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 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
171 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
174 #ifdef CONFIG_STACK_GROWSUP
175 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
177 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
181 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
182 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
183 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
184 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
187 * mapping from the currently active vm_flags protection bits (the
188 * low four bits) to a page protection mask..
190 extern pgprot_t protection_map[16];
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
198 struct vm_operations_struct {
199 void (*open)(struct vm_area_struct * area);
200 void (*close)(struct vm_area_struct * area);
201 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
202 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
204 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
205 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
214 * Each physical page in the system has a struct page associated with
215 * it to keep track of whatever it is we are using the page for at the
216 * moment. Note that we have no way to track which tasks are using
220 unsigned long flags; /* Atomic flags, some possibly
221 * updated asynchronously */
222 atomic_t _count; /* Usage count, see below. */
223 atomic_t _mapcount; /* Count of ptes mapped in mms,
224 * to show when page is mapped
225 * & limit reverse map searches.
229 unsigned long private; /* Mapping-private opaque data:
230 * usually used for buffer_heads
231 * if PagePrivate set; used for
232 * swp_entry_t if PageSwapCache.
233 * When page is free, this
234 * indicates order in the buddy
237 struct address_space *mapping; /* If low bit clear, points to
238 * inode address_space, or NULL.
239 * If page mapped as anonymous
240 * memory, low bit is set, and
241 * it points to anon_vma object:
242 * see PAGE_MAPPING_ANON below.
245 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
249 pgoff_t index; /* Our offset within mapping. */
250 struct list_head lru; /* Pageout list, eg. active_list
251 * protected by zone->lru_lock !
254 * On machines where all RAM is mapped into kernel address space,
255 * we can simply calculate the virtual address. On machines with
256 * highmem some memory is mapped into kernel virtual memory
257 * dynamically, so we need a place to store that address.
258 * Note that this field could be 16 bits on x86 ... ;)
260 * Architectures with slow multiplication can define
261 * WANT_PAGE_VIRTUAL in asm/page.h
263 #if defined(WANT_PAGE_VIRTUAL)
264 void *virtual; /* Kernel virtual address (NULL if
265 not kmapped, ie. highmem) */
266 #endif /* WANT_PAGE_VIRTUAL */
269 #define page_private(page) ((page)->private)
270 #define set_page_private(page, v) ((page)->private = (v))
273 * FIXME: take this include out, include page-flags.h in
274 * files which need it (119 of them)
276 #include <linux/page-flags.h>
279 * Methods to modify the page usage count.
281 * What counts for a page usage:
282 * - cache mapping (page->mapping)
283 * - private data (page->private)
284 * - page mapped in a task's page tables, each mapping
285 * is counted separately
287 * Also, many kernel routines increase the page count before a critical
288 * routine so they can be sure the page doesn't go away from under them.
290 * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we
291 * can use atomic_add_negative(-1, page->_count) to detect when the page
292 * becomes free and so that we can also use atomic_inc_and_test to atomically
293 * detect when we just tried to grab a ref on a page which some other CPU has
294 * already deemed to be freeable.
296 * NO code should make assumptions about this internal detail! Use the provided
297 * macros which retain the old rules: page_count(page) == 0 is a free page.
301 * Drop a ref, return true if the logical refcount fell to zero (the page has
304 #define put_page_testzero(p) \
306 BUG_ON(page_count(p) == 0); \
307 atomic_add_negative(-1, &(p)->_count); \
311 * Grab a ref, return true if the page previously had a logical refcount of
312 * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page
314 #define get_page_testone(p) atomic_inc_and_test(&(p)->_count)
316 #define set_page_count(p,v) atomic_set(&(p)->_count, (v) - 1)
317 #define __put_page(p) atomic_dec(&(p)->_count)
319 extern void FASTCALL(__page_cache_release(struct page *));
321 static inline int page_count(struct page *page)
323 if (PageCompound(page))
324 page = (struct page *)page_private(page);
325 return atomic_read(&page->_count) + 1;
328 static inline void get_page(struct page *page)
330 if (unlikely(PageCompound(page)))
331 page = (struct page *)page_private(page);
332 atomic_inc(&page->_count);
335 void put_page(struct page *page);
338 * Multiple processes may "see" the same page. E.g. for untouched
339 * mappings of /dev/null, all processes see the same page full of
340 * zeroes, and text pages of executables and shared libraries have
341 * only one copy in memory, at most, normally.
343 * For the non-reserved pages, page_count(page) denotes a reference count.
344 * page_count() == 0 means the page is free. page->lru is then used for
345 * freelist management in the buddy allocator.
346 * page_count() == 1 means the page is used for exactly one purpose
347 * (e.g. a private data page of one process).
349 * A page may be used for kmalloc() or anyone else who does a
350 * __get_free_page(). In this case the page_count() is at least 1, and
351 * all other fields are unused but should be 0 or NULL. The
352 * management of this page is the responsibility of the one who uses
355 * The other pages (we may call them "process pages") are completely
356 * managed by the Linux memory manager: I/O, buffers, swapping etc.
357 * The following discussion applies only to them.
359 * A page may belong to an inode's memory mapping. In this case,
360 * page->mapping is the pointer to the inode, and page->index is the
361 * file offset of the page, in units of PAGE_CACHE_SIZE.
363 * A page contains an opaque `private' member, which belongs to the
364 * page's address_space. Usually, this is the address of a circular
365 * list of the page's disk buffers.
367 * For pages belonging to inodes, the page_count() is the number of
368 * attaches, plus 1 if `private' contains something, plus one for
369 * the page cache itself.
371 * Instead of keeping dirty/clean pages in per address-space lists, we instead
372 * now tag pages as dirty/under writeback in the radix tree.
374 * There is also a per-mapping radix tree mapping index to the page
375 * in memory if present. The tree is rooted at mapping->root.
377 * All process pages can do I/O:
378 * - inode pages may need to be read from disk,
379 * - inode pages which have been modified and are MAP_SHARED may need
380 * to be written to disk,
381 * - private pages which have been modified may need to be swapped out
382 * to swap space and (later) to be read back into memory.
386 * The zone field is never updated after free_area_init_core()
387 * sets it, so none of the operations on it need to be atomic.
392 * page->flags layout:
394 * There are three possibilities for how page->flags get
395 * laid out. The first is for the normal case, without
396 * sparsemem. The second is for sparsemem when there is
397 * plenty of space for node and section. The last is when
398 * we have run out of space and have to fall back to an
399 * alternate (slower) way of determining the node.
401 * No sparsemem: | NODE | ZONE | ... | FLAGS |
402 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
403 * no space for node: | SECTION | ZONE | ... | FLAGS |
405 #ifdef CONFIG_SPARSEMEM
406 #define SECTIONS_WIDTH SECTIONS_SHIFT
408 #define SECTIONS_WIDTH 0
411 #define ZONES_WIDTH ZONES_SHIFT
413 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
414 #define NODES_WIDTH NODES_SHIFT
416 #define NODES_WIDTH 0
419 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
420 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
421 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
422 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
425 * We are going to use the flags for the page to node mapping if its in
426 * there. This includes the case where there is no node, so it is implicit.
428 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
430 #ifndef PFN_SECTION_SHIFT
431 #define PFN_SECTION_SHIFT 0
435 * Define the bit shifts to access each section. For non-existant
436 * sections we define the shift as 0; that plus a 0 mask ensures
437 * the compiler will optimise away reference to them.
439 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
440 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
441 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
443 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
445 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
447 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
449 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
451 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
452 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
455 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
456 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
457 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
458 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
460 static inline unsigned long page_zonenum(struct page *page)
462 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
466 extern struct zone *zone_table[];
468 static inline struct zone *page_zone(struct page *page)
470 return zone_table[(page->flags >> ZONETABLE_PGSHIFT) &
474 static inline unsigned long page_to_nid(struct page *page)
477 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
479 return page_zone(page)->zone_pgdat->node_id;
481 static inline unsigned long page_to_section(struct page *page)
483 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
486 static inline void set_page_zone(struct page *page, unsigned long zone)
488 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
489 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
491 static inline void set_page_node(struct page *page, unsigned long node)
493 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
494 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
496 static inline void set_page_section(struct page *page, unsigned long section)
498 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
499 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
502 static inline void set_page_links(struct page *page, unsigned long zone,
503 unsigned long node, unsigned long pfn)
505 set_page_zone(page, zone);
506 set_page_node(page, node);
507 set_page_section(page, pfn_to_section_nr(pfn));
510 #ifndef CONFIG_DISCONTIGMEM
511 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
512 extern struct page *mem_map;
515 static __always_inline void *lowmem_page_address(struct page *page)
517 return __va(page_to_pfn(page) << PAGE_SHIFT);
520 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
521 #define HASHED_PAGE_VIRTUAL
524 #if defined(WANT_PAGE_VIRTUAL)
525 #define page_address(page) ((page)->virtual)
526 #define set_page_address(page, address) \
528 (page)->virtual = (address); \
530 #define page_address_init() do { } while(0)
533 #if defined(HASHED_PAGE_VIRTUAL)
534 void *page_address(struct page *page);
535 void set_page_address(struct page *page, void *virtual);
536 void page_address_init(void);
539 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
540 #define page_address(page) lowmem_page_address(page)
541 #define set_page_address(page, address) do { } while(0)
542 #define page_address_init() do { } while(0)
546 * On an anonymous page mapped into a user virtual memory area,
547 * page->mapping points to its anon_vma, not to a struct address_space;
548 * with the PAGE_MAPPING_ANON bit set to distinguish it.
550 * Please note that, confusingly, "page_mapping" refers to the inode
551 * address_space which maps the page from disk; whereas "page_mapped"
552 * refers to user virtual address space into which the page is mapped.
554 #define PAGE_MAPPING_ANON 1
556 extern struct address_space swapper_space;
557 static inline struct address_space *page_mapping(struct page *page)
559 struct address_space *mapping = page->mapping;
561 if (unlikely(PageSwapCache(page)))
562 mapping = &swapper_space;
563 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
568 static inline int PageAnon(struct page *page)
570 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
574 * Return the pagecache index of the passed page. Regular pagecache pages
575 * use ->index whereas swapcache pages use ->private
577 static inline pgoff_t page_index(struct page *page)
579 if (unlikely(PageSwapCache(page)))
580 return page_private(page);
585 * The atomic page->_mapcount, like _count, starts from -1:
586 * so that transitions both from it and to it can be tracked,
587 * using atomic_inc_and_test and atomic_add_negative(-1).
589 static inline void reset_page_mapcount(struct page *page)
591 atomic_set(&(page)->_mapcount, -1);
594 static inline int page_mapcount(struct page *page)
596 return atomic_read(&(page)->_mapcount) + 1;
600 * Return true if this page is mapped into pagetables.
602 static inline int page_mapped(struct page *page)
604 return atomic_read(&(page)->_mapcount) >= 0;
608 * Error return values for the *_nopage functions
610 #define NOPAGE_SIGBUS (NULL)
611 #define NOPAGE_OOM ((struct page *) (-1))
614 * Different kinds of faults, as returned by handle_mm_fault().
615 * Used to decide whether a process gets delivered SIGBUS or
616 * just gets major/minor fault counters bumped up.
618 #define VM_FAULT_OOM 0x00
619 #define VM_FAULT_SIGBUS 0x01
620 #define VM_FAULT_MINOR 0x02
621 #define VM_FAULT_MAJOR 0x03
624 * Special case for get_user_pages.
625 * Must be in a distinct bit from the above VM_FAULT_ flags.
627 #define VM_FAULT_WRITE 0x10
629 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
631 extern void show_free_areas(void);
634 struct page *shmem_nopage(struct vm_area_struct *vma,
635 unsigned long address, int *type);
636 int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
637 struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
639 int shmem_lock(struct file *file, int lock, struct user_struct *user);
641 #define shmem_nopage filemap_nopage
643 static inline int shmem_lock(struct file *file, int lock,
644 struct user_struct *user)
649 static inline int shmem_set_policy(struct vm_area_struct *vma,
650 struct mempolicy *new)
655 static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
661 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
662 extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
664 int shmem_zero_setup(struct vm_area_struct *);
667 extern unsigned long shmem_get_unmapped_area(struct file *file,
671 unsigned long flags);
674 static inline int can_do_mlock(void)
676 if (capable(CAP_IPC_LOCK))
678 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
682 extern int user_shm_lock(size_t, struct user_struct *);
683 extern void user_shm_unlock(size_t, struct user_struct *);
686 * Parameter block passed down to zap_pte_range in exceptional cases.
689 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
690 struct address_space *check_mapping; /* Check page->mapping if set */
691 pgoff_t first_index; /* Lowest page->index to unmap */
692 pgoff_t last_index; /* Highest page->index to unmap */
693 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
694 unsigned long truncate_count; /* Compare vm_truncate_count */
697 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
698 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
699 unsigned long size, struct zap_details *);
700 unsigned long unmap_vmas(struct mmu_gather **tlb,
701 struct vm_area_struct *start_vma, unsigned long start_addr,
702 unsigned long end_addr, unsigned long *nr_accounted,
703 struct zap_details *);
704 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
705 unsigned long end, unsigned long floor, unsigned long ceiling);
706 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
707 unsigned long floor, unsigned long ceiling);
708 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
709 struct vm_area_struct *vma);
710 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
711 unsigned long size, pgprot_t prot);
712 void unmap_mapping_range(struct address_space *mapping,
713 loff_t const holebegin, loff_t const holelen, int even_cows);
715 static inline void unmap_shared_mapping_range(struct address_space *mapping,
716 loff_t const holebegin, loff_t const holelen)
718 unmap_mapping_range(mapping, holebegin, holelen, 0);
721 extern int vmtruncate(struct inode * inode, loff_t offset);
722 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
723 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
724 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
727 extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
728 unsigned long address, int write_access);
730 static inline int handle_mm_fault(struct mm_struct *mm,
731 struct vm_area_struct *vma, unsigned long address,
734 return __handle_mm_fault(mm, vma, address, write_access) &
738 static inline int handle_mm_fault(struct mm_struct *mm,
739 struct vm_area_struct *vma, unsigned long address,
742 /* should never happen if there's no MMU */
744 return VM_FAULT_SIGBUS;
748 extern int make_pages_present(unsigned long addr, unsigned long end);
749 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
750 void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
752 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
753 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
754 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
756 int __set_page_dirty_buffers(struct page *page);
757 int __set_page_dirty_nobuffers(struct page *page);
758 int redirty_page_for_writepage(struct writeback_control *wbc,
760 int FASTCALL(set_page_dirty(struct page *page));
761 int set_page_dirty_lock(struct page *page);
762 int clear_page_dirty_for_io(struct page *page);
764 extern unsigned long do_mremap(unsigned long addr,
765 unsigned long old_len, unsigned long new_len,
766 unsigned long flags, unsigned long new_addr);
769 * Prototype to add a shrinker callback for ageable caches.
771 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
772 * scan `nr_to_scan' objects, attempting to free them.
774 * The callback must return the number of objects which remain in the cache.
776 * The callback will be passed nr_to_scan == 0 when the VM is querying the
777 * cache size, so a fastpath for that case is appropriate.
779 typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
782 * Add an aging callback. The int is the number of 'seeks' it takes
783 * to recreate one of the objects that these functions age.
786 #define DEFAULT_SEEKS 2
788 extern struct shrinker *set_shrinker(int, shrinker_t);
789 extern void remove_shrinker(struct shrinker *shrinker);
791 extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
793 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
794 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
795 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
796 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
799 * The following ifdef needed to get the 4level-fixup.h header to work.
800 * Remove it when 4level-fixup.h has been removed.
802 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
803 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
805 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
806 NULL: pud_offset(pgd, address);
809 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
811 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
812 NULL: pmd_offset(pud, address);
814 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
816 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
818 * We tuck a spinlock to guard each pagetable page into its struct page,
819 * at page->private, with BUILD_BUG_ON to make sure that this will not
820 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
821 * When freeing, reset page->mapping so free_pages_check won't complain.
823 #define __pte_lockptr(page) &((page)->ptl)
824 #define pte_lock_init(_page) do { \
825 spin_lock_init(__pte_lockptr(_page)); \
827 #define pte_lock_deinit(page) ((page)->mapping = NULL)
828 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
831 * We use mm->page_table_lock to guard all pagetable pages of the mm.
833 #define pte_lock_init(page) do {} while (0)
834 #define pte_lock_deinit(page) do {} while (0)
835 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
836 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
838 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
840 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
841 pte_t *__pte = pte_offset_map(pmd, address); \
847 #define pte_unmap_unlock(pte, ptl) do { \
852 #define pte_alloc_map(mm, pmd, address) \
853 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
854 NULL: pte_offset_map(pmd, address))
856 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
857 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
858 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
860 #define pte_alloc_kernel(pmd, address) \
861 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
862 NULL: pte_offset_kernel(pmd, address))
864 extern void free_area_init(unsigned long * zones_size);
865 extern void free_area_init_node(int nid, pg_data_t *pgdat,
866 unsigned long * zones_size, unsigned long zone_start_pfn,
867 unsigned long *zholes_size);
868 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
869 extern void setup_per_zone_pages_min(void);
870 extern void mem_init(void);
871 extern void show_mem(void);
872 extern void si_meminfo(struct sysinfo * val);
873 extern void si_meminfo_node(struct sysinfo *val, int nid);
876 extern void setup_per_cpu_pageset(void);
878 static inline void setup_per_cpu_pageset(void) {}
882 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
883 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
884 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
885 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
886 struct prio_tree_iter *iter);
888 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
889 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
890 (vma = vma_prio_tree_next(vma, iter)); )
892 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
893 struct list_head *list)
895 vma->shared.vm_set.parent = NULL;
896 list_add_tail(&vma->shared.vm_set.list, list);
900 extern int __vm_enough_memory(long pages, int cap_sys_admin);
901 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
902 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
903 extern struct vm_area_struct *vma_merge(struct mm_struct *,
904 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
905 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
907 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
908 extern int split_vma(struct mm_struct *,
909 struct vm_area_struct *, unsigned long addr, int new_below);
910 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
911 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
912 struct rb_node **, struct rb_node *);
913 extern void unlink_file_vma(struct vm_area_struct *);
914 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
915 unsigned long addr, unsigned long len, pgoff_t pgoff);
916 extern void exit_mmap(struct mm_struct *);
917 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
919 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
921 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
922 unsigned long len, unsigned long prot,
923 unsigned long flag, unsigned long pgoff);
925 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
926 unsigned long len, unsigned long prot,
927 unsigned long flag, unsigned long offset)
929 unsigned long ret = -EINVAL;
930 if ((offset + PAGE_ALIGN(len)) < offset)
932 if (!(offset & ~PAGE_MASK))
933 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
938 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
940 extern unsigned long do_brk(unsigned long, unsigned long);
943 extern unsigned long page_unuse(struct page *);
944 extern void truncate_inode_pages(struct address_space *, loff_t);
945 extern void truncate_inode_pages_range(struct address_space *,
946 loff_t lstart, loff_t lend);
948 /* generic vm_area_ops exported for stackable file systems */
949 extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
950 extern int filemap_populate(struct vm_area_struct *, unsigned long,
951 unsigned long, pgprot_t, unsigned long, int);
953 /* mm/page-writeback.c */
954 int write_one_page(struct page *page, int wait);
957 #define VM_MAX_READAHEAD 128 /* kbytes */
958 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
959 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
960 * turning readahead off */
962 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
963 pgoff_t offset, unsigned long nr_to_read);
964 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
965 pgoff_t offset, unsigned long nr_to_read);
966 unsigned long page_cache_readahead(struct address_space *mapping,
967 struct file_ra_state *ra,
971 void handle_ra_miss(struct address_space *mapping,
972 struct file_ra_state *ra, pgoff_t offset);
973 unsigned long max_sane_readahead(unsigned long nr);
975 /* Do stack extension */
976 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
978 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
981 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
982 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
983 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
984 struct vm_area_struct **pprev);
986 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
987 NULL if none. Assume start_addr < end_addr. */
988 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
990 struct vm_area_struct * vma = find_vma(mm,start_addr);
992 if (vma && end_addr <= vma->vm_start)
997 static inline unsigned long vma_pages(struct vm_area_struct *vma)
999 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1002 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1003 struct page *vmalloc_to_page(void *addr);
1004 unsigned long vmalloc_to_pfn(void *addr);
1005 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1006 unsigned long pfn, unsigned long size, pgprot_t);
1007 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1009 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1010 unsigned int foll_flags);
1011 #define FOLL_WRITE 0x01 /* check pte is writable */
1012 #define FOLL_TOUCH 0x02 /* mark page accessed */
1013 #define FOLL_GET 0x04 /* do get_page on page */
1014 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1016 #ifdef CONFIG_PROC_FS
1017 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1019 static inline void vm_stat_account(struct mm_struct *mm,
1020 unsigned long flags, struct file *file, long pages)
1023 #endif /* CONFIG_PROC_FS */
1025 #ifndef CONFIG_DEBUG_PAGEALLOC
1027 kernel_map_pages(struct page *page, int numpages, int enable)
1029 if (!PageHighMem(page) && !enable)
1030 mutex_debug_check_no_locks_freed(page_address(page),
1031 numpages * PAGE_SIZE);
1035 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1036 #ifdef __HAVE_ARCH_GATE_AREA
1037 int in_gate_area_no_task(unsigned long addr);
1038 int in_gate_area(struct task_struct *task, unsigned long addr);
1040 int in_gate_area_no_task(unsigned long addr);
1041 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1042 #endif /* __HAVE_ARCH_GATE_AREA */
1044 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1045 #define OOM_DISABLE -17
1047 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1048 void __user *, size_t *, loff_t *);
1049 int shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1050 unsigned long lru_pages);
1051 void drop_pagecache(void);
1052 void drop_slab(void);
1054 #endif /* __KERNEL__ */
1055 #endif /* _LINUX_MM_H */