page *page);
1039 
1040 extern struct address_space *__page_file_mapping(struct page *);
1041 
1042 static inline
1043 struct address_space *page_file_mapping(struct page *page)
1044 {
1045         if (unlikely(PageSwapCache(page)))
1046                 return __page_file_mapping(page);
1047 
1048         return page->mapping;
1049 }
1050 
1051 /*
1052  * Return the pagecache index of the passed page.  Regular pagecache pages
1053  * use ->index whereas swapcache pages use ->private
1054  */
1055 static inline pgoff_t page_index(struct page *page)
1056 {
1057         if (unlikely(PageSwapCache(page)))
1058                 return page_private(page);
1059         return page->index;
1060 }
1061 
1062 extern pgoff_t __page_file_index(struct page *page);
1063 
1064 /*
1065  * Return the file index of the page. Regular pagecache pages use ->index
1066  * whereas swapcache pages use swp_offset(->private)
1067  */
1068 static inline pgoff_t page_file_index(struct page *page)
1069 {
1070         if (unlikely(PageSwapCache(page)))
1071                 return __page_file_index(page);
1072 
1073         return page->index;
1074 }
1075 
1076 bool page_mapped(struct page *page);
1077 struct address_space *page_mapping(struct page *page);
1078 
1079 /*
1080  * Return true only if the page has been allocated with
1081  * ALLOC_NO_WATERMARKS and the low watermark was not
1082  * met implying that the system is under some pressure.
1083  */
1084 static inline bool page_is_pfmemalloc(struct page *page)
1085 {
1086         /*
1087          * Page index cannot be this large so this must be
1088          * a pfmemalloc page.
1089          */
1090         return page->index == -1UL;
1091 }
1092 
1093 /*
1094  * Only to be called by the page allocator on a freshly allocated
1095  * page.
1096  */
1097 static inline void set_page_pfmemalloc(struct page *page)
1098 {
1099         page->index = -1UL;
1100 }
1101 
1102 static inline void clear_page_pfmemalloc(struct page *page)
1103 {
1104         page->index = 0;
1105 }
1106 
1107 /*
1108  * Different kinds of faults, as returned by handle_mm_fault().
1109  * Used to decide whether a process gets delivered SIGBUS or
1110  * just gets major/minor fault counters bumped up.
1111  */
1112 
1113 #define VM_FAULT_OOM    0x0001
1114 #define VM_FAULT_SIGBUS 0x0002
1115 #define VM_FAULT_MAJOR  0x0004
1116 #define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
1117 #define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
1118 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1119 #define VM_FAULT_SIGSEGV 0x0040
1120 
1121 #define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
1122 #define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
1123 #define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
1124 #define VM_FAULT_FALLBACK 0x0800        /* huge page fault failed, fall back to small */
1125 #define VM_FAULT_DAX_LOCKED 0x1000      /* ->fault has locked DAX entry */
1126 
1127 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1128 
1129 #define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1130                          VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1131                          VM_FAULT_FALLBACK)
1132 
1133 /* Encode hstate index for a hwpoisoned large page */
1134 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1135 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1136 
1137 /*
1138  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1139  */
1140 extern void pagefault_out_of_memory(void);
1141 
1142 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1143 
1144 /*
1145  * Flags passed to show_mem() and show_free_areas() to suppress output in
1146  * various contexts.
1147  */
1148 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1149 
1150 extern void show_free_areas(unsigned int flags);
1151 extern bool skip_free_areas_node(unsigned int flags, int nid);
1152 
1153 int shmem_zero_setup(struct vm_area_struct *);
1154 #ifdef CONFIG_SHMEM
1155 bool shmem_mapping(struct address_space *mapping);
1156 #else
1157 static inline bool shmem_mapping(struct address_space *mapping)
1158 {
1159         return false;
1160 }
1161 #endif
1162 
1163 extern bool can_do_mlock(void);
1164 extern int user_shm_lock(size_t, struct user_struct *);
1165 extern void user_shm_unlock(size_t, struct user_struct *);
1166 
1167 /*
1168  * Parameter block passed down to zap_pte_range in exceptional cases.
1169  */
1170 struct zap_details {
1171         struct address_space *check_mapping;    /* Check page->mapping if set */
1172         pgoff_t first_index;                    /* Lowest page->index to unmap */
1173         pgoff_t last_index;                     /* Highest page->index to unmap */
1174         bool ignore_dirty;                      /* Ignore dirty pages */
1175         bool check_swap_entries;                /* Check also swap entries */
1176 };
1177 
1178 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1179                 pte_t pte);
1180 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1181                                 pmd_t pmd);
1182 
1183 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1184                 unsigned long size);
1185 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1186                 unsigned long size, struct zap_details *);
1187 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1188                 unsigned long start, unsigned long end);
1189 
1190 /**
1191  * mm_walk - callbacks for walk_page_range
1192  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1193  *             this handler is required to be able to handle
1194  *             pmd_trans_huge() pmds.  They may simply choose to
1195  *             split_huge_page() instead of handling it explicitly.
1196  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1197  * @pte_hole: if set, called for each hole at all levels
1198  * @hugetlb_entry: if set, called for each hugetlb entry
1199  * @test_walk: caller specific callback function to determine whether
1200  *             we walk over the current vma or not. A positive returned
1201  *             value means "do page table walk over the current vma,"
1202  *             and a negative one means "abort current page table walk
1203  *             right now." 0 means "skip the current vma."
1204  * @mm:        mm_struct representing the target process of page table walk
1205  * @vma:       vma currently walked (NULL if walking outside vmas)
1206  * @private:   private data for callbacks' usage
1207  *
1208  * (see the comment on walk_page_range() for more details)
1209  */
1210 struct mm_walk {
1211         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1212                          unsigned long next, struct mm_walk *walk);
1213         int (*pte_entry)(pte_t *pte, unsigned long addr,
1214                          unsigned long next, struct mm_walk *walk);
1215         int (*pte_hole)(unsigned long addr, unsigned long next,
1216                         struct mm_walk *walk);
1217         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1218                              unsigned long addr, unsigned long next,
1219                              struct mm_walk *walk);
1220         int (*test_walk)(unsigned long addr, unsigned long next,
1221                         struct mm_walk *walk);
1222         struct mm_struct *mm;
1223         struct vm_area_struct *vma;
1224         void *private;
1225 };
1226 
1227 int walk_page_range(unsigned long addr, unsigned long end,
1228                 struct mm_walk *walk);
1229 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1230 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1231                 unsigned long end, unsigned long floor, unsigned long ceiling);
1232 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1233                         struct vm_area_struct *vma);
1234 void unmap_mapping_range(struct address_space *mapping,
1235                 loff_t const holebegin, loff_t const holelen, int even_cows);
1236 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1237         unsigned long *pfn);
1238 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1239                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1240 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1241                         void *buf, int len, int write);
1242 
1243 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1244                 loff_t const holebegin, loff_t const holelen)
1245 {
1246         unmap_mapping_range(mapping, holebegin, holelen, 0);
1247 }
1248 
1249 extern void truncate_pagecache(struct inode *inode, loff_t new);
1250 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1251 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1252 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1253 int truncate_inode_page(struct address_space *mapping, struct page *page);
1254 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1255 int invalidate_inode_page(struct page *page);
1256 
1257 #ifdef CONFIG_MMU
1258 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1259                 unsigned int flags);
1260 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1261                             unsigned long address, unsigned int fault_flags,
1262                             bool *unlocked);
1263 #else
1264 static inline int handle_mm_fault(struct vm_area_struct *vma,
1265                 unsigned long address, unsigned int flags)
1266 {
1267         /* should never happen if there's no MMU */
1268         BUG();
1269         return VM_FAULT_SIGBUS;
1270 }
1271 static inline int fixup_user_fault(struct task_struct *tsk,
1272                 struct mm_struct *mm, unsigned long address,
1273                 unsigned int fault_flags, bool *unlocked)
1274 {
1275         /* should never happen if there's no MMU */
1276         BUG();
1277         return -EFAULT;
1278 }
1279 #endif
1280 
1281 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1282 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1283                 void *buf, int len, int write);
1284 
1285 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1286                       unsigned long start, unsigned long nr_pages,
1287                       unsigned int foll_flags, struct page **pages,
1288                       struct vm_area_struct **vmas, int *nonblocking);
1289 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1290                             unsigned long start, unsigned long nr_pages,
1291                             int write, int force, struct page **pages,
1292                             struct vm_area_struct **vmas);
1293 long get_user_pages(unsigned long start, unsigned long nr_pages,
1294                             int write, int force, struct page **pages,
1295                             struct vm_area_struct **vmas);
1296 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1297                     int write, int force, struct page **pages, int *locked);
1298 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1299                                unsigned long start, unsigned long nr_pages,
1300                                int write, int force, struct page **pages,
1301                                unsigned int gup_flags);
1302 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1303                     int write, int force, struct page **pages);
1304 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1305                         struct page **pages);
1306 
1307 /* Container for pinned pfns / pages */
1308 struct frame_vector {
1309         unsigned int nr_allocated;      /* Number of frames we have space for */
1310         unsigned int nr_frames; /* Number of frames stored in ptrs array */
1311         bool got_ref;           /* Did we pin pages by getting page ref? */
1312         bool is_pfns;           /* Does array contain pages or pfns? */
1313         void *ptrs[0];          /* Array of pinned pfns / pages. Use
1314                                  * pfns_vector_pages() or pfns_vector_pfns()
1315                                  * for access */
1316 };
1317 
1318 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1319 void frame_vector_destroy(struct frame_vector *vec);
1320 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1321                      bool write, bool force, struct frame_vector *vec);
1322 void put_vaddr_frames(struct frame_vector *vec);
1323 int frame_vector_to_pages(struct frame_vector *vec);
1324 void frame_vector_to_pfns(struct frame_vector *vec);
1325 
1326 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1327 {
1328         return vec->nr_frames;
1329 }
1330 
1331 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1332 {
1333         if (vec->is_pfns) {
1334                 int err = frame_vector_to_pages(vec);
1335 
1336                 if (err)
1337                         return ERR_PTR(err);
1338         }
1339         return (struct page **)(vec->ptrs);
1340 }
1341 
1342 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1343 {
1344         if (!vec->is_pfns)
1345                 frame_vector_to_pfns(vec);
1346         return (unsigned long *)(vec->ptrs);
1347 }
1348 
1349 struct kvec;
1350 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1351                         struct page **pages);
1352 int get_kernel_page(unsigned long start, int write, struct page **pages);
1353 struct page *get_dump_page(unsigned long addr);
1354 
1355 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1356 extern void do_invalidatepage(struct page *page, unsigned int offset,
1357                               unsigned int length);
1358 
1359 int __set_page_dirty_nobuffers(struct page *page);
1360 int __set_page_dirty_no_writeback(struct page *page);
1361 int redirty_page_for_writepage(struct writeback_control *wbc,
1362                                 struct page *page);
1363 void account_page_dirtied(struct page *page, struct address_space *mapping);
1364 void account_page_cleaned(struct page *page, struct address_space *mapping,
1365                           struct bdi_writeback *wb);
1366 int set_page_dirty(struct page *page);
1367 int set_page_dirty_lock(struct page *page);
1368 void cancel_dirty_page(struct page *page);
1369 int clear_page_dirty_for_io(struct page *page);
1370 
1371 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1372 
1373 /* Is the vma a continuation of the stack vma above it? */
1374 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1375 {
1376         return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1377 }
1378 
1379 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1380 {
1381         return !vma->vm_ops;
1382 }
1383 
1384 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1385                                              unsigned long addr)
1386 {
1387         return (vma->vm_flags & VM_GROWSDOWN) &&
1388                 (vma->vm_start == addr) &&
1389                 !vma_growsdown(vma->vm_prev, addr);
1390 }
1391 
1392 /* Is the vma a continuation of the stack vma below it? */
1393 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1394 {
1395         return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1396 }
1397 
1398 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1399                                            unsigned long addr)
1400 {
1401         return (vma->vm_flags & VM_GROWSUP) &&
1402                 (vma->vm_end == addr) &&
1403                 !vma_growsup(vma->vm_next, addr);
1404 }
1405 
1406 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1407 
1408 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1409                 unsigned long old_addr, struct vm_area_struct *new_vma,
1410                 unsigned long new_addr, unsigned long len,
1411                 bool need_rmap_locks);
1412 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1413                               unsigned long end, pgprot_t newprot,
1414                               int dirty_accountable, int prot_numa);
1415 extern int mprotect_fixup(struct vm_area_struct *vma,
1416                           struct vm_area_struct **pprev, unsigned long start,
1417                           unsigned long end, unsigned long newflags);
1418 
1419 /*
1420  * doesn't attempt to fault and will return short.
1421  */
1422 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1423                           struct page **pages);
1424 /*
1425  * per-process(per-mm_struct) statistics.
1426  */
1427 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1428 {
1429         long val = atomic_long_read(&mm->rss_stat.count[member]);
1430 
1431 #ifdef SPLIT_RSS_COUNTING
1432         /*
1433          * counter is updated in asynchronous manner and may go to minus.
1434          * But it's never be expected number for users.
1435          */
1436         if (val < 0)
1437                 val = 0;
1438 #endif
1439         return (unsigned long)val;
1440 }
1441 
1442 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1443 {
1444         atomic_long_add(value, &mm->rss_stat.count[member]);
1445 }
1446 
1447 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1448 {
1449         atomic_long_inc(&mm->rss_stat.count[member]);
1450 }
1451 
1452 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1453 {
1454         atomic_long_dec(&mm->rss_stat.count[member]);
1455 }
1456 
1457 /* Optimized variant when page is already known not to be PageAnon */
1458 static inline int mm_counter_file(struct page *page)
1459 {
1460         if (PageSwapBacked(page))
1461                 return MM_SHMEMPAGES;
1462         return MM_FILEPAGES;
1463 }
1464 
1465 static inline int mm_counter(struct page *page)
1466 {
1467         if (PageAnon(page))
1468                 return MM_ANONPAGES;
1469         return mm_counter_file(page);
1470 }
1471 
1472 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1473 {
1474         return get_mm_counter(mm, MM_FILEPAGES) +
1475                 get_mm_counter(mm, MM_ANONPAGES) +
1476                 get_mm_counter(mm, MM_SHMEMPAGES);
1477 }
1478 
1479 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1480 {
1481         return max(mm->hiwater_rss, get_mm_rss(mm));
1482 }
1483 
1484 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1485 {
1486         return max(mm->hiwater_vm, mm->total_vm);
1487 }
1488 
1489 static inline void update_hiwater_rss(struct mm_struct *mm)
1490 {
1491         unsigned long _rss = get_mm_rss(mm);
1492 
1493         if ((mm)->hiwater_rss < _rss)
1494                 (mm)->hiwater_rss = _rss;
1495 }
1496 
1497 static inline void update_hiwater_vm(struct mm_struct *mm)
1498 {
1499         if (mm->hiwater_vm < mm->total_vm)
1500                 mm->hiwater_vm = mm->total_vm;
1501 }
1502 
1503 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1504 {
1505         mm->hiwater_rss = get_mm_rss(mm);
1506 }
1507 
1508 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1509                                          struct mm_struct *mm)
1510 {
1511         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1512 
1513         if (*maxrss < hiwater_rss)
1514                 *maxrss = hiwater_rss;
1515 }
1516 
1517 #if defined(SPLIT_RSS_COUNTING)
1518 void sync_mm_rss(struct mm_struct *mm);
1519 #else
1520 static inline void sync_mm_rss(struct mm_struct *mm)
1521 {
1522 }
1523 #endif
1524 
1525 #ifndef __HAVE_ARCH_PTE_DEVMAP
1526 static inline int pte_devmap(pte_t pte)
1527 {
1528         return 0;
1529 }
1530 #endif
1531 
1532 int vma_wants_writenotify(struct vm_area_struct *vma);
1533 
1534 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1535                                spinlock_t **ptl);
1536 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1537                                     spinlock_t **ptl)
1538 {
1539         pte_t *ptep;
1540         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1541         return ptep;
1542 }
1543 
1544 #ifdef __PAGETABLE_PUD_FOLDED
1545 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1546                                                 unsigned long address)
1547 {
1548         return 0;
1549 }
1550 #else
1551 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1552 #endif
1553 
1554 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1555 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1556                                                 unsigned long address)
1557 {
1558         return 0;
1559 }
1560 
1561 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1562 
1563 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1564 {
1565         return 0;
1566 }
1567 
1568 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1569 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1570 
1571 #else
1572 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1573 
1574 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1575 {
1576         atomic_long_set(&mm->nr_pmds, 0);
1577 }
1578 
1579 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1580 {
1581         return atomic_long_read(&mm->nr_pmds);
1582 }
1583 
1584 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1585 {
1586         atomic_long_inc(&mm->nr_pmds);
1587 }
1588 
1589 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1590 {
1591         atomic_long_dec(&mm->nr_pmds);
1592 }
1593 #endif
1594 
1595 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1596 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1597 
1598 /*
1599  * The following ifdef needed to get the 4level-fixup.h header to work.
1600  * Remove it when 4level-fixup.h has been removed.
1601  */
1602 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1603 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1604 {
1605         return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1606                 NULL: pud_offset(pgd, address);
1607 }
1608 
1609 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1610 {
1611         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1612                 NULL: pmd_offset(pud, address);
1613 }
1614 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1615 
1616 #if USE_SPLIT_PTE_PTLOCKS
1617 #if ALLOC_SPLIT_PTLOCKS
1618 void __init ptlock_cache_init(void);
1619 extern bool ptlock_alloc(struct page *page);
1620 extern void ptlock_free(struct page *page);
1621 
1622 static inline spinlock_t *ptlock_ptr(struct page *page)
1623 {
1624         return page->ptl;
1625 }
1626 #else /* ALLOC_SPLIT_PTLOCKS */
1627 static inline void ptlock_cache_init(void)
1628 {
1629 }
1630 
1631 static inline bool ptlock_alloc(struct page *page)
1632 {
1633         return true;
1634 }
1635 
1636 static inline void ptlock_free(struct page *page)
1637 {
1638 }
1639 
1640 static inline spinlock_t *ptlock_ptr(struct page *page)
1641 {
1642         return &page->ptl;
1643 }
1644 #endif /* ALLOC_SPLIT_PTLOCKS */
1645 
1646 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1647 {
1648         return ptlock_ptr(pmd_page(*pmd));
1649 }
1650 
1651 static inline bool ptlock_init(struct page *page)
1652 {
1653         /*
1654          * prep_new_page() initialize page->private (and therefore page->ptl)
1655          * with 0. Make sure nobody took it in use in between.
1656          *
1657          * It can happen if arch try to use slab for page table allocation:
1658          * slab code uses page->slab_cache, which share storage with page->ptl.
1659          */
1660         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1661         if (!ptlock_alloc(page))
1662                 return false;
1663         spin_lock_init(ptlock_ptr(page));
1664         return true;
1665 }
1666 
1667 /* Reset page->mapping so free_pages_check won't complain. */
1668 static inline void pte_lock_deinit(struct page *page)
1669 {
1670         page->mapping = NULL;
1671         ptlock_free(page);
1672 }
1673 
1674 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1675 /*
1676  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1677  */
1678 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1679 {
1680         return &mm->page_table_lock;
1681 }
1682 static inline void ptlock_cache_init(void) {}
1683 static inline bool ptlock_init(struct page *page) { return true; }
1684 static inline void pte_lock_deinit(struct page *page) {}
1685 #endif /* USE_SPLIT_PTE_PTLOCKS */
1686 
1687 static inline void pgtable_init(void)
1688 {
1689         ptlock_cache_init();
1690         pgtable_cache_init();
1691 }
1692 
1693 static inline bool pgtable_page_ctor(struct page *page)
1694 {
1695         if (!ptlock_init(page))
1696                 return false;
1697         inc_zone_page_state(page, NR_PAGETABLE);
1698         return true;
1699 }
1700 
1701 static inline void pgtable_page_dtor(struct page *page)
1702 {
1703         pte_lock_deinit(page);
1704         dec_zone_page_state(page, NR_PAGETABLE);
1705 }
1706 
1707 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1708 ({                                                      \
1709         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1710         pte_t *__pte = pte_offset_map(pmd, address);    \
1711         *(ptlp) = __ptl;                                \
1712         spin_lock(__ptl);                               \
1713         __pte;                                          \
1714 })
1715 
1716 #define pte_unmap_unlock(pte, ptl)      do {            \
1717         spin_unlock(ptl);                               \
1718         pte_unmap(pte);                                 \
1719 } while (0)
1720 
1721 #define pte_alloc(mm, pmd, address)                     \
1722         (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1723 
1724 #define pte_alloc_map(mm, pmd, address)                 \
1725         (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1726 
1727 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1728         (pte_alloc(mm, pmd, address) ?                  \
1729                  NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1730 
1731 #define pte_alloc_kernel(pmd, address)                  \
1732         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1733                 NULL: pte_offset_kernel(pmd, address))
1734 
1735 #if USE_SPLIT_PMD_PTLOCKS
1736 
1737 static struct page *pmd_to_page(pmd_t *pmd)
1738 {
1739         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1740         return virt_to_page((void *)((unsigned long) pmd & mask));
1741 }
1742 
1743 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1744 {
1745         return ptlock_ptr(pmd_to_page(pmd));
1746 }
1747 
1748 static inline bool pgtable_pmd_page_ctor(struct page *page)
1749 {
1750 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1751         page->pmd_huge_pte = NULL;
1752 #endif
1753         return ptlock_init(page);
1754 }
1755 
1756 static inline void pgtable_pmd_page_dtor(struct page *page)
1757 {
1758 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1759         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1760 #endif
1761         ptlock_free(page);
1762 }
1763 
1764 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1765 
1766 #else
1767 
1768 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1769 {
1770         return &mm->page_table_lock;
1771 }
1772 
1773 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1774 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1775 
1776 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1777 
1778 #endif
1779 
1780 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1781 {
1782         spinlock_t *ptl = pmd_lockptr(mm, pmd);
1783         spin_lock(ptl);
1784         return ptl;
1785 }
1786 
1787 extern void free_area_init(unsigned long * zones_size);
1788 extern void free_area_init_node(int nid, unsigned long * zones_size,
1789                 unsigned long zone_start_pfn, unsigned long *zholes_size);
1790 extern void free_initmem(void);
1791 
1792 /*
1793  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1794  * into the buddy system. The freed pages will be poisoned with pattern
1795  * "poison" if it's within range [0, UCHAR_MAX].
1796  * Return pages freed into the buddy system.
1797  */
1798 extern unsigned long free_reserved_area(void *start, void *end,
1799                                         int poison, char *s);
1800 
1801 #ifdef  CONFIG_HIGHMEM
1802 /*
1803  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1804  * and totalram_pages.
1805  */
1806 extern void free_highmem_page(struct page *page);
1807 #endif
1808 
1809 extern void adjust_managed_page_count(struct page *page, long count);
1810 extern void mem_init_print_info(const char *str);
1811 
1812 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1813 
1814 /* Free the reserved page into the buddy system, so it gets managed. */
1815 static inline void __free_reserved_page(struct page *page)
1816 {
1817         ClearPageReserved(page);
1818         init_page_count(page);
1819         __free_page(page);
1820 }
1821 
1822 static inline void free_reserved_page(struct page *page)
1823 {
1824         __free_reserved_page(page);
1825         adjust_managed_page_count(page, 1);
1826 }
1827 
1828 static inline void mark_page_reserved(struct page *page)
1829 {
1830         SetPageReserved(page);
1831         adjust_managed_page_count(page, -1);
1832 }
1833 
1834 /*
1835  * Default method to free all the __init memory into the buddy system.
1836  * The freed pages will be poisoned with pattern "poison" if it's within
1837  * range [0, UCHAR_MAX].
1838  * Return pages freed into the buddy system.
1839  */
1840 static inline unsigned long free_initmem_default(int poison)
1841 {
1842         extern char __init_begin[], __init_end[];
1843 
1844         return free_reserved_area(&__init_begin, &__init_end,
1845                                   poison, "unused kernel");
1846 }
1847 
1848 static inline unsigned long get_num_physpages(void)
1849 {
1850         int nid;
1851         unsigned long phys_pages = 0;
1852 
1853         for_each_online_node(nid)
1854                 phys_pages += node_present_pages(nid);
1855 
1856         return phys_pages;
1857 }
1858 
1859 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1860 /*
1861  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1862  * zones, allocate the backing mem_map and account for memory holes in a more
1863  * architecture independent manner. This is a substitute for creating the
1864  * zone_sizes[] and zholes_size[] arrays and passing them to
1865  * free_area_init_node()
1866  *
1867  * An architecture is expected to register range of page frames backed by
1868  * physical memory with memblock_add[_node]() before calling
1869  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1870  * usage, an architecture is expected to do something like
1871  *
1872  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1873  *                                                       max_highmem_pfn};
1874  * for_each_valid_physical_page_range()
1875  *      memblock_add_node(base, size, nid)
1876  * free_area_init_nodes(max_zone_pfns);
1877  *
1878  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1879  * registered physical page range.  Similarly
1880  * sparse_memory_present_with_active_regions() calls memory_present() for
1881  * each range when SPARSEMEM is enabled.
1882  *
1883  * See mm/page_alloc.c for more information on each function exposed by
1884  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1885  */
1886 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1887 unsigned long node_map_pfn_alignment(void);
1888 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1889                                                 unsigned long end_pfn);
1890 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1891                                                 unsigned long end_pfn);
1892 extern void get_pfn_range_for_nid(unsigned int nid,
1893                         unsigned long *start_pfn, unsigned long *end_pfn);
1894 extern unsigned long find_min_pfn_with_active_regions(void);
1895 extern void free_bootmem_with_active_regions(int nid,
1896                                                 unsigned long max_low_pfn);
1897 extern void sparse_memory_present_with_active_regions(int nid);
1898 
1899 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1900 
1901 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1902     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1903 static inline int __early_pfn_to_nid(unsigned long pfn,
1904                                         struct mminit_pfnnid_cache *state)
1905 {
1906         return 0;
1907 }
1908 #else
1909 /* please see mm/page_alloc.c */
1910 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1911 /* there is a per-arch backend function. */
1912 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1913                                         struct mminit_pfnnid_cache *state);
1914 #endif
1915 
1916 extern void set_dma_reserve(unsigned long new_dma_reserve);
1917 extern void memmap_init_zone(unsigned long, int, unsigned long,
1918                                 unsigned long, enum memmap_context);
1919 extern void setup_per_zone_wmarks(void);
1920 extern int __meminit init_per_zone_wmark_min(void);
1921 extern void mem_init(void);
1922 extern void __init mmap_init(void);
1923 extern void show_mem(unsigned int flags);
1924 extern long si_mem_available(void);
1925 extern void si_meminfo(struct sysinfo * val);
1926 extern void si_meminfo_node(struct sysinfo *val, int nid);
1927 
1928 extern __printf(3, 4)
1929 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1930                 const char *fmt, ...);
1931 
1932 extern void setup_per_cpu_pageset(void);
1933 
1934 extern void zone_pcp_update(struct zone *zone);
1935 extern void zone_pcp_reset(struct zone *zone);
1936 
1937 /* page_alloc.c */
1938 extern int min_free_kbytes;
1939 extern int watermark_scale_factor;
1940 
1941 /* nommu.c */
1942 extern atomic_long_t mmap_pages_allocated;
1943 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1944 
1945 /* interval_tree.c */
1946 void vma_interval_tree_insert(struct vm_area_struct *node,
1947                               struct rb_root *root);
1948 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1949                                     struct vm_area_struct *prev,
1950                                     struct rb_root *root);
1951 void vma_interval_tree_remove(struct vm_area_struct *node,
1952                               struct rb_root *root);
1953 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1954                                 unsigned long start, unsigned long last);
1955 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1956                                 unsigned long start, unsigned long last);
1957 
1958 #define vma_interval_tree_foreach(vma, root, start, last)               \
1959         for (vma = vma_interval_tree_iter_first(root, start, last);     \
1960              vma; vma = vma_interval_tree_iter_next(vma, start, last))
1961 
1962 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1963                                    struct rb_root *root);
1964 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1965                                    struct rb_root *root);
1966 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1967         struct rb_root *root, unsigned long start, unsigned long last);
1968 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1969         struct anon_vma_chain *node, unsigned long start, unsigned long last);
1970 #ifdef CONFIG_DEBUG_VM_RB
1971 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1972 #endif
1973 
1974 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1975         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1976              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1977 
1978 /* mmap.c */
1979 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1980 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1981         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1982 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1983         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1984         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1985         struct mempolicy *, struct vm_userfaultfd_ctx);
1986 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1987 extern int split_vma(struct mm_struct *,
1988         struct vm_area_struct *, unsigned long addr, int new_below);
1989 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1990 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1991         struct rb_node **, struct rb_node *);
1992 extern void unlink_file_vma(struct vm_area_struct *);
1993 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1994         unsigned long addr, unsigned long len, pgoff_t pgoff,
1995         bool *need_rmap_locks);
1996 extern void exit_mmap(struct mm_struct *);
1997 
1998 static inline int check_data_rlimit(unsigned long rlim,
1999                                     unsigned long new,
2000                                     unsigned long start,
2001                                     unsigned long end_data,
2002                                     unsigned long start_data)
2003 {
2004         if (rlim < RLIM_INFINITY) {
2005                 if (((new - start) + (end_data - start_data)) > rlim)
2006                         return -ENOSPC;
2007         }
2008 
2009         return 0;
2010 }
2011 
2012 extern int mm_take_all_locks(struct mm_struct *mm);
2013 extern void mm_drop_all_locks(struct mm_struct *mm);
2014 
2015 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2016 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2017 extern struct file *get_task_exe_file(struct task_struct *task);
2018 
2019 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2020 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2021 
2022 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2023                                    unsigned long addr, unsigned long len,
2024                                    unsigned long flags,
2025                                    const struct vm_special_mapping *spec);
2026 /* This is an obsolete alternative to _install_special_mapping. */
2027 extern int install_special_mapping(struct mm_struct *mm,
2028                                    unsigned long addr, unsigned long len,
2029                                    unsigned long flags, struct page **pages);
2030 
2031 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2032 
2033 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2034         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2035 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2036         unsigned long len, unsigned long prot, unsigned long flags,
2037         vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2038 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2039 
2040 static inline unsigned long
2041 do_mmap_pgoff(struct file *file, unsigned long addr,
2042         unsigned long len, unsigned long prot, unsigned long flags,
2043         unsigned long pgoff, unsigned long *populate)
2044 {
2045         return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2046 }
2047 
2048 #ifdef CONFIG_MMU
2049 extern int __mm_populate(unsigned long addr, unsigned long len,
2050                          int ignore_errors);
2051 static inline void mm_populate(unsigned long addr, unsigned long len)
2052 {
2053         /* Ignore errors */
2054         (void) __mm_populate(addr, len, 1);
2055 }
2056 #else
2057 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2058 #endif
2059 
2060 /* These take the mm semaphore themselves */
2061 extern int __must_check vm_brk(unsigned long, unsigned long);
2062 extern int vm_munmap(unsigned long, size_t);
2063 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2064         unsigned long, unsigned long,
2065         unsigned long, unsigned long);
2066 
2067 struct vm_unmapped_area_info {
2068 #define VM_UNMAPPED_AREA_TOPDOWN 1
2069         unsigned long flags;
2070         unsigned long length;
2071         unsigned long low_limit;
2072         unsigned long high_limit;
2073         unsigned long align_mask;
2074         unsigned long align_offset;
2075 };
2076 
2077 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2078 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2079 
2080 /*
2081  * Search for an unmapped address range.
2082  *
2083  * We are looking for a range that:
2084  * - does not intersect with any VMA;
2085  * - is contained within the [low_limit, high_limit) interval;
2086  * - is at least the desired size.
2087  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2088  */
2089 static inline unsigned long
2090 vm_unmapped_area(struct vm_unmapped_area_info *info)
2091 {
2092         if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2093                 return unmapped_area_topdown(info);
2094         else
2095                 return unmapped_area(info);
2096 }
2097 
2098 /* truncate.c */
2099 extern void truncate_inode_pages(struct address_space *, loff_t);
2100 extern void truncate_inode_pages_range(struct address_space *,
2101                                        loff_t lstart, loff_t lend);
2102 extern void truncate_inode_pages_final(struct address_space *);
2103 
2104 /* generic vm_area_ops exported for stackable file systems */
2105 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2106 extern void filemap_map_pages(struct fault_env *fe,
2107                 pgoff_t start_pgoff, pgoff_t end_pgoff);
2108 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2109 
2110 /* mm/page-writeback.c */
2111 int write_one_page(struct page *page, int wait);
2112 void task_dirty_inc(struct task_struct *tsk);
2113 
2114 /* readahead.c */
2115 #define VM_MAX_READAHEAD        128     /* kbytes */
2116 #define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
2117 
2118 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2119                         pgoff_t offset, unsigned long nr_to_read);
2120 
2121 void page_cache_sync_readahead(struct address_space *mapping,
2122                                struct file_ra_state *ra,
2123                                struct file *filp,
2124                                pgoff_t offset,
2125                                unsigned long size);
2126 
2127 void page_cache_async_readahead(struct address_space *mapping,
2128                                 struct file_ra_state *ra,
2129                                 struct file *filp,
2130                                 struct page *pg,
2131                                 pgoff_t offset,
2132                                 unsigned long size);
2133 
2134 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2135 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2136 
2137 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2138 extern int expand_downwards(struct vm_area_struct *vma,
2139                 unsigned long address);
2140 #if VM_GROWSUP
2141 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2142 #else
2143   #define expand_upwards(vma, address) (0)
2144 #endif
2145 
2146 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2147 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2148 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2149                                              struct vm_area_struct **pprev);
2150 
2151 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2152    NULL if none.  Assume start_addr < end_addr. */
2153 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2154 {
2155         struct vm_area_struct * vma = find_vma(mm,start_addr);
2156 
2157         if (vma && end_addr <= vma->vm_start)
2158                 vma = NULL;
2159         return vma;
2160 }
2161 
2162 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2163 {
2164         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2165 }
2166 
2167 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2168 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2169                                 unsigned long vm_start, unsigned long vm_end)
2170 {
2171         struct vm_area_struct *vma = find_vma(mm, vm_start);
2172 
2173         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2174                 vma = NULL;
2175 
2176         return vma;
2177 }
2178 
2179 #ifdef CONFIG_MMU
2180 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2181 void vma_set_page_prot(struct vm_area_struct *vma);
2182 #else
2183 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2184 {
2185         return __pgprot(0);
2186 }
2187 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2188 {
2189         vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2190 }
2191 #endif
2192 
2193 #ifdef CONFIG_NUMA_BALANCING
2194 unsigned long change_prot_numa(struct vm_area_struct *vma,
2195                         unsigned long start, unsigned long end);
2196 #endif
2197 
2198 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2199 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2200                         unsigned long pfn, unsigned long size, pgprot_t);
2201 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2202 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2203                         unsigned long pfn);
2204 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2205                         unsigned long pfn, pgprot_t pgprot);
2206 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2207                         pfn_t pfn);
2208 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2209 
2210 
2211 struct page *follow_page_mask(struct vm_area_struct *vma,
2212                               unsigned long address, unsigned int foll_flags,
2213                               unsigned int *page_mask);
2214 
2215 static inline struct page *follow_page(struct vm_area_struct *vma,
2216                 unsigned long address, unsigned int foll_flags)
2217 {
2218         unsigned int unused_page_mask;
2219         return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2220 }
2221 
2222 #define FOLL_WRITE      0x01    /* check pte is writable */
2223 #define FOLL_TOUCH      0x02    /* mark page accessed */
2224 #define FOLL_GET        0x04    /* do get_page on page */
2225 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2226 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2227 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2228                                  * and return without waiting upon it */
2229 #define FOLL_POPULATE   0x40    /* fault in page */
2230 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2231 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2232 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2233 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2234 #define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2235 #define FOLL_MLOCK      0x1000  /* lock present pages */
2236 #define FOLL_REMOTE     0x2000  /* we are working on non-current tsk/mm */
2237 
2238 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2239                         void *data);
2240 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2241                                unsigned long size, pte_fn_t fn, void *data);
2242 
2243 
2244 #ifdef CONFIG_PAGE_POISONING
2245 extern bool page_poisoning_enabled(void);
2246 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2247 extern bool page_is_poisoned(struct page *page);
2248 #else
2249 static inline bool page_poisoning_enabled(void) { return false; }
2250 static inline void kernel_poison_pages(struct page *page, int numpages,
2251                                         int enable) { }
2252 static inline bool page_is_poisoned(struct page *page) { return false; }
2253 #endif
2254 
2255 #ifdef CONFIG_DEBUG_PAGEALLOC
2256 extern bool _debug_pagealloc_enabled;
2257 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2258 
2259 static inline bool debug_pagealloc_enabled(void)
2260 {
2261         return _debug_pagealloc_enabled;
2262 }
2263 
2264 static inline void
2265 kernel_map_pages(struct page *page, int numpages, int enable)
2266 {
2267         if (!debug_pagealloc_enabled())
2268                 return;
2269 
2270         __kernel_map_pages(page, numpages, enable);
2271 }
2272 #ifdef CONFIG_HIBERNATION
2273 extern bool kernel_page_present(struct page *page);
2274 #endif  /* CONFIG_HIBERNATION */
2275 #else   /* CONFIG_DEBUG_PAGEALLOC */
2276 static inline void
2277 kernel_map_pages(struct page *page, int numpages, int enable) {}
2278 #ifdef CONFIG_HIBERNATION
2279 static inline bool kernel_page_present(struct page *page) { return true; }
2280 #endif  /* CONFIG_HIBERNATION */
2281 static inline bool debug_pagealloc_enabled(void)
2282 {
2283         return false;
2284 }
2285 #endif  /* CONFIG_DEBUG_PAGEALLOC */
2286 
2287 #ifdef __HAVE_ARCH_GATE_AREA
2288 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2289 extern int in_gate_area_no_mm(unsigned long addr);
2290 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2291 #else
2292 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2293 {
2294         return NULL;
2295 }
2296 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2297 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2298 {
2299         return 0;
2300 }
2301 #endif  /* __HAVE_ARCH_GATE_AREA */
2302 
2303 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2304 
2305 #ifdef CONFIG_SYSCTL
2306 extern int sysctl_drop_caches;
2307 int drop_caches_sysctl_handler(struct ctl_table *, int,
2308                                         void __user *, size_t *, loff_t *);
2309 #endif
2310 
2311 void drop_slab(void);
2312 void drop_slab_node(int nid);
2313 
2314 #ifndef CONFIG_MMU
2315 #define randomize_va_space 0
2316 #else
2317 extern int randomize_va_space;
2318 #endif
2319 
2320 const char * arch_vma_name(struct vm_area_struct *vma);
2321 void print_vma_addr(char *prefix, unsigned long rip);
2322 
2323 void sparse_mem_maps_populate_node(struct page **map_map,
2324                                    unsigned long pnum_begin,
2325                                    unsigned long pnum_end,
2326                                    unsigned long map_count,
2327                                    int nodeid);
2328 
2329 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2330 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2331 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2332 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2333 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2334 void *vmemmap_alloc_block(unsigned long size, int node);
2335 struct vmem_altmap;
2336 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2337                 struct vmem_altmap *altmap);
2338 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2339 {
2340         return __vmemmap_alloc_block_buf(size, node, NULL);
2341 }
2342 
2343 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2344 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2345                                int node);
2346 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2347 void vmemmap_populate_print_last(void);
2348 #ifdef CONFIG_MEMORY_HOTPLUG
2349 void vmemmap_free(unsigned long start, unsigned long end);
2350 #endif
2351 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2352                                   unsigned long size);
2353 
2354 enum mf_flags {
2355         MF_COUNT_INCREASED = 1 << 0,
2356         MF_ACTION_REQUIRED = 1 << 1,
2357         MF_MUST_KILL = 1 << 2,
2358         MF_SOFT_OFFLINE = 1 << 3,
2359 };
2360 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2361 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2362 extern int unpoison_memory(unsigned long pfn);
2363 extern int get_hwpoison_page(struct page *page);
2364 #define put_hwpoison_page(page) put_page(page)
2365 extern int sysctl_memory_failure_early_kill;
2366 extern int sysctl_memory_failure_recovery;
2367 extern void shake_page(struct page *p, int access);
2368 extern atomic_long_t num_poisoned_pages;
2369 extern int soft_offline_page(struct page *page, int flags);
2370 
2371 
2372 /*
2373  * Error handlers for various types of pages.
2374  */
2375 enum mf_result {
2376         MF_IGNORED,     /* Error: cannot be handled */
2377         MF_FAILED,      /* Error: handling failed */
2378         MF_DELAYED,     /* Will be handled later */
2379         MF_RECOVERED,   /* Successfully recovered */
2380 };
2381 
2382 enum mf_action_page_type {
2383         MF_MSG_KERNEL,
2384         MF_MSG_KERNEL_HIGH_ORDER,
2385         MF_MSG_SLAB,
2386         MF_MSG_DIFFERENT_COMPOUND,
2387         MF_MSG_POISONED_HUGE,
2388         MF_MSG_HUGE,
2389         MF_MSG_FREE_HUGE,
2390         MF_MSG_UNMAP_FAILED,
2391         MF_MSG_DIRTY_SWAPCACHE,
2392         MF_MSG_CLEAN_SWAPCACHE,
2393         MF_MSG_DIRTY_MLOCKED_LRU,
2394         MF_MSG_CLEAN_MLOCKED_LRU,
2395         MF_MSG_DIRTY_UNEVICTABLE_LRU,
2396         MF_MSG_CLEAN_UNEVICTABLE_LRU,
2397         MF_MSG_DIRTY_LRU,
2398         MF_MSG_CLEAN_LRU,
2399         MF_MSG_TRUNCATED_LRU,
2400         MF_MSG_BUDDY,
2401         MF_MSG_BUDDY_2ND,
2402         MF_MSG_UNKNOWN,
2403 };
2404 
2405 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2406 extern void clear_huge_page(struct page *page,
2407                             unsigned long addr,
2408                             unsigned int pages_per_huge_page);
2409 extern void copy_user_huge_page(struct page *dst, struct page *src,
2410                                 unsigned long addr, struct vm_area_struct *vma,
2411                                 unsigned int pages_per_huge_page);
2412 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2413 
2414 extern struct page_ext_operations debug_guardpage_ops;
2415 extern struct page_ext_operations page_poisoning_ops;
2416 
2417 #ifdef CONFIG_DEBUG_PAGEALLOC
2418 extern unsigned int _debug_guardpage_minorder;
2419 extern bool _debug_guardpage_enabled;
2420 
2421 static inline unsigned int debug_guardpage_minorder(void)
2422 {
2423         return _debug_guardpage_minorder;
2424 }
2425 
2426 static inline bool debug_guardpage_enabled(void)
2427 {
2428         return _debug_guardpage_enabled;
2429 }
2430 
2431 static inline bool page_is_guard(struct page *page)
2432 {
2433         struct page_ext *page_ext;
2434 
2435         if (!debug_guardpage_enabled())
2436                 return false;
2437 
2438         page_ext = lookup_page_ext(page);
2439         if (unlikely(!page_ext))
2440                 return false;
2441 
2442         return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2443 }
2444 #else
2445 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2446 static inline bool debug_guardpage_enabled(void) { return false; }
2447 static inline bool page_is_guard(struct page *page) { return false; }
2448 #endif /* CONFIG_DEBUG_PAGEALLOC */
2449 
2450 #if MAX_NUMNODES > 1
2451 void __init setup_nr_node_ids(void);
2452 #else
2453 static inline void setup_nr_node_ids(void) {}
2454 #endif
2455 
2456 #endif /* __KERNEL__ */
2457 #endif /* _LINUX_MM_H */
2458 

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