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>
32 struct anon_vma_chain;
35 struct writeback_control;
38 void init_mm_internals(void);
40 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
41 extern unsigned long max_mapnr;
43 static inline void set_max_mapnr(unsigned long limit)
48 static inline void set_max_mapnr(unsigned long limit) { }
51 extern unsigned long totalram_pages;
52 extern void * high_memory;
53 extern int page_cluster;
56 extern int sysctl_legacy_va_layout;
58 #define sysctl_legacy_va_layout 0
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
62 extern const int mmap_rnd_bits_min;
63 extern const int mmap_rnd_bits_max;
64 extern int mmap_rnd_bits __read_mostly;
66 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
67 extern const int mmap_rnd_compat_bits_min;
68 extern const int mmap_rnd_compat_bits_max;
69 extern int mmap_rnd_compat_bits __read_mostly;
73 #include <asm/pgtable.h>
74 #include <asm/processor.h>
77 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
81 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
85 #define lm_alias(x) __va(__pa_symbol(x))
89 * To prevent common memory management code establishing
90 * a zero page mapping on a read fault.
91 * This macro should be defined within <asm/pgtable.h>.
92 * s390 does this to prevent multiplexing of hardware bits
93 * related to the physical page in case of virtualization.
95 #ifndef mm_forbids_zeropage
96 #define mm_forbids_zeropage(X) (0)
100 * On some architectures it is expensive to call memset() for small sizes.
101 * Those architectures should provide their own implementation of "struct page"
102 * zeroing by defining this macro in <asm/pgtable.h>.
104 #ifndef mm_zero_struct_page
105 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
109 * Default maximum number of active map areas, this limits the number of vmas
110 * per mm struct. Users can overwrite this number by sysctl but there is a
113 * When a program's coredump is generated as ELF format, a section is created
114 * per a vma. In ELF, the number of sections is represented in unsigned short.
115 * This means the number of sections should be smaller than 65535 at coredump.
116 * Because the kernel adds some informative sections to a image of program at
117 * generating coredump, we need some margin. The number of extra sections is
118 * 1-3 now and depends on arch. We use "5" as safe margin, here.
120 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
121 * not a hard limit any more. Although some userspace tools can be surprised by
124 #define MAPCOUNT_ELF_CORE_MARGIN (5)
125 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
127 extern int sysctl_max_map_count;
129 extern unsigned long sysctl_user_reserve_kbytes;
130 extern unsigned long sysctl_admin_reserve_kbytes;
132 extern int sysctl_overcommit_memory;
133 extern int sysctl_overcommit_ratio;
134 extern unsigned long sysctl_overcommit_kbytes;
136 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
138 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
141 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
143 /* to align the pointer to the (next) page boundary */
144 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
146 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
147 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
150 * Linux kernel virtual memory manager primitives.
151 * The idea being to have a "virtual" mm in the same way
152 * we have a virtual fs - giving a cleaner interface to the
153 * mm details, and allowing different kinds of memory mappings
154 * (from shared memory to executable loading to arbitrary
158 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
159 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
160 void vm_area_free(struct vm_area_struct *);
163 extern struct rb_root nommu_region_tree;
164 extern struct rw_semaphore nommu_region_sem;
166 extern unsigned int kobjsize(const void *objp);
170 * vm_flags in vm_area_struct, see mm_types.h.
171 * When changing, update also include/trace/events/mmflags.h
173 #define VM_NONE 0x00000000
175 #define VM_READ 0x00000001 /* currently active flags */
176 #define VM_WRITE 0x00000002
177 #define VM_EXEC 0x00000004
178 #define VM_SHARED 0x00000008
180 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
181 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
182 #define VM_MAYWRITE 0x00000020
183 #define VM_MAYEXEC 0x00000040
184 #define VM_MAYSHARE 0x00000080
186 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
187 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
188 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
189 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
190 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
192 #define VM_LOCKED 0x00002000
193 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
195 /* Used by sys_madvise() */
196 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
197 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
199 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
200 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
201 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
202 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
203 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
204 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
205 #define VM_SYNC 0x00800000 /* Synchronous page faults */
206 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
207 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
208 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
210 #ifdef CONFIG_MEM_SOFT_DIRTY
211 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
213 # define VM_SOFTDIRTY 0
216 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
217 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
218 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
219 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
221 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
222 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
225 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
226 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
227 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
228 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
229 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
230 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
231 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
232 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
234 #ifdef CONFIG_ARCH_HAS_PKEYS
235 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
236 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
237 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
238 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
239 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
241 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
243 # define VM_PKEY_BIT4 0
245 #endif /* CONFIG_ARCH_HAS_PKEYS */
247 #if defined(CONFIG_X86)
248 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
249 #elif defined(CONFIG_PPC)
250 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
251 #elif defined(CONFIG_PARISC)
252 # define VM_GROWSUP VM_ARCH_1
253 #elif defined(CONFIG_IA64)
254 # define VM_GROWSUP VM_ARCH_1
255 #elif defined(CONFIG_SPARC64)
256 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
257 # define VM_ARCH_CLEAR VM_SPARC_ADI
258 #elif !defined(CONFIG_MMU)
259 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
262 #if defined(CONFIG_X86_INTEL_MPX)
263 /* MPX specific bounds table or bounds directory */
264 # define VM_MPX VM_HIGH_ARCH_4
266 # define VM_MPX VM_NONE
270 # define VM_GROWSUP VM_NONE
273 /* Bits set in the VMA until the stack is in its final location */
274 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
276 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
277 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
280 #ifdef CONFIG_STACK_GROWSUP
281 #define VM_STACK VM_GROWSUP
283 #define VM_STACK VM_GROWSDOWN
286 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
289 * Special vmas that are non-mergable, non-mlock()able.
290 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
292 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
294 /* This mask defines which mm->def_flags a process can inherit its parent */
295 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
297 /* This mask is used to clear all the VMA flags used by mlock */
298 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
300 /* Arch-specific flags to clear when updating VM flags on protection change */
301 #ifndef VM_ARCH_CLEAR
302 # define VM_ARCH_CLEAR VM_NONE
304 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
307 * mapping from the currently active vm_flags protection bits (the
308 * low four bits) to a page protection mask..
310 extern pgprot_t protection_map[16];
312 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
313 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
314 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
315 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
316 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
317 #define FAULT_FLAG_TRIED 0x20 /* Second try */
318 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
319 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
320 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
322 #define FAULT_FLAG_TRACE \
323 { FAULT_FLAG_WRITE, "WRITE" }, \
324 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
325 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
326 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
327 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
328 { FAULT_FLAG_TRIED, "TRIED" }, \
329 { FAULT_FLAG_USER, "USER" }, \
330 { FAULT_FLAG_REMOTE, "REMOTE" }, \
331 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
334 * vm_fault is filled by the the pagefault handler and passed to the vma's
335 * ->fault function. The vma's ->fault is responsible for returning a bitmask
336 * of VM_FAULT_xxx flags that give details about how the fault was handled.
338 * MM layer fills up gfp_mask for page allocations but fault handler might
339 * alter it if its implementation requires a different allocation context.
341 * pgoff should be used in favour of virtual_address, if possible.
344 struct vm_area_struct *vma; /* Target VMA */
345 unsigned int flags; /* FAULT_FLAG_xxx flags */
346 gfp_t gfp_mask; /* gfp mask to be used for allocations */
347 pgoff_t pgoff; /* Logical page offset based on vma */
348 unsigned long address; /* Faulting virtual address */
349 pmd_t *pmd; /* Pointer to pmd entry matching
351 pud_t *pud; /* Pointer to pud entry matching
354 pte_t orig_pte; /* Value of PTE at the time of fault */
356 struct page *cow_page; /* Page handler may use for COW fault */
357 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
358 struct page *page; /* ->fault handlers should return a
359 * page here, unless VM_FAULT_NOPAGE
360 * is set (which is also implied by
363 /* These three entries are valid only while holding ptl lock */
364 pte_t *pte; /* Pointer to pte entry matching
365 * the 'address'. NULL if the page
366 * table hasn't been allocated.
368 spinlock_t *ptl; /* Page table lock.
369 * Protects pte page table if 'pte'
370 * is not NULL, otherwise pmd.
372 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
373 * vm_ops->map_pages() calls
374 * alloc_set_pte() from atomic context.
375 * do_fault_around() pre-allocates
376 * page table to avoid allocation from
381 /* page entry size for vm->huge_fault() */
382 enum page_entry_size {
389 * These are the virtual MM functions - opening of an area, closing and
390 * unmapping it (needed to keep files on disk up-to-date etc), pointer
391 * to the functions called when a no-page or a wp-page exception occurs.
393 struct vm_operations_struct {
394 void (*open)(struct vm_area_struct * area);
395 void (*close)(struct vm_area_struct * area);
396 int (*split)(struct vm_area_struct * area, unsigned long addr);
397 int (*mremap)(struct vm_area_struct * area);
398 vm_fault_t (*fault)(struct vm_fault *vmf);
399 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
400 enum page_entry_size pe_size);
401 void (*map_pages)(struct vm_fault *vmf,
402 pgoff_t start_pgoff, pgoff_t end_pgoff);
403 unsigned long (*pagesize)(struct vm_area_struct * area);
405 /* notification that a previously read-only page is about to become
406 * writable, if an error is returned it will cause a SIGBUS */
407 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
409 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
410 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
412 /* called by access_process_vm when get_user_pages() fails, typically
413 * for use by special VMAs that can switch between memory and hardware
415 int (*access)(struct vm_area_struct *vma, unsigned long addr,
416 void *buf, int len, int write);
418 /* Called by the /proc/PID/maps code to ask the vma whether it
419 * has a special name. Returning non-NULL will also cause this
420 * vma to be dumped unconditionally. */
421 const char *(*name)(struct vm_area_struct *vma);
425 * set_policy() op must add a reference to any non-NULL @new mempolicy
426 * to hold the policy upon return. Caller should pass NULL @new to
427 * remove a policy and fall back to surrounding context--i.e. do not
428 * install a MPOL_DEFAULT policy, nor the task or system default
431 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
434 * get_policy() op must add reference [mpol_get()] to any policy at
435 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
436 * in mm/mempolicy.c will do this automatically.
437 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
438 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
439 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
440 * must return NULL--i.e., do not "fallback" to task or system default
443 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
447 * Called by vm_normal_page() for special PTEs to find the
448 * page for @addr. This is useful if the default behavior
449 * (using pte_page()) would not find the correct page.
451 struct page *(*find_special_page)(struct vm_area_struct *vma,
455 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
457 static const struct vm_operations_struct dummy_vm_ops = {};
460 vma->vm_ops = &dummy_vm_ops;
461 INIT_LIST_HEAD(&vma->anon_vma_chain);
464 static inline void vma_set_anonymous(struct vm_area_struct *vma)
472 #define page_private(page) ((page)->private)
473 #define set_page_private(page, v) ((page)->private = (v))
475 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
476 static inline int pmd_devmap(pmd_t pmd)
480 static inline int pud_devmap(pud_t pud)
484 static inline int pgd_devmap(pgd_t pgd)
491 * FIXME: take this include out, include page-flags.h in
492 * files which need it (119 of them)
494 #include <linux/page-flags.h>
495 #include <linux/huge_mm.h>
498 * Methods to modify the page usage count.
500 * What counts for a page usage:
501 * - cache mapping (page->mapping)
502 * - private data (page->private)
503 * - page mapped in a task's page tables, each mapping
504 * is counted separately
506 * Also, many kernel routines increase the page count before a critical
507 * routine so they can be sure the page doesn't go away from under them.
511 * Drop a ref, return true if the refcount fell to zero (the page has no users)
513 static inline int put_page_testzero(struct page *page)
515 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
516 return page_ref_dec_and_test(page);
520 * Try to grab a ref unless the page has a refcount of zero, return false if
522 * This can be called when MMU is off so it must not access
523 * any of the virtual mappings.
525 static inline int get_page_unless_zero(struct page *page)
527 return page_ref_add_unless(page, 1, 0);
530 extern int page_is_ram(unsigned long pfn);
538 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
541 /* Support for virtually mapped pages */
542 struct page *vmalloc_to_page(const void *addr);
543 unsigned long vmalloc_to_pfn(const void *addr);
546 * Determine if an address is within the vmalloc range
548 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
549 * is no special casing required.
551 static inline bool is_vmalloc_addr(const void *x)
554 unsigned long addr = (unsigned long)x;
556 return addr >= VMALLOC_START && addr < VMALLOC_END;
562 extern int is_vmalloc_or_module_addr(const void *x);
564 static inline int is_vmalloc_or_module_addr(const void *x)
570 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
571 static inline void *kvmalloc(size_t size, gfp_t flags)
573 return kvmalloc_node(size, flags, NUMA_NO_NODE);
575 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
577 return kvmalloc_node(size, flags | __GFP_ZERO, node);
579 static inline void *kvzalloc(size_t size, gfp_t flags)
581 return kvmalloc(size, flags | __GFP_ZERO);
584 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
588 if (unlikely(check_mul_overflow(n, size, &bytes)))
591 return kvmalloc(bytes, flags);
594 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
596 return kvmalloc_array(n, size, flags | __GFP_ZERO);
599 extern void kvfree(const void *addr);
601 static inline atomic_t *compound_mapcount_ptr(struct page *page)
603 return &page[1].compound_mapcount;
606 static inline int compound_mapcount(struct page *page)
608 VM_BUG_ON_PAGE(!PageCompound(page), page);
609 page = compound_head(page);
610 return atomic_read(compound_mapcount_ptr(page)) + 1;
614 * The atomic page->_mapcount, starts from -1: so that transitions
615 * both from it and to it can be tracked, using atomic_inc_and_test
616 * and atomic_add_negative(-1).
618 static inline void page_mapcount_reset(struct page *page)
620 atomic_set(&(page)->_mapcount, -1);
623 int __page_mapcount(struct page *page);
625 static inline int page_mapcount(struct page *page)
627 VM_BUG_ON_PAGE(PageSlab(page), page);
629 if (unlikely(PageCompound(page)))
630 return __page_mapcount(page);
631 return atomic_read(&page->_mapcount) + 1;
634 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
635 int total_mapcount(struct page *page);
636 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
638 static inline int total_mapcount(struct page *page)
640 return page_mapcount(page);
642 static inline int page_trans_huge_mapcount(struct page *page,
645 int mapcount = page_mapcount(page);
647 *total_mapcount = mapcount;
652 static inline struct page *virt_to_head_page(const void *x)
654 struct page *page = virt_to_page(x);
656 return compound_head(page);
659 void __put_page(struct page *page);
661 void put_pages_list(struct list_head *pages);
663 void split_page(struct page *page, unsigned int order);
666 * Compound pages have a destructor function. Provide a
667 * prototype for that function and accessor functions.
668 * These are _only_ valid on the head of a compound page.
670 typedef void compound_page_dtor(struct page *);
672 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
673 enum compound_dtor_id {
676 #ifdef CONFIG_HUGETLB_PAGE
679 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
684 extern compound_page_dtor * const compound_page_dtors[];
686 static inline void set_compound_page_dtor(struct page *page,
687 enum compound_dtor_id compound_dtor)
689 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
690 page[1].compound_dtor = compound_dtor;
693 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
695 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
696 return compound_page_dtors[page[1].compound_dtor];
699 static inline unsigned int compound_order(struct page *page)
703 return page[1].compound_order;
706 static inline void set_compound_order(struct page *page, unsigned int order)
708 page[1].compound_order = order;
711 void free_compound_page(struct page *page);
715 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
716 * servicing faults for write access. In the normal case, do always want
717 * pte_mkwrite. But get_user_pages can cause write faults for mappings
718 * that do not have writing enabled, when used by access_process_vm.
720 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
722 if (likely(vma->vm_flags & VM_WRITE))
723 pte = pte_mkwrite(pte);
727 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
729 int finish_fault(struct vm_fault *vmf);
730 int finish_mkwrite_fault(struct vm_fault *vmf);
734 * Multiple processes may "see" the same page. E.g. for untouched
735 * mappings of /dev/null, all processes see the same page full of
736 * zeroes, and text pages of executables and shared libraries have
737 * only one copy in memory, at most, normally.
739 * For the non-reserved pages, page_count(page) denotes a reference count.
740 * page_count() == 0 means the page is free. page->lru is then used for
741 * freelist management in the buddy allocator.
742 * page_count() > 0 means the page has been allocated.
744 * Pages are allocated by the slab allocator in order to provide memory
745 * to kmalloc and kmem_cache_alloc. In this case, the management of the
746 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
747 * unless a particular usage is carefully commented. (the responsibility of
748 * freeing the kmalloc memory is the caller's, of course).
750 * A page may be used by anyone else who does a __get_free_page().
751 * In this case, page_count still tracks the references, and should only
752 * be used through the normal accessor functions. The top bits of page->flags
753 * and page->virtual store page management information, but all other fields
754 * are unused and could be used privately, carefully. The management of this
755 * page is the responsibility of the one who allocated it, and those who have
756 * subsequently been given references to it.
758 * The other pages (we may call them "pagecache pages") are completely
759 * managed by the Linux memory manager: I/O, buffers, swapping etc.
760 * The following discussion applies only to them.
762 * A pagecache page contains an opaque `private' member, which belongs to the
763 * page's address_space. Usually, this is the address of a circular list of
764 * the page's disk buffers. PG_private must be set to tell the VM to call
765 * into the filesystem to release these pages.
767 * A page may belong to an inode's memory mapping. In this case, page->mapping
768 * is the pointer to the inode, and page->index is the file offset of the page,
769 * in units of PAGE_SIZE.
771 * If pagecache pages are not associated with an inode, they are said to be
772 * anonymous pages. These may become associated with the swapcache, and in that
773 * case PG_swapcache is set, and page->private is an offset into the swapcache.
775 * In either case (swapcache or inode backed), the pagecache itself holds one
776 * reference to the page. Setting PG_private should also increment the
777 * refcount. The each user mapping also has a reference to the page.
779 * The pagecache pages are stored in a per-mapping radix tree, which is
780 * rooted at mapping->i_pages, and indexed by offset.
781 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
782 * lists, we instead now tag pages as dirty/writeback in the radix tree.
784 * All pagecache pages may be subject to I/O:
785 * - inode pages may need to be read from disk,
786 * - inode pages which have been modified and are MAP_SHARED may need
787 * to be written back to the inode on disk,
788 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
789 * modified may need to be swapped out to swap space and (later) to be read
794 * The zone field is never updated after free_area_init_core()
795 * sets it, so none of the operations on it need to be atomic.
798 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
799 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
800 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
801 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
802 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
805 * Define the bit shifts to access each section. For non-existent
806 * sections we define the shift as 0; that plus a 0 mask ensures
807 * the compiler will optimise away reference to them.
809 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
810 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
811 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
812 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
814 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
815 #ifdef NODE_NOT_IN_PAGE_FLAGS
816 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
817 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
818 SECTIONS_PGOFF : ZONES_PGOFF)
820 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
821 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
822 NODES_PGOFF : ZONES_PGOFF)
825 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
827 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
828 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
831 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
832 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
833 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
834 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
835 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
837 static inline enum zone_type page_zonenum(const struct page *page)
839 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
842 #ifdef CONFIG_ZONE_DEVICE
843 static inline bool is_zone_device_page(const struct page *page)
845 return page_zonenum(page) == ZONE_DEVICE;
848 static inline bool is_zone_device_page(const struct page *page)
854 #ifdef CONFIG_DEV_PAGEMAP_OPS
855 void dev_pagemap_get_ops(void);
856 void dev_pagemap_put_ops(void);
857 void __put_devmap_managed_page(struct page *page);
858 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
859 static inline bool put_devmap_managed_page(struct page *page)
861 if (!static_branch_unlikely(&devmap_managed_key))
863 if (!is_zone_device_page(page))
865 switch (page->pgmap->type) {
866 case MEMORY_DEVICE_PRIVATE:
867 case MEMORY_DEVICE_PUBLIC:
868 case MEMORY_DEVICE_FS_DAX:
869 __put_devmap_managed_page(page);
877 static inline bool is_device_private_page(const struct page *page)
879 return is_zone_device_page(page) &&
880 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
883 static inline bool is_device_public_page(const struct page *page)
885 return is_zone_device_page(page) &&
886 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
889 #else /* CONFIG_DEV_PAGEMAP_OPS */
890 static inline void dev_pagemap_get_ops(void)
894 static inline void dev_pagemap_put_ops(void)
898 static inline bool put_devmap_managed_page(struct page *page)
903 static inline bool is_device_private_page(const struct page *page)
908 static inline bool is_device_public_page(const struct page *page)
912 #endif /* CONFIG_DEV_PAGEMAP_OPS */
914 static inline void get_page(struct page *page)
916 page = compound_head(page);
918 * Getting a normal page or the head of a compound page
919 * requires to already have an elevated page->_refcount.
921 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
925 static inline void put_page(struct page *page)
927 page = compound_head(page);
930 * For devmap managed pages we need to catch refcount transition from
931 * 2 to 1, when refcount reach one it means the page is free and we
932 * need to inform the device driver through callback. See
933 * include/linux/memremap.h and HMM for details.
935 if (put_devmap_managed_page(page))
938 if (put_page_testzero(page))
942 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
943 #define SECTION_IN_PAGE_FLAGS
947 * The identification function is mainly used by the buddy allocator for
948 * determining if two pages could be buddies. We are not really identifying
949 * the zone since we could be using the section number id if we do not have
950 * node id available in page flags.
951 * We only guarantee that it will return the same value for two combinable
954 static inline int page_zone_id(struct page *page)
956 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
959 static inline int zone_to_nid(struct zone *zone)
968 #ifdef NODE_NOT_IN_PAGE_FLAGS
969 extern int page_to_nid(const struct page *page);
971 static inline int page_to_nid(const struct page *page)
973 struct page *p = (struct page *)page;
975 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
979 #ifdef CONFIG_NUMA_BALANCING
980 static inline int cpu_pid_to_cpupid(int cpu, int pid)
982 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
985 static inline int cpupid_to_pid(int cpupid)
987 return cpupid & LAST__PID_MASK;
990 static inline int cpupid_to_cpu(int cpupid)
992 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
995 static inline int cpupid_to_nid(int cpupid)
997 return cpu_to_node(cpupid_to_cpu(cpupid));
1000 static inline bool cpupid_pid_unset(int cpupid)
1002 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1005 static inline bool cpupid_cpu_unset(int cpupid)
1007 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1010 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1012 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1015 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1016 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1017 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1019 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1022 static inline int page_cpupid_last(struct page *page)
1024 return page->_last_cpupid;
1026 static inline void page_cpupid_reset_last(struct page *page)
1028 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1031 static inline int page_cpupid_last(struct page *page)
1033 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1036 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1038 static inline void page_cpupid_reset_last(struct page *page)
1040 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1042 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1043 #else /* !CONFIG_NUMA_BALANCING */
1044 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1046 return page_to_nid(page); /* XXX */
1049 static inline int page_cpupid_last(struct page *page)
1051 return page_to_nid(page); /* XXX */
1054 static inline int cpupid_to_nid(int cpupid)
1059 static inline int cpupid_to_pid(int cpupid)
1064 static inline int cpupid_to_cpu(int cpupid)
1069 static inline int cpu_pid_to_cpupid(int nid, int pid)
1074 static inline bool cpupid_pid_unset(int cpupid)
1079 static inline void page_cpupid_reset_last(struct page *page)
1083 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1087 #endif /* CONFIG_NUMA_BALANCING */
1089 static inline struct zone *page_zone(const struct page *page)
1091 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1094 static inline pg_data_t *page_pgdat(const struct page *page)
1096 return NODE_DATA(page_to_nid(page));
1099 #ifdef SECTION_IN_PAGE_FLAGS
1100 static inline void set_page_section(struct page *page, unsigned long section)
1102 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1103 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1106 static inline unsigned long page_to_section(const struct page *page)
1108 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1112 static inline void set_page_zone(struct page *page, enum zone_type zone)
1114 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1115 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1118 static inline void set_page_node(struct page *page, unsigned long node)
1120 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1121 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1124 static inline void set_page_links(struct page *page, enum zone_type zone,
1125 unsigned long node, unsigned long pfn)
1127 set_page_zone(page, zone);
1128 set_page_node(page, node);
1129 #ifdef SECTION_IN_PAGE_FLAGS
1130 set_page_section(page, pfn_to_section_nr(pfn));
1135 static inline struct mem_cgroup *page_memcg(struct page *page)
1137 return page->mem_cgroup;
1139 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1141 WARN_ON_ONCE(!rcu_read_lock_held());
1142 return READ_ONCE(page->mem_cgroup);
1145 static inline struct mem_cgroup *page_memcg(struct page *page)
1149 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1151 WARN_ON_ONCE(!rcu_read_lock_held());
1157 * Some inline functions in vmstat.h depend on page_zone()
1159 #include <linux/vmstat.h>
1161 static __always_inline void *lowmem_page_address(const struct page *page)
1163 return page_to_virt(page);
1166 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1167 #define HASHED_PAGE_VIRTUAL
1170 #if defined(WANT_PAGE_VIRTUAL)
1171 static inline void *page_address(const struct page *page)
1173 return page->virtual;
1175 static inline void set_page_address(struct page *page, void *address)
1177 page->virtual = address;
1179 #define page_address_init() do { } while(0)
1182 #if defined(HASHED_PAGE_VIRTUAL)
1183 void *page_address(const struct page *page);
1184 void set_page_address(struct page *page, void *virtual);
1185 void page_address_init(void);
1188 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1189 #define page_address(page) lowmem_page_address(page)
1190 #define set_page_address(page, address) do { } while(0)
1191 #define page_address_init() do { } while(0)
1194 extern void *page_rmapping(struct page *page);
1195 extern struct anon_vma *page_anon_vma(struct page *page);
1196 extern struct address_space *page_mapping(struct page *page);
1198 extern struct address_space *__page_file_mapping(struct page *);
1201 struct address_space *page_file_mapping(struct page *page)
1203 if (unlikely(PageSwapCache(page)))
1204 return __page_file_mapping(page);
1206 return page->mapping;
1209 extern pgoff_t __page_file_index(struct page *page);
1212 * Return the pagecache index of the passed page. Regular pagecache pages
1213 * use ->index whereas swapcache pages use swp_offset(->private)
1215 static inline pgoff_t page_index(struct page *page)
1217 if (unlikely(PageSwapCache(page)))
1218 return __page_file_index(page);
1222 bool page_mapped(struct page *page);
1223 struct address_space *page_mapping(struct page *page);
1224 struct address_space *page_mapping_file(struct page *page);
1227 * Return true only if the page has been allocated with
1228 * ALLOC_NO_WATERMARKS and the low watermark was not
1229 * met implying that the system is under some pressure.
1231 static inline bool page_is_pfmemalloc(struct page *page)
1234 * Page index cannot be this large so this must be
1235 * a pfmemalloc page.
1237 return page->index == -1UL;
1241 * Only to be called by the page allocator on a freshly allocated
1244 static inline void set_page_pfmemalloc(struct page *page)
1249 static inline void clear_page_pfmemalloc(struct page *page)
1255 * Different kinds of faults, as returned by handle_mm_fault().
1256 * Used to decide whether a process gets delivered SIGBUS or
1257 * just gets major/minor fault counters bumped up.
1260 #define VM_FAULT_OOM 0x0001
1261 #define VM_FAULT_SIGBUS 0x0002
1262 #define VM_FAULT_MAJOR 0x0004
1263 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1264 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1265 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1266 #define VM_FAULT_SIGSEGV 0x0040
1268 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1269 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1270 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1271 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1272 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1273 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1274 * and needs fsync() to complete (for
1275 * synchronous page faults in DAX) */
1277 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1278 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1281 #define VM_FAULT_RESULT_TRACE \
1282 { VM_FAULT_OOM, "OOM" }, \
1283 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1284 { VM_FAULT_MAJOR, "MAJOR" }, \
1285 { VM_FAULT_WRITE, "WRITE" }, \
1286 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1287 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1288 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1289 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1290 { VM_FAULT_LOCKED, "LOCKED" }, \
1291 { VM_FAULT_RETRY, "RETRY" }, \
1292 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1293 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1294 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1296 /* Encode hstate index for a hwpoisoned large page */
1297 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1298 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1301 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1303 extern void pagefault_out_of_memory(void);
1305 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1308 * Flags passed to show_mem() and show_free_areas() to suppress output in
1311 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1313 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1315 extern bool can_do_mlock(void);
1316 extern int user_shm_lock(size_t, struct user_struct *);
1317 extern void user_shm_unlock(size_t, struct user_struct *);
1320 * Parameter block passed down to zap_pte_range in exceptional cases.
1322 struct zap_details {
1323 struct address_space *check_mapping; /* Check page->mapping if set */
1324 pgoff_t first_index; /* Lowest page->index to unmap */
1325 pgoff_t last_index; /* Highest page->index to unmap */
1328 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1329 pte_t pte, bool with_public_device);
1330 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1332 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1335 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1336 unsigned long size);
1337 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1338 unsigned long size);
1339 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1340 unsigned long start, unsigned long end);
1343 * mm_walk - callbacks for walk_page_range
1344 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1345 * this handler should only handle pud_trans_huge() puds.
1346 * the pmd_entry or pte_entry callbacks will be used for
1348 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1349 * this handler is required to be able to handle
1350 * pmd_trans_huge() pmds. They may simply choose to
1351 * split_huge_page() instead of handling it explicitly.
1352 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1353 * @pte_hole: if set, called for each hole at all levels
1354 * @hugetlb_entry: if set, called for each hugetlb entry
1355 * @test_walk: caller specific callback function to determine whether
1356 * we walk over the current vma or not. Returning 0
1357 * value means "do page table walk over the current vma,"
1358 * and a negative one means "abort current page table walk
1359 * right now." 1 means "skip the current vma."
1360 * @mm: mm_struct representing the target process of page table walk
1361 * @vma: vma currently walked (NULL if walking outside vmas)
1362 * @private: private data for callbacks' usage
1364 * (see the comment on walk_page_range() for more details)
1367 int (*pud_entry)(pud_t *pud, unsigned long addr,
1368 unsigned long next, struct mm_walk *walk);
1369 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1370 unsigned long next, struct mm_walk *walk);
1371 int (*pte_entry)(pte_t *pte, unsigned long addr,
1372 unsigned long next, struct mm_walk *walk);
1373 int (*pte_hole)(unsigned long addr, unsigned long next,
1374 struct mm_walk *walk);
1375 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1376 unsigned long addr, unsigned long next,
1377 struct mm_walk *walk);
1378 int (*test_walk)(unsigned long addr, unsigned long next,
1379 struct mm_walk *walk);
1380 struct mm_struct *mm;
1381 struct vm_area_struct *vma;
1385 int walk_page_range(unsigned long addr, unsigned long end,
1386 struct mm_walk *walk);
1387 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1388 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1389 unsigned long end, unsigned long floor, unsigned long ceiling);
1390 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1391 struct vm_area_struct *vma);
1392 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1393 unsigned long *start, unsigned long *end,
1394 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1395 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1396 unsigned long *pfn);
1397 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1398 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1399 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1400 void *buf, int len, int write);
1402 extern void truncate_pagecache(struct inode *inode, loff_t new);
1403 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1404 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1405 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1406 int truncate_inode_page(struct address_space *mapping, struct page *page);
1407 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1408 int invalidate_inode_page(struct page *page);
1411 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1412 unsigned int flags);
1413 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1414 unsigned long address, unsigned int fault_flags,
1416 void unmap_mapping_pages(struct address_space *mapping,
1417 pgoff_t start, pgoff_t nr, bool even_cows);
1418 void unmap_mapping_range(struct address_space *mapping,
1419 loff_t const holebegin, loff_t const holelen, int even_cows);
1421 static inline int handle_mm_fault(struct vm_area_struct *vma,
1422 unsigned long address, unsigned int flags)
1424 /* should never happen if there's no MMU */
1426 return VM_FAULT_SIGBUS;
1428 static inline int fixup_user_fault(struct task_struct *tsk,
1429 struct mm_struct *mm, unsigned long address,
1430 unsigned int fault_flags, bool *unlocked)
1432 /* should never happen if there's no MMU */
1436 static inline void unmap_mapping_pages(struct address_space *mapping,
1437 pgoff_t start, pgoff_t nr, bool even_cows) { }
1438 static inline void unmap_mapping_range(struct address_space *mapping,
1439 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1442 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1443 loff_t const holebegin, loff_t const holelen)
1445 unmap_mapping_range(mapping, holebegin, holelen, 0);
1448 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1449 void *buf, int len, unsigned int gup_flags);
1450 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1451 void *buf, int len, unsigned int gup_flags);
1452 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1453 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1455 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1456 unsigned long start, unsigned long nr_pages,
1457 unsigned int gup_flags, struct page **pages,
1458 struct vm_area_struct **vmas, int *locked);
1459 long get_user_pages(unsigned long start, unsigned long nr_pages,
1460 unsigned int gup_flags, struct page **pages,
1461 struct vm_area_struct **vmas);
1462 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1463 unsigned int gup_flags, struct page **pages, int *locked);
1464 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1465 struct page **pages, unsigned int gup_flags);
1466 #ifdef CONFIG_FS_DAX
1467 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1468 unsigned int gup_flags, struct page **pages,
1469 struct vm_area_struct **vmas);
1471 static inline long get_user_pages_longterm(unsigned long start,
1472 unsigned long nr_pages, unsigned int gup_flags,
1473 struct page **pages, struct vm_area_struct **vmas)
1475 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1477 #endif /* CONFIG_FS_DAX */
1479 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1480 struct page **pages);
1482 /* Container for pinned pfns / pages */
1483 struct frame_vector {
1484 unsigned int nr_allocated; /* Number of frames we have space for */
1485 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1486 bool got_ref; /* Did we pin pages by getting page ref? */
1487 bool is_pfns; /* Does array contain pages or pfns? */
1488 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1489 * pfns_vector_pages() or pfns_vector_pfns()
1493 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1494 void frame_vector_destroy(struct frame_vector *vec);
1495 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1496 unsigned int gup_flags, struct frame_vector *vec);
1497 void put_vaddr_frames(struct frame_vector *vec);
1498 int frame_vector_to_pages(struct frame_vector *vec);
1499 void frame_vector_to_pfns(struct frame_vector *vec);
1501 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1503 return vec->nr_frames;
1506 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1509 int err = frame_vector_to_pages(vec);
1512 return ERR_PTR(err);
1514 return (struct page **)(vec->ptrs);
1517 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1520 frame_vector_to_pfns(vec);
1521 return (unsigned long *)(vec->ptrs);
1525 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1526 struct page **pages);
1527 int get_kernel_page(unsigned long start, int write, struct page **pages);
1528 struct page *get_dump_page(unsigned long addr);
1530 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1531 extern void do_invalidatepage(struct page *page, unsigned int offset,
1532 unsigned int length);
1534 void __set_page_dirty(struct page *, struct address_space *, int warn);
1535 int __set_page_dirty_nobuffers(struct page *page);
1536 int __set_page_dirty_no_writeback(struct page *page);
1537 int redirty_page_for_writepage(struct writeback_control *wbc,
1539 void account_page_dirtied(struct page *page, struct address_space *mapping);
1540 void account_page_cleaned(struct page *page, struct address_space *mapping,
1541 struct bdi_writeback *wb);
1542 int set_page_dirty(struct page *page);
1543 int set_page_dirty_lock(struct page *page);
1544 void __cancel_dirty_page(struct page *page);
1545 static inline void cancel_dirty_page(struct page *page)
1547 /* Avoid atomic ops, locking, etc. when not actually needed. */
1548 if (PageDirty(page))
1549 __cancel_dirty_page(page);
1551 int clear_page_dirty_for_io(struct page *page);
1553 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1555 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1557 return !vma->vm_ops;
1562 * The vma_is_shmem is not inline because it is used only by slow
1563 * paths in userfault.
1565 bool vma_is_shmem(struct vm_area_struct *vma);
1567 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1570 int vma_is_stack_for_current(struct vm_area_struct *vma);
1572 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1573 unsigned long old_addr, struct vm_area_struct *new_vma,
1574 unsigned long new_addr, unsigned long len,
1575 bool need_rmap_locks);
1576 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1577 unsigned long end, pgprot_t newprot,
1578 int dirty_accountable, int prot_numa);
1579 extern int mprotect_fixup(struct vm_area_struct *vma,
1580 struct vm_area_struct **pprev, unsigned long start,
1581 unsigned long end, unsigned long newflags);
1584 * doesn't attempt to fault and will return short.
1586 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1587 struct page **pages);
1589 * per-process(per-mm_struct) statistics.
1591 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1593 long val = atomic_long_read(&mm->rss_stat.count[member]);
1595 #ifdef SPLIT_RSS_COUNTING
1597 * counter is updated in asynchronous manner and may go to minus.
1598 * But it's never be expected number for users.
1603 return (unsigned long)val;
1606 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1608 atomic_long_add(value, &mm->rss_stat.count[member]);
1611 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1613 atomic_long_inc(&mm->rss_stat.count[member]);
1616 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1618 atomic_long_dec(&mm->rss_stat.count[member]);
1621 /* Optimized variant when page is already known not to be PageAnon */
1622 static inline int mm_counter_file(struct page *page)
1624 if (PageSwapBacked(page))
1625 return MM_SHMEMPAGES;
1626 return MM_FILEPAGES;
1629 static inline int mm_counter(struct page *page)
1632 return MM_ANONPAGES;
1633 return mm_counter_file(page);
1636 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1638 return get_mm_counter(mm, MM_FILEPAGES) +
1639 get_mm_counter(mm, MM_ANONPAGES) +
1640 get_mm_counter(mm, MM_SHMEMPAGES);
1643 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1645 return max(mm->hiwater_rss, get_mm_rss(mm));
1648 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1650 return max(mm->hiwater_vm, mm->total_vm);
1653 static inline void update_hiwater_rss(struct mm_struct *mm)
1655 unsigned long _rss = get_mm_rss(mm);
1657 if ((mm)->hiwater_rss < _rss)
1658 (mm)->hiwater_rss = _rss;
1661 static inline void update_hiwater_vm(struct mm_struct *mm)
1663 if (mm->hiwater_vm < mm->total_vm)
1664 mm->hiwater_vm = mm->total_vm;
1667 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1669 mm->hiwater_rss = get_mm_rss(mm);
1672 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1673 struct mm_struct *mm)
1675 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1677 if (*maxrss < hiwater_rss)
1678 *maxrss = hiwater_rss;
1681 #if defined(SPLIT_RSS_COUNTING)
1682 void sync_mm_rss(struct mm_struct *mm);
1684 static inline void sync_mm_rss(struct mm_struct *mm)
1689 #ifndef __HAVE_ARCH_PTE_DEVMAP
1690 static inline int pte_devmap(pte_t pte)
1696 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1698 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1700 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1704 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1708 #ifdef __PAGETABLE_P4D_FOLDED
1709 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1710 unsigned long address)
1715 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1718 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1719 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1720 unsigned long address)
1724 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1725 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1728 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1730 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1732 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1735 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1737 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1741 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1742 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1743 unsigned long address)
1748 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1749 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1752 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1754 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1756 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1759 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1761 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1766 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1768 atomic_long_set(&mm->pgtables_bytes, 0);
1771 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1773 return atomic_long_read(&mm->pgtables_bytes);
1776 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1778 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1781 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1783 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1787 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1788 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1793 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1794 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1797 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1798 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1801 * The following ifdef needed to get the 4level-fixup.h header to work.
1802 * Remove it when 4level-fixup.h has been removed.
1804 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1806 #ifndef __ARCH_HAS_5LEVEL_HACK
1807 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1808 unsigned long address)
1810 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1811 NULL : p4d_offset(pgd, address);
1814 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1815 unsigned long address)
1817 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1818 NULL : pud_offset(p4d, address);
1820 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1822 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1824 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1825 NULL: pmd_offset(pud, address);
1827 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1829 #if USE_SPLIT_PTE_PTLOCKS
1830 #if ALLOC_SPLIT_PTLOCKS
1831 void __init ptlock_cache_init(void);
1832 extern bool ptlock_alloc(struct page *page);
1833 extern void ptlock_free(struct page *page);
1835 static inline spinlock_t *ptlock_ptr(struct page *page)
1839 #else /* ALLOC_SPLIT_PTLOCKS */
1840 static inline void ptlock_cache_init(void)
1844 static inline bool ptlock_alloc(struct page *page)
1849 static inline void ptlock_free(struct page *page)
1853 static inline spinlock_t *ptlock_ptr(struct page *page)
1857 #endif /* ALLOC_SPLIT_PTLOCKS */
1859 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1861 return ptlock_ptr(pmd_page(*pmd));
1864 static inline bool ptlock_init(struct page *page)
1867 * prep_new_page() initialize page->private (and therefore page->ptl)
1868 * with 0. Make sure nobody took it in use in between.
1870 * It can happen if arch try to use slab for page table allocation:
1871 * slab code uses page->slab_cache, which share storage with page->ptl.
1873 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1874 if (!ptlock_alloc(page))
1876 spin_lock_init(ptlock_ptr(page));
1880 /* Reset page->mapping so free_pages_check won't complain. */
1881 static inline void pte_lock_deinit(struct page *page)
1883 page->mapping = NULL;
1887 #else /* !USE_SPLIT_PTE_PTLOCKS */
1889 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1891 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1893 return &mm->page_table_lock;
1895 static inline void ptlock_cache_init(void) {}
1896 static inline bool ptlock_init(struct page *page) { return true; }
1897 static inline void pte_lock_deinit(struct page *page) {}
1898 #endif /* USE_SPLIT_PTE_PTLOCKS */
1900 static inline void pgtable_init(void)
1902 ptlock_cache_init();
1903 pgtable_cache_init();
1906 static inline bool pgtable_page_ctor(struct page *page)
1908 if (!ptlock_init(page))
1910 __SetPageTable(page);
1911 inc_zone_page_state(page, NR_PAGETABLE);
1915 static inline void pgtable_page_dtor(struct page *page)
1917 pte_lock_deinit(page);
1918 __ClearPageTable(page);
1919 dec_zone_page_state(page, NR_PAGETABLE);
1922 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1924 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1925 pte_t *__pte = pte_offset_map(pmd, address); \
1931 #define pte_unmap_unlock(pte, ptl) do { \
1936 #define pte_alloc(mm, pmd, address) \
1937 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1939 #define pte_alloc_map(mm, pmd, address) \
1940 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1942 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1943 (pte_alloc(mm, pmd, address) ? \
1944 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1946 #define pte_alloc_kernel(pmd, address) \
1947 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1948 NULL: pte_offset_kernel(pmd, address))
1950 #if USE_SPLIT_PMD_PTLOCKS
1952 static struct page *pmd_to_page(pmd_t *pmd)
1954 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1955 return virt_to_page((void *)((unsigned long) pmd & mask));
1958 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1960 return ptlock_ptr(pmd_to_page(pmd));
1963 static inline bool pgtable_pmd_page_ctor(struct page *page)
1965 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1966 page->pmd_huge_pte = NULL;
1968 return ptlock_init(page);
1971 static inline void pgtable_pmd_page_dtor(struct page *page)
1973 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1974 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1979 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1983 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1985 return &mm->page_table_lock;
1988 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1989 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1991 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1995 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1997 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2003 * No scalability reason to split PUD locks yet, but follow the same pattern
2004 * as the PMD locks to make it easier if we decide to. The VM should not be
2005 * considered ready to switch to split PUD locks yet; there may be places
2006 * which need to be converted from page_table_lock.
2008 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2010 return &mm->page_table_lock;
2013 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2015 spinlock_t *ptl = pud_lockptr(mm, pud);
2021 extern void __init pagecache_init(void);
2022 extern void free_area_init(unsigned long * zones_size);
2023 extern void free_area_init_node(int nid, unsigned long * zones_size,
2024 unsigned long zone_start_pfn, unsigned long *zholes_size);
2025 extern void free_initmem(void);
2028 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2029 * into the buddy system. The freed pages will be poisoned with pattern
2030 * "poison" if it's within range [0, UCHAR_MAX].
2031 * Return pages freed into the buddy system.
2033 extern unsigned long free_reserved_area(void *start, void *end,
2034 int poison, char *s);
2036 #ifdef CONFIG_HIGHMEM
2038 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2039 * and totalram_pages.
2041 extern void free_highmem_page(struct page *page);
2044 extern void adjust_managed_page_count(struct page *page, long count);
2045 extern void mem_init_print_info(const char *str);
2047 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2049 /* Free the reserved page into the buddy system, so it gets managed. */
2050 static inline void __free_reserved_page(struct page *page)
2052 ClearPageReserved(page);
2053 init_page_count(page);
2057 static inline void free_reserved_page(struct page *page)
2059 __free_reserved_page(page);
2060 adjust_managed_page_count(page, 1);
2063 static inline void mark_page_reserved(struct page *page)
2065 SetPageReserved(page);
2066 adjust_managed_page_count(page, -1);
2070 * Default method to free all the __init memory into the buddy system.
2071 * The freed pages will be poisoned with pattern "poison" if it's within
2072 * range [0, UCHAR_MAX].
2073 * Return pages freed into the buddy system.
2075 static inline unsigned long free_initmem_default(int poison)
2077 extern char __init_begin[], __init_end[];
2079 return free_reserved_area(&__init_begin, &__init_end,
2080 poison, "unused kernel");
2083 static inline unsigned long get_num_physpages(void)
2086 unsigned long phys_pages = 0;
2088 for_each_online_node(nid)
2089 phys_pages += node_present_pages(nid);
2094 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2096 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2097 * zones, allocate the backing mem_map and account for memory holes in a more
2098 * architecture independent manner. This is a substitute for creating the
2099 * zone_sizes[] and zholes_size[] arrays and passing them to
2100 * free_area_init_node()
2102 * An architecture is expected to register range of page frames backed by
2103 * physical memory with memblock_add[_node]() before calling
2104 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2105 * usage, an architecture is expected to do something like
2107 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2109 * for_each_valid_physical_page_range()
2110 * memblock_add_node(base, size, nid)
2111 * free_area_init_nodes(max_zone_pfns);
2113 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2114 * registered physical page range. Similarly
2115 * sparse_memory_present_with_active_regions() calls memory_present() for
2116 * each range when SPARSEMEM is enabled.
2118 * See mm/page_alloc.c for more information on each function exposed by
2119 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2121 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2122 unsigned long node_map_pfn_alignment(void);
2123 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2124 unsigned long end_pfn);
2125 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2126 unsigned long end_pfn);
2127 extern void get_pfn_range_for_nid(unsigned int nid,
2128 unsigned long *start_pfn, unsigned long *end_pfn);
2129 extern unsigned long find_min_pfn_with_active_regions(void);
2130 extern void free_bootmem_with_active_regions(int nid,
2131 unsigned long max_low_pfn);
2132 extern void sparse_memory_present_with_active_regions(int nid);
2134 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2136 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2137 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2138 static inline int __early_pfn_to_nid(unsigned long pfn,
2139 struct mminit_pfnnid_cache *state)
2144 /* please see mm/page_alloc.c */
2145 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2146 /* there is a per-arch backend function. */
2147 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2148 struct mminit_pfnnid_cache *state);
2151 #if defined(CONFIG_HAVE_MEMBLOCK) && !defined(CONFIG_FLAT_NODE_MEM_MAP)
2152 void zero_resv_unavail(void);
2154 static inline void zero_resv_unavail(void) {}
2157 extern void set_dma_reserve(unsigned long new_dma_reserve);
2158 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2159 enum memmap_context, struct vmem_altmap *);
2160 extern void setup_per_zone_wmarks(void);
2161 extern int __meminit init_per_zone_wmark_min(void);
2162 extern void mem_init(void);
2163 extern void __init mmap_init(void);
2164 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2165 extern long si_mem_available(void);
2166 extern void si_meminfo(struct sysinfo * val);
2167 extern void si_meminfo_node(struct sysinfo *val, int nid);
2168 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2169 extern unsigned long arch_reserved_kernel_pages(void);
2172 extern __printf(3, 4)
2173 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2175 extern void setup_per_cpu_pageset(void);
2177 extern void zone_pcp_update(struct zone *zone);
2178 extern void zone_pcp_reset(struct zone *zone);
2181 extern int min_free_kbytes;
2182 extern int watermark_scale_factor;
2185 extern atomic_long_t mmap_pages_allocated;
2186 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2188 /* interval_tree.c */
2189 void vma_interval_tree_insert(struct vm_area_struct *node,
2190 struct rb_root_cached *root);
2191 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2192 struct vm_area_struct *prev,
2193 struct rb_root_cached *root);
2194 void vma_interval_tree_remove(struct vm_area_struct *node,
2195 struct rb_root_cached *root);
2196 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2197 unsigned long start, unsigned long last);
2198 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2199 unsigned long start, unsigned long last);
2201 #define vma_interval_tree_foreach(vma, root, start, last) \
2202 for (vma = vma_interval_tree_iter_first(root, start, last); \
2203 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2205 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2206 struct rb_root_cached *root);
2207 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2208 struct rb_root_cached *root);
2209 struct anon_vma_chain *
2210 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2211 unsigned long start, unsigned long last);
2212 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2213 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2214 #ifdef CONFIG_DEBUG_VM_RB
2215 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2218 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2219 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2220 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2223 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2224 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2225 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2226 struct vm_area_struct *expand);
2227 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2228 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2230 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2232 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2233 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2234 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2235 struct mempolicy *, struct vm_userfaultfd_ctx);
2236 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2237 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2238 unsigned long addr, int new_below);
2239 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2240 unsigned long addr, int new_below);
2241 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2242 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2243 struct rb_node **, struct rb_node *);
2244 extern void unlink_file_vma(struct vm_area_struct *);
2245 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2246 unsigned long addr, unsigned long len, pgoff_t pgoff,
2247 bool *need_rmap_locks);
2248 extern void exit_mmap(struct mm_struct *);
2250 static inline int check_data_rlimit(unsigned long rlim,
2252 unsigned long start,
2253 unsigned long end_data,
2254 unsigned long start_data)
2256 if (rlim < RLIM_INFINITY) {
2257 if (((new - start) + (end_data - start_data)) > rlim)
2264 extern int mm_take_all_locks(struct mm_struct *mm);
2265 extern void mm_drop_all_locks(struct mm_struct *mm);
2267 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2268 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2269 extern struct file *get_task_exe_file(struct task_struct *task);
2271 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2272 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2274 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2275 const struct vm_special_mapping *sm);
2276 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2277 unsigned long addr, unsigned long len,
2278 unsigned long flags,
2279 const struct vm_special_mapping *spec);
2280 /* This is an obsolete alternative to _install_special_mapping. */
2281 extern int install_special_mapping(struct mm_struct *mm,
2282 unsigned long addr, unsigned long len,
2283 unsigned long flags, struct page **pages);
2285 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2287 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2288 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2289 struct list_head *uf);
2290 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2291 unsigned long len, unsigned long prot, unsigned long flags,
2292 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2293 struct list_head *uf);
2294 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2295 struct list_head *uf);
2297 static inline unsigned long
2298 do_mmap_pgoff(struct file *file, unsigned long addr,
2299 unsigned long len, unsigned long prot, unsigned long flags,
2300 unsigned long pgoff, unsigned long *populate,
2301 struct list_head *uf)
2303 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2307 extern int __mm_populate(unsigned long addr, unsigned long len,
2309 static inline void mm_populate(unsigned long addr, unsigned long len)
2312 (void) __mm_populate(addr, len, 1);
2315 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2318 /* These take the mm semaphore themselves */
2319 extern int __must_check vm_brk(unsigned long, unsigned long);
2320 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2321 extern int vm_munmap(unsigned long, size_t);
2322 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2323 unsigned long, unsigned long,
2324 unsigned long, unsigned long);
2326 struct vm_unmapped_area_info {
2327 #define VM_UNMAPPED_AREA_TOPDOWN 1
2328 unsigned long flags;
2329 unsigned long length;
2330 unsigned long low_limit;
2331 unsigned long high_limit;
2332 unsigned long align_mask;
2333 unsigned long align_offset;
2336 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2337 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2340 * Search for an unmapped address range.
2342 * We are looking for a range that:
2343 * - does not intersect with any VMA;
2344 * - is contained within the [low_limit, high_limit) interval;
2345 * - is at least the desired size.
2346 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2348 static inline unsigned long
2349 vm_unmapped_area(struct vm_unmapped_area_info *info)
2351 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2352 return unmapped_area_topdown(info);
2354 return unmapped_area(info);
2358 extern void truncate_inode_pages(struct address_space *, loff_t);
2359 extern void truncate_inode_pages_range(struct address_space *,
2360 loff_t lstart, loff_t lend);
2361 extern void truncate_inode_pages_final(struct address_space *);
2363 /* generic vm_area_ops exported for stackable file systems */
2364 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2365 extern void filemap_map_pages(struct vm_fault *vmf,
2366 pgoff_t start_pgoff, pgoff_t end_pgoff);
2367 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2369 /* mm/page-writeback.c */
2370 int __must_check write_one_page(struct page *page);
2371 void task_dirty_inc(struct task_struct *tsk);
2374 #define VM_MAX_READAHEAD 128 /* kbytes */
2375 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2377 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2378 pgoff_t offset, unsigned long nr_to_read);
2380 void page_cache_sync_readahead(struct address_space *mapping,
2381 struct file_ra_state *ra,
2384 unsigned long size);
2386 void page_cache_async_readahead(struct address_space *mapping,
2387 struct file_ra_state *ra,
2391 unsigned long size);
2393 extern unsigned long stack_guard_gap;
2394 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2395 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2397 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2398 extern int expand_downwards(struct vm_area_struct *vma,
2399 unsigned long address);
2401 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2403 #define expand_upwards(vma, address) (0)
2406 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2407 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2408 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2409 struct vm_area_struct **pprev);
2411 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2412 NULL if none. Assume start_addr < end_addr. */
2413 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2415 struct vm_area_struct * vma = find_vma(mm,start_addr);
2417 if (vma && end_addr <= vma->vm_start)
2422 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2424 unsigned long vm_start = vma->vm_start;
2426 if (vma->vm_flags & VM_GROWSDOWN) {
2427 vm_start -= stack_guard_gap;
2428 if (vm_start > vma->vm_start)
2434 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2436 unsigned long vm_end = vma->vm_end;
2438 if (vma->vm_flags & VM_GROWSUP) {
2439 vm_end += stack_guard_gap;
2440 if (vm_end < vma->vm_end)
2441 vm_end = -PAGE_SIZE;
2446 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2448 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2451 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2452 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2453 unsigned long vm_start, unsigned long vm_end)
2455 struct vm_area_struct *vma = find_vma(mm, vm_start);
2457 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2464 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2465 void vma_set_page_prot(struct vm_area_struct *vma);
2467 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2471 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2473 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2477 #ifdef CONFIG_NUMA_BALANCING
2478 unsigned long change_prot_numa(struct vm_area_struct *vma,
2479 unsigned long start, unsigned long end);
2482 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2483 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2484 unsigned long pfn, unsigned long size, pgprot_t);
2485 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2486 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2488 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2489 unsigned long pfn, pgprot_t pgprot);
2490 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2492 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2493 unsigned long addr, pfn_t pfn);
2494 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2496 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2497 unsigned long addr, struct page *page)
2499 int err = vm_insert_page(vma, addr, page);
2502 return VM_FAULT_OOM;
2503 if (err < 0 && err != -EBUSY)
2504 return VM_FAULT_SIGBUS;
2506 return VM_FAULT_NOPAGE;
2509 static inline vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma,
2510 unsigned long addr, pfn_t pfn)
2512 int err = vm_insert_mixed(vma, addr, pfn);
2515 return VM_FAULT_OOM;
2516 if (err < 0 && err != -EBUSY)
2517 return VM_FAULT_SIGBUS;
2519 return VM_FAULT_NOPAGE;
2522 static inline vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma,
2523 unsigned long addr, unsigned long pfn)
2525 int err = vm_insert_pfn(vma, addr, pfn);
2528 return VM_FAULT_OOM;
2529 if (err < 0 && err != -EBUSY)
2530 return VM_FAULT_SIGBUS;
2532 return VM_FAULT_NOPAGE;
2535 static inline vm_fault_t vmf_error(int err)
2538 return VM_FAULT_OOM;
2539 return VM_FAULT_SIGBUS;
2542 struct page *follow_page_mask(struct vm_area_struct *vma,
2543 unsigned long address, unsigned int foll_flags,
2544 unsigned int *page_mask);
2546 static inline struct page *follow_page(struct vm_area_struct *vma,
2547 unsigned long address, unsigned int foll_flags)
2549 unsigned int unused_page_mask;
2550 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2553 #define FOLL_WRITE 0x01 /* check pte is writable */
2554 #define FOLL_TOUCH 0x02 /* mark page accessed */
2555 #define FOLL_GET 0x04 /* do get_page on page */
2556 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2557 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2558 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2559 * and return without waiting upon it */
2560 #define FOLL_POPULATE 0x40 /* fault in page */
2561 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2562 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2563 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2564 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2565 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2566 #define FOLL_MLOCK 0x1000 /* lock present pages */
2567 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2568 #define FOLL_COW 0x4000 /* internal GUP flag */
2569 #define FOLL_ANON 0x8000 /* don't do file mappings */
2571 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2573 if (vm_fault & VM_FAULT_OOM)
2575 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2576 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2577 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2582 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2584 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2585 unsigned long size, pte_fn_t fn, void *data);
2588 #ifdef CONFIG_PAGE_POISONING
2589 extern bool page_poisoning_enabled(void);
2590 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2592 static inline bool page_poisoning_enabled(void) { return false; }
2593 static inline void kernel_poison_pages(struct page *page, int numpages,
2597 #ifdef CONFIG_DEBUG_PAGEALLOC
2598 extern bool _debug_pagealloc_enabled;
2599 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2601 static inline bool debug_pagealloc_enabled(void)
2603 return _debug_pagealloc_enabled;
2607 kernel_map_pages(struct page *page, int numpages, int enable)
2609 if (!debug_pagealloc_enabled())
2612 __kernel_map_pages(page, numpages, enable);
2614 #ifdef CONFIG_HIBERNATION
2615 extern bool kernel_page_present(struct page *page);
2616 #endif /* CONFIG_HIBERNATION */
2617 #else /* CONFIG_DEBUG_PAGEALLOC */
2619 kernel_map_pages(struct page *page, int numpages, int enable) {}
2620 #ifdef CONFIG_HIBERNATION
2621 static inline bool kernel_page_present(struct page *page) { return true; }
2622 #endif /* CONFIG_HIBERNATION */
2623 static inline bool debug_pagealloc_enabled(void)
2627 #endif /* CONFIG_DEBUG_PAGEALLOC */
2629 #ifdef __HAVE_ARCH_GATE_AREA
2630 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2631 extern int in_gate_area_no_mm(unsigned long addr);
2632 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2634 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2638 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2639 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2643 #endif /* __HAVE_ARCH_GATE_AREA */
2645 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2647 #ifdef CONFIG_SYSCTL
2648 extern int sysctl_drop_caches;
2649 int drop_caches_sysctl_handler(struct ctl_table *, int,
2650 void __user *, size_t *, loff_t *);
2653 void drop_slab(void);
2654 void drop_slab_node(int nid);
2657 #define randomize_va_space 0
2659 extern int randomize_va_space;
2662 const char * arch_vma_name(struct vm_area_struct *vma);
2663 void print_vma_addr(char *prefix, unsigned long rip);
2665 void sparse_mem_maps_populate_node(struct page **map_map,
2666 unsigned long pnum_begin,
2667 unsigned long pnum_end,
2668 unsigned long map_count,
2671 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2672 struct vmem_altmap *altmap);
2673 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2674 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2675 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2676 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2677 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2678 void *vmemmap_alloc_block(unsigned long size, int node);
2680 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2681 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2682 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2683 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2685 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2686 struct vmem_altmap *altmap);
2687 void vmemmap_populate_print_last(void);
2688 #ifdef CONFIG_MEMORY_HOTPLUG
2689 void vmemmap_free(unsigned long start, unsigned long end,
2690 struct vmem_altmap *altmap);
2692 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2693 unsigned long nr_pages);
2696 MF_COUNT_INCREASED = 1 << 0,
2697 MF_ACTION_REQUIRED = 1 << 1,
2698 MF_MUST_KILL = 1 << 2,
2699 MF_SOFT_OFFLINE = 1 << 3,
2701 extern int memory_failure(unsigned long pfn, int flags);
2702 extern void memory_failure_queue(unsigned long pfn, int flags);
2703 extern int unpoison_memory(unsigned long pfn);
2704 extern int get_hwpoison_page(struct page *page);
2705 #define put_hwpoison_page(page) put_page(page)
2706 extern int sysctl_memory_failure_early_kill;
2707 extern int sysctl_memory_failure_recovery;
2708 extern void shake_page(struct page *p, int access);
2709 extern atomic_long_t num_poisoned_pages __read_mostly;
2710 extern int soft_offline_page(struct page *page, int flags);
2714 * Error handlers for various types of pages.
2717 MF_IGNORED, /* Error: cannot be handled */
2718 MF_FAILED, /* Error: handling failed */
2719 MF_DELAYED, /* Will be handled later */
2720 MF_RECOVERED, /* Successfully recovered */
2723 enum mf_action_page_type {
2725 MF_MSG_KERNEL_HIGH_ORDER,
2727 MF_MSG_DIFFERENT_COMPOUND,
2728 MF_MSG_POISONED_HUGE,
2731 MF_MSG_NON_PMD_HUGE,
2732 MF_MSG_UNMAP_FAILED,
2733 MF_MSG_DIRTY_SWAPCACHE,
2734 MF_MSG_CLEAN_SWAPCACHE,
2735 MF_MSG_DIRTY_MLOCKED_LRU,
2736 MF_MSG_CLEAN_MLOCKED_LRU,
2737 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2738 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2741 MF_MSG_TRUNCATED_LRU,
2747 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2748 extern void clear_huge_page(struct page *page,
2749 unsigned long addr_hint,
2750 unsigned int pages_per_huge_page);
2751 extern void copy_user_huge_page(struct page *dst, struct page *src,
2752 unsigned long addr, struct vm_area_struct *vma,
2753 unsigned int pages_per_huge_page);
2754 extern long copy_huge_page_from_user(struct page *dst_page,
2755 const void __user *usr_src,
2756 unsigned int pages_per_huge_page,
2757 bool allow_pagefault);
2758 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2760 extern struct page_ext_operations debug_guardpage_ops;
2762 #ifdef CONFIG_DEBUG_PAGEALLOC
2763 extern unsigned int _debug_guardpage_minorder;
2764 extern bool _debug_guardpage_enabled;
2766 static inline unsigned int debug_guardpage_minorder(void)
2768 return _debug_guardpage_minorder;
2771 static inline bool debug_guardpage_enabled(void)
2773 return _debug_guardpage_enabled;
2776 static inline bool page_is_guard(struct page *page)
2778 struct page_ext *page_ext;
2780 if (!debug_guardpage_enabled())
2783 page_ext = lookup_page_ext(page);
2784 if (unlikely(!page_ext))
2787 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2790 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2791 static inline bool debug_guardpage_enabled(void) { return false; }
2792 static inline bool page_is_guard(struct page *page) { return false; }
2793 #endif /* CONFIG_DEBUG_PAGEALLOC */
2795 #if MAX_NUMNODES > 1
2796 void __init setup_nr_node_ids(void);
2798 static inline void setup_nr_node_ids(void) {}
2801 #endif /* __KERNEL__ */
2802 #endif /* _LINUX_MM_H */