4 #include <linux/sched.h>
5 #include <linux/errno.h>
6 #include <linux/capability.h>
10 #include <linux/config.h>
11 #include <linux/gfp.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/prio_tree.h>
17 #include <linux/mutex.h>
22 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23 extern unsigned long max_mapnr;
26 extern unsigned long num_physpages;
27 extern void * high_memory;
28 extern unsigned long vmalloc_earlyreserve;
29 extern int page_cluster;
32 extern int sysctl_legacy_va_layout;
34 #define sysctl_legacy_va_layout 0
38 #include <asm/pgtable.h>
39 #include <asm/processor.h>
40 #include <asm/atomic.h>
42 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
54 * This struct defines a memory VMM memory area. There is one of these
55 * per VM-area/task. A VM area is any part of the process virtual memory
56 * space that has a special rule for the page-fault handlers (ie a shared
57 * library, the executable area etc).
59 struct vm_area_struct {
60 struct mm_struct * vm_mm; /* The address space we belong to. */
61 unsigned long vm_start; /* Our start address within vm_mm. */
62 unsigned long vm_end; /* The first byte after our end address
65 /* linked list of VM areas per task, sorted by address */
66 struct vm_area_struct *vm_next;
68 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
69 unsigned long vm_flags; /* Flags, listed below. */
74 * For areas with an address space and backing store,
75 * linkage into the address_space->i_mmap prio tree, or
76 * linkage to the list of like vmas hanging off its node, or
77 * linkage of vma in the address_space->i_mmap_nonlinear list.
81 struct list_head list;
82 void *parent; /* aligns with prio_tree_node parent */
83 struct vm_area_struct *head;
86 struct raw_prio_tree_node prio_tree_node;
90 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
91 * list, after a COW of one of the file pages. A MAP_SHARED vma
92 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
93 * or brk vma (with NULL file) can only be in an anon_vma list.
95 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
96 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
98 /* Function pointers to deal with this struct. */
99 struct vm_operations_struct * vm_ops;
101 /* Information about our backing store: */
102 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
103 units, *not* PAGE_CACHE_SIZE */
104 struct file * vm_file; /* File we map to (can be NULL). */
105 void * vm_private_data; /* was vm_pte (shared mem) */
106 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
109 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
112 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
117 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
118 * disabled, then there's a single shared list of VMAs maintained by the
119 * system, and mm's subscribe to these individually
121 struct vm_list_struct {
122 struct vm_list_struct *next;
123 struct vm_area_struct *vma;
127 extern struct rb_root nommu_vma_tree;
128 extern struct rw_semaphore nommu_vma_sem;
130 extern unsigned int kobjsize(const void *objp);
136 #define VM_READ 0x00000001 /* currently active flags */
137 #define VM_WRITE 0x00000002
138 #define VM_EXEC 0x00000004
139 #define VM_SHARED 0x00000008
141 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
142 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
143 #define VM_MAYWRITE 0x00000020
144 #define VM_MAYEXEC 0x00000040
145 #define VM_MAYSHARE 0x00000080
147 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
148 #define VM_GROWSUP 0x00000200
149 #define VM_SHM 0x00000000 /* Means nothing: delete it later */
150 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
151 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
153 #define VM_EXECUTABLE 0x00001000
154 #define VM_LOCKED 0x00002000
155 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
157 /* Used by sys_madvise() */
158 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
161 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
163 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
164 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
168 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
170 #define VM_FOREIGN 0x04000000 /* Has pages belonging to another VM */
173 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
174 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
177 #ifdef CONFIG_STACK_GROWSUP
178 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
183 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
184 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
185 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
186 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
187 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
190 * mapping from the currently active vm_flags protection bits (the
191 * low four bits) to a page protection mask..
193 extern pgprot_t protection_map[16];
197 * These are the virtual MM functions - opening of an area, closing and
198 * unmapping it (needed to keep files on disk up-to-date etc), pointer
199 * to the functions called when a no-page or a wp-page exception occurs.
201 struct vm_operations_struct {
202 void (*open)(struct vm_area_struct * area);
203 void (*close)(struct vm_area_struct * area);
204 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
205 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
207 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
208 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
217 * Each physical page in the system has a struct page associated with
218 * it to keep track of whatever it is we are using the page for at the
219 * moment. Note that we have no way to track which tasks are using
223 unsigned long flags; /* Atomic flags, some possibly
224 * updated asynchronously */
225 atomic_t _count; /* Usage count, see below. */
226 atomic_t _mapcount; /* Count of ptes mapped in mms,
227 * to show when page is mapped
228 * & limit reverse map searches.
232 unsigned long private; /* Mapping-private opaque data:
233 * usually used for buffer_heads
234 * if PagePrivate set; used for
235 * swp_entry_t if PageSwapCache;
236 * indicates order in the buddy
237 * system if PG_buddy is set.
239 struct address_space *mapping; /* If low bit clear, points to
240 * inode address_space, or NULL.
241 * If page mapped as anonymous
242 * memory, low bit is set, and
243 * it points to anon_vma object:
244 * see PAGE_MAPPING_ANON below.
247 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
251 pgoff_t index; /* Our offset within mapping. */
252 struct list_head lru; /* Pageout list, eg. active_list
253 * protected by zone->lru_lock !
256 * On machines where all RAM is mapped into kernel address space,
257 * we can simply calculate the virtual address. On machines with
258 * highmem some memory is mapped into kernel virtual memory
259 * dynamically, so we need a place to store that address.
260 * Note that this field could be 16 bits on x86 ... ;)
262 * Architectures with slow multiplication can define
263 * WANT_PAGE_VIRTUAL in asm/page.h
265 #if defined(WANT_PAGE_VIRTUAL)
266 void *virtual; /* Kernel virtual address (NULL if
267 not kmapped, ie. highmem) */
268 #endif /* WANT_PAGE_VIRTUAL */
271 #define page_private(page) ((page)->private)
272 #define set_page_private(page, v) ((page)->private = (v))
275 * FIXME: take this include out, include page-flags.h in
276 * files which need it (119 of them)
278 #include <linux/page-flags.h>
281 * Methods to modify the page usage count.
283 * What counts for a page usage:
284 * - cache mapping (page->mapping)
285 * - private data (page->private)
286 * - page mapped in a task's page tables, each mapping
287 * is counted separately
289 * Also, many kernel routines increase the page count before a critical
290 * routine so they can be sure the page doesn't go away from under them.
294 * Drop a ref, return true if the logical refcount fell to zero (the page has
297 static inline int put_page_testzero(struct page *page)
299 BUG_ON(atomic_read(&page->_count) == 0);
300 return atomic_dec_and_test(&page->_count);
304 * Try to grab a ref unless the page has a refcount of zero, return false if
307 static inline int get_page_unless_zero(struct page *page)
309 return atomic_inc_not_zero(&page->_count);
312 extern void FASTCALL(__page_cache_release(struct page *));
314 static inline int page_count(struct page *page)
316 if (unlikely(PageCompound(page)))
317 page = (struct page *)page_private(page);
318 return atomic_read(&page->_count);
321 static inline void get_page(struct page *page)
323 if (unlikely(PageCompound(page)))
324 page = (struct page *)page_private(page);
325 atomic_inc(&page->_count);
329 * Setup the page count before being freed into the page allocator for
330 * the first time (boot or memory hotplug)
332 static inline void init_page_count(struct page *page)
334 atomic_set(&page->_count, 1);
337 void put_page(struct page *page);
339 void split_page(struct page *page, unsigned int order);
342 * Multiple processes may "see" the same page. E.g. for untouched
343 * mappings of /dev/null, all processes see the same page full of
344 * zeroes, and text pages of executables and shared libraries have
345 * only one copy in memory, at most, normally.
347 * For the non-reserved pages, page_count(page) denotes a reference count.
348 * page_count() == 0 means the page is free. page->lru is then used for
349 * freelist management in the buddy allocator.
350 * page_count() == 1 means the page is used for exactly one purpose
351 * (e.g. a private data page of one process).
353 * A page may be used for kmalloc() or anyone else who does a
354 * __get_free_page(). In this case the page_count() is at least 1, and
355 * all other fields are unused but should be 0 or NULL. The
356 * management of this page is the responsibility of the one who uses
359 * The other pages (we may call them "process pages") are completely
360 * managed by the Linux memory manager: I/O, buffers, swapping etc.
361 * The following discussion applies only to them.
363 * A page may belong to an inode's memory mapping. In this case,
364 * page->mapping is the pointer to the inode, and page->index is the
365 * file offset of the page, in units of PAGE_CACHE_SIZE.
367 * A page contains an opaque `private' member, which belongs to the
368 * page's address_space. Usually, this is the address of a circular
369 * list of the page's disk buffers.
371 * For pages belonging to inodes, the page_count() is the number of
372 * attaches, plus 1 if `private' contains something, plus one for
373 * the page cache itself.
375 * Instead of keeping dirty/clean pages in per address-space lists, we instead
376 * now tag pages as dirty/under writeback in the radix tree.
378 * There is also a per-mapping radix tree mapping index to the page
379 * in memory if present. The tree is rooted at mapping->root.
381 * All process pages can do I/O:
382 * - inode pages may need to be read from disk,
383 * - inode pages which have been modified and are MAP_SHARED may need
384 * to be written to disk,
385 * - private pages which have been modified may need to be swapped out
386 * to swap space and (later) to be read back into memory.
390 * The zone field is never updated after free_area_init_core()
391 * sets it, so none of the operations on it need to be atomic.
396 * page->flags layout:
398 * There are three possibilities for how page->flags get
399 * laid out. The first is for the normal case, without
400 * sparsemem. The second is for sparsemem when there is
401 * plenty of space for node and section. The last is when
402 * we have run out of space and have to fall back to an
403 * alternate (slower) way of determining the node.
405 * No sparsemem: | NODE | ZONE | ... | FLAGS |
406 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
407 * no space for node: | SECTION | ZONE | ... | FLAGS |
409 #ifdef CONFIG_SPARSEMEM
410 #define SECTIONS_WIDTH SECTIONS_SHIFT
412 #define SECTIONS_WIDTH 0
415 #define ZONES_WIDTH ZONES_SHIFT
417 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
418 #define NODES_WIDTH NODES_SHIFT
420 #define NODES_WIDTH 0
423 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
424 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
425 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
426 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
429 * We are going to use the flags for the page to node mapping if its in
430 * there. This includes the case where there is no node, so it is implicit.
432 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
434 #ifndef PFN_SECTION_SHIFT
435 #define PFN_SECTION_SHIFT 0
439 * Define the bit shifts to access each section. For non-existant
440 * sections we define the shift as 0; that plus a 0 mask ensures
441 * the compiler will optimise away reference to them.
443 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
444 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
445 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
447 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
449 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
451 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
453 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
455 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
456 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
459 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
460 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
461 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
462 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
464 static inline unsigned long page_zonenum(struct page *page)
466 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
470 extern struct zone *zone_table[];
472 static inline struct zone *page_zone(struct page *page)
474 return zone_table[(page->flags >> ZONETABLE_PGSHIFT) &
478 static inline unsigned long page_to_nid(struct page *page)
481 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
483 return page_zone(page)->zone_pgdat->node_id;
485 static inline unsigned long page_to_section(struct page *page)
487 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
490 static inline void set_page_zone(struct page *page, unsigned long zone)
492 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
493 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
495 static inline void set_page_node(struct page *page, unsigned long node)
497 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
498 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
500 static inline void set_page_section(struct page *page, unsigned long section)
502 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
503 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
506 static inline void set_page_links(struct page *page, unsigned long zone,
507 unsigned long node, unsigned long pfn)
509 set_page_zone(page, zone);
510 set_page_node(page, node);
511 set_page_section(page, pfn_to_section_nr(pfn));
514 #ifndef CONFIG_DISCONTIGMEM
515 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
516 extern struct page *mem_map;
519 static __always_inline void *lowmem_page_address(struct page *page)
521 return __va(page_to_pfn(page) << PAGE_SHIFT);
524 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
525 #define HASHED_PAGE_VIRTUAL
528 #if defined(WANT_PAGE_VIRTUAL)
529 #define page_address(page) ((page)->virtual)
530 #define set_page_address(page, address) \
532 (page)->virtual = (address); \
534 #define page_address_init() do { } while(0)
537 #if defined(HASHED_PAGE_VIRTUAL)
538 void *page_address(struct page *page);
539 void set_page_address(struct page *page, void *virtual);
540 void page_address_init(void);
543 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
544 #define page_address(page) lowmem_page_address(page)
545 #define set_page_address(page, address) do { } while(0)
546 #define page_address_init() do { } while(0)
550 * On an anonymous page mapped into a user virtual memory area,
551 * page->mapping points to its anon_vma, not to a struct address_space;
552 * with the PAGE_MAPPING_ANON bit set to distinguish it.
554 * Please note that, confusingly, "page_mapping" refers to the inode
555 * address_space which maps the page from disk; whereas "page_mapped"
556 * refers to user virtual address space into which the page is mapped.
558 #define PAGE_MAPPING_ANON 1
560 extern struct address_space swapper_space;
561 static inline struct address_space *page_mapping(struct page *page)
563 struct address_space *mapping = page->mapping;
565 if (unlikely(PageSwapCache(page)))
566 mapping = &swapper_space;
567 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
572 static inline int PageAnon(struct page *page)
574 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
578 * Return the pagecache index of the passed page. Regular pagecache pages
579 * use ->index whereas swapcache pages use ->private
581 static inline pgoff_t page_index(struct page *page)
583 if (unlikely(PageSwapCache(page)))
584 return page_private(page);
589 * The atomic page->_mapcount, like _count, starts from -1:
590 * so that transitions both from it and to it can be tracked,
591 * using atomic_inc_and_test and atomic_add_negative(-1).
593 static inline void reset_page_mapcount(struct page *page)
595 atomic_set(&(page)->_mapcount, -1);
598 static inline int page_mapcount(struct page *page)
600 return atomic_read(&(page)->_mapcount) + 1;
604 * Return true if this page is mapped into pagetables.
606 static inline int page_mapped(struct page *page)
608 return atomic_read(&(page)->_mapcount) >= 0;
612 * Error return values for the *_nopage functions
614 #define NOPAGE_SIGBUS (NULL)
615 #define NOPAGE_OOM ((struct page *) (-1))
618 * Different kinds of faults, as returned by handle_mm_fault().
619 * Used to decide whether a process gets delivered SIGBUS or
620 * just gets major/minor fault counters bumped up.
622 #define VM_FAULT_OOM 0x00
623 #define VM_FAULT_SIGBUS 0x01
624 #define VM_FAULT_MINOR 0x02
625 #define VM_FAULT_MAJOR 0x03
628 * Special case for get_user_pages.
629 * Must be in a distinct bit from the above VM_FAULT_ flags.
631 #define VM_FAULT_WRITE 0x10
633 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
635 extern void show_free_areas(void);
638 struct page *shmem_nopage(struct vm_area_struct *vma,
639 unsigned long address, int *type);
640 int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
641 struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
643 int shmem_lock(struct file *file, int lock, struct user_struct *user);
645 #define shmem_nopage filemap_nopage
647 static inline int shmem_lock(struct file *file, int lock,
648 struct user_struct *user)
653 static inline int shmem_set_policy(struct vm_area_struct *vma,
654 struct mempolicy *new)
659 static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
665 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
666 extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
668 int shmem_zero_setup(struct vm_area_struct *);
671 extern unsigned long shmem_get_unmapped_area(struct file *file,
675 unsigned long flags);
678 static inline int can_do_mlock(void)
680 if (capable(CAP_IPC_LOCK))
682 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
686 extern int user_shm_lock(size_t, struct user_struct *);
687 extern void user_shm_unlock(size_t, struct user_struct *);
690 * Parameter block passed down to zap_pte_range in exceptional cases.
693 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
694 struct address_space *check_mapping; /* Check page->mapping if set */
695 pgoff_t first_index; /* Lowest page->index to unmap */
696 pgoff_t last_index; /* Highest page->index to unmap */
697 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
698 unsigned long truncate_count; /* Compare vm_truncate_count */
701 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
702 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
703 unsigned long size, struct zap_details *);
704 unsigned long unmap_vmas(struct mmu_gather **tlb,
705 struct vm_area_struct *start_vma, unsigned long start_addr,
706 unsigned long end_addr, unsigned long *nr_accounted,
707 struct zap_details *);
708 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
709 unsigned long end, unsigned long floor, unsigned long ceiling);
710 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
711 unsigned long floor, unsigned long ceiling);
712 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
713 struct vm_area_struct *vma);
714 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
715 unsigned long size, pgprot_t prot);
716 void unmap_mapping_range(struct address_space *mapping,
717 loff_t const holebegin, loff_t const holelen, int even_cows);
719 static inline void unmap_shared_mapping_range(struct address_space *mapping,
720 loff_t const holebegin, loff_t const holelen)
722 unmap_mapping_range(mapping, holebegin, holelen, 0);
725 extern int vmtruncate(struct inode * inode, loff_t offset);
726 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
727 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
728 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
731 extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
732 unsigned long address, int write_access);
734 static inline int handle_mm_fault(struct mm_struct *mm,
735 struct vm_area_struct *vma, unsigned long address,
738 return __handle_mm_fault(mm, vma, address, write_access) &
742 static inline int handle_mm_fault(struct mm_struct *mm,
743 struct vm_area_struct *vma, unsigned long address,
746 /* should never happen if there's no MMU */
748 return VM_FAULT_SIGBUS;
752 extern int make_pages_present(unsigned long addr, unsigned long end);
753 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
754 void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
756 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
757 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
758 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
760 int __set_page_dirty_buffers(struct page *page);
761 int __set_page_dirty_nobuffers(struct page *page);
762 int redirty_page_for_writepage(struct writeback_control *wbc,
764 int FASTCALL(set_page_dirty(struct page *page));
765 int set_page_dirty_lock(struct page *page);
766 int clear_page_dirty_for_io(struct page *page);
768 extern unsigned long do_mremap(unsigned long addr,
769 unsigned long old_len, unsigned long new_len,
770 unsigned long flags, unsigned long new_addr);
773 * Prototype to add a shrinker callback for ageable caches.
775 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
776 * scan `nr_to_scan' objects, attempting to free them.
778 * The callback must return the number of objects which remain in the cache.
780 * The callback will be passed nr_to_scan == 0 when the VM is querying the
781 * cache size, so a fastpath for that case is appropriate.
783 typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
786 * Add an aging callback. The int is the number of 'seeks' it takes
787 * to recreate one of the objects that these functions age.
790 #define DEFAULT_SEEKS 2
792 extern struct shrinker *set_shrinker(int, shrinker_t);
793 extern void remove_shrinker(struct shrinker *shrinker);
795 extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
797 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
798 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
799 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
800 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
803 * The following ifdef needed to get the 4level-fixup.h header to work.
804 * Remove it when 4level-fixup.h has been removed.
806 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
807 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
809 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
810 NULL: pud_offset(pgd, address);
813 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
815 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
816 NULL: pmd_offset(pud, address);
818 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
820 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
822 * We tuck a spinlock to guard each pagetable page into its struct page,
823 * at page->private, with BUILD_BUG_ON to make sure that this will not
824 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
825 * When freeing, reset page->mapping so free_pages_check won't complain.
827 #define __pte_lockptr(page) &((page)->ptl)
828 #define pte_lock_init(_page) do { \
829 spin_lock_init(__pte_lockptr(_page)); \
831 #define pte_lock_deinit(page) ((page)->mapping = NULL)
832 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
835 * We use mm->page_table_lock to guard all pagetable pages of the mm.
837 #define pte_lock_init(page) do {} while (0)
838 #define pte_lock_deinit(page) do {} while (0)
839 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
840 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
842 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
844 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
845 pte_t *__pte = pte_offset_map(pmd, address); \
851 #define pte_unmap_unlock(pte, ptl) do { \
856 #define pte_alloc_map(mm, pmd, address) \
857 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
858 NULL: pte_offset_map(pmd, address))
860 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
861 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
862 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
864 #define pte_alloc_kernel(pmd, address) \
865 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
866 NULL: pte_offset_kernel(pmd, address))
868 extern void free_area_init(unsigned long * zones_size);
869 extern void free_area_init_node(int nid, pg_data_t *pgdat,
870 unsigned long * zones_size, unsigned long zone_start_pfn,
871 unsigned long *zholes_size);
872 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
873 extern void setup_per_zone_pages_min(void);
874 extern void mem_init(void);
875 extern void show_mem(void);
876 extern void si_meminfo(struct sysinfo * val);
877 extern void si_meminfo_node(struct sysinfo *val, int nid);
880 extern void setup_per_cpu_pageset(void);
882 static inline void setup_per_cpu_pageset(void) {}
886 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
887 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
888 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
889 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
890 struct prio_tree_iter *iter);
892 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
893 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
894 (vma = vma_prio_tree_next(vma, iter)); )
896 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
897 struct list_head *list)
899 vma->shared.vm_set.parent = NULL;
900 list_add_tail(&vma->shared.vm_set.list, list);
904 extern int __vm_enough_memory(long pages, int cap_sys_admin);
905 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
906 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
907 extern struct vm_area_struct *vma_merge(struct mm_struct *,
908 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
909 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
911 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
912 extern int split_vma(struct mm_struct *,
913 struct vm_area_struct *, unsigned long addr, int new_below);
914 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
915 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
916 struct rb_node **, struct rb_node *);
917 extern void unlink_file_vma(struct vm_area_struct *);
918 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
919 unsigned long addr, unsigned long len, pgoff_t pgoff);
920 extern void exit_mmap(struct mm_struct *);
921 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
923 extern unsigned long get_unmapped_area_prot(struct file *, unsigned long, unsigned long, unsigned long, unsigned long, int);
926 static inline unsigned long get_unmapped_area(struct file * file, unsigned long addr,
927 unsigned long len, unsigned long pgoff, unsigned long flags)
929 return get_unmapped_area_prot(file, addr, len, pgoff, flags, 0);
932 extern int install_special_mapping(struct mm_struct *mm,
933 unsigned long addr, unsigned long len,
934 unsigned long vm_flags, pgprot_t pgprot,
935 struct page **pages);
937 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
938 unsigned long len, unsigned long prot,
939 unsigned long flag, unsigned long pgoff);
941 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
942 unsigned long len, unsigned long prot,
943 unsigned long flag, unsigned long offset)
945 unsigned long ret = -EINVAL;
946 if ((offset + PAGE_ALIGN(len)) < offset)
948 if (!(offset & ~PAGE_MASK))
949 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
954 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
956 extern unsigned long do_brk(unsigned long, unsigned long);
959 extern unsigned long page_unuse(struct page *);
960 extern void truncate_inode_pages(struct address_space *, loff_t);
961 extern void truncate_inode_pages_range(struct address_space *,
962 loff_t lstart, loff_t lend);
964 /* generic vm_area_ops exported for stackable file systems */
965 extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
966 extern int filemap_populate(struct vm_area_struct *, unsigned long,
967 unsigned long, pgprot_t, unsigned long, int);
969 /* mm/page-writeback.c */
970 int write_one_page(struct page *page, int wait);
973 #define VM_MAX_READAHEAD 128 /* kbytes */
974 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
975 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
976 * turning readahead off */
978 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
979 pgoff_t offset, unsigned long nr_to_read);
980 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
981 pgoff_t offset, unsigned long nr_to_read);
982 unsigned long page_cache_readahead(struct address_space *mapping,
983 struct file_ra_state *ra,
987 void handle_ra_miss(struct address_space *mapping,
988 struct file_ra_state *ra, pgoff_t offset);
989 unsigned long max_sane_readahead(unsigned long nr);
991 /* Do stack extension */
992 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
994 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
997 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
998 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
999 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1000 struct vm_area_struct **pprev);
1002 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1003 NULL if none. Assume start_addr < end_addr. */
1004 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1006 struct vm_area_struct * vma = find_vma(mm,start_addr);
1008 if (vma && end_addr <= vma->vm_start)
1013 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1015 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1018 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1019 struct page *vmalloc_to_page(void *addr);
1020 unsigned long vmalloc_to_pfn(void *addr);
1021 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1022 unsigned long pfn, unsigned long size, pgprot_t);
1023 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1025 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1026 unsigned int foll_flags);
1027 #define FOLL_WRITE 0x01 /* check pte is writable */
1028 #define FOLL_TOUCH 0x02 /* mark page accessed */
1029 #define FOLL_GET 0x04 /* do get_page on page */
1030 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1033 typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr,
1035 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1036 unsigned long size, pte_fn_t fn, void *data);
1039 #ifdef CONFIG_PROC_FS
1040 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1042 static inline void vm_stat_account(struct mm_struct *mm,
1043 unsigned long flags, struct file *file, long pages)
1046 #endif /* CONFIG_PROC_FS */
1048 #ifndef CONFIG_DEBUG_PAGEALLOC
1050 kernel_map_pages(struct page *page, int numpages, int enable)
1052 if (!PageHighMem(page) && !enable)
1053 mutex_debug_check_no_locks_freed(page_address(page),
1054 numpages * PAGE_SIZE);
1058 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1059 #ifdef __HAVE_ARCH_GATE_AREA
1060 int in_gate_area_no_task(unsigned long addr);
1061 int in_gate_area(struct task_struct *task, unsigned long addr);
1063 int in_gate_area_no_task(unsigned long addr);
1064 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1065 #endif /* __HAVE_ARCH_GATE_AREA */
1067 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1068 #define OOM_DISABLE -17
1070 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1071 void __user *, size_t *, loff_t *);
1072 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1073 unsigned long lru_pages);
1074 void drop_pagecache(void);
1075 void drop_slab(void);
1078 #define randomize_va_space 0
1080 extern int randomize_va_space;
1083 #endif /* __KERNEL__ */
1084 #endif /* _LINUX_MM_H */