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 #if defined(CONFIG_X86)
233 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
234 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
235 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
236 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
237 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
238 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
239 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
241 #elif defined(CONFIG_PPC)
242 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
243 #elif defined(CONFIG_PARISC)
244 # define VM_GROWSUP VM_ARCH_1
245 #elif defined(CONFIG_IA64)
246 # define VM_GROWSUP VM_ARCH_1
247 #elif defined(CONFIG_SPARC64)
248 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
249 # define VM_ARCH_CLEAR VM_SPARC_ADI
250 #elif !defined(CONFIG_MMU)
251 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
254 #if defined(CONFIG_X86_INTEL_MPX)
255 /* MPX specific bounds table or bounds directory */
256 # define VM_MPX VM_HIGH_ARCH_4
258 # define VM_MPX VM_NONE
262 # define VM_GROWSUP VM_NONE
265 /* Bits set in the VMA until the stack is in its final location */
266 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
268 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
269 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
272 #ifdef CONFIG_STACK_GROWSUP
273 #define VM_STACK VM_GROWSUP
275 #define VM_STACK VM_GROWSDOWN
278 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
281 * Special vmas that are non-mergable, non-mlock()able.
282 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
284 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
286 /* This mask defines which mm->def_flags a process can inherit its parent */
287 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
289 /* This mask is used to clear all the VMA flags used by mlock */
290 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
292 /* Arch-specific flags to clear when updating VM flags on protection change */
293 #ifndef VM_ARCH_CLEAR
294 # define VM_ARCH_CLEAR VM_NONE
296 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
299 * mapping from the currently active vm_flags protection bits (the
300 * low four bits) to a page protection mask..
302 extern pgprot_t protection_map[16];
304 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
305 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
306 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
307 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
308 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
309 #define FAULT_FLAG_TRIED 0x20 /* Second try */
310 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
311 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
312 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
314 #define FAULT_FLAG_TRACE \
315 { FAULT_FLAG_WRITE, "WRITE" }, \
316 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
317 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
318 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
319 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
320 { FAULT_FLAG_TRIED, "TRIED" }, \
321 { FAULT_FLAG_USER, "USER" }, \
322 { FAULT_FLAG_REMOTE, "REMOTE" }, \
323 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
326 * vm_fault is filled by the the pagefault handler and passed to the vma's
327 * ->fault function. The vma's ->fault is responsible for returning a bitmask
328 * of VM_FAULT_xxx flags that give details about how the fault was handled.
330 * MM layer fills up gfp_mask for page allocations but fault handler might
331 * alter it if its implementation requires a different allocation context.
333 * pgoff should be used in favour of virtual_address, if possible.
336 struct vm_area_struct *vma; /* Target VMA */
337 unsigned int flags; /* FAULT_FLAG_xxx flags */
338 gfp_t gfp_mask; /* gfp mask to be used for allocations */
339 pgoff_t pgoff; /* Logical page offset based on vma */
340 unsigned long address; /* Faulting virtual address */
341 pmd_t *pmd; /* Pointer to pmd entry matching
343 pud_t *pud; /* Pointer to pud entry matching
346 pte_t orig_pte; /* Value of PTE at the time of fault */
348 struct page *cow_page; /* Page handler may use for COW fault */
349 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
350 struct page *page; /* ->fault handlers should return a
351 * page here, unless VM_FAULT_NOPAGE
352 * is set (which is also implied by
355 /* These three entries are valid only while holding ptl lock */
356 pte_t *pte; /* Pointer to pte entry matching
357 * the 'address'. NULL if the page
358 * table hasn't been allocated.
360 spinlock_t *ptl; /* Page table lock.
361 * Protects pte page table if 'pte'
362 * is not NULL, otherwise pmd.
364 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
365 * vm_ops->map_pages() calls
366 * alloc_set_pte() from atomic context.
367 * do_fault_around() pre-allocates
368 * page table to avoid allocation from
373 /* page entry size for vm->huge_fault() */
374 enum page_entry_size {
381 * These are the virtual MM functions - opening of an area, closing and
382 * unmapping it (needed to keep files on disk up-to-date etc), pointer
383 * to the functions called when a no-page or a wp-page exception occurs.
385 struct vm_operations_struct {
386 void (*open)(struct vm_area_struct * area);
387 void (*close)(struct vm_area_struct * area);
388 int (*split)(struct vm_area_struct * area, unsigned long addr);
389 int (*mremap)(struct vm_area_struct * area);
390 vm_fault_t (*fault)(struct vm_fault *vmf);
391 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
392 enum page_entry_size pe_size);
393 void (*map_pages)(struct vm_fault *vmf,
394 pgoff_t start_pgoff, pgoff_t end_pgoff);
395 unsigned long (*pagesize)(struct vm_area_struct * area);
397 /* notification that a previously read-only page is about to become
398 * writable, if an error is returned it will cause a SIGBUS */
399 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
401 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
402 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
404 /* called by access_process_vm when get_user_pages() fails, typically
405 * for use by special VMAs that can switch between memory and hardware
407 int (*access)(struct vm_area_struct *vma, unsigned long addr,
408 void *buf, int len, int write);
410 /* Called by the /proc/PID/maps code to ask the vma whether it
411 * has a special name. Returning non-NULL will also cause this
412 * vma to be dumped unconditionally. */
413 const char *(*name)(struct vm_area_struct *vma);
417 * set_policy() op must add a reference to any non-NULL @new mempolicy
418 * to hold the policy upon return. Caller should pass NULL @new to
419 * remove a policy and fall back to surrounding context--i.e. do not
420 * install a MPOL_DEFAULT policy, nor the task or system default
423 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
426 * get_policy() op must add reference [mpol_get()] to any policy at
427 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
428 * in mm/mempolicy.c will do this automatically.
429 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
430 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
431 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
432 * must return NULL--i.e., do not "fallback" to task or system default
435 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
439 * Called by vm_normal_page() for special PTEs to find the
440 * page for @addr. This is useful if the default behavior
441 * (using pte_page()) would not find the correct page.
443 struct page *(*find_special_page)(struct vm_area_struct *vma,
450 #define page_private(page) ((page)->private)
451 #define set_page_private(page, v) ((page)->private = (v))
453 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
454 static inline int pmd_devmap(pmd_t pmd)
458 static inline int pud_devmap(pud_t pud)
462 static inline int pgd_devmap(pgd_t pgd)
469 * FIXME: take this include out, include page-flags.h in
470 * files which need it (119 of them)
472 #include <linux/page-flags.h>
473 #include <linux/huge_mm.h>
476 * Methods to modify the page usage count.
478 * What counts for a page usage:
479 * - cache mapping (page->mapping)
480 * - private data (page->private)
481 * - page mapped in a task's page tables, each mapping
482 * is counted separately
484 * Also, many kernel routines increase the page count before a critical
485 * routine so they can be sure the page doesn't go away from under them.
489 * Drop a ref, return true if the refcount fell to zero (the page has no users)
491 static inline int put_page_testzero(struct page *page)
493 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
494 return page_ref_dec_and_test(page);
498 * Try to grab a ref unless the page has a refcount of zero, return false if
500 * This can be called when MMU is off so it must not access
501 * any of the virtual mappings.
503 static inline int get_page_unless_zero(struct page *page)
505 return page_ref_add_unless(page, 1, 0);
508 extern int page_is_ram(unsigned long pfn);
516 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
519 /* Support for virtually mapped pages */
520 struct page *vmalloc_to_page(const void *addr);
521 unsigned long vmalloc_to_pfn(const void *addr);
524 * Determine if an address is within the vmalloc range
526 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
527 * is no special casing required.
529 static inline bool is_vmalloc_addr(const void *x)
532 unsigned long addr = (unsigned long)x;
534 return addr >= VMALLOC_START && addr < VMALLOC_END;
540 extern int is_vmalloc_or_module_addr(const void *x);
542 static inline int is_vmalloc_or_module_addr(const void *x)
548 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
549 static inline void *kvmalloc(size_t size, gfp_t flags)
551 return kvmalloc_node(size, flags, NUMA_NO_NODE);
553 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
555 return kvmalloc_node(size, flags | __GFP_ZERO, node);
557 static inline void *kvzalloc(size_t size, gfp_t flags)
559 return kvmalloc(size, flags | __GFP_ZERO);
562 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
566 if (unlikely(check_mul_overflow(n, size, &bytes)))
569 return kvmalloc(bytes, flags);
572 extern void kvfree(const void *addr);
574 static inline atomic_t *compound_mapcount_ptr(struct page *page)
576 return &page[1].compound_mapcount;
579 static inline int compound_mapcount(struct page *page)
581 VM_BUG_ON_PAGE(!PageCompound(page), page);
582 page = compound_head(page);
583 return atomic_read(compound_mapcount_ptr(page)) + 1;
587 * The atomic page->_mapcount, starts from -1: so that transitions
588 * both from it and to it can be tracked, using atomic_inc_and_test
589 * and atomic_add_negative(-1).
591 static inline void page_mapcount_reset(struct page *page)
593 atomic_set(&(page)->_mapcount, -1);
596 int __page_mapcount(struct page *page);
598 static inline int page_mapcount(struct page *page)
600 VM_BUG_ON_PAGE(PageSlab(page), page);
602 if (unlikely(PageCompound(page)))
603 return __page_mapcount(page);
604 return atomic_read(&page->_mapcount) + 1;
607 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
608 int total_mapcount(struct page *page);
609 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
611 static inline int total_mapcount(struct page *page)
613 return page_mapcount(page);
615 static inline int page_trans_huge_mapcount(struct page *page,
618 int mapcount = page_mapcount(page);
620 *total_mapcount = mapcount;
625 static inline struct page *virt_to_head_page(const void *x)
627 struct page *page = virt_to_page(x);
629 return compound_head(page);
632 void __put_page(struct page *page);
634 void put_pages_list(struct list_head *pages);
636 void split_page(struct page *page, unsigned int order);
639 * Compound pages have a destructor function. Provide a
640 * prototype for that function and accessor functions.
641 * These are _only_ valid on the head of a compound page.
643 typedef void compound_page_dtor(struct page *);
645 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
646 enum compound_dtor_id {
649 #ifdef CONFIG_HUGETLB_PAGE
652 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
657 extern compound_page_dtor * const compound_page_dtors[];
659 static inline void set_compound_page_dtor(struct page *page,
660 enum compound_dtor_id compound_dtor)
662 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
663 page[1].compound_dtor = compound_dtor;
666 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
668 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
669 return compound_page_dtors[page[1].compound_dtor];
672 static inline unsigned int compound_order(struct page *page)
676 return page[1].compound_order;
679 static inline void set_compound_order(struct page *page, unsigned int order)
681 page[1].compound_order = order;
684 void free_compound_page(struct page *page);
688 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
689 * servicing faults for write access. In the normal case, do always want
690 * pte_mkwrite. But get_user_pages can cause write faults for mappings
691 * that do not have writing enabled, when used by access_process_vm.
693 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
695 if (likely(vma->vm_flags & VM_WRITE))
696 pte = pte_mkwrite(pte);
700 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
702 int finish_fault(struct vm_fault *vmf);
703 int finish_mkwrite_fault(struct vm_fault *vmf);
707 * Multiple processes may "see" the same page. E.g. for untouched
708 * mappings of /dev/null, all processes see the same page full of
709 * zeroes, and text pages of executables and shared libraries have
710 * only one copy in memory, at most, normally.
712 * For the non-reserved pages, page_count(page) denotes a reference count.
713 * page_count() == 0 means the page is free. page->lru is then used for
714 * freelist management in the buddy allocator.
715 * page_count() > 0 means the page has been allocated.
717 * Pages are allocated by the slab allocator in order to provide memory
718 * to kmalloc and kmem_cache_alloc. In this case, the management of the
719 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
720 * unless a particular usage is carefully commented. (the responsibility of
721 * freeing the kmalloc memory is the caller's, of course).
723 * A page may be used by anyone else who does a __get_free_page().
724 * In this case, page_count still tracks the references, and should only
725 * be used through the normal accessor functions. The top bits of page->flags
726 * and page->virtual store page management information, but all other fields
727 * are unused and could be used privately, carefully. The management of this
728 * page is the responsibility of the one who allocated it, and those who have
729 * subsequently been given references to it.
731 * The other pages (we may call them "pagecache pages") are completely
732 * managed by the Linux memory manager: I/O, buffers, swapping etc.
733 * The following discussion applies only to them.
735 * A pagecache page contains an opaque `private' member, which belongs to the
736 * page's address_space. Usually, this is the address of a circular list of
737 * the page's disk buffers. PG_private must be set to tell the VM to call
738 * into the filesystem to release these pages.
740 * A page may belong to an inode's memory mapping. In this case, page->mapping
741 * is the pointer to the inode, and page->index is the file offset of the page,
742 * in units of PAGE_SIZE.
744 * If pagecache pages are not associated with an inode, they are said to be
745 * anonymous pages. These may become associated with the swapcache, and in that
746 * case PG_swapcache is set, and page->private is an offset into the swapcache.
748 * In either case (swapcache or inode backed), the pagecache itself holds one
749 * reference to the page. Setting PG_private should also increment the
750 * refcount. The each user mapping also has a reference to the page.
752 * The pagecache pages are stored in a per-mapping radix tree, which is
753 * rooted at mapping->i_pages, and indexed by offset.
754 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
755 * lists, we instead now tag pages as dirty/writeback in the radix tree.
757 * All pagecache pages may be subject to I/O:
758 * - inode pages may need to be read from disk,
759 * - inode pages which have been modified and are MAP_SHARED may need
760 * to be written back to the inode on disk,
761 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
762 * modified may need to be swapped out to swap space and (later) to be read
767 * The zone field is never updated after free_area_init_core()
768 * sets it, so none of the operations on it need to be atomic.
771 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
772 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
773 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
774 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
775 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
778 * Define the bit shifts to access each section. For non-existent
779 * sections we define the shift as 0; that plus a 0 mask ensures
780 * the compiler will optimise away reference to them.
782 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
783 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
784 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
785 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
787 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
788 #ifdef NODE_NOT_IN_PAGE_FLAGS
789 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
790 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
791 SECTIONS_PGOFF : ZONES_PGOFF)
793 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
794 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
795 NODES_PGOFF : ZONES_PGOFF)
798 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
800 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
801 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
804 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
805 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
806 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
807 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
808 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
810 static inline enum zone_type page_zonenum(const struct page *page)
812 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
815 #ifdef CONFIG_ZONE_DEVICE
816 static inline bool is_zone_device_page(const struct page *page)
818 return page_zonenum(page) == ZONE_DEVICE;
821 static inline bool is_zone_device_page(const struct page *page)
827 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
828 void put_zone_device_private_or_public_page(struct page *page);
829 DECLARE_STATIC_KEY_FALSE(device_private_key);
830 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
831 static inline bool is_device_private_page(const struct page *page);
832 static inline bool is_device_public_page(const struct page *page);
833 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
834 static inline void put_zone_device_private_or_public_page(struct page *page)
837 #define IS_HMM_ENABLED 0
838 static inline bool is_device_private_page(const struct page *page)
842 static inline bool is_device_public_page(const struct page *page)
846 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
849 static inline void get_page(struct page *page)
851 page = compound_head(page);
853 * Getting a normal page or the head of a compound page
854 * requires to already have an elevated page->_refcount.
856 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
860 static inline void put_page(struct page *page)
862 page = compound_head(page);
865 * For private device pages we need to catch refcount transition from
866 * 2 to 1, when refcount reach one it means the private device page is
867 * free and we need to inform the device driver through callback. See
868 * include/linux/memremap.h and HMM for details.
870 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
871 unlikely(is_device_public_page(page)))) {
872 put_zone_device_private_or_public_page(page);
876 if (put_page_testzero(page))
880 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
881 #define SECTION_IN_PAGE_FLAGS
885 * The identification function is mainly used by the buddy allocator for
886 * determining if two pages could be buddies. We are not really identifying
887 * the zone since we could be using the section number id if we do not have
888 * node id available in page flags.
889 * We only guarantee that it will return the same value for two combinable
892 static inline int page_zone_id(struct page *page)
894 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
897 static inline int zone_to_nid(struct zone *zone)
906 #ifdef NODE_NOT_IN_PAGE_FLAGS
907 extern int page_to_nid(const struct page *page);
909 static inline int page_to_nid(const struct page *page)
911 struct page *p = (struct page *)page;
913 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
917 #ifdef CONFIG_NUMA_BALANCING
918 static inline int cpu_pid_to_cpupid(int cpu, int pid)
920 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
923 static inline int cpupid_to_pid(int cpupid)
925 return cpupid & LAST__PID_MASK;
928 static inline int cpupid_to_cpu(int cpupid)
930 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
933 static inline int cpupid_to_nid(int cpupid)
935 return cpu_to_node(cpupid_to_cpu(cpupid));
938 static inline bool cpupid_pid_unset(int cpupid)
940 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
943 static inline bool cpupid_cpu_unset(int cpupid)
945 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
948 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
950 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
953 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
954 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
955 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
957 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
960 static inline int page_cpupid_last(struct page *page)
962 return page->_last_cpupid;
964 static inline void page_cpupid_reset_last(struct page *page)
966 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
969 static inline int page_cpupid_last(struct page *page)
971 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
974 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
976 static inline void page_cpupid_reset_last(struct page *page)
978 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
980 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
981 #else /* !CONFIG_NUMA_BALANCING */
982 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
984 return page_to_nid(page); /* XXX */
987 static inline int page_cpupid_last(struct page *page)
989 return page_to_nid(page); /* XXX */
992 static inline int cpupid_to_nid(int cpupid)
997 static inline int cpupid_to_pid(int cpupid)
1002 static inline int cpupid_to_cpu(int cpupid)
1007 static inline int cpu_pid_to_cpupid(int nid, int pid)
1012 static inline bool cpupid_pid_unset(int cpupid)
1017 static inline void page_cpupid_reset_last(struct page *page)
1021 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1025 #endif /* CONFIG_NUMA_BALANCING */
1027 static inline struct zone *page_zone(const struct page *page)
1029 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1032 static inline pg_data_t *page_pgdat(const struct page *page)
1034 return NODE_DATA(page_to_nid(page));
1037 #ifdef SECTION_IN_PAGE_FLAGS
1038 static inline void set_page_section(struct page *page, unsigned long section)
1040 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1041 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1044 static inline unsigned long page_to_section(const struct page *page)
1046 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1050 static inline void set_page_zone(struct page *page, enum zone_type zone)
1052 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1053 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1056 static inline void set_page_node(struct page *page, unsigned long node)
1058 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1059 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1062 static inline void set_page_links(struct page *page, enum zone_type zone,
1063 unsigned long node, unsigned long pfn)
1065 set_page_zone(page, zone);
1066 set_page_node(page, node);
1067 #ifdef SECTION_IN_PAGE_FLAGS
1068 set_page_section(page, pfn_to_section_nr(pfn));
1073 static inline struct mem_cgroup *page_memcg(struct page *page)
1075 return page->mem_cgroup;
1077 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1079 WARN_ON_ONCE(!rcu_read_lock_held());
1080 return READ_ONCE(page->mem_cgroup);
1083 static inline struct mem_cgroup *page_memcg(struct page *page)
1087 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1089 WARN_ON_ONCE(!rcu_read_lock_held());
1095 * Some inline functions in vmstat.h depend on page_zone()
1097 #include <linux/vmstat.h>
1099 static __always_inline void *lowmem_page_address(const struct page *page)
1101 return page_to_virt(page);
1104 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1105 #define HASHED_PAGE_VIRTUAL
1108 #if defined(WANT_PAGE_VIRTUAL)
1109 static inline void *page_address(const struct page *page)
1111 return page->virtual;
1113 static inline void set_page_address(struct page *page, void *address)
1115 page->virtual = address;
1117 #define page_address_init() do { } while(0)
1120 #if defined(HASHED_PAGE_VIRTUAL)
1121 void *page_address(const struct page *page);
1122 void set_page_address(struct page *page, void *virtual);
1123 void page_address_init(void);
1126 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1127 #define page_address(page) lowmem_page_address(page)
1128 #define set_page_address(page, address) do { } while(0)
1129 #define page_address_init() do { } while(0)
1132 extern void *page_rmapping(struct page *page);
1133 extern struct anon_vma *page_anon_vma(struct page *page);
1134 extern struct address_space *page_mapping(struct page *page);
1136 extern struct address_space *__page_file_mapping(struct page *);
1139 struct address_space *page_file_mapping(struct page *page)
1141 if (unlikely(PageSwapCache(page)))
1142 return __page_file_mapping(page);
1144 return page->mapping;
1147 extern pgoff_t __page_file_index(struct page *page);
1150 * Return the pagecache index of the passed page. Regular pagecache pages
1151 * use ->index whereas swapcache pages use swp_offset(->private)
1153 static inline pgoff_t page_index(struct page *page)
1155 if (unlikely(PageSwapCache(page)))
1156 return __page_file_index(page);
1160 bool page_mapped(struct page *page);
1161 struct address_space *page_mapping(struct page *page);
1162 struct address_space *page_mapping_file(struct page *page);
1165 * Return true only if the page has been allocated with
1166 * ALLOC_NO_WATERMARKS and the low watermark was not
1167 * met implying that the system is under some pressure.
1169 static inline bool page_is_pfmemalloc(struct page *page)
1172 * Page index cannot be this large so this must be
1173 * a pfmemalloc page.
1175 return page->index == -1UL;
1179 * Only to be called by the page allocator on a freshly allocated
1182 static inline void set_page_pfmemalloc(struct page *page)
1187 static inline void clear_page_pfmemalloc(struct page *page)
1193 * Different kinds of faults, as returned by handle_mm_fault().
1194 * Used to decide whether a process gets delivered SIGBUS or
1195 * just gets major/minor fault counters bumped up.
1198 #define VM_FAULT_OOM 0x0001
1199 #define VM_FAULT_SIGBUS 0x0002
1200 #define VM_FAULT_MAJOR 0x0004
1201 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1202 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1203 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1204 #define VM_FAULT_SIGSEGV 0x0040
1206 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1207 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1208 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1209 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1210 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1211 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1212 * and needs fsync() to complete (for
1213 * synchronous page faults in DAX) */
1215 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1216 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1219 #define VM_FAULT_RESULT_TRACE \
1220 { VM_FAULT_OOM, "OOM" }, \
1221 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1222 { VM_FAULT_MAJOR, "MAJOR" }, \
1223 { VM_FAULT_WRITE, "WRITE" }, \
1224 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1225 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1226 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1227 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1228 { VM_FAULT_LOCKED, "LOCKED" }, \
1229 { VM_FAULT_RETRY, "RETRY" }, \
1230 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1231 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1232 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1234 /* Encode hstate index for a hwpoisoned large page */
1235 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1236 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1239 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1241 extern void pagefault_out_of_memory(void);
1243 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1246 * Flags passed to show_mem() and show_free_areas() to suppress output in
1249 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1251 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1253 extern bool can_do_mlock(void);
1254 extern int user_shm_lock(size_t, struct user_struct *);
1255 extern void user_shm_unlock(size_t, struct user_struct *);
1258 * Parameter block passed down to zap_pte_range in exceptional cases.
1260 struct zap_details {
1261 struct address_space *check_mapping; /* Check page->mapping if set */
1262 pgoff_t first_index; /* Lowest page->index to unmap */
1263 pgoff_t last_index; /* Highest page->index to unmap */
1266 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1267 pte_t pte, bool with_public_device);
1268 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1270 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1273 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1274 unsigned long size);
1275 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1276 unsigned long size);
1277 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1278 unsigned long start, unsigned long end);
1281 * mm_walk - callbacks for walk_page_range
1282 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1283 * this handler should only handle pud_trans_huge() puds.
1284 * the pmd_entry or pte_entry callbacks will be used for
1286 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1287 * this handler is required to be able to handle
1288 * pmd_trans_huge() pmds. They may simply choose to
1289 * split_huge_page() instead of handling it explicitly.
1290 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1291 * @pte_hole: if set, called for each hole at all levels
1292 * @hugetlb_entry: if set, called for each hugetlb entry
1293 * @test_walk: caller specific callback function to determine whether
1294 * we walk over the current vma or not. Returning 0
1295 * value means "do page table walk over the current vma,"
1296 * and a negative one means "abort current page table walk
1297 * right now." 1 means "skip the current vma."
1298 * @mm: mm_struct representing the target process of page table walk
1299 * @vma: vma currently walked (NULL if walking outside vmas)
1300 * @private: private data for callbacks' usage
1302 * (see the comment on walk_page_range() for more details)
1305 int (*pud_entry)(pud_t *pud, unsigned long addr,
1306 unsigned long next, struct mm_walk *walk);
1307 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1308 unsigned long next, struct mm_walk *walk);
1309 int (*pte_entry)(pte_t *pte, unsigned long addr,
1310 unsigned long next, struct mm_walk *walk);
1311 int (*pte_hole)(unsigned long addr, unsigned long next,
1312 struct mm_walk *walk);
1313 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1314 unsigned long addr, unsigned long next,
1315 struct mm_walk *walk);
1316 int (*test_walk)(unsigned long addr, unsigned long next,
1317 struct mm_walk *walk);
1318 struct mm_struct *mm;
1319 struct vm_area_struct *vma;
1323 int walk_page_range(unsigned long addr, unsigned long end,
1324 struct mm_walk *walk);
1325 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1326 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1327 unsigned long end, unsigned long floor, unsigned long ceiling);
1328 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1329 struct vm_area_struct *vma);
1330 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1331 unsigned long *start, unsigned long *end,
1332 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1333 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1334 unsigned long *pfn);
1335 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1336 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1337 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1338 void *buf, int len, int write);
1340 extern void truncate_pagecache(struct inode *inode, loff_t new);
1341 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1342 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1343 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1344 int truncate_inode_page(struct address_space *mapping, struct page *page);
1345 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1346 int invalidate_inode_page(struct page *page);
1349 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1350 unsigned int flags);
1351 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1352 unsigned long address, unsigned int fault_flags,
1354 void unmap_mapping_pages(struct address_space *mapping,
1355 pgoff_t start, pgoff_t nr, bool even_cows);
1356 void unmap_mapping_range(struct address_space *mapping,
1357 loff_t const holebegin, loff_t const holelen, int even_cows);
1359 static inline int handle_mm_fault(struct vm_area_struct *vma,
1360 unsigned long address, unsigned int flags)
1362 /* should never happen if there's no MMU */
1364 return VM_FAULT_SIGBUS;
1366 static inline int fixup_user_fault(struct task_struct *tsk,
1367 struct mm_struct *mm, unsigned long address,
1368 unsigned int fault_flags, bool *unlocked)
1370 /* should never happen if there's no MMU */
1374 static inline void unmap_mapping_pages(struct address_space *mapping,
1375 pgoff_t start, pgoff_t nr, bool even_cows) { }
1376 static inline void unmap_mapping_range(struct address_space *mapping,
1377 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1380 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1381 loff_t const holebegin, loff_t const holelen)
1383 unmap_mapping_range(mapping, holebegin, holelen, 0);
1386 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1387 void *buf, int len, unsigned int gup_flags);
1388 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1389 void *buf, int len, unsigned int gup_flags);
1390 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1391 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1393 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1394 unsigned long start, unsigned long nr_pages,
1395 unsigned int gup_flags, struct page **pages,
1396 struct vm_area_struct **vmas, int *locked);
1397 long get_user_pages(unsigned long start, unsigned long nr_pages,
1398 unsigned int gup_flags, struct page **pages,
1399 struct vm_area_struct **vmas);
1400 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1401 unsigned int gup_flags, struct page **pages, int *locked);
1402 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1403 struct page **pages, unsigned int gup_flags);
1404 #ifdef CONFIG_FS_DAX
1405 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1406 unsigned int gup_flags, struct page **pages,
1407 struct vm_area_struct **vmas);
1409 static inline long get_user_pages_longterm(unsigned long start,
1410 unsigned long nr_pages, unsigned int gup_flags,
1411 struct page **pages, struct vm_area_struct **vmas)
1413 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1415 #endif /* CONFIG_FS_DAX */
1417 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1418 struct page **pages);
1420 /* Container for pinned pfns / pages */
1421 struct frame_vector {
1422 unsigned int nr_allocated; /* Number of frames we have space for */
1423 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1424 bool got_ref; /* Did we pin pages by getting page ref? */
1425 bool is_pfns; /* Does array contain pages or pfns? */
1426 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1427 * pfns_vector_pages() or pfns_vector_pfns()
1431 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1432 void frame_vector_destroy(struct frame_vector *vec);
1433 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1434 unsigned int gup_flags, struct frame_vector *vec);
1435 void put_vaddr_frames(struct frame_vector *vec);
1436 int frame_vector_to_pages(struct frame_vector *vec);
1437 void frame_vector_to_pfns(struct frame_vector *vec);
1439 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1441 return vec->nr_frames;
1444 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1447 int err = frame_vector_to_pages(vec);
1450 return ERR_PTR(err);
1452 return (struct page **)(vec->ptrs);
1455 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1458 frame_vector_to_pfns(vec);
1459 return (unsigned long *)(vec->ptrs);
1463 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1464 struct page **pages);
1465 int get_kernel_page(unsigned long start, int write, struct page **pages);
1466 struct page *get_dump_page(unsigned long addr);
1468 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1469 extern void do_invalidatepage(struct page *page, unsigned int offset,
1470 unsigned int length);
1472 void __set_page_dirty(struct page *, struct address_space *, int warn);
1473 int __set_page_dirty_nobuffers(struct page *page);
1474 int __set_page_dirty_no_writeback(struct page *page);
1475 int redirty_page_for_writepage(struct writeback_control *wbc,
1477 void account_page_dirtied(struct page *page, struct address_space *mapping);
1478 void account_page_cleaned(struct page *page, struct address_space *mapping,
1479 struct bdi_writeback *wb);
1480 int set_page_dirty(struct page *page);
1481 int set_page_dirty_lock(struct page *page);
1482 void __cancel_dirty_page(struct page *page);
1483 static inline void cancel_dirty_page(struct page *page)
1485 /* Avoid atomic ops, locking, etc. when not actually needed. */
1486 if (PageDirty(page))
1487 __cancel_dirty_page(page);
1489 int clear_page_dirty_for_io(struct page *page);
1491 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1493 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1495 return !vma->vm_ops;
1500 * The vma_is_shmem is not inline because it is used only by slow
1501 * paths in userfault.
1503 bool vma_is_shmem(struct vm_area_struct *vma);
1505 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1508 int vma_is_stack_for_current(struct vm_area_struct *vma);
1510 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1511 unsigned long old_addr, struct vm_area_struct *new_vma,
1512 unsigned long new_addr, unsigned long len,
1513 bool need_rmap_locks);
1514 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1515 unsigned long end, pgprot_t newprot,
1516 int dirty_accountable, int prot_numa);
1517 extern int mprotect_fixup(struct vm_area_struct *vma,
1518 struct vm_area_struct **pprev, unsigned long start,
1519 unsigned long end, unsigned long newflags);
1522 * doesn't attempt to fault and will return short.
1524 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1525 struct page **pages);
1527 * per-process(per-mm_struct) statistics.
1529 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1531 long val = atomic_long_read(&mm->rss_stat.count[member]);
1533 #ifdef SPLIT_RSS_COUNTING
1535 * counter is updated in asynchronous manner and may go to minus.
1536 * But it's never be expected number for users.
1541 return (unsigned long)val;
1544 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1546 atomic_long_add(value, &mm->rss_stat.count[member]);
1549 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1551 atomic_long_inc(&mm->rss_stat.count[member]);
1554 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1556 atomic_long_dec(&mm->rss_stat.count[member]);
1559 /* Optimized variant when page is already known not to be PageAnon */
1560 static inline int mm_counter_file(struct page *page)
1562 if (PageSwapBacked(page))
1563 return MM_SHMEMPAGES;
1564 return MM_FILEPAGES;
1567 static inline int mm_counter(struct page *page)
1570 return MM_ANONPAGES;
1571 return mm_counter_file(page);
1574 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1576 return get_mm_counter(mm, MM_FILEPAGES) +
1577 get_mm_counter(mm, MM_ANONPAGES) +
1578 get_mm_counter(mm, MM_SHMEMPAGES);
1581 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1583 return max(mm->hiwater_rss, get_mm_rss(mm));
1586 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1588 return max(mm->hiwater_vm, mm->total_vm);
1591 static inline void update_hiwater_rss(struct mm_struct *mm)
1593 unsigned long _rss = get_mm_rss(mm);
1595 if ((mm)->hiwater_rss < _rss)
1596 (mm)->hiwater_rss = _rss;
1599 static inline void update_hiwater_vm(struct mm_struct *mm)
1601 if (mm->hiwater_vm < mm->total_vm)
1602 mm->hiwater_vm = mm->total_vm;
1605 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1607 mm->hiwater_rss = get_mm_rss(mm);
1610 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1611 struct mm_struct *mm)
1613 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1615 if (*maxrss < hiwater_rss)
1616 *maxrss = hiwater_rss;
1619 #if defined(SPLIT_RSS_COUNTING)
1620 void sync_mm_rss(struct mm_struct *mm);
1622 static inline void sync_mm_rss(struct mm_struct *mm)
1627 #ifndef __HAVE_ARCH_PTE_DEVMAP
1628 static inline int pte_devmap(pte_t pte)
1634 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1636 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1638 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1642 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1646 #ifdef __PAGETABLE_P4D_FOLDED
1647 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1648 unsigned long address)
1653 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1656 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1657 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1658 unsigned long address)
1662 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1663 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1666 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1668 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1670 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1673 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1675 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1679 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1680 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1681 unsigned long address)
1686 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1687 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1690 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1692 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1694 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1697 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1699 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1704 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1706 atomic_long_set(&mm->pgtables_bytes, 0);
1709 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1711 return atomic_long_read(&mm->pgtables_bytes);
1714 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1716 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1719 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1721 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1725 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1726 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1731 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1732 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1735 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1736 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1739 * The following ifdef needed to get the 4level-fixup.h header to work.
1740 * Remove it when 4level-fixup.h has been removed.
1742 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1744 #ifndef __ARCH_HAS_5LEVEL_HACK
1745 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1746 unsigned long address)
1748 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1749 NULL : p4d_offset(pgd, address);
1752 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1753 unsigned long address)
1755 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1756 NULL : pud_offset(p4d, address);
1758 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1760 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1762 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1763 NULL: pmd_offset(pud, address);
1765 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1767 #if USE_SPLIT_PTE_PTLOCKS
1768 #if ALLOC_SPLIT_PTLOCKS
1769 void __init ptlock_cache_init(void);
1770 extern bool ptlock_alloc(struct page *page);
1771 extern void ptlock_free(struct page *page);
1773 static inline spinlock_t *ptlock_ptr(struct page *page)
1777 #else /* ALLOC_SPLIT_PTLOCKS */
1778 static inline void ptlock_cache_init(void)
1782 static inline bool ptlock_alloc(struct page *page)
1787 static inline void ptlock_free(struct page *page)
1791 static inline spinlock_t *ptlock_ptr(struct page *page)
1795 #endif /* ALLOC_SPLIT_PTLOCKS */
1797 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1799 return ptlock_ptr(pmd_page(*pmd));
1802 static inline bool ptlock_init(struct page *page)
1805 * prep_new_page() initialize page->private (and therefore page->ptl)
1806 * with 0. Make sure nobody took it in use in between.
1808 * It can happen if arch try to use slab for page table allocation:
1809 * slab code uses page->slab_cache, which share storage with page->ptl.
1811 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1812 if (!ptlock_alloc(page))
1814 spin_lock_init(ptlock_ptr(page));
1818 /* Reset page->mapping so free_pages_check won't complain. */
1819 static inline void pte_lock_deinit(struct page *page)
1821 page->mapping = NULL;
1825 #else /* !USE_SPLIT_PTE_PTLOCKS */
1827 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1829 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1831 return &mm->page_table_lock;
1833 static inline void ptlock_cache_init(void) {}
1834 static inline bool ptlock_init(struct page *page) { return true; }
1835 static inline void pte_lock_deinit(struct page *page) {}
1836 #endif /* USE_SPLIT_PTE_PTLOCKS */
1838 static inline void pgtable_init(void)
1840 ptlock_cache_init();
1841 pgtable_cache_init();
1844 static inline bool pgtable_page_ctor(struct page *page)
1846 if (!ptlock_init(page))
1848 inc_zone_page_state(page, NR_PAGETABLE);
1852 static inline void pgtable_page_dtor(struct page *page)
1854 pte_lock_deinit(page);
1855 dec_zone_page_state(page, NR_PAGETABLE);
1858 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1860 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1861 pte_t *__pte = pte_offset_map(pmd, address); \
1867 #define pte_unmap_unlock(pte, ptl) do { \
1872 #define pte_alloc(mm, pmd, address) \
1873 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1875 #define pte_alloc_map(mm, pmd, address) \
1876 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1878 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1879 (pte_alloc(mm, pmd, address) ? \
1880 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1882 #define pte_alloc_kernel(pmd, address) \
1883 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1884 NULL: pte_offset_kernel(pmd, address))
1886 #if USE_SPLIT_PMD_PTLOCKS
1888 static struct page *pmd_to_page(pmd_t *pmd)
1890 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1891 return virt_to_page((void *)((unsigned long) pmd & mask));
1894 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1896 return ptlock_ptr(pmd_to_page(pmd));
1899 static inline bool pgtable_pmd_page_ctor(struct page *page)
1901 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1902 page->pmd_huge_pte = NULL;
1904 return ptlock_init(page);
1907 static inline void pgtable_pmd_page_dtor(struct page *page)
1909 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1910 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1915 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1919 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1921 return &mm->page_table_lock;
1924 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1925 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1927 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1931 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1933 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1939 * No scalability reason to split PUD locks yet, but follow the same pattern
1940 * as the PMD locks to make it easier if we decide to. The VM should not be
1941 * considered ready to switch to split PUD locks yet; there may be places
1942 * which need to be converted from page_table_lock.
1944 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1946 return &mm->page_table_lock;
1949 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1951 spinlock_t *ptl = pud_lockptr(mm, pud);
1957 extern void __init pagecache_init(void);
1958 extern void free_area_init(unsigned long * zones_size);
1959 extern void free_area_init_node(int nid, unsigned long * zones_size,
1960 unsigned long zone_start_pfn, unsigned long *zholes_size);
1961 extern void free_initmem(void);
1964 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1965 * into the buddy system. The freed pages will be poisoned with pattern
1966 * "poison" if it's within range [0, UCHAR_MAX].
1967 * Return pages freed into the buddy system.
1969 extern unsigned long free_reserved_area(void *start, void *end,
1970 int poison, char *s);
1972 #ifdef CONFIG_HIGHMEM
1974 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1975 * and totalram_pages.
1977 extern void free_highmem_page(struct page *page);
1980 extern void adjust_managed_page_count(struct page *page, long count);
1981 extern void mem_init_print_info(const char *str);
1983 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1985 /* Free the reserved page into the buddy system, so it gets managed. */
1986 static inline void __free_reserved_page(struct page *page)
1988 ClearPageReserved(page);
1989 init_page_count(page);
1993 static inline void free_reserved_page(struct page *page)
1995 __free_reserved_page(page);
1996 adjust_managed_page_count(page, 1);
1999 static inline void mark_page_reserved(struct page *page)
2001 SetPageReserved(page);
2002 adjust_managed_page_count(page, -1);
2006 * Default method to free all the __init memory into the buddy system.
2007 * The freed pages will be poisoned with pattern "poison" if it's within
2008 * range [0, UCHAR_MAX].
2009 * Return pages freed into the buddy system.
2011 static inline unsigned long free_initmem_default(int poison)
2013 extern char __init_begin[], __init_end[];
2015 return free_reserved_area(&__init_begin, &__init_end,
2016 poison, "unused kernel");
2019 static inline unsigned long get_num_physpages(void)
2022 unsigned long phys_pages = 0;
2024 for_each_online_node(nid)
2025 phys_pages += node_present_pages(nid);
2030 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2032 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2033 * zones, allocate the backing mem_map and account for memory holes in a more
2034 * architecture independent manner. This is a substitute for creating the
2035 * zone_sizes[] and zholes_size[] arrays and passing them to
2036 * free_area_init_node()
2038 * An architecture is expected to register range of page frames backed by
2039 * physical memory with memblock_add[_node]() before calling
2040 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2041 * usage, an architecture is expected to do something like
2043 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2045 * for_each_valid_physical_page_range()
2046 * memblock_add_node(base, size, nid)
2047 * free_area_init_nodes(max_zone_pfns);
2049 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2050 * registered physical page range. Similarly
2051 * sparse_memory_present_with_active_regions() calls memory_present() for
2052 * each range when SPARSEMEM is enabled.
2054 * See mm/page_alloc.c for more information on each function exposed by
2055 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2057 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2058 unsigned long node_map_pfn_alignment(void);
2059 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2060 unsigned long end_pfn);
2061 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2062 unsigned long end_pfn);
2063 extern void get_pfn_range_for_nid(unsigned int nid,
2064 unsigned long *start_pfn, unsigned long *end_pfn);
2065 extern unsigned long find_min_pfn_with_active_regions(void);
2066 extern void free_bootmem_with_active_regions(int nid,
2067 unsigned long max_low_pfn);
2068 extern void sparse_memory_present_with_active_regions(int nid);
2070 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2072 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2073 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2074 static inline int __early_pfn_to_nid(unsigned long pfn,
2075 struct mminit_pfnnid_cache *state)
2080 /* please see mm/page_alloc.c */
2081 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2082 /* there is a per-arch backend function. */
2083 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2084 struct mminit_pfnnid_cache *state);
2087 #ifdef CONFIG_HAVE_MEMBLOCK
2088 void zero_resv_unavail(void);
2090 static inline void zero_resv_unavail(void) {}
2093 extern void set_dma_reserve(unsigned long new_dma_reserve);
2094 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2095 enum memmap_context, struct vmem_altmap *);
2096 extern void setup_per_zone_wmarks(void);
2097 extern int __meminit init_per_zone_wmark_min(void);
2098 extern void mem_init(void);
2099 extern void __init mmap_init(void);
2100 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2101 extern long si_mem_available(void);
2102 extern void si_meminfo(struct sysinfo * val);
2103 extern void si_meminfo_node(struct sysinfo *val, int nid);
2104 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2105 extern unsigned long arch_reserved_kernel_pages(void);
2108 extern __printf(3, 4)
2109 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2111 extern void setup_per_cpu_pageset(void);
2113 extern void zone_pcp_update(struct zone *zone);
2114 extern void zone_pcp_reset(struct zone *zone);
2117 extern int min_free_kbytes;
2118 extern int watermark_scale_factor;
2121 extern atomic_long_t mmap_pages_allocated;
2122 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2124 /* interval_tree.c */
2125 void vma_interval_tree_insert(struct vm_area_struct *node,
2126 struct rb_root_cached *root);
2127 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2128 struct vm_area_struct *prev,
2129 struct rb_root_cached *root);
2130 void vma_interval_tree_remove(struct vm_area_struct *node,
2131 struct rb_root_cached *root);
2132 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2133 unsigned long start, unsigned long last);
2134 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2135 unsigned long start, unsigned long last);
2137 #define vma_interval_tree_foreach(vma, root, start, last) \
2138 for (vma = vma_interval_tree_iter_first(root, start, last); \
2139 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2141 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2142 struct rb_root_cached *root);
2143 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2144 struct rb_root_cached *root);
2145 struct anon_vma_chain *
2146 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2147 unsigned long start, unsigned long last);
2148 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2149 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2150 #ifdef CONFIG_DEBUG_VM_RB
2151 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2154 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2155 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2156 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2159 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2160 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2161 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2162 struct vm_area_struct *expand);
2163 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2164 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2166 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2168 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2169 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2170 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2171 struct mempolicy *, struct vm_userfaultfd_ctx);
2172 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2173 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2174 unsigned long addr, int new_below);
2175 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2176 unsigned long addr, int new_below);
2177 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2178 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2179 struct rb_node **, struct rb_node *);
2180 extern void unlink_file_vma(struct vm_area_struct *);
2181 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2182 unsigned long addr, unsigned long len, pgoff_t pgoff,
2183 bool *need_rmap_locks);
2184 extern void exit_mmap(struct mm_struct *);
2186 static inline int check_data_rlimit(unsigned long rlim,
2188 unsigned long start,
2189 unsigned long end_data,
2190 unsigned long start_data)
2192 if (rlim < RLIM_INFINITY) {
2193 if (((new - start) + (end_data - start_data)) > rlim)
2200 extern int mm_take_all_locks(struct mm_struct *mm);
2201 extern void mm_drop_all_locks(struct mm_struct *mm);
2203 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2204 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2205 extern struct file *get_task_exe_file(struct task_struct *task);
2207 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2208 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2210 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2211 const struct vm_special_mapping *sm);
2212 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2213 unsigned long addr, unsigned long len,
2214 unsigned long flags,
2215 const struct vm_special_mapping *spec);
2216 /* This is an obsolete alternative to _install_special_mapping. */
2217 extern int install_special_mapping(struct mm_struct *mm,
2218 unsigned long addr, unsigned long len,
2219 unsigned long flags, struct page **pages);
2221 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2223 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2224 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2225 struct list_head *uf);
2226 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2227 unsigned long len, unsigned long prot, unsigned long flags,
2228 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2229 struct list_head *uf);
2230 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2231 struct list_head *uf);
2233 static inline unsigned long
2234 do_mmap_pgoff(struct file *file, unsigned long addr,
2235 unsigned long len, unsigned long prot, unsigned long flags,
2236 unsigned long pgoff, unsigned long *populate,
2237 struct list_head *uf)
2239 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2243 extern int __mm_populate(unsigned long addr, unsigned long len,
2245 static inline void mm_populate(unsigned long addr, unsigned long len)
2248 (void) __mm_populate(addr, len, 1);
2251 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2254 /* These take the mm semaphore themselves */
2255 extern int __must_check vm_brk(unsigned long, unsigned long);
2256 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2257 extern int vm_munmap(unsigned long, size_t);
2258 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2259 unsigned long, unsigned long,
2260 unsigned long, unsigned long);
2262 struct vm_unmapped_area_info {
2263 #define VM_UNMAPPED_AREA_TOPDOWN 1
2264 unsigned long flags;
2265 unsigned long length;
2266 unsigned long low_limit;
2267 unsigned long high_limit;
2268 unsigned long align_mask;
2269 unsigned long align_offset;
2272 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2273 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2276 * Search for an unmapped address range.
2278 * We are looking for a range that:
2279 * - does not intersect with any VMA;
2280 * - is contained within the [low_limit, high_limit) interval;
2281 * - is at least the desired size.
2282 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2284 static inline unsigned long
2285 vm_unmapped_area(struct vm_unmapped_area_info *info)
2287 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2288 return unmapped_area_topdown(info);
2290 return unmapped_area(info);
2294 extern void truncate_inode_pages(struct address_space *, loff_t);
2295 extern void truncate_inode_pages_range(struct address_space *,
2296 loff_t lstart, loff_t lend);
2297 extern void truncate_inode_pages_final(struct address_space *);
2299 /* generic vm_area_ops exported for stackable file systems */
2300 extern int filemap_fault(struct vm_fault *vmf);
2301 extern void filemap_map_pages(struct vm_fault *vmf,
2302 pgoff_t start_pgoff, pgoff_t end_pgoff);
2303 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2305 /* mm/page-writeback.c */
2306 int __must_check write_one_page(struct page *page);
2307 void task_dirty_inc(struct task_struct *tsk);
2310 #define VM_MAX_READAHEAD 128 /* kbytes */
2311 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2313 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2314 pgoff_t offset, unsigned long nr_to_read);
2316 void page_cache_sync_readahead(struct address_space *mapping,
2317 struct file_ra_state *ra,
2320 unsigned long size);
2322 void page_cache_async_readahead(struct address_space *mapping,
2323 struct file_ra_state *ra,
2327 unsigned long size);
2329 extern unsigned long stack_guard_gap;
2330 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2331 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2333 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2334 extern int expand_downwards(struct vm_area_struct *vma,
2335 unsigned long address);
2337 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2339 #define expand_upwards(vma, address) (0)
2342 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2343 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2344 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2345 struct vm_area_struct **pprev);
2347 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2348 NULL if none. Assume start_addr < end_addr. */
2349 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2351 struct vm_area_struct * vma = find_vma(mm,start_addr);
2353 if (vma && end_addr <= vma->vm_start)
2358 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2360 unsigned long vm_start = vma->vm_start;
2362 if (vma->vm_flags & VM_GROWSDOWN) {
2363 vm_start -= stack_guard_gap;
2364 if (vm_start > vma->vm_start)
2370 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2372 unsigned long vm_end = vma->vm_end;
2374 if (vma->vm_flags & VM_GROWSUP) {
2375 vm_end += stack_guard_gap;
2376 if (vm_end < vma->vm_end)
2377 vm_end = -PAGE_SIZE;
2382 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2384 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2387 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2388 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2389 unsigned long vm_start, unsigned long vm_end)
2391 struct vm_area_struct *vma = find_vma(mm, vm_start);
2393 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2400 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2401 void vma_set_page_prot(struct vm_area_struct *vma);
2403 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2407 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2409 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2413 #ifdef CONFIG_NUMA_BALANCING
2414 unsigned long change_prot_numa(struct vm_area_struct *vma,
2415 unsigned long start, unsigned long end);
2418 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2419 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2420 unsigned long pfn, unsigned long size, pgprot_t);
2421 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2422 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2424 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2425 unsigned long pfn, pgprot_t pgprot);
2426 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2428 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2430 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2432 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2433 unsigned long addr, struct page *page)
2435 int err = vm_insert_page(vma, addr, page);
2438 return VM_FAULT_OOM;
2439 if (err < 0 && err != -EBUSY)
2440 return VM_FAULT_SIGBUS;
2442 return VM_FAULT_NOPAGE;
2445 static inline vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma,
2446 unsigned long addr, pfn_t pfn)
2448 int err = vm_insert_mixed(vma, addr, pfn);
2451 return VM_FAULT_OOM;
2452 if (err < 0 && err != -EBUSY)
2453 return VM_FAULT_SIGBUS;
2455 return VM_FAULT_NOPAGE;
2458 static inline vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma,
2459 unsigned long addr, unsigned long pfn)
2461 int err = vm_insert_pfn(vma, addr, pfn);
2464 return VM_FAULT_OOM;
2465 if (err < 0 && err != -EBUSY)
2466 return VM_FAULT_SIGBUS;
2468 return VM_FAULT_NOPAGE;
2471 static inline vm_fault_t vmf_error(int err)
2474 return VM_FAULT_OOM;
2475 return VM_FAULT_SIGBUS;
2478 struct page *follow_page_mask(struct vm_area_struct *vma,
2479 unsigned long address, unsigned int foll_flags,
2480 unsigned int *page_mask);
2482 static inline struct page *follow_page(struct vm_area_struct *vma,
2483 unsigned long address, unsigned int foll_flags)
2485 unsigned int unused_page_mask;
2486 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2489 #define FOLL_WRITE 0x01 /* check pte is writable */
2490 #define FOLL_TOUCH 0x02 /* mark page accessed */
2491 #define FOLL_GET 0x04 /* do get_page on page */
2492 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2493 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2494 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2495 * and return without waiting upon it */
2496 #define FOLL_POPULATE 0x40 /* fault in page */
2497 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2498 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2499 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2500 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2501 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2502 #define FOLL_MLOCK 0x1000 /* lock present pages */
2503 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2504 #define FOLL_COW 0x4000 /* internal GUP flag */
2505 #define FOLL_ANON 0x8000 /* don't do file mappings */
2507 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2509 if (vm_fault & VM_FAULT_OOM)
2511 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2512 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2513 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2518 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2520 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2521 unsigned long size, pte_fn_t fn, void *data);
2524 #ifdef CONFIG_PAGE_POISONING
2525 extern bool page_poisoning_enabled(void);
2526 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2527 extern bool page_is_poisoned(struct page *page);
2529 static inline bool page_poisoning_enabled(void) { return false; }
2530 static inline void kernel_poison_pages(struct page *page, int numpages,
2532 static inline bool page_is_poisoned(struct page *page) { return false; }
2535 #ifdef CONFIG_DEBUG_PAGEALLOC
2536 extern bool _debug_pagealloc_enabled;
2537 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2539 static inline bool debug_pagealloc_enabled(void)
2541 return _debug_pagealloc_enabled;
2545 kernel_map_pages(struct page *page, int numpages, int enable)
2547 if (!debug_pagealloc_enabled())
2550 __kernel_map_pages(page, numpages, enable);
2552 #ifdef CONFIG_HIBERNATION
2553 extern bool kernel_page_present(struct page *page);
2554 #endif /* CONFIG_HIBERNATION */
2555 #else /* CONFIG_DEBUG_PAGEALLOC */
2557 kernel_map_pages(struct page *page, int numpages, int enable) {}
2558 #ifdef CONFIG_HIBERNATION
2559 static inline bool kernel_page_present(struct page *page) { return true; }
2560 #endif /* CONFIG_HIBERNATION */
2561 static inline bool debug_pagealloc_enabled(void)
2565 #endif /* CONFIG_DEBUG_PAGEALLOC */
2567 #ifdef __HAVE_ARCH_GATE_AREA
2568 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2569 extern int in_gate_area_no_mm(unsigned long addr);
2570 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2572 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2576 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2577 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2581 #endif /* __HAVE_ARCH_GATE_AREA */
2583 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2585 #ifdef CONFIG_SYSCTL
2586 extern int sysctl_drop_caches;
2587 int drop_caches_sysctl_handler(struct ctl_table *, int,
2588 void __user *, size_t *, loff_t *);
2591 void drop_slab(void);
2592 void drop_slab_node(int nid);
2595 #define randomize_va_space 0
2597 extern int randomize_va_space;
2600 const char * arch_vma_name(struct vm_area_struct *vma);
2601 void print_vma_addr(char *prefix, unsigned long rip);
2603 void sparse_mem_maps_populate_node(struct page **map_map,
2604 unsigned long pnum_begin,
2605 unsigned long pnum_end,
2606 unsigned long map_count,
2609 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2610 struct vmem_altmap *altmap);
2611 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2612 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2613 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2614 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2615 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2616 void *vmemmap_alloc_block(unsigned long size, int node);
2618 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2619 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2620 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2621 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2623 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2624 struct vmem_altmap *altmap);
2625 void vmemmap_populate_print_last(void);
2626 #ifdef CONFIG_MEMORY_HOTPLUG
2627 void vmemmap_free(unsigned long start, unsigned long end,
2628 struct vmem_altmap *altmap);
2630 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2631 unsigned long nr_pages);
2634 MF_COUNT_INCREASED = 1 << 0,
2635 MF_ACTION_REQUIRED = 1 << 1,
2636 MF_MUST_KILL = 1 << 2,
2637 MF_SOFT_OFFLINE = 1 << 3,
2639 extern int memory_failure(unsigned long pfn, int flags);
2640 extern void memory_failure_queue(unsigned long pfn, int flags);
2641 extern int unpoison_memory(unsigned long pfn);
2642 extern int get_hwpoison_page(struct page *page);
2643 #define put_hwpoison_page(page) put_page(page)
2644 extern int sysctl_memory_failure_early_kill;
2645 extern int sysctl_memory_failure_recovery;
2646 extern void shake_page(struct page *p, int access);
2647 extern atomic_long_t num_poisoned_pages __read_mostly;
2648 extern int soft_offline_page(struct page *page, int flags);
2652 * Error handlers for various types of pages.
2655 MF_IGNORED, /* Error: cannot be handled */
2656 MF_FAILED, /* Error: handling failed */
2657 MF_DELAYED, /* Will be handled later */
2658 MF_RECOVERED, /* Successfully recovered */
2661 enum mf_action_page_type {
2663 MF_MSG_KERNEL_HIGH_ORDER,
2665 MF_MSG_DIFFERENT_COMPOUND,
2666 MF_MSG_POISONED_HUGE,
2669 MF_MSG_NON_PMD_HUGE,
2670 MF_MSG_UNMAP_FAILED,
2671 MF_MSG_DIRTY_SWAPCACHE,
2672 MF_MSG_CLEAN_SWAPCACHE,
2673 MF_MSG_DIRTY_MLOCKED_LRU,
2674 MF_MSG_CLEAN_MLOCKED_LRU,
2675 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2676 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2679 MF_MSG_TRUNCATED_LRU,
2685 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2686 extern void clear_huge_page(struct page *page,
2687 unsigned long addr_hint,
2688 unsigned int pages_per_huge_page);
2689 extern void copy_user_huge_page(struct page *dst, struct page *src,
2690 unsigned long addr, struct vm_area_struct *vma,
2691 unsigned int pages_per_huge_page);
2692 extern long copy_huge_page_from_user(struct page *dst_page,
2693 const void __user *usr_src,
2694 unsigned int pages_per_huge_page,
2695 bool allow_pagefault);
2696 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2698 extern struct page_ext_operations debug_guardpage_ops;
2700 #ifdef CONFIG_DEBUG_PAGEALLOC
2701 extern unsigned int _debug_guardpage_minorder;
2702 extern bool _debug_guardpage_enabled;
2704 static inline unsigned int debug_guardpage_minorder(void)
2706 return _debug_guardpage_minorder;
2709 static inline bool debug_guardpage_enabled(void)
2711 return _debug_guardpage_enabled;
2714 static inline bool page_is_guard(struct page *page)
2716 struct page_ext *page_ext;
2718 if (!debug_guardpage_enabled())
2721 page_ext = lookup_page_ext(page);
2722 if (unlikely(!page_ext))
2725 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2728 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2729 static inline bool debug_guardpage_enabled(void) { return false; }
2730 static inline bool page_is_guard(struct page *page) { return false; }
2731 #endif /* CONFIG_DEBUG_PAGEALLOC */
2733 #if MAX_NUMNODES > 1
2734 void __init setup_nr_node_ids(void);
2736 static inline void setup_nr_node_ids(void) {}
2739 #endif /* __KERNEL__ */
2740 #endif /* _LINUX_MM_H */