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>
31 struct anon_vma_chain;
34 struct writeback_control;
37 void init_mm_internals(void);
39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr;
42 static inline void set_max_mapnr(unsigned long limit)
47 static inline void set_max_mapnr(unsigned long limit) { }
50 extern unsigned long totalram_pages;
51 extern void * high_memory;
52 extern int page_cluster;
55 extern int sysctl_legacy_va_layout;
57 #define sysctl_legacy_va_layout 0
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min;
62 extern const int mmap_rnd_bits_max;
63 extern int mmap_rnd_bits __read_mostly;
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min;
67 extern const int mmap_rnd_compat_bits_max;
68 extern int mmap_rnd_compat_bits __read_mostly;
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
84 #define lm_alias(x) __va(__pa_symbol(x))
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X) (0)
99 * On some architectures it is expensive to call memset() for small sizes.
100 * Those architectures should provide their own implementation of "struct page"
101 * zeroing by defining this macro in <asm/pgtable.h>.
103 #ifndef mm_zero_struct_page
104 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
108 * Default maximum number of active map areas, this limits the number of vmas
109 * per mm struct. Users can overwrite this number by sysctl but there is a
112 * When a program's coredump is generated as ELF format, a section is created
113 * per a vma. In ELF, the number of sections is represented in unsigned short.
114 * This means the number of sections should be smaller than 65535 at coredump.
115 * Because the kernel adds some informative sections to a image of program at
116 * generating coredump, we need some margin. The number of extra sections is
117 * 1-3 now and depends on arch. We use "5" as safe margin, here.
119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120 * not a hard limit any more. Although some userspace tools can be surprised by
123 #define MAPCOUNT_ELF_CORE_MARGIN (5)
124 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
126 extern int sysctl_max_map_count;
128 extern unsigned long sysctl_user_reserve_kbytes;
129 extern unsigned long sysctl_admin_reserve_kbytes;
131 extern int sysctl_overcommit_memory;
132 extern int sysctl_overcommit_ratio;
133 extern unsigned long sysctl_overcommit_kbytes;
135 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
137 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
140 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
142 /* to align the pointer to the (next) page boundary */
143 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
145 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
146 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
149 * Linux kernel virtual memory manager primitives.
150 * The idea being to have a "virtual" mm in the same way
151 * we have a virtual fs - giving a cleaner interface to the
152 * mm details, and allowing different kinds of memory mappings
153 * (from shared memory to executable loading to arbitrary
157 extern struct kmem_cache *vm_area_cachep;
160 extern struct rb_root nommu_region_tree;
161 extern struct rw_semaphore nommu_region_sem;
163 extern unsigned int kobjsize(const void *objp);
167 * vm_flags in vm_area_struct, see mm_types.h.
168 * When changing, update also include/trace/events/mmflags.h
170 #define VM_NONE 0x00000000
172 #define VM_READ 0x00000001 /* currently active flags */
173 #define VM_WRITE 0x00000002
174 #define VM_EXEC 0x00000004
175 #define VM_SHARED 0x00000008
177 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
178 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
179 #define VM_MAYWRITE 0x00000020
180 #define VM_MAYEXEC 0x00000040
181 #define VM_MAYSHARE 0x00000080
183 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
184 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
185 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
186 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
187 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
189 #define VM_LOCKED 0x00002000
190 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
192 /* Used by sys_madvise() */
193 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
194 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
196 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
197 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
198 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
199 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
200 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
201 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
202 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
203 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
204 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
206 #ifdef CONFIG_MEM_SOFT_DIRTY
207 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
209 # define VM_SOFTDIRTY 0
212 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
213 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
214 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
215 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
217 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
218 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
219 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
220 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
221 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
222 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
224 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
225 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
226 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
227 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
228 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
230 #if defined(CONFIG_X86)
231 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
232 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
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
236 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
237 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
239 #elif defined(CONFIG_PPC)
240 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
241 #elif defined(CONFIG_PARISC)
242 # define VM_GROWSUP VM_ARCH_1
243 #elif defined(CONFIG_METAG)
244 # define VM_GROWSUP VM_ARCH_1
245 #elif defined(CONFIG_IA64)
246 # define VM_GROWSUP VM_ARCH_1
247 #elif !defined(CONFIG_MMU)
248 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
251 #if defined(CONFIG_X86_INTEL_MPX)
252 /* MPX specific bounds table or bounds directory */
253 # define VM_MPX VM_HIGH_ARCH_4
255 # define VM_MPX VM_NONE
259 # define VM_GROWSUP VM_NONE
262 /* Bits set in the VMA until the stack is in its final location */
263 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
265 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
266 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
269 #ifdef CONFIG_STACK_GROWSUP
270 #define VM_STACK VM_GROWSUP
272 #define VM_STACK VM_GROWSDOWN
275 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
278 * Special vmas that are non-mergable, non-mlock()able.
279 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
281 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
283 /* This mask defines which mm->def_flags a process can inherit its parent */
284 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
286 /* This mask is used to clear all the VMA flags used by mlock */
287 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
290 * mapping from the currently active vm_flags protection bits (the
291 * low four bits) to a page protection mask..
293 extern pgprot_t protection_map[16];
295 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
296 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
297 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
298 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
299 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
300 #define FAULT_FLAG_TRIED 0x20 /* Second try */
301 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
302 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
303 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
305 #define FAULT_FLAG_TRACE \
306 { FAULT_FLAG_WRITE, "WRITE" }, \
307 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
308 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
309 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
310 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
311 { FAULT_FLAG_TRIED, "TRIED" }, \
312 { FAULT_FLAG_USER, "USER" }, \
313 { FAULT_FLAG_REMOTE, "REMOTE" }, \
314 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
317 * vm_fault is filled by the the pagefault handler and passed to the vma's
318 * ->fault function. The vma's ->fault is responsible for returning a bitmask
319 * of VM_FAULT_xxx flags that give details about how the fault was handled.
321 * MM layer fills up gfp_mask for page allocations but fault handler might
322 * alter it if its implementation requires a different allocation context.
324 * pgoff should be used in favour of virtual_address, if possible.
327 struct vm_area_struct *vma; /* Target VMA */
328 unsigned int flags; /* FAULT_FLAG_xxx flags */
329 gfp_t gfp_mask; /* gfp mask to be used for allocations */
330 pgoff_t pgoff; /* Logical page offset based on vma */
331 unsigned long address; /* Faulting virtual address */
332 pmd_t *pmd; /* Pointer to pmd entry matching
334 pud_t *pud; /* Pointer to pud entry matching
337 pte_t orig_pte; /* Value of PTE at the time of fault */
339 struct page *cow_page; /* Page handler may use for COW fault */
340 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
341 struct page *page; /* ->fault handlers should return a
342 * page here, unless VM_FAULT_NOPAGE
343 * is set (which is also implied by
346 /* These three entries are valid only while holding ptl lock */
347 pte_t *pte; /* Pointer to pte entry matching
348 * the 'address'. NULL if the page
349 * table hasn't been allocated.
351 spinlock_t *ptl; /* Page table lock.
352 * Protects pte page table if 'pte'
353 * is not NULL, otherwise pmd.
355 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
356 * vm_ops->map_pages() calls
357 * alloc_set_pte() from atomic context.
358 * do_fault_around() pre-allocates
359 * page table to avoid allocation from
364 /* page entry size for vm->huge_fault() */
365 enum page_entry_size {
372 * These are the virtual MM functions - opening of an area, closing and
373 * unmapping it (needed to keep files on disk up-to-date etc), pointer
374 * to the functions called when a no-page or a wp-page exception occurs.
376 struct vm_operations_struct {
377 void (*open)(struct vm_area_struct * area);
378 void (*close)(struct vm_area_struct * area);
379 int (*mremap)(struct vm_area_struct * area);
380 int (*fault)(struct vm_fault *vmf);
381 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
382 void (*map_pages)(struct vm_fault *vmf,
383 pgoff_t start_pgoff, pgoff_t end_pgoff);
385 /* notification that a previously read-only page is about to become
386 * writable, if an error is returned it will cause a SIGBUS */
387 int (*page_mkwrite)(struct vm_fault *vmf);
389 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
390 int (*pfn_mkwrite)(struct vm_fault *vmf);
392 /* called by access_process_vm when get_user_pages() fails, typically
393 * for use by special VMAs that can switch between memory and hardware
395 int (*access)(struct vm_area_struct *vma, unsigned long addr,
396 void *buf, int len, int write);
398 /* Called by the /proc/PID/maps code to ask the vma whether it
399 * has a special name. Returning non-NULL will also cause this
400 * vma to be dumped unconditionally. */
401 const char *(*name)(struct vm_area_struct *vma);
405 * set_policy() op must add a reference to any non-NULL @new mempolicy
406 * to hold the policy upon return. Caller should pass NULL @new to
407 * remove a policy and fall back to surrounding context--i.e. do not
408 * install a MPOL_DEFAULT policy, nor the task or system default
411 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
414 * get_policy() op must add reference [mpol_get()] to any policy at
415 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
416 * in mm/mempolicy.c will do this automatically.
417 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
418 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
419 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
420 * must return NULL--i.e., do not "fallback" to task or system default
423 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
427 * Called by vm_normal_page() for special PTEs to find the
428 * page for @addr. This is useful if the default behavior
429 * (using pte_page()) would not find the correct page.
431 struct page *(*find_special_page)(struct vm_area_struct *vma,
438 #define page_private(page) ((page)->private)
439 #define set_page_private(page, v) ((page)->private = (v))
441 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
442 static inline int pmd_devmap(pmd_t pmd)
446 static inline int pud_devmap(pud_t pud)
450 static inline int pgd_devmap(pgd_t pgd)
457 * FIXME: take this include out, include page-flags.h in
458 * files which need it (119 of them)
460 #include <linux/page-flags.h>
461 #include <linux/huge_mm.h>
464 * Methods to modify the page usage count.
466 * What counts for a page usage:
467 * - cache mapping (page->mapping)
468 * - private data (page->private)
469 * - page mapped in a task's page tables, each mapping
470 * is counted separately
472 * Also, many kernel routines increase the page count before a critical
473 * routine so they can be sure the page doesn't go away from under them.
477 * Drop a ref, return true if the refcount fell to zero (the page has no users)
479 static inline int put_page_testzero(struct page *page)
481 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
482 return page_ref_dec_and_test(page);
486 * Try to grab a ref unless the page has a refcount of zero, return false if
488 * This can be called when MMU is off so it must not access
489 * any of the virtual mappings.
491 static inline int get_page_unless_zero(struct page *page)
493 return page_ref_add_unless(page, 1, 0);
496 extern int page_is_ram(unsigned long pfn);
504 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
507 /* Support for virtually mapped pages */
508 struct page *vmalloc_to_page(const void *addr);
509 unsigned long vmalloc_to_pfn(const void *addr);
512 * Determine if an address is within the vmalloc range
514 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
515 * is no special casing required.
517 static inline bool is_vmalloc_addr(const void *x)
520 unsigned long addr = (unsigned long)x;
522 return addr >= VMALLOC_START && addr < VMALLOC_END;
528 extern int is_vmalloc_or_module_addr(const void *x);
530 static inline int is_vmalloc_or_module_addr(const void *x)
536 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
537 static inline void *kvmalloc(size_t size, gfp_t flags)
539 return kvmalloc_node(size, flags, NUMA_NO_NODE);
541 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
543 return kvmalloc_node(size, flags | __GFP_ZERO, node);
545 static inline void *kvzalloc(size_t size, gfp_t flags)
547 return kvmalloc(size, flags | __GFP_ZERO);
550 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
552 if (size != 0 && n > SIZE_MAX / size)
555 return kvmalloc(n * size, flags);
558 extern void kvfree(const void *addr);
560 static inline atomic_t *compound_mapcount_ptr(struct page *page)
562 return &page[1].compound_mapcount;
565 static inline int compound_mapcount(struct page *page)
567 VM_BUG_ON_PAGE(!PageCompound(page), page);
568 page = compound_head(page);
569 return atomic_read(compound_mapcount_ptr(page)) + 1;
573 * The atomic page->_mapcount, starts from -1: so that transitions
574 * both from it and to it can be tracked, using atomic_inc_and_test
575 * and atomic_add_negative(-1).
577 static inline void page_mapcount_reset(struct page *page)
579 atomic_set(&(page)->_mapcount, -1);
582 int __page_mapcount(struct page *page);
584 static inline int page_mapcount(struct page *page)
586 VM_BUG_ON_PAGE(PageSlab(page), page);
588 if (unlikely(PageCompound(page)))
589 return __page_mapcount(page);
590 return atomic_read(&page->_mapcount) + 1;
593 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
594 int total_mapcount(struct page *page);
595 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
597 static inline int total_mapcount(struct page *page)
599 return page_mapcount(page);
601 static inline int page_trans_huge_mapcount(struct page *page,
604 int mapcount = page_mapcount(page);
606 *total_mapcount = mapcount;
611 static inline struct page *virt_to_head_page(const void *x)
613 struct page *page = virt_to_page(x);
615 return compound_head(page);
618 void __put_page(struct page *page);
620 void put_pages_list(struct list_head *pages);
622 void split_page(struct page *page, unsigned int order);
625 * Compound pages have a destructor function. Provide a
626 * prototype for that function and accessor functions.
627 * These are _only_ valid on the head of a compound page.
629 typedef void compound_page_dtor(struct page *);
631 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
632 enum compound_dtor_id {
635 #ifdef CONFIG_HUGETLB_PAGE
638 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
643 extern compound_page_dtor * const compound_page_dtors[];
645 static inline void set_compound_page_dtor(struct page *page,
646 enum compound_dtor_id compound_dtor)
648 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
649 page[1].compound_dtor = compound_dtor;
652 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
654 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
655 return compound_page_dtors[page[1].compound_dtor];
658 static inline unsigned int compound_order(struct page *page)
662 return page[1].compound_order;
665 static inline void set_compound_order(struct page *page, unsigned int order)
667 page[1].compound_order = order;
670 void free_compound_page(struct page *page);
674 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
675 * servicing faults for write access. In the normal case, do always want
676 * pte_mkwrite. But get_user_pages can cause write faults for mappings
677 * that do not have writing enabled, when used by access_process_vm.
679 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
681 if (likely(vma->vm_flags & VM_WRITE))
682 pte = pte_mkwrite(pte);
686 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
688 int finish_fault(struct vm_fault *vmf);
689 int finish_mkwrite_fault(struct vm_fault *vmf);
693 * Multiple processes may "see" the same page. E.g. for untouched
694 * mappings of /dev/null, all processes see the same page full of
695 * zeroes, and text pages of executables and shared libraries have
696 * only one copy in memory, at most, normally.
698 * For the non-reserved pages, page_count(page) denotes a reference count.
699 * page_count() == 0 means the page is free. page->lru is then used for
700 * freelist management in the buddy allocator.
701 * page_count() > 0 means the page has been allocated.
703 * Pages are allocated by the slab allocator in order to provide memory
704 * to kmalloc and kmem_cache_alloc. In this case, the management of the
705 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
706 * unless a particular usage is carefully commented. (the responsibility of
707 * freeing the kmalloc memory is the caller's, of course).
709 * A page may be used by anyone else who does a __get_free_page().
710 * In this case, page_count still tracks the references, and should only
711 * be used through the normal accessor functions. The top bits of page->flags
712 * and page->virtual store page management information, but all other fields
713 * are unused and could be used privately, carefully. The management of this
714 * page is the responsibility of the one who allocated it, and those who have
715 * subsequently been given references to it.
717 * The other pages (we may call them "pagecache pages") are completely
718 * managed by the Linux memory manager: I/O, buffers, swapping etc.
719 * The following discussion applies only to them.
721 * A pagecache page contains an opaque `private' member, which belongs to the
722 * page's address_space. Usually, this is the address of a circular list of
723 * the page's disk buffers. PG_private must be set to tell the VM to call
724 * into the filesystem to release these pages.
726 * A page may belong to an inode's memory mapping. In this case, page->mapping
727 * is the pointer to the inode, and page->index is the file offset of the page,
728 * in units of PAGE_SIZE.
730 * If pagecache pages are not associated with an inode, they are said to be
731 * anonymous pages. These may become associated with the swapcache, and in that
732 * case PG_swapcache is set, and page->private is an offset into the swapcache.
734 * In either case (swapcache or inode backed), the pagecache itself holds one
735 * reference to the page. Setting PG_private should also increment the
736 * refcount. The each user mapping also has a reference to the page.
738 * The pagecache pages are stored in a per-mapping radix tree, which is
739 * rooted at mapping->page_tree, and indexed by offset.
740 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
741 * lists, we instead now tag pages as dirty/writeback in the radix tree.
743 * All pagecache pages may be subject to I/O:
744 * - inode pages may need to be read from disk,
745 * - inode pages which have been modified and are MAP_SHARED may need
746 * to be written back to the inode on disk,
747 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
748 * modified may need to be swapped out to swap space and (later) to be read
753 * The zone field is never updated after free_area_init_core()
754 * sets it, so none of the operations on it need to be atomic.
757 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
758 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
759 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
760 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
761 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
764 * Define the bit shifts to access each section. For non-existent
765 * sections we define the shift as 0; that plus a 0 mask ensures
766 * the compiler will optimise away reference to them.
768 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
769 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
770 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
771 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
773 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
774 #ifdef NODE_NOT_IN_PAGE_FLAGS
775 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
776 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
777 SECTIONS_PGOFF : ZONES_PGOFF)
779 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
780 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
781 NODES_PGOFF : ZONES_PGOFF)
784 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
786 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
787 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
790 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
791 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
792 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
793 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
794 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
796 static inline enum zone_type page_zonenum(const struct page *page)
798 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
801 #ifdef CONFIG_ZONE_DEVICE
802 static inline bool is_zone_device_page(const struct page *page)
804 return page_zonenum(page) == ZONE_DEVICE;
807 static inline bool is_zone_device_page(const struct page *page)
813 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
814 void put_zone_device_private_or_public_page(struct page *page);
815 DECLARE_STATIC_KEY_FALSE(device_private_key);
816 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
817 static inline bool is_device_private_page(const struct page *page);
818 static inline bool is_device_public_page(const struct page *page);
819 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
820 static inline void put_zone_device_private_or_public_page(struct page *page)
823 #define IS_HMM_ENABLED 0
824 static inline bool is_device_private_page(const struct page *page)
828 static inline bool is_device_public_page(const struct page *page)
832 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
835 static inline void get_page(struct page *page)
837 page = compound_head(page);
839 * Getting a normal page or the head of a compound page
840 * requires to already have an elevated page->_refcount.
842 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
846 static inline void put_page(struct page *page)
848 page = compound_head(page);
851 * For private device pages we need to catch refcount transition from
852 * 2 to 1, when refcount reach one it means the private device page is
853 * free and we need to inform the device driver through callback. See
854 * include/linux/memremap.h and HMM for details.
856 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
857 unlikely(is_device_public_page(page)))) {
858 put_zone_device_private_or_public_page(page);
862 if (put_page_testzero(page))
866 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
867 #define SECTION_IN_PAGE_FLAGS
871 * The identification function is mainly used by the buddy allocator for
872 * determining if two pages could be buddies. We are not really identifying
873 * the zone since we could be using the section number id if we do not have
874 * node id available in page flags.
875 * We only guarantee that it will return the same value for two combinable
878 static inline int page_zone_id(struct page *page)
880 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
883 static inline int zone_to_nid(struct zone *zone)
892 #ifdef NODE_NOT_IN_PAGE_FLAGS
893 extern int page_to_nid(const struct page *page);
895 static inline int page_to_nid(const struct page *page)
897 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
901 #ifdef CONFIG_NUMA_BALANCING
902 static inline int cpu_pid_to_cpupid(int cpu, int pid)
904 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
907 static inline int cpupid_to_pid(int cpupid)
909 return cpupid & LAST__PID_MASK;
912 static inline int cpupid_to_cpu(int cpupid)
914 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
917 static inline int cpupid_to_nid(int cpupid)
919 return cpu_to_node(cpupid_to_cpu(cpupid));
922 static inline bool cpupid_pid_unset(int cpupid)
924 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
927 static inline bool cpupid_cpu_unset(int cpupid)
929 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
932 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
934 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
937 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
938 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
939 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
941 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
944 static inline int page_cpupid_last(struct page *page)
946 return page->_last_cpupid;
948 static inline void page_cpupid_reset_last(struct page *page)
950 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
953 static inline int page_cpupid_last(struct page *page)
955 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
958 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
960 static inline void page_cpupid_reset_last(struct page *page)
962 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
964 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
965 #else /* !CONFIG_NUMA_BALANCING */
966 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
968 return page_to_nid(page); /* XXX */
971 static inline int page_cpupid_last(struct page *page)
973 return page_to_nid(page); /* XXX */
976 static inline int cpupid_to_nid(int cpupid)
981 static inline int cpupid_to_pid(int cpupid)
986 static inline int cpupid_to_cpu(int cpupid)
991 static inline int cpu_pid_to_cpupid(int nid, int pid)
996 static inline bool cpupid_pid_unset(int cpupid)
1001 static inline void page_cpupid_reset_last(struct page *page)
1005 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1009 #endif /* CONFIG_NUMA_BALANCING */
1011 static inline struct zone *page_zone(const struct page *page)
1013 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1016 static inline pg_data_t *page_pgdat(const struct page *page)
1018 return NODE_DATA(page_to_nid(page));
1021 #ifdef SECTION_IN_PAGE_FLAGS
1022 static inline void set_page_section(struct page *page, unsigned long section)
1024 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1025 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1028 static inline unsigned long page_to_section(const struct page *page)
1030 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1034 static inline void set_page_zone(struct page *page, enum zone_type zone)
1036 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1037 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1040 static inline void set_page_node(struct page *page, unsigned long node)
1042 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1043 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1046 static inline void set_page_links(struct page *page, enum zone_type zone,
1047 unsigned long node, unsigned long pfn)
1049 set_page_zone(page, zone);
1050 set_page_node(page, node);
1051 #ifdef SECTION_IN_PAGE_FLAGS
1052 set_page_section(page, pfn_to_section_nr(pfn));
1057 static inline struct mem_cgroup *page_memcg(struct page *page)
1059 return page->mem_cgroup;
1061 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1063 WARN_ON_ONCE(!rcu_read_lock_held());
1064 return READ_ONCE(page->mem_cgroup);
1067 static inline struct mem_cgroup *page_memcg(struct page *page)
1071 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1073 WARN_ON_ONCE(!rcu_read_lock_held());
1079 * Some inline functions in vmstat.h depend on page_zone()
1081 #include <linux/vmstat.h>
1083 static __always_inline void *lowmem_page_address(const struct page *page)
1085 return page_to_virt(page);
1088 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1089 #define HASHED_PAGE_VIRTUAL
1092 #if defined(WANT_PAGE_VIRTUAL)
1093 static inline void *page_address(const struct page *page)
1095 return page->virtual;
1097 static inline void set_page_address(struct page *page, void *address)
1099 page->virtual = address;
1101 #define page_address_init() do { } while(0)
1104 #if defined(HASHED_PAGE_VIRTUAL)
1105 void *page_address(const struct page *page);
1106 void set_page_address(struct page *page, void *virtual);
1107 void page_address_init(void);
1110 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1111 #define page_address(page) lowmem_page_address(page)
1112 #define set_page_address(page, address) do { } while(0)
1113 #define page_address_init() do { } while(0)
1116 extern void *page_rmapping(struct page *page);
1117 extern struct anon_vma *page_anon_vma(struct page *page);
1118 extern struct address_space *page_mapping(struct page *page);
1120 extern struct address_space *__page_file_mapping(struct page *);
1123 struct address_space *page_file_mapping(struct page *page)
1125 if (unlikely(PageSwapCache(page)))
1126 return __page_file_mapping(page);
1128 return page->mapping;
1131 extern pgoff_t __page_file_index(struct page *page);
1134 * Return the pagecache index of the passed page. Regular pagecache pages
1135 * use ->index whereas swapcache pages use swp_offset(->private)
1137 static inline pgoff_t page_index(struct page *page)
1139 if (unlikely(PageSwapCache(page)))
1140 return __page_file_index(page);
1144 bool page_mapped(struct page *page);
1145 struct address_space *page_mapping(struct page *page);
1148 * Return true only if the page has been allocated with
1149 * ALLOC_NO_WATERMARKS and the low watermark was not
1150 * met implying that the system is under some pressure.
1152 static inline bool page_is_pfmemalloc(struct page *page)
1155 * Page index cannot be this large so this must be
1156 * a pfmemalloc page.
1158 return page->index == -1UL;
1162 * Only to be called by the page allocator on a freshly allocated
1165 static inline void set_page_pfmemalloc(struct page *page)
1170 static inline void clear_page_pfmemalloc(struct page *page)
1176 * Different kinds of faults, as returned by handle_mm_fault().
1177 * Used to decide whether a process gets delivered SIGBUS or
1178 * just gets major/minor fault counters bumped up.
1181 #define VM_FAULT_OOM 0x0001
1182 #define VM_FAULT_SIGBUS 0x0002
1183 #define VM_FAULT_MAJOR 0x0004
1184 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1185 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1186 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1187 #define VM_FAULT_SIGSEGV 0x0040
1189 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1190 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1191 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1192 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1193 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1195 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1197 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1198 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1201 #define VM_FAULT_RESULT_TRACE \
1202 { VM_FAULT_OOM, "OOM" }, \
1203 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1204 { VM_FAULT_MAJOR, "MAJOR" }, \
1205 { VM_FAULT_WRITE, "WRITE" }, \
1206 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1207 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1208 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1209 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1210 { VM_FAULT_LOCKED, "LOCKED" }, \
1211 { VM_FAULT_RETRY, "RETRY" }, \
1212 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1213 { VM_FAULT_DONE_COW, "DONE_COW" }
1215 /* Encode hstate index for a hwpoisoned large page */
1216 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1217 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1220 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1222 extern void pagefault_out_of_memory(void);
1224 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1227 * Flags passed to show_mem() and show_free_areas() to suppress output in
1230 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1232 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1234 extern bool can_do_mlock(void);
1235 extern int user_shm_lock(size_t, struct user_struct *);
1236 extern void user_shm_unlock(size_t, struct user_struct *);
1239 * Parameter block passed down to zap_pte_range in exceptional cases.
1241 struct zap_details {
1242 struct address_space *check_mapping; /* Check page->mapping if set */
1243 pgoff_t first_index; /* Lowest page->index to unmap */
1244 pgoff_t last_index; /* Highest page->index to unmap */
1247 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1248 pte_t pte, bool with_public_device);
1249 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1251 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1254 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1255 unsigned long size);
1256 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1257 unsigned long size);
1258 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1259 unsigned long start, unsigned long end);
1262 * mm_walk - callbacks for walk_page_range
1263 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1264 * this handler should only handle pud_trans_huge() puds.
1265 * the pmd_entry or pte_entry callbacks will be used for
1267 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1268 * this handler is required to be able to handle
1269 * pmd_trans_huge() pmds. They may simply choose to
1270 * split_huge_page() instead of handling it explicitly.
1271 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1272 * @pte_hole: if set, called for each hole at all levels
1273 * @hugetlb_entry: if set, called for each hugetlb entry
1274 * @test_walk: caller specific callback function to determine whether
1275 * we walk over the current vma or not. Returning 0
1276 * value means "do page table walk over the current vma,"
1277 * and a negative one means "abort current page table walk
1278 * right now." 1 means "skip the current vma."
1279 * @mm: mm_struct representing the target process of page table walk
1280 * @vma: vma currently walked (NULL if walking outside vmas)
1281 * @private: private data for callbacks' usage
1283 * (see the comment on walk_page_range() for more details)
1286 int (*pud_entry)(pud_t *pud, unsigned long addr,
1287 unsigned long next, struct mm_walk *walk);
1288 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1289 unsigned long next, struct mm_walk *walk);
1290 int (*pte_entry)(pte_t *pte, unsigned long addr,
1291 unsigned long next, struct mm_walk *walk);
1292 int (*pte_hole)(unsigned long addr, unsigned long next,
1293 struct mm_walk *walk);
1294 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1295 unsigned long addr, unsigned long next,
1296 struct mm_walk *walk);
1297 int (*test_walk)(unsigned long addr, unsigned long next,
1298 struct mm_walk *walk);
1299 struct mm_struct *mm;
1300 struct vm_area_struct *vma;
1304 int walk_page_range(unsigned long addr, unsigned long end,
1305 struct mm_walk *walk);
1306 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1307 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1308 unsigned long end, unsigned long floor, unsigned long ceiling);
1309 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1310 struct vm_area_struct *vma);
1311 void unmap_mapping_range(struct address_space *mapping,
1312 loff_t const holebegin, loff_t const holelen, int even_cows);
1313 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1314 unsigned long *start, unsigned long *end,
1315 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1316 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1317 unsigned long *pfn);
1318 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1319 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1320 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1321 void *buf, int len, int write);
1323 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1324 loff_t const holebegin, loff_t const holelen)
1326 unmap_mapping_range(mapping, holebegin, holelen, 0);
1329 extern void truncate_pagecache(struct inode *inode, loff_t new);
1330 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1331 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1332 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1333 int truncate_inode_page(struct address_space *mapping, struct page *page);
1334 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1335 int invalidate_inode_page(struct page *page);
1338 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1339 unsigned int flags);
1340 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1341 unsigned long address, unsigned int fault_flags,
1344 static inline int handle_mm_fault(struct vm_area_struct *vma,
1345 unsigned long address, unsigned int flags)
1347 /* should never happen if there's no MMU */
1349 return VM_FAULT_SIGBUS;
1351 static inline int fixup_user_fault(struct task_struct *tsk,
1352 struct mm_struct *mm, unsigned long address,
1353 unsigned int fault_flags, bool *unlocked)
1355 /* should never happen if there's no MMU */
1361 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1362 unsigned int gup_flags);
1363 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1364 void *buf, int len, unsigned int gup_flags);
1365 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1366 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1368 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1369 unsigned long start, unsigned long nr_pages,
1370 unsigned int gup_flags, struct page **pages,
1371 struct vm_area_struct **vmas, int *locked);
1372 long get_user_pages(unsigned long start, unsigned long nr_pages,
1373 unsigned int gup_flags, struct page **pages,
1374 struct vm_area_struct **vmas);
1375 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1376 unsigned int gup_flags, struct page **pages, int *locked);
1377 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1378 struct page **pages, unsigned int gup_flags);
1379 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1380 struct page **pages);
1382 /* Container for pinned pfns / pages */
1383 struct frame_vector {
1384 unsigned int nr_allocated; /* Number of frames we have space for */
1385 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1386 bool got_ref; /* Did we pin pages by getting page ref? */
1387 bool is_pfns; /* Does array contain pages or pfns? */
1388 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1389 * pfns_vector_pages() or pfns_vector_pfns()
1393 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1394 void frame_vector_destroy(struct frame_vector *vec);
1395 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1396 unsigned int gup_flags, struct frame_vector *vec);
1397 void put_vaddr_frames(struct frame_vector *vec);
1398 int frame_vector_to_pages(struct frame_vector *vec);
1399 void frame_vector_to_pfns(struct frame_vector *vec);
1401 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1403 return vec->nr_frames;
1406 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1409 int err = frame_vector_to_pages(vec);
1412 return ERR_PTR(err);
1414 return (struct page **)(vec->ptrs);
1417 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1420 frame_vector_to_pfns(vec);
1421 return (unsigned long *)(vec->ptrs);
1425 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1426 struct page **pages);
1427 int get_kernel_page(unsigned long start, int write, struct page **pages);
1428 struct page *get_dump_page(unsigned long addr);
1430 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1431 extern void do_invalidatepage(struct page *page, unsigned int offset,
1432 unsigned int length);
1434 int __set_page_dirty_nobuffers(struct page *page);
1435 int __set_page_dirty_no_writeback(struct page *page);
1436 int redirty_page_for_writepage(struct writeback_control *wbc,
1438 void account_page_dirtied(struct page *page, struct address_space *mapping);
1439 void account_page_cleaned(struct page *page, struct address_space *mapping,
1440 struct bdi_writeback *wb);
1441 int set_page_dirty(struct page *page);
1442 int set_page_dirty_lock(struct page *page);
1443 void __cancel_dirty_page(struct page *page);
1444 static inline void cancel_dirty_page(struct page *page)
1446 /* Avoid atomic ops, locking, etc. when not actually needed. */
1447 if (PageDirty(page))
1448 __cancel_dirty_page(page);
1450 int clear_page_dirty_for_io(struct page *page);
1452 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1454 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1456 return !vma->vm_ops;
1461 * The vma_is_shmem is not inline because it is used only by slow
1462 * paths in userfault.
1464 bool vma_is_shmem(struct vm_area_struct *vma);
1466 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1469 int vma_is_stack_for_current(struct vm_area_struct *vma);
1471 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1472 unsigned long old_addr, struct vm_area_struct *new_vma,
1473 unsigned long new_addr, unsigned long len,
1474 bool need_rmap_locks);
1475 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1476 unsigned long end, pgprot_t newprot,
1477 int dirty_accountable, int prot_numa);
1478 extern int mprotect_fixup(struct vm_area_struct *vma,
1479 struct vm_area_struct **pprev, unsigned long start,
1480 unsigned long end, unsigned long newflags);
1483 * doesn't attempt to fault and will return short.
1485 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1486 struct page **pages);
1488 * per-process(per-mm_struct) statistics.
1490 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1492 long val = atomic_long_read(&mm->rss_stat.count[member]);
1494 #ifdef SPLIT_RSS_COUNTING
1496 * counter is updated in asynchronous manner and may go to minus.
1497 * But it's never be expected number for users.
1502 return (unsigned long)val;
1505 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1507 atomic_long_add(value, &mm->rss_stat.count[member]);
1510 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1512 atomic_long_inc(&mm->rss_stat.count[member]);
1515 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1517 atomic_long_dec(&mm->rss_stat.count[member]);
1520 /* Optimized variant when page is already known not to be PageAnon */
1521 static inline int mm_counter_file(struct page *page)
1523 if (PageSwapBacked(page))
1524 return MM_SHMEMPAGES;
1525 return MM_FILEPAGES;
1528 static inline int mm_counter(struct page *page)
1531 return MM_ANONPAGES;
1532 return mm_counter_file(page);
1535 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1537 return get_mm_counter(mm, MM_FILEPAGES) +
1538 get_mm_counter(mm, MM_ANONPAGES) +
1539 get_mm_counter(mm, MM_SHMEMPAGES);
1542 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1544 return max(mm->hiwater_rss, get_mm_rss(mm));
1547 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1549 return max(mm->hiwater_vm, mm->total_vm);
1552 static inline void update_hiwater_rss(struct mm_struct *mm)
1554 unsigned long _rss = get_mm_rss(mm);
1556 if ((mm)->hiwater_rss < _rss)
1557 (mm)->hiwater_rss = _rss;
1560 static inline void update_hiwater_vm(struct mm_struct *mm)
1562 if (mm->hiwater_vm < mm->total_vm)
1563 mm->hiwater_vm = mm->total_vm;
1566 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1568 mm->hiwater_rss = get_mm_rss(mm);
1571 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1572 struct mm_struct *mm)
1574 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1576 if (*maxrss < hiwater_rss)
1577 *maxrss = hiwater_rss;
1580 #if defined(SPLIT_RSS_COUNTING)
1581 void sync_mm_rss(struct mm_struct *mm);
1583 static inline void sync_mm_rss(struct mm_struct *mm)
1588 #ifndef __HAVE_ARCH_PTE_DEVMAP
1589 static inline int pte_devmap(pte_t pte)
1595 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1597 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1599 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1603 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1607 #ifdef __PAGETABLE_P4D_FOLDED
1608 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1609 unsigned long address)
1614 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1617 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1618 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1619 unsigned long address)
1623 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1624 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1627 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1629 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1631 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1634 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1636 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1640 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1641 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1642 unsigned long address)
1647 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1648 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1651 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1653 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1655 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1658 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1660 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1665 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1667 atomic_long_set(&mm->pgtables_bytes, 0);
1670 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1672 return atomic_long_read(&mm->pgtables_bytes);
1675 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1677 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1680 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1682 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1686 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1687 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1692 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1693 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1696 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1697 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1700 * The following ifdef needed to get the 4level-fixup.h header to work.
1701 * Remove it when 4level-fixup.h has been removed.
1703 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1705 #ifndef __ARCH_HAS_5LEVEL_HACK
1706 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1707 unsigned long address)
1709 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1710 NULL : p4d_offset(pgd, address);
1713 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1714 unsigned long address)
1716 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1717 NULL : pud_offset(p4d, address);
1719 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1721 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1723 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1724 NULL: pmd_offset(pud, address);
1726 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1728 #if USE_SPLIT_PTE_PTLOCKS
1729 #if ALLOC_SPLIT_PTLOCKS
1730 void __init ptlock_cache_init(void);
1731 extern bool ptlock_alloc(struct page *page);
1732 extern void ptlock_free(struct page *page);
1734 static inline spinlock_t *ptlock_ptr(struct page *page)
1738 #else /* ALLOC_SPLIT_PTLOCKS */
1739 static inline void ptlock_cache_init(void)
1743 static inline bool ptlock_alloc(struct page *page)
1748 static inline void ptlock_free(struct page *page)
1752 static inline spinlock_t *ptlock_ptr(struct page *page)
1756 #endif /* ALLOC_SPLIT_PTLOCKS */
1758 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1760 return ptlock_ptr(pmd_page(*pmd));
1763 static inline bool ptlock_init(struct page *page)
1766 * prep_new_page() initialize page->private (and therefore page->ptl)
1767 * with 0. Make sure nobody took it in use in between.
1769 * It can happen if arch try to use slab for page table allocation:
1770 * slab code uses page->slab_cache, which share storage with page->ptl.
1772 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1773 if (!ptlock_alloc(page))
1775 spin_lock_init(ptlock_ptr(page));
1779 /* Reset page->mapping so free_pages_check won't complain. */
1780 static inline void pte_lock_deinit(struct page *page)
1782 page->mapping = NULL;
1786 #else /* !USE_SPLIT_PTE_PTLOCKS */
1788 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1790 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1792 return &mm->page_table_lock;
1794 static inline void ptlock_cache_init(void) {}
1795 static inline bool ptlock_init(struct page *page) { return true; }
1796 static inline void pte_lock_deinit(struct page *page) {}
1797 #endif /* USE_SPLIT_PTE_PTLOCKS */
1799 static inline void pgtable_init(void)
1801 ptlock_cache_init();
1802 pgtable_cache_init();
1805 static inline bool pgtable_page_ctor(struct page *page)
1807 if (!ptlock_init(page))
1809 inc_zone_page_state(page, NR_PAGETABLE);
1813 static inline void pgtable_page_dtor(struct page *page)
1815 pte_lock_deinit(page);
1816 dec_zone_page_state(page, NR_PAGETABLE);
1819 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1821 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1822 pte_t *__pte = pte_offset_map(pmd, address); \
1828 #define pte_unmap_unlock(pte, ptl) do { \
1833 #define pte_alloc(mm, pmd, address) \
1834 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1836 #define pte_alloc_map(mm, pmd, address) \
1837 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1839 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1840 (pte_alloc(mm, pmd, address) ? \
1841 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1843 #define pte_alloc_kernel(pmd, address) \
1844 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1845 NULL: pte_offset_kernel(pmd, address))
1847 #if USE_SPLIT_PMD_PTLOCKS
1849 static struct page *pmd_to_page(pmd_t *pmd)
1851 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1852 return virt_to_page((void *)((unsigned long) pmd & mask));
1855 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1857 return ptlock_ptr(pmd_to_page(pmd));
1860 static inline bool pgtable_pmd_page_ctor(struct page *page)
1862 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1863 page->pmd_huge_pte = NULL;
1865 return ptlock_init(page);
1868 static inline void pgtable_pmd_page_dtor(struct page *page)
1870 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1871 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1876 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1880 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1882 return &mm->page_table_lock;
1885 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1886 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1888 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1892 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1894 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1900 * No scalability reason to split PUD locks yet, but follow the same pattern
1901 * as the PMD locks to make it easier if we decide to. The VM should not be
1902 * considered ready to switch to split PUD locks yet; there may be places
1903 * which need to be converted from page_table_lock.
1905 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1907 return &mm->page_table_lock;
1910 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1912 spinlock_t *ptl = pud_lockptr(mm, pud);
1918 extern void __init pagecache_init(void);
1919 extern void free_area_init(unsigned long * zones_size);
1920 extern void free_area_init_node(int nid, unsigned long * zones_size,
1921 unsigned long zone_start_pfn, unsigned long *zholes_size);
1922 extern void free_initmem(void);
1925 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1926 * into the buddy system. The freed pages will be poisoned with pattern
1927 * "poison" if it's within range [0, UCHAR_MAX].
1928 * Return pages freed into the buddy system.
1930 extern unsigned long free_reserved_area(void *start, void *end,
1931 int poison, char *s);
1933 #ifdef CONFIG_HIGHMEM
1935 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1936 * and totalram_pages.
1938 extern void free_highmem_page(struct page *page);
1941 extern void adjust_managed_page_count(struct page *page, long count);
1942 extern void mem_init_print_info(const char *str);
1944 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1946 /* Free the reserved page into the buddy system, so it gets managed. */
1947 static inline void __free_reserved_page(struct page *page)
1949 ClearPageReserved(page);
1950 init_page_count(page);
1954 static inline void free_reserved_page(struct page *page)
1956 __free_reserved_page(page);
1957 adjust_managed_page_count(page, 1);
1960 static inline void mark_page_reserved(struct page *page)
1962 SetPageReserved(page);
1963 adjust_managed_page_count(page, -1);
1967 * Default method to free all the __init memory into the buddy system.
1968 * The freed pages will be poisoned with pattern "poison" if it's within
1969 * range [0, UCHAR_MAX].
1970 * Return pages freed into the buddy system.
1972 static inline unsigned long free_initmem_default(int poison)
1974 extern char __init_begin[], __init_end[];
1976 return free_reserved_area(&__init_begin, &__init_end,
1977 poison, "unused kernel");
1980 static inline unsigned long get_num_physpages(void)
1983 unsigned long phys_pages = 0;
1985 for_each_online_node(nid)
1986 phys_pages += node_present_pages(nid);
1991 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1993 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1994 * zones, allocate the backing mem_map and account for memory holes in a more
1995 * architecture independent manner. This is a substitute for creating the
1996 * zone_sizes[] and zholes_size[] arrays and passing them to
1997 * free_area_init_node()
1999 * An architecture is expected to register range of page frames backed by
2000 * physical memory with memblock_add[_node]() before calling
2001 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2002 * usage, an architecture is expected to do something like
2004 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2006 * for_each_valid_physical_page_range()
2007 * memblock_add_node(base, size, nid)
2008 * free_area_init_nodes(max_zone_pfns);
2010 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2011 * registered physical page range. Similarly
2012 * sparse_memory_present_with_active_regions() calls memory_present() for
2013 * each range when SPARSEMEM is enabled.
2015 * See mm/page_alloc.c for more information on each function exposed by
2016 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2018 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2019 unsigned long node_map_pfn_alignment(void);
2020 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2021 unsigned long end_pfn);
2022 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2023 unsigned long end_pfn);
2024 extern void get_pfn_range_for_nid(unsigned int nid,
2025 unsigned long *start_pfn, unsigned long *end_pfn);
2026 extern unsigned long find_min_pfn_with_active_regions(void);
2027 extern void free_bootmem_with_active_regions(int nid,
2028 unsigned long max_low_pfn);
2029 extern void sparse_memory_present_with_active_regions(int nid);
2031 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2033 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2034 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2035 static inline int __early_pfn_to_nid(unsigned long pfn,
2036 struct mminit_pfnnid_cache *state)
2041 /* please see mm/page_alloc.c */
2042 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2043 /* there is a per-arch backend function. */
2044 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2045 struct mminit_pfnnid_cache *state);
2048 #ifdef CONFIG_HAVE_MEMBLOCK
2049 void zero_resv_unavail(void);
2051 static inline void zero_resv_unavail(void) {}
2054 extern void set_dma_reserve(unsigned long new_dma_reserve);
2055 extern void memmap_init_zone(unsigned long, int, unsigned long,
2056 unsigned long, enum memmap_context);
2057 extern void setup_per_zone_wmarks(void);
2058 extern int __meminit init_per_zone_wmark_min(void);
2059 extern void mem_init(void);
2060 extern void __init mmap_init(void);
2061 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2062 extern long si_mem_available(void);
2063 extern void si_meminfo(struct sysinfo * val);
2064 extern void si_meminfo_node(struct sysinfo *val, int nid);
2065 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2066 extern unsigned long arch_reserved_kernel_pages(void);
2069 extern __printf(3, 4)
2070 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2072 extern void setup_per_cpu_pageset(void);
2074 extern void zone_pcp_update(struct zone *zone);
2075 extern void zone_pcp_reset(struct zone *zone);
2078 extern int min_free_kbytes;
2079 extern int watermark_scale_factor;
2082 extern atomic_long_t mmap_pages_allocated;
2083 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2085 /* interval_tree.c */
2086 void vma_interval_tree_insert(struct vm_area_struct *node,
2087 struct rb_root_cached *root);
2088 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2089 struct vm_area_struct *prev,
2090 struct rb_root_cached *root);
2091 void vma_interval_tree_remove(struct vm_area_struct *node,
2092 struct rb_root_cached *root);
2093 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2094 unsigned long start, unsigned long last);
2095 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2096 unsigned long start, unsigned long last);
2098 #define vma_interval_tree_foreach(vma, root, start, last) \
2099 for (vma = vma_interval_tree_iter_first(root, start, last); \
2100 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2102 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2103 struct rb_root_cached *root);
2104 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2105 struct rb_root_cached *root);
2106 struct anon_vma_chain *
2107 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2108 unsigned long start, unsigned long last);
2109 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2110 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2111 #ifdef CONFIG_DEBUG_VM_RB
2112 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2115 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2116 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2117 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2120 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2121 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2122 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2123 struct vm_area_struct *expand);
2124 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2125 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2127 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2129 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2130 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2131 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2132 struct mempolicy *, struct vm_userfaultfd_ctx);
2133 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2134 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2135 unsigned long addr, int new_below);
2136 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2137 unsigned long addr, int new_below);
2138 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2139 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2140 struct rb_node **, struct rb_node *);
2141 extern void unlink_file_vma(struct vm_area_struct *);
2142 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2143 unsigned long addr, unsigned long len, pgoff_t pgoff,
2144 bool *need_rmap_locks);
2145 extern void exit_mmap(struct mm_struct *);
2147 static inline int check_data_rlimit(unsigned long rlim,
2149 unsigned long start,
2150 unsigned long end_data,
2151 unsigned long start_data)
2153 if (rlim < RLIM_INFINITY) {
2154 if (((new - start) + (end_data - start_data)) > rlim)
2161 extern int mm_take_all_locks(struct mm_struct *mm);
2162 extern void mm_drop_all_locks(struct mm_struct *mm);
2164 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2165 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2166 extern struct file *get_task_exe_file(struct task_struct *task);
2168 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2169 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2171 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2172 const struct vm_special_mapping *sm);
2173 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2174 unsigned long addr, unsigned long len,
2175 unsigned long flags,
2176 const struct vm_special_mapping *spec);
2177 /* This is an obsolete alternative to _install_special_mapping. */
2178 extern int install_special_mapping(struct mm_struct *mm,
2179 unsigned long addr, unsigned long len,
2180 unsigned long flags, struct page **pages);
2182 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2184 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2185 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2186 struct list_head *uf);
2187 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2188 unsigned long len, unsigned long prot, unsigned long flags,
2189 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2190 struct list_head *uf);
2191 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2192 struct list_head *uf);
2194 static inline unsigned long
2195 do_mmap_pgoff(struct file *file, unsigned long addr,
2196 unsigned long len, unsigned long prot, unsigned long flags,
2197 unsigned long pgoff, unsigned long *populate,
2198 struct list_head *uf)
2200 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2204 extern int __mm_populate(unsigned long addr, unsigned long len,
2206 static inline void mm_populate(unsigned long addr, unsigned long len)
2209 (void) __mm_populate(addr, len, 1);
2212 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2215 /* These take the mm semaphore themselves */
2216 extern int __must_check vm_brk(unsigned long, unsigned long);
2217 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2218 extern int vm_munmap(unsigned long, size_t);
2219 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2220 unsigned long, unsigned long,
2221 unsigned long, unsigned long);
2223 struct vm_unmapped_area_info {
2224 #define VM_UNMAPPED_AREA_TOPDOWN 1
2225 unsigned long flags;
2226 unsigned long length;
2227 unsigned long low_limit;
2228 unsigned long high_limit;
2229 unsigned long align_mask;
2230 unsigned long align_offset;
2233 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2234 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2237 * Search for an unmapped address range.
2239 * We are looking for a range that:
2240 * - does not intersect with any VMA;
2241 * - is contained within the [low_limit, high_limit) interval;
2242 * - is at least the desired size.
2243 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2245 static inline unsigned long
2246 vm_unmapped_area(struct vm_unmapped_area_info *info)
2248 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2249 return unmapped_area_topdown(info);
2251 return unmapped_area(info);
2255 extern void truncate_inode_pages(struct address_space *, loff_t);
2256 extern void truncate_inode_pages_range(struct address_space *,
2257 loff_t lstart, loff_t lend);
2258 extern void truncate_inode_pages_final(struct address_space *);
2260 /* generic vm_area_ops exported for stackable file systems */
2261 extern int filemap_fault(struct vm_fault *vmf);
2262 extern void filemap_map_pages(struct vm_fault *vmf,
2263 pgoff_t start_pgoff, pgoff_t end_pgoff);
2264 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2266 /* mm/page-writeback.c */
2267 int __must_check write_one_page(struct page *page);
2268 void task_dirty_inc(struct task_struct *tsk);
2271 #define VM_MAX_READAHEAD 128 /* kbytes */
2272 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2274 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2275 pgoff_t offset, unsigned long nr_to_read);
2277 void page_cache_sync_readahead(struct address_space *mapping,
2278 struct file_ra_state *ra,
2281 unsigned long size);
2283 void page_cache_async_readahead(struct address_space *mapping,
2284 struct file_ra_state *ra,
2288 unsigned long size);
2290 extern unsigned long stack_guard_gap;
2291 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2292 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2294 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2295 extern int expand_downwards(struct vm_area_struct *vma,
2296 unsigned long address);
2298 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2300 #define expand_upwards(vma, address) (0)
2303 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2304 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2305 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2306 struct vm_area_struct **pprev);
2308 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2309 NULL if none. Assume start_addr < end_addr. */
2310 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2312 struct vm_area_struct * vma = find_vma(mm,start_addr);
2314 if (vma && end_addr <= vma->vm_start)
2319 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2321 unsigned long vm_start = vma->vm_start;
2323 if (vma->vm_flags & VM_GROWSDOWN) {
2324 vm_start -= stack_guard_gap;
2325 if (vm_start > vma->vm_start)
2331 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2333 unsigned long vm_end = vma->vm_end;
2335 if (vma->vm_flags & VM_GROWSUP) {
2336 vm_end += stack_guard_gap;
2337 if (vm_end < vma->vm_end)
2338 vm_end = -PAGE_SIZE;
2343 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2345 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2348 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2349 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2350 unsigned long vm_start, unsigned long vm_end)
2352 struct vm_area_struct *vma = find_vma(mm, vm_start);
2354 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2361 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2362 void vma_set_page_prot(struct vm_area_struct *vma);
2364 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2368 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2370 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2374 #ifdef CONFIG_NUMA_BALANCING
2375 unsigned long change_prot_numa(struct vm_area_struct *vma,
2376 unsigned long start, unsigned long end);
2379 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2380 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2381 unsigned long pfn, unsigned long size, pgprot_t);
2382 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2383 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2385 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2386 unsigned long pfn, pgprot_t pgprot);
2387 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2389 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2391 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2394 struct page *follow_page_mask(struct vm_area_struct *vma,
2395 unsigned long address, unsigned int foll_flags,
2396 unsigned int *page_mask);
2398 static inline struct page *follow_page(struct vm_area_struct *vma,
2399 unsigned long address, unsigned int foll_flags)
2401 unsigned int unused_page_mask;
2402 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2405 #define FOLL_WRITE 0x01 /* check pte is writable */
2406 #define FOLL_TOUCH 0x02 /* mark page accessed */
2407 #define FOLL_GET 0x04 /* do get_page on page */
2408 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2409 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2410 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2411 * and return without waiting upon it */
2412 #define FOLL_POPULATE 0x40 /* fault in page */
2413 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2414 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2415 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2416 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2417 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2418 #define FOLL_MLOCK 0x1000 /* lock present pages */
2419 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2420 #define FOLL_COW 0x4000 /* internal GUP flag */
2422 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2424 if (vm_fault & VM_FAULT_OOM)
2426 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2427 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2428 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2433 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2435 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2436 unsigned long size, pte_fn_t fn, void *data);
2439 #ifdef CONFIG_PAGE_POISONING
2440 extern bool page_poisoning_enabled(void);
2441 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2442 extern bool page_is_poisoned(struct page *page);
2444 static inline bool page_poisoning_enabled(void) { return false; }
2445 static inline void kernel_poison_pages(struct page *page, int numpages,
2447 static inline bool page_is_poisoned(struct page *page) { return false; }
2450 #ifdef CONFIG_DEBUG_PAGEALLOC
2451 extern bool _debug_pagealloc_enabled;
2452 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2454 static inline bool debug_pagealloc_enabled(void)
2456 return _debug_pagealloc_enabled;
2460 kernel_map_pages(struct page *page, int numpages, int enable)
2462 if (!debug_pagealloc_enabled())
2465 __kernel_map_pages(page, numpages, enable);
2467 #ifdef CONFIG_HIBERNATION
2468 extern bool kernel_page_present(struct page *page);
2469 #endif /* CONFIG_HIBERNATION */
2470 #else /* CONFIG_DEBUG_PAGEALLOC */
2472 kernel_map_pages(struct page *page, int numpages, int enable) {}
2473 #ifdef CONFIG_HIBERNATION
2474 static inline bool kernel_page_present(struct page *page) { return true; }
2475 #endif /* CONFIG_HIBERNATION */
2476 static inline bool debug_pagealloc_enabled(void)
2480 #endif /* CONFIG_DEBUG_PAGEALLOC */
2482 #ifdef __HAVE_ARCH_GATE_AREA
2483 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2484 extern int in_gate_area_no_mm(unsigned long addr);
2485 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2487 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2491 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2492 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2496 #endif /* __HAVE_ARCH_GATE_AREA */
2498 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2500 #ifdef CONFIG_SYSCTL
2501 extern int sysctl_drop_caches;
2502 int drop_caches_sysctl_handler(struct ctl_table *, int,
2503 void __user *, size_t *, loff_t *);
2506 void drop_slab(void);
2507 void drop_slab_node(int nid);
2510 #define randomize_va_space 0
2512 extern int randomize_va_space;
2515 const char * arch_vma_name(struct vm_area_struct *vma);
2516 void print_vma_addr(char *prefix, unsigned long rip);
2518 void sparse_mem_maps_populate_node(struct page **map_map,
2519 unsigned long pnum_begin,
2520 unsigned long pnum_end,
2521 unsigned long map_count,
2524 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2525 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2526 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2527 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2528 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2529 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2530 void *vmemmap_alloc_block(unsigned long size, int node);
2532 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2533 struct vmem_altmap *altmap);
2534 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2536 return __vmemmap_alloc_block_buf(size, node, NULL);
2539 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2540 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2542 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2543 void vmemmap_populate_print_last(void);
2544 #ifdef CONFIG_MEMORY_HOTPLUG
2545 void vmemmap_free(unsigned long start, unsigned long end);
2547 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2548 unsigned long nr_pages);
2551 MF_COUNT_INCREASED = 1 << 0,
2552 MF_ACTION_REQUIRED = 1 << 1,
2553 MF_MUST_KILL = 1 << 2,
2554 MF_SOFT_OFFLINE = 1 << 3,
2556 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2557 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2558 extern int unpoison_memory(unsigned long pfn);
2559 extern int get_hwpoison_page(struct page *page);
2560 #define put_hwpoison_page(page) put_page(page)
2561 extern int sysctl_memory_failure_early_kill;
2562 extern int sysctl_memory_failure_recovery;
2563 extern void shake_page(struct page *p, int access);
2564 extern atomic_long_t num_poisoned_pages;
2565 extern int soft_offline_page(struct page *page, int flags);
2569 * Error handlers for various types of pages.
2572 MF_IGNORED, /* Error: cannot be handled */
2573 MF_FAILED, /* Error: handling failed */
2574 MF_DELAYED, /* Will be handled later */
2575 MF_RECOVERED, /* Successfully recovered */
2578 enum mf_action_page_type {
2580 MF_MSG_KERNEL_HIGH_ORDER,
2582 MF_MSG_DIFFERENT_COMPOUND,
2583 MF_MSG_POISONED_HUGE,
2586 MF_MSG_UNMAP_FAILED,
2587 MF_MSG_DIRTY_SWAPCACHE,
2588 MF_MSG_CLEAN_SWAPCACHE,
2589 MF_MSG_DIRTY_MLOCKED_LRU,
2590 MF_MSG_CLEAN_MLOCKED_LRU,
2591 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2592 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2595 MF_MSG_TRUNCATED_LRU,
2601 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2602 extern void clear_huge_page(struct page *page,
2603 unsigned long addr_hint,
2604 unsigned int pages_per_huge_page);
2605 extern void copy_user_huge_page(struct page *dst, struct page *src,
2606 unsigned long addr, struct vm_area_struct *vma,
2607 unsigned int pages_per_huge_page);
2608 extern long copy_huge_page_from_user(struct page *dst_page,
2609 const void __user *usr_src,
2610 unsigned int pages_per_huge_page,
2611 bool allow_pagefault);
2612 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2614 extern struct page_ext_operations debug_guardpage_ops;
2616 #ifdef CONFIG_DEBUG_PAGEALLOC
2617 extern unsigned int _debug_guardpage_minorder;
2618 extern bool _debug_guardpage_enabled;
2620 static inline unsigned int debug_guardpage_minorder(void)
2622 return _debug_guardpage_minorder;
2625 static inline bool debug_guardpage_enabled(void)
2627 return _debug_guardpage_enabled;
2630 static inline bool page_is_guard(struct page *page)
2632 struct page_ext *page_ext;
2634 if (!debug_guardpage_enabled())
2637 page_ext = lookup_page_ext(page);
2638 if (unlikely(!page_ext))
2641 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2644 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2645 static inline bool debug_guardpage_enabled(void) { return false; }
2646 static inline bool page_is_guard(struct page *page) { return false; }
2647 #endif /* CONFIG_DEBUG_PAGEALLOC */
2649 #if MAX_NUMNODES > 1
2650 void __init setup_nr_node_ids(void);
2652 static inline void setup_nr_node_ids(void) {}
2655 #endif /* __KERNEL__ */
2656 #endif /* _LINUX_MM_H */