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 */
289 * vm_fault is filled by the the pagefault handler and passed to the vma's
290 * ->fault function. The vma's ->fault is responsible for returning a bitmask
291 * of VM_FAULT_xxx flags that give details about how the fault was handled.
293 * MM layer fills up gfp_mask for page allocations but fault handler might
294 * alter it if its implementation requires a different allocation context.
296 * pgoff should be used in favour of virtual_address, if possible.
299 unsigned int flags; /* FAULT_FLAG_xxx flags */
300 gfp_t gfp_mask; /* gfp mask to be used for allocations */
301 pgoff_t pgoff; /* Logical page offset based on vma */
302 void __user *virtual_address; /* Faulting virtual address */
304 struct page *cow_page; /* Handler may choose to COW */
305 struct page *page; /* ->fault handlers should return a
306 * page here, unless VM_FAULT_NOPAGE
307 * is set (which is also implied by
310 void *entry; /* ->fault handler can alternatively
311 * return locked DAX entry. In that
312 * case handler should return
313 * VM_FAULT_DAX_LOCKED and fill in
319 * Page fault context: passes though page fault handler instead of endless list
320 * of function arguments.
323 struct vm_area_struct *vma; /* Target VMA */
324 unsigned long address; /* Faulting virtual address */
325 unsigned int flags; /* FAULT_FLAG_xxx flags */
326 pmd_t *pmd; /* Pointer to pmd entry matching
329 pte_t *pte; /* Pointer to pte entry matching
330 * the 'address'. NULL if the page
331 * table hasn't been allocated.
333 spinlock_t *ptl; /* Page table lock.
334 * Protects pte page table if 'pte'
335 * is not NULL, otherwise pmd.
337 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
338 * vm_ops->map_pages() calls
339 * alloc_set_pte() from atomic context.
340 * do_fault_around() pre-allocates
341 * page table to avoid allocation from
347 * These are the virtual MM functions - opening of an area, closing and
348 * unmapping it (needed to keep files on disk up-to-date etc), pointer
349 * to the functions called when a no-page or a wp-page exception occurs.
351 struct vm_operations_struct {
352 void (*open)(struct vm_area_struct * area);
353 void (*close)(struct vm_area_struct * area);
354 int (*split)(struct vm_area_struct * area, unsigned long addr);
355 int (*mremap)(struct vm_area_struct * area);
356 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
357 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
358 pmd_t *, unsigned int flags);
359 void (*map_pages)(struct fault_env *fe,
360 pgoff_t start_pgoff, pgoff_t end_pgoff);
362 /* notification that a previously read-only page is about to become
363 * writable, if an error is returned it will cause a SIGBUS */
364 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
366 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
367 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
369 /* called by access_process_vm when get_user_pages() fails, typically
370 * for use by special VMAs that can switch between memory and hardware
372 int (*access)(struct vm_area_struct *vma, unsigned long addr,
373 void *buf, int len, int write);
375 /* Called by the /proc/PID/maps code to ask the vma whether it
376 * has a special name. Returning non-NULL will also cause this
377 * vma to be dumped unconditionally. */
378 const char *(*name)(struct vm_area_struct *vma);
382 * set_policy() op must add a reference to any non-NULL @new mempolicy
383 * to hold the policy upon return. Caller should pass NULL @new to
384 * remove a policy and fall back to surrounding context--i.e. do not
385 * install a MPOL_DEFAULT policy, nor the task or system default
388 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
391 * get_policy() op must add reference [mpol_get()] to any policy at
392 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
393 * in mm/mempolicy.c will do this automatically.
394 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
395 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
396 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
397 * must return NULL--i.e., do not "fallback" to task or system default
400 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
404 * Called by vm_normal_page() for special PTEs to find the
405 * page for @addr. This is useful if the default behavior
406 * (using pte_page()) would not find the correct page.
408 struct page *(*find_special_page)(struct vm_area_struct *vma,
415 #define page_private(page) ((page)->private)
416 #define set_page_private(page, v) ((page)->private = (v))
418 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
419 static inline int pmd_devmap(pmd_t pmd)
426 * FIXME: take this include out, include page-flags.h in
427 * files which need it (119 of them)
429 #include <linux/page-flags.h>
430 #include <linux/huge_mm.h>
433 * Methods to modify the page usage count.
435 * What counts for a page usage:
436 * - cache mapping (page->mapping)
437 * - private data (page->private)
438 * - page mapped in a task's page tables, each mapping
439 * is counted separately
441 * Also, many kernel routines increase the page count before a critical
442 * routine so they can be sure the page doesn't go away from under them.
446 * Drop a ref, return true if the refcount fell to zero (the page has no users)
448 static inline int put_page_testzero(struct page *page)
450 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
451 return page_ref_dec_and_test(page);
455 * Try to grab a ref unless the page has a refcount of zero, return false if
457 * This can be called when MMU is off so it must not access
458 * any of the virtual mappings.
460 static inline int get_page_unless_zero(struct page *page)
462 return page_ref_add_unless(page, 1, 0);
465 extern int page_is_ram(unsigned long pfn);
473 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
476 /* Support for virtually mapped pages */
477 struct page *vmalloc_to_page(const void *addr);
478 unsigned long vmalloc_to_pfn(const void *addr);
481 * Determine if an address is within the vmalloc range
483 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
484 * is no special casing required.
486 static inline bool is_vmalloc_addr(const void *x)
489 unsigned long addr = (unsigned long)x;
491 return addr >= VMALLOC_START && addr < VMALLOC_END;
497 extern int is_vmalloc_or_module_addr(const void *x);
499 static inline int is_vmalloc_or_module_addr(const void *x)
505 extern void kvfree(const void *addr);
508 * Mapcount of compound page as a whole, does not include mapped sub-pages.
510 * Must be called only for compound pages or any their tail sub-pages.
512 static inline int compound_mapcount(struct page *page)
514 VM_BUG_ON_PAGE(!PageCompound(page), page);
515 page = compound_head(page);
516 return atomic_read(compound_mapcount_ptr(page)) + 1;
520 * The atomic page->_mapcount, starts from -1: so that transitions
521 * both from it and to it can be tracked, using atomic_inc_and_test
522 * and atomic_add_negative(-1).
524 static inline void page_mapcount_reset(struct page *page)
526 atomic_set(&(page)->_mapcount, -1);
529 int __page_mapcount(struct page *page);
532 * Mapcount of 0-order page; when compound sub-page, includes
533 * compound_mapcount().
535 * Result is undefined for pages which cannot be mapped into userspace.
536 * For example SLAB or special types of pages. See function page_has_type().
537 * They use this place in struct page differently.
539 static inline int page_mapcount(struct page *page)
541 if (unlikely(PageCompound(page)))
542 return __page_mapcount(page);
543 return atomic_read(&page->_mapcount) + 1;
546 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
547 int total_mapcount(struct page *page);
548 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
550 static inline int total_mapcount(struct page *page)
552 return page_mapcount(page);
554 static inline int page_trans_huge_mapcount(struct page *page,
557 int mapcount = page_mapcount(page);
559 *total_mapcount = mapcount;
564 static inline struct page *virt_to_head_page(const void *x)
566 struct page *page = virt_to_page(x);
568 return compound_head(page);
571 void __put_page(struct page *page);
573 void put_pages_list(struct list_head *pages);
575 void split_page(struct page *page, unsigned int order);
578 * Compound pages have a destructor function. Provide a
579 * prototype for that function and accessor functions.
580 * These are _only_ valid on the head of a compound page.
582 typedef void compound_page_dtor(struct page *);
584 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
585 enum compound_dtor_id {
588 #ifdef CONFIG_HUGETLB_PAGE
591 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
596 extern compound_page_dtor * const compound_page_dtors[];
598 static inline void set_compound_page_dtor(struct page *page,
599 enum compound_dtor_id compound_dtor)
601 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
602 page[1].compound_dtor = compound_dtor;
605 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
607 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
608 return compound_page_dtors[page[1].compound_dtor];
611 static inline unsigned int compound_order(struct page *page)
615 return page[1].compound_order;
618 static inline void set_compound_order(struct page *page, unsigned int order)
620 page[1].compound_order = order;
623 void free_compound_page(struct page *page);
627 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
628 * servicing faults for write access. In the normal case, do always want
629 * pte_mkwrite. But get_user_pages can cause write faults for mappings
630 * that do not have writing enabled, when used by access_process_vm.
632 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
634 if (likely(vma->vm_flags & VM_WRITE))
635 pte = pte_mkwrite(pte);
639 int alloc_set_pte(struct fault_env *fe, struct mem_cgroup *memcg,
644 * Multiple processes may "see" the same page. E.g. for untouched
645 * mappings of /dev/null, all processes see the same page full of
646 * zeroes, and text pages of executables and shared libraries have
647 * only one copy in memory, at most, normally.
649 * For the non-reserved pages, page_count(page) denotes a reference count.
650 * page_count() == 0 means the page is free. page->lru is then used for
651 * freelist management in the buddy allocator.
652 * page_count() > 0 means the page has been allocated.
654 * Pages are allocated by the slab allocator in order to provide memory
655 * to kmalloc and kmem_cache_alloc. In this case, the management of the
656 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
657 * unless a particular usage is carefully commented. (the responsibility of
658 * freeing the kmalloc memory is the caller's, of course).
660 * A page may be used by anyone else who does a __get_free_page().
661 * In this case, page_count still tracks the references, and should only
662 * be used through the normal accessor functions. The top bits of page->flags
663 * and page->virtual store page management information, but all other fields
664 * are unused and could be used privately, carefully. The management of this
665 * page is the responsibility of the one who allocated it, and those who have
666 * subsequently been given references to it.
668 * The other pages (we may call them "pagecache pages") are completely
669 * managed by the Linux memory manager: I/O, buffers, swapping etc.
670 * The following discussion applies only to them.
672 * A pagecache page contains an opaque `private' member, which belongs to the
673 * page's address_space. Usually, this is the address of a circular list of
674 * the page's disk buffers. PG_private must be set to tell the VM to call
675 * into the filesystem to release these pages.
677 * A page may belong to an inode's memory mapping. In this case, page->mapping
678 * is the pointer to the inode, and page->index is the file offset of the page,
679 * in units of PAGE_SIZE.
681 * If pagecache pages are not associated with an inode, they are said to be
682 * anonymous pages. These may become associated with the swapcache, and in that
683 * case PG_swapcache is set, and page->private is an offset into the swapcache.
685 * In either case (swapcache or inode backed), the pagecache itself holds one
686 * reference to the page. Setting PG_private should also increment the
687 * refcount. The each user mapping also has a reference to the page.
689 * The pagecache pages are stored in a per-mapping radix tree, which is
690 * rooted at mapping->page_tree, and indexed by offset.
691 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
692 * lists, we instead now tag pages as dirty/writeback in the radix tree.
694 * All pagecache pages may be subject to I/O:
695 * - inode pages may need to be read from disk,
696 * - inode pages which have been modified and are MAP_SHARED may need
697 * to be written back to the inode on disk,
698 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
699 * modified may need to be swapped out to swap space and (later) to be read
704 * The zone field is never updated after free_area_init_core()
705 * sets it, so none of the operations on it need to be atomic.
708 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
709 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
710 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
711 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
712 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
715 * Define the bit shifts to access each section. For non-existent
716 * sections we define the shift as 0; that plus a 0 mask ensures
717 * the compiler will optimise away reference to them.
719 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
720 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
721 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
722 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
724 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
725 #ifdef NODE_NOT_IN_PAGE_FLAGS
726 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
727 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
728 SECTIONS_PGOFF : ZONES_PGOFF)
730 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
731 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
732 NODES_PGOFF : ZONES_PGOFF)
735 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
737 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
738 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
741 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
742 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
743 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
744 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
745 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
747 static inline enum zone_type page_zonenum(const struct page *page)
749 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
752 #ifdef CONFIG_ZONE_DEVICE
753 void get_zone_device_page(struct page *page);
754 void put_zone_device_page(struct page *page);
755 static inline bool is_zone_device_page(const struct page *page)
757 return page_zonenum(page) == ZONE_DEVICE;
760 static inline void get_zone_device_page(struct page *page)
763 static inline void put_zone_device_page(struct page *page)
766 static inline bool is_zone_device_page(const struct page *page)
772 /* 127: arbitrary random number, small enough to assemble well */
773 #define page_ref_zero_or_close_to_overflow(page) \
774 ((unsigned int) page_ref_count(page) + 127u <= 127u)
776 static inline void get_page(struct page *page)
778 page = compound_head(page);
780 * Getting a normal page or the head of a compound page
781 * requires to already have an elevated page->_refcount.
783 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
786 if (unlikely(is_zone_device_page(page)))
787 get_zone_device_page(page);
790 static inline __must_check bool try_get_page(struct page *page)
792 page = compound_head(page);
793 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
799 static inline void put_page(struct page *page)
801 page = compound_head(page);
803 if (put_page_testzero(page))
806 if (unlikely(is_zone_device_page(page)))
807 put_zone_device_page(page);
810 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
811 #define SECTION_IN_PAGE_FLAGS
815 * The identification function is mainly used by the buddy allocator for
816 * determining if two pages could be buddies. We are not really identifying
817 * the zone since we could be using the section number id if we do not have
818 * node id available in page flags.
819 * We only guarantee that it will return the same value for two combinable
822 static inline int page_zone_id(struct page *page)
824 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
827 static inline int zone_to_nid(struct zone *zone)
836 #ifdef NODE_NOT_IN_PAGE_FLAGS
837 extern int page_to_nid(const struct page *page);
839 static inline int page_to_nid(const struct page *page)
841 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
845 #ifdef CONFIG_NUMA_BALANCING
846 static inline int cpu_pid_to_cpupid(int cpu, int pid)
848 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
851 static inline int cpupid_to_pid(int cpupid)
853 return cpupid & LAST__PID_MASK;
856 static inline int cpupid_to_cpu(int cpupid)
858 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
861 static inline int cpupid_to_nid(int cpupid)
863 return cpu_to_node(cpupid_to_cpu(cpupid));
866 static inline bool cpupid_pid_unset(int cpupid)
868 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
871 static inline bool cpupid_cpu_unset(int cpupid)
873 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
876 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
878 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
881 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
882 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
883 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
885 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
888 static inline int page_cpupid_last(struct page *page)
890 return page->_last_cpupid;
892 static inline void page_cpupid_reset_last(struct page *page)
894 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
897 static inline int page_cpupid_last(struct page *page)
899 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
902 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
904 static inline void page_cpupid_reset_last(struct page *page)
906 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
908 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
909 #else /* !CONFIG_NUMA_BALANCING */
910 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
912 return page_to_nid(page); /* XXX */
915 static inline int page_cpupid_last(struct page *page)
917 return page_to_nid(page); /* XXX */
920 static inline int cpupid_to_nid(int cpupid)
925 static inline int cpupid_to_pid(int cpupid)
930 static inline int cpupid_to_cpu(int cpupid)
935 static inline int cpu_pid_to_cpupid(int nid, int pid)
940 static inline bool cpupid_pid_unset(int cpupid)
945 static inline void page_cpupid_reset_last(struct page *page)
949 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
953 #endif /* CONFIG_NUMA_BALANCING */
955 static inline struct zone *page_zone(const struct page *page)
957 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
960 static inline pg_data_t *page_pgdat(const struct page *page)
962 return NODE_DATA(page_to_nid(page));
965 #ifdef SECTION_IN_PAGE_FLAGS
966 static inline void set_page_section(struct page *page, unsigned long section)
968 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
969 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
972 static inline unsigned long page_to_section(const struct page *page)
974 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
978 static inline void set_page_zone(struct page *page, enum zone_type zone)
980 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
981 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
984 static inline void set_page_node(struct page *page, unsigned long node)
986 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
987 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
990 static inline void set_page_links(struct page *page, enum zone_type zone,
991 unsigned long node, unsigned long pfn)
993 set_page_zone(page, zone);
994 set_page_node(page, node);
995 #ifdef SECTION_IN_PAGE_FLAGS
996 set_page_section(page, pfn_to_section_nr(pfn));
1001 static inline struct mem_cgroup *page_memcg(struct page *page)
1003 return page->mem_cgroup;
1005 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1007 WARN_ON_ONCE(!rcu_read_lock_held());
1008 return READ_ONCE(page->mem_cgroup);
1011 static inline struct mem_cgroup *page_memcg(struct page *page)
1015 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1017 WARN_ON_ONCE(!rcu_read_lock_held());
1023 * Some inline functions in vmstat.h depend on page_zone()
1025 #include <linux/vmstat.h>
1027 static __always_inline void *lowmem_page_address(const struct page *page)
1029 return page_to_virt(page);
1032 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1033 #define HASHED_PAGE_VIRTUAL
1036 #if defined(WANT_PAGE_VIRTUAL)
1037 static inline void *page_address(const struct page *page)
1039 return page->virtual;
1041 static inline void set_page_address(struct page *page, void *address)
1043 page->virtual = address;
1045 #define page_address_init() do { } while(0)
1048 #if defined(HASHED_PAGE_VIRTUAL)
1049 void *page_address(const struct page *page);
1050 void set_page_address(struct page *page, void *virtual);
1051 void page_address_init(void);
1054 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1055 #define page_address(page) lowmem_page_address(page)
1056 #define set_page_address(page, address) do { } while(0)
1057 #define page_address_init() do { } while(0)
1060 extern void *page_rmapping(struct page *page);
1061 extern struct anon_vma *page_anon_vma(struct page *page);
1062 extern struct address_space *page_mapping(struct page *page);
1064 extern struct address_space *__page_file_mapping(struct page *);
1067 struct address_space *page_file_mapping(struct page *page)
1069 if (unlikely(PageSwapCache(page)))
1070 return __page_file_mapping(page);
1072 return page->mapping;
1075 extern pgoff_t __page_file_index(struct page *page);
1078 * Return the pagecache index of the passed page. Regular pagecache pages
1079 * use ->index whereas swapcache pages use swp_offset(->private)
1081 static inline pgoff_t page_index(struct page *page)
1083 if (unlikely(PageSwapCache(page)))
1084 return __page_file_index(page);
1088 bool page_mapped(struct page *page);
1089 struct address_space *page_mapping(struct page *page);
1092 * Return true only if the page has been allocated with
1093 * ALLOC_NO_WATERMARKS and the low watermark was not
1094 * met implying that the system is under some pressure.
1096 static inline bool page_is_pfmemalloc(struct page *page)
1099 * Page index cannot be this large so this must be
1100 * a pfmemalloc page.
1102 return page->index == -1UL;
1106 * Only to be called by the page allocator on a freshly allocated
1109 static inline void set_page_pfmemalloc(struct page *page)
1114 static inline void clear_page_pfmemalloc(struct page *page)
1120 * Different kinds of faults, as returned by handle_mm_fault().
1121 * Used to decide whether a process gets delivered SIGBUS or
1122 * just gets major/minor fault counters bumped up.
1125 #define VM_FAULT_OOM 0x0001
1126 #define VM_FAULT_SIGBUS 0x0002
1127 #define VM_FAULT_MAJOR 0x0004
1128 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1129 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1130 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1131 #define VM_FAULT_SIGSEGV 0x0040
1133 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1134 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1135 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1136 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1137 #define VM_FAULT_DAX_LOCKED 0x1000 /* ->fault has locked DAX entry */
1139 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1141 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1142 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1145 /* Encode hstate index for a hwpoisoned large page */
1146 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1147 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1150 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1152 extern void pagefault_out_of_memory(void);
1154 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1157 * Flags passed to show_mem() and show_free_areas() to suppress output in
1160 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1162 extern void show_free_areas(unsigned int flags);
1163 extern bool skip_free_areas_node(unsigned int flags, int nid);
1165 int shmem_zero_setup(struct vm_area_struct *);
1167 bool shmem_mapping(struct address_space *mapping);
1169 static inline bool shmem_mapping(struct address_space *mapping)
1175 extern bool can_do_mlock(void);
1176 extern int user_shm_lock(size_t, struct user_struct *);
1177 extern void user_shm_unlock(size_t, struct user_struct *);
1180 * Parameter block passed down to zap_pte_range in exceptional cases.
1182 struct zap_details {
1183 struct address_space *check_mapping; /* Check page->mapping if set */
1184 pgoff_t first_index; /* Lowest page->index to unmap */
1185 pgoff_t last_index; /* Highest page->index to unmap */
1186 bool ignore_dirty; /* Ignore dirty pages */
1187 bool check_swap_entries; /* Check also swap entries */
1190 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1192 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1195 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1196 unsigned long size);
1197 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1198 unsigned long size, struct zap_details *);
1199 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1200 unsigned long start, unsigned long end);
1202 * This has to be called after a get_task_mm()/mmget_not_zero()
1203 * followed by taking the mmap_sem for writing before modifying the
1204 * vmas or anything the coredump pretends not to change from under it.
1206 * It also has to be called when mmgrab() is used in the context of
1207 * the process, but then the mm_count refcount is transferred outside
1208 * the context of the process to run down_write() on that pinned mm.
1210 * NOTE: find_extend_vma() called from GUP context is the only place
1211 * that can modify the "mm" (notably the vm_start/end) under mmap_sem
1212 * for reading and outside the context of the process, so it is also
1213 * the only case that holds the mmap_sem for reading that must call
1214 * this function. Generally if the mmap_sem is hold for reading
1215 * there's no need of this check after get_task_mm()/mmget_not_zero().
1217 * This function can be obsoleted and the check can be removed, after
1218 * the coredump code will hold the mmap_sem for writing before
1219 * invoking the ->core_dump methods.
1221 static inline bool mmget_still_valid(struct mm_struct *mm)
1223 return likely(!mm->core_state);
1227 * mm_walk - callbacks for walk_page_range
1228 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1229 * this handler is required to be able to handle
1230 * pmd_trans_huge() pmds. They may simply choose to
1231 * split_huge_page() instead of handling it explicitly.
1232 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1233 * @pte_hole: if set, called for each hole at all levels
1234 * @hugetlb_entry: if set, called for each hugetlb entry
1235 * @test_walk: caller specific callback function to determine whether
1236 * we walk over the current vma or not. Returning 0
1237 * value means "do page table walk over the current vma,"
1238 * and a negative one means "abort current page table walk
1239 * right now." 1 means "skip the current vma."
1240 * @mm: mm_struct representing the target process of page table walk
1241 * @vma: vma currently walked (NULL if walking outside vmas)
1242 * @private: private data for callbacks' usage
1244 * (see the comment on walk_page_range() for more details)
1247 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1248 unsigned long next, struct mm_walk *walk);
1249 int (*pte_entry)(pte_t *pte, unsigned long addr,
1250 unsigned long next, struct mm_walk *walk);
1251 int (*pte_hole)(unsigned long addr, unsigned long next,
1252 struct mm_walk *walk);
1253 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1254 unsigned long addr, unsigned long next,
1255 struct mm_walk *walk);
1256 int (*test_walk)(unsigned long addr, unsigned long next,
1257 struct mm_walk *walk);
1258 struct mm_struct *mm;
1259 struct vm_area_struct *vma;
1263 int walk_page_range(unsigned long addr, unsigned long end,
1264 struct mm_walk *walk);
1265 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1266 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1267 unsigned long end, unsigned long floor, unsigned long ceiling);
1268 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1269 struct vm_area_struct *vma);
1270 void unmap_mapping_range(struct address_space *mapping,
1271 loff_t const holebegin, loff_t const holelen, int even_cows);
1272 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1273 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1274 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1275 unsigned long *pfn);
1276 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1277 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1278 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1279 void *buf, int len, int write);
1281 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1282 loff_t const holebegin, loff_t const holelen)
1284 unmap_mapping_range(mapping, holebegin, holelen, 0);
1287 extern void truncate_pagecache(struct inode *inode, loff_t new);
1288 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1289 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1290 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1291 int truncate_inode_page(struct address_space *mapping, struct page *page);
1292 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1293 int invalidate_inode_page(struct page *page);
1296 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1297 unsigned int flags);
1298 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1299 unsigned long address, unsigned int fault_flags,
1302 static inline int handle_mm_fault(struct vm_area_struct *vma,
1303 unsigned long address, unsigned int flags)
1305 /* should never happen if there's no MMU */
1307 return VM_FAULT_SIGBUS;
1309 static inline int fixup_user_fault(struct task_struct *tsk,
1310 struct mm_struct *mm, unsigned long address,
1311 unsigned int fault_flags, bool *unlocked)
1313 /* should never happen if there's no MMU */
1319 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1320 unsigned int gup_flags);
1321 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1322 void *buf, int len, unsigned int gup_flags);
1323 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1324 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1326 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1327 unsigned long start, unsigned long nr_pages,
1328 unsigned int gup_flags, struct page **pages,
1329 struct vm_area_struct **vmas);
1330 long get_user_pages(unsigned long start, unsigned long nr_pages,
1331 unsigned int gup_flags, struct page **pages,
1332 struct vm_area_struct **vmas);
1333 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1334 unsigned int gup_flags, struct page **pages, int *locked);
1335 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1336 unsigned long start, unsigned long nr_pages,
1337 struct page **pages, unsigned int gup_flags);
1338 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1339 struct page **pages, unsigned int gup_flags);
1340 #ifdef CONFIG_FS_DAX
1341 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1342 unsigned int gup_flags, struct page **pages,
1343 struct vm_area_struct **vmas);
1345 static inline long get_user_pages_longterm(unsigned long start,
1346 unsigned long nr_pages, unsigned int gup_flags,
1347 struct page **pages, struct vm_area_struct **vmas)
1349 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1351 #endif /* CONFIG_FS_DAX */
1353 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1354 struct page **pages);
1356 /* Container for pinned pfns / pages */
1357 struct frame_vector {
1358 unsigned int nr_allocated; /* Number of frames we have space for */
1359 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1360 bool got_ref; /* Did we pin pages by getting page ref? */
1361 bool is_pfns; /* Does array contain pages or pfns? */
1362 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1363 * pfns_vector_pages() or pfns_vector_pfns()
1367 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1368 void frame_vector_destroy(struct frame_vector *vec);
1369 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1370 unsigned int gup_flags, struct frame_vector *vec);
1371 void put_vaddr_frames(struct frame_vector *vec);
1372 int frame_vector_to_pages(struct frame_vector *vec);
1373 void frame_vector_to_pfns(struct frame_vector *vec);
1375 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1377 return vec->nr_frames;
1380 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1383 int err = frame_vector_to_pages(vec);
1386 return ERR_PTR(err);
1388 return (struct page **)(vec->ptrs);
1391 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1394 frame_vector_to_pfns(vec);
1395 return (unsigned long *)(vec->ptrs);
1399 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1400 struct page **pages);
1401 int get_kernel_page(unsigned long start, int write, struct page **pages);
1402 struct page *get_dump_page(unsigned long addr);
1404 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1405 extern void do_invalidatepage(struct page *page, unsigned int offset,
1406 unsigned int length);
1408 int __set_page_dirty_nobuffers(struct page *page);
1409 int __set_page_dirty_no_writeback(struct page *page);
1410 int redirty_page_for_writepage(struct writeback_control *wbc,
1412 void account_page_dirtied(struct page *page, struct address_space *mapping);
1413 void account_page_cleaned(struct page *page, struct address_space *mapping,
1414 struct bdi_writeback *wb);
1415 int set_page_dirty(struct page *page);
1416 int set_page_dirty_lock(struct page *page);
1417 void cancel_dirty_page(struct page *page);
1418 int clear_page_dirty_for_io(struct page *page);
1420 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1422 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1424 return !vma->vm_ops;
1427 int vma_is_stack_for_current(struct vm_area_struct *vma);
1429 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1430 unsigned long old_addr, struct vm_area_struct *new_vma,
1431 unsigned long new_addr, unsigned long len,
1432 bool need_rmap_locks);
1433 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1434 unsigned long end, pgprot_t newprot,
1435 int dirty_accountable, int prot_numa);
1436 extern int mprotect_fixup(struct vm_area_struct *vma,
1437 struct vm_area_struct **pprev, unsigned long start,
1438 unsigned long end, unsigned long newflags);
1441 * doesn't attempt to fault and will return short.
1443 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1444 struct page **pages);
1446 * per-process(per-mm_struct) statistics.
1448 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1450 long val = atomic_long_read(&mm->rss_stat.count[member]);
1452 #ifdef SPLIT_RSS_COUNTING
1454 * counter is updated in asynchronous manner and may go to minus.
1455 * But it's never be expected number for users.
1460 return (unsigned long)val;
1463 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1465 atomic_long_add(value, &mm->rss_stat.count[member]);
1468 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1470 atomic_long_inc(&mm->rss_stat.count[member]);
1473 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1475 atomic_long_dec(&mm->rss_stat.count[member]);
1478 /* Optimized variant when page is already known not to be PageAnon */
1479 static inline int mm_counter_file(struct page *page)
1481 if (PageSwapBacked(page))
1482 return MM_SHMEMPAGES;
1483 return MM_FILEPAGES;
1486 static inline int mm_counter(struct page *page)
1489 return MM_ANONPAGES;
1490 return mm_counter_file(page);
1493 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1495 return get_mm_counter(mm, MM_FILEPAGES) +
1496 get_mm_counter(mm, MM_ANONPAGES) +
1497 get_mm_counter(mm, MM_SHMEMPAGES);
1500 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1502 return max(mm->hiwater_rss, get_mm_rss(mm));
1505 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1507 return max(mm->hiwater_vm, mm->total_vm);
1510 static inline void update_hiwater_rss(struct mm_struct *mm)
1512 unsigned long _rss = get_mm_rss(mm);
1514 if ((mm)->hiwater_rss < _rss)
1515 (mm)->hiwater_rss = _rss;
1518 static inline void update_hiwater_vm(struct mm_struct *mm)
1520 if (mm->hiwater_vm < mm->total_vm)
1521 mm->hiwater_vm = mm->total_vm;
1524 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1526 mm->hiwater_rss = get_mm_rss(mm);
1529 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1530 struct mm_struct *mm)
1532 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1534 if (*maxrss < hiwater_rss)
1535 *maxrss = hiwater_rss;
1538 #if defined(SPLIT_RSS_COUNTING)
1539 void sync_mm_rss(struct mm_struct *mm);
1541 static inline void sync_mm_rss(struct mm_struct *mm)
1546 #ifndef __HAVE_ARCH_PTE_DEVMAP
1547 static inline int pte_devmap(pte_t pte)
1553 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1555 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1557 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1561 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1565 #ifdef __PAGETABLE_PUD_FOLDED
1566 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1567 unsigned long address)
1572 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1575 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1576 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1577 unsigned long address)
1582 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1584 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1589 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1590 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1593 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1595 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1597 atomic_long_set(&mm->nr_pmds, 0);
1600 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1602 return atomic_long_read(&mm->nr_pmds);
1605 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1607 atomic_long_inc(&mm->nr_pmds);
1610 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1612 atomic_long_dec(&mm->nr_pmds);
1616 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1617 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1620 * The following ifdef needed to get the 4level-fixup.h header to work.
1621 * Remove it when 4level-fixup.h has been removed.
1623 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1624 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1626 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1627 NULL: pud_offset(pgd, address);
1630 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1632 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1633 NULL: pmd_offset(pud, address);
1635 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1637 #if USE_SPLIT_PTE_PTLOCKS
1638 #if ALLOC_SPLIT_PTLOCKS
1639 void __init ptlock_cache_init(void);
1640 extern bool ptlock_alloc(struct page *page);
1641 extern void ptlock_free(struct page *page);
1643 static inline spinlock_t *ptlock_ptr(struct page *page)
1647 #else /* ALLOC_SPLIT_PTLOCKS */
1648 static inline void ptlock_cache_init(void)
1652 static inline bool ptlock_alloc(struct page *page)
1657 static inline void ptlock_free(struct page *page)
1661 static inline spinlock_t *ptlock_ptr(struct page *page)
1665 #endif /* ALLOC_SPLIT_PTLOCKS */
1667 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1669 return ptlock_ptr(pmd_page(*pmd));
1672 static inline bool ptlock_init(struct page *page)
1675 * prep_new_page() initialize page->private (and therefore page->ptl)
1676 * with 0. Make sure nobody took it in use in between.
1678 * It can happen if arch try to use slab for page table allocation:
1679 * slab code uses page->slab_cache, which share storage with page->ptl.
1681 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1682 if (!ptlock_alloc(page))
1684 spin_lock_init(ptlock_ptr(page));
1688 /* Reset page->mapping so free_pages_check won't complain. */
1689 static inline void pte_lock_deinit(struct page *page)
1691 page->mapping = NULL;
1695 #else /* !USE_SPLIT_PTE_PTLOCKS */
1697 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1699 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1701 return &mm->page_table_lock;
1703 static inline void ptlock_cache_init(void) {}
1704 static inline bool ptlock_init(struct page *page) { return true; }
1705 static inline void pte_lock_deinit(struct page *page) {}
1706 #endif /* USE_SPLIT_PTE_PTLOCKS */
1708 static inline void pgtable_init(void)
1710 ptlock_cache_init();
1711 pgtable_cache_init();
1714 static inline bool pgtable_page_ctor(struct page *page)
1716 if (!ptlock_init(page))
1718 inc_zone_page_state(page, NR_PAGETABLE);
1722 static inline void pgtable_page_dtor(struct page *page)
1724 pte_lock_deinit(page);
1725 dec_zone_page_state(page, NR_PAGETABLE);
1728 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1730 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1731 pte_t *__pte = pte_offset_map(pmd, address); \
1737 #define pte_unmap_unlock(pte, ptl) do { \
1742 #define pte_alloc(mm, pmd, address) \
1743 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1745 #define pte_alloc_map(mm, pmd, address) \
1746 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1748 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1749 (pte_alloc(mm, pmd, address) ? \
1750 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1752 #define pte_alloc_kernel(pmd, address) \
1753 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1754 NULL: pte_offset_kernel(pmd, address))
1756 #if USE_SPLIT_PMD_PTLOCKS
1758 static struct page *pmd_to_page(pmd_t *pmd)
1760 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1761 return virt_to_page((void *)((unsigned long) pmd & mask));
1764 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1766 return ptlock_ptr(pmd_to_page(pmd));
1769 static inline bool pgtable_pmd_page_ctor(struct page *page)
1771 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1772 page->pmd_huge_pte = NULL;
1774 return ptlock_init(page);
1777 static inline void pgtable_pmd_page_dtor(struct page *page)
1779 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1780 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1785 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1789 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1791 return &mm->page_table_lock;
1794 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1795 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1797 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1801 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1803 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1808 extern void free_area_init(unsigned long * zones_size);
1809 extern void free_area_init_node(int nid, unsigned long * zones_size,
1810 unsigned long zone_start_pfn, unsigned long *zholes_size);
1811 extern void free_initmem(void);
1814 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1815 * into the buddy system. The freed pages will be poisoned with pattern
1816 * "poison" if it's within range [0, UCHAR_MAX].
1817 * Return pages freed into the buddy system.
1819 extern unsigned long free_reserved_area(void *start, void *end,
1820 int poison, char *s);
1822 #ifdef CONFIG_HIGHMEM
1824 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1825 * and totalram_pages.
1827 extern void free_highmem_page(struct page *page);
1830 extern void adjust_managed_page_count(struct page *page, long count);
1831 extern void mem_init_print_info(const char *str);
1833 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1835 /* Free the reserved page into the buddy system, so it gets managed. */
1836 static inline void __free_reserved_page(struct page *page)
1838 ClearPageReserved(page);
1839 init_page_count(page);
1843 static inline void free_reserved_page(struct page *page)
1845 __free_reserved_page(page);
1846 adjust_managed_page_count(page, 1);
1849 static inline void mark_page_reserved(struct page *page)
1851 SetPageReserved(page);
1852 adjust_managed_page_count(page, -1);
1856 * Default method to free all the __init memory into the buddy system.
1857 * The freed pages will be poisoned with pattern "poison" if it's within
1858 * range [0, UCHAR_MAX].
1859 * Return pages freed into the buddy system.
1861 static inline unsigned long free_initmem_default(int poison)
1863 extern char __init_begin[], __init_end[];
1865 return free_reserved_area(&__init_begin, &__init_end,
1866 poison, "unused kernel");
1869 static inline unsigned long get_num_physpages(void)
1872 unsigned long phys_pages = 0;
1874 for_each_online_node(nid)
1875 phys_pages += node_present_pages(nid);
1880 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1882 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1883 * zones, allocate the backing mem_map and account for memory holes in a more
1884 * architecture independent manner. This is a substitute for creating the
1885 * zone_sizes[] and zholes_size[] arrays and passing them to
1886 * free_area_init_node()
1888 * An architecture is expected to register range of page frames backed by
1889 * physical memory with memblock_add[_node]() before calling
1890 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1891 * usage, an architecture is expected to do something like
1893 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1895 * for_each_valid_physical_page_range()
1896 * memblock_add_node(base, size, nid)
1897 * free_area_init_nodes(max_zone_pfns);
1899 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1900 * registered physical page range. Similarly
1901 * sparse_memory_present_with_active_regions() calls memory_present() for
1902 * each range when SPARSEMEM is enabled.
1904 * See mm/page_alloc.c for more information on each function exposed by
1905 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1907 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1908 unsigned long node_map_pfn_alignment(void);
1909 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1910 unsigned long end_pfn);
1911 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1912 unsigned long end_pfn);
1913 extern void get_pfn_range_for_nid(unsigned int nid,
1914 unsigned long *start_pfn, unsigned long *end_pfn);
1915 extern unsigned long find_min_pfn_with_active_regions(void);
1916 extern void free_bootmem_with_active_regions(int nid,
1917 unsigned long max_low_pfn);
1918 extern void sparse_memory_present_with_active_regions(int nid);
1920 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1922 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1923 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1924 static inline int __early_pfn_to_nid(unsigned long pfn,
1925 struct mminit_pfnnid_cache *state)
1930 /* please see mm/page_alloc.c */
1931 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1932 /* there is a per-arch backend function. */
1933 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1934 struct mminit_pfnnid_cache *state);
1937 extern void set_dma_reserve(unsigned long new_dma_reserve);
1938 extern void memmap_init_zone(unsigned long, int, unsigned long,
1939 unsigned long, enum memmap_context);
1940 extern void setup_per_zone_wmarks(void);
1941 extern int __meminit init_per_zone_wmark_min(void);
1942 extern void mem_init(void);
1943 extern void __init mmap_init(void);
1944 extern void show_mem(unsigned int flags);
1945 extern long si_mem_available(void);
1946 extern void si_meminfo(struct sysinfo * val);
1947 extern void si_meminfo_node(struct sysinfo *val, int nid);
1948 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1949 extern unsigned long arch_reserved_kernel_pages(void);
1952 extern __printf(2, 3)
1953 void warn_alloc(gfp_t gfp_mask, const char *fmt, ...);
1955 extern void setup_per_cpu_pageset(void);
1957 extern void zone_pcp_update(struct zone *zone);
1958 extern void zone_pcp_reset(struct zone *zone);
1961 extern int min_free_kbytes;
1962 extern int watermark_scale_factor;
1965 extern atomic_long_t mmap_pages_allocated;
1966 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1968 /* interval_tree.c */
1969 void vma_interval_tree_insert(struct vm_area_struct *node,
1970 struct rb_root *root);
1971 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1972 struct vm_area_struct *prev,
1973 struct rb_root *root);
1974 void vma_interval_tree_remove(struct vm_area_struct *node,
1975 struct rb_root *root);
1976 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1977 unsigned long start, unsigned long last);
1978 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1979 unsigned long start, unsigned long last);
1981 #define vma_interval_tree_foreach(vma, root, start, last) \
1982 for (vma = vma_interval_tree_iter_first(root, start, last); \
1983 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1985 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1986 struct rb_root *root);
1987 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1988 struct rb_root *root);
1989 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1990 struct rb_root *root, unsigned long start, unsigned long last);
1991 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1992 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1993 #ifdef CONFIG_DEBUG_VM_RB
1994 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1997 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1998 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1999 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2002 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2003 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2004 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2005 struct vm_area_struct *expand);
2006 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2007 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2009 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2011 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2012 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2013 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2014 struct mempolicy *, struct vm_userfaultfd_ctx);
2015 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2016 extern int split_vma(struct mm_struct *,
2017 struct vm_area_struct *, unsigned long addr, int new_below);
2018 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2019 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2020 struct rb_node **, struct rb_node *);
2021 extern void unlink_file_vma(struct vm_area_struct *);
2022 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2023 unsigned long addr, unsigned long len, pgoff_t pgoff,
2024 bool *need_rmap_locks);
2025 extern void exit_mmap(struct mm_struct *);
2027 static inline int check_data_rlimit(unsigned long rlim,
2029 unsigned long start,
2030 unsigned long end_data,
2031 unsigned long start_data)
2033 if (rlim < RLIM_INFINITY) {
2034 if (((new - start) + (end_data - start_data)) > rlim)
2041 extern int mm_take_all_locks(struct mm_struct *mm);
2042 extern void mm_drop_all_locks(struct mm_struct *mm);
2044 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2045 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2046 extern struct file *get_task_exe_file(struct task_struct *task);
2048 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2049 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2051 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2052 const struct vm_special_mapping *sm);
2053 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2054 unsigned long addr, unsigned long len,
2055 unsigned long flags,
2056 const struct vm_special_mapping *spec);
2057 /* This is an obsolete alternative to _install_special_mapping. */
2058 extern int install_special_mapping(struct mm_struct *mm,
2059 unsigned long addr, unsigned long len,
2060 unsigned long flags, struct page **pages);
2062 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2064 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2065 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2066 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2067 unsigned long len, unsigned long prot, unsigned long flags,
2068 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2069 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2071 static inline unsigned long
2072 do_mmap_pgoff(struct file *file, unsigned long addr,
2073 unsigned long len, unsigned long prot, unsigned long flags,
2074 unsigned long pgoff, unsigned long *populate)
2076 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2080 extern int __mm_populate(unsigned long addr, unsigned long len,
2082 static inline void mm_populate(unsigned long addr, unsigned long len)
2085 (void) __mm_populate(addr, len, 1);
2088 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2091 /* These take the mm semaphore themselves */
2092 extern int __must_check vm_brk(unsigned long, unsigned long);
2093 extern int vm_munmap(unsigned long, size_t);
2094 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2095 unsigned long, unsigned long,
2096 unsigned long, unsigned long);
2098 struct vm_unmapped_area_info {
2099 #define VM_UNMAPPED_AREA_TOPDOWN 1
2100 unsigned long flags;
2101 unsigned long length;
2102 unsigned long low_limit;
2103 unsigned long high_limit;
2104 unsigned long align_mask;
2105 unsigned long align_offset;
2108 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2109 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2112 * Search for an unmapped address range.
2114 * We are looking for a range that:
2115 * - does not intersect with any VMA;
2116 * - is contained within the [low_limit, high_limit) interval;
2117 * - is at least the desired size.
2118 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2120 static inline unsigned long
2121 vm_unmapped_area(struct vm_unmapped_area_info *info)
2123 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2124 return unmapped_area_topdown(info);
2126 return unmapped_area(info);
2130 extern void truncate_inode_pages(struct address_space *, loff_t);
2131 extern void truncate_inode_pages_range(struct address_space *,
2132 loff_t lstart, loff_t lend);
2133 extern void truncate_inode_pages_final(struct address_space *);
2135 /* generic vm_area_ops exported for stackable file systems */
2136 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2137 extern void filemap_map_pages(struct fault_env *fe,
2138 pgoff_t start_pgoff, pgoff_t end_pgoff);
2139 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2141 /* mm/page-writeback.c */
2142 int write_one_page(struct page *page, int wait);
2143 void task_dirty_inc(struct task_struct *tsk);
2146 #define VM_MAX_READAHEAD 128 /* kbytes */
2147 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2149 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2150 pgoff_t offset, unsigned long nr_to_read);
2152 void page_cache_sync_readahead(struct address_space *mapping,
2153 struct file_ra_state *ra,
2156 unsigned long size);
2158 void page_cache_async_readahead(struct address_space *mapping,
2159 struct file_ra_state *ra,
2163 unsigned long size);
2165 extern unsigned long stack_guard_gap;
2166 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2167 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2169 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2170 extern int expand_downwards(struct vm_area_struct *vma,
2171 unsigned long address);
2173 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2175 #define expand_upwards(vma, address) (0)
2178 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2179 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2180 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2181 struct vm_area_struct **pprev);
2183 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2184 NULL if none. Assume start_addr < end_addr. */
2185 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2187 struct vm_area_struct * vma = find_vma(mm,start_addr);
2189 if (vma && end_addr <= vma->vm_start)
2194 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2196 unsigned long vm_start = vma->vm_start;
2198 if (vma->vm_flags & VM_GROWSDOWN) {
2199 vm_start -= stack_guard_gap;
2200 if (vm_start > vma->vm_start)
2206 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2208 unsigned long vm_end = vma->vm_end;
2210 if (vma->vm_flags & VM_GROWSUP) {
2211 vm_end += stack_guard_gap;
2212 if (vm_end < vma->vm_end)
2213 vm_end = -PAGE_SIZE;
2218 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2220 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2223 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2224 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2225 unsigned long vm_start, unsigned long vm_end)
2227 struct vm_area_struct *vma = find_vma(mm, vm_start);
2229 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2235 static inline bool range_in_vma(struct vm_area_struct *vma,
2236 unsigned long start, unsigned long end)
2238 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2242 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2243 void vma_set_page_prot(struct vm_area_struct *vma);
2245 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2249 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2251 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2255 #ifdef CONFIG_NUMA_BALANCING
2256 unsigned long change_prot_numa(struct vm_area_struct *vma,
2257 unsigned long start, unsigned long end);
2260 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2261 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2262 unsigned long pfn, unsigned long size, pgprot_t);
2263 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2264 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2266 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2267 unsigned long pfn, pgprot_t pgprot);
2268 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2270 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2273 struct page *follow_page_mask(struct vm_area_struct *vma,
2274 unsigned long address, unsigned int foll_flags,
2275 unsigned int *page_mask);
2277 static inline struct page *follow_page(struct vm_area_struct *vma,
2278 unsigned long address, unsigned int foll_flags)
2280 unsigned int unused_page_mask;
2281 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2284 #define FOLL_WRITE 0x01 /* check pte is writable */
2285 #define FOLL_TOUCH 0x02 /* mark page accessed */
2286 #define FOLL_GET 0x04 /* do get_page on page */
2287 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2288 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2289 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2290 * and return without waiting upon it */
2291 #define FOLL_POPULATE 0x40 /* fault in page */
2292 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2293 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2294 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2295 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2296 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2297 #define FOLL_MLOCK 0x1000 /* lock present pages */
2298 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2299 #define FOLL_COW 0x4000 /* internal GUP flag */
2300 #define FOLL_ANON 0x8000 /* don't do file mappings */
2302 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2304 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2305 unsigned long size, pte_fn_t fn, void *data);
2308 #ifdef CONFIG_PAGE_POISONING
2309 extern bool page_poisoning_enabled(void);
2310 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2311 extern bool page_is_poisoned(struct page *page);
2313 static inline bool page_poisoning_enabled(void) { return false; }
2314 static inline void kernel_poison_pages(struct page *page, int numpages,
2316 static inline bool page_is_poisoned(struct page *page) { return false; }
2319 #ifdef CONFIG_DEBUG_PAGEALLOC
2320 extern bool _debug_pagealloc_enabled;
2321 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2323 static inline bool debug_pagealloc_enabled(void)
2325 return _debug_pagealloc_enabled;
2329 kernel_map_pages(struct page *page, int numpages, int enable)
2331 if (!debug_pagealloc_enabled())
2334 __kernel_map_pages(page, numpages, enable);
2336 #ifdef CONFIG_HIBERNATION
2337 extern bool kernel_page_present(struct page *page);
2338 #endif /* CONFIG_HIBERNATION */
2339 #else /* CONFIG_DEBUG_PAGEALLOC */
2341 kernel_map_pages(struct page *page, int numpages, int enable) {}
2342 #ifdef CONFIG_HIBERNATION
2343 static inline bool kernel_page_present(struct page *page) { return true; }
2344 #endif /* CONFIG_HIBERNATION */
2345 static inline bool debug_pagealloc_enabled(void)
2349 #endif /* CONFIG_DEBUG_PAGEALLOC */
2351 #ifdef __HAVE_ARCH_GATE_AREA
2352 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2353 extern int in_gate_area_no_mm(unsigned long addr);
2354 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2356 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2360 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2361 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2365 #endif /* __HAVE_ARCH_GATE_AREA */
2367 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2369 #ifdef CONFIG_SYSCTL
2370 extern int sysctl_drop_caches;
2371 int drop_caches_sysctl_handler(struct ctl_table *, int,
2372 void __user *, size_t *, loff_t *);
2375 void drop_slab(void);
2376 void drop_slab_node(int nid);
2379 #define randomize_va_space 0
2381 extern int randomize_va_space;
2384 const char * arch_vma_name(struct vm_area_struct *vma);
2385 void print_vma_addr(char *prefix, unsigned long rip);
2387 void sparse_mem_maps_populate_node(struct page **map_map,
2388 unsigned long pnum_begin,
2389 unsigned long pnum_end,
2390 unsigned long map_count,
2393 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2394 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2395 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2396 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2397 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2398 void *vmemmap_alloc_block(unsigned long size, int node);
2400 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2401 struct vmem_altmap *altmap);
2402 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2404 return __vmemmap_alloc_block_buf(size, node, NULL);
2407 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2408 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2410 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2411 void vmemmap_populate_print_last(void);
2412 #ifdef CONFIG_MEMORY_HOTPLUG
2413 void vmemmap_free(unsigned long start, unsigned long end);
2415 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2416 unsigned long size);
2419 MF_COUNT_INCREASED = 1 << 0,
2420 MF_ACTION_REQUIRED = 1 << 1,
2421 MF_MUST_KILL = 1 << 2,
2422 MF_SOFT_OFFLINE = 1 << 3,
2424 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2425 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2426 extern int unpoison_memory(unsigned long pfn);
2427 extern int get_hwpoison_page(struct page *page);
2428 #define put_hwpoison_page(page) put_page(page)
2429 extern int sysctl_memory_failure_early_kill;
2430 extern int sysctl_memory_failure_recovery;
2431 extern void shake_page(struct page *p, int access);
2432 extern atomic_long_t num_poisoned_pages;
2433 extern int soft_offline_page(struct page *page, int flags);
2437 * Error handlers for various types of pages.
2440 MF_IGNORED, /* Error: cannot be handled */
2441 MF_FAILED, /* Error: handling failed */
2442 MF_DELAYED, /* Will be handled later */
2443 MF_RECOVERED, /* Successfully recovered */
2446 enum mf_action_page_type {
2448 MF_MSG_KERNEL_HIGH_ORDER,
2450 MF_MSG_DIFFERENT_COMPOUND,
2451 MF_MSG_POISONED_HUGE,
2454 MF_MSG_UNMAP_FAILED,
2455 MF_MSG_DIRTY_SWAPCACHE,
2456 MF_MSG_CLEAN_SWAPCACHE,
2457 MF_MSG_DIRTY_MLOCKED_LRU,
2458 MF_MSG_CLEAN_MLOCKED_LRU,
2459 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2460 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2463 MF_MSG_TRUNCATED_LRU,
2469 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2470 extern void clear_huge_page(struct page *page,
2472 unsigned int pages_per_huge_page);
2473 extern void copy_user_huge_page(struct page *dst, struct page *src,
2474 unsigned long addr, struct vm_area_struct *vma,
2475 unsigned int pages_per_huge_page);
2476 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2478 extern struct page_ext_operations debug_guardpage_ops;
2479 extern struct page_ext_operations page_poisoning_ops;
2481 #ifdef CONFIG_DEBUG_PAGEALLOC
2482 extern unsigned int _debug_guardpage_minorder;
2483 extern bool _debug_guardpage_enabled;
2485 static inline unsigned int debug_guardpage_minorder(void)
2487 return _debug_guardpage_minorder;
2490 static inline bool debug_guardpage_enabled(void)
2492 return _debug_guardpage_enabled;
2495 static inline bool page_is_guard(struct page *page)
2497 struct page_ext *page_ext;
2499 if (!debug_guardpage_enabled())
2502 page_ext = lookup_page_ext(page);
2503 if (unlikely(!page_ext))
2506 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2509 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2510 static inline bool debug_guardpage_enabled(void) { return false; }
2511 static inline bool page_is_guard(struct page *page) { return false; }
2512 #endif /* CONFIG_DEBUG_PAGEALLOC */
2514 #if MAX_NUMNODES > 1
2515 void __init setup_nr_node_ids(void);
2517 static inline void setup_nr_node_ids(void) {}
2520 #endif /* __KERNEL__ */
2521 #endif /* _LINUX_MM_H */