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>
15 #include <linux/range.h>
16 #include <linux/pfn.h>
17 #include <linux/bit_spinlock.h>
23 struct writeback_control;
25 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
26 extern unsigned long max_mapnr;
29 extern unsigned long num_physpages;
30 extern unsigned long totalram_pages;
31 extern void * high_memory;
32 extern int page_cluster;
35 extern int sysctl_legacy_va_layout;
37 #define sysctl_legacy_va_layout 0
41 #include <asm/pgtable.h>
42 #include <asm/processor.h>
44 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
46 /* to align the pointer to the (next) page boundary */
47 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
50 * Linux kernel virtual memory manager primitives.
51 * The idea being to have a "virtual" mm in the same way
52 * we have a virtual fs - giving a cleaner interface to the
53 * mm details, and allowing different kinds of memory mappings
54 * (from shared memory to executable loading to arbitrary
58 extern struct kmem_cache *vm_area_cachep;
61 extern struct rb_root nommu_region_tree;
62 extern struct rw_semaphore nommu_region_sem;
64 extern unsigned int kobjsize(const void *objp);
68 * vm_flags in vm_area_struct, see mm_types.h.
70 #define VM_READ 0x00000001 /* currently active flags */
71 #define VM_WRITE 0x00000002
72 #define VM_EXEC 0x00000004
73 #define VM_SHARED 0x00000008
75 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
76 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
77 #define VM_MAYWRITE 0x00000020
78 #define VM_MAYEXEC 0x00000040
79 #define VM_MAYSHARE 0x00000080
81 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
82 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
83 #define VM_GROWSUP 0x00000200
85 #define VM_GROWSUP 0x00000000
87 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
88 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90 #define VM_EXECUTABLE 0x00001000
91 #define VM_LOCKED 0x00002000
92 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94 /* Used by sys_madvise() */
95 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
96 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
99 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
100 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
101 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
102 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
103 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
104 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
105 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
106 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
108 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
110 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
111 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
113 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
114 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
115 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
116 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
117 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
119 /* Bits set in the VMA until the stack is in its final location */
120 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
122 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
123 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
126 #ifdef CONFIG_STACK_GROWSUP
127 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
129 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
132 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
133 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
134 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
135 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
136 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
139 * special vmas that are non-mergable, non-mlock()able
141 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
144 * mapping from the currently active vm_flags protection bits (the
145 * low four bits) to a page protection mask..
147 extern pgprot_t protection_map[16];
149 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
150 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
151 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
152 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
155 * This interface is used by x86 PAT code to identify a pfn mapping that is
156 * linear over entire vma. This is to optimize PAT code that deals with
157 * marking the physical region with a particular prot. This is not for generic
158 * mm use. Note also that this check will not work if the pfn mapping is
159 * linear for a vma starting at physical address 0. In which case PAT code
160 * falls back to slow path of reserving physical range page by page.
162 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
164 return (vma->vm_flags & VM_PFN_AT_MMAP);
167 static inline int is_pfn_mapping(struct vm_area_struct *vma)
169 return (vma->vm_flags & VM_PFNMAP);
173 * vm_fault is filled by the the pagefault handler and passed to the vma's
174 * ->fault function. The vma's ->fault is responsible for returning a bitmask
175 * of VM_FAULT_xxx flags that give details about how the fault was handled.
177 * pgoff should be used in favour of virtual_address, if possible. If pgoff
178 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
182 unsigned int flags; /* FAULT_FLAG_xxx flags */
183 pgoff_t pgoff; /* Logical page offset based on vma */
184 void __user *virtual_address; /* Faulting virtual address */
186 struct page *page; /* ->fault handlers should return a
187 * page here, unless VM_FAULT_NOPAGE
188 * is set (which is also implied by
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
198 struct vm_operations_struct {
199 void (*open)(struct vm_area_struct * area);
200 void (*close)(struct vm_area_struct * area);
201 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
203 /* notification that a previously read-only page is about to become
204 * writable, if an error is returned it will cause a SIGBUS */
205 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
207 /* called by access_process_vm when get_user_pages() fails, typically
208 * for use by special VMAs that can switch between memory and hardware
210 int (*access)(struct vm_area_struct *vma, unsigned long addr,
211 void *buf, int len, int write);
214 * set_policy() op must add a reference to any non-NULL @new mempolicy
215 * to hold the policy upon return. Caller should pass NULL @new to
216 * remove a policy and fall back to surrounding context--i.e. do not
217 * install a MPOL_DEFAULT policy, nor the task or system default
220 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
223 * get_policy() op must add reference [mpol_get()] to any policy at
224 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
225 * in mm/mempolicy.c will do this automatically.
226 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
227 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
228 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
229 * must return NULL--i.e., do not "fallback" to task or system default
232 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
234 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
235 const nodemask_t *to, unsigned long flags);
242 #define page_private(page) ((page)->private)
243 #define set_page_private(page, v) ((page)->private = (v))
246 * FIXME: take this include out, include page-flags.h in
247 * files which need it (119 of them)
249 #include <linux/page-flags.h>
250 #include <linux/huge_mm.h>
253 * Methods to modify the page usage count.
255 * What counts for a page usage:
256 * - cache mapping (page->mapping)
257 * - private data (page->private)
258 * - page mapped in a task's page tables, each mapping
259 * is counted separately
261 * Also, many kernel routines increase the page count before a critical
262 * routine so they can be sure the page doesn't go away from under them.
266 * Drop a ref, return true if the refcount fell to zero (the page has no users)
268 static inline int put_page_testzero(struct page *page)
270 VM_BUG_ON(atomic_read(&page->_count) == 0);
271 return atomic_dec_and_test(&page->_count);
275 * Try to grab a ref unless the page has a refcount of zero, return false if
278 static inline int get_page_unless_zero(struct page *page)
280 return atomic_inc_not_zero(&page->_count);
283 extern int page_is_ram(unsigned long pfn);
285 /* Support for virtually mapped pages */
286 struct page *vmalloc_to_page(const void *addr);
287 unsigned long vmalloc_to_pfn(const void *addr);
290 * Determine if an address is within the vmalloc range
292 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
293 * is no special casing required.
295 static inline int is_vmalloc_addr(const void *x)
298 unsigned long addr = (unsigned long)x;
300 return addr >= VMALLOC_START && addr < VMALLOC_END;
306 extern int is_vmalloc_or_module_addr(const void *x);
308 static inline int is_vmalloc_or_module_addr(const void *x)
314 static inline void compound_lock(struct page *page)
316 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
317 bit_spin_lock(PG_compound_lock, &page->flags);
321 static inline void compound_unlock(struct page *page)
323 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
324 bit_spin_unlock(PG_compound_lock, &page->flags);
328 static inline unsigned long compound_lock_irqsave(struct page *page)
330 unsigned long uninitialized_var(flags);
331 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
332 local_irq_save(flags);
338 static inline void compound_unlock_irqrestore(struct page *page,
341 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
342 compound_unlock(page);
343 local_irq_restore(flags);
347 static inline struct page *compound_head(struct page *page)
349 if (unlikely(PageTail(page)))
350 return page->first_page;
354 static inline int page_count(struct page *page)
356 return atomic_read(&compound_head(page)->_count);
359 static inline void get_page(struct page *page)
362 * Getting a normal page or the head of a compound page
363 * requires to already have an elevated page->_count. Only if
364 * we're getting a tail page, the elevated page->_count is
365 * required only in the head page, so for tail pages the
366 * bugcheck only verifies that the page->_count isn't
369 VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
370 atomic_inc(&page->_count);
372 * Getting a tail page will elevate both the head and tail
375 if (unlikely(PageTail(page))) {
377 * This is safe only because
378 * __split_huge_page_refcount can't run under
381 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
382 atomic_inc(&page->first_page->_count);
386 static inline struct page *virt_to_head_page(const void *x)
388 struct page *page = virt_to_page(x);
389 return compound_head(page);
393 * Setup the page count before being freed into the page allocator for
394 * the first time (boot or memory hotplug)
396 static inline void init_page_count(struct page *page)
398 atomic_set(&page->_count, 1);
402 * PageBuddy() indicate that the page is free and in the buddy system
403 * (see mm/page_alloc.c).
405 static inline int PageBuddy(struct page *page)
407 return atomic_read(&page->_mapcount) == -2;
410 static inline void __SetPageBuddy(struct page *page)
412 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
413 atomic_set(&page->_mapcount, -2);
416 static inline void __ClearPageBuddy(struct page *page)
418 VM_BUG_ON(!PageBuddy(page));
419 atomic_set(&page->_mapcount, -1);
422 void put_page(struct page *page);
423 void put_pages_list(struct list_head *pages);
425 void split_page(struct page *page, unsigned int order);
426 int split_free_page(struct page *page);
429 * Compound pages have a destructor function. Provide a
430 * prototype for that function and accessor functions.
431 * These are _only_ valid on the head of a PG_compound page.
433 typedef void compound_page_dtor(struct page *);
435 static inline void set_compound_page_dtor(struct page *page,
436 compound_page_dtor *dtor)
438 page[1].lru.next = (void *)dtor;
441 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
443 return (compound_page_dtor *)page[1].lru.next;
446 static inline int compound_order(struct page *page)
450 return (unsigned long)page[1].lru.prev;
453 static inline int compound_trans_order(struct page *page)
461 flags = compound_lock_irqsave(page);
462 order = compound_order(page);
463 compound_unlock_irqrestore(page, flags);
467 static inline void set_compound_order(struct page *page, unsigned long order)
469 page[1].lru.prev = (void *)order;
473 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
474 * servicing faults for write access. In the normal case, do always want
475 * pte_mkwrite. But get_user_pages can cause write faults for mappings
476 * that do not have writing enabled, when used by access_process_vm.
478 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
480 if (likely(vma->vm_flags & VM_WRITE))
481 pte = pte_mkwrite(pte);
486 * Multiple processes may "see" the same page. E.g. for untouched
487 * mappings of /dev/null, all processes see the same page full of
488 * zeroes, and text pages of executables and shared libraries have
489 * only one copy in memory, at most, normally.
491 * For the non-reserved pages, page_count(page) denotes a reference count.
492 * page_count() == 0 means the page is free. page->lru is then used for
493 * freelist management in the buddy allocator.
494 * page_count() > 0 means the page has been allocated.
496 * Pages are allocated by the slab allocator in order to provide memory
497 * to kmalloc and kmem_cache_alloc. In this case, the management of the
498 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
499 * unless a particular usage is carefully commented. (the responsibility of
500 * freeing the kmalloc memory is the caller's, of course).
502 * A page may be used by anyone else who does a __get_free_page().
503 * In this case, page_count still tracks the references, and should only
504 * be used through the normal accessor functions. The top bits of page->flags
505 * and page->virtual store page management information, but all other fields
506 * are unused and could be used privately, carefully. The management of this
507 * page is the responsibility of the one who allocated it, and those who have
508 * subsequently been given references to it.
510 * The other pages (we may call them "pagecache pages") are completely
511 * managed by the Linux memory manager: I/O, buffers, swapping etc.
512 * The following discussion applies only to them.
514 * A pagecache page contains an opaque `private' member, which belongs to the
515 * page's address_space. Usually, this is the address of a circular list of
516 * the page's disk buffers. PG_private must be set to tell the VM to call
517 * into the filesystem to release these pages.
519 * A page may belong to an inode's memory mapping. In this case, page->mapping
520 * is the pointer to the inode, and page->index is the file offset of the page,
521 * in units of PAGE_CACHE_SIZE.
523 * If pagecache pages are not associated with an inode, they are said to be
524 * anonymous pages. These may become associated with the swapcache, and in that
525 * case PG_swapcache is set, and page->private is an offset into the swapcache.
527 * In either case (swapcache or inode backed), the pagecache itself holds one
528 * reference to the page. Setting PG_private should also increment the
529 * refcount. The each user mapping also has a reference to the page.
531 * The pagecache pages are stored in a per-mapping radix tree, which is
532 * rooted at mapping->page_tree, and indexed by offset.
533 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
534 * lists, we instead now tag pages as dirty/writeback in the radix tree.
536 * All pagecache pages may be subject to I/O:
537 * - inode pages may need to be read from disk,
538 * - inode pages which have been modified and are MAP_SHARED may need
539 * to be written back to the inode on disk,
540 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
541 * modified may need to be swapped out to swap space and (later) to be read
546 * The zone field is never updated after free_area_init_core()
547 * sets it, so none of the operations on it need to be atomic.
552 * page->flags layout:
554 * There are three possibilities for how page->flags get
555 * laid out. The first is for the normal case, without
556 * sparsemem. The second is for sparsemem when there is
557 * plenty of space for node and section. The last is when
558 * we have run out of space and have to fall back to an
559 * alternate (slower) way of determining the node.
561 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
562 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
563 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
565 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
566 #define SECTIONS_WIDTH SECTIONS_SHIFT
568 #define SECTIONS_WIDTH 0
571 #define ZONES_WIDTH ZONES_SHIFT
573 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
574 #define NODES_WIDTH NODES_SHIFT
576 #ifdef CONFIG_SPARSEMEM_VMEMMAP
577 #error "Vmemmap: No space for nodes field in page flags"
579 #define NODES_WIDTH 0
582 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
583 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
584 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
585 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
588 * We are going to use the flags for the page to node mapping if its in
589 * there. This includes the case where there is no node, so it is implicit.
591 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
592 #define NODE_NOT_IN_PAGE_FLAGS
595 #ifndef PFN_SECTION_SHIFT
596 #define PFN_SECTION_SHIFT 0
600 * Define the bit shifts to access each section. For non-existant
601 * sections we define the shift as 0; that plus a 0 mask ensures
602 * the compiler will optimise away reference to them.
604 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
605 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
606 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
608 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
609 #ifdef NODE_NOT_IN_PAGE_FLAGS
610 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
611 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
612 SECTIONS_PGOFF : ZONES_PGOFF)
614 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
615 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
616 NODES_PGOFF : ZONES_PGOFF)
619 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
621 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
622 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
625 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
626 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
627 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
628 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
630 static inline enum zone_type page_zonenum(struct page *page)
632 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
636 * The identification function is only used by the buddy allocator for
637 * determining if two pages could be buddies. We are not really
638 * identifying a zone since we could be using a the section number
639 * id if we have not node id available in page flags.
640 * We guarantee only that it will return the same value for two
641 * combinable pages in a zone.
643 static inline int page_zone_id(struct page *page)
645 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
648 static inline int zone_to_nid(struct zone *zone)
657 #ifdef NODE_NOT_IN_PAGE_FLAGS
658 extern int page_to_nid(struct page *page);
660 static inline int page_to_nid(struct page *page)
662 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
666 static inline struct zone *page_zone(struct page *page)
668 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
671 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
672 static inline unsigned long page_to_section(struct page *page)
674 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
678 static inline void set_page_zone(struct page *page, enum zone_type zone)
680 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
681 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
684 static inline void set_page_node(struct page *page, unsigned long node)
686 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
687 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
690 static inline void set_page_section(struct page *page, unsigned long section)
692 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
693 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
696 static inline void set_page_links(struct page *page, enum zone_type zone,
697 unsigned long node, unsigned long pfn)
699 set_page_zone(page, zone);
700 set_page_node(page, node);
701 set_page_section(page, pfn_to_section_nr(pfn));
705 * Some inline functions in vmstat.h depend on page_zone()
707 #include <linux/vmstat.h>
709 static __always_inline void *lowmem_page_address(struct page *page)
711 return __va(PFN_PHYS(page_to_pfn(page)));
714 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
715 #define HASHED_PAGE_VIRTUAL
718 #if defined(WANT_PAGE_VIRTUAL)
719 #define page_address(page) ((page)->virtual)
720 #define set_page_address(page, address) \
722 (page)->virtual = (address); \
724 #define page_address_init() do { } while(0)
727 #if defined(HASHED_PAGE_VIRTUAL)
728 void *page_address(struct page *page);
729 void set_page_address(struct page *page, void *virtual);
730 void page_address_init(void);
733 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
734 #define page_address(page) lowmem_page_address(page)
735 #define set_page_address(page, address) do { } while(0)
736 #define page_address_init() do { } while(0)
740 * On an anonymous page mapped into a user virtual memory area,
741 * page->mapping points to its anon_vma, not to a struct address_space;
742 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
744 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
745 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
746 * and then page->mapping points, not to an anon_vma, but to a private
747 * structure which KSM associates with that merged page. See ksm.h.
749 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
751 * Please note that, confusingly, "page_mapping" refers to the inode
752 * address_space which maps the page from disk; whereas "page_mapped"
753 * refers to user virtual address space into which the page is mapped.
755 #define PAGE_MAPPING_ANON 1
756 #define PAGE_MAPPING_KSM 2
757 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
759 extern struct address_space swapper_space;
760 static inline struct address_space *page_mapping(struct page *page)
762 struct address_space *mapping = page->mapping;
764 VM_BUG_ON(PageSlab(page));
765 if (unlikely(PageSwapCache(page)))
766 mapping = &swapper_space;
767 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
772 /* Neutral page->mapping pointer to address_space or anon_vma or other */
773 static inline void *page_rmapping(struct page *page)
775 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
778 static inline int PageAnon(struct page *page)
780 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
784 * Return the pagecache index of the passed page. Regular pagecache pages
785 * use ->index whereas swapcache pages use ->private
787 static inline pgoff_t page_index(struct page *page)
789 if (unlikely(PageSwapCache(page)))
790 return page_private(page);
795 * The atomic page->_mapcount, like _count, starts from -1:
796 * so that transitions both from it and to it can be tracked,
797 * using atomic_inc_and_test and atomic_add_negative(-1).
799 static inline void reset_page_mapcount(struct page *page)
801 atomic_set(&(page)->_mapcount, -1);
804 static inline int page_mapcount(struct page *page)
806 return atomic_read(&(page)->_mapcount) + 1;
810 * Return true if this page is mapped into pagetables.
812 static inline int page_mapped(struct page *page)
814 return atomic_read(&(page)->_mapcount) >= 0;
818 * Different kinds of faults, as returned by handle_mm_fault().
819 * Used to decide whether a process gets delivered SIGBUS or
820 * just gets major/minor fault counters bumped up.
823 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
825 #define VM_FAULT_OOM 0x0001
826 #define VM_FAULT_SIGBUS 0x0002
827 #define VM_FAULT_MAJOR 0x0004
828 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
829 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
830 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
832 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
833 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
834 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
836 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
838 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
839 VM_FAULT_HWPOISON_LARGE)
841 /* Encode hstate index for a hwpoisoned large page */
842 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
843 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
846 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
848 extern void pagefault_out_of_memory(void);
850 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
852 extern void show_free_areas(void);
854 int shmem_lock(struct file *file, int lock, struct user_struct *user);
855 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
856 int shmem_zero_setup(struct vm_area_struct *);
859 extern unsigned long shmem_get_unmapped_area(struct file *file,
863 unsigned long flags);
866 extern int can_do_mlock(void);
867 extern int user_shm_lock(size_t, struct user_struct *);
868 extern void user_shm_unlock(size_t, struct user_struct *);
871 * Parameter block passed down to zap_pte_range in exceptional cases.
874 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
875 struct address_space *check_mapping; /* Check page->mapping if set */
876 pgoff_t first_index; /* Lowest page->index to unmap */
877 pgoff_t last_index; /* Highest page->index to unmap */
878 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
879 unsigned long truncate_count; /* Compare vm_truncate_count */
882 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
885 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
887 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
888 unsigned long size, struct zap_details *);
889 unsigned long unmap_vmas(struct mmu_gather **tlb,
890 struct vm_area_struct *start_vma, unsigned long start_addr,
891 unsigned long end_addr, unsigned long *nr_accounted,
892 struct zap_details *);
895 * mm_walk - callbacks for walk_page_range
896 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
897 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
898 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
899 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
900 * @pte_hole: if set, called for each hole at all levels
901 * @hugetlb_entry: if set, called for each hugetlb entry
903 * (see walk_page_range for more details)
906 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
907 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
908 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
909 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
910 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
911 int (*hugetlb_entry)(pte_t *, unsigned long,
912 unsigned long, unsigned long, struct mm_walk *);
913 struct mm_struct *mm;
917 int walk_page_range(unsigned long addr, unsigned long end,
918 struct mm_walk *walk);
919 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
920 unsigned long end, unsigned long floor, unsigned long ceiling);
921 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
922 struct vm_area_struct *vma);
923 void unmap_mapping_range(struct address_space *mapping,
924 loff_t const holebegin, loff_t const holelen, int even_cows);
925 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
927 int follow_phys(struct vm_area_struct *vma, unsigned long address,
928 unsigned int flags, unsigned long *prot, resource_size_t *phys);
929 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
930 void *buf, int len, int write);
932 static inline void unmap_shared_mapping_range(struct address_space *mapping,
933 loff_t const holebegin, loff_t const holelen)
935 unmap_mapping_range(mapping, holebegin, holelen, 0);
938 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
939 extern void truncate_setsize(struct inode *inode, loff_t newsize);
940 extern int vmtruncate(struct inode *inode, loff_t offset);
941 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
943 int truncate_inode_page(struct address_space *mapping, struct page *page);
944 int generic_error_remove_page(struct address_space *mapping, struct page *page);
946 int invalidate_inode_page(struct page *page);
949 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
950 unsigned long address, unsigned int flags);
952 static inline int handle_mm_fault(struct mm_struct *mm,
953 struct vm_area_struct *vma, unsigned long address,
956 /* should never happen if there's no MMU */
958 return VM_FAULT_SIGBUS;
962 extern int make_pages_present(unsigned long addr, unsigned long end);
963 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
965 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
966 unsigned long start, int nr_pages, int write, int force,
967 struct page **pages, struct vm_area_struct **vmas);
968 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
969 struct page **pages);
970 struct page *get_dump_page(unsigned long addr);
972 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
973 extern void do_invalidatepage(struct page *page, unsigned long offset);
975 int __set_page_dirty_nobuffers(struct page *page);
976 int __set_page_dirty_no_writeback(struct page *page);
977 int redirty_page_for_writepage(struct writeback_control *wbc,
979 void account_page_dirtied(struct page *page, struct address_space *mapping);
980 void account_page_writeback(struct page *page);
981 int set_page_dirty(struct page *page);
982 int set_page_dirty_lock(struct page *page);
983 int clear_page_dirty_for_io(struct page *page);
985 /* Is the vma a continuation of the stack vma above it? */
986 static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr)
988 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
991 extern unsigned long move_page_tables(struct vm_area_struct *vma,
992 unsigned long old_addr, struct vm_area_struct *new_vma,
993 unsigned long new_addr, unsigned long len);
994 extern unsigned long do_mremap(unsigned long addr,
995 unsigned long old_len, unsigned long new_len,
996 unsigned long flags, unsigned long new_addr);
997 extern int mprotect_fixup(struct vm_area_struct *vma,
998 struct vm_area_struct **pprev, unsigned long start,
999 unsigned long end, unsigned long newflags);
1002 * doesn't attempt to fault and will return short.
1004 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1005 struct page **pages);
1007 * per-process(per-mm_struct) statistics.
1009 #if defined(SPLIT_RSS_COUNTING)
1011 * The mm counters are not protected by its page_table_lock,
1012 * so must be incremented atomically.
1014 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1016 atomic_long_set(&mm->rss_stat.count[member], value);
1019 unsigned long get_mm_counter(struct mm_struct *mm, int member);
1021 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1023 atomic_long_add(value, &mm->rss_stat.count[member]);
1026 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1028 atomic_long_inc(&mm->rss_stat.count[member]);
1031 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1033 atomic_long_dec(&mm->rss_stat.count[member]);
1036 #else /* !USE_SPLIT_PTLOCKS */
1038 * The mm counters are protected by its page_table_lock,
1039 * so can be incremented directly.
1041 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1043 mm->rss_stat.count[member] = value;
1046 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1048 return mm->rss_stat.count[member];
1051 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1053 mm->rss_stat.count[member] += value;
1056 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1058 mm->rss_stat.count[member]++;
1061 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1063 mm->rss_stat.count[member]--;
1066 #endif /* !USE_SPLIT_PTLOCKS */
1068 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1070 return get_mm_counter(mm, MM_FILEPAGES) +
1071 get_mm_counter(mm, MM_ANONPAGES);
1074 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1076 return max(mm->hiwater_rss, get_mm_rss(mm));
1079 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1081 return max(mm->hiwater_vm, mm->total_vm);
1084 static inline void update_hiwater_rss(struct mm_struct *mm)
1086 unsigned long _rss = get_mm_rss(mm);
1088 if ((mm)->hiwater_rss < _rss)
1089 (mm)->hiwater_rss = _rss;
1092 static inline void update_hiwater_vm(struct mm_struct *mm)
1094 if (mm->hiwater_vm < mm->total_vm)
1095 mm->hiwater_vm = mm->total_vm;
1098 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1099 struct mm_struct *mm)
1101 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1103 if (*maxrss < hiwater_rss)
1104 *maxrss = hiwater_rss;
1107 #if defined(SPLIT_RSS_COUNTING)
1108 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
1110 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1116 * A callback you can register to apply pressure to ageable caches.
1118 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
1119 * look through the least-recently-used 'nr_to_scan' entries and
1120 * attempt to free them up. It should return the number of objects
1121 * which remain in the cache. If it returns -1, it means it cannot do
1122 * any scanning at this time (eg. there is a risk of deadlock).
1124 * The 'gfpmask' refers to the allocation we are currently trying to
1127 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1128 * querying the cache size, so a fastpath for that case is appropriate.
1131 int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask);
1132 int seeks; /* seeks to recreate an obj */
1134 /* These are for internal use */
1135 struct list_head list;
1136 long nr; /* objs pending delete */
1138 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1139 extern void register_shrinker(struct shrinker *);
1140 extern void unregister_shrinker(struct shrinker *);
1142 int vma_wants_writenotify(struct vm_area_struct *vma);
1144 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1146 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1150 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1154 #ifdef __PAGETABLE_PUD_FOLDED
1155 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1156 unsigned long address)
1161 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1164 #ifdef __PAGETABLE_PMD_FOLDED
1165 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1166 unsigned long address)
1171 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1174 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1175 pmd_t *pmd, unsigned long address);
1176 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1179 * The following ifdef needed to get the 4level-fixup.h header to work.
1180 * Remove it when 4level-fixup.h has been removed.
1182 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1183 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1185 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1186 NULL: pud_offset(pgd, address);
1189 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1191 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1192 NULL: pmd_offset(pud, address);
1194 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1196 #if USE_SPLIT_PTLOCKS
1198 * We tuck a spinlock to guard each pagetable page into its struct page,
1199 * at page->private, with BUILD_BUG_ON to make sure that this will not
1200 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1201 * When freeing, reset page->mapping so free_pages_check won't complain.
1203 #define __pte_lockptr(page) &((page)->ptl)
1204 #define pte_lock_init(_page) do { \
1205 spin_lock_init(__pte_lockptr(_page)); \
1207 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1208 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1209 #else /* !USE_SPLIT_PTLOCKS */
1211 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1213 #define pte_lock_init(page) do {} while (0)
1214 #define pte_lock_deinit(page) do {} while (0)
1215 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1216 #endif /* USE_SPLIT_PTLOCKS */
1218 static inline void pgtable_page_ctor(struct page *page)
1220 pte_lock_init(page);
1221 inc_zone_page_state(page, NR_PAGETABLE);
1224 static inline void pgtable_page_dtor(struct page *page)
1226 pte_lock_deinit(page);
1227 dec_zone_page_state(page, NR_PAGETABLE);
1230 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1232 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1233 pte_t *__pte = pte_offset_map(pmd, address); \
1239 #define pte_unmap_unlock(pte, ptl) do { \
1244 #define pte_alloc_map(mm, vma, pmd, address) \
1245 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1247 NULL: pte_offset_map(pmd, address))
1249 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1250 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1252 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1254 #define pte_alloc_kernel(pmd, address) \
1255 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1256 NULL: pte_offset_kernel(pmd, address))
1258 extern void free_area_init(unsigned long * zones_size);
1259 extern void free_area_init_node(int nid, unsigned long * zones_size,
1260 unsigned long zone_start_pfn, unsigned long *zholes_size);
1261 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1263 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1264 * zones, allocate the backing mem_map and account for memory holes in a more
1265 * architecture independent manner. This is a substitute for creating the
1266 * zone_sizes[] and zholes_size[] arrays and passing them to
1267 * free_area_init_node()
1269 * An architecture is expected to register range of page frames backed by
1270 * physical memory with add_active_range() before calling
1271 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1272 * usage, an architecture is expected to do something like
1274 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1276 * for_each_valid_physical_page_range()
1277 * add_active_range(node_id, start_pfn, end_pfn)
1278 * free_area_init_nodes(max_zone_pfns);
1280 * If the architecture guarantees that there are no holes in the ranges
1281 * registered with add_active_range(), free_bootmem_active_regions()
1282 * will call free_bootmem_node() for each registered physical page range.
1283 * Similarly sparse_memory_present_with_active_regions() calls
1284 * memory_present() for each range when SPARSEMEM is enabled.
1286 * See mm/page_alloc.c for more information on each function exposed by
1287 * CONFIG_ARCH_POPULATES_NODE_MAP
1289 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1290 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1291 unsigned long end_pfn);
1292 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1293 unsigned long end_pfn);
1294 extern void remove_all_active_ranges(void);
1295 void sort_node_map(void);
1296 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1297 unsigned long end_pfn);
1298 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1299 unsigned long end_pfn);
1300 extern void get_pfn_range_for_nid(unsigned int nid,
1301 unsigned long *start_pfn, unsigned long *end_pfn);
1302 extern unsigned long find_min_pfn_with_active_regions(void);
1303 extern void free_bootmem_with_active_regions(int nid,
1304 unsigned long max_low_pfn);
1305 int add_from_early_node_map(struct range *range, int az,
1306 int nr_range, int nid);
1307 u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1308 u64 goal, u64 limit);
1309 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
1310 u64 goal, u64 limit);
1311 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1312 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1313 extern void sparse_memory_present_with_active_regions(int nid);
1314 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1316 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1317 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1318 static inline int __early_pfn_to_nid(unsigned long pfn)
1323 /* please see mm/page_alloc.c */
1324 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1325 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1326 /* there is a per-arch backend function. */
1327 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1328 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1331 extern void set_dma_reserve(unsigned long new_dma_reserve);
1332 extern void memmap_init_zone(unsigned long, int, unsigned long,
1333 unsigned long, enum memmap_context);
1334 extern void setup_per_zone_wmarks(void);
1335 extern void calculate_zone_inactive_ratio(struct zone *zone);
1336 extern void mem_init(void);
1337 extern void __init mmap_init(void);
1338 extern void show_mem(void);
1339 extern void si_meminfo(struct sysinfo * val);
1340 extern void si_meminfo_node(struct sysinfo *val, int nid);
1341 extern int after_bootmem;
1343 extern void setup_per_cpu_pageset(void);
1345 extern void zone_pcp_update(struct zone *zone);
1348 extern atomic_long_t mmap_pages_allocated;
1349 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1352 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1353 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1354 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1355 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1356 struct prio_tree_iter *iter);
1358 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1359 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1360 (vma = vma_prio_tree_next(vma, iter)); )
1362 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1363 struct list_head *list)
1365 vma->shared.vm_set.parent = NULL;
1366 list_add_tail(&vma->shared.vm_set.list, list);
1370 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1371 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1372 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1373 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1374 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1375 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1376 struct mempolicy *);
1377 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1378 extern int split_vma(struct mm_struct *,
1379 struct vm_area_struct *, unsigned long addr, int new_below);
1380 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1381 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1382 struct rb_node **, struct rb_node *);
1383 extern void unlink_file_vma(struct vm_area_struct *);
1384 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1385 unsigned long addr, unsigned long len, pgoff_t pgoff);
1386 extern void exit_mmap(struct mm_struct *);
1388 extern int mm_take_all_locks(struct mm_struct *mm);
1389 extern void mm_drop_all_locks(struct mm_struct *mm);
1391 #ifdef CONFIG_PROC_FS
1392 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1393 extern void added_exe_file_vma(struct mm_struct *mm);
1394 extern void removed_exe_file_vma(struct mm_struct *mm);
1396 static inline void added_exe_file_vma(struct mm_struct *mm)
1399 static inline void removed_exe_file_vma(struct mm_struct *mm)
1401 #endif /* CONFIG_PROC_FS */
1403 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1404 extern int install_special_mapping(struct mm_struct *mm,
1405 unsigned long addr, unsigned long len,
1406 unsigned long flags, struct page **pages);
1408 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1410 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1411 unsigned long len, unsigned long prot,
1412 unsigned long flag, unsigned long pgoff);
1413 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1414 unsigned long len, unsigned long flags,
1415 unsigned int vm_flags, unsigned long pgoff);
1417 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1418 unsigned long len, unsigned long prot,
1419 unsigned long flag, unsigned long offset)
1421 unsigned long ret = -EINVAL;
1422 if ((offset + PAGE_ALIGN(len)) < offset)
1424 if (!(offset & ~PAGE_MASK))
1425 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1430 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1432 extern unsigned long do_brk(unsigned long, unsigned long);
1435 extern unsigned long page_unuse(struct page *);
1436 extern void truncate_inode_pages(struct address_space *, loff_t);
1437 extern void truncate_inode_pages_range(struct address_space *,
1438 loff_t lstart, loff_t lend);
1440 /* generic vm_area_ops exported for stackable file systems */
1441 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1443 /* mm/page-writeback.c */
1444 int write_one_page(struct page *page, int wait);
1445 void task_dirty_inc(struct task_struct *tsk);
1448 #define VM_MAX_READAHEAD 128 /* kbytes */
1449 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1451 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1452 pgoff_t offset, unsigned long nr_to_read);
1454 void page_cache_sync_readahead(struct address_space *mapping,
1455 struct file_ra_state *ra,
1458 unsigned long size);
1460 void page_cache_async_readahead(struct address_space *mapping,
1461 struct file_ra_state *ra,
1465 unsigned long size);
1467 unsigned long max_sane_readahead(unsigned long nr);
1468 unsigned long ra_submit(struct file_ra_state *ra,
1469 struct address_space *mapping,
1472 /* Do stack extension */
1473 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1475 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1477 #define expand_upwards(vma, address) do { } while (0)
1479 extern int expand_stack_downwards(struct vm_area_struct *vma,
1480 unsigned long address);
1482 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1483 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1484 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1485 struct vm_area_struct **pprev);
1487 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1488 NULL if none. Assume start_addr < end_addr. */
1489 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1491 struct vm_area_struct * vma = find_vma(mm,start_addr);
1493 if (vma && end_addr <= vma->vm_start)
1498 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1500 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1504 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1506 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1512 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1513 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1514 unsigned long pfn, unsigned long size, pgprot_t);
1515 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1516 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1518 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1521 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1522 unsigned int foll_flags);
1523 #define FOLL_WRITE 0x01 /* check pte is writable */
1524 #define FOLL_TOUCH 0x02 /* mark page accessed */
1525 #define FOLL_GET 0x04 /* do get_page on page */
1526 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1527 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1528 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1529 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1531 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1533 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1534 unsigned long size, pte_fn_t fn, void *data);
1536 #ifdef CONFIG_PROC_FS
1537 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1539 static inline void vm_stat_account(struct mm_struct *mm,
1540 unsigned long flags, struct file *file, long pages)
1543 #endif /* CONFIG_PROC_FS */
1545 #ifdef CONFIG_DEBUG_PAGEALLOC
1546 extern int debug_pagealloc_enabled;
1548 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1550 static inline void enable_debug_pagealloc(void)
1552 debug_pagealloc_enabled = 1;
1554 #ifdef CONFIG_HIBERNATION
1555 extern bool kernel_page_present(struct page *page);
1556 #endif /* CONFIG_HIBERNATION */
1559 kernel_map_pages(struct page *page, int numpages, int enable) {}
1560 static inline void enable_debug_pagealloc(void)
1563 #ifdef CONFIG_HIBERNATION
1564 static inline bool kernel_page_present(struct page *page) { return true; }
1565 #endif /* CONFIG_HIBERNATION */
1568 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1569 #ifdef __HAVE_ARCH_GATE_AREA
1570 int in_gate_area_no_task(unsigned long addr);
1571 int in_gate_area(struct task_struct *task, unsigned long addr);
1573 int in_gate_area_no_task(unsigned long addr);
1574 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1575 #endif /* __HAVE_ARCH_GATE_AREA */
1577 int drop_caches_sysctl_handler(struct ctl_table *, int,
1578 void __user *, size_t *, loff_t *);
1579 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1580 unsigned long lru_pages);
1583 #define randomize_va_space 0
1585 extern int randomize_va_space;
1588 const char * arch_vma_name(struct vm_area_struct *vma);
1589 void print_vma_addr(char *prefix, unsigned long rip);
1591 void sparse_mem_maps_populate_node(struct page **map_map,
1592 unsigned long pnum_begin,
1593 unsigned long pnum_end,
1594 unsigned long map_count,
1597 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1598 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1599 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1600 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1601 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1602 void *vmemmap_alloc_block(unsigned long size, int node);
1603 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1604 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1605 int vmemmap_populate_basepages(struct page *start_page,
1606 unsigned long pages, int node);
1607 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1608 void vmemmap_populate_print_last(void);
1612 MF_COUNT_INCREASED = 1 << 0,
1614 extern void memory_failure(unsigned long pfn, int trapno);
1615 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1616 extern int unpoison_memory(unsigned long pfn);
1617 extern int sysctl_memory_failure_early_kill;
1618 extern int sysctl_memory_failure_recovery;
1619 extern void shake_page(struct page *p, int access);
1620 extern atomic_long_t mce_bad_pages;
1621 extern int soft_offline_page(struct page *page, int flags);
1622 #ifdef CONFIG_MEMORY_FAILURE
1623 int is_hwpoison_address(unsigned long addr);
1625 static inline int is_hwpoison_address(unsigned long addr)
1631 extern void dump_page(struct page *page);
1633 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1634 extern void clear_huge_page(struct page *page,
1636 unsigned int pages_per_huge_page);
1637 extern void copy_user_huge_page(struct page *dst, struct page *src,
1638 unsigned long addr, struct vm_area_struct *vma,
1639 unsigned int pages_per_huge_page);
1640 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1642 #endif /* __KERNEL__ */
1643 #endif /* _LINUX_MM_H */