4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmdebug.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/prio_tree.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
21 struct writeback_control;
23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24 extern unsigned long max_mapnr;
27 extern unsigned long num_physpages;
28 extern void * high_memory;
29 extern int page_cluster;
32 extern int sysctl_legacy_va_layout;
34 #define sysctl_legacy_va_layout 0
37 extern unsigned long mmap_min_addr;
40 #include <asm/pgtable.h>
41 #include <asm/processor.h>
43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45 /* to align the pointer to the (next) page boundary */
46 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
49 * Linux kernel virtual memory manager primitives.
50 * The idea being to have a "virtual" mm in the same way
51 * we have a virtual fs - giving a cleaner interface to the
52 * mm details, and allowing different kinds of memory mappings
53 * (from shared memory to executable loading to arbitrary
57 extern struct kmem_cache *vm_area_cachep;
60 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
61 * disabled, then there's a single shared list of VMAs maintained by the
62 * system, and mm's subscribe to these individually
64 struct vm_list_struct {
65 struct vm_list_struct *next;
66 struct vm_area_struct *vma;
70 extern struct rb_root nommu_vma_tree;
71 extern struct rw_semaphore nommu_vma_sem;
73 extern unsigned int kobjsize(const void *objp);
77 * vm_flags in vm_area_struct, see mm_types.h.
79 #define VM_READ 0x00000001 /* currently active flags */
80 #define VM_WRITE 0x00000002
81 #define VM_EXEC 0x00000004
82 #define VM_SHARED 0x00000008
84 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
85 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
86 #define VM_MAYWRITE 0x00000020
87 #define VM_MAYEXEC 0x00000040
88 #define VM_MAYSHARE 0x00000080
90 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
91 #define VM_GROWSUP 0x00000200
92 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
93 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
95 #define VM_EXECUTABLE 0x00001000
96 #define VM_LOCKED 0x00002000
97 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
99 /* Used by sys_madvise() */
100 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
101 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
103 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
104 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
105 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
106 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
107 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
108 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
109 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
110 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
111 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
112 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
114 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
115 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
116 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
118 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
119 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
122 #ifdef CONFIG_STACK_GROWSUP
123 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
125 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
128 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
129 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
130 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
131 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
132 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
135 * mapping from the currently active vm_flags protection bits (the
136 * low four bits) to a page protection mask..
138 extern pgprot_t protection_map[16];
140 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
141 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
145 * vm_fault is filled by the the pagefault handler and passed to the vma's
146 * ->fault function. The vma's ->fault is responsible for returning a bitmask
147 * of VM_FAULT_xxx flags that give details about how the fault was handled.
149 * pgoff should be used in favour of virtual_address, if possible. If pgoff
150 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
154 unsigned int flags; /* FAULT_FLAG_xxx flags */
155 pgoff_t pgoff; /* Logical page offset based on vma */
156 void __user *virtual_address; /* Faulting virtual address */
158 struct page *page; /* ->fault handlers should return a
159 * page here, unless VM_FAULT_NOPAGE
160 * is set (which is also implied by
166 * These are the virtual MM functions - opening of an area, closing and
167 * unmapping it (needed to keep files on disk up-to-date etc), pointer
168 * to the functions called when a no-page or a wp-page exception occurs.
170 struct vm_operations_struct {
171 void (*open)(struct vm_area_struct * area);
172 void (*close)(struct vm_area_struct * area);
173 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
175 /* notification that a previously read-only page is about to become
176 * writable, if an error is returned it will cause a SIGBUS */
177 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
179 /* called by access_process_vm when get_user_pages() fails, typically
180 * for use by special VMAs that can switch between memory and hardware
182 int (*access)(struct vm_area_struct *vma, unsigned long addr,
183 void *buf, int len, int write);
186 * set_policy() op must add a reference to any non-NULL @new mempolicy
187 * to hold the policy upon return. Caller should pass NULL @new to
188 * remove a policy and fall back to surrounding context--i.e. do not
189 * install a MPOL_DEFAULT policy, nor the task or system default
192 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
195 * get_policy() op must add reference [mpol_get()] to any policy at
196 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
197 * in mm/mempolicy.c will do this automatically.
198 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
199 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
200 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
201 * must return NULL--i.e., do not "fallback" to task or system default
204 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
206 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
207 const nodemask_t *to, unsigned long flags);
214 #define page_private(page) ((page)->private)
215 #define set_page_private(page, v) ((page)->private = (v))
218 * FIXME: take this include out, include page-flags.h in
219 * files which need it (119 of them)
221 #include <linux/page-flags.h>
224 * Methods to modify the page usage count.
226 * What counts for a page usage:
227 * - cache mapping (page->mapping)
228 * - private data (page->private)
229 * - page mapped in a task's page tables, each mapping
230 * is counted separately
232 * Also, many kernel routines increase the page count before a critical
233 * routine so they can be sure the page doesn't go away from under them.
237 * Drop a ref, return true if the refcount fell to zero (the page has no users)
239 static inline int put_page_testzero(struct page *page)
241 VM_BUG_ON(atomic_read(&page->_count) == 0);
242 return atomic_dec_and_test(&page->_count);
246 * Try to grab a ref unless the page has a refcount of zero, return false if
249 static inline int get_page_unless_zero(struct page *page)
251 VM_BUG_ON(PageTail(page));
252 return atomic_inc_not_zero(&page->_count);
255 /* Support for virtually mapped pages */
256 struct page *vmalloc_to_page(const void *addr);
257 unsigned long vmalloc_to_pfn(const void *addr);
260 * Determine if an address is within the vmalloc range
262 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
263 * is no special casing required.
265 static inline int is_vmalloc_addr(const void *x)
268 unsigned long addr = (unsigned long)x;
270 return addr >= VMALLOC_START && addr < VMALLOC_END;
276 static inline struct page *compound_head(struct page *page)
278 if (unlikely(PageTail(page)))
279 return page->first_page;
283 static inline int page_count(struct page *page)
285 return atomic_read(&compound_head(page)->_count);
288 static inline void get_page(struct page *page)
290 page = compound_head(page);
291 VM_BUG_ON(atomic_read(&page->_count) == 0);
292 atomic_inc(&page->_count);
295 static inline struct page *virt_to_head_page(const void *x)
297 struct page *page = virt_to_page(x);
298 return compound_head(page);
302 * Setup the page count before being freed into the page allocator for
303 * the first time (boot or memory hotplug)
305 static inline void init_page_count(struct page *page)
307 atomic_set(&page->_count, 1);
310 void put_page(struct page *page);
311 void put_pages_list(struct list_head *pages);
313 void split_page(struct page *page, unsigned int order);
316 * Compound pages have a destructor function. Provide a
317 * prototype for that function and accessor functions.
318 * These are _only_ valid on the head of a PG_compound page.
320 typedef void compound_page_dtor(struct page *);
322 static inline void set_compound_page_dtor(struct page *page,
323 compound_page_dtor *dtor)
325 page[1].lru.next = (void *)dtor;
328 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
330 return (compound_page_dtor *)page[1].lru.next;
333 static inline int compound_order(struct page *page)
337 return (unsigned long)page[1].lru.prev;
340 static inline void set_compound_order(struct page *page, unsigned long order)
342 page[1].lru.prev = (void *)order;
346 * Multiple processes may "see" the same page. E.g. for untouched
347 * mappings of /dev/null, all processes see the same page full of
348 * zeroes, and text pages of executables and shared libraries have
349 * only one copy in memory, at most, normally.
351 * For the non-reserved pages, page_count(page) denotes a reference count.
352 * page_count() == 0 means the page is free. page->lru is then used for
353 * freelist management in the buddy allocator.
354 * page_count() > 0 means the page has been allocated.
356 * Pages are allocated by the slab allocator in order to provide memory
357 * to kmalloc and kmem_cache_alloc. In this case, the management of the
358 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
359 * unless a particular usage is carefully commented. (the responsibility of
360 * freeing the kmalloc memory is the caller's, of course).
362 * A page may be used by anyone else who does a __get_free_page().
363 * In this case, page_count still tracks the references, and should only
364 * be used through the normal accessor functions. The top bits of page->flags
365 * and page->virtual store page management information, but all other fields
366 * are unused and could be used privately, carefully. The management of this
367 * page is the responsibility of the one who allocated it, and those who have
368 * subsequently been given references to it.
370 * The other pages (we may call them "pagecache pages") are completely
371 * managed by the Linux memory manager: I/O, buffers, swapping etc.
372 * The following discussion applies only to them.
374 * A pagecache page contains an opaque `private' member, which belongs to the
375 * page's address_space. Usually, this is the address of a circular list of
376 * the page's disk buffers. PG_private must be set to tell the VM to call
377 * into the filesystem to release these pages.
379 * A page may belong to an inode's memory mapping. In this case, page->mapping
380 * is the pointer to the inode, and page->index is the file offset of the page,
381 * in units of PAGE_CACHE_SIZE.
383 * If pagecache pages are not associated with an inode, they are said to be
384 * anonymous pages. These may become associated with the swapcache, and in that
385 * case PG_swapcache is set, and page->private is an offset into the swapcache.
387 * In either case (swapcache or inode backed), the pagecache itself holds one
388 * reference to the page. Setting PG_private should also increment the
389 * refcount. The each user mapping also has a reference to the page.
391 * The pagecache pages are stored in a per-mapping radix tree, which is
392 * rooted at mapping->page_tree, and indexed by offset.
393 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
394 * lists, we instead now tag pages as dirty/writeback in the radix tree.
396 * All pagecache pages may be subject to I/O:
397 * - inode pages may need to be read from disk,
398 * - inode pages which have been modified and are MAP_SHARED may need
399 * to be written back to the inode on disk,
400 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
401 * modified may need to be swapped out to swap space and (later) to be read
406 * The zone field is never updated after free_area_init_core()
407 * sets it, so none of the operations on it need to be atomic.
412 * page->flags layout:
414 * There are three possibilities for how page->flags get
415 * laid out. The first is for the normal case, without
416 * sparsemem. The second is for sparsemem when there is
417 * plenty of space for node and section. The last is when
418 * we have run out of space and have to fall back to an
419 * alternate (slower) way of determining the node.
421 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
422 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
423 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
425 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
426 #define SECTIONS_WIDTH SECTIONS_SHIFT
428 #define SECTIONS_WIDTH 0
431 #define ZONES_WIDTH ZONES_SHIFT
433 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
434 #define NODES_WIDTH NODES_SHIFT
436 #ifdef CONFIG_SPARSEMEM_VMEMMAP
437 #error "Vmemmap: No space for nodes field in page flags"
439 #define NODES_WIDTH 0
442 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
443 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
444 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
445 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
448 * We are going to use the flags for the page to node mapping if its in
449 * there. This includes the case where there is no node, so it is implicit.
451 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
452 #define NODE_NOT_IN_PAGE_FLAGS
455 #ifndef PFN_SECTION_SHIFT
456 #define PFN_SECTION_SHIFT 0
460 * Define the bit shifts to access each section. For non-existant
461 * sections we define the shift as 0; that plus a 0 mask ensures
462 * the compiler will optimise away reference to them.
464 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
465 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
466 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
468 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
469 #ifdef NODE_NOT_IN_PAGEFLAGS
470 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
471 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
472 SECTIONS_PGOFF : ZONES_PGOFF)
474 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
475 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
476 NODES_PGOFF : ZONES_PGOFF)
479 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
481 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
482 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
485 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
486 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
487 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
488 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
490 static inline enum zone_type page_zonenum(struct page *page)
492 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
496 * The identification function is only used by the buddy allocator for
497 * determining if two pages could be buddies. We are not really
498 * identifying a zone since we could be using a the section number
499 * id if we have not node id available in page flags.
500 * We guarantee only that it will return the same value for two
501 * combinable pages in a zone.
503 static inline int page_zone_id(struct page *page)
505 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
508 static inline int zone_to_nid(struct zone *zone)
517 #ifdef NODE_NOT_IN_PAGE_FLAGS
518 extern int page_to_nid(struct page *page);
520 static inline int page_to_nid(struct page *page)
522 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
526 static inline struct zone *page_zone(struct page *page)
528 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
531 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
532 static inline unsigned long page_to_section(struct page *page)
534 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
538 static inline void set_page_zone(struct page *page, enum zone_type zone)
540 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
541 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
544 static inline void set_page_node(struct page *page, unsigned long node)
546 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
547 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
550 static inline void set_page_section(struct page *page, unsigned long section)
552 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
553 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
556 static inline void set_page_links(struct page *page, enum zone_type zone,
557 unsigned long node, unsigned long pfn)
559 set_page_zone(page, zone);
560 set_page_node(page, node);
561 set_page_section(page, pfn_to_section_nr(pfn));
565 * If a hint addr is less than mmap_min_addr change hint to be as
566 * low as possible but still greater than mmap_min_addr
568 static inline unsigned long round_hint_to_min(unsigned long hint)
570 #ifdef CONFIG_SECURITY
572 if (((void *)hint != NULL) &&
573 (hint < mmap_min_addr))
574 return PAGE_ALIGN(mmap_min_addr);
580 * Some inline functions in vmstat.h depend on page_zone()
582 #include <linux/vmstat.h>
584 static __always_inline void *lowmem_page_address(struct page *page)
586 return __va(page_to_pfn(page) << PAGE_SHIFT);
589 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
590 #define HASHED_PAGE_VIRTUAL
593 #if defined(WANT_PAGE_VIRTUAL)
594 #define page_address(page) ((page)->virtual)
595 #define set_page_address(page, address) \
597 (page)->virtual = (address); \
599 #define page_address_init() do { } while(0)
602 #if defined(HASHED_PAGE_VIRTUAL)
603 void *page_address(struct page *page);
604 void set_page_address(struct page *page, void *virtual);
605 void page_address_init(void);
608 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
609 #define page_address(page) lowmem_page_address(page)
610 #define set_page_address(page, address) do { } while(0)
611 #define page_address_init() do { } while(0)
615 * On an anonymous page mapped into a user virtual memory area,
616 * page->mapping points to its anon_vma, not to a struct address_space;
617 * with the PAGE_MAPPING_ANON bit set to distinguish it.
619 * Please note that, confusingly, "page_mapping" refers to the inode
620 * address_space which maps the page from disk; whereas "page_mapped"
621 * refers to user virtual address space into which the page is mapped.
623 #define PAGE_MAPPING_ANON 1
625 extern struct address_space swapper_space;
626 static inline struct address_space *page_mapping(struct page *page)
628 struct address_space *mapping = page->mapping;
630 VM_BUG_ON(PageSlab(page));
632 if (unlikely(PageSwapCache(page)))
633 mapping = &swapper_space;
636 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
641 static inline int PageAnon(struct page *page)
643 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
647 * Return the pagecache index of the passed page. Regular pagecache pages
648 * use ->index whereas swapcache pages use ->private
650 static inline pgoff_t page_index(struct page *page)
652 if (unlikely(PageSwapCache(page)))
653 return page_private(page);
658 * The atomic page->_mapcount, like _count, starts from -1:
659 * so that transitions both from it and to it can be tracked,
660 * using atomic_inc_and_test and atomic_add_negative(-1).
662 static inline void reset_page_mapcount(struct page *page)
664 atomic_set(&(page)->_mapcount, -1);
667 static inline int page_mapcount(struct page *page)
669 return atomic_read(&(page)->_mapcount) + 1;
673 * Return true if this page is mapped into pagetables.
675 static inline int page_mapped(struct page *page)
677 return atomic_read(&(page)->_mapcount) >= 0;
681 * Different kinds of faults, as returned by handle_mm_fault().
682 * Used to decide whether a process gets delivered SIGBUS or
683 * just gets major/minor fault counters bumped up.
686 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
688 #define VM_FAULT_OOM 0x0001
689 #define VM_FAULT_SIGBUS 0x0002
690 #define VM_FAULT_MAJOR 0x0004
691 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
693 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
694 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
696 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
698 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
700 extern void show_free_areas(void);
703 int shmem_lock(struct file *file, int lock, struct user_struct *user);
705 static inline int shmem_lock(struct file *file, int lock,
706 struct user_struct *user)
711 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
713 int shmem_zero_setup(struct vm_area_struct *);
716 extern unsigned long shmem_get_unmapped_area(struct file *file,
720 unsigned long flags);
723 extern int can_do_mlock(void);
724 extern int user_shm_lock(size_t, struct user_struct *);
725 extern void user_shm_unlock(size_t, struct user_struct *);
728 * Parameter block passed down to zap_pte_range in exceptional cases.
731 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
732 struct address_space *check_mapping; /* Check page->mapping if set */
733 pgoff_t first_index; /* Lowest page->index to unmap */
734 pgoff_t last_index; /* Highest page->index to unmap */
735 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
736 unsigned long truncate_count; /* Compare vm_truncate_count */
739 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
742 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
744 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
745 unsigned long size, struct zap_details *);
746 unsigned long unmap_vmas(struct mmu_gather **tlb,
747 struct vm_area_struct *start_vma, unsigned long start_addr,
748 unsigned long end_addr, unsigned long *nr_accounted,
749 struct zap_details *);
752 * mm_walk - callbacks for walk_page_range
753 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
754 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
755 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
756 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
757 * @pte_hole: if set, called for each hole at all levels
759 * (see walk_page_range for more details)
762 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
763 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
764 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
765 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
766 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
767 struct mm_struct *mm;
771 int walk_page_range(unsigned long addr, unsigned long end,
772 struct mm_walk *walk);
773 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
774 unsigned long end, unsigned long floor, unsigned long ceiling);
775 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
776 struct vm_area_struct *vma);
777 void unmap_mapping_range(struct address_space *mapping,
778 loff_t const holebegin, loff_t const holelen, int even_cows);
779 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
780 void *buf, int len, int write);
782 static inline void unmap_shared_mapping_range(struct address_space *mapping,
783 loff_t const holebegin, loff_t const holelen)
785 unmap_mapping_range(mapping, holebegin, holelen, 0);
788 extern int vmtruncate(struct inode * inode, loff_t offset);
789 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
792 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
793 unsigned long address, int write_access);
795 static inline int handle_mm_fault(struct mm_struct *mm,
796 struct vm_area_struct *vma, unsigned long address,
799 /* should never happen if there's no MMU */
801 return VM_FAULT_SIGBUS;
805 extern int make_pages_present(unsigned long addr, unsigned long end);
806 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
808 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
809 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
811 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
812 extern void do_invalidatepage(struct page *page, unsigned long offset);
814 int __set_page_dirty_nobuffers(struct page *page);
815 int __set_page_dirty_no_writeback(struct page *page);
816 int redirty_page_for_writepage(struct writeback_control *wbc,
818 int set_page_dirty(struct page *page);
819 int set_page_dirty_lock(struct page *page);
820 int clear_page_dirty_for_io(struct page *page);
822 extern unsigned long move_page_tables(struct vm_area_struct *vma,
823 unsigned long old_addr, struct vm_area_struct *new_vma,
824 unsigned long new_addr, unsigned long len);
825 extern unsigned long do_mremap(unsigned long addr,
826 unsigned long old_len, unsigned long new_len,
827 unsigned long flags, unsigned long new_addr);
828 extern int mprotect_fixup(struct vm_area_struct *vma,
829 struct vm_area_struct **pprev, unsigned long start,
830 unsigned long end, unsigned long newflags);
833 * get_user_pages_fast provides equivalent functionality to get_user_pages,
834 * operating on current and current->mm (force=0 and doesn't return any vmas).
836 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
837 * can be made about locking. get_user_pages_fast is to be implemented in a
838 * way that is advantageous (vs get_user_pages()) when the user memory area is
839 * already faulted in and present in ptes. However if the pages have to be
840 * faulted in, it may turn out to be slightly slower).
842 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
843 struct page **pages);
846 * A callback you can register to apply pressure to ageable caches.
848 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
849 * look through the least-recently-used 'nr_to_scan' entries and
850 * attempt to free them up. It should return the number of objects
851 * which remain in the cache. If it returns -1, it means it cannot do
852 * any scanning at this time (eg. there is a risk of deadlock).
854 * The 'gfpmask' refers to the allocation we are currently trying to
857 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
858 * querying the cache size, so a fastpath for that case is appropriate.
861 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
862 int seeks; /* seeks to recreate an obj */
864 /* These are for internal use */
865 struct list_head list;
866 long nr; /* objs pending delete */
868 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
869 extern void register_shrinker(struct shrinker *);
870 extern void unregister_shrinker(struct shrinker *);
872 int vma_wants_writenotify(struct vm_area_struct *vma);
874 extern pte_t *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 USE_SPLIT_PTLOCKS
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)));})
930 #else /* !USE_SPLIT_PTLOCKS */
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 /* USE_SPLIT_PTLOCKS */
939 static inline void pgtable_page_ctor(struct page *page)
942 inc_zone_page_state(page, NR_PAGETABLE);
945 static inline void pgtable_page_dtor(struct page *page)
947 pte_lock_deinit(page);
948 dec_zone_page_state(page, NR_PAGETABLE);
951 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
953 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
954 pte_t *__pte = pte_offset_map(pmd, address); \
960 #define pte_unmap_unlock(pte, ptl) do { \
965 #define pte_alloc_map(mm, pmd, address) \
966 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
967 NULL: pte_offset_map(pmd, address))
969 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
970 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
971 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
973 #define pte_alloc_kernel(pmd, address) \
974 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
975 NULL: pte_offset_kernel(pmd, address))
977 extern void free_area_init(unsigned long * zones_size);
978 extern void free_area_init_node(int nid, unsigned long * zones_size,
979 unsigned long zone_start_pfn, unsigned long *zholes_size);
980 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
982 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
983 * zones, allocate the backing mem_map and account for memory holes in a more
984 * architecture independent manner. This is a substitute for creating the
985 * zone_sizes[] and zholes_size[] arrays and passing them to
986 * free_area_init_node()
988 * An architecture is expected to register range of page frames backed by
989 * physical memory with add_active_range() before calling
990 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
991 * usage, an architecture is expected to do something like
993 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
995 * for_each_valid_physical_page_range()
996 * add_active_range(node_id, start_pfn, end_pfn)
997 * free_area_init_nodes(max_zone_pfns);
999 * If the architecture guarantees that there are no holes in the ranges
1000 * registered with add_active_range(), free_bootmem_active_regions()
1001 * will call free_bootmem_node() for each registered physical page range.
1002 * Similarly sparse_memory_present_with_active_regions() calls
1003 * memory_present() for each range when SPARSEMEM is enabled.
1005 * See mm/page_alloc.c for more information on each function exposed by
1006 * CONFIG_ARCH_POPULATES_NODE_MAP
1008 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1009 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1010 unsigned long end_pfn);
1011 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1012 unsigned long end_pfn);
1013 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1014 unsigned long end_pfn);
1015 extern void remove_all_active_ranges(void);
1016 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1017 unsigned long end_pfn);
1018 extern void get_pfn_range_for_nid(unsigned int nid,
1019 unsigned long *start_pfn, unsigned long *end_pfn);
1020 extern unsigned long find_min_pfn_with_active_regions(void);
1021 extern void free_bootmem_with_active_regions(int nid,
1022 unsigned long max_low_pfn);
1023 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1024 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1025 extern void sparse_memory_present_with_active_regions(int nid);
1026 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1027 extern int early_pfn_to_nid(unsigned long pfn);
1028 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1029 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1030 extern void set_dma_reserve(unsigned long new_dma_reserve);
1031 extern void memmap_init_zone(unsigned long, int, unsigned long,
1032 unsigned long, enum memmap_context);
1033 extern void setup_per_zone_pages_min(void);
1034 extern void mem_init(void);
1035 extern void show_mem(void);
1036 extern void si_meminfo(struct sysinfo * val);
1037 extern void si_meminfo_node(struct sysinfo *val, int nid);
1038 extern int after_bootmem;
1041 extern void setup_per_cpu_pageset(void);
1043 static inline void setup_per_cpu_pageset(void) {}
1047 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1048 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1049 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1050 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1051 struct prio_tree_iter *iter);
1053 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1054 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1055 (vma = vma_prio_tree_next(vma, iter)); )
1057 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1058 struct list_head *list)
1060 vma->shared.vm_set.parent = NULL;
1061 list_add_tail(&vma->shared.vm_set.list, list);
1065 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1066 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1067 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1068 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1069 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1070 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1071 struct mempolicy *);
1072 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1073 extern int split_vma(struct mm_struct *,
1074 struct vm_area_struct *, unsigned long addr, int new_below);
1075 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1076 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1077 struct rb_node **, struct rb_node *);
1078 extern void unlink_file_vma(struct vm_area_struct *);
1079 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1080 unsigned long addr, unsigned long len, pgoff_t pgoff);
1081 extern void exit_mmap(struct mm_struct *);
1083 extern int mm_take_all_locks(struct mm_struct *mm);
1084 extern void mm_drop_all_locks(struct mm_struct *mm);
1086 #ifdef CONFIG_PROC_FS
1087 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1088 extern void added_exe_file_vma(struct mm_struct *mm);
1089 extern void removed_exe_file_vma(struct mm_struct *mm);
1091 static inline void added_exe_file_vma(struct mm_struct *mm)
1094 static inline void removed_exe_file_vma(struct mm_struct *mm)
1096 #endif /* CONFIG_PROC_FS */
1098 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1099 extern int install_special_mapping(struct mm_struct *mm,
1100 unsigned long addr, unsigned long len,
1101 unsigned long flags, struct page **pages);
1103 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1105 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1106 unsigned long len, unsigned long prot,
1107 unsigned long flag, unsigned long pgoff);
1108 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1109 unsigned long len, unsigned long flags,
1110 unsigned int vm_flags, unsigned long pgoff,
1113 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1114 unsigned long len, unsigned long prot,
1115 unsigned long flag, unsigned long offset)
1117 unsigned long ret = -EINVAL;
1118 if ((offset + PAGE_ALIGN(len)) < offset)
1120 if (!(offset & ~PAGE_MASK))
1121 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1126 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1128 extern unsigned long do_brk(unsigned long, unsigned long);
1131 extern unsigned long page_unuse(struct page *);
1132 extern void truncate_inode_pages(struct address_space *, loff_t);
1133 extern void truncate_inode_pages_range(struct address_space *,
1134 loff_t lstart, loff_t lend);
1136 /* generic vm_area_ops exported for stackable file systems */
1137 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1139 /* mm/page-writeback.c */
1140 int write_one_page(struct page *page, int wait);
1143 #define VM_MAX_READAHEAD 128 /* kbytes */
1144 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1146 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1147 pgoff_t offset, unsigned long nr_to_read);
1148 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1149 pgoff_t offset, unsigned long nr_to_read);
1151 void page_cache_sync_readahead(struct address_space *mapping,
1152 struct file_ra_state *ra,
1155 unsigned long size);
1157 void page_cache_async_readahead(struct address_space *mapping,
1158 struct file_ra_state *ra,
1162 unsigned long size);
1164 unsigned long max_sane_readahead(unsigned long nr);
1166 /* Do stack extension */
1167 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1169 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1171 extern int expand_stack_downwards(struct vm_area_struct *vma,
1172 unsigned long address);
1174 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1175 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1176 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1177 struct vm_area_struct **pprev);
1179 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1180 NULL if none. Assume start_addr < end_addr. */
1181 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1183 struct vm_area_struct * vma = find_vma(mm,start_addr);
1185 if (vma && end_addr <= vma->vm_start)
1190 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1192 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1195 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1196 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1197 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1198 unsigned long pfn, unsigned long size, pgprot_t);
1199 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1200 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1202 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1205 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1206 unsigned int foll_flags);
1207 #define FOLL_WRITE 0x01 /* check pte is writable */
1208 #define FOLL_TOUCH 0x02 /* mark page accessed */
1209 #define FOLL_GET 0x04 /* do get_page on page */
1210 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1212 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1214 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1215 unsigned long size, pte_fn_t fn, void *data);
1217 #ifdef CONFIG_PROC_FS
1218 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1220 static inline void vm_stat_account(struct mm_struct *mm,
1221 unsigned long flags, struct file *file, long pages)
1224 #endif /* CONFIG_PROC_FS */
1226 #ifdef CONFIG_DEBUG_PAGEALLOC
1227 extern int debug_pagealloc_enabled;
1229 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1231 static inline void enable_debug_pagealloc(void)
1233 debug_pagealloc_enabled = 1;
1235 #ifdef CONFIG_HIBERNATION
1236 extern bool kernel_page_present(struct page *page);
1237 #endif /* CONFIG_HIBERNATION */
1240 kernel_map_pages(struct page *page, int numpages, int enable) {}
1241 static inline void enable_debug_pagealloc(void)
1244 #ifdef CONFIG_HIBERNATION
1245 static inline bool kernel_page_present(struct page *page) { return true; }
1246 #endif /* CONFIG_HIBERNATION */
1249 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1250 #ifdef __HAVE_ARCH_GATE_AREA
1251 int in_gate_area_no_task(unsigned long addr);
1252 int in_gate_area(struct task_struct *task, unsigned long addr);
1254 int in_gate_area_no_task(unsigned long addr);
1255 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1256 #endif /* __HAVE_ARCH_GATE_AREA */
1258 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1259 void __user *, size_t *, loff_t *);
1260 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1261 unsigned long lru_pages);
1264 #define randomize_va_space 0
1266 extern int randomize_va_space;
1269 const char * arch_vma_name(struct vm_area_struct *vma);
1270 void print_vma_addr(char *prefix, unsigned long rip);
1272 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1273 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1274 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1275 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1276 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1277 void *vmemmap_alloc_block(unsigned long size, int node);
1278 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1279 int vmemmap_populate_basepages(struct page *start_page,
1280 unsigned long pages, int node);
1281 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1282 void vmemmap_populate_print_last(void);
1284 #endif /* __KERNEL__ */
1285 #endif /* _LINUX_MM_H */