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>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
20 struct writeback_control;
22 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23 extern unsigned long max_mapnr;
26 extern unsigned long num_physpages;
27 extern void * high_memory;
28 extern int page_cluster;
31 extern int sysctl_legacy_va_layout;
33 #define sysctl_legacy_va_layout 0
36 extern unsigned long mmap_min_addr;
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
53 extern struct kmem_cache *vm_area_cachep;
56 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
57 * disabled, then there's a single shared list of VMAs maintained by the
58 * system, and mm's subscribe to these individually
60 struct vm_list_struct {
61 struct vm_list_struct *next;
62 struct vm_area_struct *vma;
66 extern struct rb_root nommu_vma_tree;
67 extern struct rw_semaphore nommu_vma_sem;
69 extern unsigned int kobjsize(const void *objp);
75 #define VM_READ 0x00000001 /* currently active flags */
76 #define VM_WRITE 0x00000002
77 #define VM_EXEC 0x00000004
78 #define VM_SHARED 0x00000008
80 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
81 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82 #define VM_MAYWRITE 0x00000020
83 #define VM_MAYEXEC 0x00000040
84 #define VM_MAYSHARE 0x00000080
86 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87 #define VM_GROWSUP 0x00000200
88 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
89 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
91 #define VM_EXECUTABLE 0x00001000
92 #define VM_LOCKED 0x00002000
93 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
95 /* Used by sys_madvise() */
96 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
99 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
101 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
102 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
104 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
105 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
106 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
107 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
109 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
111 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
112 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
115 #ifdef CONFIG_STACK_GROWSUP
116 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
118 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
121 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
122 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
123 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
124 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
125 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
128 * mapping from the currently active vm_flags protection bits (the
129 * low four bits) to a page protection mask..
131 extern pgprot_t protection_map[16];
133 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
134 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
138 * vm_fault is filled by the the pagefault handler and passed to the vma's
139 * ->fault function. The vma's ->fault is responsible for returning a bitmask
140 * of VM_FAULT_xxx flags that give details about how the fault was handled.
142 * pgoff should be used in favour of virtual_address, if possible. If pgoff
143 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
147 unsigned int flags; /* FAULT_FLAG_xxx flags */
148 pgoff_t pgoff; /* Logical page offset based on vma */
149 void __user *virtual_address; /* Faulting virtual address */
151 struct page *page; /* ->fault handlers should return a
152 * page here, unless VM_FAULT_NOPAGE
153 * is set (which is also implied by
159 * These are the virtual MM functions - opening of an area, closing and
160 * unmapping it (needed to keep files on disk up-to-date etc), pointer
161 * to the functions called when a no-page or a wp-page exception occurs.
163 struct vm_operations_struct {
164 void (*open)(struct vm_area_struct * area);
165 void (*close)(struct vm_area_struct * area);
166 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
167 unsigned long (*nopfn)(struct vm_area_struct *area,
168 unsigned long address);
170 /* notification that a previously read-only page is about to become
171 * writable, if an error is returned it will cause a SIGBUS */
172 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
174 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
175 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
177 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
178 const nodemask_t *to, unsigned long flags);
185 #define page_private(page) ((page)->private)
186 #define set_page_private(page, v) ((page)->private = (v))
189 * FIXME: take this include out, include page-flags.h in
190 * files which need it (119 of them)
192 #include <linux/page-flags.h>
194 #ifdef CONFIG_DEBUG_VM
195 #define VM_BUG_ON(cond) BUG_ON(cond)
197 #define VM_BUG_ON(condition) do { } while(0)
201 * Methods to modify the page usage count.
203 * What counts for a page usage:
204 * - cache mapping (page->mapping)
205 * - private data (page->private)
206 * - page mapped in a task's page tables, each mapping
207 * is counted separately
209 * Also, many kernel routines increase the page count before a critical
210 * routine so they can be sure the page doesn't go away from under them.
214 * Drop a ref, return true if the refcount fell to zero (the page has no users)
216 static inline int put_page_testzero(struct page *page)
218 VM_BUG_ON(atomic_read(&page->_count) == 0);
219 return atomic_dec_and_test(&page->_count);
223 * Try to grab a ref unless the page has a refcount of zero, return false if
226 static inline int get_page_unless_zero(struct page *page)
228 VM_BUG_ON(PageTail(page));
229 return atomic_inc_not_zero(&page->_count);
232 /* Support for virtually mapped pages */
233 struct page *vmalloc_to_page(const void *addr);
234 unsigned long vmalloc_to_pfn(const void *addr);
237 * Determine if an address is within the vmalloc range
239 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
240 * is no special casing required.
242 static inline int is_vmalloc_addr(const void *x)
245 unsigned long addr = (unsigned long)x;
247 return addr >= VMALLOC_START && addr < VMALLOC_END;
253 static inline struct page *compound_head(struct page *page)
255 if (unlikely(PageTail(page)))
256 return page->first_page;
260 static inline int page_count(struct page *page)
262 return atomic_read(&compound_head(page)->_count);
265 static inline void get_page(struct page *page)
267 page = compound_head(page);
268 VM_BUG_ON(atomic_read(&page->_count) == 0);
269 atomic_inc(&page->_count);
272 static inline struct page *virt_to_head_page(const void *x)
274 struct page *page = virt_to_page(x);
275 return compound_head(page);
279 * Setup the page count before being freed into the page allocator for
280 * the first time (boot or memory hotplug)
282 static inline void init_page_count(struct page *page)
284 atomic_set(&page->_count, 1);
287 void put_page(struct page *page);
288 void put_pages_list(struct list_head *pages);
290 void split_page(struct page *page, unsigned int order);
293 * Compound pages have a destructor function. Provide a
294 * prototype for that function and accessor functions.
295 * These are _only_ valid on the head of a PG_compound page.
297 typedef void compound_page_dtor(struct page *);
299 static inline void set_compound_page_dtor(struct page *page,
300 compound_page_dtor *dtor)
302 page[1].lru.next = (void *)dtor;
305 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
307 return (compound_page_dtor *)page[1].lru.next;
310 static inline int compound_order(struct page *page)
314 return (unsigned long)page[1].lru.prev;
317 static inline void set_compound_order(struct page *page, unsigned long order)
319 page[1].lru.prev = (void *)order;
323 * Multiple processes may "see" the same page. E.g. for untouched
324 * mappings of /dev/null, all processes see the same page full of
325 * zeroes, and text pages of executables and shared libraries have
326 * only one copy in memory, at most, normally.
328 * For the non-reserved pages, page_count(page) denotes a reference count.
329 * page_count() == 0 means the page is free. page->lru is then used for
330 * freelist management in the buddy allocator.
331 * page_count() > 0 means the page has been allocated.
333 * Pages are allocated by the slab allocator in order to provide memory
334 * to kmalloc and kmem_cache_alloc. In this case, the management of the
335 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
336 * unless a particular usage is carefully commented. (the responsibility of
337 * freeing the kmalloc memory is the caller's, of course).
339 * A page may be used by anyone else who does a __get_free_page().
340 * In this case, page_count still tracks the references, and should only
341 * be used through the normal accessor functions. The top bits of page->flags
342 * and page->virtual store page management information, but all other fields
343 * are unused and could be used privately, carefully. The management of this
344 * page is the responsibility of the one who allocated it, and those who have
345 * subsequently been given references to it.
347 * The other pages (we may call them "pagecache pages") are completely
348 * managed by the Linux memory manager: I/O, buffers, swapping etc.
349 * The following discussion applies only to them.
351 * A pagecache page contains an opaque `private' member, which belongs to the
352 * page's address_space. Usually, this is the address of a circular list of
353 * the page's disk buffers. PG_private must be set to tell the VM to call
354 * into the filesystem to release these pages.
356 * A page may belong to an inode's memory mapping. In this case, page->mapping
357 * is the pointer to the inode, and page->index is the file offset of the page,
358 * in units of PAGE_CACHE_SIZE.
360 * If pagecache pages are not associated with an inode, they are said to be
361 * anonymous pages. These may become associated with the swapcache, and in that
362 * case PG_swapcache is set, and page->private is an offset into the swapcache.
364 * In either case (swapcache or inode backed), the pagecache itself holds one
365 * reference to the page. Setting PG_private should also increment the
366 * refcount. The each user mapping also has a reference to the page.
368 * The pagecache pages are stored in a per-mapping radix tree, which is
369 * rooted at mapping->page_tree, and indexed by offset.
370 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
371 * lists, we instead now tag pages as dirty/writeback in the radix tree.
373 * All pagecache pages may be subject to I/O:
374 * - inode pages may need to be read from disk,
375 * - inode pages which have been modified and are MAP_SHARED may need
376 * to be written back to the inode on disk,
377 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
378 * modified may need to be swapped out to swap space and (later) to be read
383 * The zone field is never updated after free_area_init_core()
384 * sets it, so none of the operations on it need to be atomic.
389 * page->flags layout:
391 * There are three possibilities for how page->flags get
392 * laid out. The first is for the normal case, without
393 * sparsemem. The second is for sparsemem when there is
394 * plenty of space for node and section. The last is when
395 * we have run out of space and have to fall back to an
396 * alternate (slower) way of determining the node.
398 * No sparsemem: | NODE | ZONE | ... | FLAGS |
399 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
400 * no space for node: | SECTION | ZONE | ... | FLAGS |
402 #ifdef CONFIG_SPARSEMEM
403 #define SECTIONS_WIDTH SECTIONS_SHIFT
405 #define SECTIONS_WIDTH 0
408 #define ZONES_WIDTH ZONES_SHIFT
410 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
411 #define NODES_WIDTH NODES_SHIFT
413 #define NODES_WIDTH 0
416 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
417 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
418 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
419 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
422 * We are going to use the flags for the page to node mapping if its in
423 * there. This includes the case where there is no node, so it is implicit.
425 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
426 #define NODE_NOT_IN_PAGE_FLAGS
429 #ifndef PFN_SECTION_SHIFT
430 #define PFN_SECTION_SHIFT 0
434 * Define the bit shifts to access each section. For non-existant
435 * sections we define the shift as 0; that plus a 0 mask ensures
436 * the compiler will optimise away reference to them.
438 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
439 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
440 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
442 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
443 #ifdef NODE_NOT_IN_PAGEFLAGS
444 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
445 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
446 SECTIONS_PGOFF : ZONES_PGOFF)
448 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
449 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
450 NODES_PGOFF : ZONES_PGOFF)
453 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
455 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
456 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
459 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
460 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
461 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
462 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
464 static inline enum zone_type page_zonenum(struct page *page)
466 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
470 * The identification function is only used by the buddy allocator for
471 * determining if two pages could be buddies. We are not really
472 * identifying a zone since we could be using a the section number
473 * id if we have not node id available in page flags.
474 * We guarantee only that it will return the same value for two
475 * combinable pages in a zone.
477 static inline int page_zone_id(struct page *page)
479 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
482 static inline int zone_to_nid(struct zone *zone)
491 #ifdef NODE_NOT_IN_PAGE_FLAGS
492 extern int page_to_nid(struct page *page);
494 static inline int page_to_nid(struct page *page)
496 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
500 static inline struct zone *page_zone(struct page *page)
502 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
505 static inline unsigned long page_to_section(struct page *page)
507 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
510 static inline void set_page_zone(struct page *page, enum zone_type zone)
512 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
513 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
516 static inline void set_page_node(struct page *page, unsigned long node)
518 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
519 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
522 static inline void set_page_section(struct page *page, unsigned long section)
524 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
525 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
528 static inline void set_page_links(struct page *page, enum zone_type zone,
529 unsigned long node, unsigned long pfn)
531 set_page_zone(page, zone);
532 set_page_node(page, node);
533 set_page_section(page, pfn_to_section_nr(pfn));
537 * If a hint addr is less than mmap_min_addr change hint to be as
538 * low as possible but still greater than mmap_min_addr
540 static inline unsigned long round_hint_to_min(unsigned long hint)
542 #ifdef CONFIG_SECURITY
544 if (((void *)hint != NULL) &&
545 (hint < mmap_min_addr))
546 return PAGE_ALIGN(mmap_min_addr);
552 * Some inline functions in vmstat.h depend on page_zone()
554 #include <linux/vmstat.h>
556 static __always_inline void *lowmem_page_address(struct page *page)
558 return __va(page_to_pfn(page) << PAGE_SHIFT);
561 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
562 #define HASHED_PAGE_VIRTUAL
565 #if defined(WANT_PAGE_VIRTUAL)
566 #define page_address(page) ((page)->virtual)
567 #define set_page_address(page, address) \
569 (page)->virtual = (address); \
571 #define page_address_init() do { } while(0)
574 #if defined(HASHED_PAGE_VIRTUAL)
575 void *page_address(struct page *page);
576 void set_page_address(struct page *page, void *virtual);
577 void page_address_init(void);
580 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
581 #define page_address(page) lowmem_page_address(page)
582 #define set_page_address(page, address) do { } while(0)
583 #define page_address_init() do { } while(0)
587 * On an anonymous page mapped into a user virtual memory area,
588 * page->mapping points to its anon_vma, not to a struct address_space;
589 * with the PAGE_MAPPING_ANON bit set to distinguish it.
591 * Please note that, confusingly, "page_mapping" refers to the inode
592 * address_space which maps the page from disk; whereas "page_mapped"
593 * refers to user virtual address space into which the page is mapped.
595 #define PAGE_MAPPING_ANON 1
597 extern struct address_space swapper_space;
598 static inline struct address_space *page_mapping(struct page *page)
600 struct address_space *mapping = page->mapping;
602 VM_BUG_ON(PageSlab(page));
603 if (unlikely(PageSwapCache(page)))
604 mapping = &swapper_space;
605 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
610 static inline int PageAnon(struct page *page)
612 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
616 * Return the pagecache index of the passed page. Regular pagecache pages
617 * use ->index whereas swapcache pages use ->private
619 static inline pgoff_t page_index(struct page *page)
621 if (unlikely(PageSwapCache(page)))
622 return page_private(page);
627 * The atomic page->_mapcount, like _count, starts from -1:
628 * so that transitions both from it and to it can be tracked,
629 * using atomic_inc_and_test and atomic_add_negative(-1).
631 static inline void reset_page_mapcount(struct page *page)
633 atomic_set(&(page)->_mapcount, -1);
636 static inline int page_mapcount(struct page *page)
638 return atomic_read(&(page)->_mapcount) + 1;
642 * Return true if this page is mapped into pagetables.
644 static inline int page_mapped(struct page *page)
646 return atomic_read(&(page)->_mapcount) >= 0;
650 * Error return values for the *_nopfn functions
652 #define NOPFN_SIGBUS ((unsigned long) -1)
653 #define NOPFN_OOM ((unsigned long) -2)
654 #define NOPFN_REFAULT ((unsigned long) -3)
657 * Different kinds of faults, as returned by handle_mm_fault().
658 * Used to decide whether a process gets delivered SIGBUS or
659 * just gets major/minor fault counters bumped up.
662 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
664 #define VM_FAULT_OOM 0x0001
665 #define VM_FAULT_SIGBUS 0x0002
666 #define VM_FAULT_MAJOR 0x0004
667 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
669 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
670 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
672 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
674 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
676 extern void show_free_areas(void);
679 int shmem_lock(struct file *file, int lock, struct user_struct *user);
681 static inline int shmem_lock(struct file *file, int lock,
682 struct user_struct *user)
687 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
689 int shmem_zero_setup(struct vm_area_struct *);
692 extern unsigned long shmem_get_unmapped_area(struct file *file,
696 unsigned long flags);
699 extern int can_do_mlock(void);
700 extern int user_shm_lock(size_t, struct user_struct *);
701 extern void user_shm_unlock(size_t, struct user_struct *);
704 * Parameter block passed down to zap_pte_range in exceptional cases.
707 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
708 struct address_space *check_mapping; /* Check page->mapping if set */
709 pgoff_t first_index; /* Lowest page->index to unmap */
710 pgoff_t last_index; /* Highest page->index to unmap */
711 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
712 unsigned long truncate_count; /* Compare vm_truncate_count */
715 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
716 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
717 unsigned long size, struct zap_details *);
718 unsigned long unmap_vmas(struct mmu_gather **tlb,
719 struct vm_area_struct *start_vma, unsigned long start_addr,
720 unsigned long end_addr, unsigned long *nr_accounted,
721 struct zap_details *);
724 * mm_walk - callbacks for walk_page_range
725 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
726 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
727 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
728 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
729 * @pte_hole: if set, called for each hole at all levels
731 * (see walk_page_range for more details)
734 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, void *);
735 int (*pud_entry)(pud_t *, unsigned long, unsigned long, void *);
736 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, void *);
737 int (*pte_entry)(pte_t *, unsigned long, unsigned long, void *);
738 int (*pte_hole)(unsigned long, unsigned long, void *);
741 int walk_page_range(const struct mm_struct *, unsigned long addr,
742 unsigned long end, const struct mm_walk *walk,
744 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
745 unsigned long end, unsigned long floor, unsigned long ceiling);
746 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
747 unsigned long floor, unsigned long ceiling);
748 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
749 struct vm_area_struct *vma);
750 void unmap_mapping_range(struct address_space *mapping,
751 loff_t const holebegin, loff_t const holelen, int even_cows);
753 static inline void unmap_shared_mapping_range(struct address_space *mapping,
754 loff_t const holebegin, loff_t const holelen)
756 unmap_mapping_range(mapping, holebegin, holelen, 0);
759 extern int vmtruncate(struct inode * inode, loff_t offset);
760 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
763 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
764 unsigned long address, int write_access);
766 static inline int handle_mm_fault(struct mm_struct *mm,
767 struct vm_area_struct *vma, unsigned long address,
770 /* should never happen if there's no MMU */
772 return VM_FAULT_SIGBUS;
776 extern int make_pages_present(unsigned long addr, unsigned long end);
777 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
779 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
780 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
781 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
783 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
784 extern void do_invalidatepage(struct page *page, unsigned long offset);
786 int __set_page_dirty_nobuffers(struct page *page);
787 int __set_page_dirty_no_writeback(struct page *page);
788 int redirty_page_for_writepage(struct writeback_control *wbc,
790 int set_page_dirty(struct page *page);
791 int set_page_dirty_lock(struct page *page);
792 int clear_page_dirty_for_io(struct page *page);
794 extern unsigned long move_page_tables(struct vm_area_struct *vma,
795 unsigned long old_addr, struct vm_area_struct *new_vma,
796 unsigned long new_addr, unsigned long len);
797 extern unsigned long do_mremap(unsigned long addr,
798 unsigned long old_len, unsigned long new_len,
799 unsigned long flags, unsigned long new_addr);
800 extern int mprotect_fixup(struct vm_area_struct *vma,
801 struct vm_area_struct **pprev, unsigned long start,
802 unsigned long end, unsigned long newflags);
805 * A callback you can register to apply pressure to ageable caches.
807 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
808 * look through the least-recently-used 'nr_to_scan' entries and
809 * attempt to free them up. It should return the number of objects
810 * which remain in the cache. If it returns -1, it means it cannot do
811 * any scanning at this time (eg. there is a risk of deadlock).
813 * The 'gfpmask' refers to the allocation we are currently trying to
816 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
817 * querying the cache size, so a fastpath for that case is appropriate.
820 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
821 int seeks; /* seeks to recreate an obj */
823 /* These are for internal use */
824 struct list_head list;
825 long nr; /* objs pending delete */
827 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
828 extern void register_shrinker(struct shrinker *);
829 extern void unregister_shrinker(struct shrinker *);
831 int vma_wants_writenotify(struct vm_area_struct *vma);
833 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
835 #ifdef __PAGETABLE_PUD_FOLDED
836 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
837 unsigned long address)
842 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
845 #ifdef __PAGETABLE_PMD_FOLDED
846 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
847 unsigned long address)
852 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
855 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
856 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
859 * The following ifdef needed to get the 4level-fixup.h header to work.
860 * Remove it when 4level-fixup.h has been removed.
862 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
863 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
865 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
866 NULL: pud_offset(pgd, address);
869 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
871 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
872 NULL: pmd_offset(pud, address);
874 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
876 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
878 * We tuck a spinlock to guard each pagetable page into its struct page,
879 * at page->private, with BUILD_BUG_ON to make sure that this will not
880 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
881 * When freeing, reset page->mapping so free_pages_check won't complain.
883 #define __pte_lockptr(page) &((page)->ptl)
884 #define pte_lock_init(_page) do { \
885 spin_lock_init(__pte_lockptr(_page)); \
887 #define pte_lock_deinit(page) ((page)->mapping = NULL)
888 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
891 * We use mm->page_table_lock to guard all pagetable pages of the mm.
893 #define pte_lock_init(page) do {} while (0)
894 #define pte_lock_deinit(page) do {} while (0)
895 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
896 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
898 static inline void pgtable_page_ctor(struct page *page)
901 inc_zone_page_state(page, NR_PAGETABLE);
904 static inline void pgtable_page_dtor(struct page *page)
906 pte_lock_deinit(page);
907 dec_zone_page_state(page, NR_PAGETABLE);
910 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
912 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
913 pte_t *__pte = pte_offset_map(pmd, address); \
919 #define pte_unmap_unlock(pte, ptl) do { \
924 #define pte_alloc_map(mm, pmd, address) \
925 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
926 NULL: pte_offset_map(pmd, address))
928 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
929 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
930 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
932 #define pte_alloc_kernel(pmd, address) \
933 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
934 NULL: pte_offset_kernel(pmd, address))
936 extern void free_area_init(unsigned long * zones_size);
937 extern void free_area_init_node(int nid, pg_data_t *pgdat,
938 unsigned long * zones_size, unsigned long zone_start_pfn,
939 unsigned long *zholes_size);
940 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
942 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
943 * zones, allocate the backing mem_map and account for memory holes in a more
944 * architecture independent manner. This is a substitute for creating the
945 * zone_sizes[] and zholes_size[] arrays and passing them to
946 * free_area_init_node()
948 * An architecture is expected to register range of page frames backed by
949 * physical memory with add_active_range() before calling
950 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
951 * usage, an architecture is expected to do something like
953 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
955 * for_each_valid_physical_page_range()
956 * add_active_range(node_id, start_pfn, end_pfn)
957 * free_area_init_nodes(max_zone_pfns);
959 * If the architecture guarantees that there are no holes in the ranges
960 * registered with add_active_range(), free_bootmem_active_regions()
961 * will call free_bootmem_node() for each registered physical page range.
962 * Similarly sparse_memory_present_with_active_regions() calls
963 * memory_present() for each range when SPARSEMEM is enabled.
965 * See mm/page_alloc.c for more information on each function exposed by
966 * CONFIG_ARCH_POPULATES_NODE_MAP
968 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
969 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
970 unsigned long end_pfn);
971 extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
972 unsigned long new_end_pfn);
973 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
974 unsigned long end_pfn);
975 extern void remove_all_active_ranges(void);
976 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
977 unsigned long end_pfn);
978 extern void get_pfn_range_for_nid(unsigned int nid,
979 unsigned long *start_pfn, unsigned long *end_pfn);
980 extern unsigned long find_min_pfn_with_active_regions(void);
981 extern unsigned long find_max_pfn_with_active_regions(void);
982 extern void free_bootmem_with_active_regions(int nid,
983 unsigned long max_low_pfn);
984 extern void sparse_memory_present_with_active_regions(int nid);
985 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
986 extern int early_pfn_to_nid(unsigned long pfn);
987 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
988 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
989 extern void set_dma_reserve(unsigned long new_dma_reserve);
990 extern void memmap_init_zone(unsigned long, int, unsigned long,
991 unsigned long, enum memmap_context);
992 extern void setup_per_zone_pages_min(void);
993 extern void mem_init(void);
994 extern void show_mem(void);
995 extern void si_meminfo(struct sysinfo * val);
996 extern void si_meminfo_node(struct sysinfo *val, int nid);
999 extern void setup_per_cpu_pageset(void);
1001 static inline void setup_per_cpu_pageset(void) {}
1005 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1006 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1007 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1008 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1009 struct prio_tree_iter *iter);
1011 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1012 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1013 (vma = vma_prio_tree_next(vma, iter)); )
1015 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1016 struct list_head *list)
1018 vma->shared.vm_set.parent = NULL;
1019 list_add_tail(&vma->shared.vm_set.list, list);
1023 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1024 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1025 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1026 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1027 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1028 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1029 struct mempolicy *);
1030 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1031 extern int split_vma(struct mm_struct *,
1032 struct vm_area_struct *, unsigned long addr, int new_below);
1033 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1034 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1035 struct rb_node **, struct rb_node *);
1036 extern void unlink_file_vma(struct vm_area_struct *);
1037 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1038 unsigned long addr, unsigned long len, pgoff_t pgoff);
1039 extern void exit_mmap(struct mm_struct *);
1040 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1041 extern int install_special_mapping(struct mm_struct *mm,
1042 unsigned long addr, unsigned long len,
1043 unsigned long flags, struct page **pages);
1045 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1047 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1048 unsigned long len, unsigned long prot,
1049 unsigned long flag, unsigned long pgoff);
1050 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1051 unsigned long len, unsigned long flags,
1052 unsigned int vm_flags, unsigned long pgoff,
1055 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1056 unsigned long len, unsigned long prot,
1057 unsigned long flag, unsigned long offset)
1059 unsigned long ret = -EINVAL;
1060 if ((offset + PAGE_ALIGN(len)) < offset)
1062 if (!(offset & ~PAGE_MASK))
1063 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1068 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1070 extern unsigned long do_brk(unsigned long, unsigned long);
1073 extern unsigned long page_unuse(struct page *);
1074 extern void truncate_inode_pages(struct address_space *, loff_t);
1075 extern void truncate_inode_pages_range(struct address_space *,
1076 loff_t lstart, loff_t lend);
1078 /* generic vm_area_ops exported for stackable file systems */
1079 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1081 /* mm/page-writeback.c */
1082 int write_one_page(struct page *page, int wait);
1085 #define VM_MAX_READAHEAD 128 /* kbytes */
1086 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1088 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1089 pgoff_t offset, unsigned long nr_to_read);
1090 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1091 pgoff_t offset, unsigned long nr_to_read);
1093 void page_cache_sync_readahead(struct address_space *mapping,
1094 struct file_ra_state *ra,
1097 unsigned long size);
1099 void page_cache_async_readahead(struct address_space *mapping,
1100 struct file_ra_state *ra,
1104 unsigned long size);
1106 unsigned long max_sane_readahead(unsigned long nr);
1108 /* Do stack extension */
1109 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1111 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1113 extern int expand_stack_downwards(struct vm_area_struct *vma,
1114 unsigned long address);
1116 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1117 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1118 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1119 struct vm_area_struct **pprev);
1121 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1122 NULL if none. Assume start_addr < end_addr. */
1123 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1125 struct vm_area_struct * vma = find_vma(mm,start_addr);
1127 if (vma && end_addr <= vma->vm_start)
1132 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1134 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1137 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1138 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1139 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1140 unsigned long pfn, unsigned long size, pgprot_t);
1141 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1142 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1145 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1146 unsigned int foll_flags);
1147 #define FOLL_WRITE 0x01 /* check pte is writable */
1148 #define FOLL_TOUCH 0x02 /* mark page accessed */
1149 #define FOLL_GET 0x04 /* do get_page on page */
1150 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1152 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1154 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1155 unsigned long size, pte_fn_t fn, void *data);
1157 #ifdef CONFIG_PROC_FS
1158 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1160 static inline void vm_stat_account(struct mm_struct *mm,
1161 unsigned long flags, struct file *file, long pages)
1164 #endif /* CONFIG_PROC_FS */
1166 #ifdef CONFIG_DEBUG_PAGEALLOC
1167 extern int debug_pagealloc_enabled;
1169 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1171 static inline void enable_debug_pagealloc(void)
1173 debug_pagealloc_enabled = 1;
1175 #ifdef CONFIG_HIBERNATION
1176 extern bool kernel_page_present(struct page *page);
1177 #endif /* CONFIG_HIBERNATION */
1180 kernel_map_pages(struct page *page, int numpages, int enable) {}
1181 static inline void enable_debug_pagealloc(void)
1184 #ifdef CONFIG_HIBERNATION
1185 static inline bool kernel_page_present(struct page *page) { return true; }
1186 #endif /* CONFIG_HIBERNATION */
1189 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1190 #ifdef __HAVE_ARCH_GATE_AREA
1191 int in_gate_area_no_task(unsigned long addr);
1192 int in_gate_area(struct task_struct *task, unsigned long addr);
1194 int in_gate_area_no_task(unsigned long addr);
1195 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1196 #endif /* __HAVE_ARCH_GATE_AREA */
1198 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1199 void __user *, size_t *, loff_t *);
1200 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1201 unsigned long lru_pages);
1202 void drop_pagecache(void);
1203 void drop_slab(void);
1206 #define randomize_va_space 0
1208 extern int randomize_va_space;
1211 const char * arch_vma_name(struct vm_area_struct *vma);
1212 void print_vma_addr(char *prefix, unsigned long rip);
1214 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1215 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1216 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1217 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1218 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1219 void *vmemmap_alloc_block(unsigned long size, int node);
1220 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1221 int vmemmap_populate_basepages(struct page *start_page,
1222 unsigned long pages, int node);
1223 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1224 void vmemmap_populate_print_last(void);
1226 #endif /* __KERNEL__ */
1227 #endif /* _LINUX_MM_H */