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 */
112 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
113 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
116 #ifdef CONFIG_STACK_GROWSUP
117 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
119 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
122 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
123 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
124 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
125 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
126 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
129 * mapping from the currently active vm_flags protection bits (the
130 * low four bits) to a page protection mask..
132 extern pgprot_t protection_map[16];
134 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
135 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
139 * vm_fault is filled by the the pagefault handler and passed to the vma's
140 * ->fault function. The vma's ->fault is responsible for returning a bitmask
141 * of VM_FAULT_xxx flags that give details about how the fault was handled.
143 * pgoff should be used in favour of virtual_address, if possible. If pgoff
144 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
148 unsigned int flags; /* FAULT_FLAG_xxx flags */
149 pgoff_t pgoff; /* Logical page offset based on vma */
150 void __user *virtual_address; /* Faulting virtual address */
152 struct page *page; /* ->fault handlers should return a
153 * page here, unless VM_FAULT_NOPAGE
154 * is set (which is also implied by
160 * These are the virtual MM functions - opening of an area, closing and
161 * unmapping it (needed to keep files on disk up-to-date etc), pointer
162 * to the functions called when a no-page or a wp-page exception occurs.
164 struct vm_operations_struct {
165 void (*open)(struct vm_area_struct * area);
166 void (*close)(struct vm_area_struct * area);
167 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
168 unsigned long (*nopfn)(struct vm_area_struct *area,
169 unsigned long address);
171 /* notification that a previously read-only page is about to become
172 * writable, if an error is returned it will cause a SIGBUS */
173 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
176 * set_policy() op must add a reference to any non-NULL @new mempolicy
177 * to hold the policy upon return. Caller should pass NULL @new to
178 * remove a policy and fall back to surrounding context--i.e. do not
179 * install a MPOL_DEFAULT policy, nor the task or system default
182 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
185 * get_policy() op must add reference [mpol_get()] to any policy at
186 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
187 * in mm/mempolicy.c will do this automatically.
188 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
189 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
190 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
191 * must return NULL--i.e., do not "fallback" to task or system default
194 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
196 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
197 const nodemask_t *to, unsigned long flags);
204 #define page_private(page) ((page)->private)
205 #define set_page_private(page, v) ((page)->private = (v))
208 * FIXME: take this include out, include page-flags.h in
209 * files which need it (119 of them)
211 #include <linux/page-flags.h>
213 #ifdef CONFIG_DEBUG_VM
214 #define VM_BUG_ON(cond) BUG_ON(cond)
216 #define VM_BUG_ON(condition) do { } while(0)
220 * Methods to modify the page usage count.
222 * What counts for a page usage:
223 * - cache mapping (page->mapping)
224 * - private data (page->private)
225 * - page mapped in a task's page tables, each mapping
226 * is counted separately
228 * Also, many kernel routines increase the page count before a critical
229 * routine so they can be sure the page doesn't go away from under them.
233 * Drop a ref, return true if the refcount fell to zero (the page has no users)
235 static inline int put_page_testzero(struct page *page)
237 VM_BUG_ON(atomic_read(&page->_count) == 0);
238 return atomic_dec_and_test(&page->_count);
242 * Try to grab a ref unless the page has a refcount of zero, return false if
245 static inline int get_page_unless_zero(struct page *page)
247 VM_BUG_ON(PageTail(page));
248 return atomic_inc_not_zero(&page->_count);
251 /* Support for virtually mapped pages */
252 struct page *vmalloc_to_page(const void *addr);
253 unsigned long vmalloc_to_pfn(const void *addr);
256 * Determine if an address is within the vmalloc range
258 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
259 * is no special casing required.
261 static inline int is_vmalloc_addr(const void *x)
264 unsigned long addr = (unsigned long)x;
266 return addr >= VMALLOC_START && addr < VMALLOC_END;
272 static inline struct page *compound_head(struct page *page)
274 if (unlikely(PageTail(page)))
275 return page->first_page;
279 static inline int page_count(struct page *page)
281 return atomic_read(&compound_head(page)->_count);
284 static inline void get_page(struct page *page)
286 page = compound_head(page);
287 VM_BUG_ON(atomic_read(&page->_count) == 0);
288 atomic_inc(&page->_count);
291 static inline struct page *virt_to_head_page(const void *x)
293 struct page *page = virt_to_page(x);
294 return compound_head(page);
298 * Setup the page count before being freed into the page allocator for
299 * the first time (boot or memory hotplug)
301 static inline void init_page_count(struct page *page)
303 atomic_set(&page->_count, 1);
306 void put_page(struct page *page);
307 void put_pages_list(struct list_head *pages);
309 void split_page(struct page *page, unsigned int order);
312 * Compound pages have a destructor function. Provide a
313 * prototype for that function and accessor functions.
314 * These are _only_ valid on the head of a PG_compound page.
316 typedef void compound_page_dtor(struct page *);
318 static inline void set_compound_page_dtor(struct page *page,
319 compound_page_dtor *dtor)
321 page[1].lru.next = (void *)dtor;
324 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
326 return (compound_page_dtor *)page[1].lru.next;
329 static inline int compound_order(struct page *page)
333 return (unsigned long)page[1].lru.prev;
336 static inline void set_compound_order(struct page *page, unsigned long order)
338 page[1].lru.prev = (void *)order;
342 * Multiple processes may "see" the same page. E.g. for untouched
343 * mappings of /dev/null, all processes see the same page full of
344 * zeroes, and text pages of executables and shared libraries have
345 * only one copy in memory, at most, normally.
347 * For the non-reserved pages, page_count(page) denotes a reference count.
348 * page_count() == 0 means the page is free. page->lru is then used for
349 * freelist management in the buddy allocator.
350 * page_count() > 0 means the page has been allocated.
352 * Pages are allocated by the slab allocator in order to provide memory
353 * to kmalloc and kmem_cache_alloc. In this case, the management of the
354 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
355 * unless a particular usage is carefully commented. (the responsibility of
356 * freeing the kmalloc memory is the caller's, of course).
358 * A page may be used by anyone else who does a __get_free_page().
359 * In this case, page_count still tracks the references, and should only
360 * be used through the normal accessor functions. The top bits of page->flags
361 * and page->virtual store page management information, but all other fields
362 * are unused and could be used privately, carefully. The management of this
363 * page is the responsibility of the one who allocated it, and those who have
364 * subsequently been given references to it.
366 * The other pages (we may call them "pagecache pages") are completely
367 * managed by the Linux memory manager: I/O, buffers, swapping etc.
368 * The following discussion applies only to them.
370 * A pagecache page contains an opaque `private' member, which belongs to the
371 * page's address_space. Usually, this is the address of a circular list of
372 * the page's disk buffers. PG_private must be set to tell the VM to call
373 * into the filesystem to release these pages.
375 * A page may belong to an inode's memory mapping. In this case, page->mapping
376 * is the pointer to the inode, and page->index is the file offset of the page,
377 * in units of PAGE_CACHE_SIZE.
379 * If pagecache pages are not associated with an inode, they are said to be
380 * anonymous pages. These may become associated with the swapcache, and in that
381 * case PG_swapcache is set, and page->private is an offset into the swapcache.
383 * In either case (swapcache or inode backed), the pagecache itself holds one
384 * reference to the page. Setting PG_private should also increment the
385 * refcount. The each user mapping also has a reference to the page.
387 * The pagecache pages are stored in a per-mapping radix tree, which is
388 * rooted at mapping->page_tree, and indexed by offset.
389 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
390 * lists, we instead now tag pages as dirty/writeback in the radix tree.
392 * All pagecache pages may be subject to I/O:
393 * - inode pages may need to be read from disk,
394 * - inode pages which have been modified and are MAP_SHARED may need
395 * to be written back to the inode on disk,
396 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
397 * modified may need to be swapped out to swap space and (later) to be read
402 * The zone field is never updated after free_area_init_core()
403 * sets it, so none of the operations on it need to be atomic.
408 * page->flags layout:
410 * There are three possibilities for how page->flags get
411 * laid out. The first is for the normal case, without
412 * sparsemem. The second is for sparsemem when there is
413 * plenty of space for node and section. The last is when
414 * we have run out of space and have to fall back to an
415 * alternate (slower) way of determining the node.
417 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
418 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
419 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
421 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
422 #define SECTIONS_WIDTH SECTIONS_SHIFT
424 #define SECTIONS_WIDTH 0
427 #define ZONES_WIDTH ZONES_SHIFT
429 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
430 #define NODES_WIDTH NODES_SHIFT
432 #ifdef CONFIG_SPARSEMEM_VMEMMAP
433 #error "Vmemmap: No space for nodes field in page flags"
435 #define NODES_WIDTH 0
438 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
439 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
440 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
441 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
444 * We are going to use the flags for the page to node mapping if its in
445 * there. This includes the case where there is no node, so it is implicit.
447 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
448 #define NODE_NOT_IN_PAGE_FLAGS
451 #ifndef PFN_SECTION_SHIFT
452 #define PFN_SECTION_SHIFT 0
456 * Define the bit shifts to access each section. For non-existant
457 * sections we define the shift as 0; that plus a 0 mask ensures
458 * the compiler will optimise away reference to them.
460 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
461 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
462 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
464 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
465 #ifdef NODE_NOT_IN_PAGEFLAGS
466 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
467 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
468 SECTIONS_PGOFF : ZONES_PGOFF)
470 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
471 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
472 NODES_PGOFF : ZONES_PGOFF)
475 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
477 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
478 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
481 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
482 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
483 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
484 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
486 static inline enum zone_type page_zonenum(struct page *page)
488 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
492 * The identification function is only used by the buddy allocator for
493 * determining if two pages could be buddies. We are not really
494 * identifying a zone since we could be using a the section number
495 * id if we have not node id available in page flags.
496 * We guarantee only that it will return the same value for two
497 * combinable pages in a zone.
499 static inline int page_zone_id(struct page *page)
501 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
504 static inline int zone_to_nid(struct zone *zone)
513 #ifdef NODE_NOT_IN_PAGE_FLAGS
514 extern int page_to_nid(struct page *page);
516 static inline int page_to_nid(struct page *page)
518 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
522 static inline struct zone *page_zone(struct page *page)
524 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
527 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
528 static inline unsigned long page_to_section(struct page *page)
530 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
534 static inline void set_page_zone(struct page *page, enum zone_type zone)
536 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
537 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
540 static inline void set_page_node(struct page *page, unsigned long node)
542 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
543 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
546 static inline void set_page_section(struct page *page, unsigned long section)
548 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
549 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
552 static inline void set_page_links(struct page *page, enum zone_type zone,
553 unsigned long node, unsigned long pfn)
555 set_page_zone(page, zone);
556 set_page_node(page, node);
557 set_page_section(page, pfn_to_section_nr(pfn));
561 * If a hint addr is less than mmap_min_addr change hint to be as
562 * low as possible but still greater than mmap_min_addr
564 static inline unsigned long round_hint_to_min(unsigned long hint)
566 #ifdef CONFIG_SECURITY
568 if (((void *)hint != NULL) &&
569 (hint < mmap_min_addr))
570 return PAGE_ALIGN(mmap_min_addr);
576 * Some inline functions in vmstat.h depend on page_zone()
578 #include <linux/vmstat.h>
580 static __always_inline void *lowmem_page_address(struct page *page)
582 return __va(page_to_pfn(page) << PAGE_SHIFT);
585 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
586 #define HASHED_PAGE_VIRTUAL
589 #if defined(WANT_PAGE_VIRTUAL)
590 #define page_address(page) ((page)->virtual)
591 #define set_page_address(page, address) \
593 (page)->virtual = (address); \
595 #define page_address_init() do { } while(0)
598 #if defined(HASHED_PAGE_VIRTUAL)
599 void *page_address(struct page *page);
600 void set_page_address(struct page *page, void *virtual);
601 void page_address_init(void);
604 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
605 #define page_address(page) lowmem_page_address(page)
606 #define set_page_address(page, address) do { } while(0)
607 #define page_address_init() do { } while(0)
611 * On an anonymous page mapped into a user virtual memory area,
612 * page->mapping points to its anon_vma, not to a struct address_space;
613 * with the PAGE_MAPPING_ANON bit set to distinguish it.
615 * Please note that, confusingly, "page_mapping" refers to the inode
616 * address_space which maps the page from disk; whereas "page_mapped"
617 * refers to user virtual address space into which the page is mapped.
619 #define PAGE_MAPPING_ANON 1
621 extern struct address_space swapper_space;
622 static inline struct address_space *page_mapping(struct page *page)
624 struct address_space *mapping = page->mapping;
626 VM_BUG_ON(PageSlab(page));
628 if (unlikely(PageSwapCache(page)))
629 mapping = &swapper_space;
632 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
637 static inline int PageAnon(struct page *page)
639 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
643 * Return the pagecache index of the passed page. Regular pagecache pages
644 * use ->index whereas swapcache pages use ->private
646 static inline pgoff_t page_index(struct page *page)
648 if (unlikely(PageSwapCache(page)))
649 return page_private(page);
654 * The atomic page->_mapcount, like _count, starts from -1:
655 * so that transitions both from it and to it can be tracked,
656 * using atomic_inc_and_test and atomic_add_negative(-1).
658 static inline void reset_page_mapcount(struct page *page)
660 atomic_set(&(page)->_mapcount, -1);
663 static inline int page_mapcount(struct page *page)
665 return atomic_read(&(page)->_mapcount) + 1;
669 * Return true if this page is mapped into pagetables.
671 static inline int page_mapped(struct page *page)
673 return atomic_read(&(page)->_mapcount) >= 0;
677 * Error return values for the *_nopfn functions
679 #define NOPFN_SIGBUS ((unsigned long) -1)
680 #define NOPFN_OOM ((unsigned long) -2)
681 #define NOPFN_REFAULT ((unsigned long) -3)
684 * Different kinds of faults, as returned by handle_mm_fault().
685 * Used to decide whether a process gets delivered SIGBUS or
686 * just gets major/minor fault counters bumped up.
689 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
691 #define VM_FAULT_OOM 0x0001
692 #define VM_FAULT_SIGBUS 0x0002
693 #define VM_FAULT_MAJOR 0x0004
694 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
696 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
697 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
699 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
701 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
703 extern void show_free_areas(void);
706 int shmem_lock(struct file *file, int lock, struct user_struct *user);
708 static inline int shmem_lock(struct file *file, int lock,
709 struct user_struct *user)
714 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
716 int shmem_zero_setup(struct vm_area_struct *);
719 extern unsigned long shmem_get_unmapped_area(struct file *file,
723 unsigned long flags);
726 extern int can_do_mlock(void);
727 extern int user_shm_lock(size_t, struct user_struct *);
728 extern void user_shm_unlock(size_t, struct user_struct *);
731 * Parameter block passed down to zap_pte_range in exceptional cases.
734 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
735 struct address_space *check_mapping; /* Check page->mapping if set */
736 pgoff_t first_index; /* Lowest page->index to unmap */
737 pgoff_t last_index; /* Highest page->index to unmap */
738 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
739 unsigned long truncate_count; /* Compare vm_truncate_count */
742 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
745 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
746 unsigned long size, struct zap_details *);
747 unsigned long unmap_vmas(struct mmu_gather **tlb,
748 struct vm_area_struct *start_vma, unsigned long start_addr,
749 unsigned long end_addr, unsigned long *nr_accounted,
750 struct zap_details *);
753 * mm_walk - callbacks for walk_page_range
754 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
755 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
756 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
757 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
758 * @pte_hole: if set, called for each hole at all levels
760 * (see walk_page_range for more details)
763 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, void *);
764 int (*pud_entry)(pud_t *, unsigned long, unsigned long, void *);
765 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, void *);
766 int (*pte_entry)(pte_t *, unsigned long, unsigned long, void *);
767 int (*pte_hole)(unsigned long, unsigned long, void *);
770 int walk_page_range(const struct mm_struct *, unsigned long addr,
771 unsigned long end, const 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 void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
776 unsigned long floor, unsigned long ceiling);
777 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
778 struct vm_area_struct *vma);
779 void unmap_mapping_range(struct address_space *mapping,
780 loff_t const holebegin, loff_t const holelen, int even_cows);
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);
810 void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
812 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
813 extern void do_invalidatepage(struct page *page, unsigned long offset);
815 int __set_page_dirty_nobuffers(struct page *page);
816 int __set_page_dirty_no_writeback(struct page *page);
817 int redirty_page_for_writepage(struct writeback_control *wbc,
819 int set_page_dirty(struct page *page);
820 int set_page_dirty_lock(struct page *page);
821 int clear_page_dirty_for_io(struct page *page);
823 extern unsigned long move_page_tables(struct vm_area_struct *vma,
824 unsigned long old_addr, struct vm_area_struct *new_vma,
825 unsigned long new_addr, unsigned long len);
826 extern unsigned long do_mremap(unsigned long addr,
827 unsigned long old_len, unsigned long new_len,
828 unsigned long flags, unsigned long new_addr);
829 extern int mprotect_fixup(struct vm_area_struct *vma,
830 struct vm_area_struct **pprev, unsigned long start,
831 unsigned long end, unsigned long newflags);
834 * A callback you can register to apply pressure to ageable caches.
836 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
837 * look through the least-recently-used 'nr_to_scan' entries and
838 * attempt to free them up. It should return the number of objects
839 * which remain in the cache. If it returns -1, it means it cannot do
840 * any scanning at this time (eg. there is a risk of deadlock).
842 * The 'gfpmask' refers to the allocation we are currently trying to
845 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
846 * querying the cache size, so a fastpath for that case is appropriate.
849 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
850 int seeks; /* seeks to recreate an obj */
852 /* These are for internal use */
853 struct list_head list;
854 long nr; /* objs pending delete */
856 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
857 extern void register_shrinker(struct shrinker *);
858 extern void unregister_shrinker(struct shrinker *);
860 int vma_wants_writenotify(struct vm_area_struct *vma);
862 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
864 #ifdef __PAGETABLE_PUD_FOLDED
865 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
866 unsigned long address)
871 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
874 #ifdef __PAGETABLE_PMD_FOLDED
875 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
876 unsigned long address)
881 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
884 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
885 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
888 * The following ifdef needed to get the 4level-fixup.h header to work.
889 * Remove it when 4level-fixup.h has been removed.
891 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
892 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
894 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
895 NULL: pud_offset(pgd, address);
898 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
900 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
901 NULL: pmd_offset(pud, address);
903 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
905 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
907 * We tuck a spinlock to guard each pagetable page into its struct page,
908 * at page->private, with BUILD_BUG_ON to make sure that this will not
909 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
910 * When freeing, reset page->mapping so free_pages_check won't complain.
912 #define __pte_lockptr(page) &((page)->ptl)
913 #define pte_lock_init(_page) do { \
914 spin_lock_init(__pte_lockptr(_page)); \
916 #define pte_lock_deinit(page) ((page)->mapping = NULL)
917 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
920 * We use mm->page_table_lock to guard all pagetable pages of the mm.
922 #define pte_lock_init(page) do {} while (0)
923 #define pte_lock_deinit(page) do {} while (0)
924 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
925 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
927 static inline void pgtable_page_ctor(struct page *page)
930 inc_zone_page_state(page, NR_PAGETABLE);
933 static inline void pgtable_page_dtor(struct page *page)
935 pte_lock_deinit(page);
936 dec_zone_page_state(page, NR_PAGETABLE);
939 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
941 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
942 pte_t *__pte = pte_offset_map(pmd, address); \
948 #define pte_unmap_unlock(pte, ptl) do { \
953 #define pte_alloc_map(mm, pmd, address) \
954 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
955 NULL: pte_offset_map(pmd, address))
957 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
958 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
959 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
961 #define pte_alloc_kernel(pmd, address) \
962 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
963 NULL: pte_offset_kernel(pmd, address))
965 extern void free_area_init(unsigned long * zones_size);
966 extern void free_area_init_node(int nid, pg_data_t *pgdat,
967 unsigned long * zones_size, unsigned long zone_start_pfn,
968 unsigned long *zholes_size);
969 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
971 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
972 * zones, allocate the backing mem_map and account for memory holes in a more
973 * architecture independent manner. This is a substitute for creating the
974 * zone_sizes[] and zholes_size[] arrays and passing them to
975 * free_area_init_node()
977 * An architecture is expected to register range of page frames backed by
978 * physical memory with add_active_range() before calling
979 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
980 * usage, an architecture is expected to do something like
982 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
984 * for_each_valid_physical_page_range()
985 * add_active_range(node_id, start_pfn, end_pfn)
986 * free_area_init_nodes(max_zone_pfns);
988 * If the architecture guarantees that there are no holes in the ranges
989 * registered with add_active_range(), free_bootmem_active_regions()
990 * will call free_bootmem_node() for each registered physical page range.
991 * Similarly sparse_memory_present_with_active_regions() calls
992 * memory_present() for each range when SPARSEMEM is enabled.
994 * See mm/page_alloc.c for more information on each function exposed by
995 * CONFIG_ARCH_POPULATES_NODE_MAP
997 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
998 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
999 unsigned long end_pfn);
1000 extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn,
1001 unsigned long new_end_pfn);
1002 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1003 unsigned long end_pfn);
1004 extern void remove_all_active_ranges(void);
1005 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1006 unsigned long end_pfn);
1007 extern void get_pfn_range_for_nid(unsigned int nid,
1008 unsigned long *start_pfn, unsigned long *end_pfn);
1009 extern unsigned long find_min_pfn_with_active_regions(void);
1010 extern unsigned long find_max_pfn_with_active_regions(void);
1011 extern void free_bootmem_with_active_regions(int nid,
1012 unsigned long max_low_pfn);
1013 extern void sparse_memory_present_with_active_regions(int nid);
1014 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1015 extern int early_pfn_to_nid(unsigned long pfn);
1016 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1017 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1018 extern void set_dma_reserve(unsigned long new_dma_reserve);
1019 extern void memmap_init_zone(unsigned long, int, unsigned long,
1020 unsigned long, enum memmap_context);
1021 extern void setup_per_zone_pages_min(void);
1022 extern void mem_init(void);
1023 extern void show_mem(void);
1024 extern void si_meminfo(struct sysinfo * val);
1025 extern void si_meminfo_node(struct sysinfo *val, int nid);
1028 extern void setup_per_cpu_pageset(void);
1030 static inline void setup_per_cpu_pageset(void) {}
1034 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1035 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1036 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1037 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1038 struct prio_tree_iter *iter);
1040 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1041 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1042 (vma = vma_prio_tree_next(vma, iter)); )
1044 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1045 struct list_head *list)
1047 vma->shared.vm_set.parent = NULL;
1048 list_add_tail(&vma->shared.vm_set.list, list);
1052 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1053 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1054 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1055 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1056 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1057 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1058 struct mempolicy *);
1059 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1060 extern int split_vma(struct mm_struct *,
1061 struct vm_area_struct *, unsigned long addr, int new_below);
1062 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1063 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1064 struct rb_node **, struct rb_node *);
1065 extern void unlink_file_vma(struct vm_area_struct *);
1066 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1067 unsigned long addr, unsigned long len, pgoff_t pgoff);
1068 extern void exit_mmap(struct mm_struct *);
1070 #ifdef CONFIG_PROC_FS
1071 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1072 extern void added_exe_file_vma(struct mm_struct *mm);
1073 extern void removed_exe_file_vma(struct mm_struct *mm);
1075 static inline void added_exe_file_vma(struct mm_struct *mm)
1078 static inline void removed_exe_file_vma(struct mm_struct *mm)
1080 #endif /* CONFIG_PROC_FS */
1082 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1083 extern int install_special_mapping(struct mm_struct *mm,
1084 unsigned long addr, unsigned long len,
1085 unsigned long flags, struct page **pages);
1087 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1089 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1090 unsigned long len, unsigned long prot,
1091 unsigned long flag, unsigned long pgoff);
1092 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1093 unsigned long len, unsigned long flags,
1094 unsigned int vm_flags, unsigned long pgoff,
1097 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1098 unsigned long len, unsigned long prot,
1099 unsigned long flag, unsigned long offset)
1101 unsigned long ret = -EINVAL;
1102 if ((offset + PAGE_ALIGN(len)) < offset)
1104 if (!(offset & ~PAGE_MASK))
1105 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1110 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1112 extern unsigned long do_brk(unsigned long, unsigned long);
1115 extern unsigned long page_unuse(struct page *);
1116 extern void truncate_inode_pages(struct address_space *, loff_t);
1117 extern void truncate_inode_pages_range(struct address_space *,
1118 loff_t lstart, loff_t lend);
1120 /* generic vm_area_ops exported for stackable file systems */
1121 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1123 /* mm/page-writeback.c */
1124 int write_one_page(struct page *page, int wait);
1127 #define VM_MAX_READAHEAD 128 /* kbytes */
1128 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1130 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1131 pgoff_t offset, unsigned long nr_to_read);
1132 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1133 pgoff_t offset, unsigned long nr_to_read);
1135 void page_cache_sync_readahead(struct address_space *mapping,
1136 struct file_ra_state *ra,
1139 unsigned long size);
1141 void page_cache_async_readahead(struct address_space *mapping,
1142 struct file_ra_state *ra,
1146 unsigned long size);
1148 unsigned long max_sane_readahead(unsigned long nr);
1150 /* Do stack extension */
1151 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1153 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1155 extern int expand_stack_downwards(struct vm_area_struct *vma,
1156 unsigned long address);
1158 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1159 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1160 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1161 struct vm_area_struct **pprev);
1163 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1164 NULL if none. Assume start_addr < end_addr. */
1165 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1167 struct vm_area_struct * vma = find_vma(mm,start_addr);
1169 if (vma && end_addr <= vma->vm_start)
1174 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1176 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1179 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1180 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1181 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1182 unsigned long pfn, unsigned long size, pgprot_t);
1183 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1184 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1186 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1189 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1190 unsigned int foll_flags);
1191 #define FOLL_WRITE 0x01 /* check pte is writable */
1192 #define FOLL_TOUCH 0x02 /* mark page accessed */
1193 #define FOLL_GET 0x04 /* do get_page on page */
1194 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1196 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1198 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1199 unsigned long size, pte_fn_t fn, void *data);
1201 #ifdef CONFIG_PROC_FS
1202 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1204 static inline void vm_stat_account(struct mm_struct *mm,
1205 unsigned long flags, struct file *file, long pages)
1208 #endif /* CONFIG_PROC_FS */
1210 #ifdef CONFIG_DEBUG_PAGEALLOC
1211 extern int debug_pagealloc_enabled;
1213 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1215 static inline void enable_debug_pagealloc(void)
1217 debug_pagealloc_enabled = 1;
1219 #ifdef CONFIG_HIBERNATION
1220 extern bool kernel_page_present(struct page *page);
1221 #endif /* CONFIG_HIBERNATION */
1224 kernel_map_pages(struct page *page, int numpages, int enable) {}
1225 static inline void enable_debug_pagealloc(void)
1228 #ifdef CONFIG_HIBERNATION
1229 static inline bool kernel_page_present(struct page *page) { return true; }
1230 #endif /* CONFIG_HIBERNATION */
1233 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1234 #ifdef __HAVE_ARCH_GATE_AREA
1235 int in_gate_area_no_task(unsigned long addr);
1236 int in_gate_area(struct task_struct *task, unsigned long addr);
1238 int in_gate_area_no_task(unsigned long addr);
1239 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1240 #endif /* __HAVE_ARCH_GATE_AREA */
1242 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1243 void __user *, size_t *, loff_t *);
1244 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1245 unsigned long lru_pages);
1248 #define randomize_va_space 0
1250 extern int randomize_va_space;
1253 const char * arch_vma_name(struct vm_area_struct *vma);
1254 void print_vma_addr(char *prefix, unsigned long rip);
1256 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1257 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1258 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1259 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1260 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1261 void *vmemmap_alloc_block(unsigned long size, int node);
1262 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1263 int vmemmap_populate_basepages(struct page *start_page,
1264 unsigned long pages, int node);
1265 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1266 void vmemmap_populate_print_last(void);
1268 #endif /* __KERNEL__ */
1269 #endif /* _LINUX_MM_H */