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 atomic_long_t _totalram_pages;
52 static inline unsigned long totalram_pages(void)
54 return (unsigned long)atomic_long_read(&_totalram_pages);
57 static inline void totalram_pages_inc(void)
59 atomic_long_inc(&_totalram_pages);
62 static inline void totalram_pages_dec(void)
64 atomic_long_dec(&_totalram_pages);
67 static inline void totalram_pages_add(long count)
69 atomic_long_add(count, &_totalram_pages);
72 static inline void totalram_pages_set(long val)
74 atomic_long_set(&_totalram_pages, val);
77 extern void * high_memory;
78 extern int page_cluster;
81 extern int sysctl_legacy_va_layout;
83 #define sysctl_legacy_va_layout 0
86 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
87 extern const int mmap_rnd_bits_min;
88 extern const int mmap_rnd_bits_max;
89 extern int mmap_rnd_bits __read_mostly;
91 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
92 extern const int mmap_rnd_compat_bits_min;
93 extern const int mmap_rnd_compat_bits_max;
94 extern int mmap_rnd_compat_bits __read_mostly;
98 #include <asm/pgtable.h>
99 #include <asm/processor.h>
102 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
106 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
110 #define lm_alias(x) __va(__pa_symbol(x))
114 * To prevent common memory management code establishing
115 * a zero page mapping on a read fault.
116 * This macro should be defined within <asm/pgtable.h>.
117 * s390 does this to prevent multiplexing of hardware bits
118 * related to the physical page in case of virtualization.
120 #ifndef mm_forbids_zeropage
121 #define mm_forbids_zeropage(X) (0)
125 * On some architectures it is expensive to call memset() for small sizes.
126 * Those architectures should provide their own implementation of "struct page"
127 * zeroing by defining this macro in <asm/pgtable.h>.
129 #ifndef mm_zero_struct_page
130 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
134 * Default maximum number of active map areas, this limits the number of vmas
135 * per mm struct. Users can overwrite this number by sysctl but there is a
138 * When a program's coredump is generated as ELF format, a section is created
139 * per a vma. In ELF, the number of sections is represented in unsigned short.
140 * This means the number of sections should be smaller than 65535 at coredump.
141 * Because the kernel adds some informative sections to a image of program at
142 * generating coredump, we need some margin. The number of extra sections is
143 * 1-3 now and depends on arch. We use "5" as safe margin, here.
145 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
146 * not a hard limit any more. Although some userspace tools can be surprised by
149 #define MAPCOUNT_ELF_CORE_MARGIN (5)
150 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
152 extern int sysctl_max_map_count;
154 extern unsigned long sysctl_user_reserve_kbytes;
155 extern unsigned long sysctl_admin_reserve_kbytes;
157 extern int sysctl_overcommit_memory;
158 extern int sysctl_overcommit_ratio;
159 extern unsigned long sysctl_overcommit_kbytes;
161 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
163 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
166 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
168 /* to align the pointer to the (next) page boundary */
169 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
171 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
172 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
174 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
177 * Linux kernel virtual memory manager primitives.
178 * The idea being to have a "virtual" mm in the same way
179 * we have a virtual fs - giving a cleaner interface to the
180 * mm details, and allowing different kinds of memory mappings
181 * (from shared memory to executable loading to arbitrary
185 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
186 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
187 void vm_area_free(struct vm_area_struct *);
190 extern struct rb_root nommu_region_tree;
191 extern struct rw_semaphore nommu_region_sem;
193 extern unsigned int kobjsize(const void *objp);
197 * vm_flags in vm_area_struct, see mm_types.h.
198 * When changing, update also include/trace/events/mmflags.h
200 #define VM_NONE 0x00000000
202 #define VM_READ 0x00000001 /* currently active flags */
203 #define VM_WRITE 0x00000002
204 #define VM_EXEC 0x00000004
205 #define VM_SHARED 0x00000008
207 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
208 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
209 #define VM_MAYWRITE 0x00000020
210 #define VM_MAYEXEC 0x00000040
211 #define VM_MAYSHARE 0x00000080
213 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
214 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
215 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
216 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
217 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
219 #define VM_LOCKED 0x00002000
220 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
222 /* Used by sys_madvise() */
223 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
224 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
226 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
227 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
228 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
229 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
230 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
231 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
232 #define VM_SYNC 0x00800000 /* Synchronous page faults */
233 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
234 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
235 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
237 #ifdef CONFIG_MEM_SOFT_DIRTY
238 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
240 # define VM_SOFTDIRTY 0
243 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
244 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
245 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
246 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
248 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
249 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
250 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
251 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
252 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
253 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
254 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
255 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
256 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
257 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
258 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
259 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
261 #ifdef CONFIG_ARCH_HAS_PKEYS
262 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
263 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
264 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
265 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
266 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
268 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
270 # define VM_PKEY_BIT4 0
272 #endif /* CONFIG_ARCH_HAS_PKEYS */
274 #if defined(CONFIG_X86)
275 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
276 #elif defined(CONFIG_PPC)
277 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
278 #elif defined(CONFIG_PARISC)
279 # define VM_GROWSUP VM_ARCH_1
280 #elif defined(CONFIG_IA64)
281 # define VM_GROWSUP VM_ARCH_1
282 #elif defined(CONFIG_SPARC64)
283 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
284 # define VM_ARCH_CLEAR VM_SPARC_ADI
285 #elif !defined(CONFIG_MMU)
286 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
289 #if defined(CONFIG_X86_INTEL_MPX)
290 /* MPX specific bounds table or bounds directory */
291 # define VM_MPX VM_HIGH_ARCH_4
293 # define VM_MPX VM_NONE
297 # define VM_GROWSUP VM_NONE
300 /* Bits set in the VMA until the stack is in its final location */
301 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
303 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
304 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
307 #ifdef CONFIG_STACK_GROWSUP
308 #define VM_STACK VM_GROWSUP
310 #define VM_STACK VM_GROWSDOWN
313 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
316 * Special vmas that are non-mergable, non-mlock()able.
317 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
319 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
321 /* This mask defines which mm->def_flags a process can inherit its parent */
322 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
324 /* This mask is used to clear all the VMA flags used by mlock */
325 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
327 /* Arch-specific flags to clear when updating VM flags on protection change */
328 #ifndef VM_ARCH_CLEAR
329 # define VM_ARCH_CLEAR VM_NONE
331 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
334 * mapping from the currently active vm_flags protection bits (the
335 * low four bits) to a page protection mask..
337 extern pgprot_t protection_map[16];
339 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
340 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
341 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
342 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
343 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
344 #define FAULT_FLAG_TRIED 0x20 /* Second try */
345 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
346 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
347 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
349 #define FAULT_FLAG_TRACE \
350 { FAULT_FLAG_WRITE, "WRITE" }, \
351 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
352 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
353 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
354 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
355 { FAULT_FLAG_TRIED, "TRIED" }, \
356 { FAULT_FLAG_USER, "USER" }, \
357 { FAULT_FLAG_REMOTE, "REMOTE" }, \
358 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
361 * vm_fault is filled by the the pagefault handler and passed to the vma's
362 * ->fault function. The vma's ->fault is responsible for returning a bitmask
363 * of VM_FAULT_xxx flags that give details about how the fault was handled.
365 * MM layer fills up gfp_mask for page allocations but fault handler might
366 * alter it if its implementation requires a different allocation context.
368 * pgoff should be used in favour of virtual_address, if possible.
371 struct vm_area_struct *vma; /* Target VMA */
372 unsigned int flags; /* FAULT_FLAG_xxx flags */
373 gfp_t gfp_mask; /* gfp mask to be used for allocations */
374 pgoff_t pgoff; /* Logical page offset based on vma */
375 unsigned long address; /* Faulting virtual address */
376 pmd_t *pmd; /* Pointer to pmd entry matching
378 pud_t *pud; /* Pointer to pud entry matching
381 pte_t orig_pte; /* Value of PTE at the time of fault */
383 struct page *cow_page; /* Page handler may use for COW fault */
384 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
385 struct page *page; /* ->fault handlers should return a
386 * page here, unless VM_FAULT_NOPAGE
387 * is set (which is also implied by
390 /* These three entries are valid only while holding ptl lock */
391 pte_t *pte; /* Pointer to pte entry matching
392 * the 'address'. NULL if the page
393 * table hasn't been allocated.
395 spinlock_t *ptl; /* Page table lock.
396 * Protects pte page table if 'pte'
397 * is not NULL, otherwise pmd.
399 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
400 * vm_ops->map_pages() calls
401 * alloc_set_pte() from atomic context.
402 * do_fault_around() pre-allocates
403 * page table to avoid allocation from
408 /* page entry size for vm->huge_fault() */
409 enum page_entry_size {
416 * These are the virtual MM functions - opening of an area, closing and
417 * unmapping it (needed to keep files on disk up-to-date etc), pointer
418 * to the functions called when a no-page or a wp-page exception occurs.
420 struct vm_operations_struct {
421 void (*open)(struct vm_area_struct * area);
422 void (*close)(struct vm_area_struct * area);
423 int (*split)(struct vm_area_struct * area, unsigned long addr);
424 int (*mremap)(struct vm_area_struct * area);
425 vm_fault_t (*fault)(struct vm_fault *vmf);
426 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
427 enum page_entry_size pe_size);
428 void (*map_pages)(struct vm_fault *vmf,
429 pgoff_t start_pgoff, pgoff_t end_pgoff);
430 unsigned long (*pagesize)(struct vm_area_struct * area);
432 /* notification that a previously read-only page is about to become
433 * writable, if an error is returned it will cause a SIGBUS */
434 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
436 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
437 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
439 /* called by access_process_vm when get_user_pages() fails, typically
440 * for use by special VMAs that can switch between memory and hardware
442 int (*access)(struct vm_area_struct *vma, unsigned long addr,
443 void *buf, int len, int write);
445 /* Called by the /proc/PID/maps code to ask the vma whether it
446 * has a special name. Returning non-NULL will also cause this
447 * vma to be dumped unconditionally. */
448 const char *(*name)(struct vm_area_struct *vma);
452 * set_policy() op must add a reference to any non-NULL @new mempolicy
453 * to hold the policy upon return. Caller should pass NULL @new to
454 * remove a policy and fall back to surrounding context--i.e. do not
455 * install a MPOL_DEFAULT policy, nor the task or system default
458 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
461 * get_policy() op must add reference [mpol_get()] to any policy at
462 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
463 * in mm/mempolicy.c will do this automatically.
464 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
465 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
466 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
467 * must return NULL--i.e., do not "fallback" to task or system default
470 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
474 * Called by vm_normal_page() for special PTEs to find the
475 * page for @addr. This is useful if the default behavior
476 * (using pte_page()) would not find the correct page.
478 struct page *(*find_special_page)(struct vm_area_struct *vma,
482 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
484 static const struct vm_operations_struct dummy_vm_ops = {};
486 memset(vma, 0, sizeof(*vma));
488 vma->vm_ops = &dummy_vm_ops;
489 INIT_LIST_HEAD(&vma->anon_vma_chain);
492 static inline void vma_set_anonymous(struct vm_area_struct *vma)
497 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
498 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
503 #define page_private(page) ((page)->private)
504 #define set_page_private(page, v) ((page)->private = (v))
506 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
507 static inline int pmd_devmap(pmd_t pmd)
511 static inline int pud_devmap(pud_t pud)
515 static inline int pgd_devmap(pgd_t pgd)
522 * FIXME: take this include out, include page-flags.h in
523 * files which need it (119 of them)
525 #include <linux/page-flags.h>
526 #include <linux/huge_mm.h>
529 * Methods to modify the page usage count.
531 * What counts for a page usage:
532 * - cache mapping (page->mapping)
533 * - private data (page->private)
534 * - page mapped in a task's page tables, each mapping
535 * is counted separately
537 * Also, many kernel routines increase the page count before a critical
538 * routine so they can be sure the page doesn't go away from under them.
542 * Drop a ref, return true if the refcount fell to zero (the page has no users)
544 static inline int put_page_testzero(struct page *page)
546 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
547 return page_ref_dec_and_test(page);
551 * Try to grab a ref unless the page has a refcount of zero, return false if
553 * This can be called when MMU is off so it must not access
554 * any of the virtual mappings.
556 static inline int get_page_unless_zero(struct page *page)
558 return page_ref_add_unless(page, 1, 0);
561 extern int page_is_ram(unsigned long pfn);
569 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
572 /* Support for virtually mapped pages */
573 struct page *vmalloc_to_page(const void *addr);
574 unsigned long vmalloc_to_pfn(const void *addr);
577 * Determine if an address is within the vmalloc range
579 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
580 * is no special casing required.
582 static inline bool is_vmalloc_addr(const void *x)
585 unsigned long addr = (unsigned long)x;
587 return addr >= VMALLOC_START && addr < VMALLOC_END;
593 extern int is_vmalloc_or_module_addr(const void *x);
595 static inline int is_vmalloc_or_module_addr(const void *x)
601 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
602 static inline void *kvmalloc(size_t size, gfp_t flags)
604 return kvmalloc_node(size, flags, NUMA_NO_NODE);
606 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
608 return kvmalloc_node(size, flags | __GFP_ZERO, node);
610 static inline void *kvzalloc(size_t size, gfp_t flags)
612 return kvmalloc(size, flags | __GFP_ZERO);
615 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
619 if (unlikely(check_mul_overflow(n, size, &bytes)))
622 return kvmalloc(bytes, flags);
625 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
627 return kvmalloc_array(n, size, flags | __GFP_ZERO);
630 extern void kvfree(const void *addr);
632 static inline atomic_t *compound_mapcount_ptr(struct page *page)
634 return &page[1].compound_mapcount;
637 static inline int compound_mapcount(struct page *page)
639 VM_BUG_ON_PAGE(!PageCompound(page), page);
640 page = compound_head(page);
641 return atomic_read(compound_mapcount_ptr(page)) + 1;
645 * The atomic page->_mapcount, starts from -1: so that transitions
646 * both from it and to it can be tracked, using atomic_inc_and_test
647 * and atomic_add_negative(-1).
649 static inline void page_mapcount_reset(struct page *page)
651 atomic_set(&(page)->_mapcount, -1);
654 int __page_mapcount(struct page *page);
656 static inline int page_mapcount(struct page *page)
658 VM_BUG_ON_PAGE(PageSlab(page), page);
660 if (unlikely(PageCompound(page)))
661 return __page_mapcount(page);
662 return atomic_read(&page->_mapcount) + 1;
665 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
666 int total_mapcount(struct page *page);
667 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
669 static inline int total_mapcount(struct page *page)
671 return page_mapcount(page);
673 static inline int page_trans_huge_mapcount(struct page *page,
676 int mapcount = page_mapcount(page);
678 *total_mapcount = mapcount;
683 static inline struct page *virt_to_head_page(const void *x)
685 struct page *page = virt_to_page(x);
687 return compound_head(page);
690 void __put_page(struct page *page);
692 void put_pages_list(struct list_head *pages);
694 void split_page(struct page *page, unsigned int order);
697 * Compound pages have a destructor function. Provide a
698 * prototype for that function and accessor functions.
699 * These are _only_ valid on the head of a compound page.
701 typedef void compound_page_dtor(struct page *);
703 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
704 enum compound_dtor_id {
707 #ifdef CONFIG_HUGETLB_PAGE
710 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
715 extern compound_page_dtor * const compound_page_dtors[];
717 static inline void set_compound_page_dtor(struct page *page,
718 enum compound_dtor_id compound_dtor)
720 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
721 page[1].compound_dtor = compound_dtor;
724 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
726 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
727 return compound_page_dtors[page[1].compound_dtor];
730 static inline unsigned int compound_order(struct page *page)
734 return page[1].compound_order;
737 static inline void set_compound_order(struct page *page, unsigned int order)
739 page[1].compound_order = order;
742 void free_compound_page(struct page *page);
746 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
747 * servicing faults for write access. In the normal case, do always want
748 * pte_mkwrite. But get_user_pages can cause write faults for mappings
749 * that do not have writing enabled, when used by access_process_vm.
751 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
753 if (likely(vma->vm_flags & VM_WRITE))
754 pte = pte_mkwrite(pte);
758 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
760 vm_fault_t finish_fault(struct vm_fault *vmf);
761 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
765 * Multiple processes may "see" the same page. E.g. for untouched
766 * mappings of /dev/null, all processes see the same page full of
767 * zeroes, and text pages of executables and shared libraries have
768 * only one copy in memory, at most, normally.
770 * For the non-reserved pages, page_count(page) denotes a reference count.
771 * page_count() == 0 means the page is free. page->lru is then used for
772 * freelist management in the buddy allocator.
773 * page_count() > 0 means the page has been allocated.
775 * Pages are allocated by the slab allocator in order to provide memory
776 * to kmalloc and kmem_cache_alloc. In this case, the management of the
777 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
778 * unless a particular usage is carefully commented. (the responsibility of
779 * freeing the kmalloc memory is the caller's, of course).
781 * A page may be used by anyone else who does a __get_free_page().
782 * In this case, page_count still tracks the references, and should only
783 * be used through the normal accessor functions. The top bits of page->flags
784 * and page->virtual store page management information, but all other fields
785 * are unused and could be used privately, carefully. The management of this
786 * page is the responsibility of the one who allocated it, and those who have
787 * subsequently been given references to it.
789 * The other pages (we may call them "pagecache pages") are completely
790 * managed by the Linux memory manager: I/O, buffers, swapping etc.
791 * The following discussion applies only to them.
793 * A pagecache page contains an opaque `private' member, which belongs to the
794 * page's address_space. Usually, this is the address of a circular list of
795 * the page's disk buffers. PG_private must be set to tell the VM to call
796 * into the filesystem to release these pages.
798 * A page may belong to an inode's memory mapping. In this case, page->mapping
799 * is the pointer to the inode, and page->index is the file offset of the page,
800 * in units of PAGE_SIZE.
802 * If pagecache pages are not associated with an inode, they are said to be
803 * anonymous pages. These may become associated with the swapcache, and in that
804 * case PG_swapcache is set, and page->private is an offset into the swapcache.
806 * In either case (swapcache or inode backed), the pagecache itself holds one
807 * reference to the page. Setting PG_private should also increment the
808 * refcount. The each user mapping also has a reference to the page.
810 * The pagecache pages are stored in a per-mapping radix tree, which is
811 * rooted at mapping->i_pages, and indexed by offset.
812 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
813 * lists, we instead now tag pages as dirty/writeback in the radix tree.
815 * All pagecache pages may be subject to I/O:
816 * - inode pages may need to be read from disk,
817 * - inode pages which have been modified and are MAP_SHARED may need
818 * to be written back to the inode on disk,
819 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
820 * modified may need to be swapped out to swap space and (later) to be read
825 * The zone field is never updated after free_area_init_core()
826 * sets it, so none of the operations on it need to be atomic.
829 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
830 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
831 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
832 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
833 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
834 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
837 * Define the bit shifts to access each section. For non-existent
838 * sections we define the shift as 0; that plus a 0 mask ensures
839 * the compiler will optimise away reference to them.
841 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
842 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
843 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
844 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
845 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
847 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
848 #ifdef NODE_NOT_IN_PAGE_FLAGS
849 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
850 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
851 SECTIONS_PGOFF : ZONES_PGOFF)
853 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
854 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
855 NODES_PGOFF : ZONES_PGOFF)
858 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
860 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
861 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
864 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
865 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
866 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
867 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
868 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
869 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
871 static inline enum zone_type page_zonenum(const struct page *page)
873 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
876 #ifdef CONFIG_ZONE_DEVICE
877 static inline bool is_zone_device_page(const struct page *page)
879 return page_zonenum(page) == ZONE_DEVICE;
881 extern void memmap_init_zone_device(struct zone *, unsigned long,
882 unsigned long, struct dev_pagemap *);
884 static inline bool is_zone_device_page(const struct page *page)
890 #ifdef CONFIG_DEV_PAGEMAP_OPS
891 void dev_pagemap_get_ops(void);
892 void dev_pagemap_put_ops(void);
893 void __put_devmap_managed_page(struct page *page);
894 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
895 static inline bool put_devmap_managed_page(struct page *page)
897 if (!static_branch_unlikely(&devmap_managed_key))
899 if (!is_zone_device_page(page))
901 switch (page->pgmap->type) {
902 case MEMORY_DEVICE_PRIVATE:
903 case MEMORY_DEVICE_PUBLIC:
904 case MEMORY_DEVICE_FS_DAX:
905 __put_devmap_managed_page(page);
913 static inline bool is_device_private_page(const struct page *page)
915 return is_zone_device_page(page) &&
916 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
919 static inline bool is_device_public_page(const struct page *page)
921 return is_zone_device_page(page) &&
922 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
925 #ifdef CONFIG_PCI_P2PDMA
926 static inline bool is_pci_p2pdma_page(const struct page *page)
928 return is_zone_device_page(page) &&
929 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
931 #else /* CONFIG_PCI_P2PDMA */
932 static inline bool is_pci_p2pdma_page(const struct page *page)
936 #endif /* CONFIG_PCI_P2PDMA */
938 #else /* CONFIG_DEV_PAGEMAP_OPS */
939 static inline void dev_pagemap_get_ops(void)
943 static inline void dev_pagemap_put_ops(void)
947 static inline bool put_devmap_managed_page(struct page *page)
952 static inline bool is_device_private_page(const struct page *page)
957 static inline bool is_device_public_page(const struct page *page)
962 static inline bool is_pci_p2pdma_page(const struct page *page)
966 #endif /* CONFIG_DEV_PAGEMAP_OPS */
968 static inline void get_page(struct page *page)
970 page = compound_head(page);
972 * Getting a normal page or the head of a compound page
973 * requires to already have an elevated page->_refcount.
975 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
979 static inline void put_page(struct page *page)
981 page = compound_head(page);
984 * For devmap managed pages we need to catch refcount transition from
985 * 2 to 1, when refcount reach one it means the page is free and we
986 * need to inform the device driver through callback. See
987 * include/linux/memremap.h and HMM for details.
989 if (put_devmap_managed_page(page))
992 if (put_page_testzero(page))
996 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
997 #define SECTION_IN_PAGE_FLAGS
1001 * The identification function is mainly used by the buddy allocator for
1002 * determining if two pages could be buddies. We are not really identifying
1003 * the zone since we could be using the section number id if we do not have
1004 * node id available in page flags.
1005 * We only guarantee that it will return the same value for two combinable
1008 static inline int page_zone_id(struct page *page)
1010 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1013 #ifdef NODE_NOT_IN_PAGE_FLAGS
1014 extern int page_to_nid(const struct page *page);
1016 static inline int page_to_nid(const struct page *page)
1018 struct page *p = (struct page *)page;
1020 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1024 #ifdef CONFIG_NUMA_BALANCING
1025 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1027 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1030 static inline int cpupid_to_pid(int cpupid)
1032 return cpupid & LAST__PID_MASK;
1035 static inline int cpupid_to_cpu(int cpupid)
1037 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1040 static inline int cpupid_to_nid(int cpupid)
1042 return cpu_to_node(cpupid_to_cpu(cpupid));
1045 static inline bool cpupid_pid_unset(int cpupid)
1047 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1050 static inline bool cpupid_cpu_unset(int cpupid)
1052 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1055 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1057 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1060 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1061 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1062 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1064 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1067 static inline int page_cpupid_last(struct page *page)
1069 return page->_last_cpupid;
1071 static inline void page_cpupid_reset_last(struct page *page)
1073 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1076 static inline int page_cpupid_last(struct page *page)
1078 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1081 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1083 static inline void page_cpupid_reset_last(struct page *page)
1085 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1087 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1088 #else /* !CONFIG_NUMA_BALANCING */
1089 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1091 return page_to_nid(page); /* XXX */
1094 static inline int page_cpupid_last(struct page *page)
1096 return page_to_nid(page); /* XXX */
1099 static inline int cpupid_to_nid(int cpupid)
1104 static inline int cpupid_to_pid(int cpupid)
1109 static inline int cpupid_to_cpu(int cpupid)
1114 static inline int cpu_pid_to_cpupid(int nid, int pid)
1119 static inline bool cpupid_pid_unset(int cpupid)
1124 static inline void page_cpupid_reset_last(struct page *page)
1128 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1132 #endif /* CONFIG_NUMA_BALANCING */
1134 #ifdef CONFIG_KASAN_SW_TAGS
1135 static inline u8 page_kasan_tag(const struct page *page)
1137 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1140 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1142 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1143 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1146 static inline void page_kasan_tag_reset(struct page *page)
1148 page_kasan_tag_set(page, 0xff);
1151 static inline u8 page_kasan_tag(const struct page *page)
1156 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1157 static inline void page_kasan_tag_reset(struct page *page) { }
1160 static inline struct zone *page_zone(const struct page *page)
1162 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1165 static inline pg_data_t *page_pgdat(const struct page *page)
1167 return NODE_DATA(page_to_nid(page));
1170 #ifdef SECTION_IN_PAGE_FLAGS
1171 static inline void set_page_section(struct page *page, unsigned long section)
1173 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1174 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1177 static inline unsigned long page_to_section(const struct page *page)
1179 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1183 static inline void set_page_zone(struct page *page, enum zone_type zone)
1185 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1186 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1189 static inline void set_page_node(struct page *page, unsigned long node)
1191 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1192 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1195 static inline void set_page_links(struct page *page, enum zone_type zone,
1196 unsigned long node, unsigned long pfn)
1198 set_page_zone(page, zone);
1199 set_page_node(page, node);
1200 #ifdef SECTION_IN_PAGE_FLAGS
1201 set_page_section(page, pfn_to_section_nr(pfn));
1206 static inline struct mem_cgroup *page_memcg(struct page *page)
1208 return page->mem_cgroup;
1210 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1212 WARN_ON_ONCE(!rcu_read_lock_held());
1213 return READ_ONCE(page->mem_cgroup);
1216 static inline struct mem_cgroup *page_memcg(struct page *page)
1220 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1222 WARN_ON_ONCE(!rcu_read_lock_held());
1228 * Some inline functions in vmstat.h depend on page_zone()
1230 #include <linux/vmstat.h>
1232 static __always_inline void *lowmem_page_address(const struct page *page)
1234 return page_to_virt(page);
1237 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1238 #define HASHED_PAGE_VIRTUAL
1241 #if defined(WANT_PAGE_VIRTUAL)
1242 static inline void *page_address(const struct page *page)
1244 return page->virtual;
1246 static inline void set_page_address(struct page *page, void *address)
1248 page->virtual = address;
1250 #define page_address_init() do { } while(0)
1253 #if defined(HASHED_PAGE_VIRTUAL)
1254 void *page_address(const struct page *page);
1255 void set_page_address(struct page *page, void *virtual);
1256 void page_address_init(void);
1259 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1260 #define page_address(page) lowmem_page_address(page)
1261 #define set_page_address(page, address) do { } while(0)
1262 #define page_address_init() do { } while(0)
1265 extern void *page_rmapping(struct page *page);
1266 extern struct anon_vma *page_anon_vma(struct page *page);
1267 extern struct address_space *page_mapping(struct page *page);
1269 extern struct address_space *__page_file_mapping(struct page *);
1272 struct address_space *page_file_mapping(struct page *page)
1274 if (unlikely(PageSwapCache(page)))
1275 return __page_file_mapping(page);
1277 return page->mapping;
1280 extern pgoff_t __page_file_index(struct page *page);
1283 * Return the pagecache index of the passed page. Regular pagecache pages
1284 * use ->index whereas swapcache pages use swp_offset(->private)
1286 static inline pgoff_t page_index(struct page *page)
1288 if (unlikely(PageSwapCache(page)))
1289 return __page_file_index(page);
1293 bool page_mapped(struct page *page);
1294 struct address_space *page_mapping(struct page *page);
1295 struct address_space *page_mapping_file(struct page *page);
1298 * Return true only if the page has been allocated with
1299 * ALLOC_NO_WATERMARKS and the low watermark was not
1300 * met implying that the system is under some pressure.
1302 static inline bool page_is_pfmemalloc(struct page *page)
1305 * Page index cannot be this large so this must be
1306 * a pfmemalloc page.
1308 return page->index == -1UL;
1312 * Only to be called by the page allocator on a freshly allocated
1315 static inline void set_page_pfmemalloc(struct page *page)
1320 static inline void clear_page_pfmemalloc(struct page *page)
1326 * Different kinds of faults, as returned by handle_mm_fault().
1327 * Used to decide whether a process gets delivered SIGBUS or
1328 * just gets major/minor fault counters bumped up.
1331 #define VM_FAULT_OOM 0x0001
1332 #define VM_FAULT_SIGBUS 0x0002
1333 #define VM_FAULT_MAJOR 0x0004
1334 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1335 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1336 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1337 #define VM_FAULT_SIGSEGV 0x0040
1339 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1340 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1341 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1342 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1343 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1344 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1345 * and needs fsync() to complete (for
1346 * synchronous page faults in DAX) */
1348 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1349 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1352 #define VM_FAULT_RESULT_TRACE \
1353 { VM_FAULT_OOM, "OOM" }, \
1354 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1355 { VM_FAULT_MAJOR, "MAJOR" }, \
1356 { VM_FAULT_WRITE, "WRITE" }, \
1357 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1358 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1359 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1360 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1361 { VM_FAULT_LOCKED, "LOCKED" }, \
1362 { VM_FAULT_RETRY, "RETRY" }, \
1363 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1364 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1365 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1367 /* Encode hstate index for a hwpoisoned large page */
1368 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1369 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1372 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1374 extern void pagefault_out_of_memory(void);
1376 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1379 * Flags passed to show_mem() and show_free_areas() to suppress output in
1382 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1384 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1386 extern bool can_do_mlock(void);
1387 extern int user_shm_lock(size_t, struct user_struct *);
1388 extern void user_shm_unlock(size_t, struct user_struct *);
1391 * Parameter block passed down to zap_pte_range in exceptional cases.
1393 struct zap_details {
1394 struct address_space *check_mapping; /* Check page->mapping if set */
1395 pgoff_t first_index; /* Lowest page->index to unmap */
1396 pgoff_t last_index; /* Highest page->index to unmap */
1399 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1400 pte_t pte, bool with_public_device);
1401 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1403 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1406 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1407 unsigned long size);
1408 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1409 unsigned long size);
1410 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1411 unsigned long start, unsigned long end);
1414 * mm_walk - callbacks for walk_page_range
1415 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1416 * this handler should only handle pud_trans_huge() puds.
1417 * the pmd_entry or pte_entry callbacks will be used for
1419 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1420 * this handler is required to be able to handle
1421 * pmd_trans_huge() pmds. They may simply choose to
1422 * split_huge_page() instead of handling it explicitly.
1423 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1424 * @pte_hole: if set, called for each hole at all levels
1425 * @hugetlb_entry: if set, called for each hugetlb entry
1426 * @test_walk: caller specific callback function to determine whether
1427 * we walk over the current vma or not. Returning 0
1428 * value means "do page table walk over the current vma,"
1429 * and a negative one means "abort current page table walk
1430 * right now." 1 means "skip the current vma."
1431 * @mm: mm_struct representing the target process of page table walk
1432 * @vma: vma currently walked (NULL if walking outside vmas)
1433 * @private: private data for callbacks' usage
1435 * (see the comment on walk_page_range() for more details)
1438 int (*pud_entry)(pud_t *pud, unsigned long addr,
1439 unsigned long next, struct mm_walk *walk);
1440 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1441 unsigned long next, struct mm_walk *walk);
1442 int (*pte_entry)(pte_t *pte, unsigned long addr,
1443 unsigned long next, struct mm_walk *walk);
1444 int (*pte_hole)(unsigned long addr, unsigned long next,
1445 struct mm_walk *walk);
1446 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1447 unsigned long addr, unsigned long next,
1448 struct mm_walk *walk);
1449 int (*test_walk)(unsigned long addr, unsigned long next,
1450 struct mm_walk *walk);
1451 struct mm_struct *mm;
1452 struct vm_area_struct *vma;
1456 struct mmu_notifier_range;
1458 int walk_page_range(unsigned long addr, unsigned long end,
1459 struct mm_walk *walk);
1460 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1461 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1462 unsigned long end, unsigned long floor, unsigned long ceiling);
1463 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1464 struct vm_area_struct *vma);
1465 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1466 struct mmu_notifier_range *range,
1467 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1468 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1469 unsigned long *pfn);
1470 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1471 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1472 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1473 void *buf, int len, int write);
1475 extern void truncate_pagecache(struct inode *inode, loff_t new);
1476 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1477 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1478 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1479 int truncate_inode_page(struct address_space *mapping, struct page *page);
1480 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1481 int invalidate_inode_page(struct page *page);
1484 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1485 unsigned long address, unsigned int flags);
1486 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1487 unsigned long address, unsigned int fault_flags,
1489 void unmap_mapping_pages(struct address_space *mapping,
1490 pgoff_t start, pgoff_t nr, bool even_cows);
1491 void unmap_mapping_range(struct address_space *mapping,
1492 loff_t const holebegin, loff_t const holelen, int even_cows);
1494 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1495 unsigned long address, unsigned int flags)
1497 /* should never happen if there's no MMU */
1499 return VM_FAULT_SIGBUS;
1501 static inline int fixup_user_fault(struct task_struct *tsk,
1502 struct mm_struct *mm, unsigned long address,
1503 unsigned int fault_flags, bool *unlocked)
1505 /* should never happen if there's no MMU */
1509 static inline void unmap_mapping_pages(struct address_space *mapping,
1510 pgoff_t start, pgoff_t nr, bool even_cows) { }
1511 static inline void unmap_mapping_range(struct address_space *mapping,
1512 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1515 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1516 loff_t const holebegin, loff_t const holelen)
1518 unmap_mapping_range(mapping, holebegin, holelen, 0);
1521 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1522 void *buf, int len, unsigned int gup_flags);
1523 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1524 void *buf, int len, unsigned int gup_flags);
1525 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1526 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1528 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1529 unsigned long start, unsigned long nr_pages,
1530 unsigned int gup_flags, struct page **pages,
1531 struct vm_area_struct **vmas, int *locked);
1532 long get_user_pages(unsigned long start, unsigned long nr_pages,
1533 unsigned int gup_flags, struct page **pages,
1534 struct vm_area_struct **vmas);
1535 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1536 unsigned int gup_flags, struct page **pages, int *locked);
1537 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1538 struct page **pages, unsigned int gup_flags);
1540 #if defined(CONFIG_FS_DAX) || defined(CONFIG_CMA)
1541 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1542 unsigned int gup_flags, struct page **pages,
1543 struct vm_area_struct **vmas);
1545 static inline long get_user_pages_longterm(unsigned long start,
1546 unsigned long nr_pages, unsigned int gup_flags,
1547 struct page **pages, struct vm_area_struct **vmas)
1549 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1551 #endif /* CONFIG_FS_DAX */
1553 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1554 struct page **pages);
1556 /* Container for pinned pfns / pages */
1557 struct frame_vector {
1558 unsigned int nr_allocated; /* Number of frames we have space for */
1559 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1560 bool got_ref; /* Did we pin pages by getting page ref? */
1561 bool is_pfns; /* Does array contain pages or pfns? */
1562 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1563 * pfns_vector_pages() or pfns_vector_pfns()
1567 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1568 void frame_vector_destroy(struct frame_vector *vec);
1569 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1570 unsigned int gup_flags, struct frame_vector *vec);
1571 void put_vaddr_frames(struct frame_vector *vec);
1572 int frame_vector_to_pages(struct frame_vector *vec);
1573 void frame_vector_to_pfns(struct frame_vector *vec);
1575 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1577 return vec->nr_frames;
1580 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1583 int err = frame_vector_to_pages(vec);
1586 return ERR_PTR(err);
1588 return (struct page **)(vec->ptrs);
1591 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1594 frame_vector_to_pfns(vec);
1595 return (unsigned long *)(vec->ptrs);
1599 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1600 struct page **pages);
1601 int get_kernel_page(unsigned long start, int write, struct page **pages);
1602 struct page *get_dump_page(unsigned long addr);
1604 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1605 extern void do_invalidatepage(struct page *page, unsigned int offset,
1606 unsigned int length);
1608 void __set_page_dirty(struct page *, struct address_space *, int warn);
1609 int __set_page_dirty_nobuffers(struct page *page);
1610 int __set_page_dirty_no_writeback(struct page *page);
1611 int redirty_page_for_writepage(struct writeback_control *wbc,
1613 void account_page_dirtied(struct page *page, struct address_space *mapping);
1614 void account_page_cleaned(struct page *page, struct address_space *mapping,
1615 struct bdi_writeback *wb);
1616 int set_page_dirty(struct page *page);
1617 int set_page_dirty_lock(struct page *page);
1618 void __cancel_dirty_page(struct page *page);
1619 static inline void cancel_dirty_page(struct page *page)
1621 /* Avoid atomic ops, locking, etc. when not actually needed. */
1622 if (PageDirty(page))
1623 __cancel_dirty_page(page);
1625 int clear_page_dirty_for_io(struct page *page);
1627 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1629 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1631 return !vma->vm_ops;
1636 * The vma_is_shmem is not inline because it is used only by slow
1637 * paths in userfault.
1639 bool vma_is_shmem(struct vm_area_struct *vma);
1641 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1644 int vma_is_stack_for_current(struct vm_area_struct *vma);
1646 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1647 unsigned long old_addr, struct vm_area_struct *new_vma,
1648 unsigned long new_addr, unsigned long len,
1649 bool need_rmap_locks);
1650 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1651 unsigned long end, pgprot_t newprot,
1652 int dirty_accountable, int prot_numa);
1653 extern int mprotect_fixup(struct vm_area_struct *vma,
1654 struct vm_area_struct **pprev, unsigned long start,
1655 unsigned long end, unsigned long newflags);
1658 * doesn't attempt to fault and will return short.
1660 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1661 struct page **pages);
1663 * per-process(per-mm_struct) statistics.
1665 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1667 long val = atomic_long_read(&mm->rss_stat.count[member]);
1669 #ifdef SPLIT_RSS_COUNTING
1671 * counter is updated in asynchronous manner and may go to minus.
1672 * But it's never be expected number for users.
1677 return (unsigned long)val;
1680 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1682 atomic_long_add(value, &mm->rss_stat.count[member]);
1685 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1687 atomic_long_inc(&mm->rss_stat.count[member]);
1690 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1692 atomic_long_dec(&mm->rss_stat.count[member]);
1695 /* Optimized variant when page is already known not to be PageAnon */
1696 static inline int mm_counter_file(struct page *page)
1698 if (PageSwapBacked(page))
1699 return MM_SHMEMPAGES;
1700 return MM_FILEPAGES;
1703 static inline int mm_counter(struct page *page)
1706 return MM_ANONPAGES;
1707 return mm_counter_file(page);
1710 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1712 return get_mm_counter(mm, MM_FILEPAGES) +
1713 get_mm_counter(mm, MM_ANONPAGES) +
1714 get_mm_counter(mm, MM_SHMEMPAGES);
1717 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1719 return max(mm->hiwater_rss, get_mm_rss(mm));
1722 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1724 return max(mm->hiwater_vm, mm->total_vm);
1727 static inline void update_hiwater_rss(struct mm_struct *mm)
1729 unsigned long _rss = get_mm_rss(mm);
1731 if ((mm)->hiwater_rss < _rss)
1732 (mm)->hiwater_rss = _rss;
1735 static inline void update_hiwater_vm(struct mm_struct *mm)
1737 if (mm->hiwater_vm < mm->total_vm)
1738 mm->hiwater_vm = mm->total_vm;
1741 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1743 mm->hiwater_rss = get_mm_rss(mm);
1746 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1747 struct mm_struct *mm)
1749 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1751 if (*maxrss < hiwater_rss)
1752 *maxrss = hiwater_rss;
1755 #if defined(SPLIT_RSS_COUNTING)
1756 void sync_mm_rss(struct mm_struct *mm);
1758 static inline void sync_mm_rss(struct mm_struct *mm)
1763 #ifndef __HAVE_ARCH_PTE_DEVMAP
1764 static inline int pte_devmap(pte_t pte)
1770 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1772 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1774 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1778 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1782 #ifdef __PAGETABLE_P4D_FOLDED
1783 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1784 unsigned long address)
1789 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1792 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1793 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1794 unsigned long address)
1798 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1799 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1802 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1804 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1806 if (mm_pud_folded(mm))
1808 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1811 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1813 if (mm_pud_folded(mm))
1815 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1819 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1820 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1821 unsigned long address)
1826 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1827 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1830 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1832 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1834 if (mm_pmd_folded(mm))
1836 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1839 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1841 if (mm_pmd_folded(mm))
1843 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1848 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1850 atomic_long_set(&mm->pgtables_bytes, 0);
1853 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1855 return atomic_long_read(&mm->pgtables_bytes);
1858 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1860 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1863 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1865 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1869 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1870 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1875 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1876 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1879 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1880 int __pte_alloc_kernel(pmd_t *pmd);
1883 * The following ifdef needed to get the 4level-fixup.h header to work.
1884 * Remove it when 4level-fixup.h has been removed.
1886 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1888 #ifndef __ARCH_HAS_5LEVEL_HACK
1889 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1890 unsigned long address)
1892 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1893 NULL : p4d_offset(pgd, address);
1896 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1897 unsigned long address)
1899 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1900 NULL : pud_offset(p4d, address);
1902 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1904 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1906 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1907 NULL: pmd_offset(pud, address);
1909 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1911 #if USE_SPLIT_PTE_PTLOCKS
1912 #if ALLOC_SPLIT_PTLOCKS
1913 void __init ptlock_cache_init(void);
1914 extern bool ptlock_alloc(struct page *page);
1915 extern void ptlock_free(struct page *page);
1917 static inline spinlock_t *ptlock_ptr(struct page *page)
1921 #else /* ALLOC_SPLIT_PTLOCKS */
1922 static inline void ptlock_cache_init(void)
1926 static inline bool ptlock_alloc(struct page *page)
1931 static inline void ptlock_free(struct page *page)
1935 static inline spinlock_t *ptlock_ptr(struct page *page)
1939 #endif /* ALLOC_SPLIT_PTLOCKS */
1941 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1943 return ptlock_ptr(pmd_page(*pmd));
1946 static inline bool ptlock_init(struct page *page)
1949 * prep_new_page() initialize page->private (and therefore page->ptl)
1950 * with 0. Make sure nobody took it in use in between.
1952 * It can happen if arch try to use slab for page table allocation:
1953 * slab code uses page->slab_cache, which share storage with page->ptl.
1955 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1956 if (!ptlock_alloc(page))
1958 spin_lock_init(ptlock_ptr(page));
1962 #else /* !USE_SPLIT_PTE_PTLOCKS */
1964 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1966 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1968 return &mm->page_table_lock;
1970 static inline void ptlock_cache_init(void) {}
1971 static inline bool ptlock_init(struct page *page) { return true; }
1972 static inline void ptlock_free(struct page *page) {}
1973 #endif /* USE_SPLIT_PTE_PTLOCKS */
1975 static inline void pgtable_init(void)
1977 ptlock_cache_init();
1978 pgtable_cache_init();
1981 static inline bool pgtable_page_ctor(struct page *page)
1983 if (!ptlock_init(page))
1985 __SetPageTable(page);
1986 inc_zone_page_state(page, NR_PAGETABLE);
1990 static inline void pgtable_page_dtor(struct page *page)
1993 __ClearPageTable(page);
1994 dec_zone_page_state(page, NR_PAGETABLE);
1997 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1999 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2000 pte_t *__pte = pte_offset_map(pmd, address); \
2006 #define pte_unmap_unlock(pte, ptl) do { \
2011 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2013 #define pte_alloc_map(mm, pmd, address) \
2014 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2016 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2017 (pte_alloc(mm, pmd) ? \
2018 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2020 #define pte_alloc_kernel(pmd, address) \
2021 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2022 NULL: pte_offset_kernel(pmd, address))
2024 #if USE_SPLIT_PMD_PTLOCKS
2026 static struct page *pmd_to_page(pmd_t *pmd)
2028 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2029 return virt_to_page((void *)((unsigned long) pmd & mask));
2032 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2034 return ptlock_ptr(pmd_to_page(pmd));
2037 static inline bool pgtable_pmd_page_ctor(struct page *page)
2039 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2040 page->pmd_huge_pte = NULL;
2042 return ptlock_init(page);
2045 static inline void pgtable_pmd_page_dtor(struct page *page)
2047 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2048 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2053 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2057 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2059 return &mm->page_table_lock;
2062 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2063 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2065 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2069 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2071 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2077 * No scalability reason to split PUD locks yet, but follow the same pattern
2078 * as the PMD locks to make it easier if we decide to. The VM should not be
2079 * considered ready to switch to split PUD locks yet; there may be places
2080 * which need to be converted from page_table_lock.
2082 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2084 return &mm->page_table_lock;
2087 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2089 spinlock_t *ptl = pud_lockptr(mm, pud);
2095 extern void __init pagecache_init(void);
2096 extern void free_area_init(unsigned long * zones_size);
2097 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2098 unsigned long zone_start_pfn, unsigned long *zholes_size);
2099 extern void free_initmem(void);
2102 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2103 * into the buddy system. The freed pages will be poisoned with pattern
2104 * "poison" if it's within range [0, UCHAR_MAX].
2105 * Return pages freed into the buddy system.
2107 extern unsigned long free_reserved_area(void *start, void *end,
2108 int poison, const char *s);
2110 #ifdef CONFIG_HIGHMEM
2112 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2113 * and totalram_pages.
2115 extern void free_highmem_page(struct page *page);
2118 extern void adjust_managed_page_count(struct page *page, long count);
2119 extern void mem_init_print_info(const char *str);
2121 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2123 /* Free the reserved page into the buddy system, so it gets managed. */
2124 static inline void __free_reserved_page(struct page *page)
2126 ClearPageReserved(page);
2127 init_page_count(page);
2131 static inline void free_reserved_page(struct page *page)
2133 __free_reserved_page(page);
2134 adjust_managed_page_count(page, 1);
2137 static inline void mark_page_reserved(struct page *page)
2139 SetPageReserved(page);
2140 adjust_managed_page_count(page, -1);
2144 * Default method to free all the __init memory into the buddy system.
2145 * The freed pages will be poisoned with pattern "poison" if it's within
2146 * range [0, UCHAR_MAX].
2147 * Return pages freed into the buddy system.
2149 static inline unsigned long free_initmem_default(int poison)
2151 extern char __init_begin[], __init_end[];
2153 return free_reserved_area(&__init_begin, &__init_end,
2154 poison, "unused kernel");
2157 static inline unsigned long get_num_physpages(void)
2160 unsigned long phys_pages = 0;
2162 for_each_online_node(nid)
2163 phys_pages += node_present_pages(nid);
2168 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2170 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2171 * zones, allocate the backing mem_map and account for memory holes in a more
2172 * architecture independent manner. This is a substitute for creating the
2173 * zone_sizes[] and zholes_size[] arrays and passing them to
2174 * free_area_init_node()
2176 * An architecture is expected to register range of page frames backed by
2177 * physical memory with memblock_add[_node]() before calling
2178 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2179 * usage, an architecture is expected to do something like
2181 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2183 * for_each_valid_physical_page_range()
2184 * memblock_add_node(base, size, nid)
2185 * free_area_init_nodes(max_zone_pfns);
2187 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2188 * registered physical page range. Similarly
2189 * sparse_memory_present_with_active_regions() calls memory_present() for
2190 * each range when SPARSEMEM is enabled.
2192 * See mm/page_alloc.c for more information on each function exposed by
2193 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2195 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2196 unsigned long node_map_pfn_alignment(void);
2197 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2198 unsigned long end_pfn);
2199 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2200 unsigned long end_pfn);
2201 extern void get_pfn_range_for_nid(unsigned int nid,
2202 unsigned long *start_pfn, unsigned long *end_pfn);
2203 extern unsigned long find_min_pfn_with_active_regions(void);
2204 extern void free_bootmem_with_active_regions(int nid,
2205 unsigned long max_low_pfn);
2206 extern void sparse_memory_present_with_active_regions(int nid);
2208 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2210 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2211 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2212 static inline int __early_pfn_to_nid(unsigned long pfn,
2213 struct mminit_pfnnid_cache *state)
2218 /* please see mm/page_alloc.c */
2219 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2220 /* there is a per-arch backend function. */
2221 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2222 struct mminit_pfnnid_cache *state);
2225 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2226 void zero_resv_unavail(void);
2228 static inline void zero_resv_unavail(void) {}
2231 extern void set_dma_reserve(unsigned long new_dma_reserve);
2232 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2233 enum memmap_context, struct vmem_altmap *);
2234 extern void setup_per_zone_wmarks(void);
2235 extern int __meminit init_per_zone_wmark_min(void);
2236 extern void mem_init(void);
2237 extern void __init mmap_init(void);
2238 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2239 extern long si_mem_available(void);
2240 extern void si_meminfo(struct sysinfo * val);
2241 extern void si_meminfo_node(struct sysinfo *val, int nid);
2242 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2243 extern unsigned long arch_reserved_kernel_pages(void);
2246 extern __printf(3, 4)
2247 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2249 extern void setup_per_cpu_pageset(void);
2251 extern void zone_pcp_update(struct zone *zone);
2252 extern void zone_pcp_reset(struct zone *zone);
2255 extern int min_free_kbytes;
2256 extern int watermark_boost_factor;
2257 extern int watermark_scale_factor;
2260 extern atomic_long_t mmap_pages_allocated;
2261 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2263 /* interval_tree.c */
2264 void vma_interval_tree_insert(struct vm_area_struct *node,
2265 struct rb_root_cached *root);
2266 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2267 struct vm_area_struct *prev,
2268 struct rb_root_cached *root);
2269 void vma_interval_tree_remove(struct vm_area_struct *node,
2270 struct rb_root_cached *root);
2271 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2272 unsigned long start, unsigned long last);
2273 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2274 unsigned long start, unsigned long last);
2276 #define vma_interval_tree_foreach(vma, root, start, last) \
2277 for (vma = vma_interval_tree_iter_first(root, start, last); \
2278 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2280 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2281 struct rb_root_cached *root);
2282 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2283 struct rb_root_cached *root);
2284 struct anon_vma_chain *
2285 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2286 unsigned long start, unsigned long last);
2287 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2288 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2289 #ifdef CONFIG_DEBUG_VM_RB
2290 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2293 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2294 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2295 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2298 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2299 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2300 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2301 struct vm_area_struct *expand);
2302 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2303 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2305 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2307 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2308 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2309 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2310 struct mempolicy *, struct vm_userfaultfd_ctx);
2311 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2312 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2313 unsigned long addr, int new_below);
2314 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2315 unsigned long addr, int new_below);
2316 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2317 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2318 struct rb_node **, struct rb_node *);
2319 extern void unlink_file_vma(struct vm_area_struct *);
2320 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2321 unsigned long addr, unsigned long len, pgoff_t pgoff,
2322 bool *need_rmap_locks);
2323 extern void exit_mmap(struct mm_struct *);
2325 static inline int check_data_rlimit(unsigned long rlim,
2327 unsigned long start,
2328 unsigned long end_data,
2329 unsigned long start_data)
2331 if (rlim < RLIM_INFINITY) {
2332 if (((new - start) + (end_data - start_data)) > rlim)
2339 extern int mm_take_all_locks(struct mm_struct *mm);
2340 extern void mm_drop_all_locks(struct mm_struct *mm);
2342 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2343 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2344 extern struct file *get_task_exe_file(struct task_struct *task);
2346 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2347 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2349 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2350 const struct vm_special_mapping *sm);
2351 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2352 unsigned long addr, unsigned long len,
2353 unsigned long flags,
2354 const struct vm_special_mapping *spec);
2355 /* This is an obsolete alternative to _install_special_mapping. */
2356 extern int install_special_mapping(struct mm_struct *mm,
2357 unsigned long addr, unsigned long len,
2358 unsigned long flags, struct page **pages);
2360 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2362 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2363 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2364 struct list_head *uf);
2365 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2366 unsigned long len, unsigned long prot, unsigned long flags,
2367 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2368 struct list_head *uf);
2369 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2370 struct list_head *uf, bool downgrade);
2371 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2372 struct list_head *uf);
2374 static inline unsigned long
2375 do_mmap_pgoff(struct file *file, unsigned long addr,
2376 unsigned long len, unsigned long prot, unsigned long flags,
2377 unsigned long pgoff, unsigned long *populate,
2378 struct list_head *uf)
2380 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2384 extern int __mm_populate(unsigned long addr, unsigned long len,
2386 static inline void mm_populate(unsigned long addr, unsigned long len)
2389 (void) __mm_populate(addr, len, 1);
2392 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2395 /* These take the mm semaphore themselves */
2396 extern int __must_check vm_brk(unsigned long, unsigned long);
2397 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2398 extern int vm_munmap(unsigned long, size_t);
2399 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2400 unsigned long, unsigned long,
2401 unsigned long, unsigned long);
2403 struct vm_unmapped_area_info {
2404 #define VM_UNMAPPED_AREA_TOPDOWN 1
2405 unsigned long flags;
2406 unsigned long length;
2407 unsigned long low_limit;
2408 unsigned long high_limit;
2409 unsigned long align_mask;
2410 unsigned long align_offset;
2413 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2414 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2417 * Search for an unmapped address range.
2419 * We are looking for a range that:
2420 * - does not intersect with any VMA;
2421 * - is contained within the [low_limit, high_limit) interval;
2422 * - is at least the desired size.
2423 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2425 static inline unsigned long
2426 vm_unmapped_area(struct vm_unmapped_area_info *info)
2428 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2429 return unmapped_area_topdown(info);
2431 return unmapped_area(info);
2435 extern void truncate_inode_pages(struct address_space *, loff_t);
2436 extern void truncate_inode_pages_range(struct address_space *,
2437 loff_t lstart, loff_t lend);
2438 extern void truncate_inode_pages_final(struct address_space *);
2440 /* generic vm_area_ops exported for stackable file systems */
2441 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2442 extern void filemap_map_pages(struct vm_fault *vmf,
2443 pgoff_t start_pgoff, pgoff_t end_pgoff);
2444 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2446 /* mm/page-writeback.c */
2447 int __must_check write_one_page(struct page *page);
2448 void task_dirty_inc(struct task_struct *tsk);
2451 #define VM_MAX_READAHEAD 128 /* kbytes */
2452 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2454 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2455 pgoff_t offset, unsigned long nr_to_read);
2457 void page_cache_sync_readahead(struct address_space *mapping,
2458 struct file_ra_state *ra,
2461 unsigned long size);
2463 void page_cache_async_readahead(struct address_space *mapping,
2464 struct file_ra_state *ra,
2468 unsigned long size);
2470 extern unsigned long stack_guard_gap;
2471 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2472 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2474 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2475 extern int expand_downwards(struct vm_area_struct *vma,
2476 unsigned long address);
2478 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2480 #define expand_upwards(vma, address) (0)
2483 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2484 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2485 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2486 struct vm_area_struct **pprev);
2488 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2489 NULL if none. Assume start_addr < end_addr. */
2490 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2492 struct vm_area_struct * vma = find_vma(mm,start_addr);
2494 if (vma && end_addr <= vma->vm_start)
2499 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2501 unsigned long vm_start = vma->vm_start;
2503 if (vma->vm_flags & VM_GROWSDOWN) {
2504 vm_start -= stack_guard_gap;
2505 if (vm_start > vma->vm_start)
2511 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2513 unsigned long vm_end = vma->vm_end;
2515 if (vma->vm_flags & VM_GROWSUP) {
2516 vm_end += stack_guard_gap;
2517 if (vm_end < vma->vm_end)
2518 vm_end = -PAGE_SIZE;
2523 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2525 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2528 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2529 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2530 unsigned long vm_start, unsigned long vm_end)
2532 struct vm_area_struct *vma = find_vma(mm, vm_start);
2534 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2540 static inline bool range_in_vma(struct vm_area_struct *vma,
2541 unsigned long start, unsigned long end)
2543 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2547 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2548 void vma_set_page_prot(struct vm_area_struct *vma);
2550 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2554 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2556 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2560 #ifdef CONFIG_NUMA_BALANCING
2561 unsigned long change_prot_numa(struct vm_area_struct *vma,
2562 unsigned long start, unsigned long end);
2565 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2566 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2567 unsigned long pfn, unsigned long size, pgprot_t);
2568 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2569 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2571 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2572 unsigned long pfn, pgprot_t pgprot);
2573 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2575 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2576 unsigned long addr, pfn_t pfn);
2577 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2579 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2580 unsigned long addr, struct page *page)
2582 int err = vm_insert_page(vma, addr, page);
2585 return VM_FAULT_OOM;
2586 if (err < 0 && err != -EBUSY)
2587 return VM_FAULT_SIGBUS;
2589 return VM_FAULT_NOPAGE;
2592 static inline vm_fault_t vmf_error(int err)
2595 return VM_FAULT_OOM;
2596 return VM_FAULT_SIGBUS;
2599 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2600 unsigned int foll_flags);
2602 #define FOLL_WRITE 0x01 /* check pte is writable */
2603 #define FOLL_TOUCH 0x02 /* mark page accessed */
2604 #define FOLL_GET 0x04 /* do get_page on page */
2605 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2606 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2607 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2608 * and return without waiting upon it */
2609 #define FOLL_POPULATE 0x40 /* fault in page */
2610 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2611 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2612 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2613 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2614 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2615 #define FOLL_MLOCK 0x1000 /* lock present pages */
2616 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2617 #define FOLL_COW 0x4000 /* internal GUP flag */
2618 #define FOLL_ANON 0x8000 /* don't do file mappings */
2620 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2622 if (vm_fault & VM_FAULT_OOM)
2624 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2625 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2626 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2631 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2633 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2634 unsigned long size, pte_fn_t fn, void *data);
2637 #ifdef CONFIG_PAGE_POISONING
2638 extern bool page_poisoning_enabled(void);
2639 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2641 static inline bool page_poisoning_enabled(void) { return false; }
2642 static inline void kernel_poison_pages(struct page *page, int numpages,
2646 #ifdef CONFIG_DEBUG_PAGEALLOC
2647 extern bool _debug_pagealloc_enabled;
2648 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2650 static inline bool debug_pagealloc_enabled(void)
2652 return _debug_pagealloc_enabled;
2656 kernel_map_pages(struct page *page, int numpages, int enable)
2658 if (!debug_pagealloc_enabled())
2661 __kernel_map_pages(page, numpages, enable);
2663 #ifdef CONFIG_HIBERNATION
2664 extern bool kernel_page_present(struct page *page);
2665 #endif /* CONFIG_HIBERNATION */
2666 #else /* CONFIG_DEBUG_PAGEALLOC */
2668 kernel_map_pages(struct page *page, int numpages, int enable) {}
2669 #ifdef CONFIG_HIBERNATION
2670 static inline bool kernel_page_present(struct page *page) { return true; }
2671 #endif /* CONFIG_HIBERNATION */
2672 static inline bool debug_pagealloc_enabled(void)
2676 #endif /* CONFIG_DEBUG_PAGEALLOC */
2678 #ifdef __HAVE_ARCH_GATE_AREA
2679 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2680 extern int in_gate_area_no_mm(unsigned long addr);
2681 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2683 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2687 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2688 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2692 #endif /* __HAVE_ARCH_GATE_AREA */
2694 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2696 #ifdef CONFIG_SYSCTL
2697 extern int sysctl_drop_caches;
2698 int drop_caches_sysctl_handler(struct ctl_table *, int,
2699 void __user *, size_t *, loff_t *);
2702 void drop_slab(void);
2703 void drop_slab_node(int nid);
2706 #define randomize_va_space 0
2708 extern int randomize_va_space;
2711 const char * arch_vma_name(struct vm_area_struct *vma);
2712 void print_vma_addr(char *prefix, unsigned long rip);
2714 void *sparse_buffer_alloc(unsigned long size);
2715 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2716 struct vmem_altmap *altmap);
2717 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2718 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2719 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2720 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2721 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2722 void *vmemmap_alloc_block(unsigned long size, int node);
2724 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2725 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2726 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2727 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2729 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2730 struct vmem_altmap *altmap);
2731 void vmemmap_populate_print_last(void);
2732 #ifdef CONFIG_MEMORY_HOTPLUG
2733 void vmemmap_free(unsigned long start, unsigned long end,
2734 struct vmem_altmap *altmap);
2736 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2737 unsigned long nr_pages);
2740 MF_COUNT_INCREASED = 1 << 0,
2741 MF_ACTION_REQUIRED = 1 << 1,
2742 MF_MUST_KILL = 1 << 2,
2743 MF_SOFT_OFFLINE = 1 << 3,
2745 extern int memory_failure(unsigned long pfn, int flags);
2746 extern void memory_failure_queue(unsigned long pfn, int flags);
2747 extern int unpoison_memory(unsigned long pfn);
2748 extern int get_hwpoison_page(struct page *page);
2749 #define put_hwpoison_page(page) put_page(page)
2750 extern int sysctl_memory_failure_early_kill;
2751 extern int sysctl_memory_failure_recovery;
2752 extern void shake_page(struct page *p, int access);
2753 extern atomic_long_t num_poisoned_pages __read_mostly;
2754 extern int soft_offline_page(struct page *page, int flags);
2758 * Error handlers for various types of pages.
2761 MF_IGNORED, /* Error: cannot be handled */
2762 MF_FAILED, /* Error: handling failed */
2763 MF_DELAYED, /* Will be handled later */
2764 MF_RECOVERED, /* Successfully recovered */
2767 enum mf_action_page_type {
2769 MF_MSG_KERNEL_HIGH_ORDER,
2771 MF_MSG_DIFFERENT_COMPOUND,
2772 MF_MSG_POISONED_HUGE,
2775 MF_MSG_NON_PMD_HUGE,
2776 MF_MSG_UNMAP_FAILED,
2777 MF_MSG_DIRTY_SWAPCACHE,
2778 MF_MSG_CLEAN_SWAPCACHE,
2779 MF_MSG_DIRTY_MLOCKED_LRU,
2780 MF_MSG_CLEAN_MLOCKED_LRU,
2781 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2782 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2785 MF_MSG_TRUNCATED_LRU,
2792 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2793 extern void clear_huge_page(struct page *page,
2794 unsigned long addr_hint,
2795 unsigned int pages_per_huge_page);
2796 extern void copy_user_huge_page(struct page *dst, struct page *src,
2797 unsigned long addr_hint,
2798 struct vm_area_struct *vma,
2799 unsigned int pages_per_huge_page);
2800 extern long copy_huge_page_from_user(struct page *dst_page,
2801 const void __user *usr_src,
2802 unsigned int pages_per_huge_page,
2803 bool allow_pagefault);
2804 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2806 extern struct page_ext_operations debug_guardpage_ops;
2808 #ifdef CONFIG_DEBUG_PAGEALLOC
2809 extern unsigned int _debug_guardpage_minorder;
2810 extern bool _debug_guardpage_enabled;
2812 static inline unsigned int debug_guardpage_minorder(void)
2814 return _debug_guardpage_minorder;
2817 static inline bool debug_guardpage_enabled(void)
2819 return _debug_guardpage_enabled;
2822 static inline bool page_is_guard(struct page *page)
2824 struct page_ext *page_ext;
2826 if (!debug_guardpage_enabled())
2829 page_ext = lookup_page_ext(page);
2830 if (unlikely(!page_ext))
2833 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2836 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2837 static inline bool debug_guardpage_enabled(void) { return false; }
2838 static inline bool page_is_guard(struct page *page) { return false; }
2839 #endif /* CONFIG_DEBUG_PAGEALLOC */
2841 #if MAX_NUMNODES > 1
2842 void __init setup_nr_node_ids(void);
2844 static inline void setup_nr_node_ids(void) {}
2847 #endif /* __KERNEL__ */
2848 #endif /* _LINUX_MM_H */