1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
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>
33 struct anon_vma_chain;
36 struct writeback_control;
39 void init_mm_internals(void);
41 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
42 extern unsigned long max_mapnr;
44 static inline void set_max_mapnr(unsigned long limit)
49 static inline void set_max_mapnr(unsigned long limit) { }
52 extern atomic_long_t _totalram_pages;
53 static inline unsigned long totalram_pages(void)
55 return (unsigned long)atomic_long_read(&_totalram_pages);
58 static inline void totalram_pages_inc(void)
60 atomic_long_inc(&_totalram_pages);
63 static inline void totalram_pages_dec(void)
65 atomic_long_dec(&_totalram_pages);
68 static inline void totalram_pages_add(long count)
70 atomic_long_add(count, &_totalram_pages);
73 static inline void totalram_pages_set(long val)
75 atomic_long_set(&_totalram_pages, val);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
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;
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;
99 #include <asm/pgtable.h>
100 #include <asm/processor.h>
103 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
107 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
111 #define lm_alias(x) __va(__pa_symbol(x))
115 * To prevent common memory management code establishing
116 * a zero page mapping on a read fault.
117 * This macro should be defined within <asm/pgtable.h>.
118 * s390 does this to prevent multiplexing of hardware bits
119 * related to the physical page in case of virtualization.
121 #ifndef mm_forbids_zeropage
122 #define mm_forbids_zeropage(X) (0)
126 * On some architectures it is expensive to call memset() for small sizes.
127 * Those architectures should provide their own implementation of "struct page"
128 * zeroing by defining this macro in <asm/pgtable.h>.
130 #ifndef mm_zero_struct_page
131 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
135 * Default maximum number of active map areas, this limits the number of vmas
136 * per mm struct. Users can overwrite this number by sysctl but there is a
139 * When a program's coredump is generated as ELF format, a section is created
140 * per a vma. In ELF, the number of sections is represented in unsigned short.
141 * This means the number of sections should be smaller than 65535 at coredump.
142 * Because the kernel adds some informative sections to a image of program at
143 * generating coredump, we need some margin. The number of extra sections is
144 * 1-3 now and depends on arch. We use "5" as safe margin, here.
146 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
147 * not a hard limit any more. Although some userspace tools can be surprised by
150 #define MAPCOUNT_ELF_CORE_MARGIN (5)
151 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
153 extern int sysctl_max_map_count;
155 extern unsigned long sysctl_user_reserve_kbytes;
156 extern unsigned long sysctl_admin_reserve_kbytes;
158 extern int sysctl_overcommit_memory;
159 extern int sysctl_overcommit_ratio;
160 extern unsigned long sysctl_overcommit_kbytes;
162 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
164 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
167 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
169 /* to align the pointer to the (next) page boundary */
170 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
172 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
173 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
175 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
178 * Linux kernel virtual memory manager primitives.
179 * The idea being to have a "virtual" mm in the same way
180 * we have a virtual fs - giving a cleaner interface to the
181 * mm details, and allowing different kinds of memory mappings
182 * (from shared memory to executable loading to arbitrary
186 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
187 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
188 void vm_area_free(struct vm_area_struct *);
191 extern struct rb_root nommu_region_tree;
192 extern struct rw_semaphore nommu_region_sem;
194 extern unsigned int kobjsize(const void *objp);
198 * vm_flags in vm_area_struct, see mm_types.h.
199 * When changing, update also include/trace/events/mmflags.h
201 #define VM_NONE 0x00000000
203 #define VM_READ 0x00000001 /* currently active flags */
204 #define VM_WRITE 0x00000002
205 #define VM_EXEC 0x00000004
206 #define VM_SHARED 0x00000008
208 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
209 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
210 #define VM_MAYWRITE 0x00000020
211 #define VM_MAYEXEC 0x00000040
212 #define VM_MAYSHARE 0x00000080
214 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
215 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
216 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
217 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
218 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
220 #define VM_LOCKED 0x00002000
221 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
223 /* Used by sys_madvise() */
224 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
225 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
227 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
228 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
229 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
230 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
231 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
232 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
233 #define VM_SYNC 0x00800000 /* Synchronous page faults */
234 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
235 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
236 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
238 #ifdef CONFIG_MEM_SOFT_DIRTY
239 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
241 # define VM_SOFTDIRTY 0
244 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
245 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
246 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
247 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
249 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
250 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
251 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
252 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
253 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
254 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
255 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
256 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
257 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
258 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
259 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
260 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
262 #ifdef CONFIG_ARCH_HAS_PKEYS
263 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
264 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
265 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
266 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
267 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
269 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
271 # define VM_PKEY_BIT4 0
273 #endif /* CONFIG_ARCH_HAS_PKEYS */
275 #if defined(CONFIG_X86)
276 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
277 #elif defined(CONFIG_PPC)
278 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
279 #elif defined(CONFIG_PARISC)
280 # define VM_GROWSUP VM_ARCH_1
281 #elif defined(CONFIG_IA64)
282 # define VM_GROWSUP VM_ARCH_1
283 #elif defined(CONFIG_SPARC64)
284 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
285 # define VM_ARCH_CLEAR VM_SPARC_ADI
286 #elif !defined(CONFIG_MMU)
287 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
290 #if defined(CONFIG_X86_INTEL_MPX)
291 /* MPX specific bounds table or bounds directory */
292 # define VM_MPX VM_HIGH_ARCH_4
294 # define VM_MPX VM_NONE
298 # define VM_GROWSUP VM_NONE
301 /* Bits set in the VMA until the stack is in its final location */
302 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
304 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
305 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
308 #ifdef CONFIG_STACK_GROWSUP
309 #define VM_STACK VM_GROWSUP
311 #define VM_STACK VM_GROWSDOWN
314 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
317 * Special vmas that are non-mergable, non-mlock()able.
318 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
320 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
322 /* This mask defines which mm->def_flags a process can inherit its parent */
323 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
325 /* This mask is used to clear all the VMA flags used by mlock */
326 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
328 /* Arch-specific flags to clear when updating VM flags on protection change */
329 #ifndef VM_ARCH_CLEAR
330 # define VM_ARCH_CLEAR VM_NONE
332 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
335 * mapping from the currently active vm_flags protection bits (the
336 * low four bits) to a page protection mask..
338 extern pgprot_t protection_map[16];
340 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
341 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
342 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
343 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
344 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
345 #define FAULT_FLAG_TRIED 0x20 /* Second try */
346 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
347 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
348 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
350 #define FAULT_FLAG_TRACE \
351 { FAULT_FLAG_WRITE, "WRITE" }, \
352 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
353 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
354 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
355 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
356 { FAULT_FLAG_TRIED, "TRIED" }, \
357 { FAULT_FLAG_USER, "USER" }, \
358 { FAULT_FLAG_REMOTE, "REMOTE" }, \
359 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
362 * vm_fault is filled by the the pagefault handler and passed to the vma's
363 * ->fault function. The vma's ->fault is responsible for returning a bitmask
364 * of VM_FAULT_xxx flags that give details about how the fault was handled.
366 * MM layer fills up gfp_mask for page allocations but fault handler might
367 * alter it if its implementation requires a different allocation context.
369 * pgoff should be used in favour of virtual_address, if possible.
372 struct vm_area_struct *vma; /* Target VMA */
373 unsigned int flags; /* FAULT_FLAG_xxx flags */
374 gfp_t gfp_mask; /* gfp mask to be used for allocations */
375 pgoff_t pgoff; /* Logical page offset based on vma */
376 unsigned long address; /* Faulting virtual address */
377 pmd_t *pmd; /* Pointer to pmd entry matching
379 pud_t *pud; /* Pointer to pud entry matching
382 pte_t orig_pte; /* Value of PTE at the time of fault */
384 struct page *cow_page; /* Page handler may use for COW fault */
385 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
386 struct page *page; /* ->fault handlers should return a
387 * page here, unless VM_FAULT_NOPAGE
388 * is set (which is also implied by
391 /* These three entries are valid only while holding ptl lock */
392 pte_t *pte; /* Pointer to pte entry matching
393 * the 'address'. NULL if the page
394 * table hasn't been allocated.
396 spinlock_t *ptl; /* Page table lock.
397 * Protects pte page table if 'pte'
398 * is not NULL, otherwise pmd.
400 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
401 * vm_ops->map_pages() calls
402 * alloc_set_pte() from atomic context.
403 * do_fault_around() pre-allocates
404 * page table to avoid allocation from
409 /* page entry size for vm->huge_fault() */
410 enum page_entry_size {
417 * These are the virtual MM functions - opening of an area, closing and
418 * unmapping it (needed to keep files on disk up-to-date etc), pointer
419 * to the functions called when a no-page or a wp-page exception occurs.
421 struct vm_operations_struct {
422 void (*open)(struct vm_area_struct * area);
423 void (*close)(struct vm_area_struct * area);
424 int (*split)(struct vm_area_struct * area, unsigned long addr);
425 int (*mremap)(struct vm_area_struct * area);
426 vm_fault_t (*fault)(struct vm_fault *vmf);
427 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
428 enum page_entry_size pe_size);
429 void (*map_pages)(struct vm_fault *vmf,
430 pgoff_t start_pgoff, pgoff_t end_pgoff);
431 unsigned long (*pagesize)(struct vm_area_struct * area);
433 /* notification that a previously read-only page is about to become
434 * writable, if an error is returned it will cause a SIGBUS */
435 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
437 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
438 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
440 /* called by access_process_vm when get_user_pages() fails, typically
441 * for use by special VMAs that can switch between memory and hardware
443 int (*access)(struct vm_area_struct *vma, unsigned long addr,
444 void *buf, int len, int write);
446 /* Called by the /proc/PID/maps code to ask the vma whether it
447 * has a special name. Returning non-NULL will also cause this
448 * vma to be dumped unconditionally. */
449 const char *(*name)(struct vm_area_struct *vma);
453 * set_policy() op must add a reference to any non-NULL @new mempolicy
454 * to hold the policy upon return. Caller should pass NULL @new to
455 * remove a policy and fall back to surrounding context--i.e. do not
456 * install a MPOL_DEFAULT policy, nor the task or system default
459 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
462 * get_policy() op must add reference [mpol_get()] to any policy at
463 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
464 * in mm/mempolicy.c will do this automatically.
465 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
466 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
467 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
468 * must return NULL--i.e., do not "fallback" to task or system default
471 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
475 * Called by vm_normal_page() for special PTEs to find the
476 * page for @addr. This is useful if the default behavior
477 * (using pte_page()) would not find the correct page.
479 struct page *(*find_special_page)(struct vm_area_struct *vma,
483 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
485 static const struct vm_operations_struct dummy_vm_ops = {};
487 memset(vma, 0, sizeof(*vma));
489 vma->vm_ops = &dummy_vm_ops;
490 INIT_LIST_HEAD(&vma->anon_vma_chain);
493 static inline void vma_set_anonymous(struct vm_area_struct *vma)
498 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
499 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
504 #define page_private(page) ((page)->private)
505 #define set_page_private(page, v) ((page)->private = (v))
507 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
508 static inline int pmd_devmap(pmd_t pmd)
512 static inline int pud_devmap(pud_t pud)
516 static inline int pgd_devmap(pgd_t pgd)
523 * FIXME: take this include out, include page-flags.h in
524 * files which need it (119 of them)
526 #include <linux/page-flags.h>
527 #include <linux/huge_mm.h>
530 * Methods to modify the page usage count.
532 * What counts for a page usage:
533 * - cache mapping (page->mapping)
534 * - private data (page->private)
535 * - page mapped in a task's page tables, each mapping
536 * is counted separately
538 * Also, many kernel routines increase the page count before a critical
539 * routine so they can be sure the page doesn't go away from under them.
543 * Drop a ref, return true if the refcount fell to zero (the page has no users)
545 static inline int put_page_testzero(struct page *page)
547 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
548 return page_ref_dec_and_test(page);
552 * Try to grab a ref unless the page has a refcount of zero, return false if
554 * This can be called when MMU is off so it must not access
555 * any of the virtual mappings.
557 static inline int get_page_unless_zero(struct page *page)
559 return page_ref_add_unless(page, 1, 0);
562 extern int page_is_ram(unsigned long pfn);
570 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
573 /* Support for virtually mapped pages */
574 struct page *vmalloc_to_page(const void *addr);
575 unsigned long vmalloc_to_pfn(const void *addr);
578 * Determine if an address is within the vmalloc range
580 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
581 * is no special casing required.
583 static inline bool is_vmalloc_addr(const void *x)
586 unsigned long addr = (unsigned long)x;
588 return addr >= VMALLOC_START && addr < VMALLOC_END;
594 extern int is_vmalloc_or_module_addr(const void *x);
596 static inline int is_vmalloc_or_module_addr(const void *x)
602 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
603 static inline void *kvmalloc(size_t size, gfp_t flags)
605 return kvmalloc_node(size, flags, NUMA_NO_NODE);
607 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
609 return kvmalloc_node(size, flags | __GFP_ZERO, node);
611 static inline void *kvzalloc(size_t size, gfp_t flags)
613 return kvmalloc(size, flags | __GFP_ZERO);
616 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
620 if (unlikely(check_mul_overflow(n, size, &bytes)))
623 return kvmalloc(bytes, flags);
626 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
628 return kvmalloc_array(n, size, flags | __GFP_ZERO);
631 extern void kvfree(const void *addr);
633 static inline atomic_t *compound_mapcount_ptr(struct page *page)
635 return &page[1].compound_mapcount;
638 static inline int compound_mapcount(struct page *page)
640 VM_BUG_ON_PAGE(!PageCompound(page), page);
641 page = compound_head(page);
642 return atomic_read(compound_mapcount_ptr(page)) + 1;
646 * The atomic page->_mapcount, starts from -1: so that transitions
647 * both from it and to it can be tracked, using atomic_inc_and_test
648 * and atomic_add_negative(-1).
650 static inline void page_mapcount_reset(struct page *page)
652 atomic_set(&(page)->_mapcount, -1);
655 int __page_mapcount(struct page *page);
657 static inline int page_mapcount(struct page *page)
659 VM_BUG_ON_PAGE(PageSlab(page), page);
661 if (unlikely(PageCompound(page)))
662 return __page_mapcount(page);
663 return atomic_read(&page->_mapcount) + 1;
666 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
667 int total_mapcount(struct page *page);
668 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
670 static inline int total_mapcount(struct page *page)
672 return page_mapcount(page);
674 static inline int page_trans_huge_mapcount(struct page *page,
677 int mapcount = page_mapcount(page);
679 *total_mapcount = mapcount;
684 static inline struct page *virt_to_head_page(const void *x)
686 struct page *page = virt_to_page(x);
688 return compound_head(page);
691 void __put_page(struct page *page);
693 void put_pages_list(struct list_head *pages);
695 void split_page(struct page *page, unsigned int order);
698 * Compound pages have a destructor function. Provide a
699 * prototype for that function and accessor functions.
700 * These are _only_ valid on the head of a compound page.
702 typedef void compound_page_dtor(struct page *);
704 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
705 enum compound_dtor_id {
708 #ifdef CONFIG_HUGETLB_PAGE
711 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
716 extern compound_page_dtor * const compound_page_dtors[];
718 static inline void set_compound_page_dtor(struct page *page,
719 enum compound_dtor_id compound_dtor)
721 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
722 page[1].compound_dtor = compound_dtor;
725 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
727 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
728 return compound_page_dtors[page[1].compound_dtor];
731 static inline unsigned int compound_order(struct page *page)
735 return page[1].compound_order;
738 static inline void set_compound_order(struct page *page, unsigned int order)
740 page[1].compound_order = order;
743 void free_compound_page(struct page *page);
747 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
748 * servicing faults for write access. In the normal case, do always want
749 * pte_mkwrite. But get_user_pages can cause write faults for mappings
750 * that do not have writing enabled, when used by access_process_vm.
752 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
754 if (likely(vma->vm_flags & VM_WRITE))
755 pte = pte_mkwrite(pte);
759 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
761 vm_fault_t finish_fault(struct vm_fault *vmf);
762 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
766 * Multiple processes may "see" the same page. E.g. for untouched
767 * mappings of /dev/null, all processes see the same page full of
768 * zeroes, and text pages of executables and shared libraries have
769 * only one copy in memory, at most, normally.
771 * For the non-reserved pages, page_count(page) denotes a reference count.
772 * page_count() == 0 means the page is free. page->lru is then used for
773 * freelist management in the buddy allocator.
774 * page_count() > 0 means the page has been allocated.
776 * Pages are allocated by the slab allocator in order to provide memory
777 * to kmalloc and kmem_cache_alloc. In this case, the management of the
778 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
779 * unless a particular usage is carefully commented. (the responsibility of
780 * freeing the kmalloc memory is the caller's, of course).
782 * A page may be used by anyone else who does a __get_free_page().
783 * In this case, page_count still tracks the references, and should only
784 * be used through the normal accessor functions. The top bits of page->flags
785 * and page->virtual store page management information, but all other fields
786 * are unused and could be used privately, carefully. The management of this
787 * page is the responsibility of the one who allocated it, and those who have
788 * subsequently been given references to it.
790 * The other pages (we may call them "pagecache pages") are completely
791 * managed by the Linux memory manager: I/O, buffers, swapping etc.
792 * The following discussion applies only to them.
794 * A pagecache page contains an opaque `private' member, which belongs to the
795 * page's address_space. Usually, this is the address of a circular list of
796 * the page's disk buffers. PG_private must be set to tell the VM to call
797 * into the filesystem to release these pages.
799 * A page may belong to an inode's memory mapping. In this case, page->mapping
800 * is the pointer to the inode, and page->index is the file offset of the page,
801 * in units of PAGE_SIZE.
803 * If pagecache pages are not associated with an inode, they are said to be
804 * anonymous pages. These may become associated with the swapcache, and in that
805 * case PG_swapcache is set, and page->private is an offset into the swapcache.
807 * In either case (swapcache or inode backed), the pagecache itself holds one
808 * reference to the page. Setting PG_private should also increment the
809 * refcount. The each user mapping also has a reference to the page.
811 * The pagecache pages are stored in a per-mapping radix tree, which is
812 * rooted at mapping->i_pages, and indexed by offset.
813 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
814 * lists, we instead now tag pages as dirty/writeback in the radix tree.
816 * All pagecache pages may be subject to I/O:
817 * - inode pages may need to be read from disk,
818 * - inode pages which have been modified and are MAP_SHARED may need
819 * to be written back to the inode on disk,
820 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
821 * modified may need to be swapped out to swap space and (later) to be read
826 * The zone field is never updated after free_area_init_core()
827 * sets it, so none of the operations on it need to be atomic.
830 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
831 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
832 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
833 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
834 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
835 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
838 * Define the bit shifts to access each section. For non-existent
839 * sections we define the shift as 0; that plus a 0 mask ensures
840 * the compiler will optimise away reference to them.
842 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
843 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
844 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
845 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
846 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
848 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
849 #ifdef NODE_NOT_IN_PAGE_FLAGS
850 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
851 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
852 SECTIONS_PGOFF : ZONES_PGOFF)
854 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
855 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
856 NODES_PGOFF : ZONES_PGOFF)
859 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
861 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
862 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
865 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
866 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
867 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
868 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
869 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
870 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
872 static inline enum zone_type page_zonenum(const struct page *page)
874 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
877 #ifdef CONFIG_ZONE_DEVICE
878 static inline bool is_zone_device_page(const struct page *page)
880 return page_zonenum(page) == ZONE_DEVICE;
882 extern void memmap_init_zone_device(struct zone *, unsigned long,
883 unsigned long, struct dev_pagemap *);
885 static inline bool is_zone_device_page(const struct page *page)
891 #ifdef CONFIG_DEV_PAGEMAP_OPS
892 void dev_pagemap_get_ops(void);
893 void dev_pagemap_put_ops(void);
894 void __put_devmap_managed_page(struct page *page);
895 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
896 static inline bool put_devmap_managed_page(struct page *page)
898 if (!static_branch_unlikely(&devmap_managed_key))
900 if (!is_zone_device_page(page))
902 switch (page->pgmap->type) {
903 case MEMORY_DEVICE_PRIVATE:
904 case MEMORY_DEVICE_PUBLIC:
905 case MEMORY_DEVICE_FS_DAX:
906 __put_devmap_managed_page(page);
914 static inline bool is_device_private_page(const struct page *page)
916 return is_zone_device_page(page) &&
917 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
920 static inline bool is_device_public_page(const struct page *page)
922 return is_zone_device_page(page) &&
923 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
926 #ifdef CONFIG_PCI_P2PDMA
927 static inline bool is_pci_p2pdma_page(const struct page *page)
929 return is_zone_device_page(page) &&
930 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
932 #else /* CONFIG_PCI_P2PDMA */
933 static inline bool is_pci_p2pdma_page(const struct page *page)
937 #endif /* CONFIG_PCI_P2PDMA */
939 #else /* CONFIG_DEV_PAGEMAP_OPS */
940 static inline void dev_pagemap_get_ops(void)
944 static inline void dev_pagemap_put_ops(void)
948 static inline bool put_devmap_managed_page(struct page *page)
953 static inline bool is_device_private_page(const struct page *page)
958 static inline bool is_device_public_page(const struct page *page)
963 static inline bool is_pci_p2pdma_page(const struct page *page)
967 #endif /* CONFIG_DEV_PAGEMAP_OPS */
969 /* 127: arbitrary random number, small enough to assemble well */
970 #define page_ref_zero_or_close_to_overflow(page) \
971 ((unsigned int) page_ref_count(page) + 127u <= 127u)
973 static inline void get_page(struct page *page)
975 page = compound_head(page);
977 * Getting a normal page or the head of a compound page
978 * requires to already have an elevated page->_refcount.
980 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
984 static inline __must_check bool try_get_page(struct page *page)
986 page = compound_head(page);
987 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
993 static inline void put_page(struct page *page)
995 page = compound_head(page);
998 * For devmap managed pages we need to catch refcount transition from
999 * 2 to 1, when refcount reach one it means the page is free and we
1000 * need to inform the device driver through callback. See
1001 * include/linux/memremap.h and HMM for details.
1003 if (put_devmap_managed_page(page))
1006 if (put_page_testzero(page))
1010 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1011 #define SECTION_IN_PAGE_FLAGS
1015 * The identification function is mainly used by the buddy allocator for
1016 * determining if two pages could be buddies. We are not really identifying
1017 * the zone since we could be using the section number id if we do not have
1018 * node id available in page flags.
1019 * We only guarantee that it will return the same value for two combinable
1022 static inline int page_zone_id(struct page *page)
1024 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1027 #ifdef NODE_NOT_IN_PAGE_FLAGS
1028 extern int page_to_nid(const struct page *page);
1030 static inline int page_to_nid(const struct page *page)
1032 struct page *p = (struct page *)page;
1034 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1038 #ifdef CONFIG_NUMA_BALANCING
1039 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1041 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1044 static inline int cpupid_to_pid(int cpupid)
1046 return cpupid & LAST__PID_MASK;
1049 static inline int cpupid_to_cpu(int cpupid)
1051 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1054 static inline int cpupid_to_nid(int cpupid)
1056 return cpu_to_node(cpupid_to_cpu(cpupid));
1059 static inline bool cpupid_pid_unset(int cpupid)
1061 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1064 static inline bool cpupid_cpu_unset(int cpupid)
1066 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1069 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1071 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1074 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1075 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1076 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1078 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1081 static inline int page_cpupid_last(struct page *page)
1083 return page->_last_cpupid;
1085 static inline void page_cpupid_reset_last(struct page *page)
1087 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1090 static inline int page_cpupid_last(struct page *page)
1092 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1095 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1097 static inline void page_cpupid_reset_last(struct page *page)
1099 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1101 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1102 #else /* !CONFIG_NUMA_BALANCING */
1103 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1105 return page_to_nid(page); /* XXX */
1108 static inline int page_cpupid_last(struct page *page)
1110 return page_to_nid(page); /* XXX */
1113 static inline int cpupid_to_nid(int cpupid)
1118 static inline int cpupid_to_pid(int cpupid)
1123 static inline int cpupid_to_cpu(int cpupid)
1128 static inline int cpu_pid_to_cpupid(int nid, int pid)
1133 static inline bool cpupid_pid_unset(int cpupid)
1138 static inline void page_cpupid_reset_last(struct page *page)
1142 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1146 #endif /* CONFIG_NUMA_BALANCING */
1148 #ifdef CONFIG_KASAN_SW_TAGS
1149 static inline u8 page_kasan_tag(const struct page *page)
1151 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1154 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1156 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1157 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1160 static inline void page_kasan_tag_reset(struct page *page)
1162 page_kasan_tag_set(page, 0xff);
1165 static inline u8 page_kasan_tag(const struct page *page)
1170 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1171 static inline void page_kasan_tag_reset(struct page *page) { }
1174 static inline struct zone *page_zone(const struct page *page)
1176 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1179 static inline pg_data_t *page_pgdat(const struct page *page)
1181 return NODE_DATA(page_to_nid(page));
1184 #ifdef SECTION_IN_PAGE_FLAGS
1185 static inline void set_page_section(struct page *page, unsigned long section)
1187 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1188 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1191 static inline unsigned long page_to_section(const struct page *page)
1193 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1197 static inline void set_page_zone(struct page *page, enum zone_type zone)
1199 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1200 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1203 static inline void set_page_node(struct page *page, unsigned long node)
1205 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1206 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1209 static inline void set_page_links(struct page *page, enum zone_type zone,
1210 unsigned long node, unsigned long pfn)
1212 set_page_zone(page, zone);
1213 set_page_node(page, node);
1214 #ifdef SECTION_IN_PAGE_FLAGS
1215 set_page_section(page, pfn_to_section_nr(pfn));
1220 static inline struct mem_cgroup *page_memcg(struct page *page)
1222 return page->mem_cgroup;
1224 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1226 WARN_ON_ONCE(!rcu_read_lock_held());
1227 return READ_ONCE(page->mem_cgroup);
1230 static inline struct mem_cgroup *page_memcg(struct page *page)
1234 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1236 WARN_ON_ONCE(!rcu_read_lock_held());
1242 * Some inline functions in vmstat.h depend on page_zone()
1244 #include <linux/vmstat.h>
1246 static __always_inline void *lowmem_page_address(const struct page *page)
1248 return page_to_virt(page);
1251 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1252 #define HASHED_PAGE_VIRTUAL
1255 #if defined(WANT_PAGE_VIRTUAL)
1256 static inline void *page_address(const struct page *page)
1258 return page->virtual;
1260 static inline void set_page_address(struct page *page, void *address)
1262 page->virtual = address;
1264 #define page_address_init() do { } while(0)
1267 #if defined(HASHED_PAGE_VIRTUAL)
1268 void *page_address(const struct page *page);
1269 void set_page_address(struct page *page, void *virtual);
1270 void page_address_init(void);
1273 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1274 #define page_address(page) lowmem_page_address(page)
1275 #define set_page_address(page, address) do { } while(0)
1276 #define page_address_init() do { } while(0)
1279 extern void *page_rmapping(struct page *page);
1280 extern struct anon_vma *page_anon_vma(struct page *page);
1281 extern struct address_space *page_mapping(struct page *page);
1283 extern struct address_space *__page_file_mapping(struct page *);
1286 struct address_space *page_file_mapping(struct page *page)
1288 if (unlikely(PageSwapCache(page)))
1289 return __page_file_mapping(page);
1291 return page->mapping;
1294 extern pgoff_t __page_file_index(struct page *page);
1297 * Return the pagecache index of the passed page. Regular pagecache pages
1298 * use ->index whereas swapcache pages use swp_offset(->private)
1300 static inline pgoff_t page_index(struct page *page)
1302 if (unlikely(PageSwapCache(page)))
1303 return __page_file_index(page);
1307 bool page_mapped(struct page *page);
1308 struct address_space *page_mapping(struct page *page);
1309 struct address_space *page_mapping_file(struct page *page);
1312 * Return true only if the page has been allocated with
1313 * ALLOC_NO_WATERMARKS and the low watermark was not
1314 * met implying that the system is under some pressure.
1316 static inline bool page_is_pfmemalloc(struct page *page)
1319 * Page index cannot be this large so this must be
1320 * a pfmemalloc page.
1322 return page->index == -1UL;
1326 * Only to be called by the page allocator on a freshly allocated
1329 static inline void set_page_pfmemalloc(struct page *page)
1334 static inline void clear_page_pfmemalloc(struct page *page)
1340 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1342 extern void pagefault_out_of_memory(void);
1344 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1347 * Flags passed to show_mem() and show_free_areas() to suppress output in
1350 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1352 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1354 extern bool can_do_mlock(void);
1355 extern int user_shm_lock(size_t, struct user_struct *);
1356 extern void user_shm_unlock(size_t, struct user_struct *);
1359 * Parameter block passed down to zap_pte_range in exceptional cases.
1361 struct zap_details {
1362 struct address_space *check_mapping; /* Check page->mapping if set */
1363 pgoff_t first_index; /* Lowest page->index to unmap */
1364 pgoff_t last_index; /* Highest page->index to unmap */
1367 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1368 pte_t pte, bool with_public_device);
1369 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1371 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1374 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1375 unsigned long size);
1376 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1377 unsigned long size);
1378 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1379 unsigned long start, unsigned long end);
1382 * mm_walk - callbacks for walk_page_range
1383 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1384 * this handler should only handle pud_trans_huge() puds.
1385 * the pmd_entry or pte_entry callbacks will be used for
1387 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1388 * this handler is required to be able to handle
1389 * pmd_trans_huge() pmds. They may simply choose to
1390 * split_huge_page() instead of handling it explicitly.
1391 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1392 * @pte_hole: if set, called for each hole at all levels
1393 * @hugetlb_entry: if set, called for each hugetlb entry
1394 * @test_walk: caller specific callback function to determine whether
1395 * we walk over the current vma or not. Returning 0
1396 * value means "do page table walk over the current vma,"
1397 * and a negative one means "abort current page table walk
1398 * right now." 1 means "skip the current vma."
1399 * @mm: mm_struct representing the target process of page table walk
1400 * @vma: vma currently walked (NULL if walking outside vmas)
1401 * @private: private data for callbacks' usage
1403 * (see the comment on walk_page_range() for more details)
1406 int (*pud_entry)(pud_t *pud, unsigned long addr,
1407 unsigned long next, struct mm_walk *walk);
1408 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1409 unsigned long next, struct mm_walk *walk);
1410 int (*pte_entry)(pte_t *pte, unsigned long addr,
1411 unsigned long next, struct mm_walk *walk);
1412 int (*pte_hole)(unsigned long addr, unsigned long next,
1413 struct mm_walk *walk);
1414 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1415 unsigned long addr, unsigned long next,
1416 struct mm_walk *walk);
1417 int (*test_walk)(unsigned long addr, unsigned long next,
1418 struct mm_walk *walk);
1419 struct mm_struct *mm;
1420 struct vm_area_struct *vma;
1424 struct mmu_notifier_range;
1426 int walk_page_range(unsigned long addr, unsigned long end,
1427 struct mm_walk *walk);
1428 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1429 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1430 unsigned long end, unsigned long floor, unsigned long ceiling);
1431 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1432 struct vm_area_struct *vma);
1433 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1434 struct mmu_notifier_range *range,
1435 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1436 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1437 unsigned long *pfn);
1438 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1439 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1440 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1441 void *buf, int len, int write);
1443 extern void truncate_pagecache(struct inode *inode, loff_t new);
1444 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1445 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1446 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1447 int truncate_inode_page(struct address_space *mapping, struct page *page);
1448 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1449 int invalidate_inode_page(struct page *page);
1452 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1453 unsigned long address, unsigned int flags);
1454 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1455 unsigned long address, unsigned int fault_flags,
1457 void unmap_mapping_pages(struct address_space *mapping,
1458 pgoff_t start, pgoff_t nr, bool even_cows);
1459 void unmap_mapping_range(struct address_space *mapping,
1460 loff_t const holebegin, loff_t const holelen, int even_cows);
1462 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1463 unsigned long address, unsigned int flags)
1465 /* should never happen if there's no MMU */
1467 return VM_FAULT_SIGBUS;
1469 static inline int fixup_user_fault(struct task_struct *tsk,
1470 struct mm_struct *mm, unsigned long address,
1471 unsigned int fault_flags, bool *unlocked)
1473 /* should never happen if there's no MMU */
1477 static inline void unmap_mapping_pages(struct address_space *mapping,
1478 pgoff_t start, pgoff_t nr, bool even_cows) { }
1479 static inline void unmap_mapping_range(struct address_space *mapping,
1480 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1483 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1484 loff_t const holebegin, loff_t const holelen)
1486 unmap_mapping_range(mapping, holebegin, holelen, 0);
1489 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1490 void *buf, int len, unsigned int gup_flags);
1491 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1492 void *buf, int len, unsigned int gup_flags);
1493 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1494 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1496 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1497 unsigned long start, unsigned long nr_pages,
1498 unsigned int gup_flags, struct page **pages,
1499 struct vm_area_struct **vmas, int *locked);
1500 long get_user_pages(unsigned long start, unsigned long nr_pages,
1501 unsigned int gup_flags, struct page **pages,
1502 struct vm_area_struct **vmas);
1503 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1504 unsigned int gup_flags, struct page **pages, int *locked);
1505 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1506 struct page **pages, unsigned int gup_flags);
1508 #if defined(CONFIG_FS_DAX) || defined(CONFIG_CMA)
1509 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1510 unsigned int gup_flags, struct page **pages,
1511 struct vm_area_struct **vmas);
1513 static inline long get_user_pages_longterm(unsigned long start,
1514 unsigned long nr_pages, unsigned int gup_flags,
1515 struct page **pages, struct vm_area_struct **vmas)
1517 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1519 #endif /* CONFIG_FS_DAX */
1521 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1522 struct page **pages);
1524 /* Container for pinned pfns / pages */
1525 struct frame_vector {
1526 unsigned int nr_allocated; /* Number of frames we have space for */
1527 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1528 bool got_ref; /* Did we pin pages by getting page ref? */
1529 bool is_pfns; /* Does array contain pages or pfns? */
1530 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1531 * pfns_vector_pages() or pfns_vector_pfns()
1535 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1536 void frame_vector_destroy(struct frame_vector *vec);
1537 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1538 unsigned int gup_flags, struct frame_vector *vec);
1539 void put_vaddr_frames(struct frame_vector *vec);
1540 int frame_vector_to_pages(struct frame_vector *vec);
1541 void frame_vector_to_pfns(struct frame_vector *vec);
1543 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1545 return vec->nr_frames;
1548 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1551 int err = frame_vector_to_pages(vec);
1554 return ERR_PTR(err);
1556 return (struct page **)(vec->ptrs);
1559 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1562 frame_vector_to_pfns(vec);
1563 return (unsigned long *)(vec->ptrs);
1567 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1568 struct page **pages);
1569 int get_kernel_page(unsigned long start, int write, struct page **pages);
1570 struct page *get_dump_page(unsigned long addr);
1572 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1573 extern void do_invalidatepage(struct page *page, unsigned int offset,
1574 unsigned int length);
1576 void __set_page_dirty(struct page *, struct address_space *, int warn);
1577 int __set_page_dirty_nobuffers(struct page *page);
1578 int __set_page_dirty_no_writeback(struct page *page);
1579 int redirty_page_for_writepage(struct writeback_control *wbc,
1581 void account_page_dirtied(struct page *page, struct address_space *mapping);
1582 void account_page_cleaned(struct page *page, struct address_space *mapping,
1583 struct bdi_writeback *wb);
1584 int set_page_dirty(struct page *page);
1585 int set_page_dirty_lock(struct page *page);
1586 void __cancel_dirty_page(struct page *page);
1587 static inline void cancel_dirty_page(struct page *page)
1589 /* Avoid atomic ops, locking, etc. when not actually needed. */
1590 if (PageDirty(page))
1591 __cancel_dirty_page(page);
1593 int clear_page_dirty_for_io(struct page *page);
1595 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1597 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1599 return !vma->vm_ops;
1604 * The vma_is_shmem is not inline because it is used only by slow
1605 * paths in userfault.
1607 bool vma_is_shmem(struct vm_area_struct *vma);
1609 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1612 int vma_is_stack_for_current(struct vm_area_struct *vma);
1614 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1615 unsigned long old_addr, struct vm_area_struct *new_vma,
1616 unsigned long new_addr, unsigned long len,
1617 bool need_rmap_locks);
1618 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1619 unsigned long end, pgprot_t newprot,
1620 int dirty_accountable, int prot_numa);
1621 extern int mprotect_fixup(struct vm_area_struct *vma,
1622 struct vm_area_struct **pprev, unsigned long start,
1623 unsigned long end, unsigned long newflags);
1626 * doesn't attempt to fault and will return short.
1628 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1629 struct page **pages);
1631 * per-process(per-mm_struct) statistics.
1633 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1635 long val = atomic_long_read(&mm->rss_stat.count[member]);
1637 #ifdef SPLIT_RSS_COUNTING
1639 * counter is updated in asynchronous manner and may go to minus.
1640 * But it's never be expected number for users.
1645 return (unsigned long)val;
1648 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1650 atomic_long_add(value, &mm->rss_stat.count[member]);
1653 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1655 atomic_long_inc(&mm->rss_stat.count[member]);
1658 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1660 atomic_long_dec(&mm->rss_stat.count[member]);
1663 /* Optimized variant when page is already known not to be PageAnon */
1664 static inline int mm_counter_file(struct page *page)
1666 if (PageSwapBacked(page))
1667 return MM_SHMEMPAGES;
1668 return MM_FILEPAGES;
1671 static inline int mm_counter(struct page *page)
1674 return MM_ANONPAGES;
1675 return mm_counter_file(page);
1678 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1680 return get_mm_counter(mm, MM_FILEPAGES) +
1681 get_mm_counter(mm, MM_ANONPAGES) +
1682 get_mm_counter(mm, MM_SHMEMPAGES);
1685 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1687 return max(mm->hiwater_rss, get_mm_rss(mm));
1690 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1692 return max(mm->hiwater_vm, mm->total_vm);
1695 static inline void update_hiwater_rss(struct mm_struct *mm)
1697 unsigned long _rss = get_mm_rss(mm);
1699 if ((mm)->hiwater_rss < _rss)
1700 (mm)->hiwater_rss = _rss;
1703 static inline void update_hiwater_vm(struct mm_struct *mm)
1705 if (mm->hiwater_vm < mm->total_vm)
1706 mm->hiwater_vm = mm->total_vm;
1709 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1711 mm->hiwater_rss = get_mm_rss(mm);
1714 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1715 struct mm_struct *mm)
1717 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1719 if (*maxrss < hiwater_rss)
1720 *maxrss = hiwater_rss;
1723 #if defined(SPLIT_RSS_COUNTING)
1724 void sync_mm_rss(struct mm_struct *mm);
1726 static inline void sync_mm_rss(struct mm_struct *mm)
1731 #ifndef __HAVE_ARCH_PTE_DEVMAP
1732 static inline int pte_devmap(pte_t pte)
1738 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1740 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1742 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1746 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1750 #ifdef __PAGETABLE_P4D_FOLDED
1751 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1752 unsigned long address)
1757 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1760 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1761 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1762 unsigned long address)
1766 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1767 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1770 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1772 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1774 if (mm_pud_folded(mm))
1776 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1779 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1781 if (mm_pud_folded(mm))
1783 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1787 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1788 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1789 unsigned long address)
1794 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1795 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1798 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1800 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1802 if (mm_pmd_folded(mm))
1804 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1807 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1809 if (mm_pmd_folded(mm))
1811 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1816 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1818 atomic_long_set(&mm->pgtables_bytes, 0);
1821 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1823 return atomic_long_read(&mm->pgtables_bytes);
1826 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1828 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1831 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1833 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1837 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1838 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1843 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1844 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1847 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1848 int __pte_alloc_kernel(pmd_t *pmd);
1851 * The following ifdef needed to get the 4level-fixup.h header to work.
1852 * Remove it when 4level-fixup.h has been removed.
1854 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1856 #ifndef __ARCH_HAS_5LEVEL_HACK
1857 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1858 unsigned long address)
1860 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1861 NULL : p4d_offset(pgd, address);
1864 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1865 unsigned long address)
1867 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1868 NULL : pud_offset(p4d, address);
1870 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1872 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1874 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1875 NULL: pmd_offset(pud, address);
1877 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1879 #if USE_SPLIT_PTE_PTLOCKS
1880 #if ALLOC_SPLIT_PTLOCKS
1881 void __init ptlock_cache_init(void);
1882 extern bool ptlock_alloc(struct page *page);
1883 extern void ptlock_free(struct page *page);
1885 static inline spinlock_t *ptlock_ptr(struct page *page)
1889 #else /* ALLOC_SPLIT_PTLOCKS */
1890 static inline void ptlock_cache_init(void)
1894 static inline bool ptlock_alloc(struct page *page)
1899 static inline void ptlock_free(struct page *page)
1903 static inline spinlock_t *ptlock_ptr(struct page *page)
1907 #endif /* ALLOC_SPLIT_PTLOCKS */
1909 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1911 return ptlock_ptr(pmd_page(*pmd));
1914 static inline bool ptlock_init(struct page *page)
1917 * prep_new_page() initialize page->private (and therefore page->ptl)
1918 * with 0. Make sure nobody took it in use in between.
1920 * It can happen if arch try to use slab for page table allocation:
1921 * slab code uses page->slab_cache, which share storage with page->ptl.
1923 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1924 if (!ptlock_alloc(page))
1926 spin_lock_init(ptlock_ptr(page));
1930 #else /* !USE_SPLIT_PTE_PTLOCKS */
1932 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1934 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1936 return &mm->page_table_lock;
1938 static inline void ptlock_cache_init(void) {}
1939 static inline bool ptlock_init(struct page *page) { return true; }
1940 static inline void ptlock_free(struct page *page) {}
1941 #endif /* USE_SPLIT_PTE_PTLOCKS */
1943 static inline void pgtable_init(void)
1945 ptlock_cache_init();
1946 pgtable_cache_init();
1949 static inline bool pgtable_page_ctor(struct page *page)
1951 if (!ptlock_init(page))
1953 __SetPageTable(page);
1954 inc_zone_page_state(page, NR_PAGETABLE);
1958 static inline void pgtable_page_dtor(struct page *page)
1961 __ClearPageTable(page);
1962 dec_zone_page_state(page, NR_PAGETABLE);
1965 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1967 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1968 pte_t *__pte = pte_offset_map(pmd, address); \
1974 #define pte_unmap_unlock(pte, ptl) do { \
1979 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1981 #define pte_alloc_map(mm, pmd, address) \
1982 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1984 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1985 (pte_alloc(mm, pmd) ? \
1986 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1988 #define pte_alloc_kernel(pmd, address) \
1989 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1990 NULL: pte_offset_kernel(pmd, address))
1992 #if USE_SPLIT_PMD_PTLOCKS
1994 static struct page *pmd_to_page(pmd_t *pmd)
1996 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1997 return virt_to_page((void *)((unsigned long) pmd & mask));
2000 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2002 return ptlock_ptr(pmd_to_page(pmd));
2005 static inline bool pgtable_pmd_page_ctor(struct page *page)
2007 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2008 page->pmd_huge_pte = NULL;
2010 return ptlock_init(page);
2013 static inline void pgtable_pmd_page_dtor(struct page *page)
2015 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2016 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2021 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2025 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2027 return &mm->page_table_lock;
2030 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2031 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2033 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2037 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2039 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2045 * No scalability reason to split PUD locks yet, but follow the same pattern
2046 * as the PMD locks to make it easier if we decide to. The VM should not be
2047 * considered ready to switch to split PUD locks yet; there may be places
2048 * which need to be converted from page_table_lock.
2050 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2052 return &mm->page_table_lock;
2055 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2057 spinlock_t *ptl = pud_lockptr(mm, pud);
2063 extern void __init pagecache_init(void);
2064 extern void free_area_init(unsigned long * zones_size);
2065 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2066 unsigned long zone_start_pfn, unsigned long *zholes_size);
2067 extern void free_initmem(void);
2070 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2071 * into the buddy system. The freed pages will be poisoned with pattern
2072 * "poison" if it's within range [0, UCHAR_MAX].
2073 * Return pages freed into the buddy system.
2075 extern unsigned long free_reserved_area(void *start, void *end,
2076 int poison, const char *s);
2078 #ifdef CONFIG_HIGHMEM
2080 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2081 * and totalram_pages.
2083 extern void free_highmem_page(struct page *page);
2086 extern void adjust_managed_page_count(struct page *page, long count);
2087 extern void mem_init_print_info(const char *str);
2089 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2091 /* Free the reserved page into the buddy system, so it gets managed. */
2092 static inline void __free_reserved_page(struct page *page)
2094 ClearPageReserved(page);
2095 init_page_count(page);
2099 static inline void free_reserved_page(struct page *page)
2101 __free_reserved_page(page);
2102 adjust_managed_page_count(page, 1);
2105 static inline void mark_page_reserved(struct page *page)
2107 SetPageReserved(page);
2108 adjust_managed_page_count(page, -1);
2112 * Default method to free all the __init memory into the buddy system.
2113 * The freed pages will be poisoned with pattern "poison" if it's within
2114 * range [0, UCHAR_MAX].
2115 * Return pages freed into the buddy system.
2117 static inline unsigned long free_initmem_default(int poison)
2119 extern char __init_begin[], __init_end[];
2121 return free_reserved_area(&__init_begin, &__init_end,
2122 poison, "unused kernel");
2125 static inline unsigned long get_num_physpages(void)
2128 unsigned long phys_pages = 0;
2130 for_each_online_node(nid)
2131 phys_pages += node_present_pages(nid);
2136 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2138 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2139 * zones, allocate the backing mem_map and account for memory holes in a more
2140 * architecture independent manner. This is a substitute for creating the
2141 * zone_sizes[] and zholes_size[] arrays and passing them to
2142 * free_area_init_node()
2144 * An architecture is expected to register range of page frames backed by
2145 * physical memory with memblock_add[_node]() before calling
2146 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2147 * usage, an architecture is expected to do something like
2149 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2151 * for_each_valid_physical_page_range()
2152 * memblock_add_node(base, size, nid)
2153 * free_area_init_nodes(max_zone_pfns);
2155 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2156 * registered physical page range. Similarly
2157 * sparse_memory_present_with_active_regions() calls memory_present() for
2158 * each range when SPARSEMEM is enabled.
2160 * See mm/page_alloc.c for more information on each function exposed by
2161 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2163 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2164 unsigned long node_map_pfn_alignment(void);
2165 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2166 unsigned long end_pfn);
2167 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2168 unsigned long end_pfn);
2169 extern void get_pfn_range_for_nid(unsigned int nid,
2170 unsigned long *start_pfn, unsigned long *end_pfn);
2171 extern unsigned long find_min_pfn_with_active_regions(void);
2172 extern void free_bootmem_with_active_regions(int nid,
2173 unsigned long max_low_pfn);
2174 extern void sparse_memory_present_with_active_regions(int nid);
2176 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2178 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2179 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2180 static inline int __early_pfn_to_nid(unsigned long pfn,
2181 struct mminit_pfnnid_cache *state)
2186 /* please see mm/page_alloc.c */
2187 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2188 /* there is a per-arch backend function. */
2189 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2190 struct mminit_pfnnid_cache *state);
2193 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2194 void zero_resv_unavail(void);
2196 static inline void zero_resv_unavail(void) {}
2199 extern void set_dma_reserve(unsigned long new_dma_reserve);
2200 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2201 enum memmap_context, struct vmem_altmap *);
2202 extern void setup_per_zone_wmarks(void);
2203 extern int __meminit init_per_zone_wmark_min(void);
2204 extern void mem_init(void);
2205 extern void __init mmap_init(void);
2206 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2207 extern long si_mem_available(void);
2208 extern void si_meminfo(struct sysinfo * val);
2209 extern void si_meminfo_node(struct sysinfo *val, int nid);
2210 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2211 extern unsigned long arch_reserved_kernel_pages(void);
2214 extern __printf(3, 4)
2215 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2217 extern void setup_per_cpu_pageset(void);
2219 extern void zone_pcp_update(struct zone *zone);
2220 extern void zone_pcp_reset(struct zone *zone);
2223 extern int min_free_kbytes;
2224 extern int watermark_boost_factor;
2225 extern int watermark_scale_factor;
2228 extern atomic_long_t mmap_pages_allocated;
2229 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2231 /* interval_tree.c */
2232 void vma_interval_tree_insert(struct vm_area_struct *node,
2233 struct rb_root_cached *root);
2234 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2235 struct vm_area_struct *prev,
2236 struct rb_root_cached *root);
2237 void vma_interval_tree_remove(struct vm_area_struct *node,
2238 struct rb_root_cached *root);
2239 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2240 unsigned long start, unsigned long last);
2241 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2242 unsigned long start, unsigned long last);
2244 #define vma_interval_tree_foreach(vma, root, start, last) \
2245 for (vma = vma_interval_tree_iter_first(root, start, last); \
2246 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2248 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2249 struct rb_root_cached *root);
2250 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2251 struct rb_root_cached *root);
2252 struct anon_vma_chain *
2253 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2254 unsigned long start, unsigned long last);
2255 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2256 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2257 #ifdef CONFIG_DEBUG_VM_RB
2258 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2261 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2262 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2263 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2266 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2267 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2268 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2269 struct vm_area_struct *expand);
2270 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2271 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2273 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2275 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2276 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2277 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2278 struct mempolicy *, struct vm_userfaultfd_ctx);
2279 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2280 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2281 unsigned long addr, int new_below);
2282 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2283 unsigned long addr, int new_below);
2284 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2285 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2286 struct rb_node **, struct rb_node *);
2287 extern void unlink_file_vma(struct vm_area_struct *);
2288 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2289 unsigned long addr, unsigned long len, pgoff_t pgoff,
2290 bool *need_rmap_locks);
2291 extern void exit_mmap(struct mm_struct *);
2293 static inline int check_data_rlimit(unsigned long rlim,
2295 unsigned long start,
2296 unsigned long end_data,
2297 unsigned long start_data)
2299 if (rlim < RLIM_INFINITY) {
2300 if (((new - start) + (end_data - start_data)) > rlim)
2307 extern int mm_take_all_locks(struct mm_struct *mm);
2308 extern void mm_drop_all_locks(struct mm_struct *mm);
2310 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2311 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2312 extern struct file *get_task_exe_file(struct task_struct *task);
2314 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2315 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2317 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2318 const struct vm_special_mapping *sm);
2319 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2320 unsigned long addr, unsigned long len,
2321 unsigned long flags,
2322 const struct vm_special_mapping *spec);
2323 /* This is an obsolete alternative to _install_special_mapping. */
2324 extern int install_special_mapping(struct mm_struct *mm,
2325 unsigned long addr, unsigned long len,
2326 unsigned long flags, struct page **pages);
2328 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2330 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2331 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2332 struct list_head *uf);
2333 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2334 unsigned long len, unsigned long prot, unsigned long flags,
2335 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2336 struct list_head *uf);
2337 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2338 struct list_head *uf, bool downgrade);
2339 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2340 struct list_head *uf);
2342 static inline unsigned long
2343 do_mmap_pgoff(struct file *file, unsigned long addr,
2344 unsigned long len, unsigned long prot, unsigned long flags,
2345 unsigned long pgoff, unsigned long *populate,
2346 struct list_head *uf)
2348 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2352 extern int __mm_populate(unsigned long addr, unsigned long len,
2354 static inline void mm_populate(unsigned long addr, unsigned long len)
2357 (void) __mm_populate(addr, len, 1);
2360 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2363 /* These take the mm semaphore themselves */
2364 extern int __must_check vm_brk(unsigned long, unsigned long);
2365 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2366 extern int vm_munmap(unsigned long, size_t);
2367 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2368 unsigned long, unsigned long,
2369 unsigned long, unsigned long);
2371 struct vm_unmapped_area_info {
2372 #define VM_UNMAPPED_AREA_TOPDOWN 1
2373 unsigned long flags;
2374 unsigned long length;
2375 unsigned long low_limit;
2376 unsigned long high_limit;
2377 unsigned long align_mask;
2378 unsigned long align_offset;
2381 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2382 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2385 * Search for an unmapped address range.
2387 * We are looking for a range that:
2388 * - does not intersect with any VMA;
2389 * - is contained within the [low_limit, high_limit) interval;
2390 * - is at least the desired size.
2391 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2393 static inline unsigned long
2394 vm_unmapped_area(struct vm_unmapped_area_info *info)
2396 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2397 return unmapped_area_topdown(info);
2399 return unmapped_area(info);
2403 extern void truncate_inode_pages(struct address_space *, loff_t);
2404 extern void truncate_inode_pages_range(struct address_space *,
2405 loff_t lstart, loff_t lend);
2406 extern void truncate_inode_pages_final(struct address_space *);
2408 /* generic vm_area_ops exported for stackable file systems */
2409 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2410 extern void filemap_map_pages(struct vm_fault *vmf,
2411 pgoff_t start_pgoff, pgoff_t end_pgoff);
2412 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2414 /* mm/page-writeback.c */
2415 int __must_check write_one_page(struct page *page);
2416 void task_dirty_inc(struct task_struct *tsk);
2419 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2421 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2422 pgoff_t offset, unsigned long nr_to_read);
2424 void page_cache_sync_readahead(struct address_space *mapping,
2425 struct file_ra_state *ra,
2428 unsigned long size);
2430 void page_cache_async_readahead(struct address_space *mapping,
2431 struct file_ra_state *ra,
2435 unsigned long size);
2437 extern unsigned long stack_guard_gap;
2438 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2439 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2441 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2442 extern int expand_downwards(struct vm_area_struct *vma,
2443 unsigned long address);
2445 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2447 #define expand_upwards(vma, address) (0)
2450 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2451 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2452 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2453 struct vm_area_struct **pprev);
2455 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2456 NULL if none. Assume start_addr < end_addr. */
2457 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2459 struct vm_area_struct * vma = find_vma(mm,start_addr);
2461 if (vma && end_addr <= vma->vm_start)
2466 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2468 unsigned long vm_start = vma->vm_start;
2470 if (vma->vm_flags & VM_GROWSDOWN) {
2471 vm_start -= stack_guard_gap;
2472 if (vm_start > vma->vm_start)
2478 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2480 unsigned long vm_end = vma->vm_end;
2482 if (vma->vm_flags & VM_GROWSUP) {
2483 vm_end += stack_guard_gap;
2484 if (vm_end < vma->vm_end)
2485 vm_end = -PAGE_SIZE;
2490 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2492 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2495 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2496 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2497 unsigned long vm_start, unsigned long vm_end)
2499 struct vm_area_struct *vma = find_vma(mm, vm_start);
2501 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2507 static inline bool range_in_vma(struct vm_area_struct *vma,
2508 unsigned long start, unsigned long end)
2510 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2514 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2515 void vma_set_page_prot(struct vm_area_struct *vma);
2517 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2521 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2523 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2527 #ifdef CONFIG_NUMA_BALANCING
2528 unsigned long change_prot_numa(struct vm_area_struct *vma,
2529 unsigned long start, unsigned long end);
2532 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2533 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2534 unsigned long pfn, unsigned long size, pgprot_t);
2535 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2536 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2538 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2539 unsigned long pfn, pgprot_t pgprot);
2540 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2542 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2543 unsigned long addr, pfn_t pfn);
2544 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2546 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2547 unsigned long addr, struct page *page)
2549 int err = vm_insert_page(vma, addr, page);
2552 return VM_FAULT_OOM;
2553 if (err < 0 && err != -EBUSY)
2554 return VM_FAULT_SIGBUS;
2556 return VM_FAULT_NOPAGE;
2559 static inline vm_fault_t vmf_error(int err)
2562 return VM_FAULT_OOM;
2563 return VM_FAULT_SIGBUS;
2566 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2567 unsigned int foll_flags);
2569 #define FOLL_WRITE 0x01 /* check pte is writable */
2570 #define FOLL_TOUCH 0x02 /* mark page accessed */
2571 #define FOLL_GET 0x04 /* do get_page on page */
2572 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2573 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2574 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2575 * and return without waiting upon it */
2576 #define FOLL_POPULATE 0x40 /* fault in page */
2577 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2578 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2579 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2580 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2581 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2582 #define FOLL_MLOCK 0x1000 /* lock present pages */
2583 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2584 #define FOLL_COW 0x4000 /* internal GUP flag */
2585 #define FOLL_ANON 0x8000 /* don't do file mappings */
2587 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2589 if (vm_fault & VM_FAULT_OOM)
2591 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2592 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2593 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2598 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2600 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2601 unsigned long size, pte_fn_t fn, void *data);
2604 #ifdef CONFIG_PAGE_POISONING
2605 extern bool page_poisoning_enabled(void);
2606 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2608 static inline bool page_poisoning_enabled(void) { return false; }
2609 static inline void kernel_poison_pages(struct page *page, int numpages,
2613 extern bool _debug_pagealloc_enabled;
2615 static inline bool debug_pagealloc_enabled(void)
2617 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) && _debug_pagealloc_enabled;
2620 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2621 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2624 kernel_map_pages(struct page *page, int numpages, int enable)
2626 __kernel_map_pages(page, numpages, enable);
2628 #ifdef CONFIG_HIBERNATION
2629 extern bool kernel_page_present(struct page *page);
2630 #endif /* CONFIG_HIBERNATION */
2631 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2633 kernel_map_pages(struct page *page, int numpages, int enable) {}
2634 #ifdef CONFIG_HIBERNATION
2635 static inline bool kernel_page_present(struct page *page) { return true; }
2636 #endif /* CONFIG_HIBERNATION */
2637 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2639 #ifdef __HAVE_ARCH_GATE_AREA
2640 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2641 extern int in_gate_area_no_mm(unsigned long addr);
2642 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2644 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2648 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2649 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2653 #endif /* __HAVE_ARCH_GATE_AREA */
2655 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2657 #ifdef CONFIG_SYSCTL
2658 extern int sysctl_drop_caches;
2659 int drop_caches_sysctl_handler(struct ctl_table *, int,
2660 void __user *, size_t *, loff_t *);
2663 void drop_slab(void);
2664 void drop_slab_node(int nid);
2667 #define randomize_va_space 0
2669 extern int randomize_va_space;
2672 const char * arch_vma_name(struct vm_area_struct *vma);
2673 void print_vma_addr(char *prefix, unsigned long rip);
2675 void *sparse_buffer_alloc(unsigned long size);
2676 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2677 struct vmem_altmap *altmap);
2678 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2679 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2680 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2681 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2682 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2683 void *vmemmap_alloc_block(unsigned long size, int node);
2685 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2686 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2687 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2688 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2690 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2691 struct vmem_altmap *altmap);
2692 void vmemmap_populate_print_last(void);
2693 #ifdef CONFIG_MEMORY_HOTPLUG
2694 void vmemmap_free(unsigned long start, unsigned long end,
2695 struct vmem_altmap *altmap);
2697 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2698 unsigned long nr_pages);
2701 MF_COUNT_INCREASED = 1 << 0,
2702 MF_ACTION_REQUIRED = 1 << 1,
2703 MF_MUST_KILL = 1 << 2,
2704 MF_SOFT_OFFLINE = 1 << 3,
2706 extern int memory_failure(unsigned long pfn, int flags);
2707 extern void memory_failure_queue(unsigned long pfn, int flags);
2708 extern int unpoison_memory(unsigned long pfn);
2709 extern int get_hwpoison_page(struct page *page);
2710 #define put_hwpoison_page(page) put_page(page)
2711 extern int sysctl_memory_failure_early_kill;
2712 extern int sysctl_memory_failure_recovery;
2713 extern void shake_page(struct page *p, int access);
2714 extern atomic_long_t num_poisoned_pages __read_mostly;
2715 extern int soft_offline_page(struct page *page, int flags);
2719 * Error handlers for various types of pages.
2722 MF_IGNORED, /* Error: cannot be handled */
2723 MF_FAILED, /* Error: handling failed */
2724 MF_DELAYED, /* Will be handled later */
2725 MF_RECOVERED, /* Successfully recovered */
2728 enum mf_action_page_type {
2730 MF_MSG_KERNEL_HIGH_ORDER,
2732 MF_MSG_DIFFERENT_COMPOUND,
2733 MF_MSG_POISONED_HUGE,
2736 MF_MSG_NON_PMD_HUGE,
2737 MF_MSG_UNMAP_FAILED,
2738 MF_MSG_DIRTY_SWAPCACHE,
2739 MF_MSG_CLEAN_SWAPCACHE,
2740 MF_MSG_DIRTY_MLOCKED_LRU,
2741 MF_MSG_CLEAN_MLOCKED_LRU,
2742 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2743 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2746 MF_MSG_TRUNCATED_LRU,
2753 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2754 extern void clear_huge_page(struct page *page,
2755 unsigned long addr_hint,
2756 unsigned int pages_per_huge_page);
2757 extern void copy_user_huge_page(struct page *dst, struct page *src,
2758 unsigned long addr_hint,
2759 struct vm_area_struct *vma,
2760 unsigned int pages_per_huge_page);
2761 extern long copy_huge_page_from_user(struct page *dst_page,
2762 const void __user *usr_src,
2763 unsigned int pages_per_huge_page,
2764 bool allow_pagefault);
2765 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2767 extern struct page_ext_operations debug_guardpage_ops;
2769 #ifdef CONFIG_DEBUG_PAGEALLOC
2770 extern unsigned int _debug_guardpage_minorder;
2771 extern bool _debug_guardpage_enabled;
2773 static inline unsigned int debug_guardpage_minorder(void)
2775 return _debug_guardpage_minorder;
2778 static inline bool debug_guardpage_enabled(void)
2780 return _debug_guardpage_enabled;
2783 static inline bool page_is_guard(struct page *page)
2785 struct page_ext *page_ext;
2787 if (!debug_guardpage_enabled())
2790 page_ext = lookup_page_ext(page);
2791 if (unlikely(!page_ext))
2794 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2797 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2798 static inline bool debug_guardpage_enabled(void) { return false; }
2799 static inline bool page_is_guard(struct page *page) { return false; }
2800 #endif /* CONFIG_DEBUG_PAGEALLOC */
2802 #if MAX_NUMNODES > 1
2803 void __init setup_nr_node_ids(void);
2805 static inline void setup_nr_node_ids(void) {}
2808 #endif /* __KERNEL__ */
2809 #endif /* _LINUX_MM_H */