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>
29 struct anon_vma_chain;
32 struct writeback_control;
35 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
36 extern unsigned long max_mapnr;
38 static inline void set_max_mapnr(unsigned long limit)
43 static inline void set_max_mapnr(unsigned long limit) { }
46 extern unsigned long totalram_pages;
47 extern void * high_memory;
48 extern int page_cluster;
51 extern int sysctl_legacy_va_layout;
53 #define sysctl_legacy_va_layout 0
56 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
57 extern const int mmap_rnd_bits_min;
58 extern const int mmap_rnd_bits_max;
59 extern int mmap_rnd_bits __read_mostly;
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
62 extern const int mmap_rnd_compat_bits_min;
63 extern const int mmap_rnd_compat_bits_max;
64 extern int mmap_rnd_compat_bits __read_mostly;
68 #include <asm/pgtable.h>
69 #include <asm/processor.h>
72 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
76 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
80 #define lm_alias(x) __va(__pa_symbol(x))
84 * To prevent common memory management code establishing
85 * a zero page mapping on a read fault.
86 * This macro should be defined within <asm/pgtable.h>.
87 * s390 does this to prevent multiplexing of hardware bits
88 * related to the physical page in case of virtualization.
90 #ifndef mm_forbids_zeropage
91 #define mm_forbids_zeropage(X) (0)
95 * Default maximum number of active map areas, this limits the number of vmas
96 * per mm struct. Users can overwrite this number by sysctl but there is a
99 * When a program's coredump is generated as ELF format, a section is created
100 * per a vma. In ELF, the number of sections is represented in unsigned short.
101 * This means the number of sections should be smaller than 65535 at coredump.
102 * Because the kernel adds some informative sections to a image of program at
103 * generating coredump, we need some margin. The number of extra sections is
104 * 1-3 now and depends on arch. We use "5" as safe margin, here.
106 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
107 * not a hard limit any more. Although some userspace tools can be surprised by
110 #define MAPCOUNT_ELF_CORE_MARGIN (5)
111 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
113 extern int sysctl_max_map_count;
115 extern unsigned long sysctl_user_reserve_kbytes;
116 extern unsigned long sysctl_admin_reserve_kbytes;
118 extern int sysctl_overcommit_memory;
119 extern int sysctl_overcommit_ratio;
120 extern unsigned long sysctl_overcommit_kbytes;
122 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
124 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
127 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
129 /* to align the pointer to the (next) page boundary */
130 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
132 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
133 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
136 * Linux kernel virtual memory manager primitives.
137 * The idea being to have a "virtual" mm in the same way
138 * we have a virtual fs - giving a cleaner interface to the
139 * mm details, and allowing different kinds of memory mappings
140 * (from shared memory to executable loading to arbitrary
144 extern struct kmem_cache *vm_area_cachep;
147 extern struct rb_root nommu_region_tree;
148 extern struct rw_semaphore nommu_region_sem;
150 extern unsigned int kobjsize(const void *objp);
154 * vm_flags in vm_area_struct, see mm_types.h.
155 * When changing, update also include/trace/events/mmflags.h
157 #define VM_NONE 0x00000000
159 #define VM_READ 0x00000001 /* currently active flags */
160 #define VM_WRITE 0x00000002
161 #define VM_EXEC 0x00000004
162 #define VM_SHARED 0x00000008
164 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
165 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
166 #define VM_MAYWRITE 0x00000020
167 #define VM_MAYEXEC 0x00000040
168 #define VM_MAYSHARE 0x00000080
170 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
171 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
172 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
173 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
174 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
176 #define VM_LOCKED 0x00002000
177 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
179 /* Used by sys_madvise() */
180 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
181 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
183 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
184 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
185 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
186 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
187 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
188 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
189 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
190 #define VM_ARCH_2 0x02000000
191 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
193 #ifdef CONFIG_MEM_SOFT_DIRTY
194 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
196 # define VM_SOFTDIRTY 0
199 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
200 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
201 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
202 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
204 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
205 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
206 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
207 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
208 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
209 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
210 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
211 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
212 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
213 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
215 #if defined(CONFIG_X86)
216 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
217 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
218 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
219 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
220 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
221 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
222 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
224 #elif defined(CONFIG_PPC)
225 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
226 #elif defined(CONFIG_PARISC)
227 # define VM_GROWSUP VM_ARCH_1
228 #elif defined(CONFIG_METAG)
229 # define VM_GROWSUP VM_ARCH_1
230 #elif defined(CONFIG_IA64)
231 # define VM_GROWSUP VM_ARCH_1
232 #elif !defined(CONFIG_MMU)
233 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
236 #if defined(CONFIG_X86)
237 /* MPX specific bounds table or bounds directory */
238 # define VM_MPX VM_ARCH_2
242 # define VM_GROWSUP VM_NONE
245 /* Bits set in the VMA until the stack is in its final location */
246 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
248 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
249 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
252 #ifdef CONFIG_STACK_GROWSUP
253 #define VM_STACK VM_GROWSUP
255 #define VM_STACK VM_GROWSDOWN
258 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
261 * Special vmas that are non-mergable, non-mlock()able.
262 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
264 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
266 /* This mask defines which mm->def_flags a process can inherit its parent */
267 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
269 /* This mask is used to clear all the VMA flags used by mlock */
270 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
273 * mapping from the currently active vm_flags protection bits (the
274 * low four bits) to a page protection mask..
276 extern pgprot_t protection_map[16];
278 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
279 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
280 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
281 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
282 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
283 #define FAULT_FLAG_TRIED 0x20 /* Second try */
284 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
285 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
286 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
288 #define FAULT_FLAG_TRACE \
289 { FAULT_FLAG_WRITE, "WRITE" }, \
290 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
291 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
292 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
293 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
294 { FAULT_FLAG_TRIED, "TRIED" }, \
295 { FAULT_FLAG_USER, "USER" }, \
296 { FAULT_FLAG_REMOTE, "REMOTE" }, \
297 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
300 * vm_fault is filled by the the pagefault handler and passed to the vma's
301 * ->fault function. The vma's ->fault is responsible for returning a bitmask
302 * of VM_FAULT_xxx flags that give details about how the fault was handled.
304 * MM layer fills up gfp_mask for page allocations but fault handler might
305 * alter it if its implementation requires a different allocation context.
307 * pgoff should be used in favour of virtual_address, if possible.
310 struct vm_area_struct *vma; /* Target VMA */
311 unsigned int flags; /* FAULT_FLAG_xxx flags */
312 gfp_t gfp_mask; /* gfp mask to be used for allocations */
313 pgoff_t pgoff; /* Logical page offset based on vma */
314 unsigned long address; /* Faulting virtual address */
315 pmd_t *pmd; /* Pointer to pmd entry matching
317 pte_t orig_pte; /* Value of PTE at the time of fault */
319 struct page *cow_page; /* Page handler may use for COW fault */
320 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
321 struct page *page; /* ->fault handlers should return a
322 * page here, unless VM_FAULT_NOPAGE
323 * is set (which is also implied by
326 /* These three entries are valid only while holding ptl lock */
327 pte_t *pte; /* Pointer to pte entry matching
328 * the 'address'. NULL if the page
329 * table hasn't been allocated.
331 spinlock_t *ptl; /* Page table lock.
332 * Protects pte page table if 'pte'
333 * is not NULL, otherwise pmd.
335 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
336 * vm_ops->map_pages() calls
337 * alloc_set_pte() from atomic context.
338 * do_fault_around() pre-allocates
339 * page table to avoid allocation from
345 * These are the virtual MM functions - opening of an area, closing and
346 * unmapping it (needed to keep files on disk up-to-date etc), pointer
347 * to the functions called when a no-page or a wp-page exception occurs.
349 struct vm_operations_struct {
350 void (*open)(struct vm_area_struct * area);
351 void (*close)(struct vm_area_struct * area);
352 int (*mremap)(struct vm_area_struct * area);
353 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
354 int (*pmd_fault)(struct vm_fault *vmf);
355 void (*map_pages)(struct vm_fault *vmf,
356 pgoff_t start_pgoff, pgoff_t end_pgoff);
358 /* notification that a previously read-only page is about to become
359 * writable, if an error is returned it will cause a SIGBUS */
360 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
362 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
363 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
365 /* called by access_process_vm when get_user_pages() fails, typically
366 * for use by special VMAs that can switch between memory and hardware
368 int (*access)(struct vm_area_struct *vma, unsigned long addr,
369 void *buf, int len, int write);
371 /* Called by the /proc/PID/maps code to ask the vma whether it
372 * has a special name. Returning non-NULL will also cause this
373 * vma to be dumped unconditionally. */
374 const char *(*name)(struct vm_area_struct *vma);
378 * set_policy() op must add a reference to any non-NULL @new mempolicy
379 * to hold the policy upon return. Caller should pass NULL @new to
380 * remove a policy and fall back to surrounding context--i.e. do not
381 * install a MPOL_DEFAULT policy, nor the task or system default
384 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
387 * get_policy() op must add reference [mpol_get()] to any policy at
388 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
389 * in mm/mempolicy.c will do this automatically.
390 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
391 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
392 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
393 * must return NULL--i.e., do not "fallback" to task or system default
396 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
400 * Called by vm_normal_page() for special PTEs to find the
401 * page for @addr. This is useful if the default behavior
402 * (using pte_page()) would not find the correct page.
404 struct page *(*find_special_page)(struct vm_area_struct *vma,
411 #define page_private(page) ((page)->private)
412 #define set_page_private(page, v) ((page)->private = (v))
414 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
415 static inline int pmd_devmap(pmd_t pmd)
422 * FIXME: take this include out, include page-flags.h in
423 * files which need it (119 of them)
425 #include <linux/page-flags.h>
426 #include <linux/huge_mm.h>
429 * Methods to modify the page usage count.
431 * What counts for a page usage:
432 * - cache mapping (page->mapping)
433 * - private data (page->private)
434 * - page mapped in a task's page tables, each mapping
435 * is counted separately
437 * Also, many kernel routines increase the page count before a critical
438 * routine so they can be sure the page doesn't go away from under them.
442 * Drop a ref, return true if the refcount fell to zero (the page has no users)
444 static inline int put_page_testzero(struct page *page)
446 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
447 return page_ref_dec_and_test(page);
451 * Try to grab a ref unless the page has a refcount of zero, return false if
453 * This can be called when MMU is off so it must not access
454 * any of the virtual mappings.
456 static inline int get_page_unless_zero(struct page *page)
458 return page_ref_add_unless(page, 1, 0);
461 extern int page_is_ram(unsigned long pfn);
469 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
472 /* Support for virtually mapped pages */
473 struct page *vmalloc_to_page(const void *addr);
474 unsigned long vmalloc_to_pfn(const void *addr);
477 * Determine if an address is within the vmalloc range
479 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
480 * is no special casing required.
482 static inline bool is_vmalloc_addr(const void *x)
485 unsigned long addr = (unsigned long)x;
487 return addr >= VMALLOC_START && addr < VMALLOC_END;
493 extern int is_vmalloc_or_module_addr(const void *x);
495 static inline int is_vmalloc_or_module_addr(const void *x)
501 extern void kvfree(const void *addr);
503 static inline atomic_t *compound_mapcount_ptr(struct page *page)
505 return &page[1].compound_mapcount;
508 static inline int compound_mapcount(struct page *page)
510 VM_BUG_ON_PAGE(!PageCompound(page), page);
511 page = compound_head(page);
512 return atomic_read(compound_mapcount_ptr(page)) + 1;
516 * The atomic page->_mapcount, starts from -1: so that transitions
517 * both from it and to it can be tracked, using atomic_inc_and_test
518 * and atomic_add_negative(-1).
520 static inline void page_mapcount_reset(struct page *page)
522 atomic_set(&(page)->_mapcount, -1);
525 int __page_mapcount(struct page *page);
527 static inline int page_mapcount(struct page *page)
529 VM_BUG_ON_PAGE(PageSlab(page), page);
531 if (unlikely(PageCompound(page)))
532 return __page_mapcount(page);
533 return atomic_read(&page->_mapcount) + 1;
536 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
537 int total_mapcount(struct page *page);
538 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
540 static inline int total_mapcount(struct page *page)
542 return page_mapcount(page);
544 static inline int page_trans_huge_mapcount(struct page *page,
547 int mapcount = page_mapcount(page);
549 *total_mapcount = mapcount;
554 static inline struct page *virt_to_head_page(const void *x)
556 struct page *page = virt_to_page(x);
558 return compound_head(page);
561 void __put_page(struct page *page);
563 void put_pages_list(struct list_head *pages);
565 void split_page(struct page *page, unsigned int order);
568 * Compound pages have a destructor function. Provide a
569 * prototype for that function and accessor functions.
570 * These are _only_ valid on the head of a compound page.
572 typedef void compound_page_dtor(struct page *);
574 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
575 enum compound_dtor_id {
578 #ifdef CONFIG_HUGETLB_PAGE
581 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
586 extern compound_page_dtor * const compound_page_dtors[];
588 static inline void set_compound_page_dtor(struct page *page,
589 enum compound_dtor_id compound_dtor)
591 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
592 page[1].compound_dtor = compound_dtor;
595 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
597 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
598 return compound_page_dtors[page[1].compound_dtor];
601 static inline unsigned int compound_order(struct page *page)
605 return page[1].compound_order;
608 static inline void set_compound_order(struct page *page, unsigned int order)
610 page[1].compound_order = order;
613 void free_compound_page(struct page *page);
617 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
618 * servicing faults for write access. In the normal case, do always want
619 * pte_mkwrite. But get_user_pages can cause write faults for mappings
620 * that do not have writing enabled, when used by access_process_vm.
622 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
624 if (likely(vma->vm_flags & VM_WRITE))
625 pte = pte_mkwrite(pte);
629 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
631 int finish_fault(struct vm_fault *vmf);
632 int finish_mkwrite_fault(struct vm_fault *vmf);
636 * Multiple processes may "see" the same page. E.g. for untouched
637 * mappings of /dev/null, all processes see the same page full of
638 * zeroes, and text pages of executables and shared libraries have
639 * only one copy in memory, at most, normally.
641 * For the non-reserved pages, page_count(page) denotes a reference count.
642 * page_count() == 0 means the page is free. page->lru is then used for
643 * freelist management in the buddy allocator.
644 * page_count() > 0 means the page has been allocated.
646 * Pages are allocated by the slab allocator in order to provide memory
647 * to kmalloc and kmem_cache_alloc. In this case, the management of the
648 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
649 * unless a particular usage is carefully commented. (the responsibility of
650 * freeing the kmalloc memory is the caller's, of course).
652 * A page may be used by anyone else who does a __get_free_page().
653 * In this case, page_count still tracks the references, and should only
654 * be used through the normal accessor functions. The top bits of page->flags
655 * and page->virtual store page management information, but all other fields
656 * are unused and could be used privately, carefully. The management of this
657 * page is the responsibility of the one who allocated it, and those who have
658 * subsequently been given references to it.
660 * The other pages (we may call them "pagecache pages") are completely
661 * managed by the Linux memory manager: I/O, buffers, swapping etc.
662 * The following discussion applies only to them.
664 * A pagecache page contains an opaque `private' member, which belongs to the
665 * page's address_space. Usually, this is the address of a circular list of
666 * the page's disk buffers. PG_private must be set to tell the VM to call
667 * into the filesystem to release these pages.
669 * A page may belong to an inode's memory mapping. In this case, page->mapping
670 * is the pointer to the inode, and page->index is the file offset of the page,
671 * in units of PAGE_SIZE.
673 * If pagecache pages are not associated with an inode, they are said to be
674 * anonymous pages. These may become associated with the swapcache, and in that
675 * case PG_swapcache is set, and page->private is an offset into the swapcache.
677 * In either case (swapcache or inode backed), the pagecache itself holds one
678 * reference to the page. Setting PG_private should also increment the
679 * refcount. The each user mapping also has a reference to the page.
681 * The pagecache pages are stored in a per-mapping radix tree, which is
682 * rooted at mapping->page_tree, and indexed by offset.
683 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
684 * lists, we instead now tag pages as dirty/writeback in the radix tree.
686 * All pagecache pages may be subject to I/O:
687 * - inode pages may need to be read from disk,
688 * - inode pages which have been modified and are MAP_SHARED may need
689 * to be written back to the inode on disk,
690 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
691 * modified may need to be swapped out to swap space and (later) to be read
696 * The zone field is never updated after free_area_init_core()
697 * sets it, so none of the operations on it need to be atomic.
700 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
701 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
702 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
703 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
704 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
707 * Define the bit shifts to access each section. For non-existent
708 * sections we define the shift as 0; that plus a 0 mask ensures
709 * the compiler will optimise away reference to them.
711 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
712 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
713 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
714 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
716 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
717 #ifdef NODE_NOT_IN_PAGE_FLAGS
718 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
719 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
720 SECTIONS_PGOFF : ZONES_PGOFF)
722 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
723 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
724 NODES_PGOFF : ZONES_PGOFF)
727 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
729 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
730 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
733 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
734 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
735 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
736 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
737 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
739 static inline enum zone_type page_zonenum(const struct page *page)
741 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
744 #ifdef CONFIG_ZONE_DEVICE
745 void get_zone_device_page(struct page *page);
746 void put_zone_device_page(struct page *page);
747 static inline bool is_zone_device_page(const struct page *page)
749 return page_zonenum(page) == ZONE_DEVICE;
752 static inline void get_zone_device_page(struct page *page)
755 static inline void put_zone_device_page(struct page *page)
758 static inline bool is_zone_device_page(const struct page *page)
764 static inline void get_page(struct page *page)
766 page = compound_head(page);
768 * Getting a normal page or the head of a compound page
769 * requires to already have an elevated page->_refcount.
771 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
774 if (unlikely(is_zone_device_page(page)))
775 get_zone_device_page(page);
778 static inline void put_page(struct page *page)
780 page = compound_head(page);
782 if (put_page_testzero(page))
785 if (unlikely(is_zone_device_page(page)))
786 put_zone_device_page(page);
789 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
790 #define SECTION_IN_PAGE_FLAGS
794 * The identification function is mainly used by the buddy allocator for
795 * determining if two pages could be buddies. We are not really identifying
796 * the zone since we could be using the section number id if we do not have
797 * node id available in page flags.
798 * We only guarantee that it will return the same value for two combinable
801 static inline int page_zone_id(struct page *page)
803 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
806 static inline int zone_to_nid(struct zone *zone)
815 #ifdef NODE_NOT_IN_PAGE_FLAGS
816 extern int page_to_nid(const struct page *page);
818 static inline int page_to_nid(const struct page *page)
820 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
824 #ifdef CONFIG_NUMA_BALANCING
825 static inline int cpu_pid_to_cpupid(int cpu, int pid)
827 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
830 static inline int cpupid_to_pid(int cpupid)
832 return cpupid & LAST__PID_MASK;
835 static inline int cpupid_to_cpu(int cpupid)
837 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
840 static inline int cpupid_to_nid(int cpupid)
842 return cpu_to_node(cpupid_to_cpu(cpupid));
845 static inline bool cpupid_pid_unset(int cpupid)
847 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
850 static inline bool cpupid_cpu_unset(int cpupid)
852 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
855 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
857 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
860 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
861 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
862 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
864 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
867 static inline int page_cpupid_last(struct page *page)
869 return page->_last_cpupid;
871 static inline void page_cpupid_reset_last(struct page *page)
873 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
876 static inline int page_cpupid_last(struct page *page)
878 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
881 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
883 static inline void page_cpupid_reset_last(struct page *page)
885 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
887 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
888 #else /* !CONFIG_NUMA_BALANCING */
889 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
891 return page_to_nid(page); /* XXX */
894 static inline int page_cpupid_last(struct page *page)
896 return page_to_nid(page); /* XXX */
899 static inline int cpupid_to_nid(int cpupid)
904 static inline int cpupid_to_pid(int cpupid)
909 static inline int cpupid_to_cpu(int cpupid)
914 static inline int cpu_pid_to_cpupid(int nid, int pid)
919 static inline bool cpupid_pid_unset(int cpupid)
924 static inline void page_cpupid_reset_last(struct page *page)
928 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
932 #endif /* CONFIG_NUMA_BALANCING */
934 static inline struct zone *page_zone(const struct page *page)
936 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
939 static inline pg_data_t *page_pgdat(const struct page *page)
941 return NODE_DATA(page_to_nid(page));
944 #ifdef SECTION_IN_PAGE_FLAGS
945 static inline void set_page_section(struct page *page, unsigned long section)
947 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
948 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
951 static inline unsigned long page_to_section(const struct page *page)
953 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
957 static inline void set_page_zone(struct page *page, enum zone_type zone)
959 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
960 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
963 static inline void set_page_node(struct page *page, unsigned long node)
965 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
966 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
969 static inline void set_page_links(struct page *page, enum zone_type zone,
970 unsigned long node, unsigned long pfn)
972 set_page_zone(page, zone);
973 set_page_node(page, node);
974 #ifdef SECTION_IN_PAGE_FLAGS
975 set_page_section(page, pfn_to_section_nr(pfn));
980 static inline struct mem_cgroup *page_memcg(struct page *page)
982 return page->mem_cgroup;
984 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
986 WARN_ON_ONCE(!rcu_read_lock_held());
987 return READ_ONCE(page->mem_cgroup);
990 static inline struct mem_cgroup *page_memcg(struct page *page)
994 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
996 WARN_ON_ONCE(!rcu_read_lock_held());
1002 * Some inline functions in vmstat.h depend on page_zone()
1004 #include <linux/vmstat.h>
1006 static __always_inline void *lowmem_page_address(const struct page *page)
1008 return page_to_virt(page);
1011 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1012 #define HASHED_PAGE_VIRTUAL
1015 #if defined(WANT_PAGE_VIRTUAL)
1016 static inline void *page_address(const struct page *page)
1018 return page->virtual;
1020 static inline void set_page_address(struct page *page, void *address)
1022 page->virtual = address;
1024 #define page_address_init() do { } while(0)
1027 #if defined(HASHED_PAGE_VIRTUAL)
1028 void *page_address(const struct page *page);
1029 void set_page_address(struct page *page, void *virtual);
1030 void page_address_init(void);
1033 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1034 #define page_address(page) lowmem_page_address(page)
1035 #define set_page_address(page, address) do { } while(0)
1036 #define page_address_init() do { } while(0)
1039 extern void *page_rmapping(struct page *page);
1040 extern struct anon_vma *page_anon_vma(struct page *page);
1041 extern struct address_space *page_mapping(struct page *page);
1043 extern struct address_space *__page_file_mapping(struct page *);
1046 struct address_space *page_file_mapping(struct page *page)
1048 if (unlikely(PageSwapCache(page)))
1049 return __page_file_mapping(page);
1051 return page->mapping;
1054 extern pgoff_t __page_file_index(struct page *page);
1057 * Return the pagecache index of the passed page. Regular pagecache pages
1058 * use ->index whereas swapcache pages use swp_offset(->private)
1060 static inline pgoff_t page_index(struct page *page)
1062 if (unlikely(PageSwapCache(page)))
1063 return __page_file_index(page);
1067 bool page_mapped(struct page *page);
1068 struct address_space *page_mapping(struct page *page);
1071 * Return true only if the page has been allocated with
1072 * ALLOC_NO_WATERMARKS and the low watermark was not
1073 * met implying that the system is under some pressure.
1075 static inline bool page_is_pfmemalloc(struct page *page)
1078 * Page index cannot be this large so this must be
1079 * a pfmemalloc page.
1081 return page->index == -1UL;
1085 * Only to be called by the page allocator on a freshly allocated
1088 static inline void set_page_pfmemalloc(struct page *page)
1093 static inline void clear_page_pfmemalloc(struct page *page)
1099 * Different kinds of faults, as returned by handle_mm_fault().
1100 * Used to decide whether a process gets delivered SIGBUS or
1101 * just gets major/minor fault counters bumped up.
1104 #define VM_FAULT_OOM 0x0001
1105 #define VM_FAULT_SIGBUS 0x0002
1106 #define VM_FAULT_MAJOR 0x0004
1107 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1108 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1109 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1110 #define VM_FAULT_SIGSEGV 0x0040
1112 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1113 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1114 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1115 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1116 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1118 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1120 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1121 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1124 #define VM_FAULT_RESULT_TRACE \
1125 { VM_FAULT_OOM, "OOM" }, \
1126 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1127 { VM_FAULT_MAJOR, "MAJOR" }, \
1128 { VM_FAULT_WRITE, "WRITE" }, \
1129 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1130 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1131 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1132 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1133 { VM_FAULT_LOCKED, "LOCKED" }, \
1134 { VM_FAULT_RETRY, "RETRY" }, \
1135 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1136 { VM_FAULT_DONE_COW, "DONE_COW" }
1138 /* Encode hstate index for a hwpoisoned large page */
1139 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1140 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1143 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1145 extern void pagefault_out_of_memory(void);
1147 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1150 * Flags passed to show_mem() and show_free_areas() to suppress output in
1153 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1155 extern void show_free_areas(unsigned int flags);
1156 extern bool skip_free_areas_node(unsigned int flags, int nid);
1158 int shmem_zero_setup(struct vm_area_struct *);
1160 bool shmem_mapping(struct address_space *mapping);
1162 static inline bool shmem_mapping(struct address_space *mapping)
1168 extern bool can_do_mlock(void);
1169 extern int user_shm_lock(size_t, struct user_struct *);
1170 extern void user_shm_unlock(size_t, struct user_struct *);
1173 * Parameter block passed down to zap_pte_range in exceptional cases.
1175 struct zap_details {
1176 struct address_space *check_mapping; /* Check page->mapping if set */
1177 pgoff_t first_index; /* Lowest page->index to unmap */
1178 pgoff_t last_index; /* Highest page->index to unmap */
1179 bool ignore_dirty; /* Ignore dirty pages */
1180 bool check_swap_entries; /* Check also swap entries */
1183 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1185 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1188 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1189 unsigned long size);
1190 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1191 unsigned long size, struct zap_details *);
1192 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1193 unsigned long start, unsigned long end);
1196 * mm_walk - callbacks for walk_page_range
1197 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1198 * this handler is required to be able to handle
1199 * pmd_trans_huge() pmds. They may simply choose to
1200 * split_huge_page() instead of handling it explicitly.
1201 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1202 * @pte_hole: if set, called for each hole at all levels
1203 * @hugetlb_entry: if set, called for each hugetlb entry
1204 * @test_walk: caller specific callback function to determine whether
1205 * we walk over the current vma or not. Returning 0
1206 * value means "do page table walk over the current vma,"
1207 * and a negative one means "abort current page table walk
1208 * right now." 1 means "skip the current vma."
1209 * @mm: mm_struct representing the target process of page table walk
1210 * @vma: vma currently walked (NULL if walking outside vmas)
1211 * @private: private data for callbacks' usage
1213 * (see the comment on walk_page_range() for more details)
1216 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1217 unsigned long next, struct mm_walk *walk);
1218 int (*pte_entry)(pte_t *pte, unsigned long addr,
1219 unsigned long next, struct mm_walk *walk);
1220 int (*pte_hole)(unsigned long addr, unsigned long next,
1221 struct mm_walk *walk);
1222 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1223 unsigned long addr, unsigned long next,
1224 struct mm_walk *walk);
1225 int (*test_walk)(unsigned long addr, unsigned long next,
1226 struct mm_walk *walk);
1227 struct mm_struct *mm;
1228 struct vm_area_struct *vma;
1232 int walk_page_range(unsigned long addr, unsigned long end,
1233 struct mm_walk *walk);
1234 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1235 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1236 unsigned long end, unsigned long floor, unsigned long ceiling);
1237 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1238 struct vm_area_struct *vma);
1239 void unmap_mapping_range(struct address_space *mapping,
1240 loff_t const holebegin, loff_t const holelen, int even_cows);
1241 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1242 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1243 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1244 unsigned long *pfn);
1245 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1246 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1247 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1248 void *buf, int len, int write);
1250 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1251 loff_t const holebegin, loff_t const holelen)
1253 unmap_mapping_range(mapping, holebegin, holelen, 0);
1256 extern void truncate_pagecache(struct inode *inode, loff_t new);
1257 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1258 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1259 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1260 int truncate_inode_page(struct address_space *mapping, struct page *page);
1261 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1262 int invalidate_inode_page(struct page *page);
1265 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1266 unsigned int flags);
1267 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1268 unsigned long address, unsigned int fault_flags,
1271 static inline int handle_mm_fault(struct vm_area_struct *vma,
1272 unsigned long address, unsigned int flags)
1274 /* should never happen if there's no MMU */
1276 return VM_FAULT_SIGBUS;
1278 static inline int fixup_user_fault(struct task_struct *tsk,
1279 struct mm_struct *mm, unsigned long address,
1280 unsigned int fault_flags, bool *unlocked)
1282 /* should never happen if there's no MMU */
1288 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1289 unsigned int gup_flags);
1290 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1291 void *buf, int len, unsigned int gup_flags);
1292 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1293 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1295 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1296 unsigned long start, unsigned long nr_pages,
1297 unsigned int gup_flags, struct page **pages,
1298 struct vm_area_struct **vmas, int *locked);
1299 long get_user_pages(unsigned long start, unsigned long nr_pages,
1300 unsigned int gup_flags, struct page **pages,
1301 struct vm_area_struct **vmas);
1302 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1303 unsigned int gup_flags, struct page **pages, int *locked);
1304 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1305 struct page **pages, unsigned int gup_flags);
1306 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1307 struct page **pages);
1309 /* Container for pinned pfns / pages */
1310 struct frame_vector {
1311 unsigned int nr_allocated; /* Number of frames we have space for */
1312 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1313 bool got_ref; /* Did we pin pages by getting page ref? */
1314 bool is_pfns; /* Does array contain pages or pfns? */
1315 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1316 * pfns_vector_pages() or pfns_vector_pfns()
1320 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1321 void frame_vector_destroy(struct frame_vector *vec);
1322 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1323 unsigned int gup_flags, struct frame_vector *vec);
1324 void put_vaddr_frames(struct frame_vector *vec);
1325 int frame_vector_to_pages(struct frame_vector *vec);
1326 void frame_vector_to_pfns(struct frame_vector *vec);
1328 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1330 return vec->nr_frames;
1333 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1336 int err = frame_vector_to_pages(vec);
1339 return ERR_PTR(err);
1341 return (struct page **)(vec->ptrs);
1344 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1347 frame_vector_to_pfns(vec);
1348 return (unsigned long *)(vec->ptrs);
1352 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1353 struct page **pages);
1354 int get_kernel_page(unsigned long start, int write, struct page **pages);
1355 struct page *get_dump_page(unsigned long addr);
1357 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1358 extern void do_invalidatepage(struct page *page, unsigned int offset,
1359 unsigned int length);
1361 int __set_page_dirty_nobuffers(struct page *page);
1362 int __set_page_dirty_no_writeback(struct page *page);
1363 int redirty_page_for_writepage(struct writeback_control *wbc,
1365 void account_page_dirtied(struct page *page, struct address_space *mapping);
1366 void account_page_cleaned(struct page *page, struct address_space *mapping,
1367 struct bdi_writeback *wb);
1368 int set_page_dirty(struct page *page);
1369 int set_page_dirty_lock(struct page *page);
1370 void cancel_dirty_page(struct page *page);
1371 int clear_page_dirty_for_io(struct page *page);
1373 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1375 /* Is the vma a continuation of the stack vma above it? */
1376 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1378 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1381 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1383 return !vma->vm_ops;
1388 * The vma_is_shmem is not inline because it is used only by slow
1389 * paths in userfault.
1391 bool vma_is_shmem(struct vm_area_struct *vma);
1393 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1396 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1399 return (vma->vm_flags & VM_GROWSDOWN) &&
1400 (vma->vm_start == addr) &&
1401 !vma_growsdown(vma->vm_prev, addr);
1404 /* Is the vma a continuation of the stack vma below it? */
1405 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1407 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1410 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1413 return (vma->vm_flags & VM_GROWSUP) &&
1414 (vma->vm_end == addr) &&
1415 !vma_growsup(vma->vm_next, addr);
1418 int vma_is_stack_for_current(struct vm_area_struct *vma);
1420 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1421 unsigned long old_addr, struct vm_area_struct *new_vma,
1422 unsigned long new_addr, unsigned long len,
1423 bool need_rmap_locks);
1424 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1425 unsigned long end, pgprot_t newprot,
1426 int dirty_accountable, int prot_numa);
1427 extern int mprotect_fixup(struct vm_area_struct *vma,
1428 struct vm_area_struct **pprev, unsigned long start,
1429 unsigned long end, unsigned long newflags);
1432 * doesn't attempt to fault and will return short.
1434 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1435 struct page **pages);
1437 * per-process(per-mm_struct) statistics.
1439 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1441 long val = atomic_long_read(&mm->rss_stat.count[member]);
1443 #ifdef SPLIT_RSS_COUNTING
1445 * counter is updated in asynchronous manner and may go to minus.
1446 * But it's never be expected number for users.
1451 return (unsigned long)val;
1454 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1456 atomic_long_add(value, &mm->rss_stat.count[member]);
1459 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1461 atomic_long_inc(&mm->rss_stat.count[member]);
1464 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1466 atomic_long_dec(&mm->rss_stat.count[member]);
1469 /* Optimized variant when page is already known not to be PageAnon */
1470 static inline int mm_counter_file(struct page *page)
1472 if (PageSwapBacked(page))
1473 return MM_SHMEMPAGES;
1474 return MM_FILEPAGES;
1477 static inline int mm_counter(struct page *page)
1480 return MM_ANONPAGES;
1481 return mm_counter_file(page);
1484 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1486 return get_mm_counter(mm, MM_FILEPAGES) +
1487 get_mm_counter(mm, MM_ANONPAGES) +
1488 get_mm_counter(mm, MM_SHMEMPAGES);
1491 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1493 return max(mm->hiwater_rss, get_mm_rss(mm));
1496 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1498 return max(mm->hiwater_vm, mm->total_vm);
1501 static inline void update_hiwater_rss(struct mm_struct *mm)
1503 unsigned long _rss = get_mm_rss(mm);
1505 if ((mm)->hiwater_rss < _rss)
1506 (mm)->hiwater_rss = _rss;
1509 static inline void update_hiwater_vm(struct mm_struct *mm)
1511 if (mm->hiwater_vm < mm->total_vm)
1512 mm->hiwater_vm = mm->total_vm;
1515 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1517 mm->hiwater_rss = get_mm_rss(mm);
1520 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1521 struct mm_struct *mm)
1523 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1525 if (*maxrss < hiwater_rss)
1526 *maxrss = hiwater_rss;
1529 #if defined(SPLIT_RSS_COUNTING)
1530 void sync_mm_rss(struct mm_struct *mm);
1532 static inline void sync_mm_rss(struct mm_struct *mm)
1537 #ifndef __HAVE_ARCH_PTE_DEVMAP
1538 static inline int pte_devmap(pte_t pte)
1544 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1546 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1548 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1552 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1556 #ifdef __PAGETABLE_PUD_FOLDED
1557 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1558 unsigned long address)
1563 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1566 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1567 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1568 unsigned long address)
1573 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1575 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1580 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1581 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1584 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1586 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1588 atomic_long_set(&mm->nr_pmds, 0);
1591 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1593 return atomic_long_read(&mm->nr_pmds);
1596 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1598 atomic_long_inc(&mm->nr_pmds);
1601 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1603 atomic_long_dec(&mm->nr_pmds);
1607 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1608 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1611 * The following ifdef needed to get the 4level-fixup.h header to work.
1612 * Remove it when 4level-fixup.h has been removed.
1614 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1615 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1617 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1618 NULL: pud_offset(pgd, address);
1621 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1623 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1624 NULL: pmd_offset(pud, address);
1626 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1628 #if USE_SPLIT_PTE_PTLOCKS
1629 #if ALLOC_SPLIT_PTLOCKS
1630 void __init ptlock_cache_init(void);
1631 extern bool ptlock_alloc(struct page *page);
1632 extern void ptlock_free(struct page *page);
1634 static inline spinlock_t *ptlock_ptr(struct page *page)
1638 #else /* ALLOC_SPLIT_PTLOCKS */
1639 static inline void ptlock_cache_init(void)
1643 static inline bool ptlock_alloc(struct page *page)
1648 static inline void ptlock_free(struct page *page)
1652 static inline spinlock_t *ptlock_ptr(struct page *page)
1656 #endif /* ALLOC_SPLIT_PTLOCKS */
1658 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1660 return ptlock_ptr(pmd_page(*pmd));
1663 static inline bool ptlock_init(struct page *page)
1666 * prep_new_page() initialize page->private (and therefore page->ptl)
1667 * with 0. Make sure nobody took it in use in between.
1669 * It can happen if arch try to use slab for page table allocation:
1670 * slab code uses page->slab_cache, which share storage with page->ptl.
1672 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1673 if (!ptlock_alloc(page))
1675 spin_lock_init(ptlock_ptr(page));
1679 /* Reset page->mapping so free_pages_check won't complain. */
1680 static inline void pte_lock_deinit(struct page *page)
1682 page->mapping = NULL;
1686 #else /* !USE_SPLIT_PTE_PTLOCKS */
1688 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1690 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1692 return &mm->page_table_lock;
1694 static inline void ptlock_cache_init(void) {}
1695 static inline bool ptlock_init(struct page *page) { return true; }
1696 static inline void pte_lock_deinit(struct page *page) {}
1697 #endif /* USE_SPLIT_PTE_PTLOCKS */
1699 static inline void pgtable_init(void)
1701 ptlock_cache_init();
1702 pgtable_cache_init();
1705 static inline bool pgtable_page_ctor(struct page *page)
1707 if (!ptlock_init(page))
1709 inc_zone_page_state(page, NR_PAGETABLE);
1713 static inline void pgtable_page_dtor(struct page *page)
1715 pte_lock_deinit(page);
1716 dec_zone_page_state(page, NR_PAGETABLE);
1719 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1721 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1722 pte_t *__pte = pte_offset_map(pmd, address); \
1728 #define pte_unmap_unlock(pte, ptl) do { \
1733 #define pte_alloc(mm, pmd, address) \
1734 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1736 #define pte_alloc_map(mm, pmd, address) \
1737 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1739 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1740 (pte_alloc(mm, pmd, address) ? \
1741 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1743 #define pte_alloc_kernel(pmd, address) \
1744 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1745 NULL: pte_offset_kernel(pmd, address))
1747 #if USE_SPLIT_PMD_PTLOCKS
1749 static struct page *pmd_to_page(pmd_t *pmd)
1751 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1752 return virt_to_page((void *)((unsigned long) pmd & mask));
1755 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1757 return ptlock_ptr(pmd_to_page(pmd));
1760 static inline bool pgtable_pmd_page_ctor(struct page *page)
1762 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1763 page->pmd_huge_pte = NULL;
1765 return ptlock_init(page);
1768 static inline void pgtable_pmd_page_dtor(struct page *page)
1770 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1771 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1776 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1780 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1782 return &mm->page_table_lock;
1785 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1786 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1788 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1792 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1794 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1799 extern void __init pagecache_init(void);
1801 extern void free_area_init(unsigned long * zones_size);
1802 extern void free_area_init_node(int nid, unsigned long * zones_size,
1803 unsigned long zone_start_pfn, unsigned long *zholes_size);
1804 extern void free_initmem(void);
1807 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1808 * into the buddy system. The freed pages will be poisoned with pattern
1809 * "poison" if it's within range [0, UCHAR_MAX].
1810 * Return pages freed into the buddy system.
1812 extern unsigned long free_reserved_area(void *start, void *end,
1813 int poison, char *s);
1815 #ifdef CONFIG_HIGHMEM
1817 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1818 * and totalram_pages.
1820 extern void free_highmem_page(struct page *page);
1823 extern void adjust_managed_page_count(struct page *page, long count);
1824 extern void mem_init_print_info(const char *str);
1826 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1828 /* Free the reserved page into the buddy system, so it gets managed. */
1829 static inline void __free_reserved_page(struct page *page)
1831 ClearPageReserved(page);
1832 init_page_count(page);
1836 static inline void free_reserved_page(struct page *page)
1838 __free_reserved_page(page);
1839 adjust_managed_page_count(page, 1);
1842 static inline void mark_page_reserved(struct page *page)
1844 SetPageReserved(page);
1845 adjust_managed_page_count(page, -1);
1849 * Default method to free all the __init memory into the buddy system.
1850 * The freed pages will be poisoned with pattern "poison" if it's within
1851 * range [0, UCHAR_MAX].
1852 * Return pages freed into the buddy system.
1854 static inline unsigned long free_initmem_default(int poison)
1856 extern char __init_begin[], __init_end[];
1858 return free_reserved_area(&__init_begin, &__init_end,
1859 poison, "unused kernel");
1862 static inline unsigned long get_num_physpages(void)
1865 unsigned long phys_pages = 0;
1867 for_each_online_node(nid)
1868 phys_pages += node_present_pages(nid);
1873 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1875 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1876 * zones, allocate the backing mem_map and account for memory holes in a more
1877 * architecture independent manner. This is a substitute for creating the
1878 * zone_sizes[] and zholes_size[] arrays and passing them to
1879 * free_area_init_node()
1881 * An architecture is expected to register range of page frames backed by
1882 * physical memory with memblock_add[_node]() before calling
1883 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1884 * usage, an architecture is expected to do something like
1886 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1888 * for_each_valid_physical_page_range()
1889 * memblock_add_node(base, size, nid)
1890 * free_area_init_nodes(max_zone_pfns);
1892 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1893 * registered physical page range. Similarly
1894 * sparse_memory_present_with_active_regions() calls memory_present() for
1895 * each range when SPARSEMEM is enabled.
1897 * See mm/page_alloc.c for more information on each function exposed by
1898 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1900 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1901 unsigned long node_map_pfn_alignment(void);
1902 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1903 unsigned long end_pfn);
1904 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1905 unsigned long end_pfn);
1906 extern void get_pfn_range_for_nid(unsigned int nid,
1907 unsigned long *start_pfn, unsigned long *end_pfn);
1908 extern unsigned long find_min_pfn_with_active_regions(void);
1909 extern void free_bootmem_with_active_regions(int nid,
1910 unsigned long max_low_pfn);
1911 extern void sparse_memory_present_with_active_regions(int nid);
1913 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1915 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1916 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1917 static inline int __early_pfn_to_nid(unsigned long pfn,
1918 struct mminit_pfnnid_cache *state)
1923 /* please see mm/page_alloc.c */
1924 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1925 /* there is a per-arch backend function. */
1926 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1927 struct mminit_pfnnid_cache *state);
1930 extern void set_dma_reserve(unsigned long new_dma_reserve);
1931 extern void memmap_init_zone(unsigned long, int, unsigned long,
1932 unsigned long, enum memmap_context);
1933 extern void setup_per_zone_wmarks(void);
1934 extern int __meminit init_per_zone_wmark_min(void);
1935 extern void mem_init(void);
1936 extern void __init mmap_init(void);
1937 extern void show_mem(unsigned int flags);
1938 extern long si_mem_available(void);
1939 extern void si_meminfo(struct sysinfo * val);
1940 extern void si_meminfo_node(struct sysinfo *val, int nid);
1941 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1942 extern unsigned long arch_reserved_kernel_pages(void);
1945 extern __printf(2, 3)
1946 void warn_alloc(gfp_t gfp_mask, const char *fmt, ...);
1948 extern void setup_per_cpu_pageset(void);
1950 extern void zone_pcp_update(struct zone *zone);
1951 extern void zone_pcp_reset(struct zone *zone);
1954 extern int min_free_kbytes;
1955 extern int watermark_scale_factor;
1958 extern atomic_long_t mmap_pages_allocated;
1959 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1961 /* interval_tree.c */
1962 void vma_interval_tree_insert(struct vm_area_struct *node,
1963 struct rb_root *root);
1964 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1965 struct vm_area_struct *prev,
1966 struct rb_root *root);
1967 void vma_interval_tree_remove(struct vm_area_struct *node,
1968 struct rb_root *root);
1969 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1970 unsigned long start, unsigned long last);
1971 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1972 unsigned long start, unsigned long last);
1974 #define vma_interval_tree_foreach(vma, root, start, last) \
1975 for (vma = vma_interval_tree_iter_first(root, start, last); \
1976 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1978 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1979 struct rb_root *root);
1980 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1981 struct rb_root *root);
1982 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1983 struct rb_root *root, unsigned long start, unsigned long last);
1984 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1985 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1986 #ifdef CONFIG_DEBUG_VM_RB
1987 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1990 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1991 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1992 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1995 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1996 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
1997 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
1998 struct vm_area_struct *expand);
1999 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2000 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2002 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2004 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2005 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2006 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2007 struct mempolicy *, struct vm_userfaultfd_ctx);
2008 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2009 extern int split_vma(struct mm_struct *,
2010 struct vm_area_struct *, unsigned long addr, int new_below);
2011 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2012 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2013 struct rb_node **, struct rb_node *);
2014 extern void unlink_file_vma(struct vm_area_struct *);
2015 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2016 unsigned long addr, unsigned long len, pgoff_t pgoff,
2017 bool *need_rmap_locks);
2018 extern void exit_mmap(struct mm_struct *);
2020 static inline int check_data_rlimit(unsigned long rlim,
2022 unsigned long start,
2023 unsigned long end_data,
2024 unsigned long start_data)
2026 if (rlim < RLIM_INFINITY) {
2027 if (((new - start) + (end_data - start_data)) > rlim)
2034 extern int mm_take_all_locks(struct mm_struct *mm);
2035 extern void mm_drop_all_locks(struct mm_struct *mm);
2037 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2038 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2039 extern struct file *get_task_exe_file(struct task_struct *task);
2041 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2042 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2044 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2045 const struct vm_special_mapping *sm);
2046 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2047 unsigned long addr, unsigned long len,
2048 unsigned long flags,
2049 const struct vm_special_mapping *spec);
2050 /* This is an obsolete alternative to _install_special_mapping. */
2051 extern int install_special_mapping(struct mm_struct *mm,
2052 unsigned long addr, unsigned long len,
2053 unsigned long flags, struct page **pages);
2055 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2057 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2058 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2059 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2060 unsigned long len, unsigned long prot, unsigned long flags,
2061 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2062 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2064 static inline unsigned long
2065 do_mmap_pgoff(struct file *file, unsigned long addr,
2066 unsigned long len, unsigned long prot, unsigned long flags,
2067 unsigned long pgoff, unsigned long *populate)
2069 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2073 extern int __mm_populate(unsigned long addr, unsigned long len,
2075 static inline void mm_populate(unsigned long addr, unsigned long len)
2078 (void) __mm_populate(addr, len, 1);
2081 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2084 /* These take the mm semaphore themselves */
2085 extern int __must_check vm_brk(unsigned long, unsigned long);
2086 extern int vm_munmap(unsigned long, size_t);
2087 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2088 unsigned long, unsigned long,
2089 unsigned long, unsigned long);
2091 struct vm_unmapped_area_info {
2092 #define VM_UNMAPPED_AREA_TOPDOWN 1
2093 unsigned long flags;
2094 unsigned long length;
2095 unsigned long low_limit;
2096 unsigned long high_limit;
2097 unsigned long align_mask;
2098 unsigned long align_offset;
2101 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2102 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2105 * Search for an unmapped address range.
2107 * We are looking for a range that:
2108 * - does not intersect with any VMA;
2109 * - is contained within the [low_limit, high_limit) interval;
2110 * - is at least the desired size.
2111 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2113 static inline unsigned long
2114 vm_unmapped_area(struct vm_unmapped_area_info *info)
2116 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2117 return unmapped_area_topdown(info);
2119 return unmapped_area(info);
2123 extern void truncate_inode_pages(struct address_space *, loff_t);
2124 extern void truncate_inode_pages_range(struct address_space *,
2125 loff_t lstart, loff_t lend);
2126 extern void truncate_inode_pages_final(struct address_space *);
2128 /* generic vm_area_ops exported for stackable file systems */
2129 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2130 extern void filemap_map_pages(struct vm_fault *vmf,
2131 pgoff_t start_pgoff, pgoff_t end_pgoff);
2132 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2134 /* mm/page-writeback.c */
2135 int write_one_page(struct page *page, int wait);
2136 void task_dirty_inc(struct task_struct *tsk);
2139 #define VM_MAX_READAHEAD 128 /* kbytes */
2140 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2142 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2143 pgoff_t offset, unsigned long nr_to_read);
2145 void page_cache_sync_readahead(struct address_space *mapping,
2146 struct file_ra_state *ra,
2149 unsigned long size);
2151 void page_cache_async_readahead(struct address_space *mapping,
2152 struct file_ra_state *ra,
2156 unsigned long size);
2158 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2159 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2161 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2162 extern int expand_downwards(struct vm_area_struct *vma,
2163 unsigned long address);
2165 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2167 #define expand_upwards(vma, address) (0)
2170 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2171 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2172 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2173 struct vm_area_struct **pprev);
2175 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2176 NULL if none. Assume start_addr < end_addr. */
2177 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2179 struct vm_area_struct * vma = find_vma(mm,start_addr);
2181 if (vma && end_addr <= vma->vm_start)
2186 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2188 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2191 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2192 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2193 unsigned long vm_start, unsigned long vm_end)
2195 struct vm_area_struct *vma = find_vma(mm, vm_start);
2197 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2204 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2205 void vma_set_page_prot(struct vm_area_struct *vma);
2207 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2211 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2213 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2217 #ifdef CONFIG_NUMA_BALANCING
2218 unsigned long change_prot_numa(struct vm_area_struct *vma,
2219 unsigned long start, unsigned long end);
2222 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2223 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2224 unsigned long pfn, unsigned long size, pgprot_t);
2225 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2226 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2228 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2229 unsigned long pfn, pgprot_t pgprot);
2230 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2232 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2235 struct page *follow_page_mask(struct vm_area_struct *vma,
2236 unsigned long address, unsigned int foll_flags,
2237 unsigned int *page_mask);
2239 static inline struct page *follow_page(struct vm_area_struct *vma,
2240 unsigned long address, unsigned int foll_flags)
2242 unsigned int unused_page_mask;
2243 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2246 #define FOLL_WRITE 0x01 /* check pte is writable */
2247 #define FOLL_TOUCH 0x02 /* mark page accessed */
2248 #define FOLL_GET 0x04 /* do get_page on page */
2249 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2250 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2251 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2252 * and return without waiting upon it */
2253 #define FOLL_POPULATE 0x40 /* fault in page */
2254 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2255 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2256 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2257 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2258 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2259 #define FOLL_MLOCK 0x1000 /* lock present pages */
2260 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2261 #define FOLL_COW 0x4000 /* internal GUP flag */
2263 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2265 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2266 unsigned long size, pte_fn_t fn, void *data);
2269 #ifdef CONFIG_PAGE_POISONING
2270 extern bool page_poisoning_enabled(void);
2271 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2272 extern bool page_is_poisoned(struct page *page);
2274 static inline bool page_poisoning_enabled(void) { return false; }
2275 static inline void kernel_poison_pages(struct page *page, int numpages,
2277 static inline bool page_is_poisoned(struct page *page) { return false; }
2280 #ifdef CONFIG_DEBUG_PAGEALLOC
2281 extern bool _debug_pagealloc_enabled;
2282 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2284 static inline bool debug_pagealloc_enabled(void)
2286 return _debug_pagealloc_enabled;
2290 kernel_map_pages(struct page *page, int numpages, int enable)
2292 if (!debug_pagealloc_enabled())
2295 __kernel_map_pages(page, numpages, enable);
2297 #ifdef CONFIG_HIBERNATION
2298 extern bool kernel_page_present(struct page *page);
2299 #endif /* CONFIG_HIBERNATION */
2300 #else /* CONFIG_DEBUG_PAGEALLOC */
2302 kernel_map_pages(struct page *page, int numpages, int enable) {}
2303 #ifdef CONFIG_HIBERNATION
2304 static inline bool kernel_page_present(struct page *page) { return true; }
2305 #endif /* CONFIG_HIBERNATION */
2306 static inline bool debug_pagealloc_enabled(void)
2310 #endif /* CONFIG_DEBUG_PAGEALLOC */
2312 #ifdef __HAVE_ARCH_GATE_AREA
2313 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2314 extern int in_gate_area_no_mm(unsigned long addr);
2315 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2317 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2321 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2322 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2326 #endif /* __HAVE_ARCH_GATE_AREA */
2328 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2330 #ifdef CONFIG_SYSCTL
2331 extern int sysctl_drop_caches;
2332 int drop_caches_sysctl_handler(struct ctl_table *, int,
2333 void __user *, size_t *, loff_t *);
2336 void drop_slab(void);
2337 void drop_slab_node(int nid);
2340 #define randomize_va_space 0
2342 extern int randomize_va_space;
2345 const char * arch_vma_name(struct vm_area_struct *vma);
2346 void print_vma_addr(char *prefix, unsigned long rip);
2348 void sparse_mem_maps_populate_node(struct page **map_map,
2349 unsigned long pnum_begin,
2350 unsigned long pnum_end,
2351 unsigned long map_count,
2354 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2355 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2356 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2357 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2358 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2359 void *vmemmap_alloc_block(unsigned long size, int node);
2361 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2362 struct vmem_altmap *altmap);
2363 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2365 return __vmemmap_alloc_block_buf(size, node, NULL);
2368 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2369 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2371 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2372 void vmemmap_populate_print_last(void);
2373 #ifdef CONFIG_MEMORY_HOTPLUG
2374 void vmemmap_free(unsigned long start, unsigned long end);
2376 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2377 unsigned long size);
2380 MF_COUNT_INCREASED = 1 << 0,
2381 MF_ACTION_REQUIRED = 1 << 1,
2382 MF_MUST_KILL = 1 << 2,
2383 MF_SOFT_OFFLINE = 1 << 3,
2385 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2386 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2387 extern int unpoison_memory(unsigned long pfn);
2388 extern int get_hwpoison_page(struct page *page);
2389 #define put_hwpoison_page(page) put_page(page)
2390 extern int sysctl_memory_failure_early_kill;
2391 extern int sysctl_memory_failure_recovery;
2392 extern void shake_page(struct page *p, int access);
2393 extern atomic_long_t num_poisoned_pages;
2394 extern int soft_offline_page(struct page *page, int flags);
2398 * Error handlers for various types of pages.
2401 MF_IGNORED, /* Error: cannot be handled */
2402 MF_FAILED, /* Error: handling failed */
2403 MF_DELAYED, /* Will be handled later */
2404 MF_RECOVERED, /* Successfully recovered */
2407 enum mf_action_page_type {
2409 MF_MSG_KERNEL_HIGH_ORDER,
2411 MF_MSG_DIFFERENT_COMPOUND,
2412 MF_MSG_POISONED_HUGE,
2415 MF_MSG_UNMAP_FAILED,
2416 MF_MSG_DIRTY_SWAPCACHE,
2417 MF_MSG_CLEAN_SWAPCACHE,
2418 MF_MSG_DIRTY_MLOCKED_LRU,
2419 MF_MSG_CLEAN_MLOCKED_LRU,
2420 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2421 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2424 MF_MSG_TRUNCATED_LRU,
2430 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2431 extern void clear_huge_page(struct page *page,
2433 unsigned int pages_per_huge_page);
2434 extern void copy_user_huge_page(struct page *dst, struct page *src,
2435 unsigned long addr, struct vm_area_struct *vma,
2436 unsigned int pages_per_huge_page);
2437 extern long copy_huge_page_from_user(struct page *dst_page,
2438 const void __user *usr_src,
2439 unsigned int pages_per_huge_page,
2440 bool allow_pagefault);
2441 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2443 extern struct page_ext_operations debug_guardpage_ops;
2444 extern struct page_ext_operations page_poisoning_ops;
2446 #ifdef CONFIG_DEBUG_PAGEALLOC
2447 extern unsigned int _debug_guardpage_minorder;
2448 extern bool _debug_guardpage_enabled;
2450 static inline unsigned int debug_guardpage_minorder(void)
2452 return _debug_guardpage_minorder;
2455 static inline bool debug_guardpage_enabled(void)
2457 return _debug_guardpage_enabled;
2460 static inline bool page_is_guard(struct page *page)
2462 struct page_ext *page_ext;
2464 if (!debug_guardpage_enabled())
2467 page_ext = lookup_page_ext(page);
2468 if (unlikely(!page_ext))
2471 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2474 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2475 static inline bool debug_guardpage_enabled(void) { return false; }
2476 static inline bool page_is_guard(struct page *page) { return false; }
2477 #endif /* CONFIG_DEBUG_PAGEALLOC */
2479 #if MAX_NUMNODES > 1
2480 void __init setup_nr_node_ids(void);
2482 static inline void setup_nr_node_ids(void) {}
2485 #endif /* __KERNEL__ */
2486 #endif /* _LINUX_MM_H */