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

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

Version: ~ [ linux-5.8-rc3 ] ~ [ linux-5.7.5 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.48 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.129 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.185 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.228 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.228 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /* SPDX-License-Identifier: GPL-2.0 */
  2 #ifndef _LINUX_MM_H
  3 #define _LINUX_MM_H
  4 
  5 #include <linux/errno.h>
  6 
  7 #ifdef __KERNEL__
  8 
  9 #include <linux/mmdebug.h>
 10 #include <linux/gfp.h>
 11 #include <linux/bug.h>
 12 #include <linux/list.h>
 13 #include <linux/mmzone.h>
 14 #include <linux/rbtree.h>
 15 #include <linux/atomic.h>
 16 #include <linux/debug_locks.h>
 17 #include <linux/mm_types.h>
 18 #include <linux/range.h>
 19 #include <linux/pfn.h>
 20 #include <linux/percpu-refcount.h>
 21 #include <linux/bit_spinlock.h>
 22 #include <linux/shrinker.h>
 23 #include <linux/resource.h>
 24 #include <linux/page_ext.h>
 25 #include <linux/err.h>
 26 #include <linux/page_ref.h>
 27 #include <linux/memremap.h>
 28 #include <linux/overflow.h>
 29 #include <linux/sizes.h>
 30 
 31 struct mempolicy;
 32 struct anon_vma;
 33 struct anon_vma_chain;
 34 struct file_ra_state;
 35 struct user_struct;
 36 struct writeback_control;
 37 struct bdi_writeback;
 38 
 39 void init_mm_internals(void);
 40 
 41 #ifndef CONFIG_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
 42 extern unsigned long max_mapnr;
 43 
 44 static inline void set_max_mapnr(unsigned long limit)
 45 {
 46         max_mapnr = limit;
 47 }
 48 #else
 49 static inline void set_max_mapnr(unsigned long limit) { }
 50 #endif
 51 
 52 extern atomic_long_t _totalram_pages;
 53 static inline unsigned long totalram_pages(void)
 54 {
 55         return (unsigned long)atomic_long_read(&_totalram_pages);
 56 }
 57 
 58 static inline void totalram_pages_inc(void)
 59 {
 60         atomic_long_inc(&_totalram_pages);
 61 }
 62 
 63 static inline void totalram_pages_dec(void)
 64 {
 65         atomic_long_dec(&_totalram_pages);
 66 }
 67 
 68 static inline void totalram_pages_add(long count)
 69 {
 70         atomic_long_add(count, &_totalram_pages);
 71 }
 72 
 73 static inline void totalram_pages_set(long val)
 74 {
 75         atomic_long_set(&_totalram_pages, val);
 76 }
 77 
 78 extern void * high_memory;
 79 extern int page_cluster;
 80 
 81 #ifdef CONFIG_SYSCTL
 82 extern int sysctl_legacy_va_layout;
 83 #else
 84 #define sysctl_legacy_va_layout 0
 85 #endif
 86 
 87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
 88 extern const int mmap_rnd_bits_min;
 89 extern const int mmap_rnd_bits_max;
 90 extern int mmap_rnd_bits __read_mostly;
 91 #endif
 92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
 93 extern const int mmap_rnd_compat_bits_min;
 94 extern const int mmap_rnd_compat_bits_max;
 95 extern int mmap_rnd_compat_bits __read_mostly;
 96 #endif
 97 
 98 #include <asm/page.h>
 99 #include <asm/pgtable.h>
100 #include <asm/processor.h>
101 
102 /*
103  * Architectures that support memory tagging (assigning tags to memory regions,
104  * embedding these tags into addresses that point to these memory regions, and
105  * checking that the memory and the pointer tags match on memory accesses)
106  * redefine this macro to strip tags from pointers.
107  * It's defined as noop for arcitectures that don't support memory tagging.
108  */
109 #ifndef untagged_addr
110 #define untagged_addr(addr) (addr)
111 #endif
112 
113 #ifndef __pa_symbol
114 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
115 #endif
116 
117 #ifndef page_to_virt
118 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 #endif
120 
121 #ifndef lm_alias
122 #define lm_alias(x)     __va(__pa_symbol(x))
123 #endif
124 
125 /*
126  * To prevent common memory management code establishing
127  * a zero page mapping on a read fault.
128  * This macro should be defined within <asm/pgtable.h>.
129  * s390 does this to prevent multiplexing of hardware bits
130  * related to the physical page in case of virtualization.
131  */
132 #ifndef mm_forbids_zeropage
133 #define mm_forbids_zeropage(X)  (0)
134 #endif
135 
136 /*
137  * On some architectures it is expensive to call memset() for small sizes.
138  * If an architecture decides to implement their own version of
139  * mm_zero_struct_page they should wrap the defines below in a #ifndef and
140  * define their own version of this macro in <asm/pgtable.h>
141  */
142 #if BITS_PER_LONG == 64
143 /* This function must be updated when the size of struct page grows above 80
144  * or reduces below 56. The idea that compiler optimizes out switch()
145  * statement, and only leaves move/store instructions. Also the compiler can
146  * combine write statments if they are both assignments and can be reordered,
147  * this can result in several of the writes here being dropped.
148  */
149 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
150 static inline void __mm_zero_struct_page(struct page *page)
151 {
152         unsigned long *_pp = (void *)page;
153 
154          /* Check that struct page is either 56, 64, 72, or 80 bytes */
155         BUILD_BUG_ON(sizeof(struct page) & 7);
156         BUILD_BUG_ON(sizeof(struct page) < 56);
157         BUILD_BUG_ON(sizeof(struct page) > 80);
158 
159         switch (sizeof(struct page)) {
160         case 80:
161                 _pp[9] = 0;     /* fallthrough */
162         case 72:
163                 _pp[8] = 0;     /* fallthrough */
164         case 64:
165                 _pp[7] = 0;     /* fallthrough */
166         case 56:
167                 _pp[6] = 0;
168                 _pp[5] = 0;
169                 _pp[4] = 0;
170                 _pp[3] = 0;
171                 _pp[2] = 0;
172                 _pp[1] = 0;
173                 _pp[0] = 0;
174         }
175 }
176 #else
177 #define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
178 #endif
179 
180 /*
181  * Default maximum number of active map areas, this limits the number of vmas
182  * per mm struct. Users can overwrite this number by sysctl but there is a
183  * problem.
184  *
185  * When a program's coredump is generated as ELF format, a section is created
186  * per a vma. In ELF, the number of sections is represented in unsigned short.
187  * This means the number of sections should be smaller than 65535 at coredump.
188  * Because the kernel adds some informative sections to a image of program at
189  * generating coredump, we need some margin. The number of extra sections is
190  * 1-3 now and depends on arch. We use "5" as safe margin, here.
191  *
192  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
193  * not a hard limit any more. Although some userspace tools can be surprised by
194  * that.
195  */
196 #define MAPCOUNT_ELF_CORE_MARGIN        (5)
197 #define DEFAULT_MAX_MAP_COUNT   (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
198 
199 extern int sysctl_max_map_count;
200 
201 extern unsigned long sysctl_user_reserve_kbytes;
202 extern unsigned long sysctl_admin_reserve_kbytes;
203 
204 extern int sysctl_overcommit_memory;
205 extern int sysctl_overcommit_ratio;
206 extern unsigned long sysctl_overcommit_kbytes;
207 
208 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
209                                     size_t *, loff_t *);
210 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
211                                     size_t *, loff_t *);
212 
213 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
214 
215 /* to align the pointer to the (next) page boundary */
216 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
217 
218 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
219 #define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
220 
221 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
222 
223 /*
224  * Linux kernel virtual memory manager primitives.
225  * The idea being to have a "virtual" mm in the same way
226  * we have a virtual fs - giving a cleaner interface to the
227  * mm details, and allowing different kinds of memory mappings
228  * (from shared memory to executable loading to arbitrary
229  * mmap() functions).
230  */
231 
232 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
233 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
234 void vm_area_free(struct vm_area_struct *);
235 
236 #ifndef CONFIG_MMU
237 extern struct rb_root nommu_region_tree;
238 extern struct rw_semaphore nommu_region_sem;
239 
240 extern unsigned int kobjsize(const void *objp);
241 #endif
242 
243 /*
244  * vm_flags in vm_area_struct, see mm_types.h.
245  * When changing, update also include/trace/events/mmflags.h
246  */
247 #define VM_NONE         0x00000000
248 
249 #define VM_READ         0x00000001      /* currently active flags */
250 #define VM_WRITE        0x00000002
251 #define VM_EXEC         0x00000004
252 #define VM_SHARED       0x00000008
253 
254 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
255 #define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
256 #define VM_MAYWRITE     0x00000020
257 #define VM_MAYEXEC      0x00000040
258 #define VM_MAYSHARE     0x00000080
259 
260 #define VM_GROWSDOWN    0x00000100      /* general info on the segment */
261 #define VM_UFFD_MISSING 0x00000200      /* missing pages tracking */
262 #define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
263 #define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
264 #define VM_UFFD_WP      0x00001000      /* wrprotect pages tracking */
265 
266 #define VM_LOCKED       0x00002000
267 #define VM_IO           0x00004000      /* Memory mapped I/O or similar */
268 
269                                         /* Used by sys_madvise() */
270 #define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
271 #define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
272 
273 #define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
274 #define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
275 #define VM_LOCKONFAULT  0x00080000      /* Lock the pages covered when they are faulted in */
276 #define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
277 #define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
278 #define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
279 #define VM_SYNC         0x00800000      /* Synchronous page faults */
280 #define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
281 #define VM_WIPEONFORK   0x02000000      /* Wipe VMA contents in child. */
282 #define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
283 
284 #ifdef CONFIG_MEM_SOFT_DIRTY
285 # define VM_SOFTDIRTY   0x08000000      /* Not soft dirty clean area */
286 #else
287 # define VM_SOFTDIRTY   0
288 #endif
289 
290 #define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
291 #define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
292 #define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
293 #define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
294 
295 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
296 #define VM_HIGH_ARCH_BIT_0      32      /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_BIT_1      33      /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_2      34      /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_3      35      /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_BIT_4      36      /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_0  BIT(VM_HIGH_ARCH_BIT_0)
302 #define VM_HIGH_ARCH_1  BIT(VM_HIGH_ARCH_BIT_1)
303 #define VM_HIGH_ARCH_2  BIT(VM_HIGH_ARCH_BIT_2)
304 #define VM_HIGH_ARCH_3  BIT(VM_HIGH_ARCH_BIT_3)
305 #define VM_HIGH_ARCH_4  BIT(VM_HIGH_ARCH_BIT_4)
306 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
307 
308 #ifdef CONFIG_ARCH_HAS_PKEYS
309 # define VM_PKEY_SHIFT  VM_HIGH_ARCH_BIT_0
310 # define VM_PKEY_BIT0   VM_HIGH_ARCH_0  /* A protection key is a 4-bit value */
311 # define VM_PKEY_BIT1   VM_HIGH_ARCH_1  /* on x86 and 5-bit value on ppc64   */
312 # define VM_PKEY_BIT2   VM_HIGH_ARCH_2
313 # define VM_PKEY_BIT3   VM_HIGH_ARCH_3
314 #ifdef CONFIG_PPC
315 # define VM_PKEY_BIT4  VM_HIGH_ARCH_4
316 #else
317 # define VM_PKEY_BIT4  0
318 #endif
319 #endif /* CONFIG_ARCH_HAS_PKEYS */
320 
321 #if defined(CONFIG_X86)
322 # define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
323 #elif defined(CONFIG_PPC)
324 # define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
325 #elif defined(CONFIG_PARISC)
326 # define VM_GROWSUP     VM_ARCH_1
327 #elif defined(CONFIG_IA64)
328 # define VM_GROWSUP     VM_ARCH_1
329 #elif defined(CONFIG_SPARC64)
330 # define VM_SPARC_ADI   VM_ARCH_1       /* Uses ADI tag for access control */
331 # define VM_ARCH_CLEAR  VM_SPARC_ADI
332 #elif !defined(CONFIG_MMU)
333 # define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
334 #endif
335 
336 #if defined(CONFIG_X86_INTEL_MPX)
337 /* MPX specific bounds table or bounds directory */
338 # define VM_MPX         VM_HIGH_ARCH_4
339 #else
340 # define VM_MPX         VM_NONE
341 #endif
342 
343 #ifndef VM_GROWSUP
344 # define VM_GROWSUP     VM_NONE
345 #endif
346 
347 /* Bits set in the VMA until the stack is in its final location */
348 #define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
349 
350 #ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
351 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
352 #endif
353 
354 #ifdef CONFIG_STACK_GROWSUP
355 #define VM_STACK        VM_GROWSUP
356 #else
357 #define VM_STACK        VM_GROWSDOWN
358 #endif
359 
360 #define VM_STACK_FLAGS  (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
361 
362 /*
363  * Special vmas that are non-mergable, non-mlock()able.
364  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
365  */
366 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
367 
368 /* This mask defines which mm->def_flags a process can inherit its parent */
369 #define VM_INIT_DEF_MASK        VM_NOHUGEPAGE
370 
371 /* This mask is used to clear all the VMA flags used by mlock */
372 #define VM_LOCKED_CLEAR_MASK    (~(VM_LOCKED | VM_LOCKONFAULT))
373 
374 /* Arch-specific flags to clear when updating VM flags on protection change */
375 #ifndef VM_ARCH_CLEAR
376 # define VM_ARCH_CLEAR  VM_NONE
377 #endif
378 #define VM_FLAGS_CLEAR  (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
379 
380 /*
381  * mapping from the currently active vm_flags protection bits (the
382  * low four bits) to a page protection mask..
383  */
384 extern pgprot_t protection_map[16];
385 
386 #define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
387 #define FAULT_FLAG_MKWRITE      0x02    /* Fault was mkwrite of existing pte */
388 #define FAULT_FLAG_ALLOW_RETRY  0x04    /* Retry fault if blocking */
389 #define FAULT_FLAG_RETRY_NOWAIT 0x08    /* Don't drop mmap_sem and wait when retrying */
390 #define FAULT_FLAG_KILLABLE     0x10    /* The fault task is in SIGKILL killable region */
391 #define FAULT_FLAG_TRIED        0x20    /* Second try */
392 #define FAULT_FLAG_USER         0x40    /* The fault originated in userspace */
393 #define FAULT_FLAG_REMOTE       0x80    /* faulting for non current tsk/mm */
394 #define FAULT_FLAG_INSTRUCTION  0x100   /* The fault was during an instruction fetch */
395 
396 #define FAULT_FLAG_TRACE \
397         { FAULT_FLAG_WRITE,             "WRITE" }, \
398         { FAULT_FLAG_MKWRITE,           "MKWRITE" }, \
399         { FAULT_FLAG_ALLOW_RETRY,       "ALLOW_RETRY" }, \
400         { FAULT_FLAG_RETRY_NOWAIT,      "RETRY_NOWAIT" }, \
401         { FAULT_FLAG_KILLABLE,          "KILLABLE" }, \
402         { FAULT_FLAG_TRIED,             "TRIED" }, \
403         { FAULT_FLAG_USER,              "USER" }, \
404         { FAULT_FLAG_REMOTE,            "REMOTE" }, \
405         { FAULT_FLAG_INSTRUCTION,       "INSTRUCTION" }
406 
407 /*
408  * vm_fault is filled by the the pagefault handler and passed to the vma's
409  * ->fault function. The vma's ->fault is responsible for returning a bitmask
410  * of VM_FAULT_xxx flags that give details about how the fault was handled.
411  *
412  * MM layer fills up gfp_mask for page allocations but fault handler might
413  * alter it if its implementation requires a different allocation context.
414  *
415  * pgoff should be used in favour of virtual_address, if possible.
416  */
417 struct vm_fault {
418         struct vm_area_struct *vma;     /* Target VMA */
419         unsigned int flags;             /* FAULT_FLAG_xxx flags */
420         gfp_t gfp_mask;                 /* gfp mask to be used for allocations */
421         pgoff_t pgoff;                  /* Logical page offset based on vma */
422         unsigned long address;          /* Faulting virtual address */
423         pmd_t *pmd;                     /* Pointer to pmd entry matching
424                                          * the 'address' */
425         pud_t *pud;                     /* Pointer to pud entry matching
426                                          * the 'address'
427                                          */
428         pte_t orig_pte;                 /* Value of PTE at the time of fault */
429 
430         struct page *cow_page;          /* Page handler may use for COW fault */
431         struct mem_cgroup *memcg;       /* Cgroup cow_page belongs to */
432         struct page *page;              /* ->fault handlers should return a
433                                          * page here, unless VM_FAULT_NOPAGE
434                                          * is set (which is also implied by
435                                          * VM_FAULT_ERROR).
436                                          */
437         /* These three entries are valid only while holding ptl lock */
438         pte_t *pte;                     /* Pointer to pte entry matching
439                                          * the 'address'. NULL if the page
440                                          * table hasn't been allocated.
441                                          */
442         spinlock_t *ptl;                /* Page table lock.
443                                          * Protects pte page table if 'pte'
444                                          * is not NULL, otherwise pmd.
445                                          */
446         pgtable_t prealloc_pte;         /* Pre-allocated pte page table.
447                                          * vm_ops->map_pages() calls
448                                          * alloc_set_pte() from atomic context.
449                                          * do_fault_around() pre-allocates
450                                          * page table to avoid allocation from
451                                          * atomic context.
452                                          */
453 };
454 
455 /* page entry size for vm->huge_fault() */
456 enum page_entry_size {
457         PE_SIZE_PTE = 0,
458         PE_SIZE_PMD,
459         PE_SIZE_PUD,
460 };
461 
462 /*
463  * These are the virtual MM functions - opening of an area, closing and
464  * unmapping it (needed to keep files on disk up-to-date etc), pointer
465  * to the functions called when a no-page or a wp-page exception occurs.
466  */
467 struct vm_operations_struct {
468         void (*open)(struct vm_area_struct * area);
469         void (*close)(struct vm_area_struct * area);
470         int (*split)(struct vm_area_struct * area, unsigned long addr);
471         int (*mremap)(struct vm_area_struct * area);
472         vm_fault_t (*fault)(struct vm_fault *vmf);
473         vm_fault_t (*huge_fault)(struct vm_fault *vmf,
474                         enum page_entry_size pe_size);
475         void (*map_pages)(struct vm_fault *vmf,
476                         pgoff_t start_pgoff, pgoff_t end_pgoff);
477         unsigned long (*pagesize)(struct vm_area_struct * area);
478 
479         /* notification that a previously read-only page is about to become
480          * writable, if an error is returned it will cause a SIGBUS */
481         vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
482 
483         /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
484         vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
485 
486         /* called by access_process_vm when get_user_pages() fails, typically
487          * for use by special VMAs that can switch between memory and hardware
488          */
489         int (*access)(struct vm_area_struct *vma, unsigned long addr,
490                       void *buf, int len, int write);
491 
492         /* Called by the /proc/PID/maps code to ask the vma whether it
493          * has a special name.  Returning non-NULL will also cause this
494          * vma to be dumped unconditionally. */
495         const char *(*name)(struct vm_area_struct *vma);
496 
497 #ifdef CONFIG_NUMA
498         /*
499          * set_policy() op must add a reference to any non-NULL @new mempolicy
500          * to hold the policy upon return.  Caller should pass NULL @new to
501          * remove a policy and fall back to surrounding context--i.e. do not
502          * install a MPOL_DEFAULT policy, nor the task or system default
503          * mempolicy.
504          */
505         int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
506 
507         /*
508          * get_policy() op must add reference [mpol_get()] to any policy at
509          * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
510          * in mm/mempolicy.c will do this automatically.
511          * get_policy() must NOT add a ref if the policy at (vma,addr) is not
512          * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
513          * If no [shared/vma] mempolicy exists at the addr, get_policy() op
514          * must return NULL--i.e., do not "fallback" to task or system default
515          * policy.
516          */
517         struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
518                                         unsigned long addr);
519 #endif
520         /*
521          * Called by vm_normal_page() for special PTEs to find the
522          * page for @addr.  This is useful if the default behavior
523          * (using pte_page()) would not find the correct page.
524          */
525         struct page *(*find_special_page)(struct vm_area_struct *vma,
526                                           unsigned long addr);
527 };
528 
529 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
530 {
531         static const struct vm_operations_struct dummy_vm_ops = {};
532 
533         memset(vma, 0, sizeof(*vma));
534         vma->vm_mm = mm;
535         vma->vm_ops = &dummy_vm_ops;
536         INIT_LIST_HEAD(&vma->anon_vma_chain);
537 }
538 
539 static inline void vma_set_anonymous(struct vm_area_struct *vma)
540 {
541         vma->vm_ops = NULL;
542 }
543 
544 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
545 {
546         return !vma->vm_ops;
547 }
548 
549 #ifdef CONFIG_SHMEM
550 /*
551  * The vma_is_shmem is not inline because it is used only by slow
552  * paths in userfault.
553  */
554 bool vma_is_shmem(struct vm_area_struct *vma);
555 #else
556 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
557 #endif
558 
559 int vma_is_stack_for_current(struct vm_area_struct *vma);
560 
561 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
562 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
563 
564 struct mmu_gather;
565 struct inode;
566 
567 #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
568 static inline int pmd_devmap(pmd_t pmd)
569 {
570         return 0;
571 }
572 static inline int pud_devmap(pud_t pud)
573 {
574         return 0;
575 }
576 static inline int pgd_devmap(pgd_t pgd)
577 {
578         return 0;
579 }
580 #endif
581 
582 /*
583  * FIXME: take this include out, include page-flags.h in
584  * files which need it (119 of them)
585  */
586 #include <linux/page-flags.h>
587 #include <linux/huge_mm.h>
588 
589 /*
590  * Methods to modify the page usage count.
591  *
592  * What counts for a page usage:
593  * - cache mapping   (page->mapping)
594  * - private data    (page->private)
595  * - page mapped in a task's page tables, each mapping
596  *   is counted separately
597  *
598  * Also, many kernel routines increase the page count before a critical
599  * routine so they can be sure the page doesn't go away from under them.
600  */
601 
602 /*
603  * Drop a ref, return true if the refcount fell to zero (the page has no users)
604  */
605 static inline int put_page_testzero(struct page *page)
606 {
607         VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
608         return page_ref_dec_and_test(page);
609 }
610 
611 /*
612  * Try to grab a ref unless the page has a refcount of zero, return false if
613  * that is the case.
614  * This can be called when MMU is off so it must not access
615  * any of the virtual mappings.
616  */
617 static inline int get_page_unless_zero(struct page *page)
618 {
619         return page_ref_add_unless(page, 1, 0);
620 }
621 
622 extern int page_is_ram(unsigned long pfn);
623 
624 enum {
625         REGION_INTERSECTS,
626         REGION_DISJOINT,
627         REGION_MIXED,
628 };
629 
630 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
631                       unsigned long desc);
632 
633 /* Support for virtually mapped pages */
634 struct page *vmalloc_to_page(const void *addr);
635 unsigned long vmalloc_to_pfn(const void *addr);
636 
637 /*
638  * Determine if an address is within the vmalloc range
639  *
640  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
641  * is no special casing required.
642  */
643 static inline bool is_vmalloc_addr(const void *x)
644 {
645 #ifdef CONFIG_MMU
646         unsigned long addr = (unsigned long)x;
647 
648         return addr >= VMALLOC_START && addr < VMALLOC_END;
649 #else
650         return false;
651 #endif
652 }
653 
654 #ifndef is_ioremap_addr
655 #define is_ioremap_addr(x) is_vmalloc_addr(x)
656 #endif
657 
658 #ifdef CONFIG_MMU
659 extern int is_vmalloc_or_module_addr(const void *x);
660 #else
661 static inline int is_vmalloc_or_module_addr(const void *x)
662 {
663         return 0;
664 }
665 #endif
666 
667 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
668 static inline void *kvmalloc(size_t size, gfp_t flags)
669 {
670         return kvmalloc_node(size, flags, NUMA_NO_NODE);
671 }
672 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
673 {
674         return kvmalloc_node(size, flags | __GFP_ZERO, node);
675 }
676 static inline void *kvzalloc(size_t size, gfp_t flags)
677 {
678         return kvmalloc(size, flags | __GFP_ZERO);
679 }
680 
681 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
682 {
683         size_t bytes;
684 
685         if (unlikely(check_mul_overflow(n, size, &bytes)))
686                 return NULL;
687 
688         return kvmalloc(bytes, flags);
689 }
690 
691 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
692 {
693         return kvmalloc_array(n, size, flags | __GFP_ZERO);
694 }
695 
696 extern void kvfree(const void *addr);
697 
698 static inline int compound_mapcount(struct page *page)
699 {
700         VM_BUG_ON_PAGE(!PageCompound(page), page);
701         page = compound_head(page);
702         return atomic_read(compound_mapcount_ptr(page)) + 1;
703 }
704 
705 /*
706  * The atomic page->_mapcount, starts from -1: so that transitions
707  * both from it and to it can be tracked, using atomic_inc_and_test
708  * and atomic_add_negative(-1).
709  */
710 static inline void page_mapcount_reset(struct page *page)
711 {
712         atomic_set(&(page)->_mapcount, -1);
713 }
714 
715 int __page_mapcount(struct page *page);
716 
717 static inline int page_mapcount(struct page *page)
718 {
719         VM_BUG_ON_PAGE(PageSlab(page), page);
720 
721         if (unlikely(PageCompound(page)))
722                 return __page_mapcount(page);
723         return atomic_read(&page->_mapcount) + 1;
724 }
725 
726 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
727 int total_mapcount(struct page *page);
728 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
729 #else
730 static inline int total_mapcount(struct page *page)
731 {
732         return page_mapcount(page);
733 }
734 static inline int page_trans_huge_mapcount(struct page *page,
735                                            int *total_mapcount)
736 {
737         int mapcount = page_mapcount(page);
738         if (total_mapcount)
739                 *total_mapcount = mapcount;
740         return mapcount;
741 }
742 #endif
743 
744 static inline struct page *virt_to_head_page(const void *x)
745 {
746         struct page *page = virt_to_page(x);
747 
748         return compound_head(page);
749 }
750 
751 void __put_page(struct page *page);
752 
753 void put_pages_list(struct list_head *pages);
754 
755 void split_page(struct page *page, unsigned int order);
756 
757 /*
758  * Compound pages have a destructor function.  Provide a
759  * prototype for that function and accessor functions.
760  * These are _only_ valid on the head of a compound page.
761  */
762 typedef void compound_page_dtor(struct page *);
763 
764 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
765 enum compound_dtor_id {
766         NULL_COMPOUND_DTOR,
767         COMPOUND_PAGE_DTOR,
768 #ifdef CONFIG_HUGETLB_PAGE
769         HUGETLB_PAGE_DTOR,
770 #endif
771 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
772         TRANSHUGE_PAGE_DTOR,
773 #endif
774         NR_COMPOUND_DTORS,
775 };
776 extern compound_page_dtor * const compound_page_dtors[];
777 
778 static inline void set_compound_page_dtor(struct page *page,
779                 enum compound_dtor_id compound_dtor)
780 {
781         VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
782         page[1].compound_dtor = compound_dtor;
783 }
784 
785 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
786 {
787         VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
788         return compound_page_dtors[page[1].compound_dtor];
789 }
790 
791 static inline unsigned int compound_order(struct page *page)
792 {
793         if (!PageHead(page))
794                 return 0;
795         return page[1].compound_order;
796 }
797 
798 static inline void set_compound_order(struct page *page, unsigned int order)
799 {
800         page[1].compound_order = order;
801 }
802 
803 void free_compound_page(struct page *page);
804 
805 #ifdef CONFIG_MMU
806 /*
807  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
808  * servicing faults for write access.  In the normal case, do always want
809  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
810  * that do not have writing enabled, when used by access_process_vm.
811  */
812 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
813 {
814         if (likely(vma->vm_flags & VM_WRITE))
815                 pte = pte_mkwrite(pte);
816         return pte;
817 }
818 
819 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
820                 struct page *page);
821 vm_fault_t finish_fault(struct vm_fault *vmf);
822 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
823 #endif
824 
825 /*
826  * Multiple processes may "see" the same page. E.g. for untouched
827  * mappings of /dev/null, all processes see the same page full of
828  * zeroes, and text pages of executables and shared libraries have
829  * only one copy in memory, at most, normally.
830  *
831  * For the non-reserved pages, page_count(page) denotes a reference count.
832  *   page_count() == 0 means the page is free. page->lru is then used for
833  *   freelist management in the buddy allocator.
834  *   page_count() > 0  means the page has been allocated.
835  *
836  * Pages are allocated by the slab allocator in order to provide memory
837  * to kmalloc and kmem_cache_alloc. In this case, the management of the
838  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
839  * unless a particular usage is carefully commented. (the responsibility of
840  * freeing the kmalloc memory is the caller's, of course).
841  *
842  * A page may be used by anyone else who does a __get_free_page().
843  * In this case, page_count still tracks the references, and should only
844  * be used through the normal accessor functions. The top bits of page->flags
845  * and page->virtual store page management information, but all other fields
846  * are unused and could be used privately, carefully. The management of this
847  * page is the responsibility of the one who allocated it, and those who have
848  * subsequently been given references to it.
849  *
850  * The other pages (we may call them "pagecache pages") are completely
851  * managed by the Linux memory manager: I/O, buffers, swapping etc.
852  * The following discussion applies only to them.
853  *
854  * A pagecache page contains an opaque `private' member, which belongs to the
855  * page's address_space. Usually, this is the address of a circular list of
856  * the page's disk buffers. PG_private must be set to tell the VM to call
857  * into the filesystem to release these pages.
858  *
859  * A page may belong to an inode's memory mapping. In this case, page->mapping
860  * is the pointer to the inode, and page->index is the file offset of the page,
861  * in units of PAGE_SIZE.
862  *
863  * If pagecache pages are not associated with an inode, they are said to be
864  * anonymous pages. These may become associated with the swapcache, and in that
865  * case PG_swapcache is set, and page->private is an offset into the swapcache.
866  *
867  * In either case (swapcache or inode backed), the pagecache itself holds one
868  * reference to the page. Setting PG_private should also increment the
869  * refcount. The each user mapping also has a reference to the page.
870  *
871  * The pagecache pages are stored in a per-mapping radix tree, which is
872  * rooted at mapping->i_pages, and indexed by offset.
873  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
874  * lists, we instead now tag pages as dirty/writeback in the radix tree.
875  *
876  * All pagecache pages may be subject to I/O:
877  * - inode pages may need to be read from disk,
878  * - inode pages which have been modified and are MAP_SHARED may need
879  *   to be written back to the inode on disk,
880  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
881  *   modified may need to be swapped out to swap space and (later) to be read
882  *   back into memory.
883  */
884 
885 /*
886  * The zone field is never updated after free_area_init_core()
887  * sets it, so none of the operations on it need to be atomic.
888  */
889 
890 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
891 #define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
892 #define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
893 #define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
894 #define LAST_CPUPID_PGOFF       (ZONES_PGOFF - LAST_CPUPID_WIDTH)
895 #define KASAN_TAG_PGOFF         (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
896 
897 /*
898  * Define the bit shifts to access each section.  For non-existent
899  * sections we define the shift as 0; that plus a 0 mask ensures
900  * the compiler will optimise away reference to them.
901  */
902 #define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
903 #define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
904 #define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
905 #define LAST_CPUPID_PGSHIFT     (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
906 #define KASAN_TAG_PGSHIFT       (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
907 
908 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
909 #ifdef NODE_NOT_IN_PAGE_FLAGS
910 #define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
911 #define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
912                                                 SECTIONS_PGOFF : ZONES_PGOFF)
913 #else
914 #define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
915 #define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
916                                                 NODES_PGOFF : ZONES_PGOFF)
917 #endif
918 
919 #define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
920 
921 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
922 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
923 #endif
924 
925 #define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
926 #define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
927 #define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
928 #define LAST_CPUPID_MASK        ((1UL << LAST_CPUPID_SHIFT) - 1)
929 #define KASAN_TAG_MASK          ((1UL << KASAN_TAG_WIDTH) - 1)
930 #define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
931 
932 static inline enum zone_type page_zonenum(const struct page *page)
933 {
934         return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
935 }
936 
937 #ifdef CONFIG_ZONE_DEVICE
938 static inline bool is_zone_device_page(const struct page *page)
939 {
940         return page_zonenum(page) == ZONE_DEVICE;
941 }
942 extern void memmap_init_zone_device(struct zone *, unsigned long,
943                                     unsigned long, struct dev_pagemap *);
944 #else
945 static inline bool is_zone_device_page(const struct page *page)
946 {
947         return false;
948 }
949 #endif
950 
951 #ifdef CONFIG_DEV_PAGEMAP_OPS
952 void __put_devmap_managed_page(struct page *page);
953 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
954 static inline bool put_devmap_managed_page(struct page *page)
955 {
956         if (!static_branch_unlikely(&devmap_managed_key))
957                 return false;
958         if (!is_zone_device_page(page))
959                 return false;
960         switch (page->pgmap->type) {
961         case MEMORY_DEVICE_PRIVATE:
962         case MEMORY_DEVICE_FS_DAX:
963                 __put_devmap_managed_page(page);
964                 return true;
965         default:
966                 break;
967         }
968         return false;
969 }
970 
971 #else /* CONFIG_DEV_PAGEMAP_OPS */
972 static inline bool put_devmap_managed_page(struct page *page)
973 {
974         return false;
975 }
976 #endif /* CONFIG_DEV_PAGEMAP_OPS */
977 
978 static inline bool is_device_private_page(const struct page *page)
979 {
980         return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
981                 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
982                 is_zone_device_page(page) &&
983                 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
984 }
985 
986 static inline bool is_pci_p2pdma_page(const struct page *page)
987 {
988         return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
989                 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
990                 is_zone_device_page(page) &&
991                 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
992 }
993 
994 /* 127: arbitrary random number, small enough to assemble well */
995 #define page_ref_zero_or_close_to_overflow(page) \
996         ((unsigned int) page_ref_count(page) + 127u <= 127u)
997 
998 static inline void get_page(struct page *page)
999 {
1000         page = compound_head(page);
1001         /*
1002          * Getting a normal page or the head of a compound page
1003          * requires to already have an elevated page->_refcount.
1004          */
1005         VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1006         page_ref_inc(page);
1007 }
1008 
1009 static inline __must_check bool try_get_page(struct page *page)
1010 {
1011         page = compound_head(page);
1012         if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1013                 return false;
1014         page_ref_inc(page);
1015         return true;
1016 }
1017 
1018 static inline void put_page(struct page *page)
1019 {
1020         page = compound_head(page);
1021 
1022         /*
1023          * For devmap managed pages we need to catch refcount transition from
1024          * 2 to 1, when refcount reach one it means the page is free and we
1025          * need to inform the device driver through callback. See
1026          * include/linux/memremap.h and HMM for details.
1027          */
1028         if (put_devmap_managed_page(page))
1029                 return;
1030 
1031         if (put_page_testzero(page))
1032                 __put_page(page);
1033 }
1034 
1035 /**
1036  * put_user_page() - release a gup-pinned page
1037  * @page:            pointer to page to be released
1038  *
1039  * Pages that were pinned via get_user_pages*() must be released via
1040  * either put_user_page(), or one of the put_user_pages*() routines
1041  * below. This is so that eventually, pages that are pinned via
1042  * get_user_pages*() can be separately tracked and uniquely handled. In
1043  * particular, interactions with RDMA and filesystems need special
1044  * handling.
1045  *
1046  * put_user_page() and put_page() are not interchangeable, despite this early
1047  * implementation that makes them look the same. put_user_page() calls must
1048  * be perfectly matched up with get_user_page() calls.
1049  */
1050 static inline void put_user_page(struct page *page)
1051 {
1052         put_page(page);
1053 }
1054 
1055 void put_user_pages_dirty(struct page **pages, unsigned long npages);
1056 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages);
1057 void put_user_pages(struct page **pages, unsigned long npages);
1058 
1059 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1060 #define SECTION_IN_PAGE_FLAGS
1061 #endif
1062 
1063 /*
1064  * The identification function is mainly used by the buddy allocator for
1065  * determining if two pages could be buddies. We are not really identifying
1066  * the zone since we could be using the section number id if we do not have
1067  * node id available in page flags.
1068  * We only guarantee that it will return the same value for two combinable
1069  * pages in a zone.
1070  */
1071 static inline int page_zone_id(struct page *page)
1072 {
1073         return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1074 }
1075 
1076 #ifdef NODE_NOT_IN_PAGE_FLAGS
1077 extern int page_to_nid(const struct page *page);
1078 #else
1079 static inline int page_to_nid(const struct page *page)
1080 {
1081         struct page *p = (struct page *)page;
1082 
1083         return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1084 }
1085 #endif
1086 
1087 #ifdef CONFIG_NUMA_BALANCING
1088 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1089 {
1090         return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1091 }
1092 
1093 static inline int cpupid_to_pid(int cpupid)
1094 {
1095         return cpupid & LAST__PID_MASK;
1096 }
1097 
1098 static inline int cpupid_to_cpu(int cpupid)
1099 {
1100         return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1101 }
1102 
1103 static inline int cpupid_to_nid(int cpupid)
1104 {
1105         return cpu_to_node(cpupid_to_cpu(cpupid));
1106 }
1107 
1108 static inline bool cpupid_pid_unset(int cpupid)
1109 {
1110         return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1111 }
1112 
1113 static inline bool cpupid_cpu_unset(int cpupid)
1114 {
1115         return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1116 }
1117 
1118 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1119 {
1120         return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1121 }
1122 
1123 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1124 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1125 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1126 {
1127         return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1128 }
1129 
1130 static inline int page_cpupid_last(struct page *page)
1131 {
1132         return page->_last_cpupid;
1133 }
1134 static inline void page_cpupid_reset_last(struct page *page)
1135 {
1136         page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1137 }
1138 #else
1139 static inline int page_cpupid_last(struct page *page)
1140 {
1141         return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1142 }
1143 
1144 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1145 
1146 static inline void page_cpupid_reset_last(struct page *page)
1147 {
1148         page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1149 }
1150 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1151 #else /* !CONFIG_NUMA_BALANCING */
1152 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1153 {
1154         return page_to_nid(page); /* XXX */
1155 }
1156 
1157 static inline int page_cpupid_last(struct page *page)
1158 {
1159         return page_to_nid(page); /* XXX */
1160 }
1161 
1162 static inline int cpupid_to_nid(int cpupid)
1163 {
1164         return -1;
1165 }
1166 
1167 static inline int cpupid_to_pid(int cpupid)
1168 {
1169         return -1;
1170 }
1171 
1172 static inline int cpupid_to_cpu(int cpupid)
1173 {
1174         return -1;
1175 }
1176 
1177 static inline int cpu_pid_to_cpupid(int nid, int pid)
1178 {
1179         return -1;
1180 }
1181 
1182 static inline bool cpupid_pid_unset(int cpupid)
1183 {
1184         return 1;
1185 }
1186 
1187 static inline void page_cpupid_reset_last(struct page *page)
1188 {
1189 }
1190 
1191 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1192 {
1193         return false;
1194 }
1195 #endif /* CONFIG_NUMA_BALANCING */
1196 
1197 #ifdef CONFIG_KASAN_SW_TAGS
1198 static inline u8 page_kasan_tag(const struct page *page)
1199 {
1200         return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1201 }
1202 
1203 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1204 {
1205         page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1206         page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1207 }
1208 
1209 static inline void page_kasan_tag_reset(struct page *page)
1210 {
1211         page_kasan_tag_set(page, 0xff);
1212 }
1213 #else
1214 static inline u8 page_kasan_tag(const struct page *page)
1215 {
1216         return 0xff;
1217 }
1218 
1219 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1220 static inline void page_kasan_tag_reset(struct page *page) { }
1221 #endif
1222 
1223 static inline struct zone *page_zone(const struct page *page)
1224 {
1225         return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1226 }
1227 
1228 static inline pg_data_t *page_pgdat(const struct page *page)
1229 {
1230         return NODE_DATA(page_to_nid(page));
1231 }
1232 
1233 #ifdef SECTION_IN_PAGE_FLAGS
1234 static inline void set_page_section(struct page *page, unsigned long section)
1235 {
1236         page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1237         page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1238 }
1239 
1240 static inline unsigned long page_to_section(const struct page *page)
1241 {
1242         return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1243 }
1244 #endif
1245 
1246 static inline void set_page_zone(struct page *page, enum zone_type zone)
1247 {
1248         page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1249         page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1250 }
1251 
1252 static inline void set_page_node(struct page *page, unsigned long node)
1253 {
1254         page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1255         page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1256 }
1257 
1258 static inline void set_page_links(struct page *page, enum zone_type zone,
1259         unsigned long node, unsigned long pfn)
1260 {
1261         set_page_zone(page, zone);
1262         set_page_node(page, node);
1263 #ifdef SECTION_IN_PAGE_FLAGS
1264         set_page_section(page, pfn_to_section_nr(pfn));
1265 #endif
1266 }
1267 
1268 #ifdef CONFIG_MEMCG
1269 static inline struct mem_cgroup *page_memcg(struct page *page)
1270 {
1271         return page->mem_cgroup;
1272 }
1273 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1274 {
1275         WARN_ON_ONCE(!rcu_read_lock_held());
1276         return READ_ONCE(page->mem_cgroup);
1277 }
1278 #else
1279 static inline struct mem_cgroup *page_memcg(struct page *page)
1280 {
1281         return NULL;
1282 }
1283 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1284 {
1285         WARN_ON_ONCE(!rcu_read_lock_held());
1286         return NULL;
1287 }
1288 #endif
1289 
1290 /*
1291  * Some inline functions in vmstat.h depend on page_zone()
1292  */
1293 #include <linux/vmstat.h>
1294 
1295 static __always_inline void *lowmem_page_address(const struct page *page)
1296 {
1297         return page_to_virt(page);
1298 }
1299 
1300 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1301 #define HASHED_PAGE_VIRTUAL
1302 #endif
1303 
1304 #if defined(WANT_PAGE_VIRTUAL)
1305 static inline void *page_address(const struct page *page)
1306 {
1307         return page->virtual;
1308 }
1309 static inline void set_page_address(struct page *page, void *address)
1310 {
1311         page->virtual = address;
1312 }
1313 #define page_address_init()  do { } while(0)
1314 #endif
1315 
1316 #if defined(HASHED_PAGE_VIRTUAL)
1317 void *page_address(const struct page *page);
1318 void set_page_address(struct page *page, void *virtual);
1319 void page_address_init(void);
1320 #endif
1321 
1322 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1323 #define page_address(page) lowmem_page_address(page)
1324 #define set_page_address(page, address)  do { } while(0)
1325 #define page_address_init()  do { } while(0)
1326 #endif
1327 
1328 extern void *page_rmapping(struct page *page);
1329 extern struct anon_vma *page_anon_vma(struct page *page);
1330 extern struct address_space *page_mapping(struct page *page);
1331 
1332 extern struct address_space *__page_file_mapping(struct page *);
1333 
1334 static inline
1335 struct address_space *page_file_mapping(struct page *page)
1336 {
1337         if (unlikely(PageSwapCache(page)))
1338                 return __page_file_mapping(page);
1339 
1340         return page->mapping;
1341 }
1342 
1343 extern pgoff_t __page_file_index(struct page *page);
1344 
1345 /*
1346  * Return the pagecache index of the passed page.  Regular pagecache pages
1347  * use ->index whereas swapcache pages use swp_offset(->private)
1348  */
1349 static inline pgoff_t page_index(struct page *page)
1350 {
1351         if (unlikely(PageSwapCache(page)))
1352                 return __page_file_index(page);
1353         return page->index;
1354 }
1355 
1356 bool page_mapped(struct page *page);
1357 struct address_space *page_mapping(struct page *page);
1358 struct address_space *page_mapping_file(struct page *page);
1359 
1360 /*
1361  * Return true only if the page has been allocated with
1362  * ALLOC_NO_WATERMARKS and the low watermark was not
1363  * met implying that the system is under some pressure.
1364  */
1365 static inline bool page_is_pfmemalloc(struct page *page)
1366 {
1367         /*
1368          * Page index cannot be this large so this must be
1369          * a pfmemalloc page.
1370          */
1371         return page->index == -1UL;
1372 }
1373 
1374 /*
1375  * Only to be called by the page allocator on a freshly allocated
1376  * page.
1377  */
1378 static inline void set_page_pfmemalloc(struct page *page)
1379 {
1380         page->index = -1UL;
1381 }
1382 
1383 static inline void clear_page_pfmemalloc(struct page *page)
1384 {
1385         page->index = 0;
1386 }
1387 
1388 /*
1389  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1390  */
1391 extern void pagefault_out_of_memory(void);
1392 
1393 #define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
1394 
1395 /*
1396  * Flags passed to show_mem() and show_free_areas() to suppress output in
1397  * various contexts.
1398  */
1399 #define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
1400 
1401 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1402 
1403 #ifdef CONFIG_MMU
1404 extern bool can_do_mlock(void);
1405 #else
1406 static inline bool can_do_mlock(void) { return false; }
1407 #endif
1408 extern int user_shm_lock(size_t, struct user_struct *);
1409 extern void user_shm_unlock(size_t, struct user_struct *);
1410 
1411 /*
1412  * Parameter block passed down to zap_pte_range in exceptional cases.
1413  */
1414 struct zap_details {
1415         struct address_space *check_mapping;    /* Check page->mapping if set */
1416         pgoff_t first_index;                    /* Lowest page->index to unmap */
1417         pgoff_t last_index;                     /* Highest page->index to unmap */
1418 };
1419 
1420 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1421                              pte_t pte);
1422 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1423                                 pmd_t pmd);
1424 
1425 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1426                   unsigned long size);
1427 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1428                     unsigned long size);
1429 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1430                 unsigned long start, unsigned long end);
1431 
1432 /**
1433  * mm_walk - callbacks for walk_page_range
1434  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1435  *             this handler should only handle pud_trans_huge() puds.
1436  *             the pmd_entry or pte_entry callbacks will be used for
1437  *             regular PUDs.
1438  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1439  *             this handler is required to be able to handle
1440  *             pmd_trans_huge() pmds.  They may simply choose to
1441  *             split_huge_page() instead of handling it explicitly.
1442  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1443  * @pte_hole: if set, called for each hole at all levels
1444  * @hugetlb_entry: if set, called for each hugetlb entry
1445  * @test_walk: caller specific callback function to determine whether
1446  *             we walk over the current vma or not. Returning 0
1447  *             value means "do page table walk over the current vma,"
1448  *             and a negative one means "abort current page table walk
1449  *             right now." 1 means "skip the current vma."
1450  * @mm:        mm_struct representing the target process of page table walk
1451  * @vma:       vma currently walked (NULL if walking outside vmas)
1452  * @private:   private data for callbacks' usage
1453  *
1454  * (see the comment on walk_page_range() for more details)
1455  */
1456 struct mm_walk {
1457         int (*pud_entry)(pud_t *pud, unsigned long addr,
1458                          unsigned long next, struct mm_walk *walk);
1459         int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1460                          unsigned long next, struct mm_walk *walk);
1461         int (*pte_entry)(pte_t *pte, unsigned long addr,
1462                          unsigned long next, struct mm_walk *walk);
1463         int (*pte_hole)(unsigned long addr, unsigned long next,
1464                         struct mm_walk *walk);
1465         int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1466                              unsigned long addr, unsigned long next,
1467                              struct mm_walk *walk);
1468         int (*test_walk)(unsigned long addr, unsigned long next,
1469                         struct mm_walk *walk);
1470         struct mm_struct *mm;
1471         struct vm_area_struct *vma;
1472         void *private;
1473 };
1474 
1475 struct mmu_notifier_range;
1476 
1477 int walk_page_range(unsigned long addr, unsigned long end,
1478                 struct mm_walk *walk);
1479 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1480 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1481                 unsigned long end, unsigned long floor, unsigned long ceiling);
1482 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1483                         struct vm_area_struct *vma);
1484 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1485                    struct mmu_notifier_range *range,
1486                    pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1487 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1488         unsigned long *pfn);
1489 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1490                 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1491 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1492                         void *buf, int len, int write);
1493 
1494 extern void truncate_pagecache(struct inode *inode, loff_t new);
1495 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1496 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1497 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1498 int truncate_inode_page(struct address_space *mapping, struct page *page);
1499 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1500 int invalidate_inode_page(struct page *page);
1501 
1502 #ifdef CONFIG_MMU
1503 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1504                         unsigned long address, unsigned int flags);
1505 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1506                             unsigned long address, unsigned int fault_flags,
1507                             bool *unlocked);
1508 void unmap_mapping_pages(struct address_space *mapping,
1509                 pgoff_t start, pgoff_t nr, bool even_cows);
1510 void unmap_mapping_range(struct address_space *mapping,
1511                 loff_t const holebegin, loff_t const holelen, int even_cows);
1512 #else
1513 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1514                 unsigned long address, unsigned int flags)
1515 {
1516         /* should never happen if there's no MMU */
1517         BUG();
1518         return VM_FAULT_SIGBUS;
1519 }
1520 static inline int fixup_user_fault(struct task_struct *tsk,
1521                 struct mm_struct *mm, unsigned long address,
1522                 unsigned int fault_flags, bool *unlocked)
1523 {
1524         /* should never happen if there's no MMU */
1525         BUG();
1526         return -EFAULT;
1527 }
1528 static inline void unmap_mapping_pages(struct address_space *mapping,
1529                 pgoff_t start, pgoff_t nr, bool even_cows) { }
1530 static inline void unmap_mapping_range(struct address_space *mapping,
1531                 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1532 #endif
1533 
1534 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1535                 loff_t const holebegin, loff_t const holelen)
1536 {
1537         unmap_mapping_range(mapping, holebegin, holelen, 0);
1538 }
1539 
1540 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1541                 void *buf, int len, unsigned int gup_flags);
1542 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1543                 void *buf, int len, unsigned int gup_flags);
1544 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1545                 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1546 
1547 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1548                             unsigned long start, unsigned long nr_pages,
1549                             unsigned int gup_flags, struct page **pages,
1550                             struct vm_area_struct **vmas, int *locked);
1551 long get_user_pages(unsigned long start, unsigned long nr_pages,
1552                             unsigned int gup_flags, struct page **pages,
1553                             struct vm_area_struct **vmas);
1554 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1555                     unsigned int gup_flags, struct page **pages, int *locked);
1556 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1557                     struct page **pages, unsigned int gup_flags);
1558 
1559 int get_user_pages_fast(unsigned long start, int nr_pages,
1560                         unsigned int gup_flags, struct page **pages);
1561 
1562 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1563 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1564                         struct task_struct *task, bool bypass_rlim);
1565 
1566 /* Container for pinned pfns / pages */
1567 struct frame_vector {
1568         unsigned int nr_allocated;      /* Number of frames we have space for */
1569         unsigned int nr_frames; /* Number of frames stored in ptrs array */
1570         bool got_ref;           /* Did we pin pages by getting page ref? */
1571         bool is_pfns;           /* Does array contain pages or pfns? */
1572         void *ptrs[0];          /* Array of pinned pfns / pages. Use
1573                                  * pfns_vector_pages() or pfns_vector_pfns()
1574                                  * for access */
1575 };
1576 
1577 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1578 void frame_vector_destroy(struct frame_vector *vec);
1579 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1580                      unsigned int gup_flags, struct frame_vector *vec);
1581 void put_vaddr_frames(struct frame_vector *vec);
1582 int frame_vector_to_pages(struct frame_vector *vec);
1583 void frame_vector_to_pfns(struct frame_vector *vec);
1584 
1585 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1586 {
1587         return vec->nr_frames;
1588 }
1589 
1590 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1591 {
1592         if (vec->is_pfns) {
1593                 int err = frame_vector_to_pages(vec);
1594 
1595                 if (err)
1596                         return ERR_PTR(err);
1597         }
1598         return (struct page **)(vec->ptrs);
1599 }
1600 
1601 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1602 {
1603         if (!vec->is_pfns)
1604                 frame_vector_to_pfns(vec);
1605         return (unsigned long *)(vec->ptrs);
1606 }
1607 
1608 struct kvec;
1609 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1610                         struct page **pages);
1611 int get_kernel_page(unsigned long start, int write, struct page **pages);
1612 struct page *get_dump_page(unsigned long addr);
1613 
1614 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1615 extern void do_invalidatepage(struct page *page, unsigned int offset,
1616                               unsigned int length);
1617 
1618 void __set_page_dirty(struct page *, struct address_space *, int warn);
1619 int __set_page_dirty_nobuffers(struct page *page);
1620 int __set_page_dirty_no_writeback(struct page *page);
1621 int redirty_page_for_writepage(struct writeback_control *wbc,
1622                                 struct page *page);
1623 void account_page_dirtied(struct page *page, struct address_space *mapping);
1624 void account_page_cleaned(struct page *page, struct address_space *mapping,
1625                           struct bdi_writeback *wb);
1626 int set_page_dirty(struct page *page);
1627 int set_page_dirty_lock(struct page *page);
1628 void __cancel_dirty_page(struct page *page);
1629 static inline void cancel_dirty_page(struct page *page)
1630 {
1631         /* Avoid atomic ops, locking, etc. when not actually needed. */
1632         if (PageDirty(page))
1633                 __cancel_dirty_page(page);
1634 }
1635 int clear_page_dirty_for_io(struct page *page);
1636 
1637 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1638 
1639 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1640                 unsigned long old_addr, struct vm_area_struct *new_vma,
1641                 unsigned long new_addr, unsigned long len,
1642                 bool need_rmap_locks);
1643 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1644                               unsigned long end, pgprot_t newprot,
1645                               int dirty_accountable, int prot_numa);
1646 extern int mprotect_fixup(struct vm_area_struct *vma,
1647                           struct vm_area_struct **pprev, unsigned long start,
1648                           unsigned long end, unsigned long newflags);
1649 
1650 /*
1651  * doesn't attempt to fault and will return short.
1652  */
1653 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1654                           struct page **pages);
1655 /*
1656  * per-process(per-mm_struct) statistics.
1657  */
1658 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1659 {
1660         long val = atomic_long_read(&mm->rss_stat.count[member]);
1661 
1662 #ifdef SPLIT_RSS_COUNTING
1663         /*
1664          * counter is updated in asynchronous manner and may go to minus.
1665          * But it's never be expected number for users.
1666          */
1667         if (val < 0)
1668                 val = 0;
1669 #endif
1670         return (unsigned long)val;
1671 }
1672 
1673 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1674 {
1675         atomic_long_add(value, &mm->rss_stat.count[member]);
1676 }
1677 
1678 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1679 {
1680         atomic_long_inc(&mm->rss_stat.count[member]);
1681 }
1682 
1683 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1684 {
1685         atomic_long_dec(&mm->rss_stat.count[member]);
1686 }
1687 
1688 /* Optimized variant when page is already known not to be PageAnon */
1689 static inline int mm_counter_file(struct page *page)
1690 {
1691         if (PageSwapBacked(page))
1692                 return MM_SHMEMPAGES;
1693         return MM_FILEPAGES;
1694 }
1695 
1696 static inline int mm_counter(struct page *page)
1697 {
1698         if (PageAnon(page))
1699                 return MM_ANONPAGES;
1700         return mm_counter_file(page);
1701 }
1702 
1703 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1704 {
1705         return get_mm_counter(mm, MM_FILEPAGES) +
1706                 get_mm_counter(mm, MM_ANONPAGES) +
1707                 get_mm_counter(mm, MM_SHMEMPAGES);
1708 }
1709 
1710 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1711 {
1712         return max(mm->hiwater_rss, get_mm_rss(mm));
1713 }
1714 
1715 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1716 {
1717         return max(mm->hiwater_vm, mm->total_vm);
1718 }
1719 
1720 static inline void update_hiwater_rss(struct mm_struct *mm)
1721 {
1722         unsigned long _rss = get_mm_rss(mm);
1723 
1724         if ((mm)->hiwater_rss < _rss)
1725                 (mm)->hiwater_rss = _rss;
1726 }
1727 
1728 static inline void update_hiwater_vm(struct mm_struct *mm)
1729 {
1730         if (mm->hiwater_vm < mm->total_vm)
1731                 mm->hiwater_vm = mm->total_vm;
1732 }
1733 
1734 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1735 {
1736         mm->hiwater_rss = get_mm_rss(mm);
1737 }
1738 
1739 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1740                                          struct mm_struct *mm)
1741 {
1742         unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1743 
1744         if (*maxrss < hiwater_rss)
1745                 *maxrss = hiwater_rss;
1746 }
1747 
1748 #if defined(SPLIT_RSS_COUNTING)
1749 void sync_mm_rss(struct mm_struct *mm);
1750 #else
1751 static inline void sync_mm_rss(struct mm_struct *mm)
1752 {
1753 }
1754 #endif
1755 
1756 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1757 static inline int pte_devmap(pte_t pte)
1758 {
1759         return 0;
1760 }
1761 #endif
1762 
1763 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1764 
1765 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1766                                spinlock_t **ptl);
1767 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1768                                     spinlock_t **ptl)
1769 {
1770         pte_t *ptep;
1771         __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1772         return ptep;
1773 }
1774 
1775 #ifdef __PAGETABLE_P4D_FOLDED
1776 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1777                                                 unsigned long address)
1778 {
1779         return 0;
1780 }
1781 #else
1782 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1783 #endif
1784 
1785 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1786 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1787                                                 unsigned long address)
1788 {
1789         return 0;
1790 }
1791 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1792 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1793 
1794 #else
1795 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1796 
1797 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1798 {
1799         if (mm_pud_folded(mm))
1800                 return;
1801         atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1802 }
1803 
1804 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1805 {
1806         if (mm_pud_folded(mm))
1807                 return;
1808         atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1809 }
1810 #endif
1811 
1812 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1813 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1814                                                 unsigned long address)
1815 {
1816         return 0;
1817 }
1818 
1819 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1820 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1821 
1822 #else
1823 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1824 
1825 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1826 {
1827         if (mm_pmd_folded(mm))
1828                 return;
1829         atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1830 }
1831 
1832 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1833 {
1834         if (mm_pmd_folded(mm))
1835                 return;
1836         atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1837 }
1838 #endif
1839 
1840 #ifdef CONFIG_MMU
1841 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1842 {
1843         atomic_long_set(&mm->pgtables_bytes, 0);
1844 }
1845 
1846 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1847 {
1848         return atomic_long_read(&mm->pgtables_bytes);
1849 }
1850 
1851 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1852 {
1853         atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1854 }
1855 
1856 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1857 {
1858         atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1859 }
1860 #else
1861 
1862 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1863 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1864 {
1865         return 0;
1866 }
1867 
1868 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1869 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1870 #endif
1871 
1872 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1873 int __pte_alloc_kernel(pmd_t *pmd);
1874 
1875 /*
1876  * The following ifdef needed to get the 4level-fixup.h header to work.
1877  * Remove it when 4level-fixup.h has been removed.
1878  */
1879 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1880 
1881 #ifndef __ARCH_HAS_5LEVEL_HACK
1882 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1883                 unsigned long address)
1884 {
1885         return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1886                 NULL : p4d_offset(pgd, address);
1887 }
1888 
1889 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1890                 unsigned long address)
1891 {
1892         return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1893                 NULL : pud_offset(p4d, address);
1894 }
1895 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1896 
1897 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1898 {
1899         return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1900                 NULL: pmd_offset(pud, address);
1901 }
1902 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1903 
1904 #if USE_SPLIT_PTE_PTLOCKS
1905 #if ALLOC_SPLIT_PTLOCKS
1906 void __init ptlock_cache_init(void);
1907 extern bool ptlock_alloc(struct page *page);
1908 extern void ptlock_free(struct page *page);
1909 
1910 static inline spinlock_t *ptlock_ptr(struct page *page)
1911 {
1912         return page->ptl;
1913 }
1914 #else /* ALLOC_SPLIT_PTLOCKS */
1915 static inline void ptlock_cache_init(void)
1916 {
1917 }
1918 
1919 static inline bool ptlock_alloc(struct page *page)
1920 {
1921         return true;
1922 }
1923 
1924 static inline void ptlock_free(struct page *page)
1925 {
1926 }
1927 
1928 static inline spinlock_t *ptlock_ptr(struct page *page)
1929 {
1930         return &page->ptl;
1931 }
1932 #endif /* ALLOC_SPLIT_PTLOCKS */
1933 
1934 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1935 {
1936         return ptlock_ptr(pmd_page(*pmd));
1937 }
1938 
1939 static inline bool ptlock_init(struct page *page)
1940 {
1941         /*
1942          * prep_new_page() initialize page->private (and therefore page->ptl)
1943          * with 0. Make sure nobody took it in use in between.
1944          *
1945          * It can happen if arch try to use slab for page table allocation:
1946          * slab code uses page->slab_cache, which share storage with page->ptl.
1947          */
1948         VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1949         if (!ptlock_alloc(page))
1950                 return false;
1951         spin_lock_init(ptlock_ptr(page));
1952         return true;
1953 }
1954 
1955 #else   /* !USE_SPLIT_PTE_PTLOCKS */
1956 /*
1957  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1958  */
1959 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1960 {
1961         return &mm->page_table_lock;
1962 }
1963 static inline void ptlock_cache_init(void) {}
1964 static inline bool ptlock_init(struct page *page) { return true; }
1965 static inline void ptlock_free(struct page *page) {}
1966 #endif /* USE_SPLIT_PTE_PTLOCKS */
1967 
1968 static inline void pgtable_init(void)
1969 {
1970         ptlock_cache_init();
1971         pgtable_cache_init();
1972 }
1973 
1974 static inline bool pgtable_page_ctor(struct page *page)
1975 {
1976         if (!ptlock_init(page))
1977                 return false;
1978         __SetPageTable(page);
1979         inc_zone_page_state(page, NR_PAGETABLE);
1980         return true;
1981 }
1982 
1983 static inline void pgtable_page_dtor(struct page *page)
1984 {
1985         ptlock_free(page);
1986         __ClearPageTable(page);
1987         dec_zone_page_state(page, NR_PAGETABLE);
1988 }
1989 
1990 #define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1991 ({                                                      \
1992         spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1993         pte_t *__pte = pte_offset_map(pmd, address);    \
1994         *(ptlp) = __ptl;                                \
1995         spin_lock(__ptl);                               \
1996         __pte;                                          \
1997 })
1998 
1999 #define pte_unmap_unlock(pte, ptl)      do {            \
2000         spin_unlock(ptl);                               \
2001         pte_unmap(pte);                                 \
2002 } while (0)
2003 
2004 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2005 
2006 #define pte_alloc_map(mm, pmd, address)                 \
2007         (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2008 
2009 #define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
2010         (pte_alloc(mm, pmd) ?                   \
2011                  NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2012 
2013 #define pte_alloc_kernel(pmd, address)                  \
2014         ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2015                 NULL: pte_offset_kernel(pmd, address))
2016 
2017 #if USE_SPLIT_PMD_PTLOCKS
2018 
2019 static struct page *pmd_to_page(pmd_t *pmd)
2020 {
2021         unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2022         return virt_to_page((void *)((unsigned long) pmd & mask));
2023 }
2024 
2025 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2026 {
2027         return ptlock_ptr(pmd_to_page(pmd));
2028 }
2029 
2030 static inline bool pgtable_pmd_page_ctor(struct page *page)
2031 {
2032 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2033         page->pmd_huge_pte = NULL;
2034 #endif
2035         return ptlock_init(page);
2036 }
2037 
2038 static inline void pgtable_pmd_page_dtor(struct page *page)
2039 {
2040 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2041         VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2042 #endif
2043         ptlock_free(page);
2044 }
2045 
2046 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2047 
2048 #else
2049 
2050 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2051 {
2052         return &mm->page_table_lock;
2053 }
2054 
2055 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2056 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2057 
2058 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2059 
2060 #endif
2061 
2062 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2063 {
2064         spinlock_t *ptl = pmd_lockptr(mm, pmd);
2065         spin_lock(ptl);
2066         return ptl;
2067 }
2068 
2069 /*
2070  * No scalability reason to split PUD locks yet, but follow the same pattern
2071  * as the PMD locks to make it easier if we decide to.  The VM should not be
2072  * considered ready to switch to split PUD locks yet; there may be places
2073  * which need to be converted from page_table_lock.
2074  */
2075 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2076 {
2077         return &mm->page_table_lock;
2078 }
2079 
2080 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2081 {
2082         spinlock_t *ptl = pud_lockptr(mm, pud);
2083 
2084         spin_lock(ptl);
2085         return ptl;
2086 }
2087 
2088 extern void __init pagecache_init(void);
2089 extern void free_area_init(unsigned long * zones_size);
2090 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2091                 unsigned long zone_start_pfn, unsigned long *zholes_size);
2092 extern void free_initmem(void);
2093 
2094 /*
2095  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2096  * into the buddy system. The freed pages will be poisoned with pattern
2097  * "poison" if it's within range [0, UCHAR_MAX].
2098  * Return pages freed into the buddy system.
2099  */
2100 extern unsigned long free_reserved_area(void *start, void *end,
2101                                         int poison, const char *s);
2102 
2103 #ifdef  CONFIG_HIGHMEM
2104 /*
2105  * Free a highmem page into the buddy system, adjusting totalhigh_pages
2106  * and totalram_pages.
2107  */
2108 extern void free_highmem_page(struct page *page);
2109 #endif
2110 
2111 extern void adjust_managed_page_count(struct page *page, long count);
2112 extern void mem_init_print_info(const char *str);
2113 
2114 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2115 
2116 /* Free the reserved page into the buddy system, so it gets managed. */
2117 static inline void __free_reserved_page(struct page *page)
2118 {
2119         ClearPageReserved(page);
2120         init_page_count(page);
2121         __free_page(page);
2122 }
2123 
2124 static inline void free_reserved_page(struct page *page)
2125 {
2126         __free_reserved_page(page);
2127         adjust_managed_page_count(page, 1);
2128 }
2129 
2130 static inline void mark_page_reserved(struct page *page)
2131 {
2132         SetPageReserved(page);
2133         adjust_managed_page_count(page, -1);
2134 }
2135 
2136 /*
2137  * Default method to free all the __init memory into the buddy system.
2138  * The freed pages will be poisoned with pattern "poison" if it's within
2139  * range [0, UCHAR_MAX].
2140  * Return pages freed into the buddy system.
2141  */
2142 static inline unsigned long free_initmem_default(int poison)
2143 {
2144         extern char __init_begin[], __init_end[];
2145 
2146         return free_reserved_area(&__init_begin, &__init_end,
2147                                   poison, "unused kernel");
2148 }
2149 
2150 static inline unsigned long get_num_physpages(void)
2151 {
2152         int nid;
2153         unsigned long phys_pages = 0;
2154 
2155         for_each_online_node(nid)
2156                 phys_pages += node_present_pages(nid);
2157 
2158         return phys_pages;
2159 }
2160 
2161 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2162 /*
2163  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2164  * zones, allocate the backing mem_map and account for memory holes in a more
2165  * architecture independent manner. This is a substitute for creating the
2166  * zone_sizes[] and zholes_size[] arrays and passing them to
2167  * free_area_init_node()
2168  *
2169  * An architecture is expected to register range of page frames backed by
2170  * physical memory with memblock_add[_node]() before calling
2171  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2172  * usage, an architecture is expected to do something like
2173  *
2174  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2175  *                                                       max_highmem_pfn};
2176  * for_each_valid_physical_page_range()
2177  *      memblock_add_node(base, size, nid)
2178  * free_area_init_nodes(max_zone_pfns);
2179  *
2180  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2181  * registered physical page range.  Similarly
2182  * sparse_memory_present_with_active_regions() calls memory_present() for
2183  * each range when SPARSEMEM is enabled.
2184  *
2185  * See mm/page_alloc.c for more information on each function exposed by
2186  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2187  */
2188 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2189 unsigned long node_map_pfn_alignment(void);
2190 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2191                                                 unsigned long end_pfn);
2192 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2193                                                 unsigned long end_pfn);
2194 extern void get_pfn_range_for_nid(unsigned int nid,
2195                         unsigned long *start_pfn, unsigned long *end_pfn);
2196 extern unsigned long find_min_pfn_with_active_regions(void);
2197 extern void free_bootmem_with_active_regions(int nid,
2198                                                 unsigned long max_low_pfn);
2199 extern void sparse_memory_present_with_active_regions(int nid);
2200 
2201 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2202 
2203 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2204     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2205 static inline int __early_pfn_to_nid(unsigned long pfn,
2206                                         struct mminit_pfnnid_cache *state)
2207 {
2208         return 0;
2209 }
2210 #else
2211 /* please see mm/page_alloc.c */
2212 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2213 /* there is a per-arch backend function. */
2214 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2215                                         struct mminit_pfnnid_cache *state);
2216 #endif
2217 
2218 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2219 void zero_resv_unavail(void);
2220 #else
2221 static inline void zero_resv_unavail(void) {}
2222 #endif
2223 
2224 extern void set_dma_reserve(unsigned long new_dma_reserve);
2225 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2226                 enum memmap_context, struct vmem_altmap *);
2227 extern void setup_per_zone_wmarks(void);
2228 extern int __meminit init_per_zone_wmark_min(void);
2229 extern void mem_init(void);
2230 extern void __init mmap_init(void);
2231 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2232 extern long si_mem_available(void);
2233 extern void si_meminfo(struct sysinfo * val);
2234 extern void si_meminfo_node(struct sysinfo *val, int nid);
2235 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2236 extern unsigned long arch_reserved_kernel_pages(void);
2237 #endif
2238 
2239 extern __printf(3, 4)
2240 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2241 
2242 extern void setup_per_cpu_pageset(void);
2243 
2244 extern void zone_pcp_update(struct zone *zone);
2245 extern void zone_pcp_reset(struct zone *zone);
2246 
2247 /* page_alloc.c */
2248 extern int min_free_kbytes;
2249 extern int watermark_boost_factor;
2250 extern int watermark_scale_factor;
2251 
2252 /* nommu.c */
2253 extern atomic_long_t mmap_pages_allocated;
2254 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2255 
2256 /* interval_tree.c */
2257 void vma_interval_tree_insert(struct vm_area_struct *node,
2258                               struct rb_root_cached *root);
2259 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2260                                     struct vm_area_struct *prev,
2261                                     struct rb_root_cached *root);
2262 void vma_interval_tree_remove(struct vm_area_struct *node,
2263                               struct rb_root_cached *root);
2264 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2265                                 unsigned long start, unsigned long last);
2266 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2267                                 unsigned long start, unsigned long last);
2268 
2269 #define vma_interval_tree_foreach(vma, root, start, last)               \
2270         for (vma = vma_interval_tree_iter_first(root, start, last);     \
2271              vma; vma = vma_interval_tree_iter_next(vma, start, last))
2272 
2273 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2274                                    struct rb_root_cached *root);
2275 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2276                                    struct rb_root_cached *root);
2277 struct anon_vma_chain *
2278 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2279                                   unsigned long start, unsigned long last);
2280 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2281         struct anon_vma_chain *node, unsigned long start, unsigned long last);
2282 #ifdef CONFIG_DEBUG_VM_RB
2283 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2284 #endif
2285 
2286 #define anon_vma_interval_tree_foreach(avc, root, start, last)           \
2287         for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2288              avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2289 
2290 /* mmap.c */
2291 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2292 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2293         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2294         struct vm_area_struct *expand);
2295 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2296         unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2297 {
2298         return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2299 }
2300 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2301         struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2302         unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2303         struct mempolicy *, struct vm_userfaultfd_ctx);
2304 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2305 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2306         unsigned long addr, int new_below);
2307 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2308         unsigned long addr, int new_below);
2309 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2310 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2311         struct rb_node **, struct rb_node *);
2312 extern void unlink_file_vma(struct vm_area_struct *);
2313 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2314         unsigned long addr, unsigned long len, pgoff_t pgoff,
2315         bool *need_rmap_locks);
2316 extern void exit_mmap(struct mm_struct *);
2317 
2318 static inline int check_data_rlimit(unsigned long rlim,
2319                                     unsigned long new,
2320                                     unsigned long start,
2321                                     unsigned long end_data,
2322                                     unsigned long start_data)
2323 {
2324         if (rlim < RLIM_INFINITY) {
2325                 if (((new - start) + (end_data - start_data)) > rlim)
2326                         return -ENOSPC;
2327         }
2328 
2329         return 0;
2330 }
2331 
2332 extern int mm_take_all_locks(struct mm_struct *mm);
2333 extern void mm_drop_all_locks(struct mm_struct *mm);
2334 
2335 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2336 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2337 extern struct file *get_task_exe_file(struct task_struct *task);
2338 
2339 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2340 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2341 
2342 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2343                                    const struct vm_special_mapping *sm);
2344 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2345                                    unsigned long addr, unsigned long len,
2346                                    unsigned long flags,
2347                                    const struct vm_special_mapping *spec);
2348 /* This is an obsolete alternative to _install_special_mapping. */
2349 extern int install_special_mapping(struct mm_struct *mm,
2350                                    unsigned long addr, unsigned long len,
2351                                    unsigned long flags, struct page **pages);
2352 
2353 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2354 
2355 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2356         unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2357         struct list_head *uf);
2358 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2359         unsigned long len, unsigned long prot, unsigned long flags,
2360         vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2361         struct list_head *uf);
2362 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2363                        struct list_head *uf, bool downgrade);
2364 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2365                      struct list_head *uf);
2366 
2367 static inline unsigned long
2368 do_mmap_pgoff(struct file *file, unsigned long addr,
2369         unsigned long len, unsigned long prot, unsigned long flags,
2370         unsigned long pgoff, unsigned long *populate,
2371         struct list_head *uf)
2372 {
2373         return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2374 }
2375 
2376 #ifdef CONFIG_MMU
2377 extern int __mm_populate(unsigned long addr, unsigned long len,
2378                          int ignore_errors);
2379 static inline void mm_populate(unsigned long addr, unsigned long len)
2380 {
2381         /* Ignore errors */
2382         (void) __mm_populate(addr, len, 1);
2383 }
2384 #else
2385 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2386 #endif
2387 
2388 /* These take the mm semaphore themselves */
2389 extern int __must_check vm_brk(unsigned long, unsigned long);
2390 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2391 extern int vm_munmap(unsigned long, size_t);
2392 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2393         unsigned long, unsigned long,
2394         unsigned long, unsigned long);
2395 
2396 struct vm_unmapped_area_info {
2397 #define VM_UNMAPPED_AREA_TOPDOWN 1
2398         unsigned long flags;
2399         unsigned long length;
2400         unsigned long low_limit;
2401         unsigned long high_limit;
2402         unsigned long align_mask;
2403         unsigned long align_offset;
2404 };
2405 
2406 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2407 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2408 
2409 /*
2410  * Search for an unmapped address range.
2411  *
2412  * We are looking for a range that:
2413  * - does not intersect with any VMA;
2414  * - is contained within the [low_limit, high_limit) interval;
2415  * - is at least the desired size.
2416  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2417  */
2418 static inline unsigned long
2419 vm_unmapped_area(struct vm_unmapped_area_info *info)
2420 {
2421         if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2422                 return unmapped_area_topdown(info);
2423         else
2424                 return unmapped_area(info);
2425 }
2426 
2427 /* truncate.c */
2428 extern void truncate_inode_pages(struct address_space *, loff_t);
2429 extern void truncate_inode_pages_range(struct address_space *,
2430                                        loff_t lstart, loff_t lend);
2431 extern void truncate_inode_pages_final(struct address_space *);
2432 
2433 /* generic vm_area_ops exported for stackable file systems */
2434 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2435 extern void filemap_map_pages(struct vm_fault *vmf,
2436                 pgoff_t start_pgoff, pgoff_t end_pgoff);
2437 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2438 
2439 /* mm/page-writeback.c */
2440 int __must_check write_one_page(struct page *page);
2441 void task_dirty_inc(struct task_struct *tsk);
2442 
2443 /* readahead.c */
2444 #define VM_READAHEAD_PAGES      (SZ_128K / PAGE_SIZE)
2445 
2446 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2447                         pgoff_t offset, unsigned long nr_to_read);
2448 
2449 void page_cache_sync_readahead(struct address_space *mapping,
2450                                struct file_ra_state *ra,
2451                                struct file *filp,
2452                                pgoff_t offset,
2453                                unsigned long size);
2454 
2455 void page_cache_async_readahead(struct address_space *mapping,
2456                                 struct file_ra_state *ra,
2457                                 struct file *filp,
2458                                 struct page *pg,
2459                                 pgoff_t offset,
2460                                 unsigned long size);
2461 
2462 extern unsigned long stack_guard_gap;
2463 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2464 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2465 
2466 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2467 extern int expand_downwards(struct vm_area_struct *vma,
2468                 unsigned long address);
2469 #if VM_GROWSUP
2470 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2471 #else
2472   #define expand_upwards(vma, address) (0)
2473 #endif
2474 
2475 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2476 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2477 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2478                                              struct vm_area_struct **pprev);
2479 
2480 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2481    NULL if none.  Assume start_addr < end_addr. */
2482 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2483 {
2484         struct vm_area_struct * vma = find_vma(mm,start_addr);
2485 
2486         if (vma && end_addr <= vma->vm_start)
2487                 vma = NULL;
2488         return vma;
2489 }
2490 
2491 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2492 {
2493         unsigned long vm_start = vma->vm_start;
2494 
2495         if (vma->vm_flags & VM_GROWSDOWN) {
2496                 vm_start -= stack_guard_gap;
2497                 if (vm_start > vma->vm_start)
2498                         vm_start = 0;
2499         }
2500         return vm_start;
2501 }
2502 
2503 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2504 {
2505         unsigned long vm_end = vma->vm_end;
2506 
2507         if (vma->vm_flags & VM_GROWSUP) {
2508                 vm_end += stack_guard_gap;
2509                 if (vm_end < vma->vm_end)
2510                         vm_end = -PAGE_SIZE;
2511         }
2512         return vm_end;
2513 }
2514 
2515 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2516 {
2517         return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2518 }
2519 
2520 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2521 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2522                                 unsigned long vm_start, unsigned long vm_end)
2523 {
2524         struct vm_area_struct *vma = find_vma(mm, vm_start);
2525 
2526         if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2527                 vma = NULL;
2528 
2529         return vma;
2530 }
2531 
2532 static inline bool range_in_vma(struct vm_area_struct *vma,
2533                                 unsigned long start, unsigned long end)
2534 {
2535         return (vma && vma->vm_start <= start && end <= vma->vm_end);
2536 }
2537 
2538 #ifdef CONFIG_MMU
2539 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2540 void vma_set_page_prot(struct vm_area_struct *vma);
2541 #else
2542 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2543 {
2544         return __pgprot(0);
2545 }
2546 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2547 {
2548         vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2549 }
2550 #endif
2551 
2552 #ifdef CONFIG_NUMA_BALANCING
2553 unsigned long change_prot_numa(struct vm_area_struct *vma,
2554                         unsigned long start, unsigned long end);
2555 #endif
2556 
2557 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2558 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2559                         unsigned long pfn, unsigned long size, pgprot_t);
2560 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2561 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2562                                 unsigned long num);
2563 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2564                                 unsigned long num);
2565 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2566                         unsigned long pfn);
2567 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2568                         unsigned long pfn, pgprot_t pgprot);
2569 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2570                         pfn_t pfn);
2571 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2572                 unsigned long addr, pfn_t pfn);
2573 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2574 
2575 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2576                                 unsigned long addr, struct page *page)
2577 {
2578         int err = vm_insert_page(vma, addr, page);
2579 
2580         if (err == -ENOMEM)
2581                 return VM_FAULT_OOM;
2582         if (err < 0 && err != -EBUSY)
2583                 return VM_FAULT_SIGBUS;
2584 
2585         return VM_FAULT_NOPAGE;
2586 }
2587 
2588 static inline vm_fault_t vmf_error(int err)
2589 {
2590         if (err == -ENOMEM)
2591                 return VM_FAULT_OOM;
2592         return VM_FAULT_SIGBUS;
2593 }
2594 
2595 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2596                          unsigned int foll_flags);
2597 
2598 #define FOLL_WRITE      0x01    /* check pte is writable */
2599 #define FOLL_TOUCH      0x02    /* mark page accessed */
2600 #define FOLL_GET        0x04    /* do get_page on page */
2601 #define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
2602 #define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
2603 #define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
2604                                  * and return without waiting upon it */
2605 #define FOLL_POPULATE   0x40    /* fault in page */
2606 #define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
2607 #define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
2608 #define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
2609 #define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
2610 #define FOLL_TRIED      0x800   /* a retry, previous pass started an IO */
2611 #define FOLL_MLOCK      0x1000  /* lock present pages */
2612 #define FOLL_REMOTE     0x2000  /* we are working on non-current tsk/mm */
2613 #define FOLL_COW        0x4000  /* internal GUP flag */
2614 #define FOLL_ANON       0x8000  /* don't do file mappings */
2615 #define FOLL_LONGTERM   0x10000 /* mapping lifetime is indefinite: see below */
2616 
2617 /*
2618  * NOTE on FOLL_LONGTERM:
2619  *
2620  * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2621  * period _often_ under userspace control.  This is contrasted with
2622  * iov_iter_get_pages() where usages which are transient.
2623  *
2624  * FIXME: For pages which are part of a filesystem, mappings are subject to the
2625  * lifetime enforced by the filesystem and we need guarantees that longterm
2626  * users like RDMA and V4L2 only establish mappings which coordinate usage with
2627  * the filesystem.  Ideas for this coordination include revoking the longterm
2628  * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
2629  * added after the problem with filesystems was found FS DAX VMAs are
2630  * specifically failed.  Filesystem pages are still subject to bugs and use of
2631  * FOLL_LONGTERM should be avoided on those pages.
2632  *
2633  * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2634  * Currently only get_user_pages() and get_user_pages_fast() support this flag
2635  * and calls to get_user_pages_[un]locked are specifically not allowed.  This
2636  * is due to an incompatibility with the FS DAX check and
2637  * FAULT_FLAG_ALLOW_RETRY
2638  *
2639  * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2640  * that region.  And so CMA attempts to migrate the page before pinning when
2641  * FOLL_LONGTERM is specified.
2642  */
2643 
2644 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2645 {
2646         if (vm_fault & VM_FAULT_OOM)
2647                 return -ENOMEM;
2648         if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2649                 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2650         if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2651                 return -EFAULT;
2652         return 0;
2653 }
2654 
2655 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2656 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2657                                unsigned long size, pte_fn_t fn, void *data);
2658 
2659 
2660 #ifdef CONFIG_PAGE_POISONING
2661 extern bool page_poisoning_enabled(void);
2662 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2663 #else
2664 static inline bool page_poisoning_enabled(void) { return false; }
2665 static inline void kernel_poison_pages(struct page *page, int numpages,
2666                                         int enable) { }
2667 #endif
2668 
2669 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2670 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2671 #else
2672 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2673 #endif
2674 static inline bool want_init_on_alloc(gfp_t flags)
2675 {
2676         if (static_branch_unlikely(&init_on_alloc) &&
2677             !page_poisoning_enabled())
2678                 return true;
2679         return flags & __GFP_ZERO;
2680 }
2681 
2682 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2683 DECLARE_STATIC_KEY_TRUE(init_on_free);
2684 #else
2685 DECLARE_STATIC_KEY_FALSE(init_on_free);
2686 #endif
2687 static inline bool want_init_on_free(void)
2688 {
2689         return static_branch_unlikely(&init_on_free) &&
2690                !page_poisoning_enabled();
2691 }
2692 
2693 #ifdef CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT
2694 DECLARE_STATIC_KEY_TRUE(_debug_pagealloc_enabled);
2695 #else
2696 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2697 #endif
2698 
2699 static inline bool debug_pagealloc_enabled(void)
2700 {
2701         if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2702                 return false;
2703 
2704         return static_branch_unlikely(&_debug_pagealloc_enabled);
2705 }
2706 
2707 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2708 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2709 
2710 static inline void
2711 kernel_map_pages(struct page *page, int numpages, int enable)
2712 {
2713         __kernel_map_pages(page, numpages, enable);
2714 }
2715 #ifdef CONFIG_HIBERNATION
2716 extern bool kernel_page_present(struct page *page);
2717 #endif  /* CONFIG_HIBERNATION */
2718 #else   /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2719 static inline void
2720 kernel_map_pages(struct page *page, int numpages, int enable) {}
2721 #ifdef CONFIG_HIBERNATION
2722 static inline bool kernel_page_present(struct page *page) { return true; }
2723 #endif  /* CONFIG_HIBERNATION */
2724 #endif  /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2725 
2726 #ifdef __HAVE_ARCH_GATE_AREA
2727 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2728 extern int in_gate_area_no_mm(unsigned long addr);
2729 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2730 #else
2731 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2732 {
2733         return NULL;
2734 }
2735 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2736 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2737 {
2738         return 0;
2739 }
2740 #endif  /* __HAVE_ARCH_GATE_AREA */
2741 
2742 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2743 
2744 #ifdef CONFIG_SYSCTL
2745 extern int sysctl_drop_caches;
2746 int drop_caches_sysctl_handler(struct ctl_table *, int,
2747                                         void __user *, size_t *, loff_t *);
2748 #endif
2749 
2750 void drop_slab(void);
2751 void drop_slab_node(int nid);
2752 
2753 #ifndef CONFIG_MMU
2754 #define randomize_va_space 0
2755 #else
2756 extern int randomize_va_space;
2757 #endif
2758 
2759 const char * arch_vma_name(struct vm_area_struct *vma);
2760 #ifdef CONFIG_MMU
2761 void print_vma_addr(char *prefix, unsigned long rip);
2762 #else
2763 static inline void print_vma_addr(char *prefix, unsigned long rip)
2764 {
2765 }
2766 #endif
2767 
2768 void *sparse_buffer_alloc(unsigned long size);
2769 struct page * __populate_section_memmap(unsigned long pfn,
2770                 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2771 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2772 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2773 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2774 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2775 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2776 void *vmemmap_alloc_block(unsigned long size, int node);
2777 struct vmem_altmap;
2778 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2779 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2780 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2781 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2782                                int node);
2783 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2784                 struct vmem_altmap *altmap);
2785 void vmemmap_populate_print_last(void);
2786 #ifdef CONFIG_MEMORY_HOTPLUG
2787 void vmemmap_free(unsigned long start, unsigned long end,
2788                 struct vmem_altmap *altmap);
2789 #endif
2790 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2791                                   unsigned long nr_pages);
2792 
2793 enum mf_flags {
2794         MF_COUNT_INCREASED = 1 << 0,
2795         MF_ACTION_REQUIRED = 1 << 1,
2796         MF_MUST_KILL = 1 << 2,
2797         MF_SOFT_OFFLINE = 1 << 3,
2798 };
2799 extern int memory_failure(unsigned long pfn, int flags);
2800 extern void memory_failure_queue(unsigned long pfn, int flags);
2801 extern int unpoison_memory(unsigned long pfn);
2802 extern int get_hwpoison_page(struct page *page);
2803 #define put_hwpoison_page(page) put_page(page)
2804 extern int sysctl_memory_failure_early_kill;
2805 extern int sysctl_memory_failure_recovery;
2806 extern void shake_page(struct page *p, int access);
2807 extern atomic_long_t num_poisoned_pages __read_mostly;
2808 extern int soft_offline_page(struct page *page, int flags);
2809 
2810 
2811 /*
2812  * Error handlers for various types of pages.
2813  */
2814 enum mf_result {
2815         MF_IGNORED,     /* Error: cannot be handled */
2816         MF_FAILED,      /* Error: handling failed */
2817         MF_DELAYED,     /* Will be handled later */
2818         MF_RECOVERED,   /* Successfully recovered */
2819 };
2820 
2821 enum mf_action_page_type {
2822         MF_MSG_KERNEL,
2823         MF_MSG_KERNEL_HIGH_ORDER,
2824         MF_MSG_SLAB,
2825         MF_MSG_DIFFERENT_COMPOUND,
2826         MF_MSG_POISONED_HUGE,
2827         MF_MSG_HUGE,
2828         MF_MSG_FREE_HUGE,
2829         MF_MSG_NON_PMD_HUGE,
2830         MF_MSG_UNMAP_FAILED,
2831         MF_MSG_DIRTY_SWAPCACHE,
2832         MF_MSG_CLEAN_SWAPCACHE,
2833         MF_MSG_DIRTY_MLOCKED_LRU,
2834         MF_MSG_CLEAN_MLOCKED_LRU,
2835         MF_MSG_DIRTY_UNEVICTABLE_LRU,
2836         MF_MSG_CLEAN_UNEVICTABLE_LRU,
2837         MF_MSG_DIRTY_LRU,
2838         MF_MSG_CLEAN_LRU,
2839         MF_MSG_TRUNCATED_LRU,
2840         MF_MSG_BUDDY,
2841         MF_MSG_BUDDY_2ND,
2842         MF_MSG_DAX,
2843         MF_MSG_UNKNOWN,
2844 };
2845 
2846 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2847 extern void clear_huge_page(struct page *page,
2848                             unsigned long addr_hint,
2849                             unsigned int pages_per_huge_page);
2850 extern void copy_user_huge_page(struct page *dst, struct page *src,
2851                                 unsigned long addr_hint,
2852                                 struct vm_area_struct *vma,
2853                                 unsigned int pages_per_huge_page);
2854 extern long copy_huge_page_from_user(struct page *dst_page,
2855                                 const void __user *usr_src,
2856                                 unsigned int pages_per_huge_page,
2857                                 bool allow_pagefault);
2858 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2859 
2860 #ifdef CONFIG_DEBUG_PAGEALLOC
2861 extern unsigned int _debug_guardpage_minorder;
2862 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2863 
2864 static inline unsigned int debug_guardpage_minorder(void)
2865 {
2866         return _debug_guardpage_minorder;
2867 }
2868 
2869 static inline bool debug_guardpage_enabled(void)
2870 {
2871         return static_branch_unlikely(&_debug_guardpage_enabled);
2872 }
2873 
2874 static inline bool page_is_guard(struct page *page)
2875 {
2876         if (!debug_guardpage_enabled())
2877                 return false;
2878 
2879         return PageGuard(page);
2880 }
2881 #else
2882 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2883 static inline bool debug_guardpage_enabled(void) { return false; }
2884 static inline bool page_is_guard(struct page *page) { return false; }
2885 #endif /* CONFIG_DEBUG_PAGEALLOC */
2886 
2887 #if MAX_NUMNODES > 1
2888 void __init setup_nr_node_ids(void);
2889 #else
2890 static inline void setup_nr_node_ids(void) {}
2891 #endif
2892 
2893 #endif /* __KERNEL__ */
2894 #endif /* _LINUX_MM_H */
2895 

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

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

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

osdn.jp