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>
19 #include <linux/mm_types.h>
24 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
25 extern unsigned long max_mapnr;
28 extern unsigned long num_physpages;
29 extern void * high_memory;
30 extern unsigned long vmalloc_earlyreserve;
31 extern int page_cluster;
34 extern int sysctl_legacy_va_layout;
36 #define sysctl_legacy_va_layout 0
40 #include <asm/pgtable.h>
41 #include <asm/processor.h>
43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
46 * Linux kernel virtual memory manager primitives.
47 * The idea being to have a "virtual" mm in the same way
48 * we have a virtual fs - giving a cleaner interface to the
49 * mm details, and allowing different kinds of memory mappings
50 * (from shared memory to executable loading to arbitrary
55 * This struct defines a memory VMM memory area. There is one of these
56 * per VM-area/task. A VM area is any part of the process virtual memory
57 * space that has a special rule for the page-fault handlers (ie a shared
58 * library, the executable area etc).
60 struct vm_area_struct {
61 struct mm_struct * vm_mm; /* The address space we belong to. */
62 unsigned long vm_start; /* Our start address within vm_mm. */
63 unsigned long vm_end; /* The first byte after our end address
66 /* linked list of VM areas per task, sorted by address */
67 struct vm_area_struct *vm_next;
69 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
70 unsigned long vm_flags; /* Flags, listed below. */
75 * For areas with an address space and backing store,
76 * linkage into the address_space->i_mmap prio tree, or
77 * linkage to the list of like vmas hanging off its node, or
78 * linkage of vma in the address_space->i_mmap_nonlinear list.
82 struct list_head list;
83 void *parent; /* aligns with prio_tree_node parent */
84 struct vm_area_struct *head;
87 struct raw_prio_tree_node prio_tree_node;
91 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
92 * list, after a COW of one of the file pages. A MAP_SHARED vma
93 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
94 * or brk vma (with NULL file) can only be in an anon_vma list.
96 struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
97 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
99 /* Function pointers to deal with this struct. */
100 struct vm_operations_struct * vm_ops;
102 /* Information about our backing store: */
103 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
104 units, *not* PAGE_CACHE_SIZE */
105 struct file * vm_file; /* File we map to (can be NULL). */
106 void * vm_private_data; /* was vm_pte (shared mem) */
107 unsigned long vm_truncate_count;/* truncate_count or restart_addr */
110 atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
113 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
117 extern struct kmem_cache *vm_area_cachep;
120 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
121 * disabled, then there's a single shared list of VMAs maintained by the
122 * system, and mm's subscribe to these individually
124 struct vm_list_struct {
125 struct vm_list_struct *next;
126 struct vm_area_struct *vma;
130 extern struct rb_root nommu_vma_tree;
131 extern struct rw_semaphore nommu_vma_sem;
133 extern unsigned int kobjsize(const void *objp);
139 #define VM_READ 0x00000001 /* currently active flags */
140 #define VM_WRITE 0x00000002
141 #define VM_EXEC 0x00000004
142 #define VM_SHARED 0x00000008
144 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
145 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
146 #define VM_MAYWRITE 0x00000020
147 #define VM_MAYEXEC 0x00000040
148 #define VM_MAYSHARE 0x00000080
150 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
151 #define VM_GROWSUP 0x00000200
152 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
153 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
155 #define VM_EXECUTABLE 0x00001000
156 #define VM_LOCKED 0x00002000
157 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
159 /* Used by sys_madvise() */
160 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
161 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
163 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
164 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
165 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
166 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
167 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
168 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
169 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
170 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
171 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
173 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
174 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
177 #ifdef CONFIG_STACK_GROWSUP
178 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
183 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
184 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
185 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
186 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
187 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
190 * mapping from the currently active vm_flags protection bits (the
191 * low four bits) to a page protection mask..
193 extern pgprot_t protection_map[16];
197 * These are the virtual MM functions - opening of an area, closing and
198 * unmapping it (needed to keep files on disk up-to-date etc), pointer
199 * to the functions called when a no-page or a wp-page exception occurs.
201 struct vm_operations_struct {
202 void (*open)(struct vm_area_struct * area);
203 void (*close)(struct vm_area_struct * area);
204 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
205 unsigned long (*nopfn)(struct vm_area_struct * area, unsigned long address);
206 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
208 /* notification that a previously read-only page is about to become
209 * writable, if an error is returned it will cause a SIGBUS */
210 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
212 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
213 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
215 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
216 const nodemask_t *to, unsigned long flags);
223 #define page_private(page) ((page)->private)
224 #define set_page_private(page, v) ((page)->private = (v))
227 * FIXME: take this include out, include page-flags.h in
228 * files which need it (119 of them)
230 #include <linux/page-flags.h>
232 #ifdef CONFIG_DEBUG_VM
233 #define VM_BUG_ON(cond) BUG_ON(cond)
235 #define VM_BUG_ON(condition) do { } while(0)
239 * Methods to modify the page usage count.
241 * What counts for a page usage:
242 * - cache mapping (page->mapping)
243 * - private data (page->private)
244 * - page mapped in a task's page tables, each mapping
245 * is counted separately
247 * Also, many kernel routines increase the page count before a critical
248 * routine so they can be sure the page doesn't go away from under them.
252 * Drop a ref, return true if the refcount fell to zero (the page has no users)
254 static inline int put_page_testzero(struct page *page)
256 VM_BUG_ON(atomic_read(&page->_count) == 0);
257 return atomic_dec_and_test(&page->_count);
261 * Try to grab a ref unless the page has a refcount of zero, return false if
264 static inline int get_page_unless_zero(struct page *page)
266 VM_BUG_ON(PageCompound(page));
267 return atomic_inc_not_zero(&page->_count);
270 static inline struct page *compound_head(struct page *page)
273 * We could avoid the PageCompound(page) check if
274 * we would not overload PageTail().
276 * This check has to be done in several performance critical
277 * paths of the slab etc. IMHO PageTail deserves its own flag.
279 if (unlikely(PageCompound(page) && PageTail(page)))
280 return page->first_page;
284 static inline int page_count(struct page *page)
286 return atomic_read(&compound_head(page)->_count);
289 static inline void get_page(struct page *page)
291 page = compound_head(page);
292 VM_BUG_ON(atomic_read(&page->_count) == 0);
293 atomic_inc(&page->_count);
297 * Setup the page count before being freed into the page allocator for
298 * the first time (boot or memory hotplug)
300 static inline void init_page_count(struct page *page)
302 atomic_set(&page->_count, 1);
305 void put_page(struct page *page);
306 void put_pages_list(struct list_head *pages);
308 void split_page(struct page *page, unsigned int order);
311 * Compound pages have a destructor function. Provide a
312 * prototype for that function and accessor functions.
313 * These are _only_ valid on the head of a PG_compound page.
315 typedef void compound_page_dtor(struct page *);
317 static inline void set_compound_page_dtor(struct page *page,
318 compound_page_dtor *dtor)
320 page[1].lru.next = (void *)dtor;
323 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
325 return (compound_page_dtor *)page[1].lru.next;
328 static inline int compound_order(struct page *page)
330 if (!PageCompound(page) || PageTail(page))
332 return (unsigned long)page[1].lru.prev;
335 static inline void set_compound_order(struct page *page, unsigned long order)
337 page[1].lru.prev = (void *)order;
341 * Multiple processes may "see" the same page. E.g. for untouched
342 * mappings of /dev/null, all processes see the same page full of
343 * zeroes, and text pages of executables and shared libraries have
344 * only one copy in memory, at most, normally.
346 * For the non-reserved pages, page_count(page) denotes a reference count.
347 * page_count() == 0 means the page is free. page->lru is then used for
348 * freelist management in the buddy allocator.
349 * page_count() > 0 means the page has been allocated.
351 * Pages are allocated by the slab allocator in order to provide memory
352 * to kmalloc and kmem_cache_alloc. In this case, the management of the
353 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
354 * unless a particular usage is carefully commented. (the responsibility of
355 * freeing the kmalloc memory is the caller's, of course).
357 * A page may be used by anyone else who does a __get_free_page().
358 * In this case, page_count still tracks the references, and should only
359 * be used through the normal accessor functions. The top bits of page->flags
360 * and page->virtual store page management information, but all other fields
361 * are unused and could be used privately, carefully. The management of this
362 * page is the responsibility of the one who allocated it, and those who have
363 * subsequently been given references to it.
365 * The other pages (we may call them "pagecache pages") are completely
366 * managed by the Linux memory manager: I/O, buffers, swapping etc.
367 * The following discussion applies only to them.
369 * A pagecache page contains an opaque `private' member, which belongs to the
370 * page's address_space. Usually, this is the address of a circular list of
371 * the page's disk buffers. PG_private must be set to tell the VM to call
372 * into the filesystem to release these pages.
374 * A page may belong to an inode's memory mapping. In this case, page->mapping
375 * is the pointer to the inode, and page->index is the file offset of the page,
376 * in units of PAGE_CACHE_SIZE.
378 * If pagecache pages are not associated with an inode, they are said to be
379 * anonymous pages. These may become associated with the swapcache, and in that
380 * case PG_swapcache is set, and page->private is an offset into the swapcache.
382 * In either case (swapcache or inode backed), the pagecache itself holds one
383 * reference to the page. Setting PG_private should also increment the
384 * refcount. The each user mapping also has a reference to the page.
386 * The pagecache pages are stored in a per-mapping radix tree, which is
387 * rooted at mapping->page_tree, and indexed by offset.
388 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
389 * lists, we instead now tag pages as dirty/writeback in the radix tree.
391 * All pagecache pages may be subject to I/O:
392 * - inode pages may need to be read from disk,
393 * - inode pages which have been modified and are MAP_SHARED may need
394 * to be written back to the inode on disk,
395 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
396 * modified may need to be swapped out to swap space and (later) to be read
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 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
444 #define NODE_NOT_IN_PAGE_FLAGS
447 #ifndef PFN_SECTION_SHIFT
448 #define PFN_SECTION_SHIFT 0
452 * Define the bit shifts to access each section. For non-existant
453 * sections we define the shift as 0; that plus a 0 mask ensures
454 * the compiler will optimise away reference to them.
456 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
457 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
458 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
460 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
461 #ifdef NODE_NOT_IN_PAGEFLAGS
462 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
463 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
464 SECTIONS_PGOFF : ZONES_PGOFF)
466 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
467 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
468 NODES_PGOFF : ZONES_PGOFF)
471 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
473 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
474 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
477 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
478 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
479 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
480 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
482 static inline enum zone_type page_zonenum(struct page *page)
484 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
488 * The identification function is only used by the buddy allocator for
489 * determining if two pages could be buddies. We are not really
490 * identifying a zone since we could be using a the section number
491 * id if we have not node id available in page flags.
492 * We guarantee only that it will return the same value for two
493 * combinable pages in a zone.
495 static inline int page_zone_id(struct page *page)
497 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
500 static inline int zone_to_nid(struct zone *zone)
509 #ifdef NODE_NOT_IN_PAGE_FLAGS
510 extern int page_to_nid(struct page *page);
512 static inline int page_to_nid(struct page *page)
514 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
518 static inline struct zone *page_zone(struct page *page)
520 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
523 static inline unsigned long page_to_section(struct page *page)
525 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
528 static inline void set_page_zone(struct page *page, enum zone_type zone)
530 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
531 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
534 static inline void set_page_node(struct page *page, unsigned long node)
536 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
537 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
540 static inline void set_page_section(struct page *page, unsigned long section)
542 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
543 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
546 static inline void set_page_links(struct page *page, enum zone_type zone,
547 unsigned long node, unsigned long pfn)
549 set_page_zone(page, zone);
550 set_page_node(page, node);
551 set_page_section(page, pfn_to_section_nr(pfn));
555 * Some inline functions in vmstat.h depend on page_zone()
557 #include <linux/vmstat.h>
559 static __always_inline void *lowmem_page_address(struct page *page)
561 return __va(page_to_pfn(page) << PAGE_SHIFT);
564 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
565 #define HASHED_PAGE_VIRTUAL
568 #if defined(WANT_PAGE_VIRTUAL)
569 #define page_address(page) ((page)->virtual)
570 #define set_page_address(page, address) \
572 (page)->virtual = (address); \
574 #define page_address_init() do { } while(0)
577 #if defined(HASHED_PAGE_VIRTUAL)
578 void *page_address(struct page *page);
579 void set_page_address(struct page *page, void *virtual);
580 void page_address_init(void);
583 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
584 #define page_address(page) lowmem_page_address(page)
585 #define set_page_address(page, address) do { } while(0)
586 #define page_address_init() do { } while(0)
590 * On an anonymous page mapped into a user virtual memory area,
591 * page->mapping points to its anon_vma, not to a struct address_space;
592 * with the PAGE_MAPPING_ANON bit set to distinguish it.
594 * Please note that, confusingly, "page_mapping" refers to the inode
595 * address_space which maps the page from disk; whereas "page_mapped"
596 * refers to user virtual address space into which the page is mapped.
598 #define PAGE_MAPPING_ANON 1
600 extern struct address_space swapper_space;
601 static inline struct address_space *page_mapping(struct page *page)
603 struct address_space *mapping = page->mapping;
605 if (unlikely(PageSwapCache(page)))
606 mapping = &swapper_space;
607 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
612 static inline int PageAnon(struct page *page)
614 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
618 * Return the pagecache index of the passed page. Regular pagecache pages
619 * use ->index whereas swapcache pages use ->private
621 static inline pgoff_t page_index(struct page *page)
623 if (unlikely(PageSwapCache(page)))
624 return page_private(page);
629 * The atomic page->_mapcount, like _count, starts from -1:
630 * so that transitions both from it and to it can be tracked,
631 * using atomic_inc_and_test and atomic_add_negative(-1).
633 static inline void reset_page_mapcount(struct page *page)
635 atomic_set(&(page)->_mapcount, -1);
638 static inline int page_mapcount(struct page *page)
640 return atomic_read(&(page)->_mapcount) + 1;
644 * Return true if this page is mapped into pagetables.
646 static inline int page_mapped(struct page *page)
648 return atomic_read(&(page)->_mapcount) >= 0;
652 * Error return values for the *_nopage functions
654 #define NOPAGE_SIGBUS (NULL)
655 #define NOPAGE_OOM ((struct page *) (-1))
656 #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
659 * Error return values for the *_nopfn functions
661 #define NOPFN_SIGBUS ((unsigned long) -1)
662 #define NOPFN_OOM ((unsigned long) -2)
663 #define NOPFN_REFAULT ((unsigned long) -3)
666 * Different kinds of faults, as returned by handle_mm_fault().
667 * Used to decide whether a process gets delivered SIGBUS or
668 * just gets major/minor fault counters bumped up.
670 #define VM_FAULT_OOM 0x00
671 #define VM_FAULT_SIGBUS 0x01
672 #define VM_FAULT_MINOR 0x02
673 #define VM_FAULT_MAJOR 0x03
676 * Special case for get_user_pages.
677 * Must be in a distinct bit from the above VM_FAULT_ flags.
679 #define VM_FAULT_WRITE 0x10
681 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
683 extern void show_free_areas(void);
686 int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
687 struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
689 int shmem_lock(struct file *file, int lock, struct user_struct *user);
691 static inline int shmem_lock(struct file *file, int lock,
692 struct user_struct *user)
697 static inline int shmem_set_policy(struct vm_area_struct *vma,
698 struct mempolicy *new)
703 static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
709 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
711 int shmem_zero_setup(struct vm_area_struct *);
714 extern unsigned long shmem_get_unmapped_area(struct file *file,
718 unsigned long flags);
721 static inline int can_do_mlock(void)
723 if (capable(CAP_IPC_LOCK))
725 if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
729 extern int user_shm_lock(size_t, struct user_struct *);
730 extern void user_shm_unlock(size_t, struct user_struct *);
733 * Parameter block passed down to zap_pte_range in exceptional cases.
736 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
737 struct address_space *check_mapping; /* Check page->mapping if set */
738 pgoff_t first_index; /* Lowest page->index to unmap */
739 pgoff_t last_index; /* Highest page->index to unmap */
740 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
741 unsigned long truncate_count; /* Compare vm_truncate_count */
744 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
745 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
746 unsigned long size, struct zap_details *);
747 unsigned long unmap_vmas(struct mmu_gather **tlb,
748 struct vm_area_struct *start_vma, unsigned long start_addr,
749 unsigned long end_addr, unsigned long *nr_accounted,
750 struct zap_details *);
751 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
752 unsigned long end, unsigned long floor, unsigned long ceiling);
753 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
754 unsigned long floor, unsigned long ceiling);
755 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
756 struct vm_area_struct *vma);
757 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
758 unsigned long size, pgprot_t prot);
759 void unmap_mapping_range(struct address_space *mapping,
760 loff_t const holebegin, loff_t const holelen, int even_cows);
762 static inline void unmap_shared_mapping_range(struct address_space *mapping,
763 loff_t const holebegin, loff_t const holelen)
765 unmap_mapping_range(mapping, holebegin, holelen, 0);
768 extern int vmtruncate(struct inode * inode, loff_t offset);
769 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
770 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
771 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
774 extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
775 unsigned long address, int write_access);
777 static inline int handle_mm_fault(struct mm_struct *mm,
778 struct vm_area_struct *vma, unsigned long address,
781 return __handle_mm_fault(mm, vma, address, write_access) &
785 static inline int handle_mm_fault(struct mm_struct *mm,
786 struct vm_area_struct *vma, unsigned long address,
789 /* should never happen if there's no MMU */
791 return VM_FAULT_SIGBUS;
795 extern int make_pages_present(unsigned long addr, unsigned long end);
796 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
797 void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
799 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
800 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
801 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
803 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
804 extern void do_invalidatepage(struct page *page, unsigned long offset);
806 int __set_page_dirty_nobuffers(struct page *page);
807 int __set_page_dirty_no_writeback(struct page *page);
808 int redirty_page_for_writepage(struct writeback_control *wbc,
810 int FASTCALL(set_page_dirty(struct page *page));
811 int set_page_dirty_lock(struct page *page);
812 int clear_page_dirty_for_io(struct page *page);
814 extern unsigned long do_mremap(unsigned long addr,
815 unsigned long old_len, unsigned long new_len,
816 unsigned long flags, unsigned long new_addr);
819 * Prototype to add a shrinker callback for ageable caches.
821 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
822 * scan `nr_to_scan' objects, attempting to free them.
824 * The callback must return the number of objects which remain in the cache.
826 * The callback will be passed nr_to_scan == 0 when the VM is querying the
827 * cache size, so a fastpath for that case is appropriate.
829 typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
832 * Add an aging callback. The int is the number of 'seeks' it takes
833 * to recreate one of the objects that these functions age.
836 #define DEFAULT_SEEKS 2
838 extern struct shrinker *set_shrinker(int, shrinker_t);
839 extern void remove_shrinker(struct shrinker *shrinker);
842 * Some shared mappigns will want the pages marked read-only
843 * to track write events. If so, we'll downgrade vm_page_prot
844 * to the private version (using protection_map[] without the
847 static inline int vma_wants_writenotify(struct vm_area_struct *vma)
849 unsigned int vm_flags = vma->vm_flags;
851 /* If it was private or non-writable, the write bit is already clear */
852 if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED)))
855 /* The backer wishes to know when pages are first written to? */
856 if (vma->vm_ops && vma->vm_ops->page_mkwrite)
859 /* The open routine did something to the protections already? */
860 if (pgprot_val(vma->vm_page_prot) !=
861 pgprot_val(protection_map[vm_flags &
862 (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]))
865 /* Specialty mapping? */
866 if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE))
869 /* Can the mapping track the dirty pages? */
870 return vma->vm_file && vma->vm_file->f_mapping &&
871 mapping_cap_account_dirty(vma->vm_file->f_mapping);
874 extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
876 #ifdef __PAGETABLE_PUD_FOLDED
877 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
878 unsigned long address)
883 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
886 #ifdef __PAGETABLE_PMD_FOLDED
887 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
888 unsigned long address)
893 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
896 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
897 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
900 * The following ifdef needed to get the 4level-fixup.h header to work.
901 * Remove it when 4level-fixup.h has been removed.
903 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
904 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
906 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
907 NULL: pud_offset(pgd, address);
910 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
912 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
913 NULL: pmd_offset(pud, address);
915 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
917 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
919 * We tuck a spinlock to guard each pagetable page into its struct page,
920 * at page->private, with BUILD_BUG_ON to make sure that this will not
921 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
922 * When freeing, reset page->mapping so free_pages_check won't complain.
924 #define __pte_lockptr(page) &((page)->ptl)
925 #define pte_lock_init(_page) do { \
926 spin_lock_init(__pte_lockptr(_page)); \
928 #define pte_lock_deinit(page) ((page)->mapping = NULL)
929 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
932 * We use mm->page_table_lock to guard all pagetable pages of the mm.
934 #define pte_lock_init(page) do {} while (0)
935 #define pte_lock_deinit(page) do {} while (0)
936 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
937 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
939 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
941 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
942 pte_t *__pte = pte_offset_map(pmd, address); \
948 #define pte_unmap_unlock(pte, ptl) do { \
953 #define pte_alloc_map(mm, pmd, address) \
954 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
955 NULL: pte_offset_map(pmd, address))
957 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
958 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
959 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
961 #define pte_alloc_kernel(pmd, address) \
962 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
963 NULL: pte_offset_kernel(pmd, address))
965 extern void free_area_init(unsigned long * zones_size);
966 extern void free_area_init_node(int nid, pg_data_t *pgdat,
967 unsigned long * zones_size, unsigned long zone_start_pfn,
968 unsigned long *zholes_size);
969 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
971 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
972 * zones, allocate the backing mem_map and account for memory holes in a more
973 * architecture independent manner. This is a substitute for creating the
974 * zone_sizes[] and zholes_size[] arrays and passing them to
975 * free_area_init_node()
977 * An architecture is expected to register range of page frames backed by
978 * physical memory with add_active_range() before calling
979 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
980 * usage, an architecture is expected to do something like
982 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
984 * for_each_valid_physical_page_range()
985 * add_active_range(node_id, start_pfn, end_pfn)
986 * free_area_init_nodes(max_zone_pfns);
988 * If the architecture guarantees that there are no holes in the ranges
989 * registered with add_active_range(), free_bootmem_active_regions()
990 * will call free_bootmem_node() for each registered physical page range.
991 * Similarly sparse_memory_present_with_active_regions() calls
992 * memory_present() for each range when SPARSEMEM is enabled.
994 * See mm/page_alloc.c for more information on each function exposed by
995 * CONFIG_ARCH_POPULATES_NODE_MAP
997 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
998 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
999 unsigned long end_pfn);
1000 extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
1001 unsigned long new_end_pfn);
1002 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1003 unsigned long end_pfn);
1004 extern void remove_all_active_ranges(void);
1005 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1006 unsigned long end_pfn);
1007 extern void get_pfn_range_for_nid(unsigned int nid,
1008 unsigned long *start_pfn, unsigned long *end_pfn);
1009 extern unsigned long find_min_pfn_with_active_regions(void);
1010 extern unsigned long find_max_pfn_with_active_regions(void);
1011 extern void free_bootmem_with_active_regions(int nid,
1012 unsigned long max_low_pfn);
1013 extern void sparse_memory_present_with_active_regions(int nid);
1014 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1015 extern int early_pfn_to_nid(unsigned long pfn);
1016 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1017 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1018 extern void set_dma_reserve(unsigned long new_dma_reserve);
1019 extern void memmap_init_zone(unsigned long, int, unsigned long,
1020 unsigned long, enum memmap_context);
1021 extern void setup_per_zone_pages_min(void);
1022 extern void mem_init(void);
1023 extern void show_mem(void);
1024 extern void si_meminfo(struct sysinfo * val);
1025 extern void si_meminfo_node(struct sysinfo *val, int nid);
1028 extern void setup_per_cpu_pageset(void);
1030 static inline void setup_per_cpu_pageset(void) {}
1034 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1035 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1036 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1037 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1038 struct prio_tree_iter *iter);
1040 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1041 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1042 (vma = vma_prio_tree_next(vma, iter)); )
1044 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1045 struct list_head *list)
1047 vma->shared.vm_set.parent = NULL;
1048 list_add_tail(&vma->shared.vm_set.list, list);
1052 extern int __vm_enough_memory(long pages, int cap_sys_admin);
1053 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1054 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1055 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1056 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1057 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1058 struct mempolicy *);
1059 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1060 extern int split_vma(struct mm_struct *,
1061 struct vm_area_struct *, unsigned long addr, int new_below);
1062 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1063 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1064 struct rb_node **, struct rb_node *);
1065 extern void unlink_file_vma(struct vm_area_struct *);
1066 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1067 unsigned long addr, unsigned long len, pgoff_t pgoff);
1068 extern void exit_mmap(struct mm_struct *);
1069 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1070 extern int install_special_mapping(struct mm_struct *mm,
1071 unsigned long addr, unsigned long len,
1072 unsigned long flags, struct page **pages);
1074 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1076 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1077 unsigned long len, unsigned long prot,
1078 unsigned long flag, unsigned long pgoff);
1080 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1081 unsigned long len, unsigned long prot,
1082 unsigned long flag, unsigned long offset)
1084 unsigned long ret = -EINVAL;
1085 if ((offset + PAGE_ALIGN(len)) < offset)
1087 if (!(offset & ~PAGE_MASK))
1088 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1093 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1095 extern unsigned long do_brk(unsigned long, unsigned long);
1098 extern unsigned long page_unuse(struct page *);
1099 extern void truncate_inode_pages(struct address_space *, loff_t);
1100 extern void truncate_inode_pages_range(struct address_space *,
1101 loff_t lstart, loff_t lend);
1103 /* generic vm_area_ops exported for stackable file systems */
1104 extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
1105 extern int filemap_populate(struct vm_area_struct *, unsigned long,
1106 unsigned long, pgprot_t, unsigned long, int);
1108 /* mm/page-writeback.c */
1109 int write_one_page(struct page *page, int wait);
1112 #define VM_MAX_READAHEAD 128 /* kbytes */
1113 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1114 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1115 * turning readahead off */
1117 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1118 pgoff_t offset, unsigned long nr_to_read);
1119 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1120 pgoff_t offset, unsigned long nr_to_read);
1121 unsigned long page_cache_readahead(struct address_space *mapping,
1122 struct file_ra_state *ra,
1125 unsigned long size);
1126 void handle_ra_miss(struct address_space *mapping,
1127 struct file_ra_state *ra, pgoff_t offset);
1128 unsigned long max_sane_readahead(unsigned long nr);
1130 /* Do stack extension */
1131 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1133 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1136 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1137 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1138 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1139 struct vm_area_struct **pprev);
1141 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1142 NULL if none. Assume start_addr < end_addr. */
1143 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1145 struct vm_area_struct * vma = find_vma(mm,start_addr);
1147 if (vma && end_addr <= vma->vm_start)
1152 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1154 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1157 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1158 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1159 struct page *vmalloc_to_page(void *addr);
1160 unsigned long vmalloc_to_pfn(void *addr);
1161 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1162 unsigned long pfn, unsigned long size, pgprot_t);
1163 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1164 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1167 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1168 unsigned int foll_flags);
1169 #define FOLL_WRITE 0x01 /* check pte is writable */
1170 #define FOLL_TOUCH 0x02 /* mark page accessed */
1171 #define FOLL_GET 0x04 /* do get_page on page */
1172 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1174 typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr,
1176 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1177 unsigned long size, pte_fn_t fn, void *data);
1179 #ifdef CONFIG_PROC_FS
1180 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1182 static inline void vm_stat_account(struct mm_struct *mm,
1183 unsigned long flags, struct file *file, long pages)
1186 #endif /* CONFIG_PROC_FS */
1188 #ifndef CONFIG_DEBUG_PAGEALLOC
1190 kernel_map_pages(struct page *page, int numpages, int enable) {}
1193 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1194 #ifdef __HAVE_ARCH_GATE_AREA
1195 int in_gate_area_no_task(unsigned long addr);
1196 int in_gate_area(struct task_struct *task, unsigned long addr);
1198 int in_gate_area_no_task(unsigned long addr);
1199 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1200 #endif /* __HAVE_ARCH_GATE_AREA */
1202 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1203 void __user *, size_t *, loff_t *);
1204 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1205 unsigned long lru_pages);
1206 void drop_pagecache(void);
1207 void drop_slab(void);
1210 #define randomize_va_space 0
1212 extern int randomize_va_space;
1215 __attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma);
1217 #endif /* __KERNEL__ */
1218 #endif /* _LINUX_MM_H */