4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/bit_spinlock.h>
19 #include <linux/shrinker.h>
23 struct anon_vma_chain;
26 struct writeback_control;
28 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
32 extern unsigned long num_physpages;
33 extern unsigned long totalram_pages;
34 extern void * high_memory;
35 extern int page_cluster;
38 extern int sysctl_legacy_va_layout;
40 #define sysctl_legacy_va_layout 0
44 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
49 /* to align the pointer to the (next) page boundary */
50 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
53 * Linux kernel virtual memory manager primitives.
54 * The idea being to have a "virtual" mm in the same way
55 * we have a virtual fs - giving a cleaner interface to the
56 * mm details, and allowing different kinds of memory mappings
57 * (from shared memory to executable loading to arbitrary
61 extern struct kmem_cache *vm_area_cachep;
64 extern struct rb_root nommu_region_tree;
65 extern struct rw_semaphore nommu_region_sem;
67 extern unsigned int kobjsize(const void *objp);
71 * vm_flags in vm_area_struct, see mm_types.h.
73 #define VM_NONE 0x00000000
75 #define VM_READ 0x00000001 /* currently active flags */
76 #define VM_WRITE 0x00000002
77 #define VM_EXEC 0x00000004
78 #define VM_SHARED 0x00000008
80 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
81 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82 #define VM_MAYWRITE 0x00000020
83 #define VM_MAYEXEC 0x00000040
84 #define VM_MAYSHARE 0x00000080
86 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
88 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90 #define VM_LOCKED 0x00002000
91 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
93 /* Used by sys_madvise() */
94 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
95 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
97 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
98 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
99 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
100 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
101 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
102 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
103 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
104 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
106 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
107 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
108 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
109 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
111 #if defined(CONFIG_X86)
112 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
113 #elif defined(CONFIG_PPC)
114 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
115 #elif defined(CONFIG_PARISC)
116 # define VM_GROWSUP VM_ARCH_1
117 #elif defined(CONFIG_IA64)
118 # define VM_GROWSUP VM_ARCH_1
119 #elif !defined(CONFIG_MMU)
120 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
124 # define VM_GROWSUP VM_NONE
127 /* Bits set in the VMA until the stack is in its final location */
128 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
130 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
131 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
134 #ifdef CONFIG_STACK_GROWSUP
135 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
137 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
140 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
141 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
142 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
143 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
144 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
147 * Special vmas that are non-mergable, non-mlock()able.
148 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
150 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
153 * mapping from the currently active vm_flags protection bits (the
154 * low four bits) to a page protection mask..
156 extern pgprot_t protection_map[16];
158 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
159 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
160 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
161 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
162 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
163 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
166 * vm_fault is filled by the the pagefault handler and passed to the vma's
167 * ->fault function. The vma's ->fault is responsible for returning a bitmask
168 * of VM_FAULT_xxx flags that give details about how the fault was handled.
170 * pgoff should be used in favour of virtual_address, if possible. If pgoff
171 * is used, one may implement ->remap_pages to get nonlinear mapping support.
174 unsigned int flags; /* FAULT_FLAG_xxx flags */
175 pgoff_t pgoff; /* Logical page offset based on vma */
176 void __user *virtual_address; /* Faulting virtual address */
178 struct page *page; /* ->fault handlers should return a
179 * page here, unless VM_FAULT_NOPAGE
180 * is set (which is also implied by
186 * These are the virtual MM functions - opening of an area, closing and
187 * unmapping it (needed to keep files on disk up-to-date etc), pointer
188 * to the functions called when a no-page or a wp-page exception occurs.
190 struct vm_operations_struct {
191 void (*open)(struct vm_area_struct * area);
192 void (*close)(struct vm_area_struct * area);
193 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
195 /* notification that a previously read-only page is about to become
196 * writable, if an error is returned it will cause a SIGBUS */
197 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
199 /* called by access_process_vm when get_user_pages() fails, typically
200 * for use by special VMAs that can switch between memory and hardware
202 int (*access)(struct vm_area_struct *vma, unsigned long addr,
203 void *buf, int len, int write);
206 * set_policy() op must add a reference to any non-NULL @new mempolicy
207 * to hold the policy upon return. Caller should pass NULL @new to
208 * remove a policy and fall back to surrounding context--i.e. do not
209 * install a MPOL_DEFAULT policy, nor the task or system default
212 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
215 * get_policy() op must add reference [mpol_get()] to any policy at
216 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
217 * in mm/mempolicy.c will do this automatically.
218 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
219 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
220 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
221 * must return NULL--i.e., do not "fallback" to task or system default
224 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
226 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
227 const nodemask_t *to, unsigned long flags);
229 /* called by sys_remap_file_pages() to populate non-linear mapping */
230 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
231 unsigned long size, pgoff_t pgoff);
237 #define page_private(page) ((page)->private)
238 #define set_page_private(page, v) ((page)->private = (v))
240 /* It's valid only if the page is free path or free_list */
241 static inline void set_freepage_migratetype(struct page *page, int migratetype)
243 set_page_private(page, migratetype);
246 /* It's valid only if the page is free path or free_list */
247 static inline int get_freepage_migratetype(struct page *page)
249 return page_private(page);
253 * FIXME: take this include out, include page-flags.h in
254 * files which need it (119 of them)
256 #include <linux/page-flags.h>
257 #include <linux/huge_mm.h>
260 * Methods to modify the page usage count.
262 * What counts for a page usage:
263 * - cache mapping (page->mapping)
264 * - private data (page->private)
265 * - page mapped in a task's page tables, each mapping
266 * is counted separately
268 * Also, many kernel routines increase the page count before a critical
269 * routine so they can be sure the page doesn't go away from under them.
273 * Drop a ref, return true if the refcount fell to zero (the page has no users)
275 static inline int put_page_testzero(struct page *page)
277 VM_BUG_ON(atomic_read(&page->_count) == 0);
278 return atomic_dec_and_test(&page->_count);
282 * Try to grab a ref unless the page has a refcount of zero, return false if
285 static inline int get_page_unless_zero(struct page *page)
287 return atomic_inc_not_zero(&page->_count);
290 extern int page_is_ram(unsigned long pfn);
292 /* Support for virtually mapped pages */
293 struct page *vmalloc_to_page(const void *addr);
294 unsigned long vmalloc_to_pfn(const void *addr);
297 * Determine if an address is within the vmalloc range
299 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
300 * is no special casing required.
302 static inline int is_vmalloc_addr(const void *x)
305 unsigned long addr = (unsigned long)x;
307 return addr >= VMALLOC_START && addr < VMALLOC_END;
313 extern int is_vmalloc_or_module_addr(const void *x);
315 static inline int is_vmalloc_or_module_addr(const void *x)
321 static inline void compound_lock(struct page *page)
323 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
324 VM_BUG_ON(PageSlab(page));
325 bit_spin_lock(PG_compound_lock, &page->flags);
329 static inline void compound_unlock(struct page *page)
331 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
332 VM_BUG_ON(PageSlab(page));
333 bit_spin_unlock(PG_compound_lock, &page->flags);
337 static inline unsigned long compound_lock_irqsave(struct page *page)
339 unsigned long uninitialized_var(flags);
340 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
341 local_irq_save(flags);
347 static inline void compound_unlock_irqrestore(struct page *page,
350 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
351 compound_unlock(page);
352 local_irq_restore(flags);
356 static inline struct page *compound_head(struct page *page)
358 if (unlikely(PageTail(page)))
359 return page->first_page;
364 * The atomic page->_mapcount, starts from -1: so that transitions
365 * both from it and to it can be tracked, using atomic_inc_and_test
366 * and atomic_add_negative(-1).
368 static inline void reset_page_mapcount(struct page *page)
370 atomic_set(&(page)->_mapcount, -1);
373 static inline int page_mapcount(struct page *page)
375 return atomic_read(&(page)->_mapcount) + 1;
378 static inline int page_count(struct page *page)
380 return atomic_read(&compound_head(page)->_count);
383 static inline void get_huge_page_tail(struct page *page)
386 * __split_huge_page_refcount() cannot run
389 VM_BUG_ON(page_mapcount(page) < 0);
390 VM_BUG_ON(atomic_read(&page->_count) != 0);
391 atomic_inc(&page->_mapcount);
394 extern bool __get_page_tail(struct page *page);
396 static inline void get_page(struct page *page)
398 if (unlikely(PageTail(page)))
399 if (likely(__get_page_tail(page)))
402 * Getting a normal page or the head of a compound page
403 * requires to already have an elevated page->_count.
405 VM_BUG_ON(atomic_read(&page->_count) <= 0);
406 atomic_inc(&page->_count);
409 static inline struct page *virt_to_head_page(const void *x)
411 struct page *page = virt_to_page(x);
412 return compound_head(page);
416 * Setup the page count before being freed into the page allocator for
417 * the first time (boot or memory hotplug)
419 static inline void init_page_count(struct page *page)
421 atomic_set(&page->_count, 1);
425 * PageBuddy() indicate that the page is free and in the buddy system
426 * (see mm/page_alloc.c).
428 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
429 * -2 so that an underflow of the page_mapcount() won't be mistaken
430 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
431 * efficiently by most CPU architectures.
433 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
435 static inline int PageBuddy(struct page *page)
437 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
440 static inline void __SetPageBuddy(struct page *page)
442 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
443 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
446 static inline void __ClearPageBuddy(struct page *page)
448 VM_BUG_ON(!PageBuddy(page));
449 atomic_set(&page->_mapcount, -1);
452 void put_page(struct page *page);
453 void put_pages_list(struct list_head *pages);
455 void split_page(struct page *page, unsigned int order);
456 int split_free_page(struct page *page);
457 int capture_free_page(struct page *page, int alloc_order, int migratetype);
460 * Compound pages have a destructor function. Provide a
461 * prototype for that function and accessor functions.
462 * These are _only_ valid on the head of a PG_compound page.
464 typedef void compound_page_dtor(struct page *);
466 static inline void set_compound_page_dtor(struct page *page,
467 compound_page_dtor *dtor)
469 page[1].lru.next = (void *)dtor;
472 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
474 return (compound_page_dtor *)page[1].lru.next;
477 static inline int compound_order(struct page *page)
481 return (unsigned long)page[1].lru.prev;
484 static inline int compound_trans_order(struct page *page)
492 flags = compound_lock_irqsave(page);
493 order = compound_order(page);
494 compound_unlock_irqrestore(page, flags);
498 static inline void set_compound_order(struct page *page, unsigned long order)
500 page[1].lru.prev = (void *)order;
505 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
506 * servicing faults for write access. In the normal case, do always want
507 * pte_mkwrite. But get_user_pages can cause write faults for mappings
508 * that do not have writing enabled, when used by access_process_vm.
510 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
512 if (likely(vma->vm_flags & VM_WRITE))
513 pte = pte_mkwrite(pte);
519 * Multiple processes may "see" the same page. E.g. for untouched
520 * mappings of /dev/null, all processes see the same page full of
521 * zeroes, and text pages of executables and shared libraries have
522 * only one copy in memory, at most, normally.
524 * For the non-reserved pages, page_count(page) denotes a reference count.
525 * page_count() == 0 means the page is free. page->lru is then used for
526 * freelist management in the buddy allocator.
527 * page_count() > 0 means the page has been allocated.
529 * Pages are allocated by the slab allocator in order to provide memory
530 * to kmalloc and kmem_cache_alloc. In this case, the management of the
531 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
532 * unless a particular usage is carefully commented. (the responsibility of
533 * freeing the kmalloc memory is the caller's, of course).
535 * A page may be used by anyone else who does a __get_free_page().
536 * In this case, page_count still tracks the references, and should only
537 * be used through the normal accessor functions. The top bits of page->flags
538 * and page->virtual store page management information, but all other fields
539 * are unused and could be used privately, carefully. The management of this
540 * page is the responsibility of the one who allocated it, and those who have
541 * subsequently been given references to it.
543 * The other pages (we may call them "pagecache pages") are completely
544 * managed by the Linux memory manager: I/O, buffers, swapping etc.
545 * The following discussion applies only to them.
547 * A pagecache page contains an opaque `private' member, which belongs to the
548 * page's address_space. Usually, this is the address of a circular list of
549 * the page's disk buffers. PG_private must be set to tell the VM to call
550 * into the filesystem to release these pages.
552 * A page may belong to an inode's memory mapping. In this case, page->mapping
553 * is the pointer to the inode, and page->index is the file offset of the page,
554 * in units of PAGE_CACHE_SIZE.
556 * If pagecache pages are not associated with an inode, they are said to be
557 * anonymous pages. These may become associated with the swapcache, and in that
558 * case PG_swapcache is set, and page->private is an offset into the swapcache.
560 * In either case (swapcache or inode backed), the pagecache itself holds one
561 * reference to the page. Setting PG_private should also increment the
562 * refcount. The each user mapping also has a reference to the page.
564 * The pagecache pages are stored in a per-mapping radix tree, which is
565 * rooted at mapping->page_tree, and indexed by offset.
566 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
567 * lists, we instead now tag pages as dirty/writeback in the radix tree.
569 * All pagecache pages may be subject to I/O:
570 * - inode pages may need to be read from disk,
571 * - inode pages which have been modified and are MAP_SHARED may need
572 * to be written back to the inode on disk,
573 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
574 * modified may need to be swapped out to swap space and (later) to be read
579 * The zone field is never updated after free_area_init_core()
580 * sets it, so none of the operations on it need to be atomic.
585 * page->flags layout:
587 * There are three possibilities for how page->flags get
588 * laid out. The first is for the normal case, without
589 * sparsemem. The second is for sparsemem when there is
590 * plenty of space for node and section. The last is when
591 * we have run out of space and have to fall back to an
592 * alternate (slower) way of determining the node.
594 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
595 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
596 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
598 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
599 #define SECTIONS_WIDTH SECTIONS_SHIFT
601 #define SECTIONS_WIDTH 0
604 #define ZONES_WIDTH ZONES_SHIFT
606 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
607 #define NODES_WIDTH NODES_SHIFT
609 #ifdef CONFIG_SPARSEMEM_VMEMMAP
610 #error "Vmemmap: No space for nodes field in page flags"
612 #define NODES_WIDTH 0
615 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
616 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
617 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
618 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
621 * We are going to use the flags for the page to node mapping if its in
622 * there. This includes the case where there is no node, so it is implicit.
624 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
625 #define NODE_NOT_IN_PAGE_FLAGS
629 * Define the bit shifts to access each section. For non-existent
630 * sections we define the shift as 0; that plus a 0 mask ensures
631 * the compiler will optimise away reference to them.
633 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
634 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
635 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
637 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
638 #ifdef NODE_NOT_IN_PAGE_FLAGS
639 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
640 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
641 SECTIONS_PGOFF : ZONES_PGOFF)
643 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
644 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
645 NODES_PGOFF : ZONES_PGOFF)
648 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
650 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
651 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
654 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
655 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
656 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
657 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
659 static inline enum zone_type page_zonenum(const struct page *page)
661 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
665 * The identification function is only used by the buddy allocator for
666 * determining if two pages could be buddies. We are not really
667 * identifying a zone since we could be using a the section number
668 * id if we have not node id available in page flags.
669 * We guarantee only that it will return the same value for two
670 * combinable pages in a zone.
672 static inline int page_zone_id(struct page *page)
674 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
677 static inline int zone_to_nid(struct zone *zone)
686 #ifdef NODE_NOT_IN_PAGE_FLAGS
687 extern int page_to_nid(const struct page *page);
689 static inline int page_to_nid(const struct page *page)
691 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
695 static inline struct zone *page_zone(const struct page *page)
697 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
700 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
701 static inline void set_page_section(struct page *page, unsigned long section)
703 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
704 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
707 static inline unsigned long page_to_section(const struct page *page)
709 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
713 static inline void set_page_zone(struct page *page, enum zone_type zone)
715 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
716 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
719 static inline void set_page_node(struct page *page, unsigned long node)
721 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
722 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
725 static inline void set_page_links(struct page *page, enum zone_type zone,
726 unsigned long node, unsigned long pfn)
728 set_page_zone(page, zone);
729 set_page_node(page, node);
730 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
731 set_page_section(page, pfn_to_section_nr(pfn));
736 * Some inline functions in vmstat.h depend on page_zone()
738 #include <linux/vmstat.h>
740 static __always_inline void *lowmem_page_address(const struct page *page)
742 return __va(PFN_PHYS(page_to_pfn(page)));
745 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
746 #define HASHED_PAGE_VIRTUAL
749 #if defined(WANT_PAGE_VIRTUAL)
750 #define page_address(page) ((page)->virtual)
751 #define set_page_address(page, address) \
753 (page)->virtual = (address); \
755 #define page_address_init() do { } while(0)
758 #if defined(HASHED_PAGE_VIRTUAL)
759 void *page_address(const struct page *page);
760 void set_page_address(struct page *page, void *virtual);
761 void page_address_init(void);
764 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
765 #define page_address(page) lowmem_page_address(page)
766 #define set_page_address(page, address) do { } while(0)
767 #define page_address_init() do { } while(0)
771 * On an anonymous page mapped into a user virtual memory area,
772 * page->mapping points to its anon_vma, not to a struct address_space;
773 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
775 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
776 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
777 * and then page->mapping points, not to an anon_vma, but to a private
778 * structure which KSM associates with that merged page. See ksm.h.
780 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
782 * Please note that, confusingly, "page_mapping" refers to the inode
783 * address_space which maps the page from disk; whereas "page_mapped"
784 * refers to user virtual address space into which the page is mapped.
786 #define PAGE_MAPPING_ANON 1
787 #define PAGE_MAPPING_KSM 2
788 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
790 extern struct address_space swapper_space;
791 static inline struct address_space *page_mapping(struct page *page)
793 struct address_space *mapping = page->mapping;
795 VM_BUG_ON(PageSlab(page));
796 if (unlikely(PageSwapCache(page)))
797 mapping = &swapper_space;
798 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
803 /* Neutral page->mapping pointer to address_space or anon_vma or other */
804 static inline void *page_rmapping(struct page *page)
806 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
809 extern struct address_space *__page_file_mapping(struct page *);
812 struct address_space *page_file_mapping(struct page *page)
814 if (unlikely(PageSwapCache(page)))
815 return __page_file_mapping(page);
817 return page->mapping;
820 static inline int PageAnon(struct page *page)
822 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
826 * Return the pagecache index of the passed page. Regular pagecache pages
827 * use ->index whereas swapcache pages use ->private
829 static inline pgoff_t page_index(struct page *page)
831 if (unlikely(PageSwapCache(page)))
832 return page_private(page);
836 extern pgoff_t __page_file_index(struct page *page);
839 * Return the file index of the page. Regular pagecache pages use ->index
840 * whereas swapcache pages use swp_offset(->private)
842 static inline pgoff_t page_file_index(struct page *page)
844 if (unlikely(PageSwapCache(page)))
845 return __page_file_index(page);
851 * Return true if this page is mapped into pagetables.
853 static inline int page_mapped(struct page *page)
855 return atomic_read(&(page)->_mapcount) >= 0;
859 * Different kinds of faults, as returned by handle_mm_fault().
860 * Used to decide whether a process gets delivered SIGBUS or
861 * just gets major/minor fault counters bumped up.
864 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
866 #define VM_FAULT_OOM 0x0001
867 #define VM_FAULT_SIGBUS 0x0002
868 #define VM_FAULT_MAJOR 0x0004
869 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
870 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
871 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
873 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
874 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
875 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
877 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
879 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
880 VM_FAULT_HWPOISON_LARGE)
882 /* Encode hstate index for a hwpoisoned large page */
883 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
884 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
887 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
889 extern void pagefault_out_of_memory(void);
891 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
894 * Flags passed to show_mem() and show_free_areas() to suppress output in
897 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
899 extern void show_free_areas(unsigned int flags);
900 extern bool skip_free_areas_node(unsigned int flags, int nid);
902 int shmem_zero_setup(struct vm_area_struct *);
904 extern int can_do_mlock(void);
905 extern int user_shm_lock(size_t, struct user_struct *);
906 extern void user_shm_unlock(size_t, struct user_struct *);
909 * Parameter block passed down to zap_pte_range in exceptional cases.
912 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
913 struct address_space *check_mapping; /* Check page->mapping if set */
914 pgoff_t first_index; /* Lowest page->index to unmap */
915 pgoff_t last_index; /* Highest page->index to unmap */
918 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
921 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
923 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
924 unsigned long size, struct zap_details *);
925 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
926 unsigned long start, unsigned long end);
929 * mm_walk - callbacks for walk_page_range
930 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
931 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
932 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
933 * this handler is required to be able to handle
934 * pmd_trans_huge() pmds. They may simply choose to
935 * split_huge_page() instead of handling it explicitly.
936 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
937 * @pte_hole: if set, called for each hole at all levels
938 * @hugetlb_entry: if set, called for each hugetlb entry
939 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
942 * (see walk_page_range for more details)
945 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
946 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
947 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
948 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
949 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
950 int (*hugetlb_entry)(pte_t *, unsigned long,
951 unsigned long, unsigned long, struct mm_walk *);
952 struct mm_struct *mm;
956 int walk_page_range(unsigned long addr, unsigned long end,
957 struct mm_walk *walk);
958 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
959 unsigned long end, unsigned long floor, unsigned long ceiling);
960 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
961 struct vm_area_struct *vma);
962 void unmap_mapping_range(struct address_space *mapping,
963 loff_t const holebegin, loff_t const holelen, int even_cows);
964 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
966 int follow_phys(struct vm_area_struct *vma, unsigned long address,
967 unsigned int flags, unsigned long *prot, resource_size_t *phys);
968 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
969 void *buf, int len, int write);
971 static inline void unmap_shared_mapping_range(struct address_space *mapping,
972 loff_t const holebegin, loff_t const holelen)
974 unmap_mapping_range(mapping, holebegin, holelen, 0);
977 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
978 extern void truncate_setsize(struct inode *inode, loff_t newsize);
979 extern int vmtruncate(struct inode *inode, loff_t offset);
980 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
981 int truncate_inode_page(struct address_space *mapping, struct page *page);
982 int generic_error_remove_page(struct address_space *mapping, struct page *page);
983 int invalidate_inode_page(struct page *page);
986 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
987 unsigned long address, unsigned int flags);
988 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
989 unsigned long address, unsigned int fault_flags);
991 static inline int handle_mm_fault(struct mm_struct *mm,
992 struct vm_area_struct *vma, unsigned long address,
995 /* should never happen if there's no MMU */
997 return VM_FAULT_SIGBUS;
999 static inline int fixup_user_fault(struct task_struct *tsk,
1000 struct mm_struct *mm, unsigned long address,
1001 unsigned int fault_flags)
1003 /* should never happen if there's no MMU */
1009 extern int make_pages_present(unsigned long addr, unsigned long end);
1010 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1011 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1012 void *buf, int len, int write);
1014 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1015 unsigned long start, int len, unsigned int foll_flags,
1016 struct page **pages, struct vm_area_struct **vmas,
1018 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1019 unsigned long start, int nr_pages, int write, int force,
1020 struct page **pages, struct vm_area_struct **vmas);
1021 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1022 struct page **pages);
1024 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1025 struct page **pages);
1026 int get_kernel_page(unsigned long start, int write, struct page **pages);
1027 struct page *get_dump_page(unsigned long addr);
1029 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1030 extern void do_invalidatepage(struct page *page, unsigned long offset);
1032 int __set_page_dirty_nobuffers(struct page *page);
1033 int __set_page_dirty_no_writeback(struct page *page);
1034 int redirty_page_for_writepage(struct writeback_control *wbc,
1036 void account_page_dirtied(struct page *page, struct address_space *mapping);
1037 void account_page_writeback(struct page *page);
1038 int set_page_dirty(struct page *page);
1039 int set_page_dirty_lock(struct page *page);
1040 int clear_page_dirty_for_io(struct page *page);
1042 /* Is the vma a continuation of the stack vma above it? */
1043 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1045 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1048 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1051 return (vma->vm_flags & VM_GROWSDOWN) &&
1052 (vma->vm_start == addr) &&
1053 !vma_growsdown(vma->vm_prev, addr);
1056 /* Is the vma a continuation of the stack vma below it? */
1057 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1059 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1062 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1065 return (vma->vm_flags & VM_GROWSUP) &&
1066 (vma->vm_end == addr) &&
1067 !vma_growsup(vma->vm_next, addr);
1071 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1073 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1074 unsigned long old_addr, struct vm_area_struct *new_vma,
1075 unsigned long new_addr, unsigned long len,
1076 bool need_rmap_locks);
1077 extern unsigned long do_mremap(unsigned long addr,
1078 unsigned long old_len, unsigned long new_len,
1079 unsigned long flags, unsigned long new_addr);
1080 extern int mprotect_fixup(struct vm_area_struct *vma,
1081 struct vm_area_struct **pprev, unsigned long start,
1082 unsigned long end, unsigned long newflags);
1085 * doesn't attempt to fault and will return short.
1087 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1088 struct page **pages);
1090 * per-process(per-mm_struct) statistics.
1092 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1094 long val = atomic_long_read(&mm->rss_stat.count[member]);
1096 #ifdef SPLIT_RSS_COUNTING
1098 * counter is updated in asynchronous manner and may go to minus.
1099 * But it's never be expected number for users.
1104 return (unsigned long)val;
1107 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1109 atomic_long_add(value, &mm->rss_stat.count[member]);
1112 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1114 atomic_long_inc(&mm->rss_stat.count[member]);
1117 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1119 atomic_long_dec(&mm->rss_stat.count[member]);
1122 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1124 return get_mm_counter(mm, MM_FILEPAGES) +
1125 get_mm_counter(mm, MM_ANONPAGES);
1128 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1130 return max(mm->hiwater_rss, get_mm_rss(mm));
1133 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1135 return max(mm->hiwater_vm, mm->total_vm);
1138 static inline void update_hiwater_rss(struct mm_struct *mm)
1140 unsigned long _rss = get_mm_rss(mm);
1142 if ((mm)->hiwater_rss < _rss)
1143 (mm)->hiwater_rss = _rss;
1146 static inline void update_hiwater_vm(struct mm_struct *mm)
1148 if (mm->hiwater_vm < mm->total_vm)
1149 mm->hiwater_vm = mm->total_vm;
1152 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1153 struct mm_struct *mm)
1155 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1157 if (*maxrss < hiwater_rss)
1158 *maxrss = hiwater_rss;
1161 #if defined(SPLIT_RSS_COUNTING)
1162 void sync_mm_rss(struct mm_struct *mm);
1164 static inline void sync_mm_rss(struct mm_struct *mm)
1169 int vma_wants_writenotify(struct vm_area_struct *vma);
1171 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1173 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1177 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1181 #ifdef __PAGETABLE_PUD_FOLDED
1182 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1183 unsigned long address)
1188 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1191 #ifdef __PAGETABLE_PMD_FOLDED
1192 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1193 unsigned long address)
1198 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1201 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1202 pmd_t *pmd, unsigned long address);
1203 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1206 * The following ifdef needed to get the 4level-fixup.h header to work.
1207 * Remove it when 4level-fixup.h has been removed.
1209 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1210 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1212 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1213 NULL: pud_offset(pgd, address);
1216 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1218 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1219 NULL: pmd_offset(pud, address);
1221 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1223 #if USE_SPLIT_PTLOCKS
1225 * We tuck a spinlock to guard each pagetable page into its struct page,
1226 * at page->private, with BUILD_BUG_ON to make sure that this will not
1227 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1228 * When freeing, reset page->mapping so free_pages_check won't complain.
1230 #define __pte_lockptr(page) &((page)->ptl)
1231 #define pte_lock_init(_page) do { \
1232 spin_lock_init(__pte_lockptr(_page)); \
1234 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1235 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1236 #else /* !USE_SPLIT_PTLOCKS */
1238 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1240 #define pte_lock_init(page) do {} while (0)
1241 #define pte_lock_deinit(page) do {} while (0)
1242 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1243 #endif /* USE_SPLIT_PTLOCKS */
1245 static inline void pgtable_page_ctor(struct page *page)
1247 pte_lock_init(page);
1248 inc_zone_page_state(page, NR_PAGETABLE);
1251 static inline void pgtable_page_dtor(struct page *page)
1253 pte_lock_deinit(page);
1254 dec_zone_page_state(page, NR_PAGETABLE);
1257 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1259 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1260 pte_t *__pte = pte_offset_map(pmd, address); \
1266 #define pte_unmap_unlock(pte, ptl) do { \
1271 #define pte_alloc_map(mm, vma, pmd, address) \
1272 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1274 NULL: pte_offset_map(pmd, address))
1276 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1277 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1279 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1281 #define pte_alloc_kernel(pmd, address) \
1282 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1283 NULL: pte_offset_kernel(pmd, address))
1285 extern void free_area_init(unsigned long * zones_size);
1286 extern void free_area_init_node(int nid, unsigned long * zones_size,
1287 unsigned long zone_start_pfn, unsigned long *zholes_size);
1288 extern void free_initmem(void);
1290 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1292 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1293 * zones, allocate the backing mem_map and account for memory holes in a more
1294 * architecture independent manner. This is a substitute for creating the
1295 * zone_sizes[] and zholes_size[] arrays and passing them to
1296 * free_area_init_node()
1298 * An architecture is expected to register range of page frames backed by
1299 * physical memory with memblock_add[_node]() before calling
1300 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1301 * usage, an architecture is expected to do something like
1303 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1305 * for_each_valid_physical_page_range()
1306 * memblock_add_node(base, size, nid)
1307 * free_area_init_nodes(max_zone_pfns);
1309 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1310 * registered physical page range. Similarly
1311 * sparse_memory_present_with_active_regions() calls memory_present() for
1312 * each range when SPARSEMEM is enabled.
1314 * See mm/page_alloc.c for more information on each function exposed by
1315 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1317 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1318 unsigned long node_map_pfn_alignment(void);
1319 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1320 unsigned long end_pfn);
1321 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1322 unsigned long end_pfn);
1323 extern void get_pfn_range_for_nid(unsigned int nid,
1324 unsigned long *start_pfn, unsigned long *end_pfn);
1325 extern unsigned long find_min_pfn_with_active_regions(void);
1326 extern void free_bootmem_with_active_regions(int nid,
1327 unsigned long max_low_pfn);
1328 extern void sparse_memory_present_with_active_regions(int nid);
1330 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1332 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1333 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1334 static inline int __early_pfn_to_nid(unsigned long pfn)
1339 /* please see mm/page_alloc.c */
1340 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1341 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1342 /* there is a per-arch backend function. */
1343 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1344 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1347 extern void set_dma_reserve(unsigned long new_dma_reserve);
1348 extern void memmap_init_zone(unsigned long, int, unsigned long,
1349 unsigned long, enum memmap_context);
1350 extern void setup_per_zone_wmarks(void);
1351 extern int __meminit init_per_zone_wmark_min(void);
1352 extern void mem_init(void);
1353 extern void __init mmap_init(void);
1354 extern void show_mem(unsigned int flags);
1355 extern void si_meminfo(struct sysinfo * val);
1356 extern void si_meminfo_node(struct sysinfo *val, int nid);
1357 extern int after_bootmem;
1359 extern __printf(3, 4)
1360 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1362 extern void setup_per_cpu_pageset(void);
1364 extern void zone_pcp_update(struct zone *zone);
1365 extern void zone_pcp_reset(struct zone *zone);
1368 extern atomic_long_t mmap_pages_allocated;
1369 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1371 /* interval_tree.c */
1372 void vma_interval_tree_insert(struct vm_area_struct *node,
1373 struct rb_root *root);
1374 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1375 struct vm_area_struct *prev,
1376 struct rb_root *root);
1377 void vma_interval_tree_remove(struct vm_area_struct *node,
1378 struct rb_root *root);
1379 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1380 unsigned long start, unsigned long last);
1381 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1382 unsigned long start, unsigned long last);
1384 #define vma_interval_tree_foreach(vma, root, start, last) \
1385 for (vma = vma_interval_tree_iter_first(root, start, last); \
1386 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1388 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1389 struct list_head *list)
1391 list_add_tail(&vma->shared.nonlinear, list);
1394 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1395 struct rb_root *root);
1396 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1397 struct rb_root *root);
1398 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1399 struct rb_root *root, unsigned long start, unsigned long last);
1400 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1401 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1402 #ifdef CONFIG_DEBUG_VM_RB
1403 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1406 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1407 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1408 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1411 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1412 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1413 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1414 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1415 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1416 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1417 struct mempolicy *);
1418 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1419 extern int split_vma(struct mm_struct *,
1420 struct vm_area_struct *, unsigned long addr, int new_below);
1421 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1422 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1423 struct rb_node **, struct rb_node *);
1424 extern void unlink_file_vma(struct vm_area_struct *);
1425 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1426 unsigned long addr, unsigned long len, pgoff_t pgoff,
1427 bool *need_rmap_locks);
1428 extern void exit_mmap(struct mm_struct *);
1430 extern int mm_take_all_locks(struct mm_struct *mm);
1431 extern void mm_drop_all_locks(struct mm_struct *mm);
1433 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1434 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1436 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1437 extern int install_special_mapping(struct mm_struct *mm,
1438 unsigned long addr, unsigned long len,
1439 unsigned long flags, struct page **pages);
1441 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1443 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1444 unsigned long len, unsigned long flags,
1445 vm_flags_t vm_flags, unsigned long pgoff);
1446 extern unsigned long do_mmap_pgoff(struct file *, unsigned long,
1447 unsigned long, unsigned long,
1448 unsigned long, unsigned long);
1449 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1451 /* These take the mm semaphore themselves */
1452 extern unsigned long vm_brk(unsigned long, unsigned long);
1453 extern int vm_munmap(unsigned long, size_t);
1454 extern unsigned long vm_mmap(struct file *, unsigned long,
1455 unsigned long, unsigned long,
1456 unsigned long, unsigned long);
1459 extern void truncate_inode_pages(struct address_space *, loff_t);
1460 extern void truncate_inode_pages_range(struct address_space *,
1461 loff_t lstart, loff_t lend);
1463 /* generic vm_area_ops exported for stackable file systems */
1464 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1465 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1467 /* mm/page-writeback.c */
1468 int write_one_page(struct page *page, int wait);
1469 void task_dirty_inc(struct task_struct *tsk);
1472 #define VM_MAX_READAHEAD 128 /* kbytes */
1473 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1475 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1476 pgoff_t offset, unsigned long nr_to_read);
1478 void page_cache_sync_readahead(struct address_space *mapping,
1479 struct file_ra_state *ra,
1482 unsigned long size);
1484 void page_cache_async_readahead(struct address_space *mapping,
1485 struct file_ra_state *ra,
1489 unsigned long size);
1491 unsigned long max_sane_readahead(unsigned long nr);
1492 unsigned long ra_submit(struct file_ra_state *ra,
1493 struct address_space *mapping,
1496 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1497 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1499 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1500 extern int expand_downwards(struct vm_area_struct *vma,
1501 unsigned long address);
1503 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1505 #define expand_upwards(vma, address) do { } while (0)
1508 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1509 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1510 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1511 struct vm_area_struct **pprev);
1513 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1514 NULL if none. Assume start_addr < end_addr. */
1515 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1517 struct vm_area_struct * vma = find_vma(mm,start_addr);
1519 if (vma && end_addr <= vma->vm_start)
1524 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1526 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1529 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1530 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1531 unsigned long vm_start, unsigned long vm_end)
1533 struct vm_area_struct *vma = find_vma(mm, vm_start);
1535 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1542 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1544 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1550 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1551 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1552 unsigned long pfn, unsigned long size, pgprot_t);
1553 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1554 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1556 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1559 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1560 unsigned int foll_flags);
1561 #define FOLL_WRITE 0x01 /* check pte is writable */
1562 #define FOLL_TOUCH 0x02 /* mark page accessed */
1563 #define FOLL_GET 0x04 /* do get_page on page */
1564 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1565 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1566 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1567 * and return without waiting upon it */
1568 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1569 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1570 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1572 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1574 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1575 unsigned long size, pte_fn_t fn, void *data);
1577 #ifdef CONFIG_PROC_FS
1578 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1580 static inline void vm_stat_account(struct mm_struct *mm,
1581 unsigned long flags, struct file *file, long pages)
1583 mm->total_vm += pages;
1585 #endif /* CONFIG_PROC_FS */
1587 #ifdef CONFIG_DEBUG_PAGEALLOC
1588 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1589 #ifdef CONFIG_HIBERNATION
1590 extern bool kernel_page_present(struct page *page);
1591 #endif /* CONFIG_HIBERNATION */
1594 kernel_map_pages(struct page *page, int numpages, int enable) {}
1595 #ifdef CONFIG_HIBERNATION
1596 static inline bool kernel_page_present(struct page *page) { return true; }
1597 #endif /* CONFIG_HIBERNATION */
1600 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1601 #ifdef __HAVE_ARCH_GATE_AREA
1602 int in_gate_area_no_mm(unsigned long addr);
1603 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1605 int in_gate_area_no_mm(unsigned long addr);
1606 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1607 #endif /* __HAVE_ARCH_GATE_AREA */
1609 int drop_caches_sysctl_handler(struct ctl_table *, int,
1610 void __user *, size_t *, loff_t *);
1611 unsigned long shrink_slab(struct shrink_control *shrink,
1612 unsigned long nr_pages_scanned,
1613 unsigned long lru_pages);
1616 #define randomize_va_space 0
1618 extern int randomize_va_space;
1621 const char * arch_vma_name(struct vm_area_struct *vma);
1622 void print_vma_addr(char *prefix, unsigned long rip);
1624 void sparse_mem_maps_populate_node(struct page **map_map,
1625 unsigned long pnum_begin,
1626 unsigned long pnum_end,
1627 unsigned long map_count,
1630 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1631 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1632 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1633 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1634 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1635 void *vmemmap_alloc_block(unsigned long size, int node);
1636 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1637 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1638 int vmemmap_populate_basepages(struct page *start_page,
1639 unsigned long pages, int node);
1640 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1641 void vmemmap_populate_print_last(void);
1645 MF_COUNT_INCREASED = 1 << 0,
1646 MF_ACTION_REQUIRED = 1 << 1,
1647 MF_MUST_KILL = 1 << 2,
1649 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1650 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1651 extern int unpoison_memory(unsigned long pfn);
1652 extern int sysctl_memory_failure_early_kill;
1653 extern int sysctl_memory_failure_recovery;
1654 extern void shake_page(struct page *p, int access);
1655 extern atomic_long_t mce_bad_pages;
1656 extern int soft_offline_page(struct page *page, int flags);
1658 extern void dump_page(struct page *page);
1660 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1661 extern void clear_huge_page(struct page *page,
1663 unsigned int pages_per_huge_page);
1664 extern void copy_user_huge_page(struct page *dst, struct page *src,
1665 unsigned long addr, struct vm_area_struct *vma,
1666 unsigned int pages_per_huge_page);
1667 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1669 #ifdef CONFIG_DEBUG_PAGEALLOC
1670 extern unsigned int _debug_guardpage_minorder;
1672 static inline unsigned int debug_guardpage_minorder(void)
1674 return _debug_guardpage_minorder;
1677 static inline bool page_is_guard(struct page *page)
1679 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1682 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1683 static inline bool page_is_guard(struct page *page) { return false; }
1684 #endif /* CONFIG_DEBUG_PAGEALLOC */
1686 #endif /* __KERNEL__ */
1687 #endif /* _LINUX_MM_H */