4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/percpu-refcount.h>
20 #include <linux/bit_spinlock.h>
21 #include <linux/shrinker.h>
22 #include <linux/resource.h>
23 #include <linux/page_ext.h>
24 #include <linux/err.h>
25 #include <linux/page_ref.h>
26 #include <linux/memremap.h>
30 struct anon_vma_chain;
33 struct writeback_control;
36 void init_mm_internals(void);
38 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
39 extern unsigned long max_mapnr;
41 static inline void set_max_mapnr(unsigned long limit)
46 static inline void set_max_mapnr(unsigned long limit) { }
49 extern unsigned long totalram_pages;
50 extern void * high_memory;
51 extern int page_cluster;
54 extern int sysctl_legacy_va_layout;
56 #define sysctl_legacy_va_layout 0
59 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
60 extern const int mmap_rnd_bits_min;
61 extern const int mmap_rnd_bits_max;
62 extern int mmap_rnd_bits __read_mostly;
64 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
65 extern const int mmap_rnd_compat_bits_min;
66 extern const int mmap_rnd_compat_bits_max;
67 extern int mmap_rnd_compat_bits __read_mostly;
71 #include <asm/pgtable.h>
72 #include <asm/processor.h>
75 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
79 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
83 #define lm_alias(x) __va(__pa_symbol(x))
87 * To prevent common memory management code establishing
88 * a zero page mapping on a read fault.
89 * This macro should be defined within <asm/pgtable.h>.
90 * s390 does this to prevent multiplexing of hardware bits
91 * related to the physical page in case of virtualization.
93 #ifndef mm_forbids_zeropage
94 #define mm_forbids_zeropage(X) (0)
98 * Default maximum number of active map areas, this limits the number of vmas
99 * per mm struct. Users can overwrite this number by sysctl but there is a
102 * When a program's coredump is generated as ELF format, a section is created
103 * per a vma. In ELF, the number of sections is represented in unsigned short.
104 * This means the number of sections should be smaller than 65535 at coredump.
105 * Because the kernel adds some informative sections to a image of program at
106 * generating coredump, we need some margin. The number of extra sections is
107 * 1-3 now and depends on arch. We use "5" as safe margin, here.
109 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
110 * not a hard limit any more. Although some userspace tools can be surprised by
113 #define MAPCOUNT_ELF_CORE_MARGIN (5)
114 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
116 extern int sysctl_max_map_count;
118 extern unsigned long sysctl_user_reserve_kbytes;
119 extern unsigned long sysctl_admin_reserve_kbytes;
121 extern int sysctl_overcommit_memory;
122 extern int sysctl_overcommit_ratio;
123 extern unsigned long sysctl_overcommit_kbytes;
125 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
127 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
130 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
132 /* to align the pointer to the (next) page boundary */
133 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
135 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
136 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
139 * Linux kernel virtual memory manager primitives.
140 * The idea being to have a "virtual" mm in the same way
141 * we have a virtual fs - giving a cleaner interface to the
142 * mm details, and allowing different kinds of memory mappings
143 * (from shared memory to executable loading to arbitrary
147 extern struct kmem_cache *vm_area_cachep;
150 extern struct rb_root nommu_region_tree;
151 extern struct rw_semaphore nommu_region_sem;
153 extern unsigned int kobjsize(const void *objp);
157 * vm_flags in vm_area_struct, see mm_types.h.
158 * When changing, update also include/trace/events/mmflags.h
160 #define VM_NONE 0x00000000
162 #define VM_READ 0x00000001 /* currently active flags */
163 #define VM_WRITE 0x00000002
164 #define VM_EXEC 0x00000004
165 #define VM_SHARED 0x00000008
167 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
168 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
169 #define VM_MAYWRITE 0x00000020
170 #define VM_MAYEXEC 0x00000040
171 #define VM_MAYSHARE 0x00000080
173 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
174 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
175 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
176 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
177 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
179 #define VM_LOCKED 0x00002000
180 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
182 /* Used by sys_madvise() */
183 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
184 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
186 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
187 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
188 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
189 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
190 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
191 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
192 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
193 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
194 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
196 #ifdef CONFIG_MEM_SOFT_DIRTY
197 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
199 # define VM_SOFTDIRTY 0
202 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
203 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
204 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
205 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
207 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
208 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
209 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
210 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
211 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
212 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
213 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
214 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
215 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
216 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
217 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
218 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
220 #if defined(CONFIG_X86)
221 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
222 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
223 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
224 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
225 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
226 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
227 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
229 #elif defined(CONFIG_PPC)
230 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
231 #elif defined(CONFIG_PARISC)
232 # define VM_GROWSUP VM_ARCH_1
233 #elif defined(CONFIG_METAG)
234 # define VM_GROWSUP VM_ARCH_1
235 #elif defined(CONFIG_IA64)
236 # define VM_GROWSUP VM_ARCH_1
237 #elif !defined(CONFIG_MMU)
238 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
241 #if defined(CONFIG_X86_INTEL_MPX)
242 /* MPX specific bounds table or bounds directory */
243 # define VM_MPX VM_HIGH_ARCH_BIT_4
245 # define VM_MPX VM_NONE
249 # define VM_GROWSUP VM_NONE
252 /* Bits set in the VMA until the stack is in its final location */
253 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
255 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
256 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
259 #ifdef CONFIG_STACK_GROWSUP
260 #define VM_STACK VM_GROWSUP
262 #define VM_STACK VM_GROWSDOWN
265 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
268 * Special vmas that are non-mergable, non-mlock()able.
269 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
271 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
273 /* This mask defines which mm->def_flags a process can inherit its parent */
274 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
276 /* This mask is used to clear all the VMA flags used by mlock */
277 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
280 * mapping from the currently active vm_flags protection bits (the
281 * low four bits) to a page protection mask..
283 extern pgprot_t protection_map[16];
285 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
286 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
287 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
288 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
289 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
290 #define FAULT_FLAG_TRIED 0x20 /* Second try */
291 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
292 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
293 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
295 #define FAULT_FLAG_TRACE \
296 { FAULT_FLAG_WRITE, "WRITE" }, \
297 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
298 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
299 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
300 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
301 { FAULT_FLAG_TRIED, "TRIED" }, \
302 { FAULT_FLAG_USER, "USER" }, \
303 { FAULT_FLAG_REMOTE, "REMOTE" }, \
304 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
307 * vm_fault is filled by the the pagefault handler and passed to the vma's
308 * ->fault function. The vma's ->fault is responsible for returning a bitmask
309 * of VM_FAULT_xxx flags that give details about how the fault was handled.
311 * MM layer fills up gfp_mask for page allocations but fault handler might
312 * alter it if its implementation requires a different allocation context.
314 * pgoff should be used in favour of virtual_address, if possible.
317 struct vm_area_struct *vma; /* Target VMA */
318 unsigned int flags; /* FAULT_FLAG_xxx flags */
319 gfp_t gfp_mask; /* gfp mask to be used for allocations */
320 pgoff_t pgoff; /* Logical page offset based on vma */
321 unsigned long address; /* Faulting virtual address */
322 pmd_t *pmd; /* Pointer to pmd entry matching
324 pud_t *pud; /* Pointer to pud entry matching
327 pte_t orig_pte; /* Value of PTE at the time of fault */
329 struct page *cow_page; /* Page handler may use for COW fault */
330 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
331 struct page *page; /* ->fault handlers should return a
332 * page here, unless VM_FAULT_NOPAGE
333 * is set (which is also implied by
336 /* These three entries are valid only while holding ptl lock */
337 pte_t *pte; /* Pointer to pte entry matching
338 * the 'address'. NULL if the page
339 * table hasn't been allocated.
341 spinlock_t *ptl; /* Page table lock.
342 * Protects pte page table if 'pte'
343 * is not NULL, otherwise pmd.
345 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
346 * vm_ops->map_pages() calls
347 * alloc_set_pte() from atomic context.
348 * do_fault_around() pre-allocates
349 * page table to avoid allocation from
354 /* page entry size for vm->huge_fault() */
355 enum page_entry_size {
362 * These are the virtual MM functions - opening of an area, closing and
363 * unmapping it (needed to keep files on disk up-to-date etc), pointer
364 * to the functions called when a no-page or a wp-page exception occurs.
366 struct vm_operations_struct {
367 void (*open)(struct vm_area_struct * area);
368 void (*close)(struct vm_area_struct * area);
369 int (*mremap)(struct vm_area_struct * area);
370 int (*fault)(struct vm_fault *vmf);
371 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
372 void (*map_pages)(struct vm_fault *vmf,
373 pgoff_t start_pgoff, pgoff_t end_pgoff);
375 /* notification that a previously read-only page is about to become
376 * writable, if an error is returned it will cause a SIGBUS */
377 int (*page_mkwrite)(struct vm_fault *vmf);
379 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
380 int (*pfn_mkwrite)(struct vm_fault *vmf);
382 /* called by access_process_vm when get_user_pages() fails, typically
383 * for use by special VMAs that can switch between memory and hardware
385 int (*access)(struct vm_area_struct *vma, unsigned long addr,
386 void *buf, int len, int write);
388 /* Called by the /proc/PID/maps code to ask the vma whether it
389 * has a special name. Returning non-NULL will also cause this
390 * vma to be dumped unconditionally. */
391 const char *(*name)(struct vm_area_struct *vma);
395 * set_policy() op must add a reference to any non-NULL @new mempolicy
396 * to hold the policy upon return. Caller should pass NULL @new to
397 * remove a policy and fall back to surrounding context--i.e. do not
398 * install a MPOL_DEFAULT policy, nor the task or system default
401 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
404 * get_policy() op must add reference [mpol_get()] to any policy at
405 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
406 * in mm/mempolicy.c will do this automatically.
407 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
408 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
409 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
410 * must return NULL--i.e., do not "fallback" to task or system default
413 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
417 * Called by vm_normal_page() for special PTEs to find the
418 * page for @addr. This is useful if the default behavior
419 * (using pte_page()) would not find the correct page.
421 struct page *(*find_special_page)(struct vm_area_struct *vma,
428 #define page_private(page) ((page)->private)
429 #define set_page_private(page, v) ((page)->private = (v))
431 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
432 static inline int pmd_devmap(pmd_t pmd)
436 static inline int pud_devmap(pud_t pud)
440 static inline int pgd_devmap(pgd_t pgd)
447 * FIXME: take this include out, include page-flags.h in
448 * files which need it (119 of them)
450 #include <linux/page-flags.h>
451 #include <linux/huge_mm.h>
454 * Methods to modify the page usage count.
456 * What counts for a page usage:
457 * - cache mapping (page->mapping)
458 * - private data (page->private)
459 * - page mapped in a task's page tables, each mapping
460 * is counted separately
462 * Also, many kernel routines increase the page count before a critical
463 * routine so they can be sure the page doesn't go away from under them.
467 * Drop a ref, return true if the refcount fell to zero (the page has no users)
469 static inline int put_page_testzero(struct page *page)
471 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
472 return page_ref_dec_and_test(page);
476 * Try to grab a ref unless the page has a refcount of zero, return false if
478 * This can be called when MMU is off so it must not access
479 * any of the virtual mappings.
481 static inline int get_page_unless_zero(struct page *page)
483 return page_ref_add_unless(page, 1, 0);
486 extern int page_is_ram(unsigned long pfn);
494 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
497 /* Support for virtually mapped pages */
498 struct page *vmalloc_to_page(const void *addr);
499 unsigned long vmalloc_to_pfn(const void *addr);
502 * Determine if an address is within the vmalloc range
504 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
505 * is no special casing required.
507 static inline bool is_vmalloc_addr(const void *x)
510 unsigned long addr = (unsigned long)x;
512 return addr >= VMALLOC_START && addr < VMALLOC_END;
518 extern int is_vmalloc_or_module_addr(const void *x);
520 static inline int is_vmalloc_or_module_addr(const void *x)
526 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
527 static inline void *kvmalloc(size_t size, gfp_t flags)
529 return kvmalloc_node(size, flags, NUMA_NO_NODE);
531 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
533 return kvmalloc_node(size, flags | __GFP_ZERO, node);
535 static inline void *kvzalloc(size_t size, gfp_t flags)
537 return kvmalloc(size, flags | __GFP_ZERO);
540 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
542 if (size != 0 && n > SIZE_MAX / size)
545 return kvmalloc(n * size, flags);
548 extern void kvfree(const void *addr);
550 static inline atomic_t *compound_mapcount_ptr(struct page *page)
552 return &page[1].compound_mapcount;
555 static inline int compound_mapcount(struct page *page)
557 VM_BUG_ON_PAGE(!PageCompound(page), page);
558 page = compound_head(page);
559 return atomic_read(compound_mapcount_ptr(page)) + 1;
563 * The atomic page->_mapcount, starts from -1: so that transitions
564 * both from it and to it can be tracked, using atomic_inc_and_test
565 * and atomic_add_negative(-1).
567 static inline void page_mapcount_reset(struct page *page)
569 atomic_set(&(page)->_mapcount, -1);
572 int __page_mapcount(struct page *page);
574 static inline int page_mapcount(struct page *page)
576 VM_BUG_ON_PAGE(PageSlab(page), page);
578 if (unlikely(PageCompound(page)))
579 return __page_mapcount(page);
580 return atomic_read(&page->_mapcount) + 1;
583 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
584 int total_mapcount(struct page *page);
585 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
587 static inline int total_mapcount(struct page *page)
589 return page_mapcount(page);
591 static inline int page_trans_huge_mapcount(struct page *page,
594 int mapcount = page_mapcount(page);
596 *total_mapcount = mapcount;
601 static inline struct page *virt_to_head_page(const void *x)
603 struct page *page = virt_to_page(x);
605 return compound_head(page);
608 void __put_page(struct page *page);
610 void put_pages_list(struct list_head *pages);
612 void split_page(struct page *page, unsigned int order);
615 * Compound pages have a destructor function. Provide a
616 * prototype for that function and accessor functions.
617 * These are _only_ valid on the head of a compound page.
619 typedef void compound_page_dtor(struct page *);
621 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
622 enum compound_dtor_id {
625 #ifdef CONFIG_HUGETLB_PAGE
628 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
633 extern compound_page_dtor * const compound_page_dtors[];
635 static inline void set_compound_page_dtor(struct page *page,
636 enum compound_dtor_id compound_dtor)
638 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
639 page[1].compound_dtor = compound_dtor;
642 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
644 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
645 return compound_page_dtors[page[1].compound_dtor];
648 static inline unsigned int compound_order(struct page *page)
652 return page[1].compound_order;
655 static inline void set_compound_order(struct page *page, unsigned int order)
657 page[1].compound_order = order;
660 void free_compound_page(struct page *page);
664 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
665 * servicing faults for write access. In the normal case, do always want
666 * pte_mkwrite. But get_user_pages can cause write faults for mappings
667 * that do not have writing enabled, when used by access_process_vm.
669 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
671 if (likely(vma->vm_flags & VM_WRITE))
672 pte = pte_mkwrite(pte);
676 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
678 int finish_fault(struct vm_fault *vmf);
679 int finish_mkwrite_fault(struct vm_fault *vmf);
683 * Multiple processes may "see" the same page. E.g. for untouched
684 * mappings of /dev/null, all processes see the same page full of
685 * zeroes, and text pages of executables and shared libraries have
686 * only one copy in memory, at most, normally.
688 * For the non-reserved pages, page_count(page) denotes a reference count.
689 * page_count() == 0 means the page is free. page->lru is then used for
690 * freelist management in the buddy allocator.
691 * page_count() > 0 means the page has been allocated.
693 * Pages are allocated by the slab allocator in order to provide memory
694 * to kmalloc and kmem_cache_alloc. In this case, the management of the
695 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
696 * unless a particular usage is carefully commented. (the responsibility of
697 * freeing the kmalloc memory is the caller's, of course).
699 * A page may be used by anyone else who does a __get_free_page().
700 * In this case, page_count still tracks the references, and should only
701 * be used through the normal accessor functions. The top bits of page->flags
702 * and page->virtual store page management information, but all other fields
703 * are unused and could be used privately, carefully. The management of this
704 * page is the responsibility of the one who allocated it, and those who have
705 * subsequently been given references to it.
707 * The other pages (we may call them "pagecache pages") are completely
708 * managed by the Linux memory manager: I/O, buffers, swapping etc.
709 * The following discussion applies only to them.
711 * A pagecache page contains an opaque `private' member, which belongs to the
712 * page's address_space. Usually, this is the address of a circular list of
713 * the page's disk buffers. PG_private must be set to tell the VM to call
714 * into the filesystem to release these pages.
716 * A page may belong to an inode's memory mapping. In this case, page->mapping
717 * is the pointer to the inode, and page->index is the file offset of the page,
718 * in units of PAGE_SIZE.
720 * If pagecache pages are not associated with an inode, they are said to be
721 * anonymous pages. These may become associated with the swapcache, and in that
722 * case PG_swapcache is set, and page->private is an offset into the swapcache.
724 * In either case (swapcache or inode backed), the pagecache itself holds one
725 * reference to the page. Setting PG_private should also increment the
726 * refcount. The each user mapping also has a reference to the page.
728 * The pagecache pages are stored in a per-mapping radix tree, which is
729 * rooted at mapping->page_tree, and indexed by offset.
730 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
731 * lists, we instead now tag pages as dirty/writeback in the radix tree.
733 * All pagecache pages may be subject to I/O:
734 * - inode pages may need to be read from disk,
735 * - inode pages which have been modified and are MAP_SHARED may need
736 * to be written back to the inode on disk,
737 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
738 * modified may need to be swapped out to swap space and (later) to be read
743 * The zone field is never updated after free_area_init_core()
744 * sets it, so none of the operations on it need to be atomic.
747 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
748 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
749 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
750 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
751 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
754 * Define the bit shifts to access each section. For non-existent
755 * sections we define the shift as 0; that plus a 0 mask ensures
756 * the compiler will optimise away reference to them.
758 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
759 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
760 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
761 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
763 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
764 #ifdef NODE_NOT_IN_PAGE_FLAGS
765 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
766 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
767 SECTIONS_PGOFF : ZONES_PGOFF)
769 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
770 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
771 NODES_PGOFF : ZONES_PGOFF)
774 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
776 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
777 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
780 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
781 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
782 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
783 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
784 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
786 static inline enum zone_type page_zonenum(const struct page *page)
788 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
791 #ifdef CONFIG_ZONE_DEVICE
792 static inline bool is_zone_device_page(const struct page *page)
794 return page_zonenum(page) == ZONE_DEVICE;
797 static inline bool is_zone_device_page(const struct page *page)
803 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
804 void put_zone_device_private_or_public_page(struct page *page);
805 DECLARE_STATIC_KEY_FALSE(device_private_key);
806 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
807 static inline bool is_device_private_page(const struct page *page);
808 static inline bool is_device_public_page(const struct page *page);
809 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
810 static inline void put_zone_device_private_or_public_page(struct page *page)
813 #define IS_HMM_ENABLED 0
814 static inline bool is_device_private_page(const struct page *page)
818 static inline bool is_device_public_page(const struct page *page)
822 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
825 static inline void get_page(struct page *page)
827 page = compound_head(page);
829 * Getting a normal page or the head of a compound page
830 * requires to already have an elevated page->_refcount.
832 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
836 static inline void put_page(struct page *page)
838 page = compound_head(page);
841 * For private device pages we need to catch refcount transition from
842 * 2 to 1, when refcount reach one it means the private device page is
843 * free and we need to inform the device driver through callback. See
844 * include/linux/memremap.h and HMM for details.
846 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
847 unlikely(is_device_public_page(page)))) {
848 put_zone_device_private_or_public_page(page);
852 if (put_page_testzero(page))
856 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
857 #define SECTION_IN_PAGE_FLAGS
861 * The identification function is mainly used by the buddy allocator for
862 * determining if two pages could be buddies. We are not really identifying
863 * the zone since we could be using the section number id if we do not have
864 * node id available in page flags.
865 * We only guarantee that it will return the same value for two combinable
868 static inline int page_zone_id(struct page *page)
870 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
873 static inline int zone_to_nid(struct zone *zone)
882 #ifdef NODE_NOT_IN_PAGE_FLAGS
883 extern int page_to_nid(const struct page *page);
885 static inline int page_to_nid(const struct page *page)
887 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
891 #ifdef CONFIG_NUMA_BALANCING
892 static inline int cpu_pid_to_cpupid(int cpu, int pid)
894 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
897 static inline int cpupid_to_pid(int cpupid)
899 return cpupid & LAST__PID_MASK;
902 static inline int cpupid_to_cpu(int cpupid)
904 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
907 static inline int cpupid_to_nid(int cpupid)
909 return cpu_to_node(cpupid_to_cpu(cpupid));
912 static inline bool cpupid_pid_unset(int cpupid)
914 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
917 static inline bool cpupid_cpu_unset(int cpupid)
919 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
922 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
924 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
927 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
928 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
929 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
931 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
934 static inline int page_cpupid_last(struct page *page)
936 return page->_last_cpupid;
938 static inline void page_cpupid_reset_last(struct page *page)
940 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
943 static inline int page_cpupid_last(struct page *page)
945 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
948 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
950 static inline void page_cpupid_reset_last(struct page *page)
952 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
954 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
955 #else /* !CONFIG_NUMA_BALANCING */
956 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
958 return page_to_nid(page); /* XXX */
961 static inline int page_cpupid_last(struct page *page)
963 return page_to_nid(page); /* XXX */
966 static inline int cpupid_to_nid(int cpupid)
971 static inline int cpupid_to_pid(int cpupid)
976 static inline int cpupid_to_cpu(int cpupid)
981 static inline int cpu_pid_to_cpupid(int nid, int pid)
986 static inline bool cpupid_pid_unset(int cpupid)
991 static inline void page_cpupid_reset_last(struct page *page)
995 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
999 #endif /* CONFIG_NUMA_BALANCING */
1001 static inline struct zone *page_zone(const struct page *page)
1003 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1006 static inline pg_data_t *page_pgdat(const struct page *page)
1008 return NODE_DATA(page_to_nid(page));
1011 #ifdef SECTION_IN_PAGE_FLAGS
1012 static inline void set_page_section(struct page *page, unsigned long section)
1014 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1015 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1018 static inline unsigned long page_to_section(const struct page *page)
1020 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1024 static inline void set_page_zone(struct page *page, enum zone_type zone)
1026 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1027 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1030 static inline void set_page_node(struct page *page, unsigned long node)
1032 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1033 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1036 static inline void set_page_links(struct page *page, enum zone_type zone,
1037 unsigned long node, unsigned long pfn)
1039 set_page_zone(page, zone);
1040 set_page_node(page, node);
1041 #ifdef SECTION_IN_PAGE_FLAGS
1042 set_page_section(page, pfn_to_section_nr(pfn));
1047 static inline struct mem_cgroup *page_memcg(struct page *page)
1049 return page->mem_cgroup;
1051 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1053 WARN_ON_ONCE(!rcu_read_lock_held());
1054 return READ_ONCE(page->mem_cgroup);
1057 static inline struct mem_cgroup *page_memcg(struct page *page)
1061 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1063 WARN_ON_ONCE(!rcu_read_lock_held());
1069 * Some inline functions in vmstat.h depend on page_zone()
1071 #include <linux/vmstat.h>
1073 static __always_inline void *lowmem_page_address(const struct page *page)
1075 return page_to_virt(page);
1078 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1079 #define HASHED_PAGE_VIRTUAL
1082 #if defined(WANT_PAGE_VIRTUAL)
1083 static inline void *page_address(const struct page *page)
1085 return page->virtual;
1087 static inline void set_page_address(struct page *page, void *address)
1089 page->virtual = address;
1091 #define page_address_init() do { } while(0)
1094 #if defined(HASHED_PAGE_VIRTUAL)
1095 void *page_address(const struct page *page);
1096 void set_page_address(struct page *page, void *virtual);
1097 void page_address_init(void);
1100 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1101 #define page_address(page) lowmem_page_address(page)
1102 #define set_page_address(page, address) do { } while(0)
1103 #define page_address_init() do { } while(0)
1106 extern void *page_rmapping(struct page *page);
1107 extern struct anon_vma *page_anon_vma(struct page *page);
1108 extern struct address_space *page_mapping(struct page *page);
1110 extern struct address_space *__page_file_mapping(struct page *);
1113 struct address_space *page_file_mapping(struct page *page)
1115 if (unlikely(PageSwapCache(page)))
1116 return __page_file_mapping(page);
1118 return page->mapping;
1121 extern pgoff_t __page_file_index(struct page *page);
1124 * Return the pagecache index of the passed page. Regular pagecache pages
1125 * use ->index whereas swapcache pages use swp_offset(->private)
1127 static inline pgoff_t page_index(struct page *page)
1129 if (unlikely(PageSwapCache(page)))
1130 return __page_file_index(page);
1134 bool page_mapped(struct page *page);
1135 struct address_space *page_mapping(struct page *page);
1138 * Return true only if the page has been allocated with
1139 * ALLOC_NO_WATERMARKS and the low watermark was not
1140 * met implying that the system is under some pressure.
1142 static inline bool page_is_pfmemalloc(struct page *page)
1145 * Page index cannot be this large so this must be
1146 * a pfmemalloc page.
1148 return page->index == -1UL;
1152 * Only to be called by the page allocator on a freshly allocated
1155 static inline void set_page_pfmemalloc(struct page *page)
1160 static inline void clear_page_pfmemalloc(struct page *page)
1166 * Different kinds of faults, as returned by handle_mm_fault().
1167 * Used to decide whether a process gets delivered SIGBUS or
1168 * just gets major/minor fault counters bumped up.
1171 #define VM_FAULT_OOM 0x0001
1172 #define VM_FAULT_SIGBUS 0x0002
1173 #define VM_FAULT_MAJOR 0x0004
1174 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1175 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1176 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1177 #define VM_FAULT_SIGSEGV 0x0040
1179 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1180 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1181 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1182 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1183 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1185 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1187 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1188 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1191 #define VM_FAULT_RESULT_TRACE \
1192 { VM_FAULT_OOM, "OOM" }, \
1193 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1194 { VM_FAULT_MAJOR, "MAJOR" }, \
1195 { VM_FAULT_WRITE, "WRITE" }, \
1196 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1197 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1198 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1199 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1200 { VM_FAULT_LOCKED, "LOCKED" }, \
1201 { VM_FAULT_RETRY, "RETRY" }, \
1202 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1203 { VM_FAULT_DONE_COW, "DONE_COW" }
1205 /* Encode hstate index for a hwpoisoned large page */
1206 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1207 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1210 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1212 extern void pagefault_out_of_memory(void);
1214 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1217 * Flags passed to show_mem() and show_free_areas() to suppress output in
1220 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1222 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1224 extern bool can_do_mlock(void);
1225 extern int user_shm_lock(size_t, struct user_struct *);
1226 extern void user_shm_unlock(size_t, struct user_struct *);
1229 * Parameter block passed down to zap_pte_range in exceptional cases.
1231 struct zap_details {
1232 struct address_space *check_mapping; /* Check page->mapping if set */
1233 pgoff_t first_index; /* Lowest page->index to unmap */
1234 pgoff_t last_index; /* Highest page->index to unmap */
1237 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1238 pte_t pte, bool with_public_device);
1239 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1241 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1244 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1245 unsigned long size);
1246 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1247 unsigned long size);
1248 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1249 unsigned long start, unsigned long end);
1252 * mm_walk - callbacks for walk_page_range
1253 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1254 * this handler should only handle pud_trans_huge() puds.
1255 * the pmd_entry or pte_entry callbacks will be used for
1257 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1258 * this handler is required to be able to handle
1259 * pmd_trans_huge() pmds. They may simply choose to
1260 * split_huge_page() instead of handling it explicitly.
1261 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1262 * @pte_hole: if set, called for each hole at all levels
1263 * @hugetlb_entry: if set, called for each hugetlb entry
1264 * @test_walk: caller specific callback function to determine whether
1265 * we walk over the current vma or not. Returning 0
1266 * value means "do page table walk over the current vma,"
1267 * and a negative one means "abort current page table walk
1268 * right now." 1 means "skip the current vma."
1269 * @mm: mm_struct representing the target process of page table walk
1270 * @vma: vma currently walked (NULL if walking outside vmas)
1271 * @private: private data for callbacks' usage
1273 * (see the comment on walk_page_range() for more details)
1276 int (*pud_entry)(pud_t *pud, unsigned long addr,
1277 unsigned long next, struct mm_walk *walk);
1278 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1279 unsigned long next, struct mm_walk *walk);
1280 int (*pte_entry)(pte_t *pte, unsigned long addr,
1281 unsigned long next, struct mm_walk *walk);
1282 int (*pte_hole)(unsigned long addr, unsigned long next,
1283 struct mm_walk *walk);
1284 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1285 unsigned long addr, unsigned long next,
1286 struct mm_walk *walk);
1287 int (*test_walk)(unsigned long addr, unsigned long next,
1288 struct mm_walk *walk);
1289 struct mm_struct *mm;
1290 struct vm_area_struct *vma;
1294 int walk_page_range(unsigned long addr, unsigned long end,
1295 struct mm_walk *walk);
1296 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1297 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1298 unsigned long end, unsigned long floor, unsigned long ceiling);
1299 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1300 struct vm_area_struct *vma);
1301 void unmap_mapping_range(struct address_space *mapping,
1302 loff_t const holebegin, loff_t const holelen, int even_cows);
1303 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1304 unsigned long *start, unsigned long *end,
1305 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1306 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1307 unsigned long *pfn);
1308 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1309 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1310 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1311 void *buf, int len, int write);
1313 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1314 loff_t const holebegin, loff_t const holelen)
1316 unmap_mapping_range(mapping, holebegin, holelen, 0);
1319 extern void truncate_pagecache(struct inode *inode, loff_t new);
1320 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1321 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1322 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1323 int truncate_inode_page(struct address_space *mapping, struct page *page);
1324 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1325 int invalidate_inode_page(struct page *page);
1328 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1329 unsigned int flags);
1330 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1331 unsigned long address, unsigned int fault_flags,
1334 static inline int handle_mm_fault(struct vm_area_struct *vma,
1335 unsigned long address, unsigned int flags)
1337 /* should never happen if there's no MMU */
1339 return VM_FAULT_SIGBUS;
1341 static inline int fixup_user_fault(struct task_struct *tsk,
1342 struct mm_struct *mm, unsigned long address,
1343 unsigned int fault_flags, bool *unlocked)
1345 /* should never happen if there's no MMU */
1351 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1352 unsigned int gup_flags);
1353 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1354 void *buf, int len, unsigned int gup_flags);
1355 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1356 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1358 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1359 unsigned long start, unsigned long nr_pages,
1360 unsigned int gup_flags, struct page **pages,
1361 struct vm_area_struct **vmas, int *locked);
1362 long get_user_pages(unsigned long start, unsigned long nr_pages,
1363 unsigned int gup_flags, struct page **pages,
1364 struct vm_area_struct **vmas);
1365 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1366 unsigned int gup_flags, struct page **pages, int *locked);
1367 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1368 struct page **pages, unsigned int gup_flags);
1369 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1370 struct page **pages);
1372 /* Container for pinned pfns / pages */
1373 struct frame_vector {
1374 unsigned int nr_allocated; /* Number of frames we have space for */
1375 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1376 bool got_ref; /* Did we pin pages by getting page ref? */
1377 bool is_pfns; /* Does array contain pages or pfns? */
1378 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1379 * pfns_vector_pages() or pfns_vector_pfns()
1383 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1384 void frame_vector_destroy(struct frame_vector *vec);
1385 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1386 unsigned int gup_flags, struct frame_vector *vec);
1387 void put_vaddr_frames(struct frame_vector *vec);
1388 int frame_vector_to_pages(struct frame_vector *vec);
1389 void frame_vector_to_pfns(struct frame_vector *vec);
1391 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1393 return vec->nr_frames;
1396 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1399 int err = frame_vector_to_pages(vec);
1402 return ERR_PTR(err);
1404 return (struct page **)(vec->ptrs);
1407 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1410 frame_vector_to_pfns(vec);
1411 return (unsigned long *)(vec->ptrs);
1415 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1416 struct page **pages);
1417 int get_kernel_page(unsigned long start, int write, struct page **pages);
1418 struct page *get_dump_page(unsigned long addr);
1420 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1421 extern void do_invalidatepage(struct page *page, unsigned int offset,
1422 unsigned int length);
1424 int __set_page_dirty_nobuffers(struct page *page);
1425 int __set_page_dirty_no_writeback(struct page *page);
1426 int redirty_page_for_writepage(struct writeback_control *wbc,
1428 void account_page_dirtied(struct page *page, struct address_space *mapping);
1429 void account_page_cleaned(struct page *page, struct address_space *mapping,
1430 struct bdi_writeback *wb);
1431 int set_page_dirty(struct page *page);
1432 int set_page_dirty_lock(struct page *page);
1433 void cancel_dirty_page(struct page *page);
1434 int clear_page_dirty_for_io(struct page *page);
1436 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1438 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1440 return !vma->vm_ops;
1445 * The vma_is_shmem is not inline because it is used only by slow
1446 * paths in userfault.
1448 bool vma_is_shmem(struct vm_area_struct *vma);
1450 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1453 int vma_is_stack_for_current(struct vm_area_struct *vma);
1455 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1456 unsigned long old_addr, struct vm_area_struct *new_vma,
1457 unsigned long new_addr, unsigned long len,
1458 bool need_rmap_locks);
1459 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1460 unsigned long end, pgprot_t newprot,
1461 int dirty_accountable, int prot_numa);
1462 extern int mprotect_fixup(struct vm_area_struct *vma,
1463 struct vm_area_struct **pprev, unsigned long start,
1464 unsigned long end, unsigned long newflags);
1467 * doesn't attempt to fault and will return short.
1469 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1470 struct page **pages);
1472 * per-process(per-mm_struct) statistics.
1474 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1476 long val = atomic_long_read(&mm->rss_stat.count[member]);
1478 #ifdef SPLIT_RSS_COUNTING
1480 * counter is updated in asynchronous manner and may go to minus.
1481 * But it's never be expected number for users.
1486 return (unsigned long)val;
1489 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1491 atomic_long_add(value, &mm->rss_stat.count[member]);
1494 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1496 atomic_long_inc(&mm->rss_stat.count[member]);
1499 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1501 atomic_long_dec(&mm->rss_stat.count[member]);
1504 /* Optimized variant when page is already known not to be PageAnon */
1505 static inline int mm_counter_file(struct page *page)
1507 if (PageSwapBacked(page))
1508 return MM_SHMEMPAGES;
1509 return MM_FILEPAGES;
1512 static inline int mm_counter(struct page *page)
1515 return MM_ANONPAGES;
1516 return mm_counter_file(page);
1519 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1521 return get_mm_counter(mm, MM_FILEPAGES) +
1522 get_mm_counter(mm, MM_ANONPAGES) +
1523 get_mm_counter(mm, MM_SHMEMPAGES);
1526 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1528 return max(mm->hiwater_rss, get_mm_rss(mm));
1531 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1533 return max(mm->hiwater_vm, mm->total_vm);
1536 static inline void update_hiwater_rss(struct mm_struct *mm)
1538 unsigned long _rss = get_mm_rss(mm);
1540 if ((mm)->hiwater_rss < _rss)
1541 (mm)->hiwater_rss = _rss;
1544 static inline void update_hiwater_vm(struct mm_struct *mm)
1546 if (mm->hiwater_vm < mm->total_vm)
1547 mm->hiwater_vm = mm->total_vm;
1550 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1552 mm->hiwater_rss = get_mm_rss(mm);
1555 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1556 struct mm_struct *mm)
1558 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1560 if (*maxrss < hiwater_rss)
1561 *maxrss = hiwater_rss;
1564 #if defined(SPLIT_RSS_COUNTING)
1565 void sync_mm_rss(struct mm_struct *mm);
1567 static inline void sync_mm_rss(struct mm_struct *mm)
1572 #ifndef __HAVE_ARCH_PTE_DEVMAP
1573 static inline int pte_devmap(pte_t pte)
1579 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1581 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1583 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1587 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1591 #ifdef __PAGETABLE_P4D_FOLDED
1592 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1593 unsigned long address)
1598 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1601 #ifdef __PAGETABLE_PUD_FOLDED
1602 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1603 unsigned long address)
1608 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1611 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1612 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1613 unsigned long address)
1618 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1620 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1625 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1626 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1629 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1631 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1633 atomic_long_set(&mm->nr_pmds, 0);
1636 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1638 return atomic_long_read(&mm->nr_pmds);
1641 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1643 atomic_long_inc(&mm->nr_pmds);
1646 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1648 atomic_long_dec(&mm->nr_pmds);
1652 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1653 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1656 * The following ifdef needed to get the 4level-fixup.h header to work.
1657 * Remove it when 4level-fixup.h has been removed.
1659 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1661 #ifndef __ARCH_HAS_5LEVEL_HACK
1662 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1663 unsigned long address)
1665 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1666 NULL : p4d_offset(pgd, address);
1669 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1670 unsigned long address)
1672 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1673 NULL : pud_offset(p4d, address);
1675 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1677 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1679 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1680 NULL: pmd_offset(pud, address);
1682 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1684 #if USE_SPLIT_PTE_PTLOCKS
1685 #if ALLOC_SPLIT_PTLOCKS
1686 void __init ptlock_cache_init(void);
1687 extern bool ptlock_alloc(struct page *page);
1688 extern void ptlock_free(struct page *page);
1690 static inline spinlock_t *ptlock_ptr(struct page *page)
1694 #else /* ALLOC_SPLIT_PTLOCKS */
1695 static inline void ptlock_cache_init(void)
1699 static inline bool ptlock_alloc(struct page *page)
1704 static inline void ptlock_free(struct page *page)
1708 static inline spinlock_t *ptlock_ptr(struct page *page)
1712 #endif /* ALLOC_SPLIT_PTLOCKS */
1714 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1716 return ptlock_ptr(pmd_page(*pmd));
1719 static inline bool ptlock_init(struct page *page)
1722 * prep_new_page() initialize page->private (and therefore page->ptl)
1723 * with 0. Make sure nobody took it in use in between.
1725 * It can happen if arch try to use slab for page table allocation:
1726 * slab code uses page->slab_cache, which share storage with page->ptl.
1728 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1729 if (!ptlock_alloc(page))
1731 spin_lock_init(ptlock_ptr(page));
1735 /* Reset page->mapping so free_pages_check won't complain. */
1736 static inline void pte_lock_deinit(struct page *page)
1738 page->mapping = NULL;
1742 #else /* !USE_SPLIT_PTE_PTLOCKS */
1744 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1746 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1748 return &mm->page_table_lock;
1750 static inline void ptlock_cache_init(void) {}
1751 static inline bool ptlock_init(struct page *page) { return true; }
1752 static inline void pte_lock_deinit(struct page *page) {}
1753 #endif /* USE_SPLIT_PTE_PTLOCKS */
1755 static inline void pgtable_init(void)
1757 ptlock_cache_init();
1758 pgtable_cache_init();
1761 static inline bool pgtable_page_ctor(struct page *page)
1763 if (!ptlock_init(page))
1765 inc_zone_page_state(page, NR_PAGETABLE);
1769 static inline void pgtable_page_dtor(struct page *page)
1771 pte_lock_deinit(page);
1772 dec_zone_page_state(page, NR_PAGETABLE);
1775 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1777 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1778 pte_t *__pte = pte_offset_map(pmd, address); \
1784 #define pte_unmap_unlock(pte, ptl) do { \
1789 #define pte_alloc(mm, pmd, address) \
1790 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1792 #define pte_alloc_map(mm, pmd, address) \
1793 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1795 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1796 (pte_alloc(mm, pmd, address) ? \
1797 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1799 #define pte_alloc_kernel(pmd, address) \
1800 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1801 NULL: pte_offset_kernel(pmd, address))
1803 #if USE_SPLIT_PMD_PTLOCKS
1805 static struct page *pmd_to_page(pmd_t *pmd)
1807 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1808 return virt_to_page((void *)((unsigned long) pmd & mask));
1811 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1813 return ptlock_ptr(pmd_to_page(pmd));
1816 static inline bool pgtable_pmd_page_ctor(struct page *page)
1818 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1819 page->pmd_huge_pte = NULL;
1821 return ptlock_init(page);
1824 static inline void pgtable_pmd_page_dtor(struct page *page)
1826 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1827 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1832 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1836 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1838 return &mm->page_table_lock;
1841 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1842 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1844 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1848 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1850 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1856 * No scalability reason to split PUD locks yet, but follow the same pattern
1857 * as the PMD locks to make it easier if we decide to. The VM should not be
1858 * considered ready to switch to split PUD locks yet; there may be places
1859 * which need to be converted from page_table_lock.
1861 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1863 return &mm->page_table_lock;
1866 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1868 spinlock_t *ptl = pud_lockptr(mm, pud);
1874 extern void __init pagecache_init(void);
1875 extern void free_area_init(unsigned long * zones_size);
1876 extern void free_area_init_node(int nid, unsigned long * zones_size,
1877 unsigned long zone_start_pfn, unsigned long *zholes_size);
1878 extern void free_initmem(void);
1881 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1882 * into the buddy system. The freed pages will be poisoned with pattern
1883 * "poison" if it's within range [0, UCHAR_MAX].
1884 * Return pages freed into the buddy system.
1886 extern unsigned long free_reserved_area(void *start, void *end,
1887 int poison, char *s);
1889 #ifdef CONFIG_HIGHMEM
1891 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1892 * and totalram_pages.
1894 extern void free_highmem_page(struct page *page);
1897 extern void adjust_managed_page_count(struct page *page, long count);
1898 extern void mem_init_print_info(const char *str);
1900 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1902 /* Free the reserved page into the buddy system, so it gets managed. */
1903 static inline void __free_reserved_page(struct page *page)
1905 ClearPageReserved(page);
1906 init_page_count(page);
1910 static inline void free_reserved_page(struct page *page)
1912 __free_reserved_page(page);
1913 adjust_managed_page_count(page, 1);
1916 static inline void mark_page_reserved(struct page *page)
1918 SetPageReserved(page);
1919 adjust_managed_page_count(page, -1);
1923 * Default method to free all the __init memory into the buddy system.
1924 * The freed pages will be poisoned with pattern "poison" if it's within
1925 * range [0, UCHAR_MAX].
1926 * Return pages freed into the buddy system.
1928 static inline unsigned long free_initmem_default(int poison)
1930 extern char __init_begin[], __init_end[];
1932 return free_reserved_area(&__init_begin, &__init_end,
1933 poison, "unused kernel");
1936 static inline unsigned long get_num_physpages(void)
1939 unsigned long phys_pages = 0;
1941 for_each_online_node(nid)
1942 phys_pages += node_present_pages(nid);
1947 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1949 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1950 * zones, allocate the backing mem_map and account for memory holes in a more
1951 * architecture independent manner. This is a substitute for creating the
1952 * zone_sizes[] and zholes_size[] arrays and passing them to
1953 * free_area_init_node()
1955 * An architecture is expected to register range of page frames backed by
1956 * physical memory with memblock_add[_node]() before calling
1957 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1958 * usage, an architecture is expected to do something like
1960 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1962 * for_each_valid_physical_page_range()
1963 * memblock_add_node(base, size, nid)
1964 * free_area_init_nodes(max_zone_pfns);
1966 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1967 * registered physical page range. Similarly
1968 * sparse_memory_present_with_active_regions() calls memory_present() for
1969 * each range when SPARSEMEM is enabled.
1971 * See mm/page_alloc.c for more information on each function exposed by
1972 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1974 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1975 unsigned long node_map_pfn_alignment(void);
1976 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1977 unsigned long end_pfn);
1978 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1979 unsigned long end_pfn);
1980 extern void get_pfn_range_for_nid(unsigned int nid,
1981 unsigned long *start_pfn, unsigned long *end_pfn);
1982 extern unsigned long find_min_pfn_with_active_regions(void);
1983 extern void free_bootmem_with_active_regions(int nid,
1984 unsigned long max_low_pfn);
1985 extern void sparse_memory_present_with_active_regions(int nid);
1987 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1989 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1990 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1991 static inline int __early_pfn_to_nid(unsigned long pfn,
1992 struct mminit_pfnnid_cache *state)
1997 /* please see mm/page_alloc.c */
1998 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1999 /* there is a per-arch backend function. */
2000 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2001 struct mminit_pfnnid_cache *state);
2004 extern void set_dma_reserve(unsigned long new_dma_reserve);
2005 extern void memmap_init_zone(unsigned long, int, unsigned long,
2006 unsigned long, enum memmap_context);
2007 extern void setup_per_zone_wmarks(void);
2008 extern int __meminit init_per_zone_wmark_min(void);
2009 extern void mem_init(void);
2010 extern void __init mmap_init(void);
2011 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2012 extern long si_mem_available(void);
2013 extern void si_meminfo(struct sysinfo * val);
2014 extern void si_meminfo_node(struct sysinfo *val, int nid);
2015 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2016 extern unsigned long arch_reserved_kernel_pages(void);
2019 extern __printf(3, 4)
2020 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2022 extern void setup_per_cpu_pageset(void);
2024 extern void zone_pcp_update(struct zone *zone);
2025 extern void zone_pcp_reset(struct zone *zone);
2028 extern int min_free_kbytes;
2029 extern int watermark_scale_factor;
2032 extern atomic_long_t mmap_pages_allocated;
2033 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2035 /* interval_tree.c */
2036 void vma_interval_tree_insert(struct vm_area_struct *node,
2037 struct rb_root *root);
2038 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2039 struct vm_area_struct *prev,
2040 struct rb_root *root);
2041 void vma_interval_tree_remove(struct vm_area_struct *node,
2042 struct rb_root *root);
2043 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
2044 unsigned long start, unsigned long last);
2045 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2046 unsigned long start, unsigned long last);
2048 #define vma_interval_tree_foreach(vma, root, start, last) \
2049 for (vma = vma_interval_tree_iter_first(root, start, last); \
2050 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2052 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2053 struct rb_root *root);
2054 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2055 struct rb_root *root);
2056 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
2057 struct rb_root *root, unsigned long start, unsigned long last);
2058 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2059 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2060 #ifdef CONFIG_DEBUG_VM_RB
2061 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2064 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2065 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2066 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2069 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2070 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2071 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2072 struct vm_area_struct *expand);
2073 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2074 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2076 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2078 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2079 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2080 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2081 struct mempolicy *, struct vm_userfaultfd_ctx);
2082 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2083 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2084 unsigned long addr, int new_below);
2085 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2086 unsigned long addr, int new_below);
2087 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2088 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2089 struct rb_node **, struct rb_node *);
2090 extern void unlink_file_vma(struct vm_area_struct *);
2091 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2092 unsigned long addr, unsigned long len, pgoff_t pgoff,
2093 bool *need_rmap_locks);
2094 extern void exit_mmap(struct mm_struct *);
2096 static inline int check_data_rlimit(unsigned long rlim,
2098 unsigned long start,
2099 unsigned long end_data,
2100 unsigned long start_data)
2102 if (rlim < RLIM_INFINITY) {
2103 if (((new - start) + (end_data - start_data)) > rlim)
2110 extern int mm_take_all_locks(struct mm_struct *mm);
2111 extern void mm_drop_all_locks(struct mm_struct *mm);
2113 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2114 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2115 extern struct file *get_task_exe_file(struct task_struct *task);
2117 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2118 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2120 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2121 const struct vm_special_mapping *sm);
2122 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2123 unsigned long addr, unsigned long len,
2124 unsigned long flags,
2125 const struct vm_special_mapping *spec);
2126 /* This is an obsolete alternative to _install_special_mapping. */
2127 extern int install_special_mapping(struct mm_struct *mm,
2128 unsigned long addr, unsigned long len,
2129 unsigned long flags, struct page **pages);
2131 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2133 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2134 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2135 struct list_head *uf);
2136 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2137 unsigned long len, unsigned long prot, unsigned long flags,
2138 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2139 struct list_head *uf);
2140 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2141 struct list_head *uf);
2143 static inline unsigned long
2144 do_mmap_pgoff(struct file *file, unsigned long addr,
2145 unsigned long len, unsigned long prot, unsigned long flags,
2146 unsigned long pgoff, unsigned long *populate,
2147 struct list_head *uf)
2149 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2153 extern int __mm_populate(unsigned long addr, unsigned long len,
2155 static inline void mm_populate(unsigned long addr, unsigned long len)
2158 (void) __mm_populate(addr, len, 1);
2161 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2164 /* These take the mm semaphore themselves */
2165 extern int __must_check vm_brk(unsigned long, unsigned long);
2166 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2167 extern int vm_munmap(unsigned long, size_t);
2168 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2169 unsigned long, unsigned long,
2170 unsigned long, unsigned long);
2172 struct vm_unmapped_area_info {
2173 #define VM_UNMAPPED_AREA_TOPDOWN 1
2174 unsigned long flags;
2175 unsigned long length;
2176 unsigned long low_limit;
2177 unsigned long high_limit;
2178 unsigned long align_mask;
2179 unsigned long align_offset;
2182 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2183 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2186 * Search for an unmapped address range.
2188 * We are looking for a range that:
2189 * - does not intersect with any VMA;
2190 * - is contained within the [low_limit, high_limit) interval;
2191 * - is at least the desired size.
2192 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2194 static inline unsigned long
2195 vm_unmapped_area(struct vm_unmapped_area_info *info)
2197 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2198 return unmapped_area_topdown(info);
2200 return unmapped_area(info);
2204 extern void truncate_inode_pages(struct address_space *, loff_t);
2205 extern void truncate_inode_pages_range(struct address_space *,
2206 loff_t lstart, loff_t lend);
2207 extern void truncate_inode_pages_final(struct address_space *);
2209 /* generic vm_area_ops exported for stackable file systems */
2210 extern int filemap_fault(struct vm_fault *vmf);
2211 extern void filemap_map_pages(struct vm_fault *vmf,
2212 pgoff_t start_pgoff, pgoff_t end_pgoff);
2213 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2215 /* mm/page-writeback.c */
2216 int __must_check write_one_page(struct page *page);
2217 void task_dirty_inc(struct task_struct *tsk);
2220 #define VM_MAX_READAHEAD 128 /* kbytes */
2221 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2223 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2224 pgoff_t offset, unsigned long nr_to_read);
2226 void page_cache_sync_readahead(struct address_space *mapping,
2227 struct file_ra_state *ra,
2230 unsigned long size);
2232 void page_cache_async_readahead(struct address_space *mapping,
2233 struct file_ra_state *ra,
2237 unsigned long size);
2239 extern unsigned long stack_guard_gap;
2240 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2241 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2243 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2244 extern int expand_downwards(struct vm_area_struct *vma,
2245 unsigned long address);
2247 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2249 #define expand_upwards(vma, address) (0)
2252 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2253 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2254 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2255 struct vm_area_struct **pprev);
2257 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2258 NULL if none. Assume start_addr < end_addr. */
2259 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2261 struct vm_area_struct * vma = find_vma(mm,start_addr);
2263 if (vma && end_addr <= vma->vm_start)
2268 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2270 unsigned long vm_start = vma->vm_start;
2272 if (vma->vm_flags & VM_GROWSDOWN) {
2273 vm_start -= stack_guard_gap;
2274 if (vm_start > vma->vm_start)
2280 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2282 unsigned long vm_end = vma->vm_end;
2284 if (vma->vm_flags & VM_GROWSUP) {
2285 vm_end += stack_guard_gap;
2286 if (vm_end < vma->vm_end)
2287 vm_end = -PAGE_SIZE;
2292 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2294 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2297 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2298 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2299 unsigned long vm_start, unsigned long vm_end)
2301 struct vm_area_struct *vma = find_vma(mm, vm_start);
2303 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2310 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2311 void vma_set_page_prot(struct vm_area_struct *vma);
2313 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2317 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2319 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2323 #ifdef CONFIG_NUMA_BALANCING
2324 unsigned long change_prot_numa(struct vm_area_struct *vma,
2325 unsigned long start, unsigned long end);
2328 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2329 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2330 unsigned long pfn, unsigned long size, pgprot_t);
2331 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2332 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2334 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2335 unsigned long pfn, pgprot_t pgprot);
2336 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2338 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2340 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2343 struct page *follow_page_mask(struct vm_area_struct *vma,
2344 unsigned long address, unsigned int foll_flags,
2345 unsigned int *page_mask);
2347 static inline struct page *follow_page(struct vm_area_struct *vma,
2348 unsigned long address, unsigned int foll_flags)
2350 unsigned int unused_page_mask;
2351 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2354 #define FOLL_WRITE 0x01 /* check pte is writable */
2355 #define FOLL_TOUCH 0x02 /* mark page accessed */
2356 #define FOLL_GET 0x04 /* do get_page on page */
2357 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2358 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2359 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2360 * and return without waiting upon it */
2361 #define FOLL_POPULATE 0x40 /* fault in page */
2362 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2363 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2364 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2365 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2366 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2367 #define FOLL_MLOCK 0x1000 /* lock present pages */
2368 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2369 #define FOLL_COW 0x4000 /* internal GUP flag */
2371 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2373 if (vm_fault & VM_FAULT_OOM)
2375 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2376 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2377 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2382 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2384 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2385 unsigned long size, pte_fn_t fn, void *data);
2388 #ifdef CONFIG_PAGE_POISONING
2389 extern bool page_poisoning_enabled(void);
2390 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2391 extern bool page_is_poisoned(struct page *page);
2393 static inline bool page_poisoning_enabled(void) { return false; }
2394 static inline void kernel_poison_pages(struct page *page, int numpages,
2396 static inline bool page_is_poisoned(struct page *page) { return false; }
2399 #ifdef CONFIG_DEBUG_PAGEALLOC
2400 extern bool _debug_pagealloc_enabled;
2401 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2403 static inline bool debug_pagealloc_enabled(void)
2405 return _debug_pagealloc_enabled;
2409 kernel_map_pages(struct page *page, int numpages, int enable)
2411 if (!debug_pagealloc_enabled())
2414 __kernel_map_pages(page, numpages, enable);
2416 #ifdef CONFIG_HIBERNATION
2417 extern bool kernel_page_present(struct page *page);
2418 #endif /* CONFIG_HIBERNATION */
2419 #else /* CONFIG_DEBUG_PAGEALLOC */
2421 kernel_map_pages(struct page *page, int numpages, int enable) {}
2422 #ifdef CONFIG_HIBERNATION
2423 static inline bool kernel_page_present(struct page *page) { return true; }
2424 #endif /* CONFIG_HIBERNATION */
2425 static inline bool debug_pagealloc_enabled(void)
2429 #endif /* CONFIG_DEBUG_PAGEALLOC */
2431 #ifdef __HAVE_ARCH_GATE_AREA
2432 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2433 extern int in_gate_area_no_mm(unsigned long addr);
2434 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2436 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2440 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2441 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2445 #endif /* __HAVE_ARCH_GATE_AREA */
2447 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2449 #ifdef CONFIG_SYSCTL
2450 extern int sysctl_drop_caches;
2451 int drop_caches_sysctl_handler(struct ctl_table *, int,
2452 void __user *, size_t *, loff_t *);
2455 void drop_slab(void);
2456 void drop_slab_node(int nid);
2459 #define randomize_va_space 0
2461 extern int randomize_va_space;
2464 const char * arch_vma_name(struct vm_area_struct *vma);
2465 void print_vma_addr(char *prefix, unsigned long rip);
2467 void sparse_mem_maps_populate_node(struct page **map_map,
2468 unsigned long pnum_begin,
2469 unsigned long pnum_end,
2470 unsigned long map_count,
2473 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2474 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2475 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2476 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2477 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2478 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2479 void *vmemmap_alloc_block(unsigned long size, int node);
2481 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2482 struct vmem_altmap *altmap);
2483 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2485 return __vmemmap_alloc_block_buf(size, node, NULL);
2488 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2489 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2491 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2492 void vmemmap_populate_print_last(void);
2493 #ifdef CONFIG_MEMORY_HOTPLUG
2494 void vmemmap_free(unsigned long start, unsigned long end);
2496 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2497 unsigned long size);
2500 MF_COUNT_INCREASED = 1 << 0,
2501 MF_ACTION_REQUIRED = 1 << 1,
2502 MF_MUST_KILL = 1 << 2,
2503 MF_SOFT_OFFLINE = 1 << 3,
2505 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2506 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2507 extern int unpoison_memory(unsigned long pfn);
2508 extern int get_hwpoison_page(struct page *page);
2509 #define put_hwpoison_page(page) put_page(page)
2510 extern int sysctl_memory_failure_early_kill;
2511 extern int sysctl_memory_failure_recovery;
2512 extern void shake_page(struct page *p, int access);
2513 extern atomic_long_t num_poisoned_pages;
2514 extern int soft_offline_page(struct page *page, int flags);
2518 * Error handlers for various types of pages.
2521 MF_IGNORED, /* Error: cannot be handled */
2522 MF_FAILED, /* Error: handling failed */
2523 MF_DELAYED, /* Will be handled later */
2524 MF_RECOVERED, /* Successfully recovered */
2527 enum mf_action_page_type {
2529 MF_MSG_KERNEL_HIGH_ORDER,
2531 MF_MSG_DIFFERENT_COMPOUND,
2532 MF_MSG_POISONED_HUGE,
2535 MF_MSG_UNMAP_FAILED,
2536 MF_MSG_DIRTY_SWAPCACHE,
2537 MF_MSG_CLEAN_SWAPCACHE,
2538 MF_MSG_DIRTY_MLOCKED_LRU,
2539 MF_MSG_CLEAN_MLOCKED_LRU,
2540 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2541 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2544 MF_MSG_TRUNCATED_LRU,
2550 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2551 extern void clear_huge_page(struct page *page,
2552 unsigned long addr_hint,
2553 unsigned int pages_per_huge_page);
2554 extern void copy_user_huge_page(struct page *dst, struct page *src,
2555 unsigned long addr, struct vm_area_struct *vma,
2556 unsigned int pages_per_huge_page);
2557 extern long copy_huge_page_from_user(struct page *dst_page,
2558 const void __user *usr_src,
2559 unsigned int pages_per_huge_page,
2560 bool allow_pagefault);
2561 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2563 extern struct page_ext_operations debug_guardpage_ops;
2565 #ifdef CONFIG_DEBUG_PAGEALLOC
2566 extern unsigned int _debug_guardpage_minorder;
2567 extern bool _debug_guardpage_enabled;
2569 static inline unsigned int debug_guardpage_minorder(void)
2571 return _debug_guardpage_minorder;
2574 static inline bool debug_guardpage_enabled(void)
2576 return _debug_guardpage_enabled;
2579 static inline bool page_is_guard(struct page *page)
2581 struct page_ext *page_ext;
2583 if (!debug_guardpage_enabled())
2586 page_ext = lookup_page_ext(page);
2587 if (unlikely(!page_ext))
2590 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2593 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2594 static inline bool debug_guardpage_enabled(void) { return false; }
2595 static inline bool page_is_guard(struct page *page) { return false; }
2596 #endif /* CONFIG_DEBUG_PAGEALLOC */
2598 #if MAX_NUMNODES > 1
2599 void __init setup_nr_node_ids(void);
2601 static inline void setup_nr_node_ids(void) {}
2604 #endif /* __KERNEL__ */
2605 #endif /* _LINUX_MM_H */