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 extern struct kmem_cache *vm_area_cachep;
161 extern struct rb_root nommu_region_tree;
162 extern struct rw_semaphore nommu_region_sem;
164 extern unsigned int kobjsize(const void *objp);
168 * vm_flags in vm_area_struct, see mm_types.h.
169 * When changing, update also include/trace/events/mmflags.h
171 #define VM_NONE 0x00000000
173 #define VM_READ 0x00000001 /* currently active flags */
174 #define VM_WRITE 0x00000002
175 #define VM_EXEC 0x00000004
176 #define VM_SHARED 0x00000008
178 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
179 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
180 #define VM_MAYWRITE 0x00000020
181 #define VM_MAYEXEC 0x00000040
182 #define VM_MAYSHARE 0x00000080
184 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
185 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
186 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
187 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
188 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
190 #define VM_LOCKED 0x00002000
191 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
193 /* Used by sys_madvise() */
194 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
195 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
197 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
198 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
199 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
200 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
201 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
202 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
203 #define VM_SYNC 0x00800000 /* Synchronous page faults */
204 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
205 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
206 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
208 #ifdef CONFIG_MEM_SOFT_DIRTY
209 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
211 # define VM_SOFTDIRTY 0
214 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
215 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
216 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
217 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
219 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
220 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
221 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
222 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
225 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
226 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
227 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
228 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
229 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
230 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
232 #ifdef CONFIG_ARCH_HAS_PKEYS
233 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
234 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
235 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
236 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
237 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
239 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
241 # define VM_PKEY_BIT4 0
243 #endif /* CONFIG_ARCH_HAS_PKEYS */
245 #if defined(CONFIG_X86)
246 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
247 #elif defined(CONFIG_PPC)
248 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
249 #elif defined(CONFIG_PARISC)
250 # define VM_GROWSUP VM_ARCH_1
251 #elif defined(CONFIG_IA64)
252 # define VM_GROWSUP VM_ARCH_1
253 #elif defined(CONFIG_SPARC64)
254 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
255 # define VM_ARCH_CLEAR VM_SPARC_ADI
256 #elif !defined(CONFIG_MMU)
257 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
260 #if defined(CONFIG_X86_INTEL_MPX)
261 /* MPX specific bounds table or bounds directory */
262 # define VM_MPX VM_HIGH_ARCH_4
264 # define VM_MPX VM_NONE
268 # define VM_GROWSUP VM_NONE
271 /* Bits set in the VMA until the stack is in its final location */
272 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
274 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
275 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
278 #ifdef CONFIG_STACK_GROWSUP
279 #define VM_STACK VM_GROWSUP
281 #define VM_STACK VM_GROWSDOWN
284 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
287 * Special vmas that are non-mergable, non-mlock()able.
288 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
290 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
292 /* This mask defines which mm->def_flags a process can inherit its parent */
293 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
295 /* This mask is used to clear all the VMA flags used by mlock */
296 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
298 /* Arch-specific flags to clear when updating VM flags on protection change */
299 #ifndef VM_ARCH_CLEAR
300 # define VM_ARCH_CLEAR VM_NONE
302 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
305 * mapping from the currently active vm_flags protection bits (the
306 * low four bits) to a page protection mask..
308 extern pgprot_t protection_map[16];
310 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
311 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
312 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
313 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
314 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
315 #define FAULT_FLAG_TRIED 0x20 /* Second try */
316 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
317 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
318 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
320 #define FAULT_FLAG_TRACE \
321 { FAULT_FLAG_WRITE, "WRITE" }, \
322 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
323 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
324 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
325 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
326 { FAULT_FLAG_TRIED, "TRIED" }, \
327 { FAULT_FLAG_USER, "USER" }, \
328 { FAULT_FLAG_REMOTE, "REMOTE" }, \
329 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
332 * vm_fault is filled by the the pagefault handler and passed to the vma's
333 * ->fault function. The vma's ->fault is responsible for returning a bitmask
334 * of VM_FAULT_xxx flags that give details about how the fault was handled.
336 * MM layer fills up gfp_mask for page allocations but fault handler might
337 * alter it if its implementation requires a different allocation context.
339 * pgoff should be used in favour of virtual_address, if possible.
342 struct vm_area_struct *vma; /* Target VMA */
343 unsigned int flags; /* FAULT_FLAG_xxx flags */
344 gfp_t gfp_mask; /* gfp mask to be used for allocations */
345 pgoff_t pgoff; /* Logical page offset based on vma */
346 unsigned long address; /* Faulting virtual address */
347 pmd_t *pmd; /* Pointer to pmd entry matching
349 pud_t *pud; /* Pointer to pud entry matching
352 pte_t orig_pte; /* Value of PTE at the time of fault */
354 struct page *cow_page; /* Page handler may use for COW fault */
355 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
356 struct page *page; /* ->fault handlers should return a
357 * page here, unless VM_FAULT_NOPAGE
358 * is set (which is also implied by
361 /* These three entries are valid only while holding ptl lock */
362 pte_t *pte; /* Pointer to pte entry matching
363 * the 'address'. NULL if the page
364 * table hasn't been allocated.
366 spinlock_t *ptl; /* Page table lock.
367 * Protects pte page table if 'pte'
368 * is not NULL, otherwise pmd.
370 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
371 * vm_ops->map_pages() calls
372 * alloc_set_pte() from atomic context.
373 * do_fault_around() pre-allocates
374 * page table to avoid allocation from
379 /* page entry size for vm->huge_fault() */
380 enum page_entry_size {
387 * These are the virtual MM functions - opening of an area, closing and
388 * unmapping it (needed to keep files on disk up-to-date etc), pointer
389 * to the functions called when a no-page or a wp-page exception occurs.
391 struct vm_operations_struct {
392 void (*open)(struct vm_area_struct * area);
393 void (*close)(struct vm_area_struct * area);
394 int (*split)(struct vm_area_struct * area, unsigned long addr);
395 int (*mremap)(struct vm_area_struct * area);
396 vm_fault_t (*fault)(struct vm_fault *vmf);
397 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
398 enum page_entry_size pe_size);
399 void (*map_pages)(struct vm_fault *vmf,
400 pgoff_t start_pgoff, pgoff_t end_pgoff);
401 unsigned long (*pagesize)(struct vm_area_struct * area);
403 /* notification that a previously read-only page is about to become
404 * writable, if an error is returned it will cause a SIGBUS */
405 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
407 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
408 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
410 /* called by access_process_vm when get_user_pages() fails, typically
411 * for use by special VMAs that can switch between memory and hardware
413 int (*access)(struct vm_area_struct *vma, unsigned long addr,
414 void *buf, int len, int write);
416 /* Called by the /proc/PID/maps code to ask the vma whether it
417 * has a special name. Returning non-NULL will also cause this
418 * vma to be dumped unconditionally. */
419 const char *(*name)(struct vm_area_struct *vma);
423 * set_policy() op must add a reference to any non-NULL @new mempolicy
424 * to hold the policy upon return. Caller should pass NULL @new to
425 * remove a policy and fall back to surrounding context--i.e. do not
426 * install a MPOL_DEFAULT policy, nor the task or system default
429 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
432 * get_policy() op must add reference [mpol_get()] to any policy at
433 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
434 * in mm/mempolicy.c will do this automatically.
435 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
436 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
437 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
438 * must return NULL--i.e., do not "fallback" to task or system default
441 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
445 * Called by vm_normal_page() for special PTEs to find the
446 * page for @addr. This is useful if the default behavior
447 * (using pte_page()) would not find the correct page.
449 struct page *(*find_special_page)(struct vm_area_struct *vma,
456 #define page_private(page) ((page)->private)
457 #define set_page_private(page, v) ((page)->private = (v))
459 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
460 static inline int pmd_devmap(pmd_t pmd)
464 static inline int pud_devmap(pud_t pud)
468 static inline int pgd_devmap(pgd_t pgd)
475 * FIXME: take this include out, include page-flags.h in
476 * files which need it (119 of them)
478 #include <linux/page-flags.h>
479 #include <linux/huge_mm.h>
482 * Methods to modify the page usage count.
484 * What counts for a page usage:
485 * - cache mapping (page->mapping)
486 * - private data (page->private)
487 * - page mapped in a task's page tables, each mapping
488 * is counted separately
490 * Also, many kernel routines increase the page count before a critical
491 * routine so they can be sure the page doesn't go away from under them.
495 * Drop a ref, return true if the refcount fell to zero (the page has no users)
497 static inline int put_page_testzero(struct page *page)
499 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
500 return page_ref_dec_and_test(page);
504 * Try to grab a ref unless the page has a refcount of zero, return false if
506 * This can be called when MMU is off so it must not access
507 * any of the virtual mappings.
509 static inline int get_page_unless_zero(struct page *page)
511 return page_ref_add_unless(page, 1, 0);
514 extern int page_is_ram(unsigned long pfn);
522 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
525 /* Support for virtually mapped pages */
526 struct page *vmalloc_to_page(const void *addr);
527 unsigned long vmalloc_to_pfn(const void *addr);
530 * Determine if an address is within the vmalloc range
532 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
533 * is no special casing required.
535 static inline bool is_vmalloc_addr(const void *x)
538 unsigned long addr = (unsigned long)x;
540 return addr >= VMALLOC_START && addr < VMALLOC_END;
546 extern int is_vmalloc_or_module_addr(const void *x);
548 static inline int is_vmalloc_or_module_addr(const void *x)
554 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
555 static inline void *kvmalloc(size_t size, gfp_t flags)
557 return kvmalloc_node(size, flags, NUMA_NO_NODE);
559 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
561 return kvmalloc_node(size, flags | __GFP_ZERO, node);
563 static inline void *kvzalloc(size_t size, gfp_t flags)
565 return kvmalloc(size, flags | __GFP_ZERO);
568 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
572 if (unlikely(check_mul_overflow(n, size, &bytes)))
575 return kvmalloc(bytes, flags);
578 extern void kvfree(const void *addr);
580 static inline atomic_t *compound_mapcount_ptr(struct page *page)
582 return &page[1].compound_mapcount;
585 static inline int compound_mapcount(struct page *page)
587 VM_BUG_ON_PAGE(!PageCompound(page), page);
588 page = compound_head(page);
589 return atomic_read(compound_mapcount_ptr(page)) + 1;
593 * The atomic page->_mapcount, starts from -1: so that transitions
594 * both from it and to it can be tracked, using atomic_inc_and_test
595 * and atomic_add_negative(-1).
597 static inline void page_mapcount_reset(struct page *page)
599 atomic_set(&(page)->_mapcount, -1);
602 int __page_mapcount(struct page *page);
604 static inline int page_mapcount(struct page *page)
606 VM_BUG_ON_PAGE(PageSlab(page), page);
608 if (unlikely(PageCompound(page)))
609 return __page_mapcount(page);
610 return atomic_read(&page->_mapcount) + 1;
613 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
614 int total_mapcount(struct page *page);
615 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
617 static inline int total_mapcount(struct page *page)
619 return page_mapcount(page);
621 static inline int page_trans_huge_mapcount(struct page *page,
624 int mapcount = page_mapcount(page);
626 *total_mapcount = mapcount;
631 static inline struct page *virt_to_head_page(const void *x)
633 struct page *page = virt_to_page(x);
635 return compound_head(page);
638 void __put_page(struct page *page);
640 void put_pages_list(struct list_head *pages);
642 void split_page(struct page *page, unsigned int order);
645 * Compound pages have a destructor function. Provide a
646 * prototype for that function and accessor functions.
647 * These are _only_ valid on the head of a compound page.
649 typedef void compound_page_dtor(struct page *);
651 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
652 enum compound_dtor_id {
655 #ifdef CONFIG_HUGETLB_PAGE
658 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
663 extern compound_page_dtor * const compound_page_dtors[];
665 static inline void set_compound_page_dtor(struct page *page,
666 enum compound_dtor_id compound_dtor)
668 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
669 page[1].compound_dtor = compound_dtor;
672 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
674 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
675 return compound_page_dtors[page[1].compound_dtor];
678 static inline unsigned int compound_order(struct page *page)
682 return page[1].compound_order;
685 static inline void set_compound_order(struct page *page, unsigned int order)
687 page[1].compound_order = order;
690 void free_compound_page(struct page *page);
694 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
695 * servicing faults for write access. In the normal case, do always want
696 * pte_mkwrite. But get_user_pages can cause write faults for mappings
697 * that do not have writing enabled, when used by access_process_vm.
699 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
701 if (likely(vma->vm_flags & VM_WRITE))
702 pte = pte_mkwrite(pte);
706 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
708 int finish_fault(struct vm_fault *vmf);
709 int finish_mkwrite_fault(struct vm_fault *vmf);
713 * Multiple processes may "see" the same page. E.g. for untouched
714 * mappings of /dev/null, all processes see the same page full of
715 * zeroes, and text pages of executables and shared libraries have
716 * only one copy in memory, at most, normally.
718 * For the non-reserved pages, page_count(page) denotes a reference count.
719 * page_count() == 0 means the page is free. page->lru is then used for
720 * freelist management in the buddy allocator.
721 * page_count() > 0 means the page has been allocated.
723 * Pages are allocated by the slab allocator in order to provide memory
724 * to kmalloc and kmem_cache_alloc. In this case, the management of the
725 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
726 * unless a particular usage is carefully commented. (the responsibility of
727 * freeing the kmalloc memory is the caller's, of course).
729 * A page may be used by anyone else who does a __get_free_page().
730 * In this case, page_count still tracks the references, and should only
731 * be used through the normal accessor functions. The top bits of page->flags
732 * and page->virtual store page management information, but all other fields
733 * are unused and could be used privately, carefully. The management of this
734 * page is the responsibility of the one who allocated it, and those who have
735 * subsequently been given references to it.
737 * The other pages (we may call them "pagecache pages") are completely
738 * managed by the Linux memory manager: I/O, buffers, swapping etc.
739 * The following discussion applies only to them.
741 * A pagecache page contains an opaque `private' member, which belongs to the
742 * page's address_space. Usually, this is the address of a circular list of
743 * the page's disk buffers. PG_private must be set to tell the VM to call
744 * into the filesystem to release these pages.
746 * A page may belong to an inode's memory mapping. In this case, page->mapping
747 * is the pointer to the inode, and page->index is the file offset of the page,
748 * in units of PAGE_SIZE.
750 * If pagecache pages are not associated with an inode, they are said to be
751 * anonymous pages. These may become associated with the swapcache, and in that
752 * case PG_swapcache is set, and page->private is an offset into the swapcache.
754 * In either case (swapcache or inode backed), the pagecache itself holds one
755 * reference to the page. Setting PG_private should also increment the
756 * refcount. The each user mapping also has a reference to the page.
758 * The pagecache pages are stored in a per-mapping radix tree, which is
759 * rooted at mapping->i_pages, and indexed by offset.
760 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
761 * lists, we instead now tag pages as dirty/writeback in the radix tree.
763 * All pagecache pages may be subject to I/O:
764 * - inode pages may need to be read from disk,
765 * - inode pages which have been modified and are MAP_SHARED may need
766 * to be written back to the inode on disk,
767 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
768 * modified may need to be swapped out to swap space and (later) to be read
773 * The zone field is never updated after free_area_init_core()
774 * sets it, so none of the operations on it need to be atomic.
777 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
778 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
779 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
780 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
781 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
784 * Define the bit shifts to access each section. For non-existent
785 * sections we define the shift as 0; that plus a 0 mask ensures
786 * the compiler will optimise away reference to them.
788 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
789 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
790 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
791 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
793 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
794 #ifdef NODE_NOT_IN_PAGE_FLAGS
795 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
796 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
797 SECTIONS_PGOFF : ZONES_PGOFF)
799 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
800 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
801 NODES_PGOFF : ZONES_PGOFF)
804 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
806 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
807 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
810 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
811 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
812 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
813 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
814 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
816 static inline enum zone_type page_zonenum(const struct page *page)
818 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
821 #ifdef CONFIG_ZONE_DEVICE
822 static inline bool is_zone_device_page(const struct page *page)
824 return page_zonenum(page) == ZONE_DEVICE;
827 static inline bool is_zone_device_page(const struct page *page)
833 #ifdef CONFIG_DEV_PAGEMAP_OPS
834 void dev_pagemap_get_ops(void);
835 void dev_pagemap_put_ops(void);
836 void __put_devmap_managed_page(struct page *page);
837 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
838 static inline bool put_devmap_managed_page(struct page *page)
840 if (!static_branch_unlikely(&devmap_managed_key))
842 if (!is_zone_device_page(page))
844 switch (page->pgmap->type) {
845 case MEMORY_DEVICE_PRIVATE:
846 case MEMORY_DEVICE_PUBLIC:
847 case MEMORY_DEVICE_FS_DAX:
848 __put_devmap_managed_page(page);
856 static inline bool is_device_private_page(const struct page *page)
858 return is_zone_device_page(page) &&
859 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
862 static inline bool is_device_public_page(const struct page *page)
864 return is_zone_device_page(page) &&
865 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
868 #else /* CONFIG_DEV_PAGEMAP_OPS */
869 static inline void dev_pagemap_get_ops(void)
873 static inline void dev_pagemap_put_ops(void)
877 static inline bool put_devmap_managed_page(struct page *page)
882 static inline bool is_device_private_page(const struct page *page)
887 static inline bool is_device_public_page(const struct page *page)
891 #endif /* CONFIG_DEV_PAGEMAP_OPS */
893 static inline void get_page(struct page *page)
895 page = compound_head(page);
897 * Getting a normal page or the head of a compound page
898 * requires to already have an elevated page->_refcount.
900 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
904 static inline void put_page(struct page *page)
906 page = compound_head(page);
909 * For devmap managed pages we need to catch refcount transition from
910 * 2 to 1, when refcount reach one it means the page is free and we
911 * need to inform the device driver through callback. See
912 * include/linux/memremap.h and HMM for details.
914 if (put_devmap_managed_page(page))
917 if (put_page_testzero(page))
921 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
922 #define SECTION_IN_PAGE_FLAGS
926 * The identification function is mainly used by the buddy allocator for
927 * determining if two pages could be buddies. We are not really identifying
928 * the zone since we could be using the section number id if we do not have
929 * node id available in page flags.
930 * We only guarantee that it will return the same value for two combinable
933 static inline int page_zone_id(struct page *page)
935 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
938 static inline int zone_to_nid(struct zone *zone)
947 #ifdef NODE_NOT_IN_PAGE_FLAGS
948 extern int page_to_nid(const struct page *page);
950 static inline int page_to_nid(const struct page *page)
952 struct page *p = (struct page *)page;
954 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
958 #ifdef CONFIG_NUMA_BALANCING
959 static inline int cpu_pid_to_cpupid(int cpu, int pid)
961 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
964 static inline int cpupid_to_pid(int cpupid)
966 return cpupid & LAST__PID_MASK;
969 static inline int cpupid_to_cpu(int cpupid)
971 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
974 static inline int cpupid_to_nid(int cpupid)
976 return cpu_to_node(cpupid_to_cpu(cpupid));
979 static inline bool cpupid_pid_unset(int cpupid)
981 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
984 static inline bool cpupid_cpu_unset(int cpupid)
986 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
989 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
991 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
994 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
995 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
996 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
998 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1001 static inline int page_cpupid_last(struct page *page)
1003 return page->_last_cpupid;
1005 static inline void page_cpupid_reset_last(struct page *page)
1007 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1010 static inline int page_cpupid_last(struct page *page)
1012 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1015 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1017 static inline void page_cpupid_reset_last(struct page *page)
1019 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1021 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1022 #else /* !CONFIG_NUMA_BALANCING */
1023 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1025 return page_to_nid(page); /* XXX */
1028 static inline int page_cpupid_last(struct page *page)
1030 return page_to_nid(page); /* XXX */
1033 static inline int cpupid_to_nid(int cpupid)
1038 static inline int cpupid_to_pid(int cpupid)
1043 static inline int cpupid_to_cpu(int cpupid)
1048 static inline int cpu_pid_to_cpupid(int nid, int pid)
1053 static inline bool cpupid_pid_unset(int cpupid)
1058 static inline void page_cpupid_reset_last(struct page *page)
1062 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1066 #endif /* CONFIG_NUMA_BALANCING */
1068 static inline struct zone *page_zone(const struct page *page)
1070 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1073 static inline pg_data_t *page_pgdat(const struct page *page)
1075 return NODE_DATA(page_to_nid(page));
1078 #ifdef SECTION_IN_PAGE_FLAGS
1079 static inline void set_page_section(struct page *page, unsigned long section)
1081 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1082 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1085 static inline unsigned long page_to_section(const struct page *page)
1087 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1091 static inline void set_page_zone(struct page *page, enum zone_type zone)
1093 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1094 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1097 static inline void set_page_node(struct page *page, unsigned long node)
1099 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1100 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1103 static inline void set_page_links(struct page *page, enum zone_type zone,
1104 unsigned long node, unsigned long pfn)
1106 set_page_zone(page, zone);
1107 set_page_node(page, node);
1108 #ifdef SECTION_IN_PAGE_FLAGS
1109 set_page_section(page, pfn_to_section_nr(pfn));
1114 static inline struct mem_cgroup *page_memcg(struct page *page)
1116 return page->mem_cgroup;
1118 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1120 WARN_ON_ONCE(!rcu_read_lock_held());
1121 return READ_ONCE(page->mem_cgroup);
1124 static inline struct mem_cgroup *page_memcg(struct page *page)
1128 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1130 WARN_ON_ONCE(!rcu_read_lock_held());
1136 * Some inline functions in vmstat.h depend on page_zone()
1138 #include <linux/vmstat.h>
1140 static __always_inline void *lowmem_page_address(const struct page *page)
1142 return page_to_virt(page);
1145 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1146 #define HASHED_PAGE_VIRTUAL
1149 #if defined(WANT_PAGE_VIRTUAL)
1150 static inline void *page_address(const struct page *page)
1152 return page->virtual;
1154 static inline void set_page_address(struct page *page, void *address)
1156 page->virtual = address;
1158 #define page_address_init() do { } while(0)
1161 #if defined(HASHED_PAGE_VIRTUAL)
1162 void *page_address(const struct page *page);
1163 void set_page_address(struct page *page, void *virtual);
1164 void page_address_init(void);
1167 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1168 #define page_address(page) lowmem_page_address(page)
1169 #define set_page_address(page, address) do { } while(0)
1170 #define page_address_init() do { } while(0)
1173 extern void *page_rmapping(struct page *page);
1174 extern struct anon_vma *page_anon_vma(struct page *page);
1175 extern struct address_space *page_mapping(struct page *page);
1177 extern struct address_space *__page_file_mapping(struct page *);
1180 struct address_space *page_file_mapping(struct page *page)
1182 if (unlikely(PageSwapCache(page)))
1183 return __page_file_mapping(page);
1185 return page->mapping;
1188 extern pgoff_t __page_file_index(struct page *page);
1191 * Return the pagecache index of the passed page. Regular pagecache pages
1192 * use ->index whereas swapcache pages use swp_offset(->private)
1194 static inline pgoff_t page_index(struct page *page)
1196 if (unlikely(PageSwapCache(page)))
1197 return __page_file_index(page);
1201 bool page_mapped(struct page *page);
1202 struct address_space *page_mapping(struct page *page);
1203 struct address_space *page_mapping_file(struct page *page);
1206 * Return true only if the page has been allocated with
1207 * ALLOC_NO_WATERMARKS and the low watermark was not
1208 * met implying that the system is under some pressure.
1210 static inline bool page_is_pfmemalloc(struct page *page)
1213 * Page index cannot be this large so this must be
1214 * a pfmemalloc page.
1216 return page->index == -1UL;
1220 * Only to be called by the page allocator on a freshly allocated
1223 static inline void set_page_pfmemalloc(struct page *page)
1228 static inline void clear_page_pfmemalloc(struct page *page)
1234 * Different kinds of faults, as returned by handle_mm_fault().
1235 * Used to decide whether a process gets delivered SIGBUS or
1236 * just gets major/minor fault counters bumped up.
1239 #define VM_FAULT_OOM 0x0001
1240 #define VM_FAULT_SIGBUS 0x0002
1241 #define VM_FAULT_MAJOR 0x0004
1242 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1243 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1244 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1245 #define VM_FAULT_SIGSEGV 0x0040
1247 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1248 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1249 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1250 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1251 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1252 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1253 * and needs fsync() to complete (for
1254 * synchronous page faults in DAX) */
1256 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1257 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1260 #define VM_FAULT_RESULT_TRACE \
1261 { VM_FAULT_OOM, "OOM" }, \
1262 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1263 { VM_FAULT_MAJOR, "MAJOR" }, \
1264 { VM_FAULT_WRITE, "WRITE" }, \
1265 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1266 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1267 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1268 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1269 { VM_FAULT_LOCKED, "LOCKED" }, \
1270 { VM_FAULT_RETRY, "RETRY" }, \
1271 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1272 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1273 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1275 /* Encode hstate index for a hwpoisoned large page */
1276 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1277 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1280 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1282 extern void pagefault_out_of_memory(void);
1284 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1287 * Flags passed to show_mem() and show_free_areas() to suppress output in
1290 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1292 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1294 extern bool can_do_mlock(void);
1295 extern int user_shm_lock(size_t, struct user_struct *);
1296 extern void user_shm_unlock(size_t, struct user_struct *);
1299 * Parameter block passed down to zap_pte_range in exceptional cases.
1301 struct zap_details {
1302 struct address_space *check_mapping; /* Check page->mapping if set */
1303 pgoff_t first_index; /* Lowest page->index to unmap */
1304 pgoff_t last_index; /* Highest page->index to unmap */
1307 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1308 pte_t pte, bool with_public_device);
1309 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1311 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1314 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1315 unsigned long size);
1316 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1317 unsigned long size);
1318 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1319 unsigned long start, unsigned long end);
1322 * mm_walk - callbacks for walk_page_range
1323 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1324 * this handler should only handle pud_trans_huge() puds.
1325 * the pmd_entry or pte_entry callbacks will be used for
1327 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1328 * this handler is required to be able to handle
1329 * pmd_trans_huge() pmds. They may simply choose to
1330 * split_huge_page() instead of handling it explicitly.
1331 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1332 * @pte_hole: if set, called for each hole at all levels
1333 * @hugetlb_entry: if set, called for each hugetlb entry
1334 * @test_walk: caller specific callback function to determine whether
1335 * we walk over the current vma or not. Returning 0
1336 * value means "do page table walk over the current vma,"
1337 * and a negative one means "abort current page table walk
1338 * right now." 1 means "skip the current vma."
1339 * @mm: mm_struct representing the target process of page table walk
1340 * @vma: vma currently walked (NULL if walking outside vmas)
1341 * @private: private data for callbacks' usage
1343 * (see the comment on walk_page_range() for more details)
1346 int (*pud_entry)(pud_t *pud, unsigned long addr,
1347 unsigned long next, struct mm_walk *walk);
1348 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1349 unsigned long next, struct mm_walk *walk);
1350 int (*pte_entry)(pte_t *pte, unsigned long addr,
1351 unsigned long next, struct mm_walk *walk);
1352 int (*pte_hole)(unsigned long addr, unsigned long next,
1353 struct mm_walk *walk);
1354 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1355 unsigned long addr, unsigned long next,
1356 struct mm_walk *walk);
1357 int (*test_walk)(unsigned long addr, unsigned long next,
1358 struct mm_walk *walk);
1359 struct mm_struct *mm;
1360 struct vm_area_struct *vma;
1364 int walk_page_range(unsigned long addr, unsigned long end,
1365 struct mm_walk *walk);
1366 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1367 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1368 unsigned long end, unsigned long floor, unsigned long ceiling);
1369 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1370 struct vm_area_struct *vma);
1371 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1372 unsigned long *start, unsigned long *end,
1373 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1374 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1375 unsigned long *pfn);
1376 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1377 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1378 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1379 void *buf, int len, int write);
1381 extern void truncate_pagecache(struct inode *inode, loff_t new);
1382 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1383 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1384 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1385 int truncate_inode_page(struct address_space *mapping, struct page *page);
1386 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1387 int invalidate_inode_page(struct page *page);
1390 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1391 unsigned int flags);
1392 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1393 unsigned long address, unsigned int fault_flags,
1395 void unmap_mapping_pages(struct address_space *mapping,
1396 pgoff_t start, pgoff_t nr, bool even_cows);
1397 void unmap_mapping_range(struct address_space *mapping,
1398 loff_t const holebegin, loff_t const holelen, int even_cows);
1400 static inline int handle_mm_fault(struct vm_area_struct *vma,
1401 unsigned long address, unsigned int flags)
1403 /* should never happen if there's no MMU */
1405 return VM_FAULT_SIGBUS;
1407 static inline int fixup_user_fault(struct task_struct *tsk,
1408 struct mm_struct *mm, unsigned long address,
1409 unsigned int fault_flags, bool *unlocked)
1411 /* should never happen if there's no MMU */
1415 static inline void unmap_mapping_pages(struct address_space *mapping,
1416 pgoff_t start, pgoff_t nr, bool even_cows) { }
1417 static inline void unmap_mapping_range(struct address_space *mapping,
1418 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1421 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1422 loff_t const holebegin, loff_t const holelen)
1424 unmap_mapping_range(mapping, holebegin, holelen, 0);
1427 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1428 void *buf, int len, unsigned int gup_flags);
1429 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1430 void *buf, int len, unsigned int gup_flags);
1431 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1432 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1434 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1435 unsigned long start, unsigned long nr_pages,
1436 unsigned int gup_flags, struct page **pages,
1437 struct vm_area_struct **vmas, int *locked);
1438 long get_user_pages(unsigned long start, unsigned long nr_pages,
1439 unsigned int gup_flags, struct page **pages,
1440 struct vm_area_struct **vmas);
1441 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1442 unsigned int gup_flags, struct page **pages, int *locked);
1443 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1444 struct page **pages, unsigned int gup_flags);
1445 #ifdef CONFIG_FS_DAX
1446 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1447 unsigned int gup_flags, struct page **pages,
1448 struct vm_area_struct **vmas);
1450 static inline long get_user_pages_longterm(unsigned long start,
1451 unsigned long nr_pages, unsigned int gup_flags,
1452 struct page **pages, struct vm_area_struct **vmas)
1454 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1456 #endif /* CONFIG_FS_DAX */
1458 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1459 struct page **pages);
1461 /* Container for pinned pfns / pages */
1462 struct frame_vector {
1463 unsigned int nr_allocated; /* Number of frames we have space for */
1464 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1465 bool got_ref; /* Did we pin pages by getting page ref? */
1466 bool is_pfns; /* Does array contain pages or pfns? */
1467 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1468 * pfns_vector_pages() or pfns_vector_pfns()
1472 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1473 void frame_vector_destroy(struct frame_vector *vec);
1474 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1475 unsigned int gup_flags, struct frame_vector *vec);
1476 void put_vaddr_frames(struct frame_vector *vec);
1477 int frame_vector_to_pages(struct frame_vector *vec);
1478 void frame_vector_to_pfns(struct frame_vector *vec);
1480 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1482 return vec->nr_frames;
1485 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1488 int err = frame_vector_to_pages(vec);
1491 return ERR_PTR(err);
1493 return (struct page **)(vec->ptrs);
1496 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1499 frame_vector_to_pfns(vec);
1500 return (unsigned long *)(vec->ptrs);
1504 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1505 struct page **pages);
1506 int get_kernel_page(unsigned long start, int write, struct page **pages);
1507 struct page *get_dump_page(unsigned long addr);
1509 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1510 extern void do_invalidatepage(struct page *page, unsigned int offset,
1511 unsigned int length);
1513 void __set_page_dirty(struct page *, struct address_space *, int warn);
1514 int __set_page_dirty_nobuffers(struct page *page);
1515 int __set_page_dirty_no_writeback(struct page *page);
1516 int redirty_page_for_writepage(struct writeback_control *wbc,
1518 void account_page_dirtied(struct page *page, struct address_space *mapping);
1519 void account_page_cleaned(struct page *page, struct address_space *mapping,
1520 struct bdi_writeback *wb);
1521 int set_page_dirty(struct page *page);
1522 int set_page_dirty_lock(struct page *page);
1523 void __cancel_dirty_page(struct page *page);
1524 static inline void cancel_dirty_page(struct page *page)
1526 /* Avoid atomic ops, locking, etc. when not actually needed. */
1527 if (PageDirty(page))
1528 __cancel_dirty_page(page);
1530 int clear_page_dirty_for_io(struct page *page);
1532 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1534 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1536 return !vma->vm_ops;
1541 * The vma_is_shmem is not inline because it is used only by slow
1542 * paths in userfault.
1544 bool vma_is_shmem(struct vm_area_struct *vma);
1546 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1549 int vma_is_stack_for_current(struct vm_area_struct *vma);
1551 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1552 unsigned long old_addr, struct vm_area_struct *new_vma,
1553 unsigned long new_addr, unsigned long len,
1554 bool need_rmap_locks);
1555 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1556 unsigned long end, pgprot_t newprot,
1557 int dirty_accountable, int prot_numa);
1558 extern int mprotect_fixup(struct vm_area_struct *vma,
1559 struct vm_area_struct **pprev, unsigned long start,
1560 unsigned long end, unsigned long newflags);
1563 * doesn't attempt to fault and will return short.
1565 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1566 struct page **pages);
1568 * per-process(per-mm_struct) statistics.
1570 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1572 long val = atomic_long_read(&mm->rss_stat.count[member]);
1574 #ifdef SPLIT_RSS_COUNTING
1576 * counter is updated in asynchronous manner and may go to minus.
1577 * But it's never be expected number for users.
1582 return (unsigned long)val;
1585 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1587 atomic_long_add(value, &mm->rss_stat.count[member]);
1590 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1592 atomic_long_inc(&mm->rss_stat.count[member]);
1595 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1597 atomic_long_dec(&mm->rss_stat.count[member]);
1600 /* Optimized variant when page is already known not to be PageAnon */
1601 static inline int mm_counter_file(struct page *page)
1603 if (PageSwapBacked(page))
1604 return MM_SHMEMPAGES;
1605 return MM_FILEPAGES;
1608 static inline int mm_counter(struct page *page)
1611 return MM_ANONPAGES;
1612 return mm_counter_file(page);
1615 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1617 return get_mm_counter(mm, MM_FILEPAGES) +
1618 get_mm_counter(mm, MM_ANONPAGES) +
1619 get_mm_counter(mm, MM_SHMEMPAGES);
1622 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1624 return max(mm->hiwater_rss, get_mm_rss(mm));
1627 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1629 return max(mm->hiwater_vm, mm->total_vm);
1632 static inline void update_hiwater_rss(struct mm_struct *mm)
1634 unsigned long _rss = get_mm_rss(mm);
1636 if ((mm)->hiwater_rss < _rss)
1637 (mm)->hiwater_rss = _rss;
1640 static inline void update_hiwater_vm(struct mm_struct *mm)
1642 if (mm->hiwater_vm < mm->total_vm)
1643 mm->hiwater_vm = mm->total_vm;
1646 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1648 mm->hiwater_rss = get_mm_rss(mm);
1651 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1652 struct mm_struct *mm)
1654 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1656 if (*maxrss < hiwater_rss)
1657 *maxrss = hiwater_rss;
1660 #if defined(SPLIT_RSS_COUNTING)
1661 void sync_mm_rss(struct mm_struct *mm);
1663 static inline void sync_mm_rss(struct mm_struct *mm)
1668 #ifndef __HAVE_ARCH_PTE_DEVMAP
1669 static inline int pte_devmap(pte_t pte)
1675 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1677 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1679 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1683 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1687 #ifdef __PAGETABLE_P4D_FOLDED
1688 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1689 unsigned long address)
1694 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1697 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1698 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1699 unsigned long address)
1703 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1704 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1707 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1709 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1711 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1714 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1716 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1720 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1721 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1722 unsigned long address)
1727 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1728 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1731 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1733 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1735 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1738 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1740 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1745 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1747 atomic_long_set(&mm->pgtables_bytes, 0);
1750 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1752 return atomic_long_read(&mm->pgtables_bytes);
1755 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1757 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1760 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1762 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1766 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1767 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1772 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1773 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1776 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1777 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1780 * The following ifdef needed to get the 4level-fixup.h header to work.
1781 * Remove it when 4level-fixup.h has been removed.
1783 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1785 #ifndef __ARCH_HAS_5LEVEL_HACK
1786 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1787 unsigned long address)
1789 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1790 NULL : p4d_offset(pgd, address);
1793 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1794 unsigned long address)
1796 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1797 NULL : pud_offset(p4d, address);
1799 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1801 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1803 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1804 NULL: pmd_offset(pud, address);
1806 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1808 #if USE_SPLIT_PTE_PTLOCKS
1809 #if ALLOC_SPLIT_PTLOCKS
1810 void __init ptlock_cache_init(void);
1811 extern bool ptlock_alloc(struct page *page);
1812 extern void ptlock_free(struct page *page);
1814 static inline spinlock_t *ptlock_ptr(struct page *page)
1818 #else /* ALLOC_SPLIT_PTLOCKS */
1819 static inline void ptlock_cache_init(void)
1823 static inline bool ptlock_alloc(struct page *page)
1828 static inline void ptlock_free(struct page *page)
1832 static inline spinlock_t *ptlock_ptr(struct page *page)
1836 #endif /* ALLOC_SPLIT_PTLOCKS */
1838 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1840 return ptlock_ptr(pmd_page(*pmd));
1843 static inline bool ptlock_init(struct page *page)
1846 * prep_new_page() initialize page->private (and therefore page->ptl)
1847 * with 0. Make sure nobody took it in use in between.
1849 * It can happen if arch try to use slab for page table allocation:
1850 * slab code uses page->slab_cache, which share storage with page->ptl.
1852 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1853 if (!ptlock_alloc(page))
1855 spin_lock_init(ptlock_ptr(page));
1859 /* Reset page->mapping so free_pages_check won't complain. */
1860 static inline void pte_lock_deinit(struct page *page)
1862 page->mapping = NULL;
1866 #else /* !USE_SPLIT_PTE_PTLOCKS */
1868 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1870 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1872 return &mm->page_table_lock;
1874 static inline void ptlock_cache_init(void) {}
1875 static inline bool ptlock_init(struct page *page) { return true; }
1876 static inline void pte_lock_deinit(struct page *page) {}
1877 #endif /* USE_SPLIT_PTE_PTLOCKS */
1879 static inline void pgtable_init(void)
1881 ptlock_cache_init();
1882 pgtable_cache_init();
1885 static inline bool pgtable_page_ctor(struct page *page)
1887 if (!ptlock_init(page))
1889 __SetPageTable(page);
1890 inc_zone_page_state(page, NR_PAGETABLE);
1894 static inline void pgtable_page_dtor(struct page *page)
1896 pte_lock_deinit(page);
1897 __ClearPageTable(page);
1898 dec_zone_page_state(page, NR_PAGETABLE);
1901 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1903 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1904 pte_t *__pte = pte_offset_map(pmd, address); \
1910 #define pte_unmap_unlock(pte, ptl) do { \
1915 #define pte_alloc(mm, pmd, address) \
1916 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1918 #define pte_alloc_map(mm, pmd, address) \
1919 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1921 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1922 (pte_alloc(mm, pmd, address) ? \
1923 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1925 #define pte_alloc_kernel(pmd, address) \
1926 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1927 NULL: pte_offset_kernel(pmd, address))
1929 #if USE_SPLIT_PMD_PTLOCKS
1931 static struct page *pmd_to_page(pmd_t *pmd)
1933 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1934 return virt_to_page((void *)((unsigned long) pmd & mask));
1937 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1939 return ptlock_ptr(pmd_to_page(pmd));
1942 static inline bool pgtable_pmd_page_ctor(struct page *page)
1944 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1945 page->pmd_huge_pte = NULL;
1947 return ptlock_init(page);
1950 static inline void pgtable_pmd_page_dtor(struct page *page)
1952 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1953 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1958 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1962 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1964 return &mm->page_table_lock;
1967 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1968 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1970 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1974 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1976 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1982 * No scalability reason to split PUD locks yet, but follow the same pattern
1983 * as the PMD locks to make it easier if we decide to. The VM should not be
1984 * considered ready to switch to split PUD locks yet; there may be places
1985 * which need to be converted from page_table_lock.
1987 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1989 return &mm->page_table_lock;
1992 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1994 spinlock_t *ptl = pud_lockptr(mm, pud);
2000 extern void __init pagecache_init(void);
2001 extern void free_area_init(unsigned long * zones_size);
2002 extern void free_area_init_node(int nid, unsigned long * zones_size,
2003 unsigned long zone_start_pfn, unsigned long *zholes_size);
2004 extern void free_initmem(void);
2007 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2008 * into the buddy system. The freed pages will be poisoned with pattern
2009 * "poison" if it's within range [0, UCHAR_MAX].
2010 * Return pages freed into the buddy system.
2012 extern unsigned long free_reserved_area(void *start, void *end,
2013 int poison, char *s);
2015 #ifdef CONFIG_HIGHMEM
2017 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2018 * and totalram_pages.
2020 extern void free_highmem_page(struct page *page);
2023 extern void adjust_managed_page_count(struct page *page, long count);
2024 extern void mem_init_print_info(const char *str);
2026 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2028 /* Free the reserved page into the buddy system, so it gets managed. */
2029 static inline void __free_reserved_page(struct page *page)
2031 ClearPageReserved(page);
2032 init_page_count(page);
2036 static inline void free_reserved_page(struct page *page)
2038 __free_reserved_page(page);
2039 adjust_managed_page_count(page, 1);
2042 static inline void mark_page_reserved(struct page *page)
2044 SetPageReserved(page);
2045 adjust_managed_page_count(page, -1);
2049 * Default method to free all the __init memory into the buddy system.
2050 * The freed pages will be poisoned with pattern "poison" if it's within
2051 * range [0, UCHAR_MAX].
2052 * Return pages freed into the buddy system.
2054 static inline unsigned long free_initmem_default(int poison)
2056 extern char __init_begin[], __init_end[];
2058 return free_reserved_area(&__init_begin, &__init_end,
2059 poison, "unused kernel");
2062 static inline unsigned long get_num_physpages(void)
2065 unsigned long phys_pages = 0;
2067 for_each_online_node(nid)
2068 phys_pages += node_present_pages(nid);
2073 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2075 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2076 * zones, allocate the backing mem_map and account for memory holes in a more
2077 * architecture independent manner. This is a substitute for creating the
2078 * zone_sizes[] and zholes_size[] arrays and passing them to
2079 * free_area_init_node()
2081 * An architecture is expected to register range of page frames backed by
2082 * physical memory with memblock_add[_node]() before calling
2083 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2084 * usage, an architecture is expected to do something like
2086 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2088 * for_each_valid_physical_page_range()
2089 * memblock_add_node(base, size, nid)
2090 * free_area_init_nodes(max_zone_pfns);
2092 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2093 * registered physical page range. Similarly
2094 * sparse_memory_present_with_active_regions() calls memory_present() for
2095 * each range when SPARSEMEM is enabled.
2097 * See mm/page_alloc.c for more information on each function exposed by
2098 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2100 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2101 unsigned long node_map_pfn_alignment(void);
2102 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2103 unsigned long end_pfn);
2104 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2105 unsigned long end_pfn);
2106 extern void get_pfn_range_for_nid(unsigned int nid,
2107 unsigned long *start_pfn, unsigned long *end_pfn);
2108 extern unsigned long find_min_pfn_with_active_regions(void);
2109 extern void free_bootmem_with_active_regions(int nid,
2110 unsigned long max_low_pfn);
2111 extern void sparse_memory_present_with_active_regions(int nid);
2113 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2115 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2116 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2117 static inline int __early_pfn_to_nid(unsigned long pfn,
2118 struct mminit_pfnnid_cache *state)
2123 /* please see mm/page_alloc.c */
2124 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2125 /* there is a per-arch backend function. */
2126 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2127 struct mminit_pfnnid_cache *state);
2130 #ifdef CONFIG_HAVE_MEMBLOCK
2131 void zero_resv_unavail(void);
2133 static inline void zero_resv_unavail(void) {}
2136 extern void set_dma_reserve(unsigned long new_dma_reserve);
2137 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2138 enum memmap_context, struct vmem_altmap *);
2139 extern void setup_per_zone_wmarks(void);
2140 extern int __meminit init_per_zone_wmark_min(void);
2141 extern void mem_init(void);
2142 extern void __init mmap_init(void);
2143 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2144 extern long si_mem_available(void);
2145 extern void si_meminfo(struct sysinfo * val);
2146 extern void si_meminfo_node(struct sysinfo *val, int nid);
2147 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2148 extern unsigned long arch_reserved_kernel_pages(void);
2151 extern __printf(3, 4)
2152 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2154 extern void setup_per_cpu_pageset(void);
2156 extern void zone_pcp_update(struct zone *zone);
2157 extern void zone_pcp_reset(struct zone *zone);
2160 extern int min_free_kbytes;
2161 extern int watermark_scale_factor;
2164 extern atomic_long_t mmap_pages_allocated;
2165 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2167 /* interval_tree.c */
2168 void vma_interval_tree_insert(struct vm_area_struct *node,
2169 struct rb_root_cached *root);
2170 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2171 struct vm_area_struct *prev,
2172 struct rb_root_cached *root);
2173 void vma_interval_tree_remove(struct vm_area_struct *node,
2174 struct rb_root_cached *root);
2175 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2176 unsigned long start, unsigned long last);
2177 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2178 unsigned long start, unsigned long last);
2180 #define vma_interval_tree_foreach(vma, root, start, last) \
2181 for (vma = vma_interval_tree_iter_first(root, start, last); \
2182 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2184 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2185 struct rb_root_cached *root);
2186 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2187 struct rb_root_cached *root);
2188 struct anon_vma_chain *
2189 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2190 unsigned long start, unsigned long last);
2191 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2192 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2193 #ifdef CONFIG_DEBUG_VM_RB
2194 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2197 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2198 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2199 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2202 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2203 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2204 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2205 struct vm_area_struct *expand);
2206 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2207 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2209 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2211 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2212 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2213 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2214 struct mempolicy *, struct vm_userfaultfd_ctx);
2215 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2216 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2217 unsigned long addr, int new_below);
2218 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2219 unsigned long addr, int new_below);
2220 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2221 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2222 struct rb_node **, struct rb_node *);
2223 extern void unlink_file_vma(struct vm_area_struct *);
2224 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2225 unsigned long addr, unsigned long len, pgoff_t pgoff,
2226 bool *need_rmap_locks);
2227 extern void exit_mmap(struct mm_struct *);
2229 static inline int check_data_rlimit(unsigned long rlim,
2231 unsigned long start,
2232 unsigned long end_data,
2233 unsigned long start_data)
2235 if (rlim < RLIM_INFINITY) {
2236 if (((new - start) + (end_data - start_data)) > rlim)
2243 extern int mm_take_all_locks(struct mm_struct *mm);
2244 extern void mm_drop_all_locks(struct mm_struct *mm);
2246 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2247 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2248 extern struct file *get_task_exe_file(struct task_struct *task);
2250 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2251 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2253 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2254 const struct vm_special_mapping *sm);
2255 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2256 unsigned long addr, unsigned long len,
2257 unsigned long flags,
2258 const struct vm_special_mapping *spec);
2259 /* This is an obsolete alternative to _install_special_mapping. */
2260 extern int install_special_mapping(struct mm_struct *mm,
2261 unsigned long addr, unsigned long len,
2262 unsigned long flags, struct page **pages);
2264 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2266 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2267 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2268 struct list_head *uf);
2269 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2270 unsigned long len, unsigned long prot, unsigned long flags,
2271 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2272 struct list_head *uf);
2273 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2274 struct list_head *uf);
2276 static inline unsigned long
2277 do_mmap_pgoff(struct file *file, unsigned long addr,
2278 unsigned long len, unsigned long prot, unsigned long flags,
2279 unsigned long pgoff, unsigned long *populate,
2280 struct list_head *uf)
2282 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2286 extern int __mm_populate(unsigned long addr, unsigned long len,
2288 static inline void mm_populate(unsigned long addr, unsigned long len)
2291 (void) __mm_populate(addr, len, 1);
2294 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2297 /* These take the mm semaphore themselves */
2298 extern int __must_check vm_brk(unsigned long, unsigned long);
2299 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2300 extern int vm_munmap(unsigned long, size_t);
2301 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2302 unsigned long, unsigned long,
2303 unsigned long, unsigned long);
2305 struct vm_unmapped_area_info {
2306 #define VM_UNMAPPED_AREA_TOPDOWN 1
2307 unsigned long flags;
2308 unsigned long length;
2309 unsigned long low_limit;
2310 unsigned long high_limit;
2311 unsigned long align_mask;
2312 unsigned long align_offset;
2315 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2316 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2319 * Search for an unmapped address range.
2321 * We are looking for a range that:
2322 * - does not intersect with any VMA;
2323 * - is contained within the [low_limit, high_limit) interval;
2324 * - is at least the desired size.
2325 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2327 static inline unsigned long
2328 vm_unmapped_area(struct vm_unmapped_area_info *info)
2330 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2331 return unmapped_area_topdown(info);
2333 return unmapped_area(info);
2337 extern void truncate_inode_pages(struct address_space *, loff_t);
2338 extern void truncate_inode_pages_range(struct address_space *,
2339 loff_t lstart, loff_t lend);
2340 extern void truncate_inode_pages_final(struct address_space *);
2342 /* generic vm_area_ops exported for stackable file systems */
2343 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2344 extern void filemap_map_pages(struct vm_fault *vmf,
2345 pgoff_t start_pgoff, pgoff_t end_pgoff);
2346 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2348 /* mm/page-writeback.c */
2349 int __must_check write_one_page(struct page *page);
2350 void task_dirty_inc(struct task_struct *tsk);
2353 #define VM_MAX_READAHEAD 128 /* kbytes */
2354 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2356 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2357 pgoff_t offset, unsigned long nr_to_read);
2359 void page_cache_sync_readahead(struct address_space *mapping,
2360 struct file_ra_state *ra,
2363 unsigned long size);
2365 void page_cache_async_readahead(struct address_space *mapping,
2366 struct file_ra_state *ra,
2370 unsigned long size);
2372 extern unsigned long stack_guard_gap;
2373 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2374 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2376 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2377 extern int expand_downwards(struct vm_area_struct *vma,
2378 unsigned long address);
2380 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2382 #define expand_upwards(vma, address) (0)
2385 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2386 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2387 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2388 struct vm_area_struct **pprev);
2390 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2391 NULL if none. Assume start_addr < end_addr. */
2392 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2394 struct vm_area_struct * vma = find_vma(mm,start_addr);
2396 if (vma && end_addr <= vma->vm_start)
2401 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2403 unsigned long vm_start = vma->vm_start;
2405 if (vma->vm_flags & VM_GROWSDOWN) {
2406 vm_start -= stack_guard_gap;
2407 if (vm_start > vma->vm_start)
2413 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2415 unsigned long vm_end = vma->vm_end;
2417 if (vma->vm_flags & VM_GROWSUP) {
2418 vm_end += stack_guard_gap;
2419 if (vm_end < vma->vm_end)
2420 vm_end = -PAGE_SIZE;
2425 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2427 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2430 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2431 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2432 unsigned long vm_start, unsigned long vm_end)
2434 struct vm_area_struct *vma = find_vma(mm, vm_start);
2436 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2443 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2444 void vma_set_page_prot(struct vm_area_struct *vma);
2446 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2450 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2452 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2456 #ifdef CONFIG_NUMA_BALANCING
2457 unsigned long change_prot_numa(struct vm_area_struct *vma,
2458 unsigned long start, unsigned long end);
2461 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2462 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2463 unsigned long pfn, unsigned long size, pgprot_t);
2464 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2465 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2467 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2468 unsigned long pfn, pgprot_t pgprot);
2469 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2471 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2472 unsigned long addr, pfn_t pfn);
2473 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2475 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2476 unsigned long addr, struct page *page)
2478 int err = vm_insert_page(vma, addr, page);
2481 return VM_FAULT_OOM;
2482 if (err < 0 && err != -EBUSY)
2483 return VM_FAULT_SIGBUS;
2485 return VM_FAULT_NOPAGE;
2488 static inline vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma,
2489 unsigned long addr, pfn_t pfn)
2491 int err = vm_insert_mixed(vma, addr, pfn);
2494 return VM_FAULT_OOM;
2495 if (err < 0 && err != -EBUSY)
2496 return VM_FAULT_SIGBUS;
2498 return VM_FAULT_NOPAGE;
2501 static inline vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma,
2502 unsigned long addr, unsigned long pfn)
2504 int err = vm_insert_pfn(vma, addr, pfn);
2507 return VM_FAULT_OOM;
2508 if (err < 0 && err != -EBUSY)
2509 return VM_FAULT_SIGBUS;
2511 return VM_FAULT_NOPAGE;
2514 static inline vm_fault_t vmf_error(int err)
2517 return VM_FAULT_OOM;
2518 return VM_FAULT_SIGBUS;
2521 struct page *follow_page_mask(struct vm_area_struct *vma,
2522 unsigned long address, unsigned int foll_flags,
2523 unsigned int *page_mask);
2525 static inline struct page *follow_page(struct vm_area_struct *vma,
2526 unsigned long address, unsigned int foll_flags)
2528 unsigned int unused_page_mask;
2529 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2532 #define FOLL_WRITE 0x01 /* check pte is writable */
2533 #define FOLL_TOUCH 0x02 /* mark page accessed */
2534 #define FOLL_GET 0x04 /* do get_page on page */
2535 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2536 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2537 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2538 * and return without waiting upon it */
2539 #define FOLL_POPULATE 0x40 /* fault in page */
2540 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2541 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2542 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2543 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2544 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2545 #define FOLL_MLOCK 0x1000 /* lock present pages */
2546 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2547 #define FOLL_COW 0x4000 /* internal GUP flag */
2548 #define FOLL_ANON 0x8000 /* don't do file mappings */
2550 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2552 if (vm_fault & VM_FAULT_OOM)
2554 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2555 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2556 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2561 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2563 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2564 unsigned long size, pte_fn_t fn, void *data);
2567 #ifdef CONFIG_PAGE_POISONING
2568 extern bool page_poisoning_enabled(void);
2569 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2571 static inline bool page_poisoning_enabled(void) { return false; }
2572 static inline void kernel_poison_pages(struct page *page, int numpages,
2576 #ifdef CONFIG_DEBUG_PAGEALLOC
2577 extern bool _debug_pagealloc_enabled;
2578 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2580 static inline bool debug_pagealloc_enabled(void)
2582 return _debug_pagealloc_enabled;
2586 kernel_map_pages(struct page *page, int numpages, int enable)
2588 if (!debug_pagealloc_enabled())
2591 __kernel_map_pages(page, numpages, enable);
2593 #ifdef CONFIG_HIBERNATION
2594 extern bool kernel_page_present(struct page *page);
2595 #endif /* CONFIG_HIBERNATION */
2596 #else /* CONFIG_DEBUG_PAGEALLOC */
2598 kernel_map_pages(struct page *page, int numpages, int enable) {}
2599 #ifdef CONFIG_HIBERNATION
2600 static inline bool kernel_page_present(struct page *page) { return true; }
2601 #endif /* CONFIG_HIBERNATION */
2602 static inline bool debug_pagealloc_enabled(void)
2606 #endif /* CONFIG_DEBUG_PAGEALLOC */
2608 #ifdef __HAVE_ARCH_GATE_AREA
2609 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2610 extern int in_gate_area_no_mm(unsigned long addr);
2611 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2613 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2617 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2618 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2622 #endif /* __HAVE_ARCH_GATE_AREA */
2624 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2626 #ifdef CONFIG_SYSCTL
2627 extern int sysctl_drop_caches;
2628 int drop_caches_sysctl_handler(struct ctl_table *, int,
2629 void __user *, size_t *, loff_t *);
2632 void drop_slab(void);
2633 void drop_slab_node(int nid);
2636 #define randomize_va_space 0
2638 extern int randomize_va_space;
2641 const char * arch_vma_name(struct vm_area_struct *vma);
2642 void print_vma_addr(char *prefix, unsigned long rip);
2644 void sparse_mem_maps_populate_node(struct page **map_map,
2645 unsigned long pnum_begin,
2646 unsigned long pnum_end,
2647 unsigned long map_count,
2650 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2651 struct vmem_altmap *altmap);
2652 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2653 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2654 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2655 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2656 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2657 void *vmemmap_alloc_block(unsigned long size, int node);
2659 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2660 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2661 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2662 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2664 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2665 struct vmem_altmap *altmap);
2666 void vmemmap_populate_print_last(void);
2667 #ifdef CONFIG_MEMORY_HOTPLUG
2668 void vmemmap_free(unsigned long start, unsigned long end,
2669 struct vmem_altmap *altmap);
2671 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2672 unsigned long nr_pages);
2675 MF_COUNT_INCREASED = 1 << 0,
2676 MF_ACTION_REQUIRED = 1 << 1,
2677 MF_MUST_KILL = 1 << 2,
2678 MF_SOFT_OFFLINE = 1 << 3,
2680 extern int memory_failure(unsigned long pfn, int flags);
2681 extern void memory_failure_queue(unsigned long pfn, int flags);
2682 extern int unpoison_memory(unsigned long pfn);
2683 extern int get_hwpoison_page(struct page *page);
2684 #define put_hwpoison_page(page) put_page(page)
2685 extern int sysctl_memory_failure_early_kill;
2686 extern int sysctl_memory_failure_recovery;
2687 extern void shake_page(struct page *p, int access);
2688 extern atomic_long_t num_poisoned_pages __read_mostly;
2689 extern int soft_offline_page(struct page *page, int flags);
2693 * Error handlers for various types of pages.
2696 MF_IGNORED, /* Error: cannot be handled */
2697 MF_FAILED, /* Error: handling failed */
2698 MF_DELAYED, /* Will be handled later */
2699 MF_RECOVERED, /* Successfully recovered */
2702 enum mf_action_page_type {
2704 MF_MSG_KERNEL_HIGH_ORDER,
2706 MF_MSG_DIFFERENT_COMPOUND,
2707 MF_MSG_POISONED_HUGE,
2710 MF_MSG_NON_PMD_HUGE,
2711 MF_MSG_UNMAP_FAILED,
2712 MF_MSG_DIRTY_SWAPCACHE,
2713 MF_MSG_CLEAN_SWAPCACHE,
2714 MF_MSG_DIRTY_MLOCKED_LRU,
2715 MF_MSG_CLEAN_MLOCKED_LRU,
2716 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2717 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2720 MF_MSG_TRUNCATED_LRU,
2726 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2727 extern void clear_huge_page(struct page *page,
2728 unsigned long addr_hint,
2729 unsigned int pages_per_huge_page);
2730 extern void copy_user_huge_page(struct page *dst, struct page *src,
2731 unsigned long addr, struct vm_area_struct *vma,
2732 unsigned int pages_per_huge_page);
2733 extern long copy_huge_page_from_user(struct page *dst_page,
2734 const void __user *usr_src,
2735 unsigned int pages_per_huge_page,
2736 bool allow_pagefault);
2737 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2739 extern struct page_ext_operations debug_guardpage_ops;
2741 #ifdef CONFIG_DEBUG_PAGEALLOC
2742 extern unsigned int _debug_guardpage_minorder;
2743 extern bool _debug_guardpage_enabled;
2745 static inline unsigned int debug_guardpage_minorder(void)
2747 return _debug_guardpage_minorder;
2750 static inline bool debug_guardpage_enabled(void)
2752 return _debug_guardpage_enabled;
2755 static inline bool page_is_guard(struct page *page)
2757 struct page_ext *page_ext;
2759 if (!debug_guardpage_enabled())
2762 page_ext = lookup_page_ext(page);
2763 if (unlikely(!page_ext))
2766 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2769 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2770 static inline bool debug_guardpage_enabled(void) { return false; }
2771 static inline bool page_is_guard(struct page *page) { return false; }
2772 #endif /* CONFIG_DEBUG_PAGEALLOC */
2774 #if MAX_NUMNODES > 1
2775 void __init setup_nr_node_ids(void);
2777 static inline void setup_nr_node_ids(void) {}
2780 #endif /* __KERNEL__ */
2781 #endif /* _LINUX_MM_H */