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 */
110 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
111 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
113 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
114 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
117 #ifdef CONFIG_STACK_GROWSUP
118 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
120 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
123 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
124 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
125 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
126 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
127 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
130 * mapping from the currently active vm_flags protection bits (the
131 * low four bits) to a page protection mask..
133 extern pgprot_t protection_map[16];
135 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
136 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
140 * vm_fault is filled by the the pagefault handler and passed to the vma's
141 * ->fault function. The vma's ->fault is responsible for returning a bitmask
142 * of VM_FAULT_xxx flags that give details about how the fault was handled.
144 * pgoff should be used in favour of virtual_address, if possible. If pgoff
145 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
149 unsigned int flags; /* FAULT_FLAG_xxx flags */
150 pgoff_t pgoff; /* Logical page offset based on vma */
151 void __user *virtual_address; /* Faulting virtual address */
153 struct page *page; /* ->fault handlers should return a
154 * page here, unless VM_FAULT_NOPAGE
155 * is set (which is also implied by
161 * These are the virtual MM functions - opening of an area, closing and
162 * unmapping it (needed to keep files on disk up-to-date etc), pointer
163 * to the functions called when a no-page or a wp-page exception occurs.
165 struct vm_operations_struct {
166 void (*open)(struct vm_area_struct * area);
167 void (*close)(struct vm_area_struct * area);
168 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
169 unsigned long (*nopfn)(struct vm_area_struct *area,
170 unsigned long address);
172 /* notification that a previously read-only page is about to become
173 * writable, if an error is returned it will cause a SIGBUS */
174 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
177 * set_policy() op must add a reference to any non-NULL @new mempolicy
178 * to hold the policy upon return. Caller should pass NULL @new to
179 * remove a policy and fall back to surrounding context--i.e. do not
180 * install a MPOL_DEFAULT policy, nor the task or system default
183 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
186 * get_policy() op must add reference [mpol_get()] to any policy at
187 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
188 * in mm/mempolicy.c will do this automatically.
189 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
190 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
191 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
192 * must return NULL--i.e., do not "fallback" to task or system default
195 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
197 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
198 const nodemask_t *to, unsigned long flags);
205 #define page_private(page) ((page)->private)
206 #define set_page_private(page, v) ((page)->private = (v))
209 * FIXME: take this include out, include page-flags.h in
210 * files which need it (119 of them)
212 #include <linux/page-flags.h>
214 #ifdef CONFIG_DEBUG_VM
215 #define VM_BUG_ON(cond) BUG_ON(cond)
217 #define VM_BUG_ON(condition) do { } while(0)
221 * Methods to modify the page usage count.
223 * What counts for a page usage:
224 * - cache mapping (page->mapping)
225 * - private data (page->private)
226 * - page mapped in a task's page tables, each mapping
227 * is counted separately
229 * Also, many kernel routines increase the page count before a critical
230 * routine so they can be sure the page doesn't go away from under them.
234 * Drop a ref, return true if the refcount fell to zero (the page has no users)
236 static inline int put_page_testzero(struct page *page)
238 VM_BUG_ON(atomic_read(&page->_count) == 0);
239 return atomic_dec_and_test(&page->_count);
243 * Try to grab a ref unless the page has a refcount of zero, return false if
246 static inline int get_page_unless_zero(struct page *page)
248 VM_BUG_ON(PageTail(page));
249 return atomic_inc_not_zero(&page->_count);
252 /* Support for virtually mapped pages */
253 struct page *vmalloc_to_page(const void *addr);
254 unsigned long vmalloc_to_pfn(const void *addr);
257 * Determine if an address is within the vmalloc range
259 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
260 * is no special casing required.
262 static inline int is_vmalloc_addr(const void *x)
265 unsigned long addr = (unsigned long)x;
267 return addr >= VMALLOC_START && addr < VMALLOC_END;
273 static inline struct page *compound_head(struct page *page)
275 if (unlikely(PageTail(page)))
276 return page->first_page;
280 static inline int page_count(struct page *page)
282 return atomic_read(&compound_head(page)->_count);
285 static inline void get_page(struct page *page)
287 page = compound_head(page);
288 VM_BUG_ON(atomic_read(&page->_count) == 0);
289 atomic_inc(&page->_count);
292 static inline struct page *virt_to_head_page(const void *x)
294 struct page *page = virt_to_page(x);
295 return compound_head(page);
299 * Setup the page count before being freed into the page allocator for
300 * the first time (boot or memory hotplug)
302 static inline void init_page_count(struct page *page)
304 atomic_set(&page->_count, 1);
307 void put_page(struct page *page);
308 void put_pages_list(struct list_head *pages);
310 void split_page(struct page *page, unsigned int order);
313 * Compound pages have a destructor function. Provide a
314 * prototype for that function and accessor functions.
315 * These are _only_ valid on the head of a PG_compound page.
317 typedef void compound_page_dtor(struct page *);
319 static inline void set_compound_page_dtor(struct page *page,
320 compound_page_dtor *dtor)
322 page[1].lru.next = (void *)dtor;
325 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
327 return (compound_page_dtor *)page[1].lru.next;
330 static inline int compound_order(struct page *page)
334 return (unsigned long)page[1].lru.prev;
337 static inline void set_compound_order(struct page *page, unsigned long order)
339 page[1].lru.prev = (void *)order;
343 * Multiple processes may "see" the same page. E.g. for untouched
344 * mappings of /dev/null, all processes see the same page full of
345 * zeroes, and text pages of executables and shared libraries have
346 * only one copy in memory, at most, normally.
348 * For the non-reserved pages, page_count(page) denotes a reference count.
349 * page_count() == 0 means the page is free. page->lru is then used for
350 * freelist management in the buddy allocator.
351 * page_count() > 0 means the page has been allocated.
353 * Pages are allocated by the slab allocator in order to provide memory
354 * to kmalloc and kmem_cache_alloc. In this case, the management of the
355 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
356 * unless a particular usage is carefully commented. (the responsibility of
357 * freeing the kmalloc memory is the caller's, of course).
359 * A page may be used by anyone else who does a __get_free_page().
360 * In this case, page_count still tracks the references, and should only
361 * be used through the normal accessor functions. The top bits of page->flags
362 * and page->virtual store page management information, but all other fields
363 * are unused and could be used privately, carefully. The management of this
364 * page is the responsibility of the one who allocated it, and those who have
365 * subsequently been given references to it.
367 * The other pages (we may call them "pagecache pages") are completely
368 * managed by the Linux memory manager: I/O, buffers, swapping etc.
369 * The following discussion applies only to them.
371 * A pagecache page contains an opaque `private' member, which belongs to the
372 * page's address_space. Usually, this is the address of a circular list of
373 * the page's disk buffers. PG_private must be set to tell the VM to call
374 * into the filesystem to release these pages.
376 * A page may belong to an inode's memory mapping. In this case, page->mapping
377 * is the pointer to the inode, and page->index is the file offset of the page,
378 * in units of PAGE_CACHE_SIZE.
380 * If pagecache pages are not associated with an inode, they are said to be
381 * anonymous pages. These may become associated with the swapcache, and in that
382 * case PG_swapcache is set, and page->private is an offset into the swapcache.
384 * In either case (swapcache or inode backed), the pagecache itself holds one
385 * reference to the page. Setting PG_private should also increment the
386 * refcount. The each user mapping also has a reference to the page.
388 * The pagecache pages are stored in a per-mapping radix tree, which is
389 * rooted at mapping->page_tree, and indexed by offset.
390 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
391 * lists, we instead now tag pages as dirty/writeback in the radix tree.
393 * All pagecache pages may be subject to I/O:
394 * - inode pages may need to be read from disk,
395 * - inode pages which have been modified and are MAP_SHARED may need
396 * to be written back to the inode on disk,
397 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
398 * modified may need to be swapped out to swap space and (later) to be read
403 * The zone field is never updated after free_area_init_core()
404 * sets it, so none of the operations on it need to be atomic.
409 * page->flags layout:
411 * There are three possibilities for how page->flags get
412 * laid out. The first is for the normal case, without
413 * sparsemem. The second is for sparsemem when there is
414 * plenty of space for node and section. The last is when
415 * we have run out of space and have to fall back to an
416 * alternate (slower) way of determining the node.
418 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
419 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
420 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
422 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
423 #define SECTIONS_WIDTH SECTIONS_SHIFT
425 #define SECTIONS_WIDTH 0
428 #define ZONES_WIDTH ZONES_SHIFT
430 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
431 #define NODES_WIDTH NODES_SHIFT
433 #ifdef CONFIG_SPARSEMEM_VMEMMAP
434 #error "Vmemmap: No space for nodes field in page flags"
436 #define NODES_WIDTH 0
439 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
440 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
441 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
442 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
445 * We are going to use the flags for the page to node mapping if its in
446 * there. This includes the case where there is no node, so it is implicit.
448 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
449 #define NODE_NOT_IN_PAGE_FLAGS
452 #ifndef PFN_SECTION_SHIFT
453 #define PFN_SECTION_SHIFT 0
457 * Define the bit shifts to access each section. For non-existant
458 * sections we define the shift as 0; that plus a 0 mask ensures
459 * the compiler will optimise away reference to them.
461 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
462 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
463 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
465 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
466 #ifdef NODE_NOT_IN_PAGEFLAGS
467 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
468 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
469 SECTIONS_PGOFF : ZONES_PGOFF)
471 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
472 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
473 NODES_PGOFF : ZONES_PGOFF)
476 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
478 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
479 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
482 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
483 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
484 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
485 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
487 static inline enum zone_type page_zonenum(struct page *page)
489 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
493 * The identification function is only used by the buddy allocator for
494 * determining if two pages could be buddies. We are not really
495 * identifying a zone since we could be using a the section number
496 * id if we have not node id available in page flags.
497 * We guarantee only that it will return the same value for two
498 * combinable pages in a zone.
500 static inline int page_zone_id(struct page *page)
502 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
505 static inline int zone_to_nid(struct zone *zone)
514 #ifdef NODE_NOT_IN_PAGE_FLAGS
515 extern int page_to_nid(struct page *page);
517 static inline int page_to_nid(struct page *page)
519 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
523 static inline struct zone *page_zone(struct page *page)
525 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
528 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
529 static inline unsigned long page_to_section(struct page *page)
531 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
535 static inline void set_page_zone(struct page *page, enum zone_type zone)
537 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
538 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
541 static inline void set_page_node(struct page *page, unsigned long node)
543 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
544 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
547 static inline void set_page_section(struct page *page, unsigned long section)
549 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
550 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
553 static inline void set_page_links(struct page *page, enum zone_type zone,
554 unsigned long node, unsigned long pfn)
556 set_page_zone(page, zone);
557 set_page_node(page, node);
558 set_page_section(page, pfn_to_section_nr(pfn));
562 * If a hint addr is less than mmap_min_addr change hint to be as
563 * low as possible but still greater than mmap_min_addr
565 static inline unsigned long round_hint_to_min(unsigned long hint)
567 #ifdef CONFIG_SECURITY
569 if (((void *)hint != NULL) &&
570 (hint < mmap_min_addr))
571 return PAGE_ALIGN(mmap_min_addr);
577 * Some inline functions in vmstat.h depend on page_zone()
579 #include <linux/vmstat.h>
581 static __always_inline void *lowmem_page_address(struct page *page)
583 return __va(page_to_pfn(page) << PAGE_SHIFT);
586 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
587 #define HASHED_PAGE_VIRTUAL
590 #if defined(WANT_PAGE_VIRTUAL)
591 #define page_address(page) ((page)->virtual)
592 #define set_page_address(page, address) \
594 (page)->virtual = (address); \
596 #define page_address_init() do { } while(0)
599 #if defined(HASHED_PAGE_VIRTUAL)
600 void *page_address(struct page *page);
601 void set_page_address(struct page *page, void *virtual);
602 void page_address_init(void);
605 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
606 #define page_address(page) lowmem_page_address(page)
607 #define set_page_address(page, address) do { } while(0)
608 #define page_address_init() do { } while(0)
612 * On an anonymous page mapped into a user virtual memory area,
613 * page->mapping points to its anon_vma, not to a struct address_space;
614 * with the PAGE_MAPPING_ANON bit set to distinguish it.
616 * Please note that, confusingly, "page_mapping" refers to the inode
617 * address_space which maps the page from disk; whereas "page_mapped"
618 * refers to user virtual address space into which the page is mapped.
620 #define PAGE_MAPPING_ANON 1
622 extern struct address_space swapper_space;
623 static inline struct address_space *page_mapping(struct page *page)
625 struct address_space *mapping = page->mapping;
627 VM_BUG_ON(PageSlab(page));
629 if (unlikely(PageSwapCache(page)))
630 mapping = &swapper_space;
633 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
638 static inline int PageAnon(struct page *page)
640 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
644 * Return the pagecache index of the passed page. Regular pagecache pages
645 * use ->index whereas swapcache pages use ->private
647 static inline pgoff_t page_index(struct page *page)
649 if (unlikely(PageSwapCache(page)))
650 return page_private(page);
655 * The atomic page->_mapcount, like _count, starts from -1:
656 * so that transitions both from it and to it can be tracked,
657 * using atomic_inc_and_test and atomic_add_negative(-1).
659 static inline void reset_page_mapcount(struct page *page)
661 atomic_set(&(page)->_mapcount, -1);
664 static inline int page_mapcount(struct page *page)
666 return atomic_read(&(page)->_mapcount) + 1;
670 * Return true if this page is mapped into pagetables.
672 static inline int page_mapped(struct page *page)
674 return atomic_read(&(page)->_mapcount) >= 0;
678 * Error return values for the *_nopfn functions
680 #define NOPFN_SIGBUS ((unsigned long) -1)
681 #define NOPFN_OOM ((unsigned long) -2)
682 #define NOPFN_REFAULT ((unsigned long) -3)
685 * Different kinds of faults, as returned by handle_mm_fault().
686 * Used to decide whether a process gets delivered SIGBUS or
687 * just gets major/minor fault counters bumped up.
690 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
692 #define VM_FAULT_OOM 0x0001
693 #define VM_FAULT_SIGBUS 0x0002
694 #define VM_FAULT_MAJOR 0x0004
695 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
697 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
698 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
700 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
702 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
704 extern void show_free_areas(void);
707 int shmem_lock(struct file *file, int lock, struct user_struct *user);
709 static inline int shmem_lock(struct file *file, int lock,
710 struct user_struct *user)
715 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
717 int shmem_zero_setup(struct vm_area_struct *);
720 extern unsigned long shmem_get_unmapped_area(struct file *file,
724 unsigned long flags);
727 extern int can_do_mlock(void);
728 extern int user_shm_lock(size_t, struct user_struct *);
729 extern void user_shm_unlock(size_t, struct user_struct *);
732 * Parameter block passed down to zap_pte_range in exceptional cases.
735 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
736 struct address_space *check_mapping; /* Check page->mapping if set */
737 pgoff_t first_index; /* Lowest page->index to unmap */
738 pgoff_t last_index; /* Highest page->index to unmap */
739 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
740 unsigned long truncate_count; /* Compare vm_truncate_count */
743 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
746 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
747 unsigned long size, struct zap_details *);
748 unsigned long unmap_vmas(struct mmu_gather **tlb,
749 struct vm_area_struct *start_vma, unsigned long start_addr,
750 unsigned long end_addr, unsigned long *nr_accounted,
751 struct zap_details *);
754 * mm_walk - callbacks for walk_page_range
755 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
756 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
757 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
758 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
759 * @pte_hole: if set, called for each hole at all levels
761 * (see walk_page_range for more details)
764 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
765 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
766 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
767 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
768 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
769 struct mm_struct *mm;
773 int walk_page_range(unsigned long addr, unsigned long end,
774 struct mm_walk *walk);
775 void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
776 unsigned long end, unsigned long floor, unsigned long ceiling);
777 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
778 unsigned long floor, unsigned long ceiling);
779 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
780 struct vm_area_struct *vma);
781 void unmap_mapping_range(struct address_space *mapping,
782 loff_t const holebegin, loff_t const holelen, int even_cows);
784 static inline void unmap_shared_mapping_range(struct address_space *mapping,
785 loff_t const holebegin, loff_t const holelen)
787 unmap_mapping_range(mapping, holebegin, holelen, 0);
790 extern int vmtruncate(struct inode * inode, loff_t offset);
791 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
794 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
795 unsigned long address, int write_access);
797 static inline int handle_mm_fault(struct mm_struct *mm,
798 struct vm_area_struct *vma, unsigned long address,
801 /* should never happen if there's no MMU */
803 return VM_FAULT_SIGBUS;
807 extern int make_pages_present(unsigned long addr, unsigned long end);
808 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
810 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
811 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
812 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
814 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
815 extern void do_invalidatepage(struct page *page, unsigned long offset);
817 int __set_page_dirty_nobuffers(struct page *page);
818 int __set_page_dirty_no_writeback(struct page *page);
819 int redirty_page_for_writepage(struct writeback_control *wbc,
821 int set_page_dirty(struct page *page);
822 int set_page_dirty_lock(struct page *page);
823 int clear_page_dirty_for_io(struct page *page);
825 extern unsigned long move_page_tables(struct vm_area_struct *vma,
826 unsigned long old_addr, struct vm_area_struct *new_vma,
827 unsigned long new_addr, unsigned long len);
828 extern unsigned long do_mremap(unsigned long addr,
829 unsigned long old_len, unsigned long new_len,
830 unsigned long flags, unsigned long new_addr);
831 extern int mprotect_fixup(struct vm_area_struct *vma,
832 struct vm_area_struct **pprev, unsigned long start,
833 unsigned long end, unsigned long newflags);
836 * A callback you can register to apply pressure to ageable caches.
838 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
839 * look through the least-recently-used 'nr_to_scan' entries and
840 * attempt to free them up. It should return the number of objects
841 * which remain in the cache. If it returns -1, it means it cannot do
842 * any scanning at this time (eg. there is a risk of deadlock).
844 * The 'gfpmask' refers to the allocation we are currently trying to
847 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
848 * querying the cache size, so a fastpath for that case is appropriate.
851 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
852 int seeks; /* seeks to recreate an obj */
854 /* These are for internal use */
855 struct list_head list;
856 long nr; /* objs pending delete */
858 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
859 extern void register_shrinker(struct shrinker *);
860 extern void unregister_shrinker(struct shrinker *);
862 int vma_wants_writenotify(struct vm_area_struct *vma);
864 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
866 #ifdef __PAGETABLE_PUD_FOLDED
867 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
868 unsigned long address)
873 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
876 #ifdef __PAGETABLE_PMD_FOLDED
877 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
878 unsigned long address)
883 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
886 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
887 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
890 * The following ifdef needed to get the 4level-fixup.h header to work.
891 * Remove it when 4level-fixup.h has been removed.
893 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
894 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
896 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
897 NULL: pud_offset(pgd, address);
900 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
902 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
903 NULL: pmd_offset(pud, address);
905 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
907 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
909 * We tuck a spinlock to guard each pagetable page into its struct page,
910 * at page->private, with BUILD_BUG_ON to make sure that this will not
911 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
912 * When freeing, reset page->mapping so free_pages_check won't complain.
914 #define __pte_lockptr(page) &((page)->ptl)
915 #define pte_lock_init(_page) do { \
916 spin_lock_init(__pte_lockptr(_page)); \
918 #define pte_lock_deinit(page) ((page)->mapping = NULL)
919 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
922 * We use mm->page_table_lock to guard all pagetable pages of the mm.
924 #define pte_lock_init(page) do {} while (0)
925 #define pte_lock_deinit(page) do {} while (0)
926 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
927 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
929 static inline void pgtable_page_ctor(struct page *page)
932 inc_zone_page_state(page, NR_PAGETABLE);
935 static inline void pgtable_page_dtor(struct page *page)
937 pte_lock_deinit(page);
938 dec_zone_page_state(page, NR_PAGETABLE);
941 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
943 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
944 pte_t *__pte = pte_offset_map(pmd, address); \
950 #define pte_unmap_unlock(pte, ptl) do { \
955 #define pte_alloc_map(mm, pmd, address) \
956 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
957 NULL: pte_offset_map(pmd, address))
959 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
960 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
961 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
963 #define pte_alloc_kernel(pmd, address) \
964 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
965 NULL: pte_offset_kernel(pmd, address))
967 extern void free_area_init(unsigned long * zones_size);
968 extern void free_area_init_node(int nid, pg_data_t *pgdat,
969 unsigned long * zones_size, unsigned long zone_start_pfn,
970 unsigned long *zholes_size);
971 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
973 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
974 * zones, allocate the backing mem_map and account for memory holes in a more
975 * architecture independent manner. This is a substitute for creating the
976 * zone_sizes[] and zholes_size[] arrays and passing them to
977 * free_area_init_node()
979 * An architecture is expected to register range of page frames backed by
980 * physical memory with add_active_range() before calling
981 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
982 * usage, an architecture is expected to do something like
984 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
986 * for_each_valid_physical_page_range()
987 * add_active_range(node_id, start_pfn, end_pfn)
988 * free_area_init_nodes(max_zone_pfns);
990 * If the architecture guarantees that there are no holes in the ranges
991 * registered with add_active_range(), free_bootmem_active_regions()
992 * will call free_bootmem_node() for each registered physical page range.
993 * Similarly sparse_memory_present_with_active_regions() calls
994 * memory_present() for each range when SPARSEMEM is enabled.
996 * See mm/page_alloc.c for more information on each function exposed by
997 * CONFIG_ARCH_POPULATES_NODE_MAP
999 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1000 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1001 unsigned long end_pfn);
1002 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1003 unsigned long end_pfn);
1004 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1005 unsigned long end_pfn);
1006 extern void remove_all_active_ranges(void);
1007 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1008 unsigned long end_pfn);
1009 extern void get_pfn_range_for_nid(unsigned int nid,
1010 unsigned long *start_pfn, unsigned long *end_pfn);
1011 extern unsigned long find_min_pfn_with_active_regions(void);
1012 extern unsigned long find_max_pfn_with_active_regions(void);
1013 extern void free_bootmem_with_active_regions(int nid,
1014 unsigned long max_low_pfn);
1015 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1016 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1017 extern void sparse_memory_present_with_active_regions(int nid);
1018 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1019 extern int early_pfn_to_nid(unsigned long pfn);
1020 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1021 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1022 extern void set_dma_reserve(unsigned long new_dma_reserve);
1023 extern void memmap_init_zone(unsigned long, int, unsigned long,
1024 unsigned long, enum memmap_context);
1025 extern void setup_per_zone_pages_min(void);
1026 extern void mem_init(void);
1027 extern void show_mem(void);
1028 extern void si_meminfo(struct sysinfo * val);
1029 extern void si_meminfo_node(struct sysinfo *val, int nid);
1030 extern int after_bootmem;
1033 extern void setup_per_cpu_pageset(void);
1035 static inline void setup_per_cpu_pageset(void) {}
1039 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1040 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1041 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1042 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1043 struct prio_tree_iter *iter);
1045 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1046 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1047 (vma = vma_prio_tree_next(vma, iter)); )
1049 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1050 struct list_head *list)
1052 vma->shared.vm_set.parent = NULL;
1053 list_add_tail(&vma->shared.vm_set.list, list);
1057 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1058 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1059 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1060 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1061 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1062 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1063 struct mempolicy *);
1064 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1065 extern int split_vma(struct mm_struct *,
1066 struct vm_area_struct *, unsigned long addr, int new_below);
1067 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1068 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1069 struct rb_node **, struct rb_node *);
1070 extern void unlink_file_vma(struct vm_area_struct *);
1071 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1072 unsigned long addr, unsigned long len, pgoff_t pgoff);
1073 extern void exit_mmap(struct mm_struct *);
1075 #ifdef CONFIG_PROC_FS
1076 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1077 extern void added_exe_file_vma(struct mm_struct *mm);
1078 extern void removed_exe_file_vma(struct mm_struct *mm);
1080 static inline void added_exe_file_vma(struct mm_struct *mm)
1083 static inline void removed_exe_file_vma(struct mm_struct *mm)
1085 #endif /* CONFIG_PROC_FS */
1087 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1088 extern int install_special_mapping(struct mm_struct *mm,
1089 unsigned long addr, unsigned long len,
1090 unsigned long flags, struct page **pages);
1092 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1094 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1095 unsigned long len, unsigned long prot,
1096 unsigned long flag, unsigned long pgoff);
1097 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1098 unsigned long len, unsigned long flags,
1099 unsigned int vm_flags, unsigned long pgoff,
1102 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1103 unsigned long len, unsigned long prot,
1104 unsigned long flag, unsigned long offset)
1106 unsigned long ret = -EINVAL;
1107 if ((offset + PAGE_ALIGN(len)) < offset)
1109 if (!(offset & ~PAGE_MASK))
1110 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1115 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1117 extern unsigned long do_brk(unsigned long, unsigned long);
1120 extern unsigned long page_unuse(struct page *);
1121 extern void truncate_inode_pages(struct address_space *, loff_t);
1122 extern void truncate_inode_pages_range(struct address_space *,
1123 loff_t lstart, loff_t lend);
1125 /* generic vm_area_ops exported for stackable file systems */
1126 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1128 /* mm/page-writeback.c */
1129 int write_one_page(struct page *page, int wait);
1132 #define VM_MAX_READAHEAD 128 /* kbytes */
1133 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1135 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1136 pgoff_t offset, unsigned long nr_to_read);
1137 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1138 pgoff_t offset, unsigned long nr_to_read);
1140 void page_cache_sync_readahead(struct address_space *mapping,
1141 struct file_ra_state *ra,
1144 unsigned long size);
1146 void page_cache_async_readahead(struct address_space *mapping,
1147 struct file_ra_state *ra,
1151 unsigned long size);
1153 unsigned long max_sane_readahead(unsigned long nr);
1155 /* Do stack extension */
1156 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1158 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1160 extern int expand_stack_downwards(struct vm_area_struct *vma,
1161 unsigned long address);
1163 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1164 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1165 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1166 struct vm_area_struct **pprev);
1168 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1169 NULL if none. Assume start_addr < end_addr. */
1170 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1172 struct vm_area_struct * vma = find_vma(mm,start_addr);
1174 if (vma && end_addr <= vma->vm_start)
1179 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1181 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1184 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1185 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1186 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1187 unsigned long pfn, unsigned long size, pgprot_t);
1188 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1189 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1191 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1194 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1195 unsigned int foll_flags);
1196 #define FOLL_WRITE 0x01 /* check pte is writable */
1197 #define FOLL_TOUCH 0x02 /* mark page accessed */
1198 #define FOLL_GET 0x04 /* do get_page on page */
1199 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1201 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1203 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1204 unsigned long size, pte_fn_t fn, void *data);
1206 #ifdef CONFIG_PROC_FS
1207 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1209 static inline void vm_stat_account(struct mm_struct *mm,
1210 unsigned long flags, struct file *file, long pages)
1213 #endif /* CONFIG_PROC_FS */
1215 #ifdef CONFIG_DEBUG_PAGEALLOC
1216 extern int debug_pagealloc_enabled;
1218 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1220 static inline void enable_debug_pagealloc(void)
1222 debug_pagealloc_enabled = 1;
1224 #ifdef CONFIG_HIBERNATION
1225 extern bool kernel_page_present(struct page *page);
1226 #endif /* CONFIG_HIBERNATION */
1229 kernel_map_pages(struct page *page, int numpages, int enable) {}
1230 static inline void enable_debug_pagealloc(void)
1233 #ifdef CONFIG_HIBERNATION
1234 static inline bool kernel_page_present(struct page *page) { return true; }
1235 #endif /* CONFIG_HIBERNATION */
1238 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1239 #ifdef __HAVE_ARCH_GATE_AREA
1240 int in_gate_area_no_task(unsigned long addr);
1241 int in_gate_area(struct task_struct *task, unsigned long addr);
1243 int in_gate_area_no_task(unsigned long addr);
1244 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1245 #endif /* __HAVE_ARCH_GATE_AREA */
1247 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1248 void __user *, size_t *, loff_t *);
1249 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1250 unsigned long lru_pages);
1253 #define randomize_va_space 0
1255 extern int randomize_va_space;
1258 const char * arch_vma_name(struct vm_area_struct *vma);
1259 void print_vma_addr(char *prefix, unsigned long rip);
1261 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1262 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1263 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1264 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1265 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1266 void *vmemmap_alloc_block(unsigned long size, int node);
1267 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1268 int vmemmap_populate_basepages(struct page *start_page,
1269 unsigned long pages, int node);
1270 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1271 void vmemmap_populate_print_last(void);
1273 #endif /* __KERNEL__ */
1274 #endif /* _LINUX_MM_H */