4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
14 #include <linux/mutex.h>
15 #include <linux/debug_locks.h>
16 #include <linux/backing-dev.h>
17 #include <linux/mm_types.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 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>
41 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44 * Linux kernel virtual memory manager primitives.
45 * The idea being to have a "virtual" mm in the same way
46 * we have a virtual fs - giving a cleaner interface to the
47 * mm details, and allowing different kinds of memory mappings
48 * (from shared memory to executable loading to arbitrary
53 * This struct defines a memory VMM memory area. There is one of these
54 * per VM-area/task. A VM area is any part of the process virtual memory
55 * space that has a special rule for the page-fault handlers (ie a shared
56 * library, the executable area etc).
58 struct vm_area_struct {
59 struct mm_struct * vm_mm; /* The address space we belong to. */
60 unsigned long vm_start; /* Our start address within vm_mm. */
61 unsigned long vm_end; /* The first byte after our end address
64 /* linked list of VM areas per task, sorted by address */
65 struct vm_area_struct *vm_next;
67 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
68 unsigned long vm_flags; /* Flags, listed below. */
73 * For areas with an address space and backing store,
74 * linkage into the address_space->i_mmap prio tree, or
75 * linkage to the list of like vmas hanging off its node, or
76 * linkage of vma in the address_space->i_mmap_nonlinear list.
80 struct list_head list;
81 void *parent; /* aligns with prio_tree_node parent */
82 struct vm_area_struct *head;
85 struct raw_prio_tree_node prio_tree_node;
89 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
90 * list, after a COW of one of the file pages. A MAP_SHARED vma
91 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
92 * or brk vma (with NULL file) can only be in an anon_vma list.
94 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
95 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
97 /* Function pointers to deal with this struct. */
98 struct vm_operations_struct * vm_ops;
100 /* Information about our backing store: */
101 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
102 units, *not* PAGE_CACHE_SIZE */
103 struct file * vm_file; /* File we map to (can be NULL). */
104 void * vm_private_data; /* was vm_pte (shared mem) */
105 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
108 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
111 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
115 extern struct kmem_cache *vm_area_cachep;
118 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
119 * disabled, then there's a single shared list of VMAs maintained by the
120 * system, and mm's subscribe to these individually
122 struct vm_list_struct {
123 struct vm_list_struct *next;
124 struct vm_area_struct *vma;
128 extern struct rb_root nommu_vma_tree;
129 extern struct rw_semaphore nommu_vma_sem;
131 extern unsigned int kobjsize(const void *objp);
137 #define VM_READ 0x00000001 /* currently active flags */
138 #define VM_WRITE 0x00000002
139 #define VM_EXEC 0x00000004
140 #define VM_SHARED 0x00000008
142 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
143 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
144 #define VM_MAYWRITE 0x00000020
145 #define VM_MAYEXEC 0x00000040
146 #define VM_MAYSHARE 0x00000080
148 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
149 #define VM_GROWSUP 0x00000200
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 */
169 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
171 #define VM_CAN_INVALIDATE 0x08000000 /* The mapping may be invalidated,
172 * eg. truncate or invalidate_inode_*.
173 * In this case, do_no_page must
174 * return with the page locked.
176 #define VM_CAN_NONLINEAR 0x10000000 /* Has ->fault & does nonlinear pages */
178 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
179 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
182 #ifdef CONFIG_STACK_GROWSUP
183 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
185 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
188 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
189 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
190 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
191 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
192 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
195 * mapping from the currently active vm_flags protection bits (the
196 * low four bits) to a page protection mask..
198 extern pgprot_t protection_map[16];
200 #define FAULT_FLAG_WRITE 0x01
201 #define FAULT_FLAG_NONLINEAR 0x02
204 * fault_data is filled in the the pagefault handler and passed to the
205 * vma's ->fault function. That function is responsible for filling in
206 * 'type', which is the type of fault if a page is returned, or the type
207 * of error if NULL is returned.
209 * pgoff should be used in favour of address, if possible. If pgoff is
210 * used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get
211 * nonlinear mapping support.
214 unsigned long address;
221 * These are the virtual MM functions - opening of an area, closing and
222 * unmapping it (needed to keep files on disk up-to-date etc), pointer
223 * to the functions called when a no-page or a wp-page exception occurs.
225 struct vm_operations_struct {
226 void (*open)(struct vm_area_struct * area);
227 void (*close)(struct vm_area_struct * area);
228 struct page *(*fault)(struct vm_area_struct *vma,
229 struct fault_data *fdata);
230 struct page *(*nopage)(struct vm_area_struct *area,
231 unsigned long address, int *type);
232 unsigned long (*nopfn)(struct vm_area_struct *area,
233 unsigned long address);
234 int (*populate)(struct vm_area_struct *area, unsigned long address,
235 unsigned long len, pgprot_t prot, unsigned long pgoff,
238 /* notification that a previously read-only page is about to become
239 * writable, if an error is returned it will cause a SIGBUS */
240 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
242 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
243 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
245 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
246 const nodemask_t *to, unsigned long flags);
253 #define page_private(page) ((page)->private)
254 #define set_page_private(page, v) ((page)->private = (v))
257 * FIXME: take this include out, include page-flags.h in
258 * files which need it (119 of them)
260 #include <linux/page-flags.h>
262 #ifdef CONFIG_DEBUG_VM
263 #define VM_BUG_ON(cond) BUG_ON(cond)
265 #define VM_BUG_ON(condition) do { } while(0)
269 * Methods to modify the page usage count.
271 * What counts for a page usage:
272 * - cache mapping (page->mapping)
273 * - private data (page->private)
274 * - page mapped in a task's page tables, each mapping
275 * is counted separately
277 * Also, many kernel routines increase the page count before a critical
278 * routine so they can be sure the page doesn't go away from under them.
282 * Drop a ref, return true if the refcount fell to zero (the page has no users)
284 static inline int put_page_testzero(struct page *page)
286 VM_BUG_ON(atomic_read(&page->_count) == 0);
287 return atomic_dec_and_test(&page->_count);
291 * Try to grab a ref unless the page has a refcount of zero, return false if
294 static inline int get_page_unless_zero(struct page *page)
296 VM_BUG_ON(PageCompound(page));
297 return atomic_inc_not_zero(&page->_count);
300 static inline struct page *compound_head(struct page *page)
302 if (unlikely(PageTail(page)))
303 return page->first_page;
307 static inline int page_count(struct page *page)
309 return atomic_read(&compound_head(page)->_count);
312 static inline void get_page(struct page *page)
314 page = compound_head(page);
315 VM_BUG_ON(atomic_read(&page->_count) == 0);
316 atomic_inc(&page->_count);
319 static inline struct page *virt_to_head_page(const void *x)
321 struct page *page = virt_to_page(x);
322 return compound_head(page);
326 * Setup the page count before being freed into the page allocator for
327 * the first time (boot or memory hotplug)
329 static inline void init_page_count(struct page *page)
331 atomic_set(&page->_count, 1);
334 void put_page(struct page *page);
335 void put_pages_list(struct list_head *pages);
337 void split_page(struct page *page, unsigned int order);
340 * Compound pages have a destructor function. Provide a
341 * prototype for that function and accessor functions.
342 * These are _only_ valid on the head of a PG_compound page.
344 typedef void compound_page_dtor(struct page *);
346 static inline void set_compound_page_dtor(struct page *page,
347 compound_page_dtor *dtor)
349 page[1].lru.next = (void *)dtor;
352 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
354 return (compound_page_dtor *)page[1].lru.next;
357 static inline int compound_order(struct page *page)
361 return (unsigned long)page[1].lru.prev;
364 static inline void set_compound_order(struct page *page, unsigned long order)
366 page[1].lru.prev = (void *)order;
370 * Multiple processes may "see" the same page. E.g. for untouched
371 * mappings of /dev/null, all processes see the same page full of
372 * zeroes, and text pages of executables and shared libraries have
373 * only one copy in memory, at most, normally.
375 * For the non-reserved pages, page_count(page) denotes a reference count.
376 * page_count() == 0 means the page is free. page->lru is then used for
377 * freelist management in the buddy allocator.
378 * page_count() > 0 means the page has been allocated.
380 * Pages are allocated by the slab allocator in order to provide memory
381 * to kmalloc and kmem_cache_alloc. In this case, the management of the
382 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
383 * unless a particular usage is carefully commented. (the responsibility of
384 * freeing the kmalloc memory is the caller's, of course).
386 * A page may be used by anyone else who does a __get_free_page().
387 * In this case, page_count still tracks the references, and should only
388 * be used through the normal accessor functions. The top bits of page->flags
389 * and page->virtual store page management information, but all other fields
390 * are unused and could be used privately, carefully. The management of this
391 * page is the responsibility of the one who allocated it, and those who have
392 * subsequently been given references to it.
394 * The other pages (we may call them "pagecache pages") are completely
395 * managed by the Linux memory manager: I/O, buffers, swapping etc.
396 * The following discussion applies only to them.
398 * A pagecache page contains an opaque `private' member, which belongs to the
399 * page's address_space. Usually, this is the address of a circular list of
400 * the page's disk buffers. PG_private must be set to tell the VM to call
401 * into the filesystem to release these pages.
403 * A page may belong to an inode's memory mapping. In this case, page->mapping
404 * is the pointer to the inode, and page->index is the file offset of the page,
405 * in units of PAGE_CACHE_SIZE.
407 * If pagecache pages are not associated with an inode, they are said to be
408 * anonymous pages. These may become associated with the swapcache, and in that
409 * case PG_swapcache is set, and page->private is an offset into the swapcache.
411 * In either case (swapcache or inode backed), the pagecache itself holds one
412 * reference to the page. Setting PG_private should also increment the
413 * refcount. The each user mapping also has a reference to the page.
415 * The pagecache pages are stored in a per-mapping radix tree, which is
416 * rooted at mapping->page_tree, and indexed by offset.
417 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
418 * lists, we instead now tag pages as dirty/writeback in the radix tree.
420 * All pagecache pages may be subject to I/O:
421 * - inode pages may need to be read from disk,
422 * - inode pages which have been modified and are MAP_SHARED may need
423 * to be written back to the inode on disk,
424 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
425 * modified may need to be swapped out to swap space and (later) to be read
430 * The zone field is never updated after free_area_init_core()
431 * sets it, so none of the operations on it need to be atomic.
436 * page->flags layout:
438 * There are three possibilities for how page->flags get
439 * laid out. The first is for the normal case, without
440 * sparsemem. The second is for sparsemem when there is
441 * plenty of space for node and section. The last is when
442 * we have run out of space and have to fall back to an
443 * alternate (slower) way of determining the node.
445 * No sparsemem: | NODE | ZONE | ... | FLAGS |
446 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
447 * no space for node: | SECTION | ZONE | ... | FLAGS |
449 #ifdef CONFIG_SPARSEMEM
450 #define SECTIONS_WIDTH SECTIONS_SHIFT
452 #define SECTIONS_WIDTH 0
455 #define ZONES_WIDTH ZONES_SHIFT
457 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
458 #define NODES_WIDTH NODES_SHIFT
460 #define NODES_WIDTH 0
463 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
464 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
465 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
466 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
469 * We are going to use the flags for the page to node mapping if its in
470 * there. This includes the case where there is no node, so it is implicit.
472 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
473 #define NODE_NOT_IN_PAGE_FLAGS
476 #ifndef PFN_SECTION_SHIFT
477 #define PFN_SECTION_SHIFT 0
481 * Define the bit shifts to access each section. For non-existant
482 * sections we define the shift as 0; that plus a 0 mask ensures
483 * the compiler will optimise away reference to them.
485 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
486 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
487 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
489 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
490 #ifdef NODE_NOT_IN_PAGEFLAGS
491 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
492 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
493 SECTIONS_PGOFF : ZONES_PGOFF)
495 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
496 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
497 NODES_PGOFF : ZONES_PGOFF)
500 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
502 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
503 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
506 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
507 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
508 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
509 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
511 static inline enum zone_type page_zonenum(struct page *page)
513 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
517 * The identification function is only used by the buddy allocator for
518 * determining if two pages could be buddies. We are not really
519 * identifying a zone since we could be using a the section number
520 * id if we have not node id available in page flags.
521 * We guarantee only that it will return the same value for two
522 * combinable pages in a zone.
524 static inline int page_zone_id(struct page *page)
526 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
529 static inline int zone_to_nid(struct zone *zone)
538 #ifdef NODE_NOT_IN_PAGE_FLAGS
539 extern int page_to_nid(struct page *page);
541 static inline int page_to_nid(struct page *page)
543 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
547 static inline struct zone *page_zone(struct page *page)
549 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
552 static inline unsigned long page_to_section(struct page *page)
554 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
557 static inline void set_page_zone(struct page *page, enum zone_type zone)
559 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
560 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
563 static inline void set_page_node(struct page *page, unsigned long node)
565 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
566 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
569 static inline void set_page_section(struct page *page, unsigned long section)
571 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
572 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
575 static inline void set_page_links(struct page *page, enum zone_type zone,
576 unsigned long node, unsigned long pfn)
578 set_page_zone(page, zone);
579 set_page_node(page, node);
580 set_page_section(page, pfn_to_section_nr(pfn));
584 * Some inline functions in vmstat.h depend on page_zone()
586 #include <linux/vmstat.h>
588 static __always_inline void *lowmem_page_address(struct page *page)
590 return __va(page_to_pfn(page) << PAGE_SHIFT);
593 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
594 #define HASHED_PAGE_VIRTUAL
597 #if defined(WANT_PAGE_VIRTUAL)
598 #define page_address(page) ((page)->virtual)
599 #define set_page_address(page, address) \
601 (page)->virtual = (address); \
603 #define page_address_init() do { } while(0)
606 #if defined(HASHED_PAGE_VIRTUAL)
607 void *page_address(struct page *page);
608 void set_page_address(struct page *page, void *virtual);
609 void page_address_init(void);
612 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
613 #define page_address(page) lowmem_page_address(page)
614 #define set_page_address(page, address) do { } while(0)
615 #define page_address_init() do { } while(0)
619 * On an anonymous page mapped into a user virtual memory area,
620 * page->mapping points to its anon_vma, not to a struct address_space;
621 * with the PAGE_MAPPING_ANON bit set to distinguish it.
623 * Please note that, confusingly, "page_mapping" refers to the inode
624 * address_space which maps the page from disk; whereas "page_mapped"
625 * refers to user virtual address space into which the page is mapped.
627 #define PAGE_MAPPING_ANON 1
629 extern struct address_space swapper_space;
630 static inline struct address_space *page_mapping(struct page *page)
632 struct address_space *mapping = page->mapping;
634 VM_BUG_ON(PageSlab(page));
635 if (unlikely(PageSwapCache(page)))
636 mapping = &swapper_space;
638 else if (unlikely(PageSlab(page)))
641 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
646 static inline int PageAnon(struct page *page)
648 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
652 * Return the pagecache index of the passed page. Regular pagecache pages
653 * use ->index whereas swapcache pages use ->private
655 static inline pgoff_t page_index(struct page *page)
657 if (unlikely(PageSwapCache(page)))
658 return page_private(page);
663 * The atomic page->_mapcount, like _count, starts from -1:
664 * so that transitions both from it and to it can be tracked,
665 * using atomic_inc_and_test and atomic_add_negative(-1).
667 static inline void reset_page_mapcount(struct page *page)
669 atomic_set(&(page)->_mapcount, -1);
672 static inline int page_mapcount(struct page *page)
674 return atomic_read(&(page)->_mapcount) + 1;
678 * Return true if this page is mapped into pagetables.
680 static inline int page_mapped(struct page *page)
682 return atomic_read(&(page)->_mapcount) >= 0;
686 * Error return values for the *_nopage functions
688 #define NOPAGE_SIGBUS (NULL)
689 #define NOPAGE_OOM ((struct page *) (-1))
692 * Error return values for the *_nopfn functions
694 #define NOPFN_SIGBUS ((unsigned long) -1)
695 #define NOPFN_OOM ((unsigned long) -2)
696 #define NOPFN_REFAULT ((unsigned long) -3)
699 * Different kinds of faults, as returned by handle_mm_fault().
700 * Used to decide whether a process gets delivered SIGBUS or
701 * just gets major/minor fault counters bumped up.
703 #define VM_FAULT_OOM 0x00
704 #define VM_FAULT_SIGBUS 0x01
705 #define VM_FAULT_MINOR 0x02
706 #define VM_FAULT_MAJOR 0x03
709 * Special case for get_user_pages.
710 * Must be in a distinct bit from the above VM_FAULT_ flags.
712 #define VM_FAULT_WRITE 0x10
714 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
716 extern void show_free_areas(void);
719 int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
720 struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
722 int shmem_lock(struct file *file, int lock, struct user_struct *user);
724 static inline int shmem_lock(struct file *file, int lock,
725 struct user_struct *user)
730 static inline int shmem_set_policy(struct vm_area_struct *vma,
731 struct mempolicy *new)
736 static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
742 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
744 int shmem_zero_setup(struct vm_area_struct *);
747 extern unsigned long shmem_get_unmapped_area(struct file *file,
751 unsigned long flags);
754 extern int can_do_mlock(void);
755 extern int user_shm_lock(size_t, struct user_struct *);
756 extern void user_shm_unlock(size_t, struct user_struct *);
759 * Parameter block passed down to zap_pte_range in exceptional cases.
762 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
763 struct address_space *check_mapping; /* Check page->mapping if set */
764 pgoff_t first_index; /* Lowest page->index to unmap */
765 pgoff_t last_index; /* Highest page->index to unmap */
766 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
767 unsigned long truncate_count; /* Compare vm_truncate_count */
770 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
771 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
772 unsigned long size, struct zap_details *);
773 unsigned long unmap_vmas(struct mmu_gather **tlb,
774 struct vm_area_struct *start_vma, unsigned long start_addr,
775 unsigned long end_addr, unsigned long *nr_accounted,
776 struct zap_details *);
777 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
778 unsigned long end, unsigned long floor, unsigned long ceiling);
779 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
780 unsigned long floor, unsigned long ceiling);
781 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
782 struct vm_area_struct *vma);
783 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
784 unsigned long size, pgprot_t prot);
785 void unmap_mapping_range(struct address_space *mapping,
786 loff_t const holebegin, loff_t const holelen, int even_cows);
788 static inline void unmap_shared_mapping_range(struct address_space *mapping,
789 loff_t const holebegin, loff_t const holelen)
791 unmap_mapping_range(mapping, holebegin, holelen, 0);
794 extern int vmtruncate(struct inode * inode, loff_t offset);
795 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
796 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
797 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
800 extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
801 unsigned long address, int write_access);
803 static inline int handle_mm_fault(struct mm_struct *mm,
804 struct vm_area_struct *vma, unsigned long address,
807 return __handle_mm_fault(mm, vma, address, write_access) &
811 static inline int handle_mm_fault(struct mm_struct *mm,
812 struct vm_area_struct *vma, unsigned long address,
815 /* should never happen if there's no MMU */
817 return VM_FAULT_SIGBUS;
821 extern int make_pages_present(unsigned long addr, unsigned long end);
822 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
823 void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
825 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
826 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
827 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
829 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
830 extern void do_invalidatepage(struct page *page, unsigned long offset);
832 int __set_page_dirty_nobuffers(struct page *page);
833 int __set_page_dirty_no_writeback(struct page *page);
834 int redirty_page_for_writepage(struct writeback_control *wbc,
836 int FASTCALL(set_page_dirty(struct page *page));
837 int set_page_dirty_lock(struct page *page);
838 int clear_page_dirty_for_io(struct page *page);
840 extern unsigned long do_mremap(unsigned long addr,
841 unsigned long old_len, unsigned long new_len,
842 unsigned long flags, unsigned long new_addr);
845 * A callback you can register to apply pressure to ageable caches.
847 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
848 * look through the least-recently-used 'nr_to_scan' entries and
849 * attempt to free them up. It should return the number of objects
850 * which remain in the cache. If it returns -1, it means it cannot do
851 * any scanning at this time (eg. there is a risk of deadlock).
853 * The 'gfpmask' refers to the allocation we are currently trying to
856 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
857 * querying the cache size, so a fastpath for that case is appropriate.
860 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
861 int seeks; /* seeks to recreate an obj */
863 /* These are for internal use */
864 struct list_head list;
865 long nr; /* objs pending delete */
867 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
868 extern void register_shrinker(struct shrinker *);
869 extern void unregister_shrinker(struct shrinker *);
872 * Some shared mappigns will want the pages marked read-only
873 * to track write events. If so, we'll downgrade vm_page_prot
874 * to the private version (using protection_map[] without the
877 static inline int vma_wants_writenotify(struct vm_area_struct *vma)
879 unsigned int vm_flags = vma->vm_flags;
881 /* If it was private or non-writable, the write bit is already clear */
882 if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
885 /* The backer wishes to know when pages are first written to? */
886 if (vma->vm_ops && vma->vm_ops->page_mkwrite)
889 /* The open routine did something to the protections already? */
890 if (pgprot_val(vma->vm_page_prot) !=
891 pgprot_val(protection_map[vm_flags &
892 (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]))
895 /* Specialty mapping? */
896 if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE))
899 /* Can the mapping track the dirty pages? */
900 return vma->vm_file && vma->vm_file->f_mapping &&
901 mapping_cap_account_dirty(vma->vm_file->f_mapping);
904 extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
906 #ifdef __PAGETABLE_PUD_FOLDED
907 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
908 unsigned long address)
913 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
916 #ifdef __PAGETABLE_PMD_FOLDED
917 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
918 unsigned long address)
923 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
926 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
927 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
930 * The following ifdef needed to get the 4level-fixup.h header to work.
931 * Remove it when 4level-fixup.h has been removed.
933 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
934 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
936 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
937 NULL: pud_offset(pgd, address);
940 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
942 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
943 NULL: pmd_offset(pud, address);
945 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
947 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
949 * We tuck a spinlock to guard each pagetable page into its struct page,
950 * at page->private, with BUILD_BUG_ON to make sure that this will not
951 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
952 * When freeing, reset page->mapping so free_pages_check won't complain.
954 #define __pte_lockptr(page) &((page)->ptl)
955 #define pte_lock_init(_page) do { \
956 spin_lock_init(__pte_lockptr(_page)); \
958 #define pte_lock_deinit(page) ((page)->mapping = NULL)
959 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
962 * We use mm->page_table_lock to guard all pagetable pages of the mm.
964 #define pte_lock_init(page) do {} while (0)
965 #define pte_lock_deinit(page) do {} while (0)
966 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
967 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
969 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
971 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
972 pte_t *__pte = pte_offset_map(pmd, address); \
978 #define pte_unmap_unlock(pte, ptl) do { \
983 #define pte_alloc_map(mm, pmd, address) \
984 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
985 NULL: pte_offset_map(pmd, address))
987 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
988 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
989 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
991 #define pte_alloc_kernel(pmd, address) \
992 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
993 NULL: pte_offset_kernel(pmd, address))
995 extern void free_area_init(unsigned long * zones_size);
996 extern void free_area_init_node(int nid, pg_data_t *pgdat,
997 unsigned long * zones_size, unsigned long zone_start_pfn,
998 unsigned long *zholes_size);
999 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1001 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1002 * zones, allocate the backing mem_map and account for memory holes in a more
1003 * architecture independent manner. This is a substitute for creating the
1004 * zone_sizes[] and zholes_size[] arrays and passing them to
1005 * free_area_init_node()
1007 * An architecture is expected to register range of page frames backed by
1008 * physical memory with add_active_range() before calling
1009 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1010 * usage, an architecture is expected to do something like
1012 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1014 * for_each_valid_physical_page_range()
1015 * add_active_range(node_id, start_pfn, end_pfn)
1016 * free_area_init_nodes(max_zone_pfns);
1018 * If the architecture guarantees that there are no holes in the ranges
1019 * registered with add_active_range(), free_bootmem_active_regions()
1020 * will call free_bootmem_node() for each registered physical page range.
1021 * Similarly sparse_memory_present_with_active_regions() calls
1022 * memory_present() for each range when SPARSEMEM is enabled.
1024 * See mm/page_alloc.c for more information on each function exposed by
1025 * CONFIG_ARCH_POPULATES_NODE_MAP
1027 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1028 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1029 unsigned long end_pfn);
1030 extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
1031 unsigned long new_end_pfn);
1032 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1033 unsigned long end_pfn);
1034 extern void remove_all_active_ranges(void);
1035 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1036 unsigned long end_pfn);
1037 extern void get_pfn_range_for_nid(unsigned int nid,
1038 unsigned long *start_pfn, unsigned long *end_pfn);
1039 extern unsigned long find_min_pfn_with_active_regions(void);
1040 extern unsigned long find_max_pfn_with_active_regions(void);
1041 extern void free_bootmem_with_active_regions(int nid,
1042 unsigned long max_low_pfn);
1043 extern void sparse_memory_present_with_active_regions(int nid);
1044 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1045 extern int early_pfn_to_nid(unsigned long pfn);
1046 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1047 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1048 extern void set_dma_reserve(unsigned long new_dma_reserve);
1049 extern void memmap_init_zone(unsigned long, int, unsigned long,
1050 unsigned long, enum memmap_context);
1051 extern void setup_per_zone_pages_min(void);
1052 extern void mem_init(void);
1053 extern void show_mem(void);
1054 extern void si_meminfo(struct sysinfo * val);
1055 extern void si_meminfo_node(struct sysinfo *val, int nid);
1058 extern void setup_per_cpu_pageset(void);
1060 static inline void setup_per_cpu_pageset(void) {}
1064 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1065 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1066 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1067 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1068 struct prio_tree_iter *iter);
1070 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1071 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1072 (vma = vma_prio_tree_next(vma, iter)); )
1074 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1075 struct list_head *list)
1077 vma->shared.vm_set.parent = NULL;
1078 list_add_tail(&vma->shared.vm_set.list, list);
1082 extern int __vm_enough_memory(long pages, int cap_sys_admin);
1083 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1084 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1085 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1086 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1087 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1088 struct mempolicy *);
1089 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1090 extern int split_vma(struct mm_struct *,
1091 struct vm_area_struct *, unsigned long addr, int new_below);
1092 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1093 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1094 struct rb_node **, struct rb_node *);
1095 extern void unlink_file_vma(struct vm_area_struct *);
1096 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1097 unsigned long addr, unsigned long len, pgoff_t pgoff);
1098 extern void exit_mmap(struct mm_struct *);
1099 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1100 extern int install_special_mapping(struct mm_struct *mm,
1101 unsigned long addr, unsigned long len,
1102 unsigned long flags, struct page **pages);
1104 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1106 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1107 unsigned long len, unsigned long prot,
1108 unsigned long flag, unsigned long pgoff);
1109 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1110 unsigned long len, unsigned long flags,
1111 unsigned int vm_flags, unsigned long pgoff,
1114 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1115 unsigned long len, unsigned long prot,
1116 unsigned long flag, unsigned long offset)
1118 unsigned long ret = -EINVAL;
1119 if ((offset + PAGE_ALIGN(len)) < offset)
1121 if (!(offset & ~PAGE_MASK))
1122 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1127 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1129 extern unsigned long do_brk(unsigned long, unsigned long);
1132 extern unsigned long page_unuse(struct page *);
1133 extern void truncate_inode_pages(struct address_space *, loff_t);
1134 extern void truncate_inode_pages_range(struct address_space *,
1135 loff_t lstart, loff_t lend);
1137 /* generic vm_area_ops exported for stackable file systems */
1138 extern struct page *filemap_fault(struct vm_area_struct *, struct fault_data *);
1139 extern struct page * __deprecated_for_modules
1140 filemap_nopage(struct vm_area_struct *, unsigned long, int *);
1141 extern int __deprecated_for_modules filemap_populate(struct vm_area_struct *,
1142 unsigned long, unsigned long, pgprot_t, unsigned long, int);
1144 /* mm/page-writeback.c */
1145 int write_one_page(struct page *page, int wait);
1148 #define VM_MAX_READAHEAD 128 /* kbytes */
1149 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1150 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1151 * turning readahead off */
1153 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1154 pgoff_t offset, unsigned long nr_to_read);
1155 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1156 pgoff_t offset, unsigned long nr_to_read);
1157 unsigned long page_cache_readahead(struct address_space *mapping,
1158 struct file_ra_state *ra,
1161 unsigned long size);
1162 void handle_ra_miss(struct address_space *mapping,
1163 struct file_ra_state *ra, pgoff_t offset);
1164 unsigned long max_sane_readahead(unsigned long nr);
1166 /* Do stack extension */
1167 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1169 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1172 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1173 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1174 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1175 struct vm_area_struct **pprev);
1177 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1178 NULL if none. Assume start_addr < end_addr. */
1179 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1181 struct vm_area_struct * vma = find_vma(mm,start_addr);
1183 if (vma && end_addr <= vma->vm_start)
1188 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1190 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1193 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1194 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1195 struct page *vmalloc_to_page(void *addr);
1196 unsigned long vmalloc_to_pfn(void *addr);
1197 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1198 unsigned long pfn, unsigned long size, pgprot_t);
1199 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1200 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1203 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1204 unsigned int foll_flags);
1205 #define FOLL_WRITE 0x01 /* check pte is writable */
1206 #define FOLL_TOUCH 0x02 /* mark page accessed */
1207 #define FOLL_GET 0x04 /* do get_page on page */
1208 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1210 typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr,
1212 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1213 unsigned long size, pte_fn_t fn, void *data);
1215 #ifdef CONFIG_PROC_FS
1216 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1218 static inline void vm_stat_account(struct mm_struct *mm,
1219 unsigned long flags, struct file *file, long pages)
1222 #endif /* CONFIG_PROC_FS */
1224 #ifndef CONFIG_DEBUG_PAGEALLOC
1226 kernel_map_pages(struct page *page, int numpages, int enable) {}
1229 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1230 #ifdef __HAVE_ARCH_GATE_AREA
1231 int in_gate_area_no_task(unsigned long addr);
1232 int in_gate_area(struct task_struct *task, unsigned long addr);
1234 int in_gate_area_no_task(unsigned long addr);
1235 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1236 #endif /* __HAVE_ARCH_GATE_AREA */
1238 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1239 void __user *, size_t *, loff_t *);
1240 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1241 unsigned long lru_pages);
1242 void drop_pagecache(void);
1243 void drop_slab(void);
1246 #define randomize_va_space 0
1248 extern int randomize_va_space;
1251 __attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma);
1253 #endif /* __KERNEL__ */
1254 #endif /* _LINUX_MM_H */