4 #include <linux/sched.h>
5 #include <linux/errno.h>
6 #include <linux/capability.h>
10 #include <linux/gfp.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/prio_tree.h>
16 #include <linux/mutex.h>
17 #include <linux/debug_locks.h>
18 #include <linux/backing-dev.h>
23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24 extern unsigned long max_mapnr;
27 extern unsigned long num_physpages;
28 extern void * high_memory;
29 extern unsigned long vmalloc_earlyreserve;
30 extern int page_cluster;
33 extern int sysctl_legacy_va_layout;
35 #define sysctl_legacy_va_layout 0
39 #include <asm/pgtable.h>
40 #include <asm/processor.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_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
150 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
152 #define VM_EXECUTABLE 0x00001000
153 #define VM_LOCKED 0x00002000
154 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
156 /* Used by sys_madvise() */
157 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
158 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
160 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
161 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
162 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
163 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
164 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
165 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
166 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
167 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
169 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
170 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
173 #ifdef CONFIG_STACK_GROWSUP
174 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
179 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
180 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
181 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
182 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
183 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
186 * mapping from the currently active vm_flags protection bits (the
187 * low four bits) to a page protection mask..
189 extern pgprot_t protection_map[16];
193 * These are the virtual MM functions - opening of an area, closing and
194 * unmapping it (needed to keep files on disk up-to-date etc), pointer
195 * to the functions called when a no-page or a wp-page exception occurs.
197 struct vm_operations_struct {
198 void (*open)(struct vm_area_struct * area);
199 void (*close)(struct vm_area_struct * area);
200 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
201 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
203 /* notification that a previously read-only page is about to become
204 * writable, if an error is returned it will cause a SIGBUS */
205 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
207 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
208 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
210 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
211 const nodemask_t *to, unsigned long flags);
219 * Each physical page in the system has a struct page associated with
220 * it to keep track of whatever it is we are using the page for at the
221 * moment. Note that we have no way to track which tasks are using
225 unsigned long flags; /* Atomic flags, some possibly
226 * updated asynchronously */
227 atomic_t _count; /* Usage count, see below. */
228 atomic_t _mapcount; /* Count of ptes mapped in mms,
229 * to show when page is mapped
230 * & limit reverse map searches.
234 unsigned long private; /* Mapping-private opaque data:
235 * usually used for buffer_heads
236 * if PagePrivate set; used for
237 * swp_entry_t if PageSwapCache;
238 * indicates order in the buddy
239 * system if PG_buddy is set.
241 struct address_space *mapping; /* If low bit clear, points to
242 * inode address_space, or NULL.
243 * If page mapped as anonymous
244 * memory, low bit is set, and
245 * it points to anon_vma object:
246 * see PAGE_MAPPING_ANON below.
249 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
253 pgoff_t index; /* Our offset within mapping. */
254 struct list_head lru; /* Pageout list, eg. active_list
255 * protected by zone->lru_lock !
258 * On machines where all RAM is mapped into kernel address space,
259 * we can simply calculate the virtual address. On machines with
260 * highmem some memory is mapped into kernel virtual memory
261 * dynamically, so we need a place to store that address.
262 * Note that this field could be 16 bits on x86 ... ;)
264 * Architectures with slow multiplication can define
265 * WANT_PAGE_VIRTUAL in asm/page.h
267 #if defined(WANT_PAGE_VIRTUAL)
268 void *virtual; /* Kernel virtual address (NULL if
269 not kmapped, ie. highmem) */
270 #endif /* WANT_PAGE_VIRTUAL */
273 #define page_private(page) ((page)->private)
274 #define set_page_private(page, v) ((page)->private = (v))
277 * FIXME: take this include out, include page-flags.h in
278 * files which need it (119 of them)
280 #include <linux/page-flags.h>
282 #ifdef CONFIG_DEBUG_VM
283 #define VM_BUG_ON(cond) BUG_ON(cond)
285 #define VM_BUG_ON(condition) do { } while(0)
289 * Methods to modify the page usage count.
291 * What counts for a page usage:
292 * - cache mapping (page->mapping)
293 * - private data (page->private)
294 * - page mapped in a task's page tables, each mapping
295 * is counted separately
297 * Also, many kernel routines increase the page count before a critical
298 * routine so they can be sure the page doesn't go away from under them.
302 * Drop a ref, return true if the logical refcount fell to zero (the page has
305 static inline int put_page_testzero(struct page *page)
307 VM_BUG_ON(atomic_read(&page->_count) == 0);
308 return atomic_dec_and_test(&page->_count);
312 * Try to grab a ref unless the page has a refcount of zero, return false if
315 static inline int get_page_unless_zero(struct page *page)
317 VM_BUG_ON(PageCompound(page));
318 return atomic_inc_not_zero(&page->_count);
321 extern void FASTCALL(__page_cache_release(struct page *));
323 static inline int page_count(struct page *page)
325 if (unlikely(PageCompound(page)))
326 page = (struct page *)page_private(page);
327 return atomic_read(&page->_count);
330 static inline void get_page(struct page *page)
332 if (unlikely(PageCompound(page)))
333 page = (struct page *)page_private(page);
334 VM_BUG_ON(atomic_read(&page->_count) == 0);
335 atomic_inc(&page->_count);
339 * Setup the page count before being freed into the page allocator for
340 * the first time (boot or memory hotplug)
342 static inline void init_page_count(struct page *page)
344 atomic_set(&page->_count, 1);
347 void put_page(struct page *page);
348 void put_pages_list(struct list_head *pages);
350 void split_page(struct page *page, unsigned int order);
353 * Multiple processes may "see" the same page. E.g. for untouched
354 * mappings of /dev/null, all processes see the same page full of
355 * zeroes, and text pages of executables and shared libraries have
356 * only one copy in memory, at most, normally.
358 * For the non-reserved pages, page_count(page) denotes a reference count.
359 * page_count() == 0 means the page is free. page->lru is then used for
360 * freelist management in the buddy allocator.
361 * page_count() == 1 means the page is used for exactly one purpose
362 * (e.g. a private data page of one process).
364 * A page may be used for kmalloc() or anyone else who does a
365 * __get_free_page(). In this case the page_count() is at least 1, and
366 * all other fields are unused but should be 0 or NULL. The
367 * management of this page is the responsibility of the one who uses
370 * The other pages (we may call them "process pages") are completely
371 * managed by the Linux memory manager: I/O, buffers, swapping etc.
372 * The following discussion applies only to them.
374 * A page may belong to an inode's memory mapping. In this case,
375 * page->mapping is the pointer to the inode, and page->index is the
376 * file offset of the page, in units of PAGE_CACHE_SIZE.
378 * A page contains an opaque `private' member, which belongs to the
379 * page's address_space. Usually, this is the address of a circular
380 * list of the page's disk buffers.
382 * For pages belonging to inodes, the page_count() is the number of
383 * attaches, plus 1 if `private' contains something, plus one for
384 * the page cache itself.
386 * Instead of keeping dirty/clean pages in per address-space lists, we instead
387 * now tag pages as dirty/under writeback in the radix tree.
389 * There is also a per-mapping radix tree mapping index to the page
390 * in memory if present. The tree is rooted at mapping->root.
392 * All process pages can do I/O:
393 * - inode pages may need to be read from disk,
394 * - inode pages which have been modified and are MAP_SHARED may need
395 * to be written to disk,
396 * - private pages which have been modified may need to be swapped out
397 * to swap space and (later) to be read back into memory.
401 * The zone field is never updated after free_area_init_core()
402 * sets it, so none of the operations on it need to be atomic.
407 * page->flags layout:
409 * There are three possibilities for how page->flags get
410 * laid out. The first is for the normal case, without
411 * sparsemem. The second is for sparsemem when there is
412 * plenty of space for node and section. The last is when
413 * we have run out of space and have to fall back to an
414 * alternate (slower) way of determining the node.
416 * No sparsemem: | NODE | ZONE | ... | FLAGS |
417 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
418 * no space for node: | SECTION | ZONE | ... | FLAGS |
420 #ifdef CONFIG_SPARSEMEM
421 #define SECTIONS_WIDTH SECTIONS_SHIFT
423 #define SECTIONS_WIDTH 0
426 #define ZONES_WIDTH ZONES_SHIFT
428 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
429 #define NODES_WIDTH NODES_SHIFT
431 #define NODES_WIDTH 0
434 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
435 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
436 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
437 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
440 * We are going to use the flags for the page to node mapping if its in
441 * there. This includes the case where there is no node, so it is implicit.
443 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
445 #ifndef PFN_SECTION_SHIFT
446 #define PFN_SECTION_SHIFT 0
450 * Define the bit shifts to access each section. For non-existant
451 * sections we define the shift as 0; that plus a 0 mask ensures
452 * the compiler will optimise away reference to them.
454 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
455 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
456 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
458 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
460 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
462 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
464 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
466 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
467 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
470 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
471 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
472 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
473 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
475 static inline unsigned long page_zonenum(struct page *page)
477 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
481 extern struct zone *zone_table[];
483 static inline int page_zone_id(struct page *page)
485 return (page->flags >> ZONETABLE_PGSHIFT) & ZONETABLE_MASK;
487 static inline struct zone *page_zone(struct page *page)
489 return zone_table[page_zone_id(page)];
492 static inline unsigned long page_to_nid(struct page *page)
495 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
497 return page_zone(page)->zone_pgdat->node_id;
499 static inline unsigned long page_to_section(struct page *page)
501 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
504 static inline void set_page_zone(struct page *page, unsigned long zone)
506 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
507 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
509 static inline void set_page_node(struct page *page, unsigned long node)
511 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
512 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
514 static inline void set_page_section(struct page *page, unsigned long section)
516 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
517 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
520 static inline void set_page_links(struct page *page, unsigned long zone,
521 unsigned long node, unsigned long pfn)
523 set_page_zone(page, zone);
524 set_page_node(page, node);
525 set_page_section(page, pfn_to_section_nr(pfn));
529 * Some inline functions in vmstat.h depend on page_zone()
531 #include <linux/vmstat.h>
533 #ifndef CONFIG_DISCONTIGMEM
534 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
535 extern struct page *mem_map;
538 static __always_inline void *lowmem_page_address(struct page *page)
540 return __va(page_to_pfn(page) << PAGE_SHIFT);
543 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
544 #define HASHED_PAGE_VIRTUAL
547 #if defined(WANT_PAGE_VIRTUAL)
548 #define page_address(page) ((page)->virtual)
549 #define set_page_address(page, address) \
551 (page)->virtual = (address); \
553 #define page_address_init() do { } while(0)
556 #if defined(HASHED_PAGE_VIRTUAL)
557 void *page_address(struct page *page);
558 void set_page_address(struct page *page, void *virtual);
559 void page_address_init(void);
562 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
563 #define page_address(page) lowmem_page_address(page)
564 #define set_page_address(page, address) do { } while(0)
565 #define page_address_init() do { } while(0)
569 * On an anonymous page mapped into a user virtual memory area,
570 * page->mapping points to its anon_vma, not to a struct address_space;
571 * with the PAGE_MAPPING_ANON bit set to distinguish it.
573 * Please note that, confusingly, "page_mapping" refers to the inode
574 * address_space which maps the page from disk; whereas "page_mapped"
575 * refers to user virtual address space into which the page is mapped.
577 #define PAGE_MAPPING_ANON 1
579 extern struct address_space swapper_space;
580 static inline struct address_space *page_mapping(struct page *page)
582 struct address_space *mapping = page->mapping;
584 if (unlikely(PageSwapCache(page)))
585 mapping = &swapper_space;
586 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
591 static inline int PageAnon(struct page *page)
593 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
597 * Return the pagecache index of the passed page. Regular pagecache pages
598 * use ->index whereas swapcache pages use ->private
600 static inline pgoff_t page_index(struct page *page)
602 if (unlikely(PageSwapCache(page)))
603 return page_private(page);
608 * The atomic page->_mapcount, like _count, starts from -1:
609 * so that transitions both from it and to it can be tracked,
610 * using atomic_inc_and_test and atomic_add_negative(-1).
612 static inline void reset_page_mapcount(struct page *page)
614 atomic_set(&(page)->_mapcount, -1);
617 static inline int page_mapcount(struct page *page)
619 return atomic_read(&(page)->_mapcount) + 1;
623 * Return true if this page is mapped into pagetables.
625 static inline int page_mapped(struct page *page)
627 return atomic_read(&(page)->_mapcount) >= 0;
631 * Error return values for the *_nopage functions
633 #define NOPAGE_SIGBUS (NULL)
634 #define NOPAGE_OOM ((struct page *) (-1))
637 * Different kinds of faults, as returned by handle_mm_fault().
638 * Used to decide whether a process gets delivered SIGBUS or
639 * just gets major/minor fault counters bumped up.
641 #define VM_FAULT_OOM 0x00
642 #define VM_FAULT_SIGBUS 0x01
643 #define VM_FAULT_MINOR 0x02
644 #define VM_FAULT_MAJOR 0x03
647 * Special case for get_user_pages.
648 * Must be in a distinct bit from the above VM_FAULT_ flags.
650 #define VM_FAULT_WRITE 0x10
652 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
654 extern void show_free_areas(void);
657 struct page *shmem_nopage(struct vm_area_struct *vma,
658 unsigned long address, int *type);
659 int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
660 struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
662 int shmem_lock(struct file *file, int lock, struct user_struct *user);
664 #define shmem_nopage filemap_nopage
666 static inline int shmem_lock(struct file *file, int lock,
667 struct user_struct *user)
672 static inline int shmem_set_policy(struct vm_area_struct *vma,
673 struct mempolicy *new)
678 static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
684 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
685 extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
687 int shmem_zero_setup(struct vm_area_struct *);
690 extern unsigned long shmem_get_unmapped_area(struct file *file,
694 unsigned long flags);
697 static inline int can_do_mlock(void)
699 if (capable(CAP_IPC_LOCK))
701 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
705 extern int user_shm_lock(size_t, struct user_struct *);
706 extern void user_shm_unlock(size_t, struct user_struct *);
709 * Parameter block passed down to zap_pte_range in exceptional cases.
712 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
713 struct address_space *check_mapping; /* Check page->mapping if set */
714 pgoff_t first_index; /* Lowest page->index to unmap */
715 pgoff_t last_index; /* Highest page->index to unmap */
716 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
717 unsigned long truncate_count; /* Compare vm_truncate_count */
720 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
721 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
722 unsigned long size, struct zap_details *);
723 unsigned long unmap_vmas(struct mmu_gather **tlb,
724 struct vm_area_struct *start_vma, unsigned long start_addr,
725 unsigned long end_addr, unsigned long *nr_accounted,
726 struct zap_details *);
727 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
728 unsigned long end, unsigned long floor, unsigned long ceiling);
729 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
730 unsigned long floor, unsigned long ceiling);
731 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
732 struct vm_area_struct *vma);
733 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
734 unsigned long size, pgprot_t prot);
735 void unmap_mapping_range(struct address_space *mapping,
736 loff_t const holebegin, loff_t const holelen, int even_cows);
738 static inline void unmap_shared_mapping_range(struct address_space *mapping,
739 loff_t const holebegin, loff_t const holelen)
741 unmap_mapping_range(mapping, holebegin, holelen, 0);
744 extern int vmtruncate(struct inode * inode, loff_t offset);
745 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
746 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
747 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
750 extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
751 unsigned long address, int write_access);
753 static inline int handle_mm_fault(struct mm_struct *mm,
754 struct vm_area_struct *vma, unsigned long address,
757 return __handle_mm_fault(mm, vma, address, write_access) &
761 static inline int handle_mm_fault(struct mm_struct *mm,
762 struct vm_area_struct *vma, unsigned long address,
765 /* should never happen if there's no MMU */
767 return VM_FAULT_SIGBUS;
771 extern int make_pages_present(unsigned long addr, unsigned long end);
772 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
773 void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
775 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
776 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
777 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
779 int __set_page_dirty_buffers(struct page *page);
780 int __set_page_dirty_nobuffers(struct page *page);
781 int redirty_page_for_writepage(struct writeback_control *wbc,
783 int FASTCALL(set_page_dirty(struct page *page));
784 int set_page_dirty_lock(struct page *page);
785 int clear_page_dirty_for_io(struct page *page);
787 extern unsigned long do_mremap(unsigned long addr,
788 unsigned long old_len, unsigned long new_len,
789 unsigned long flags, unsigned long new_addr);
792 * Prototype to add a shrinker callback for ageable caches.
794 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
795 * scan `nr_to_scan' objects, attempting to free them.
797 * The callback must return the number of objects which remain in the cache.
799 * The callback will be passed nr_to_scan == 0 when the VM is querying the
800 * cache size, so a fastpath for that case is appropriate.
802 typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
805 * Add an aging callback. The int is the number of 'seeks' it takes
806 * to recreate one of the objects that these functions age.
809 #define DEFAULT_SEEKS 2
811 extern struct shrinker *set_shrinker(int, shrinker_t);
812 extern void remove_shrinker(struct shrinker *shrinker);
815 * Some shared mappigns will want the pages marked read-only
816 * to track write events. If so, we'll downgrade vm_page_prot
817 * to the private version (using protection_map[] without the
820 static inline int vma_wants_writenotify(struct vm_area_struct *vma)
822 unsigned int vm_flags = vma->vm_flags;
824 /* If it was private or non-writable, the write bit is already clear */
825 if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
828 /* The backer wishes to know when pages are first written to? */
829 if (vma->vm_ops && vma->vm_ops->page_mkwrite)
832 /* The open routine did something to the protections already? */
833 if (pgprot_val(vma->vm_page_prot) !=
834 pgprot_val(protection_map[vm_flags &
835 (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]))
838 /* Specialty mapping? */
839 if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE))
842 /* Can the mapping track the dirty pages? */
843 return vma->vm_file && vma->vm_file->f_mapping &&
844 mapping_cap_account_dirty(vma->vm_file->f_mapping);
847 extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
849 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
850 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
851 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
852 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
855 * The following ifdef needed to get the 4level-fixup.h header to work.
856 * Remove it when 4level-fixup.h has been removed.
858 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
859 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
861 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
862 NULL: pud_offset(pgd, address);
865 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
867 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
868 NULL: pmd_offset(pud, address);
870 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
872 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
874 * We tuck a spinlock to guard each pagetable page into its struct page,
875 * at page->private, with BUILD_BUG_ON to make sure that this will not
876 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
877 * When freeing, reset page->mapping so free_pages_check won't complain.
879 #define __pte_lockptr(page) &((page)->ptl)
880 #define pte_lock_init(_page) do { \
881 spin_lock_init(__pte_lockptr(_page)); \
883 #define pte_lock_deinit(page) ((page)->mapping = NULL)
884 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
887 * We use mm->page_table_lock to guard all pagetable pages of the mm.
889 #define pte_lock_init(page) do {} while (0)
890 #define pte_lock_deinit(page) do {} while (0)
891 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
892 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
894 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
896 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
897 pte_t *__pte = pte_offset_map(pmd, address); \
903 #define pte_unmap_unlock(pte, ptl) do { \
908 #define pte_alloc_map(mm, pmd, address) \
909 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
910 NULL: pte_offset_map(pmd, address))
912 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
913 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
914 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
916 #define pte_alloc_kernel(pmd, address) \
917 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
918 NULL: pte_offset_kernel(pmd, address))
920 extern void free_area_init(unsigned long * zones_size);
921 extern void free_area_init_node(int nid, pg_data_t *pgdat,
922 unsigned long * zones_size, unsigned long zone_start_pfn,
923 unsigned long *zholes_size);
924 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
925 extern void setup_per_zone_pages_min(void);
926 extern void mem_init(void);
927 extern void show_mem(void);
928 extern void si_meminfo(struct sysinfo * val);
929 extern void si_meminfo_node(struct sysinfo *val, int nid);
932 extern void setup_per_cpu_pageset(void);
934 static inline void setup_per_cpu_pageset(void) {}
938 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
939 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
940 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
941 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
942 struct prio_tree_iter *iter);
944 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
945 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
946 (vma = vma_prio_tree_next(vma, iter)); )
948 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
949 struct list_head *list)
951 vma->shared.vm_set.parent = NULL;
952 list_add_tail(&vma->shared.vm_set.list, list);
956 extern int __vm_enough_memory(long pages, int cap_sys_admin);
957 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
958 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
959 extern struct vm_area_struct *vma_merge(struct mm_struct *,
960 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
961 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
963 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
964 extern int split_vma(struct mm_struct *,
965 struct vm_area_struct *, unsigned long addr, int new_below);
966 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
967 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
968 struct rb_node **, struct rb_node *);
969 extern void unlink_file_vma(struct vm_area_struct *);
970 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
971 unsigned long addr, unsigned long len, pgoff_t pgoff);
972 extern void exit_mmap(struct mm_struct *);
973 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
975 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
977 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
978 unsigned long len, unsigned long prot,
979 unsigned long flag, unsigned long pgoff);
981 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
982 unsigned long len, unsigned long prot,
983 unsigned long flag, unsigned long offset)
985 unsigned long ret = -EINVAL;
986 if ((offset + PAGE_ALIGN(len)) < offset)
988 if (!(offset & ~PAGE_MASK))
989 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
994 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
996 extern unsigned long do_brk(unsigned long, unsigned long);
999 extern unsigned long page_unuse(struct page *);
1000 extern void truncate_inode_pages(struct address_space *, loff_t);
1001 extern void truncate_inode_pages_range(struct address_space *,
1002 loff_t lstart, loff_t lend);
1004 /* generic vm_area_ops exported for stackable file systems */
1005 extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
1006 extern int filemap_populate(struct vm_area_struct *, unsigned long,
1007 unsigned long, pgprot_t, unsigned long, int);
1009 /* mm/page-writeback.c */
1010 int write_one_page(struct page *page, int wait);
1013 #define VM_MAX_READAHEAD 128 /* kbytes */
1014 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1015 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1016 * turning readahead off */
1018 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1019 pgoff_t offset, unsigned long nr_to_read);
1020 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1021 pgoff_t offset, unsigned long nr_to_read);
1022 unsigned long page_cache_readahead(struct address_space *mapping,
1023 struct file_ra_state *ra,
1026 unsigned long size);
1027 void handle_ra_miss(struct address_space *mapping,
1028 struct file_ra_state *ra, pgoff_t offset);
1029 unsigned long max_sane_readahead(unsigned long nr);
1031 /* Do stack extension */
1032 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1034 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1037 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1038 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1039 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1040 struct vm_area_struct **pprev);
1042 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1043 NULL if none. Assume start_addr < end_addr. */
1044 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1046 struct vm_area_struct * vma = find_vma(mm,start_addr);
1048 if (vma && end_addr <= vma->vm_start)
1053 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1055 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1058 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1059 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1060 struct page *vmalloc_to_page(void *addr);
1061 unsigned long vmalloc_to_pfn(void *addr);
1062 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1063 unsigned long pfn, unsigned long size, pgprot_t);
1064 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1066 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1067 unsigned int foll_flags);
1068 #define FOLL_WRITE 0x01 /* check pte is writable */
1069 #define FOLL_TOUCH 0x02 /* mark page accessed */
1070 #define FOLL_GET 0x04 /* do get_page on page */
1071 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1073 #ifdef CONFIG_PROC_FS
1074 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1076 static inline void vm_stat_account(struct mm_struct *mm,
1077 unsigned long flags, struct file *file, long pages)
1080 #endif /* CONFIG_PROC_FS */
1082 #ifndef CONFIG_DEBUG_PAGEALLOC
1084 kernel_map_pages(struct page *page, int numpages, int enable)
1086 if (!PageHighMem(page) && !enable)
1087 debug_check_no_locks_freed(page_address(page),
1088 numpages * PAGE_SIZE);
1092 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1093 #ifdef __HAVE_ARCH_GATE_AREA
1094 int in_gate_area_no_task(unsigned long addr);
1095 int in_gate_area(struct task_struct *task, unsigned long addr);
1097 int in_gate_area_no_task(unsigned long addr);
1098 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1099 #endif /* __HAVE_ARCH_GATE_AREA */
1101 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1102 #define OOM_DISABLE -17
1104 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1105 void __user *, size_t *, loff_t *);
1106 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1107 unsigned long lru_pages);
1108 void drop_pagecache(void);
1109 void drop_slab(void);
1112 #define randomize_va_space 0
1114 extern int randomize_va_space;
1117 const char *arch_vma_name(struct vm_area_struct *vma);
1119 #endif /* __KERNEL__ */
1120 #endif /* _LINUX_MM_H */