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 or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
399 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
400 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
402 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
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 #ifdef CONFIG_SPARSEMEM_VMEMMAP
414 #error "Vmemmap: No space for nodes field in page flags"
416 #define NODES_WIDTH 0
419 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
420 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
421 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
422 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
425 * We are going to use the flags for the page to node mapping if its in
426 * there. This includes the case where there is no node, so it is implicit.
428 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
429 #define NODE_NOT_IN_PAGE_FLAGS
432 #ifndef PFN_SECTION_SHIFT
433 #define PFN_SECTION_SHIFT 0
437 * Define the bit shifts to access each section. For non-existant
438 * sections we define the shift as 0; that plus a 0 mask ensures
439 * the compiler will optimise away reference to them.
441 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
442 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
443 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
445 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
446 #ifdef NODE_NOT_IN_PAGEFLAGS
447 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
448 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
449 SECTIONS_PGOFF : ZONES_PGOFF)
451 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
452 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
453 NODES_PGOFF : ZONES_PGOFF)
456 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
458 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
459 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
462 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
463 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
464 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
465 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
467 static inline enum zone_type page_zonenum(struct page *page)
469 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
473 * The identification function is only used by the buddy allocator for
474 * determining if two pages could be buddies. We are not really
475 * identifying a zone since we could be using a the section number
476 * id if we have not node id available in page flags.
477 * We guarantee only that it will return the same value for two
478 * combinable pages in a zone.
480 static inline int page_zone_id(struct page *page)
482 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
485 static inline int zone_to_nid(struct zone *zone)
494 #ifdef NODE_NOT_IN_PAGE_FLAGS
495 extern int page_to_nid(struct page *page);
497 static inline int page_to_nid(struct page *page)
499 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
503 static inline struct zone *page_zone(struct page *page)
505 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
508 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
509 static inline unsigned long page_to_section(struct page *page)
511 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
515 static inline void set_page_zone(struct page *page, enum zone_type zone)
517 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
518 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
521 static inline void set_page_node(struct page *page, unsigned long node)
523 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
524 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
527 static inline void set_page_section(struct page *page, unsigned long section)
529 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
530 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
533 static inline void set_page_links(struct page *page, enum zone_type zone,
534 unsigned long node, unsigned long pfn)
536 set_page_zone(page, zone);
537 set_page_node(page, node);
538 set_page_section(page, pfn_to_section_nr(pfn));
542 * If a hint addr is less than mmap_min_addr change hint to be as
543 * low as possible but still greater than mmap_min_addr
545 static inline unsigned long round_hint_to_min(unsigned long hint)
547 #ifdef CONFIG_SECURITY
549 if (((void *)hint != NULL) &&
550 (hint < mmap_min_addr))
551 return PAGE_ALIGN(mmap_min_addr);
557 * Some inline functions in vmstat.h depend on page_zone()
559 #include <linux/vmstat.h>
561 static __always_inline void *lowmem_page_address(struct page *page)
563 return __va(page_to_pfn(page) << PAGE_SHIFT);
566 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
567 #define HASHED_PAGE_VIRTUAL
570 #if defined(WANT_PAGE_VIRTUAL)
571 #define page_address(page) ((page)->virtual)
572 #define set_page_address(page, address) \
574 (page)->virtual = (address); \
576 #define page_address_init() do { } while(0)
579 #if defined(HASHED_PAGE_VIRTUAL)
580 void *page_address(struct page *page);
581 void set_page_address(struct page *page, void *virtual);
582 void page_address_init(void);
585 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
586 #define page_address(page) lowmem_page_address(page)
587 #define set_page_address(page, address) do { } while(0)
588 #define page_address_init() do { } while(0)
592 * On an anonymous page mapped into a user virtual memory area,
593 * page->mapping points to its anon_vma, not to a struct address_space;
594 * with the PAGE_MAPPING_ANON bit set to distinguish it.
596 * Please note that, confusingly, "page_mapping" refers to the inode
597 * address_space which maps the page from disk; whereas "page_mapped"
598 * refers to user virtual address space into which the page is mapped.
600 #define PAGE_MAPPING_ANON 1
602 extern struct address_space swapper_space;
603 static inline struct address_space *page_mapping(struct page *page)
605 struct address_space *mapping = page->mapping;
607 VM_BUG_ON(PageSlab(page));
608 if (unlikely(PageSwapCache(page)))
609 mapping = &swapper_space;
610 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
615 static inline int PageAnon(struct page *page)
617 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
621 * Return the pagecache index of the passed page. Regular pagecache pages
622 * use ->index whereas swapcache pages use ->private
624 static inline pgoff_t page_index(struct page *page)
626 if (unlikely(PageSwapCache(page)))
627 return page_private(page);
632 * The atomic page->_mapcount, like _count, starts from -1:
633 * so that transitions both from it and to it can be tracked,
634 * using atomic_inc_and_test and atomic_add_negative(-1).
636 static inline void reset_page_mapcount(struct page *page)
638 atomic_set(&(page)->_mapcount, -1);
641 static inline int page_mapcount(struct page *page)
643 return atomic_read(&(page)->_mapcount) + 1;
647 * Return true if this page is mapped into pagetables.
649 static inline int page_mapped(struct page *page)
651 return atomic_read(&(page)->_mapcount) >= 0;
655 * Error return values for the *_nopfn functions
657 #define NOPFN_SIGBUS ((unsigned long) -1)
658 #define NOPFN_OOM ((unsigned long) -2)
659 #define NOPFN_REFAULT ((unsigned long) -3)
662 * Different kinds of faults, as returned by handle_mm_fault().
663 * Used to decide whether a process gets delivered SIGBUS or
664 * just gets major/minor fault counters bumped up.
667 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
669 #define VM_FAULT_OOM 0x0001
670 #define VM_FAULT_SIGBUS 0x0002
671 #define VM_FAULT_MAJOR 0x0004
672 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
674 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
675 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
677 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
679 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
681 extern void show_free_areas(void);
684 int shmem_lock(struct file *file, int lock, struct user_struct *user);
686 static inline int shmem_lock(struct file *file, int lock,
687 struct user_struct *user)
692 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
694 int shmem_zero_setup(struct vm_area_struct *);
697 extern unsigned long shmem_get_unmapped_area(struct file *file,
701 unsigned long flags);
704 extern int can_do_mlock(void);
705 extern int user_shm_lock(size_t, struct user_struct *);
706 extern void user_shm_unlock(size_t, struct user_struct *);
709 * Parameter block passed down to zap_pte_range in exceptional cases.
712 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
713 struct address_space *check_mapping; /* Check page->mapping if set */
714 pgoff_t first_index; /* Lowest page->index to unmap */
715 pgoff_t last_index; /* Highest page->index to unmap */
716 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
717 unsigned long truncate_count; /* Compare vm_truncate_count */
720 struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
721 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
722 unsigned long size, struct zap_details *);
723 unsigned long unmap_vmas(struct mmu_gather **tlb,
724 struct vm_area_struct *start_vma, unsigned long start_addr,
725 unsigned long end_addr, unsigned long *nr_accounted,
726 struct zap_details *);
729 * mm_walk - callbacks for walk_page_range
730 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
731 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
732 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
733 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
734 * @pte_hole: if set, called for each hole at all levels
736 * (see walk_page_range for more details)
739 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, void *);
740 int (*pud_entry)(pud_t *, unsigned long, unsigned long, void *);
741 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, void *);
742 int (*pte_entry)(pte_t *, unsigned long, unsigned long, void *);
743 int (*pte_hole)(unsigned long, unsigned long, void *);
746 int walk_page_range(const struct mm_struct *, unsigned long addr,
747 unsigned long end, const struct mm_walk *walk,
749 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
750 unsigned long end, unsigned long floor, unsigned long ceiling);
751 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
752 unsigned long floor, unsigned long ceiling);
753 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
754 struct vm_area_struct *vma);
755 void unmap_mapping_range(struct address_space *mapping,
756 loff_t const holebegin, loff_t const holelen, int even_cows);
758 static inline void unmap_shared_mapping_range(struct address_space *mapping,
759 loff_t const holebegin, loff_t const holelen)
761 unmap_mapping_range(mapping, holebegin, holelen, 0);
764 extern int vmtruncate(struct inode * inode, loff_t offset);
765 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
768 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
769 unsigned long address, int write_access);
771 static inline int handle_mm_fault(struct mm_struct *mm,
772 struct vm_area_struct *vma, unsigned long address,
775 /* should never happen if there's no MMU */
777 return VM_FAULT_SIGBUS;
781 extern int make_pages_present(unsigned long addr, unsigned long end);
782 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
784 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
785 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
786 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
788 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
789 extern void do_invalidatepage(struct page *page, unsigned long offset);
791 int __set_page_dirty_nobuffers(struct page *page);
792 int __set_page_dirty_no_writeback(struct page *page);
793 int redirty_page_for_writepage(struct writeback_control *wbc,
795 int set_page_dirty(struct page *page);
796 int set_page_dirty_lock(struct page *page);
797 int clear_page_dirty_for_io(struct page *page);
799 extern unsigned long move_page_tables(struct vm_area_struct *vma,
800 unsigned long old_addr, struct vm_area_struct *new_vma,
801 unsigned long new_addr, unsigned long len);
802 extern unsigned long do_mremap(unsigned long addr,
803 unsigned long old_len, unsigned long new_len,
804 unsigned long flags, unsigned long new_addr);
805 extern int mprotect_fixup(struct vm_area_struct *vma,
806 struct vm_area_struct **pprev, unsigned long start,
807 unsigned long end, unsigned long newflags);
810 * A callback you can register to apply pressure to ageable caches.
812 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
813 * look through the least-recently-used 'nr_to_scan' entries and
814 * attempt to free them up. It should return the number of objects
815 * which remain in the cache. If it returns -1, it means it cannot do
816 * any scanning at this time (eg. there is a risk of deadlock).
818 * The 'gfpmask' refers to the allocation we are currently trying to
821 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
822 * querying the cache size, so a fastpath for that case is appropriate.
825 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
826 int seeks; /* seeks to recreate an obj */
828 /* These are for internal use */
829 struct list_head list;
830 long nr; /* objs pending delete */
832 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
833 extern void register_shrinker(struct shrinker *);
834 extern void unregister_shrinker(struct shrinker *);
836 int vma_wants_writenotify(struct vm_area_struct *vma);
838 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
840 #ifdef __PAGETABLE_PUD_FOLDED
841 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
842 unsigned long address)
847 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
850 #ifdef __PAGETABLE_PMD_FOLDED
851 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
852 unsigned long address)
857 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
860 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
861 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
864 * The following ifdef needed to get the 4level-fixup.h header to work.
865 * Remove it when 4level-fixup.h has been removed.
867 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
868 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
870 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
871 NULL: pud_offset(pgd, address);
874 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
876 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
877 NULL: pmd_offset(pud, address);
879 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
881 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
883 * We tuck a spinlock to guard each pagetable page into its struct page,
884 * at page->private, with BUILD_BUG_ON to make sure that this will not
885 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
886 * When freeing, reset page->mapping so free_pages_check won't complain.
888 #define __pte_lockptr(page) &((page)->ptl)
889 #define pte_lock_init(_page) do { \
890 spin_lock_init(__pte_lockptr(_page)); \
892 #define pte_lock_deinit(page) ((page)->mapping = NULL)
893 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
896 * We use mm->page_table_lock to guard all pagetable pages of the mm.
898 #define pte_lock_init(page) do {} while (0)
899 #define pte_lock_deinit(page) do {} while (0)
900 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
901 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
903 static inline void pgtable_page_ctor(struct page *page)
906 inc_zone_page_state(page, NR_PAGETABLE);
909 static inline void pgtable_page_dtor(struct page *page)
911 pte_lock_deinit(page);
912 dec_zone_page_state(page, NR_PAGETABLE);
915 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
917 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
918 pte_t *__pte = pte_offset_map(pmd, address); \
924 #define pte_unmap_unlock(pte, ptl) do { \
929 #define pte_alloc_map(mm, pmd, address) \
930 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
931 NULL: pte_offset_map(pmd, address))
933 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
934 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
935 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
937 #define pte_alloc_kernel(pmd, address) \
938 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
939 NULL: pte_offset_kernel(pmd, address))
941 extern void free_area_init(unsigned long * zones_size);
942 extern void free_area_init_node(int nid, pg_data_t *pgdat,
943 unsigned long * zones_size, unsigned long zone_start_pfn,
944 unsigned long *zholes_size);
945 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
947 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
948 * zones, allocate the backing mem_map and account for memory holes in a more
949 * architecture independent manner. This is a substitute for creating the
950 * zone_sizes[] and zholes_size[] arrays and passing them to
951 * free_area_init_node()
953 * An architecture is expected to register range of page frames backed by
954 * physical memory with add_active_range() before calling
955 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
956 * usage, an architecture is expected to do something like
958 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
960 * for_each_valid_physical_page_range()
961 * add_active_range(node_id, start_pfn, end_pfn)
962 * free_area_init_nodes(max_zone_pfns);
964 * If the architecture guarantees that there are no holes in the ranges
965 * registered with add_active_range(), free_bootmem_active_regions()
966 * will call free_bootmem_node() for each registered physical page range.
967 * Similarly sparse_memory_present_with_active_regions() calls
968 * memory_present() for each range when SPARSEMEM is enabled.
970 * See mm/page_alloc.c for more information on each function exposed by
971 * CONFIG_ARCH_POPULATES_NODE_MAP
973 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
974 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
975 unsigned long end_pfn);
976 extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
977 unsigned long new_end_pfn);
978 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
979 unsigned long end_pfn);
980 extern void remove_all_active_ranges(void);
981 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
982 unsigned long end_pfn);
983 extern void get_pfn_range_for_nid(unsigned int nid,
984 unsigned long *start_pfn, unsigned long *end_pfn);
985 extern unsigned long find_min_pfn_with_active_regions(void);
986 extern unsigned long find_max_pfn_with_active_regions(void);
987 extern void free_bootmem_with_active_regions(int nid,
988 unsigned long max_low_pfn);
989 extern void sparse_memory_present_with_active_regions(int nid);
990 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
991 extern int early_pfn_to_nid(unsigned long pfn);
992 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
993 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
994 extern void set_dma_reserve(unsigned long new_dma_reserve);
995 extern void memmap_init_zone(unsigned long, int, unsigned long,
996 unsigned long, enum memmap_context);
997 extern void setup_per_zone_pages_min(void);
998 extern void mem_init(void);
999 extern void show_mem(void);
1000 extern void si_meminfo(struct sysinfo * val);
1001 extern void si_meminfo_node(struct sysinfo *val, int nid);
1004 extern void setup_per_cpu_pageset(void);
1006 static inline void setup_per_cpu_pageset(void) {}
1010 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1011 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1012 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1013 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1014 struct prio_tree_iter *iter);
1016 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1017 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1018 (vma = vma_prio_tree_next(vma, iter)); )
1020 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1021 struct list_head *list)
1023 vma->shared.vm_set.parent = NULL;
1024 list_add_tail(&vma->shared.vm_set.list, list);
1028 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1029 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1030 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1031 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1032 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1033 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1034 struct mempolicy *);
1035 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1036 extern int split_vma(struct mm_struct *,
1037 struct vm_area_struct *, unsigned long addr, int new_below);
1038 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1039 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1040 struct rb_node **, struct rb_node *);
1041 extern void unlink_file_vma(struct vm_area_struct *);
1042 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1043 unsigned long addr, unsigned long len, pgoff_t pgoff);
1044 extern void exit_mmap(struct mm_struct *);
1045 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1046 extern int install_special_mapping(struct mm_struct *mm,
1047 unsigned long addr, unsigned long len,
1048 unsigned long flags, struct page **pages);
1050 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1052 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1053 unsigned long len, unsigned long prot,
1054 unsigned long flag, unsigned long pgoff);
1055 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1056 unsigned long len, unsigned long flags,
1057 unsigned int vm_flags, unsigned long pgoff,
1060 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1061 unsigned long len, unsigned long prot,
1062 unsigned long flag, unsigned long offset)
1064 unsigned long ret = -EINVAL;
1065 if ((offset + PAGE_ALIGN(len)) < offset)
1067 if (!(offset & ~PAGE_MASK))
1068 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1073 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1075 extern unsigned long do_brk(unsigned long, unsigned long);
1078 extern unsigned long page_unuse(struct page *);
1079 extern void truncate_inode_pages(struct address_space *, loff_t);
1080 extern void truncate_inode_pages_range(struct address_space *,
1081 loff_t lstart, loff_t lend);
1083 /* generic vm_area_ops exported for stackable file systems */
1084 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1086 /* mm/page-writeback.c */
1087 int write_one_page(struct page *page, int wait);
1090 #define VM_MAX_READAHEAD 128 /* kbytes */
1091 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1093 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1094 pgoff_t offset, unsigned long nr_to_read);
1095 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1096 pgoff_t offset, unsigned long nr_to_read);
1098 void page_cache_sync_readahead(struct address_space *mapping,
1099 struct file_ra_state *ra,
1102 unsigned long size);
1104 void page_cache_async_readahead(struct address_space *mapping,
1105 struct file_ra_state *ra,
1109 unsigned long size);
1111 unsigned long max_sane_readahead(unsigned long nr);
1113 /* Do stack extension */
1114 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1116 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1118 extern int expand_stack_downwards(struct vm_area_struct *vma,
1119 unsigned long address);
1121 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1122 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1123 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1124 struct vm_area_struct **pprev);
1126 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1127 NULL if none. Assume start_addr < end_addr. */
1128 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1130 struct vm_area_struct * vma = find_vma(mm,start_addr);
1132 if (vma && end_addr <= vma->vm_start)
1137 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1139 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1142 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1143 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1144 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1145 unsigned long pfn, unsigned long size, pgprot_t);
1146 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1147 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1150 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1151 unsigned int foll_flags);
1152 #define FOLL_WRITE 0x01 /* check pte is writable */
1153 #define FOLL_TOUCH 0x02 /* mark page accessed */
1154 #define FOLL_GET 0x04 /* do get_page on page */
1155 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1157 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1159 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1160 unsigned long size, pte_fn_t fn, void *data);
1162 #ifdef CONFIG_PROC_FS
1163 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1165 static inline void vm_stat_account(struct mm_struct *mm,
1166 unsigned long flags, struct file *file, long pages)
1169 #endif /* CONFIG_PROC_FS */
1171 #ifdef CONFIG_DEBUG_PAGEALLOC
1172 extern int debug_pagealloc_enabled;
1174 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1176 static inline void enable_debug_pagealloc(void)
1178 debug_pagealloc_enabled = 1;
1180 #ifdef CONFIG_HIBERNATION
1181 extern bool kernel_page_present(struct page *page);
1182 #endif /* CONFIG_HIBERNATION */
1185 kernel_map_pages(struct page *page, int numpages, int enable) {}
1186 static inline void enable_debug_pagealloc(void)
1189 #ifdef CONFIG_HIBERNATION
1190 static inline bool kernel_page_present(struct page *page) { return true; }
1191 #endif /* CONFIG_HIBERNATION */
1194 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1195 #ifdef __HAVE_ARCH_GATE_AREA
1196 int in_gate_area_no_task(unsigned long addr);
1197 int in_gate_area(struct task_struct *task, unsigned long addr);
1199 int in_gate_area_no_task(unsigned long addr);
1200 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1201 #endif /* __HAVE_ARCH_GATE_AREA */
1203 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1204 void __user *, size_t *, loff_t *);
1205 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1206 unsigned long lru_pages);
1207 void drop_pagecache(void);
1208 void drop_slab(void);
1211 #define randomize_va_space 0
1213 extern int randomize_va_space;
1216 const char * arch_vma_name(struct vm_area_struct *vma);
1217 void print_vma_addr(char *prefix, unsigned long rip);
1219 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1220 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1221 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1222 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1223 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1224 void *vmemmap_alloc_block(unsigned long size, int node);
1225 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1226 int vmemmap_populate_basepages(struct page *start_page,
1227 unsigned long pages, int node);
1228 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1229 void vmemmap_populate_print_last(void);
1231 #endif /* __KERNEL__ */
1232 #endif /* _LINUX_MM_H */