4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/bit_spinlock.h>
19 #include <linux/shrinker.h>
23 struct anon_vma_chain;
26 struct writeback_control;
28 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
32 extern unsigned long num_physpages;
33 extern unsigned long totalram_pages;
34 extern void * high_memory;
35 extern int page_cluster;
38 extern int sysctl_legacy_va_layout;
40 #define sysctl_legacy_va_layout 0
44 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 extern unsigned long sysctl_user_reserve_kbytes;
49 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
51 /* to align the pointer to the (next) page boundary */
52 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
55 * Linux kernel virtual memory manager primitives.
56 * The idea being to have a "virtual" mm in the same way
57 * we have a virtual fs - giving a cleaner interface to the
58 * mm details, and allowing different kinds of memory mappings
59 * (from shared memory to executable loading to arbitrary
63 extern struct kmem_cache *vm_area_cachep;
66 extern struct rb_root nommu_region_tree;
67 extern struct rw_semaphore nommu_region_sem;
69 extern unsigned int kobjsize(const void *objp);
73 * vm_flags in vm_area_struct, see mm_types.h.
75 #define VM_NONE 0x00000000
77 #define VM_READ 0x00000001 /* currently active flags */
78 #define VM_WRITE 0x00000002
79 #define VM_EXEC 0x00000004
80 #define VM_SHARED 0x00000008
82 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
83 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
84 #define VM_MAYWRITE 0x00000020
85 #define VM_MAYEXEC 0x00000040
86 #define VM_MAYSHARE 0x00000080
88 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
89 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
90 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
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_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 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
106 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
108 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
109 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
110 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
111 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
113 #if defined(CONFIG_X86)
114 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
115 #elif defined(CONFIG_PPC)
116 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
117 #elif defined(CONFIG_PARISC)
118 # define VM_GROWSUP VM_ARCH_1
119 #elif defined(CONFIG_METAG)
120 # define VM_GROWSUP VM_ARCH_1
121 #elif defined(CONFIG_IA64)
122 # define VM_GROWSUP VM_ARCH_1
123 #elif !defined(CONFIG_MMU)
124 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
128 # define VM_GROWSUP VM_NONE
131 /* Bits set in the VMA until the stack is in its final location */
132 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
134 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
135 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
138 #ifdef CONFIG_STACK_GROWSUP
139 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
141 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
144 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
145 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
146 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
147 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
148 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
151 * Special vmas that are non-mergable, non-mlock()able.
152 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
154 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
157 * mapping from the currently active vm_flags protection bits (the
158 * low four bits) to a page protection mask..
160 extern pgprot_t protection_map[16];
162 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
163 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
164 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
165 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
166 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
167 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
168 #define FAULT_FLAG_TRIED 0x40 /* second try */
171 * vm_fault is filled by the the pagefault handler and passed to the vma's
172 * ->fault function. The vma's ->fault is responsible for returning a bitmask
173 * of VM_FAULT_xxx flags that give details about how the fault was handled.
175 * pgoff should be used in favour of virtual_address, if possible. If pgoff
176 * is used, one may implement ->remap_pages to get nonlinear mapping support.
179 unsigned int flags; /* FAULT_FLAG_xxx flags */
180 pgoff_t pgoff; /* Logical page offset based on vma */
181 void __user *virtual_address; /* Faulting virtual address */
183 struct page *page; /* ->fault handlers should return a
184 * page here, unless VM_FAULT_NOPAGE
185 * is set (which is also implied by
191 * These are the virtual MM functions - opening of an area, closing and
192 * unmapping it (needed to keep files on disk up-to-date etc), pointer
193 * to the functions called when a no-page or a wp-page exception occurs.
195 struct vm_operations_struct {
196 void (*open)(struct vm_area_struct * area);
197 void (*close)(struct vm_area_struct * area);
198 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
200 /* notification that a previously read-only page is about to become
201 * writable, if an error is returned it will cause a SIGBUS */
202 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
204 /* called by access_process_vm when get_user_pages() fails, typically
205 * for use by special VMAs that can switch between memory and hardware
207 int (*access)(struct vm_area_struct *vma, unsigned long addr,
208 void *buf, int len, int write);
211 * set_policy() op must add a reference to any non-NULL @new mempolicy
212 * to hold the policy upon return. Caller should pass NULL @new to
213 * remove a policy and fall back to surrounding context--i.e. do not
214 * install a MPOL_DEFAULT policy, nor the task or system default
217 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
220 * get_policy() op must add reference [mpol_get()] to any policy at
221 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
222 * in mm/mempolicy.c will do this automatically.
223 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
224 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
225 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
226 * must return NULL--i.e., do not "fallback" to task or system default
229 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
231 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
232 const nodemask_t *to, unsigned long flags);
234 /* called by sys_remap_file_pages() to populate non-linear mapping */
235 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
236 unsigned long size, pgoff_t pgoff);
242 #define page_private(page) ((page)->private)
243 #define set_page_private(page, v) ((page)->private = (v))
245 /* It's valid only if the page is free path or free_list */
246 static inline void set_freepage_migratetype(struct page *page, int migratetype)
248 page->index = migratetype;
251 /* It's valid only if the page is free path or free_list */
252 static inline int get_freepage_migratetype(struct page *page)
258 * FIXME: take this include out, include page-flags.h in
259 * files which need it (119 of them)
261 #include <linux/page-flags.h>
262 #include <linux/huge_mm.h>
265 * Methods to modify the page usage count.
267 * What counts for a page usage:
268 * - cache mapping (page->mapping)
269 * - private data (page->private)
270 * - page mapped in a task's page tables, each mapping
271 * is counted separately
273 * Also, many kernel routines increase the page count before a critical
274 * routine so they can be sure the page doesn't go away from under them.
278 * Drop a ref, return true if the refcount fell to zero (the page has no users)
280 static inline int put_page_testzero(struct page *page)
282 VM_BUG_ON(atomic_read(&page->_count) == 0);
283 return atomic_dec_and_test(&page->_count);
287 * Try to grab a ref unless the page has a refcount of zero, return false if
290 static inline int get_page_unless_zero(struct page *page)
292 return atomic_inc_not_zero(&page->_count);
295 extern int page_is_ram(unsigned long pfn);
297 /* Support for virtually mapped pages */
298 struct page *vmalloc_to_page(const void *addr);
299 unsigned long vmalloc_to_pfn(const void *addr);
302 * Determine if an address is within the vmalloc range
304 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
305 * is no special casing required.
307 static inline int is_vmalloc_addr(const void *x)
310 unsigned long addr = (unsigned long)x;
312 return addr >= VMALLOC_START && addr < VMALLOC_END;
318 extern int is_vmalloc_or_module_addr(const void *x);
320 static inline int is_vmalloc_or_module_addr(const void *x)
326 static inline void compound_lock(struct page *page)
328 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
329 VM_BUG_ON(PageSlab(page));
330 bit_spin_lock(PG_compound_lock, &page->flags);
334 static inline void compound_unlock(struct page *page)
336 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
337 VM_BUG_ON(PageSlab(page));
338 bit_spin_unlock(PG_compound_lock, &page->flags);
342 static inline unsigned long compound_lock_irqsave(struct page *page)
344 unsigned long uninitialized_var(flags);
345 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
346 local_irq_save(flags);
352 static inline void compound_unlock_irqrestore(struct page *page,
355 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
356 compound_unlock(page);
357 local_irq_restore(flags);
361 static inline struct page *compound_head(struct page *page)
363 if (unlikely(PageTail(page)))
364 return page->first_page;
369 * The atomic page->_mapcount, starts from -1: so that transitions
370 * both from it and to it can be tracked, using atomic_inc_and_test
371 * and atomic_add_negative(-1).
373 static inline void page_mapcount_reset(struct page *page)
375 atomic_set(&(page)->_mapcount, -1);
378 static inline int page_mapcount(struct page *page)
380 return atomic_read(&(page)->_mapcount) + 1;
383 static inline int page_count(struct page *page)
385 return atomic_read(&compound_head(page)->_count);
388 static inline void get_huge_page_tail(struct page *page)
391 * __split_huge_page_refcount() cannot run
394 VM_BUG_ON(page_mapcount(page) < 0);
395 VM_BUG_ON(atomic_read(&page->_count) != 0);
396 atomic_inc(&page->_mapcount);
399 extern bool __get_page_tail(struct page *page);
401 static inline void get_page(struct page *page)
403 if (unlikely(PageTail(page)))
404 if (likely(__get_page_tail(page)))
407 * Getting a normal page or the head of a compound page
408 * requires to already have an elevated page->_count.
410 VM_BUG_ON(atomic_read(&page->_count) <= 0);
411 atomic_inc(&page->_count);
414 static inline struct page *virt_to_head_page(const void *x)
416 struct page *page = virt_to_page(x);
417 return compound_head(page);
421 * Setup the page count before being freed into the page allocator for
422 * the first time (boot or memory hotplug)
424 static inline void init_page_count(struct page *page)
426 atomic_set(&page->_count, 1);
430 * PageBuddy() indicate that the page is free and in the buddy system
431 * (see mm/page_alloc.c).
433 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
434 * -2 so that an underflow of the page_mapcount() won't be mistaken
435 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
436 * efficiently by most CPU architectures.
438 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
440 static inline int PageBuddy(struct page *page)
442 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
445 static inline void __SetPageBuddy(struct page *page)
447 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
448 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
451 static inline void __ClearPageBuddy(struct page *page)
453 VM_BUG_ON(!PageBuddy(page));
454 atomic_set(&page->_mapcount, -1);
457 void put_page(struct page *page);
458 void put_pages_list(struct list_head *pages);
460 void split_page(struct page *page, unsigned int order);
461 int split_free_page(struct page *page);
464 * Compound pages have a destructor function. Provide a
465 * prototype for that function and accessor functions.
466 * These are _only_ valid on the head of a PG_compound page.
468 typedef void compound_page_dtor(struct page *);
470 static inline void set_compound_page_dtor(struct page *page,
471 compound_page_dtor *dtor)
473 page[1].lru.next = (void *)dtor;
476 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
478 return (compound_page_dtor *)page[1].lru.next;
481 static inline int compound_order(struct page *page)
485 return (unsigned long)page[1].lru.prev;
488 static inline int compound_trans_order(struct page *page)
496 flags = compound_lock_irqsave(page);
497 order = compound_order(page);
498 compound_unlock_irqrestore(page, flags);
502 static inline void set_compound_order(struct page *page, unsigned long order)
504 page[1].lru.prev = (void *)order;
509 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
510 * servicing faults for write access. In the normal case, do always want
511 * pte_mkwrite. But get_user_pages can cause write faults for mappings
512 * that do not have writing enabled, when used by access_process_vm.
514 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
516 if (likely(vma->vm_flags & VM_WRITE))
517 pte = pte_mkwrite(pte);
523 * Multiple processes may "see" the same page. E.g. for untouched
524 * mappings of /dev/null, all processes see the same page full of
525 * zeroes, and text pages of executables and shared libraries have
526 * only one copy in memory, at most, normally.
528 * For the non-reserved pages, page_count(page) denotes a reference count.
529 * page_count() == 0 means the page is free. page->lru is then used for
530 * freelist management in the buddy allocator.
531 * page_count() > 0 means the page has been allocated.
533 * Pages are allocated by the slab allocator in order to provide memory
534 * to kmalloc and kmem_cache_alloc. In this case, the management of the
535 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
536 * unless a particular usage is carefully commented. (the responsibility of
537 * freeing the kmalloc memory is the caller's, of course).
539 * A page may be used by anyone else who does a __get_free_page().
540 * In this case, page_count still tracks the references, and should only
541 * be used through the normal accessor functions. The top bits of page->flags
542 * and page->virtual store page management information, but all other fields
543 * are unused and could be used privately, carefully. The management of this
544 * page is the responsibility of the one who allocated it, and those who have
545 * subsequently been given references to it.
547 * The other pages (we may call them "pagecache pages") are completely
548 * managed by the Linux memory manager: I/O, buffers, swapping etc.
549 * The following discussion applies only to them.
551 * A pagecache page contains an opaque `private' member, which belongs to the
552 * page's address_space. Usually, this is the address of a circular list of
553 * the page's disk buffers. PG_private must be set to tell the VM to call
554 * into the filesystem to release these pages.
556 * A page may belong to an inode's memory mapping. In this case, page->mapping
557 * is the pointer to the inode, and page->index is the file offset of the page,
558 * in units of PAGE_CACHE_SIZE.
560 * If pagecache pages are not associated with an inode, they are said to be
561 * anonymous pages. These may become associated with the swapcache, and in that
562 * case PG_swapcache is set, and page->private is an offset into the swapcache.
564 * In either case (swapcache or inode backed), the pagecache itself holds one
565 * reference to the page. Setting PG_private should also increment the
566 * refcount. The each user mapping also has a reference to the page.
568 * The pagecache pages are stored in a per-mapping radix tree, which is
569 * rooted at mapping->page_tree, and indexed by offset.
570 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
571 * lists, we instead now tag pages as dirty/writeback in the radix tree.
573 * All pagecache pages may be subject to I/O:
574 * - inode pages may need to be read from disk,
575 * - inode pages which have been modified and are MAP_SHARED may need
576 * to be written back to the inode on disk,
577 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
578 * modified may need to be swapped out to swap space and (later) to be read
583 * The zone field is never updated after free_area_init_core()
584 * sets it, so none of the operations on it need to be atomic.
587 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
588 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
589 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
590 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
591 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
594 * Define the bit shifts to access each section. For non-existent
595 * sections we define the shift as 0; that plus a 0 mask ensures
596 * the compiler will optimise away reference to them.
598 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
599 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
600 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
601 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
603 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
604 #ifdef NODE_NOT_IN_PAGE_FLAGS
605 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
606 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
607 SECTIONS_PGOFF : ZONES_PGOFF)
609 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
610 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
611 NODES_PGOFF : ZONES_PGOFF)
614 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
616 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
617 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
620 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
621 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
622 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
623 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
624 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
626 static inline enum zone_type page_zonenum(const struct page *page)
628 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
631 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
632 #define SECTION_IN_PAGE_FLAGS
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(const struct page *page);
660 static inline int page_to_nid(const struct page *page)
662 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
666 #ifdef CONFIG_NUMA_BALANCING
667 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
668 static inline int page_nid_xchg_last(struct page *page, int nid)
670 return xchg(&page->_last_nid, nid);
673 static inline int page_nid_last(struct page *page)
675 return page->_last_nid;
677 static inline void page_nid_reset_last(struct page *page)
679 page->_last_nid = -1;
682 static inline int page_nid_last(struct page *page)
684 return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
687 extern int page_nid_xchg_last(struct page *page, int nid);
689 static inline void page_nid_reset_last(struct page *page)
691 int nid = (1 << LAST_NID_SHIFT) - 1;
693 page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
694 page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
696 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
698 static inline int page_nid_xchg_last(struct page *page, int nid)
700 return page_to_nid(page);
703 static inline int page_nid_last(struct page *page)
705 return page_to_nid(page);
708 static inline void page_nid_reset_last(struct page *page)
713 static inline struct zone *page_zone(const struct page *page)
715 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
718 #ifdef SECTION_IN_PAGE_FLAGS
719 static inline void set_page_section(struct page *page, unsigned long section)
721 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
722 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
725 static inline unsigned long page_to_section(const struct page *page)
727 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
731 static inline void set_page_zone(struct page *page, enum zone_type zone)
733 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
734 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
737 static inline void set_page_node(struct page *page, unsigned long node)
739 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
740 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
743 static inline void set_page_links(struct page *page, enum zone_type zone,
744 unsigned long node, unsigned long pfn)
746 set_page_zone(page, zone);
747 set_page_node(page, node);
748 #ifdef SECTION_IN_PAGE_FLAGS
749 set_page_section(page, pfn_to_section_nr(pfn));
754 * Some inline functions in vmstat.h depend on page_zone()
756 #include <linux/vmstat.h>
758 static __always_inline void *lowmem_page_address(const struct page *page)
760 return __va(PFN_PHYS(page_to_pfn(page)));
763 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
764 #define HASHED_PAGE_VIRTUAL
767 #if defined(WANT_PAGE_VIRTUAL)
768 #define page_address(page) ((page)->virtual)
769 #define set_page_address(page, address) \
771 (page)->virtual = (address); \
773 #define page_address_init() do { } while(0)
776 #if defined(HASHED_PAGE_VIRTUAL)
777 void *page_address(const struct page *page);
778 void set_page_address(struct page *page, void *virtual);
779 void page_address_init(void);
782 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
783 #define page_address(page) lowmem_page_address(page)
784 #define set_page_address(page, address) do { } while(0)
785 #define page_address_init() do { } while(0)
789 * On an anonymous page mapped into a user virtual memory area,
790 * page->mapping points to its anon_vma, not to a struct address_space;
791 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
793 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
794 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
795 * and then page->mapping points, not to an anon_vma, but to a private
796 * structure which KSM associates with that merged page. See ksm.h.
798 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
800 * Please note that, confusingly, "page_mapping" refers to the inode
801 * address_space which maps the page from disk; whereas "page_mapped"
802 * refers to user virtual address space into which the page is mapped.
804 #define PAGE_MAPPING_ANON 1
805 #define PAGE_MAPPING_KSM 2
806 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
808 extern struct address_space *page_mapping(struct page *page);
810 /* Neutral page->mapping pointer to address_space or anon_vma or other */
811 static inline void *page_rmapping(struct page *page)
813 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
816 extern struct address_space *__page_file_mapping(struct page *);
819 struct address_space *page_file_mapping(struct page *page)
821 if (unlikely(PageSwapCache(page)))
822 return __page_file_mapping(page);
824 return page->mapping;
827 static inline int PageAnon(struct page *page)
829 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
833 * Return the pagecache index of the passed page. Regular pagecache pages
834 * use ->index whereas swapcache pages use ->private
836 static inline pgoff_t page_index(struct page *page)
838 if (unlikely(PageSwapCache(page)))
839 return page_private(page);
843 extern pgoff_t __page_file_index(struct page *page);
846 * Return the file index of the page. Regular pagecache pages use ->index
847 * whereas swapcache pages use swp_offset(->private)
849 static inline pgoff_t page_file_index(struct page *page)
851 if (unlikely(PageSwapCache(page)))
852 return __page_file_index(page);
858 * Return true if this page is mapped into pagetables.
860 static inline int page_mapped(struct page *page)
862 return atomic_read(&(page)->_mapcount) >= 0;
866 * Different kinds of faults, as returned by handle_mm_fault().
867 * Used to decide whether a process gets delivered SIGBUS or
868 * just gets major/minor fault counters bumped up.
871 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
873 #define VM_FAULT_OOM 0x0001
874 #define VM_FAULT_SIGBUS 0x0002
875 #define VM_FAULT_MAJOR 0x0004
876 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
877 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
878 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
880 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
881 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
882 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
884 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
886 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
887 VM_FAULT_HWPOISON_LARGE)
889 /* Encode hstate index for a hwpoisoned large page */
890 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
891 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
894 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
896 extern void pagefault_out_of_memory(void);
898 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
901 * Flags passed to show_mem() and show_free_areas() to suppress output in
904 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
905 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
907 extern void show_free_areas(unsigned int flags);
908 extern bool skip_free_areas_node(unsigned int flags, int nid);
910 int shmem_zero_setup(struct vm_area_struct *);
912 extern int can_do_mlock(void);
913 extern int user_shm_lock(size_t, struct user_struct *);
914 extern void user_shm_unlock(size_t, struct user_struct *);
917 * Parameter block passed down to zap_pte_range in exceptional cases.
920 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
921 struct address_space *check_mapping; /* Check page->mapping if set */
922 pgoff_t first_index; /* Lowest page->index to unmap */
923 pgoff_t last_index; /* Highest page->index to unmap */
926 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
929 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
931 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
932 unsigned long size, struct zap_details *);
933 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
934 unsigned long start, unsigned long end);
937 * mm_walk - callbacks for walk_page_range
938 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
939 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
940 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
941 * this handler is required to be able to handle
942 * pmd_trans_huge() pmds. They may simply choose to
943 * split_huge_page() instead of handling it explicitly.
944 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
945 * @pte_hole: if set, called for each hole at all levels
946 * @hugetlb_entry: if set, called for each hugetlb entry
947 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
950 * (see walk_page_range for more details)
953 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
954 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
955 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
956 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
957 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
958 int (*hugetlb_entry)(pte_t *, unsigned long,
959 unsigned long, unsigned long, struct mm_walk *);
960 struct mm_struct *mm;
964 int walk_page_range(unsigned long addr, unsigned long end,
965 struct mm_walk *walk);
966 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
967 unsigned long end, unsigned long floor, unsigned long ceiling);
968 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
969 struct vm_area_struct *vma);
970 void unmap_mapping_range(struct address_space *mapping,
971 loff_t const holebegin, loff_t const holelen, int even_cows);
972 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
974 int follow_phys(struct vm_area_struct *vma, unsigned long address,
975 unsigned int flags, unsigned long *prot, resource_size_t *phys);
976 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
977 void *buf, int len, int write);
979 static inline void unmap_shared_mapping_range(struct address_space *mapping,
980 loff_t const holebegin, loff_t const holelen)
982 unmap_mapping_range(mapping, holebegin, holelen, 0);
985 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
986 extern void truncate_setsize(struct inode *inode, loff_t newsize);
987 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
988 int truncate_inode_page(struct address_space *mapping, struct page *page);
989 int generic_error_remove_page(struct address_space *mapping, struct page *page);
990 int invalidate_inode_page(struct page *page);
993 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
994 unsigned long address, unsigned int flags);
995 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
996 unsigned long address, unsigned int fault_flags);
998 static inline int handle_mm_fault(struct mm_struct *mm,
999 struct vm_area_struct *vma, unsigned long address,
1002 /* should never happen if there's no MMU */
1004 return VM_FAULT_SIGBUS;
1006 static inline int fixup_user_fault(struct task_struct *tsk,
1007 struct mm_struct *mm, unsigned long address,
1008 unsigned int fault_flags)
1010 /* should never happen if there's no MMU */
1016 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1017 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1018 void *buf, int len, int write);
1020 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1021 unsigned long start, unsigned long nr_pages,
1022 unsigned int foll_flags, struct page **pages,
1023 struct vm_area_struct **vmas, int *nonblocking);
1024 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1025 unsigned long start, unsigned long nr_pages,
1026 int write, int force, struct page **pages,
1027 struct vm_area_struct **vmas);
1028 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1029 struct page **pages);
1031 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1032 struct page **pages);
1033 int get_kernel_page(unsigned long start, int write, struct page **pages);
1034 struct page *get_dump_page(unsigned long addr);
1036 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1037 extern void do_invalidatepage(struct page *page, unsigned long offset);
1039 int __set_page_dirty_nobuffers(struct page *page);
1040 int __set_page_dirty_no_writeback(struct page *page);
1041 int redirty_page_for_writepage(struct writeback_control *wbc,
1043 void account_page_dirtied(struct page *page, struct address_space *mapping);
1044 void account_page_writeback(struct page *page);
1045 int set_page_dirty(struct page *page);
1046 int set_page_dirty_lock(struct page *page);
1047 int clear_page_dirty_for_io(struct page *page);
1049 /* Is the vma a continuation of the stack vma above it? */
1050 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1052 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1055 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1058 return (vma->vm_flags & VM_GROWSDOWN) &&
1059 (vma->vm_start == addr) &&
1060 !vma_growsdown(vma->vm_prev, addr);
1063 /* Is the vma a continuation of the stack vma below it? */
1064 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1066 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1069 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1072 return (vma->vm_flags & VM_GROWSUP) &&
1073 (vma->vm_end == addr) &&
1074 !vma_growsup(vma->vm_next, addr);
1078 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1080 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1081 unsigned long old_addr, struct vm_area_struct *new_vma,
1082 unsigned long new_addr, unsigned long len,
1083 bool need_rmap_locks);
1084 extern unsigned long do_mremap(unsigned long addr,
1085 unsigned long old_len, unsigned long new_len,
1086 unsigned long flags, unsigned long new_addr);
1087 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1088 unsigned long end, pgprot_t newprot,
1089 int dirty_accountable, int prot_numa);
1090 extern int mprotect_fixup(struct vm_area_struct *vma,
1091 struct vm_area_struct **pprev, unsigned long start,
1092 unsigned long end, unsigned long newflags);
1095 * doesn't attempt to fault and will return short.
1097 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1098 struct page **pages);
1100 * per-process(per-mm_struct) statistics.
1102 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1104 long val = atomic_long_read(&mm->rss_stat.count[member]);
1106 #ifdef SPLIT_RSS_COUNTING
1108 * counter is updated in asynchronous manner and may go to minus.
1109 * But it's never be expected number for users.
1114 return (unsigned long)val;
1117 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1119 atomic_long_add(value, &mm->rss_stat.count[member]);
1122 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1124 atomic_long_inc(&mm->rss_stat.count[member]);
1127 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1129 atomic_long_dec(&mm->rss_stat.count[member]);
1132 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1134 return get_mm_counter(mm, MM_FILEPAGES) +
1135 get_mm_counter(mm, MM_ANONPAGES);
1138 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1140 return max(mm->hiwater_rss, get_mm_rss(mm));
1143 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1145 return max(mm->hiwater_vm, mm->total_vm);
1148 static inline void update_hiwater_rss(struct mm_struct *mm)
1150 unsigned long _rss = get_mm_rss(mm);
1152 if ((mm)->hiwater_rss < _rss)
1153 (mm)->hiwater_rss = _rss;
1156 static inline void update_hiwater_vm(struct mm_struct *mm)
1158 if (mm->hiwater_vm < mm->total_vm)
1159 mm->hiwater_vm = mm->total_vm;
1162 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1163 struct mm_struct *mm)
1165 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1167 if (*maxrss < hiwater_rss)
1168 *maxrss = hiwater_rss;
1171 #if defined(SPLIT_RSS_COUNTING)
1172 void sync_mm_rss(struct mm_struct *mm);
1174 static inline void sync_mm_rss(struct mm_struct *mm)
1179 int vma_wants_writenotify(struct vm_area_struct *vma);
1181 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1183 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1187 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1191 #ifdef __PAGETABLE_PUD_FOLDED
1192 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1193 unsigned long address)
1198 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1201 #ifdef __PAGETABLE_PMD_FOLDED
1202 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1203 unsigned long address)
1208 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1211 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1212 pmd_t *pmd, unsigned long address);
1213 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1216 * The following ifdef needed to get the 4level-fixup.h header to work.
1217 * Remove it when 4level-fixup.h has been removed.
1219 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1220 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1222 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1223 NULL: pud_offset(pgd, address);
1226 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1228 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1229 NULL: pmd_offset(pud, address);
1231 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1233 #if USE_SPLIT_PTLOCKS
1235 * We tuck a spinlock to guard each pagetable page into its struct page,
1236 * at page->private, with BUILD_BUG_ON to make sure that this will not
1237 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1238 * When freeing, reset page->mapping so free_pages_check won't complain.
1240 #define __pte_lockptr(page) &((page)->ptl)
1241 #define pte_lock_init(_page) do { \
1242 spin_lock_init(__pte_lockptr(_page)); \
1244 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1245 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1246 #else /* !USE_SPLIT_PTLOCKS */
1248 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1250 #define pte_lock_init(page) do {} while (0)
1251 #define pte_lock_deinit(page) do {} while (0)
1252 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1253 #endif /* USE_SPLIT_PTLOCKS */
1255 static inline void pgtable_page_ctor(struct page *page)
1257 pte_lock_init(page);
1258 inc_zone_page_state(page, NR_PAGETABLE);
1261 static inline void pgtable_page_dtor(struct page *page)
1263 pte_lock_deinit(page);
1264 dec_zone_page_state(page, NR_PAGETABLE);
1267 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1269 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1270 pte_t *__pte = pte_offset_map(pmd, address); \
1276 #define pte_unmap_unlock(pte, ptl) do { \
1281 #define pte_alloc_map(mm, vma, pmd, address) \
1282 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1284 NULL: pte_offset_map(pmd, address))
1286 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1287 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1289 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1291 #define pte_alloc_kernel(pmd, address) \
1292 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1293 NULL: pte_offset_kernel(pmd, address))
1295 extern void free_area_init(unsigned long * zones_size);
1296 extern void free_area_init_node(int nid, unsigned long * zones_size,
1297 unsigned long zone_start_pfn, unsigned long *zholes_size);
1298 extern void free_initmem(void);
1301 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1302 * into the buddy system. The freed pages will be poisoned with pattern
1303 * "poison" if it's non-zero.
1304 * Return pages freed into the buddy system.
1306 extern unsigned long free_reserved_area(unsigned long start, unsigned long end,
1307 int poison, char *s);
1308 #ifdef CONFIG_HIGHMEM
1310 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1311 * and totalram_pages.
1313 extern void free_highmem_page(struct page *page);
1316 static inline void adjust_managed_page_count(struct page *page, long count)
1318 totalram_pages += count;
1321 /* Free the reserved page into the buddy system, so it gets managed. */
1322 static inline void __free_reserved_page(struct page *page)
1324 ClearPageReserved(page);
1325 init_page_count(page);
1329 static inline void free_reserved_page(struct page *page)
1331 __free_reserved_page(page);
1332 adjust_managed_page_count(page, 1);
1335 static inline void mark_page_reserved(struct page *page)
1337 SetPageReserved(page);
1338 adjust_managed_page_count(page, -1);
1342 * Default method to free all the __init memory into the buddy system.
1343 * The freed pages will be poisoned with pattern "poison" if it is
1344 * non-zero. Return pages freed into the buddy system.
1346 static inline unsigned long free_initmem_default(int poison)
1348 extern char __init_begin[], __init_end[];
1350 return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin) ,
1351 ((unsigned long)&__init_end) & PAGE_MASK,
1352 poison, "unused kernel");
1355 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1357 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1358 * zones, allocate the backing mem_map and account for memory holes in a more
1359 * architecture independent manner. This is a substitute for creating the
1360 * zone_sizes[] and zholes_size[] arrays and passing them to
1361 * free_area_init_node()
1363 * An architecture is expected to register range of page frames backed by
1364 * physical memory with memblock_add[_node]() before calling
1365 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1366 * usage, an architecture is expected to do something like
1368 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1370 * for_each_valid_physical_page_range()
1371 * memblock_add_node(base, size, nid)
1372 * free_area_init_nodes(max_zone_pfns);
1374 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1375 * registered physical page range. Similarly
1376 * sparse_memory_present_with_active_regions() calls memory_present() for
1377 * each range when SPARSEMEM is enabled.
1379 * See mm/page_alloc.c for more information on each function exposed by
1380 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1382 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1383 unsigned long node_map_pfn_alignment(void);
1384 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1385 unsigned long end_pfn);
1386 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1387 unsigned long end_pfn);
1388 extern void get_pfn_range_for_nid(unsigned int nid,
1389 unsigned long *start_pfn, unsigned long *end_pfn);
1390 extern unsigned long find_min_pfn_with_active_regions(void);
1391 extern void free_bootmem_with_active_regions(int nid,
1392 unsigned long max_low_pfn);
1393 extern void sparse_memory_present_with_active_regions(int nid);
1395 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1397 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1398 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1399 static inline int __early_pfn_to_nid(unsigned long pfn)
1404 /* please see mm/page_alloc.c */
1405 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1406 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1407 /* there is a per-arch backend function. */
1408 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1409 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1412 extern void set_dma_reserve(unsigned long new_dma_reserve);
1413 extern void memmap_init_zone(unsigned long, int, unsigned long,
1414 unsigned long, enum memmap_context);
1415 extern void setup_per_zone_wmarks(void);
1416 extern int __meminit init_per_zone_wmark_min(void);
1417 extern void mem_init(void);
1418 extern void __init mmap_init(void);
1419 extern void show_mem(unsigned int flags);
1420 extern void si_meminfo(struct sysinfo * val);
1421 extern void si_meminfo_node(struct sysinfo *val, int nid);
1423 extern __printf(3, 4)
1424 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1426 extern void setup_per_cpu_pageset(void);
1428 extern void zone_pcp_update(struct zone *zone);
1429 extern void zone_pcp_reset(struct zone *zone);
1432 extern int min_free_kbytes;
1435 extern atomic_long_t mmap_pages_allocated;
1436 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1438 /* interval_tree.c */
1439 void vma_interval_tree_insert(struct vm_area_struct *node,
1440 struct rb_root *root);
1441 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1442 struct vm_area_struct *prev,
1443 struct rb_root *root);
1444 void vma_interval_tree_remove(struct vm_area_struct *node,
1445 struct rb_root *root);
1446 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1447 unsigned long start, unsigned long last);
1448 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1449 unsigned long start, unsigned long last);
1451 #define vma_interval_tree_foreach(vma, root, start, last) \
1452 for (vma = vma_interval_tree_iter_first(root, start, last); \
1453 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1455 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1456 struct list_head *list)
1458 list_add_tail(&vma->shared.nonlinear, list);
1461 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1462 struct rb_root *root);
1463 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1464 struct rb_root *root);
1465 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1466 struct rb_root *root, unsigned long start, unsigned long last);
1467 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1468 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1469 #ifdef CONFIG_DEBUG_VM_RB
1470 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1473 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1474 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1475 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1478 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1479 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1480 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1481 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1482 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1483 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1484 struct mempolicy *);
1485 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1486 extern int split_vma(struct mm_struct *,
1487 struct vm_area_struct *, unsigned long addr, int new_below);
1488 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1489 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1490 struct rb_node **, struct rb_node *);
1491 extern void unlink_file_vma(struct vm_area_struct *);
1492 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1493 unsigned long addr, unsigned long len, pgoff_t pgoff,
1494 bool *need_rmap_locks);
1495 extern void exit_mmap(struct mm_struct *);
1497 extern int mm_take_all_locks(struct mm_struct *mm);
1498 extern void mm_drop_all_locks(struct mm_struct *mm);
1500 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1501 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1503 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1504 extern int install_special_mapping(struct mm_struct *mm,
1505 unsigned long addr, unsigned long len,
1506 unsigned long flags, struct page **pages);
1508 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1510 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1511 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1512 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1513 unsigned long len, unsigned long prot, unsigned long flags,
1514 unsigned long pgoff, unsigned long *populate);
1515 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1518 extern int __mm_populate(unsigned long addr, unsigned long len,
1520 static inline void mm_populate(unsigned long addr, unsigned long len)
1523 (void) __mm_populate(addr, len, 1);
1526 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1529 /* These take the mm semaphore themselves */
1530 extern unsigned long vm_brk(unsigned long, unsigned long);
1531 extern int vm_munmap(unsigned long, size_t);
1532 extern unsigned long vm_mmap(struct file *, unsigned long,
1533 unsigned long, unsigned long,
1534 unsigned long, unsigned long);
1536 struct vm_unmapped_area_info {
1537 #define VM_UNMAPPED_AREA_TOPDOWN 1
1538 unsigned long flags;
1539 unsigned long length;
1540 unsigned long low_limit;
1541 unsigned long high_limit;
1542 unsigned long align_mask;
1543 unsigned long align_offset;
1546 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1547 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1550 * Search for an unmapped address range.
1552 * We are looking for a range that:
1553 * - does not intersect with any VMA;
1554 * - is contained within the [low_limit, high_limit) interval;
1555 * - is at least the desired size.
1556 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1558 static inline unsigned long
1559 vm_unmapped_area(struct vm_unmapped_area_info *info)
1561 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1562 return unmapped_area(info);
1564 return unmapped_area_topdown(info);
1568 extern void truncate_inode_pages(struct address_space *, loff_t);
1569 extern void truncate_inode_pages_range(struct address_space *,
1570 loff_t lstart, loff_t lend);
1572 /* generic vm_area_ops exported for stackable file systems */
1573 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1574 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1576 /* mm/page-writeback.c */
1577 int write_one_page(struct page *page, int wait);
1578 void task_dirty_inc(struct task_struct *tsk);
1581 #define VM_MAX_READAHEAD 128 /* kbytes */
1582 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1584 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1585 pgoff_t offset, unsigned long nr_to_read);
1587 void page_cache_sync_readahead(struct address_space *mapping,
1588 struct file_ra_state *ra,
1591 unsigned long size);
1593 void page_cache_async_readahead(struct address_space *mapping,
1594 struct file_ra_state *ra,
1598 unsigned long size);
1600 unsigned long max_sane_readahead(unsigned long nr);
1601 unsigned long ra_submit(struct file_ra_state *ra,
1602 struct address_space *mapping,
1605 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1606 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1608 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1609 extern int expand_downwards(struct vm_area_struct *vma,
1610 unsigned long address);
1612 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1614 #define expand_upwards(vma, address) do { } while (0)
1617 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1618 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1619 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1620 struct vm_area_struct **pprev);
1622 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1623 NULL if none. Assume start_addr < end_addr. */
1624 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1626 struct vm_area_struct * vma = find_vma(mm,start_addr);
1628 if (vma && end_addr <= vma->vm_start)
1633 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1635 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1638 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1639 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1640 unsigned long vm_start, unsigned long vm_end)
1642 struct vm_area_struct *vma = find_vma(mm, vm_start);
1644 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1651 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1653 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1659 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1660 unsigned long change_prot_numa(struct vm_area_struct *vma,
1661 unsigned long start, unsigned long end);
1664 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1665 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1666 unsigned long pfn, unsigned long size, pgprot_t);
1667 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1668 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1670 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1672 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1675 struct page *follow_page_mask(struct vm_area_struct *vma,
1676 unsigned long address, unsigned int foll_flags,
1677 unsigned int *page_mask);
1679 static inline struct page *follow_page(struct vm_area_struct *vma,
1680 unsigned long address, unsigned int foll_flags)
1682 unsigned int unused_page_mask;
1683 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1686 #define FOLL_WRITE 0x01 /* check pte is writable */
1687 #define FOLL_TOUCH 0x02 /* mark page accessed */
1688 #define FOLL_GET 0x04 /* do get_page on page */
1689 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1690 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1691 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1692 * and return without waiting upon it */
1693 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1694 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1695 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1696 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1697 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1699 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1701 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1702 unsigned long size, pte_fn_t fn, void *data);
1704 #ifdef CONFIG_PROC_FS
1705 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1707 static inline void vm_stat_account(struct mm_struct *mm,
1708 unsigned long flags, struct file *file, long pages)
1710 mm->total_vm += pages;
1712 #endif /* CONFIG_PROC_FS */
1714 #ifdef CONFIG_DEBUG_PAGEALLOC
1715 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1716 #ifdef CONFIG_HIBERNATION
1717 extern bool kernel_page_present(struct page *page);
1718 #endif /* CONFIG_HIBERNATION */
1721 kernel_map_pages(struct page *page, int numpages, int enable) {}
1722 #ifdef CONFIG_HIBERNATION
1723 static inline bool kernel_page_present(struct page *page) { return true; }
1724 #endif /* CONFIG_HIBERNATION */
1727 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1728 #ifdef __HAVE_ARCH_GATE_AREA
1729 int in_gate_area_no_mm(unsigned long addr);
1730 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1732 int in_gate_area_no_mm(unsigned long addr);
1733 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1734 #endif /* __HAVE_ARCH_GATE_AREA */
1736 #ifdef CONFIG_SYSCTL
1737 extern int sysctl_drop_caches;
1738 int drop_caches_sysctl_handler(struct ctl_table *, int,
1739 void __user *, size_t *, loff_t *);
1742 unsigned long shrink_slab(struct shrink_control *shrink,
1743 unsigned long nr_pages_scanned,
1744 unsigned long lru_pages);
1747 #define randomize_va_space 0
1749 extern int randomize_va_space;
1752 const char * arch_vma_name(struct vm_area_struct *vma);
1753 void print_vma_addr(char *prefix, unsigned long rip);
1755 void sparse_mem_maps_populate_node(struct page **map_map,
1756 unsigned long pnum_begin,
1757 unsigned long pnum_end,
1758 unsigned long map_count,
1761 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1762 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1763 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1764 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1765 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1766 void *vmemmap_alloc_block(unsigned long size, int node);
1767 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1768 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1769 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1771 int vmemmap_populate(unsigned long start, unsigned long end, int node);
1772 void vmemmap_populate_print_last(void);
1773 #ifdef CONFIG_MEMORY_HOTPLUG
1774 void vmemmap_free(unsigned long start, unsigned long end);
1776 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1777 unsigned long size);
1780 MF_COUNT_INCREASED = 1 << 0,
1781 MF_ACTION_REQUIRED = 1 << 1,
1782 MF_MUST_KILL = 1 << 2,
1784 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1785 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1786 extern int unpoison_memory(unsigned long pfn);
1787 extern int sysctl_memory_failure_early_kill;
1788 extern int sysctl_memory_failure_recovery;
1789 extern void shake_page(struct page *p, int access);
1790 extern atomic_long_t num_poisoned_pages;
1791 extern int soft_offline_page(struct page *page, int flags);
1793 extern void dump_page(struct page *page);
1795 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1796 extern void clear_huge_page(struct page *page,
1798 unsigned int pages_per_huge_page);
1799 extern void copy_user_huge_page(struct page *dst, struct page *src,
1800 unsigned long addr, struct vm_area_struct *vma,
1801 unsigned int pages_per_huge_page);
1802 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1804 #ifdef CONFIG_DEBUG_PAGEALLOC
1805 extern unsigned int _debug_guardpage_minorder;
1807 static inline unsigned int debug_guardpage_minorder(void)
1809 return _debug_guardpage_minorder;
1812 static inline bool page_is_guard(struct page *page)
1814 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1817 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1818 static inline bool page_is_guard(struct page *page) { return false; }
1819 #endif /* CONFIG_DEBUG_PAGEALLOC */
1821 #if MAX_NUMNODES > 1
1822 void __init setup_nr_node_ids(void);
1824 static inline void setup_nr_node_ids(void) {}
1827 #endif /* __KERNEL__ */
1828 #endif /* _LINUX_MM_H */