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 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
49 /* to align the pointer to the (next) page boundary */
50 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
53 * Linux kernel virtual memory manager primitives.
54 * The idea being to have a "virtual" mm in the same way
55 * we have a virtual fs - giving a cleaner interface to the
56 * mm details, and allowing different kinds of memory mappings
57 * (from shared memory to executable loading to arbitrary
61 extern struct kmem_cache *vm_area_cachep;
64 extern struct rb_root nommu_region_tree;
65 extern struct rw_semaphore nommu_region_sem;
67 extern unsigned int kobjsize(const void *objp);
71 * vm_flags in vm_area_struct, see mm_types.h.
73 #define VM_NONE 0x00000000
75 #define VM_READ 0x00000001 /* currently active flags */
76 #define VM_WRITE 0x00000002
77 #define VM_EXEC 0x00000004
78 #define VM_SHARED 0x00000008
80 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
81 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82 #define VM_MAYWRITE 0x00000020
83 #define VM_MAYEXEC 0x00000040
84 #define VM_MAYSHARE 0x00000080
86 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
88 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
90 #define VM_LOCKED 0x00002000
91 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
93 /* Used by sys_madvise() */
94 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
95 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
97 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
98 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
99 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
100 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
101 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
102 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
103 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
104 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
106 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
107 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
108 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
109 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
111 #if defined(CONFIG_X86)
112 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
113 #elif defined(CONFIG_PPC)
114 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
115 #elif defined(CONFIG_PARISC)
116 # define VM_GROWSUP VM_ARCH_1
117 #elif defined(CONFIG_IA64)
118 # define VM_GROWSUP VM_ARCH_1
119 #elif !defined(CONFIG_MMU)
120 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
124 # define VM_GROWSUP VM_NONE
127 /* Bits set in the VMA until the stack is in its final location */
128 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
130 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
131 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
134 #ifdef CONFIG_STACK_GROWSUP
135 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
137 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
140 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
141 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
142 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
143 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
144 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
147 * Special vmas that are non-mergable, non-mlock()able.
148 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
150 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
153 * mapping from the currently active vm_flags protection bits (the
154 * low four bits) to a page protection mask..
156 extern pgprot_t protection_map[16];
158 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
159 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
160 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
161 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
162 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
163 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
164 #define FAULT_FLAG_TRIED 0x40 /* second try */
167 * Some architectures (such as x86) may need to preserve certain pgprot
168 * bits, without complicating generic pgprot code.
170 * Most architectures don't care:
172 #ifndef pgprot_modify
173 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
180 * vm_fault is filled by the the pagefault handler and passed to the vma's
181 * ->fault function. The vma's ->fault is responsible for returning a bitmask
182 * of VM_FAULT_xxx flags that give details about how the fault was handled.
184 * pgoff should be used in favour of virtual_address, if possible. If pgoff
185 * is used, one may implement ->remap_pages to get nonlinear mapping support.
188 unsigned int flags; /* FAULT_FLAG_xxx flags */
189 pgoff_t pgoff; /* Logical page offset based on vma */
190 void __user *virtual_address; /* Faulting virtual address */
192 struct page *page; /* ->fault handlers should return a
193 * page here, unless VM_FAULT_NOPAGE
194 * is set (which is also implied by
200 * These are the virtual MM functions - opening of an area, closing and
201 * unmapping it (needed to keep files on disk up-to-date etc), pointer
202 * to the functions called when a no-page or a wp-page exception occurs.
204 struct vm_operations_struct {
205 void (*open)(struct vm_area_struct * area);
206 void (*close)(struct vm_area_struct * area);
207 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
209 /* notification that a previously read-only page is about to become
210 * writable, if an error is returned it will cause a SIGBUS */
211 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
213 /* called by access_process_vm when get_user_pages() fails, typically
214 * for use by special VMAs that can switch between memory and hardware
216 int (*access)(struct vm_area_struct *vma, unsigned long addr,
217 void *buf, int len, int write);
220 * set_policy() op must add a reference to any non-NULL @new mempolicy
221 * to hold the policy upon return. Caller should pass NULL @new to
222 * remove a policy and fall back to surrounding context--i.e. do not
223 * install a MPOL_DEFAULT policy, nor the task or system default
226 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
229 * get_policy() op must add reference [mpol_get()] to any policy at
230 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
231 * in mm/mempolicy.c will do this automatically.
232 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
233 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
234 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
235 * must return NULL--i.e., do not "fallback" to task or system default
238 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
240 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
241 const nodemask_t *to, unsigned long flags);
243 /* called by sys_remap_file_pages() to populate non-linear mapping */
244 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
245 unsigned long size, pgoff_t pgoff);
251 #define page_private(page) ((page)->private)
252 #define set_page_private(page, v) ((page)->private = (v))
254 /* It's valid only if the page is free path or free_list */
255 static inline void set_freepage_migratetype(struct page *page, int migratetype)
257 page->index = migratetype;
260 /* It's valid only if the page is free path or free_list */
261 static inline int get_freepage_migratetype(struct page *page)
267 * FIXME: take this include out, include page-flags.h in
268 * files which need it (119 of them)
270 #include <linux/page-flags.h>
271 #include <linux/huge_mm.h>
274 * Methods to modify the page usage count.
276 * What counts for a page usage:
277 * - cache mapping (page->mapping)
278 * - private data (page->private)
279 * - page mapped in a task's page tables, each mapping
280 * is counted separately
282 * Also, many kernel routines increase the page count before a critical
283 * routine so they can be sure the page doesn't go away from under them.
287 * Drop a ref, return true if the refcount fell to zero (the page has no users)
289 static inline int put_page_testzero(struct page *page)
291 VM_BUG_ON(atomic_read(&page->_count) == 0);
292 return atomic_dec_and_test(&page->_count);
296 * Try to grab a ref unless the page has a refcount of zero, return false if
299 static inline int get_page_unless_zero(struct page *page)
301 return atomic_inc_not_zero(&page->_count);
304 extern int page_is_ram(unsigned long pfn);
306 /* Support for virtually mapped pages */
307 struct page *vmalloc_to_page(const void *addr);
308 unsigned long vmalloc_to_pfn(const void *addr);
311 * Determine if an address is within the vmalloc range
313 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
314 * is no special casing required.
316 static inline int is_vmalloc_addr(const void *x)
319 unsigned long addr = (unsigned long)x;
321 return addr >= VMALLOC_START && addr < VMALLOC_END;
327 extern int is_vmalloc_or_module_addr(const void *x);
329 static inline int is_vmalloc_or_module_addr(const void *x)
335 static inline void compound_lock(struct page *page)
337 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
338 VM_BUG_ON(PageSlab(page));
339 bit_spin_lock(PG_compound_lock, &page->flags);
343 static inline void compound_unlock(struct page *page)
345 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
346 VM_BUG_ON(PageSlab(page));
347 bit_spin_unlock(PG_compound_lock, &page->flags);
351 static inline unsigned long compound_lock_irqsave(struct page *page)
353 unsigned long uninitialized_var(flags);
354 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
355 local_irq_save(flags);
361 static inline void compound_unlock_irqrestore(struct page *page,
364 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
365 compound_unlock(page);
366 local_irq_restore(flags);
370 static inline struct page *compound_head(struct page *page)
372 if (unlikely(PageTail(page)))
373 return page->first_page;
378 * The atomic page->_mapcount, starts from -1: so that transitions
379 * both from it and to it can be tracked, using atomic_inc_and_test
380 * and atomic_add_negative(-1).
382 static inline void reset_page_mapcount(struct page *page)
384 atomic_set(&(page)->_mapcount, -1);
387 static inline int page_mapcount(struct page *page)
389 return atomic_read(&(page)->_mapcount) + 1;
392 static inline int page_count(struct page *page)
394 return atomic_read(&compound_head(page)->_count);
397 static inline void get_huge_page_tail(struct page *page)
400 * __split_huge_page_refcount() cannot run
403 VM_BUG_ON(page_mapcount(page) < 0);
404 VM_BUG_ON(atomic_read(&page->_count) != 0);
405 atomic_inc(&page->_mapcount);
408 extern bool __get_page_tail(struct page *page);
410 static inline void get_page(struct page *page)
412 if (unlikely(PageTail(page)))
413 if (likely(__get_page_tail(page)))
416 * Getting a normal page or the head of a compound page
417 * requires to already have an elevated page->_count.
419 VM_BUG_ON(atomic_read(&page->_count) <= 0);
420 atomic_inc(&page->_count);
423 static inline struct page *virt_to_head_page(const void *x)
425 struct page *page = virt_to_page(x);
426 return compound_head(page);
430 * Setup the page count before being freed into the page allocator for
431 * the first time (boot or memory hotplug)
433 static inline void init_page_count(struct page *page)
435 atomic_set(&page->_count, 1);
439 * PageBuddy() indicate that the page is free and in the buddy system
440 * (see mm/page_alloc.c).
442 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
443 * -2 so that an underflow of the page_mapcount() won't be mistaken
444 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
445 * efficiently by most CPU architectures.
447 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
449 static inline int PageBuddy(struct page *page)
451 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
454 static inline void __SetPageBuddy(struct page *page)
456 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
457 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
460 static inline void __ClearPageBuddy(struct page *page)
462 VM_BUG_ON(!PageBuddy(page));
463 atomic_set(&page->_mapcount, -1);
466 void put_page(struct page *page);
467 void put_pages_list(struct list_head *pages);
469 void split_page(struct page *page, unsigned int order);
470 int split_free_page(struct page *page);
471 int capture_free_page(struct page *page, int alloc_order, int migratetype);
474 * Compound pages have a destructor function. Provide a
475 * prototype for that function and accessor functions.
476 * These are _only_ valid on the head of a PG_compound page.
478 typedef void compound_page_dtor(struct page *);
480 static inline void set_compound_page_dtor(struct page *page,
481 compound_page_dtor *dtor)
483 page[1].lru.next = (void *)dtor;
486 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
488 return (compound_page_dtor *)page[1].lru.next;
491 static inline int compound_order(struct page *page)
495 return (unsigned long)page[1].lru.prev;
498 static inline int compound_trans_order(struct page *page)
506 flags = compound_lock_irqsave(page);
507 order = compound_order(page);
508 compound_unlock_irqrestore(page, flags);
512 static inline void set_compound_order(struct page *page, unsigned long order)
514 page[1].lru.prev = (void *)order;
519 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
520 * servicing faults for write access. In the normal case, do always want
521 * pte_mkwrite. But get_user_pages can cause write faults for mappings
522 * that do not have writing enabled, when used by access_process_vm.
524 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
526 if (likely(vma->vm_flags & VM_WRITE))
527 pte = pte_mkwrite(pte);
533 * Multiple processes may "see" the same page. E.g. for untouched
534 * mappings of /dev/null, all processes see the same page full of
535 * zeroes, and text pages of executables and shared libraries have
536 * only one copy in memory, at most, normally.
538 * For the non-reserved pages, page_count(page) denotes a reference count.
539 * page_count() == 0 means the page is free. page->lru is then used for
540 * freelist management in the buddy allocator.
541 * page_count() > 0 means the page has been allocated.
543 * Pages are allocated by the slab allocator in order to provide memory
544 * to kmalloc and kmem_cache_alloc. In this case, the management of the
545 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
546 * unless a particular usage is carefully commented. (the responsibility of
547 * freeing the kmalloc memory is the caller's, of course).
549 * A page may be used by anyone else who does a __get_free_page().
550 * In this case, page_count still tracks the references, and should only
551 * be used through the normal accessor functions. The top bits of page->flags
552 * and page->virtual store page management information, but all other fields
553 * are unused and could be used privately, carefully. The management of this
554 * page is the responsibility of the one who allocated it, and those who have
555 * subsequently been given references to it.
557 * The other pages (we may call them "pagecache pages") are completely
558 * managed by the Linux memory manager: I/O, buffers, swapping etc.
559 * The following discussion applies only to them.
561 * A pagecache page contains an opaque `private' member, which belongs to the
562 * page's address_space. Usually, this is the address of a circular list of
563 * the page's disk buffers. PG_private must be set to tell the VM to call
564 * into the filesystem to release these pages.
566 * A page may belong to an inode's memory mapping. In this case, page->mapping
567 * is the pointer to the inode, and page->index is the file offset of the page,
568 * in units of PAGE_CACHE_SIZE.
570 * If pagecache pages are not associated with an inode, they are said to be
571 * anonymous pages. These may become associated with the swapcache, and in that
572 * case PG_swapcache is set, and page->private is an offset into the swapcache.
574 * In either case (swapcache or inode backed), the pagecache itself holds one
575 * reference to the page. Setting PG_private should also increment the
576 * refcount. The each user mapping also has a reference to the page.
578 * The pagecache pages are stored in a per-mapping radix tree, which is
579 * rooted at mapping->page_tree, and indexed by offset.
580 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
581 * lists, we instead now tag pages as dirty/writeback in the radix tree.
583 * All pagecache pages may be subject to I/O:
584 * - inode pages may need to be read from disk,
585 * - inode pages which have been modified and are MAP_SHARED may need
586 * to be written back to the inode on disk,
587 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
588 * modified may need to be swapped out to swap space and (later) to be read
593 * The zone field is never updated after free_area_init_core()
594 * sets it, so none of the operations on it need to be atomic.
597 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
598 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
599 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
600 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
601 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
604 * Define the bit shifts to access each section. For non-existent
605 * sections we define the shift as 0; that plus a 0 mask ensures
606 * the compiler will optimise away reference to them.
608 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
609 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
610 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
611 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
613 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
614 #ifdef NODE_NOT_IN_PAGE_FLAGS
615 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
616 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
617 SECTIONS_PGOFF : ZONES_PGOFF)
619 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
620 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
621 NODES_PGOFF : ZONES_PGOFF)
624 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
626 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
627 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
630 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
631 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
632 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
633 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
634 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
636 static inline enum zone_type page_zonenum(const struct page *page)
638 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
642 * The identification function is only used by the buddy allocator for
643 * determining if two pages could be buddies. We are not really
644 * identifying a zone since we could be using a the section number
645 * id if we have not node id available in page flags.
646 * We guarantee only that it will return the same value for two
647 * combinable pages in a zone.
649 static inline int page_zone_id(struct page *page)
651 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
654 static inline int zone_to_nid(struct zone *zone)
663 #ifdef NODE_NOT_IN_PAGE_FLAGS
664 extern int page_to_nid(const struct page *page);
666 static inline int page_to_nid(const struct page *page)
668 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
672 #ifdef CONFIG_SCHED_NUMA
673 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
674 static inline int page_xchg_last_nid(struct page *page, int nid)
676 return xchg(&page->_last_nid, nid);
679 static inline int page_last_nid(struct page *page)
681 return page->_last_nid;
684 static inline int page_xchg_last_nid(struct page *page, int nid)
686 unsigned long old_flags, flags;
690 old_flags = flags = page->flags;
691 last_nid = (flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
693 flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
694 flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
695 } while (unlikely(cmpxchg(&page->flags, old_flags, flags) != old_flags));
700 static inline int page_last_nid(struct page *page)
702 return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
704 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
705 #else /* CONFIG_SCHED_NUMA */
706 static inline int page_xchg_last_nid(struct page *page, int nid)
708 return page_to_nid(page);
711 static inline int page_last_nid(struct page *page)
713 return page_to_nid(page);
715 #endif /* CONFIG_SCHED_NUMA */
717 static inline struct zone *page_zone(const struct page *page)
719 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
722 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
723 static inline void set_page_section(struct page *page, unsigned long section)
725 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
726 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
729 static inline unsigned long page_to_section(const struct page *page)
731 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
735 static inline void set_page_zone(struct page *page, enum zone_type zone)
737 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
738 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
741 static inline void set_page_node(struct page *page, unsigned long node)
743 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
744 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
747 static inline void set_page_links(struct page *page, enum zone_type zone,
748 unsigned long node, unsigned long pfn)
750 set_page_zone(page, zone);
751 set_page_node(page, node);
752 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
753 set_page_section(page, pfn_to_section_nr(pfn));
758 * Some inline functions in vmstat.h depend on page_zone()
760 #include <linux/vmstat.h>
762 static __always_inline void *lowmem_page_address(const struct page *page)
764 return __va(PFN_PHYS(page_to_pfn(page)));
767 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
768 #define HASHED_PAGE_VIRTUAL
771 #if defined(WANT_PAGE_VIRTUAL)
772 #define page_address(page) ((page)->virtual)
773 #define set_page_address(page, address) \
775 (page)->virtual = (address); \
777 #define page_address_init() do { } while(0)
780 #if defined(HASHED_PAGE_VIRTUAL)
781 void *page_address(const struct page *page);
782 void set_page_address(struct page *page, void *virtual);
783 void page_address_init(void);
786 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
787 #define page_address(page) lowmem_page_address(page)
788 #define set_page_address(page, address) do { } while(0)
789 #define page_address_init() do { } while(0)
793 * On an anonymous page mapped into a user virtual memory area,
794 * page->mapping points to its anon_vma, not to a struct address_space;
795 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
797 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
798 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
799 * and then page->mapping points, not to an anon_vma, but to a private
800 * structure which KSM associates with that merged page. See ksm.h.
802 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
804 * Please note that, confusingly, "page_mapping" refers to the inode
805 * address_space which maps the page from disk; whereas "page_mapped"
806 * refers to user virtual address space into which the page is mapped.
808 #define PAGE_MAPPING_ANON 1
809 #define PAGE_MAPPING_KSM 2
810 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
812 extern struct address_space swapper_space;
813 static inline struct address_space *page_mapping(struct page *page)
815 struct address_space *mapping = page->mapping;
817 VM_BUG_ON(PageSlab(page));
818 if (unlikely(PageSwapCache(page)))
819 mapping = &swapper_space;
820 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
825 /* Neutral page->mapping pointer to address_space or anon_vma or other */
826 static inline void *page_rmapping(struct page *page)
828 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
831 extern struct address_space *__page_file_mapping(struct page *);
834 struct address_space *page_file_mapping(struct page *page)
836 if (unlikely(PageSwapCache(page)))
837 return __page_file_mapping(page);
839 return page->mapping;
842 static inline int PageAnon(struct page *page)
844 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
848 * Return the pagecache index of the passed page. Regular pagecache pages
849 * use ->index whereas swapcache pages use ->private
851 static inline pgoff_t page_index(struct page *page)
853 if (unlikely(PageSwapCache(page)))
854 return page_private(page);
858 extern pgoff_t __page_file_index(struct page *page);
861 * Return the file index of the page. Regular pagecache pages use ->index
862 * whereas swapcache pages use swp_offset(->private)
864 static inline pgoff_t page_file_index(struct page *page)
866 if (unlikely(PageSwapCache(page)))
867 return __page_file_index(page);
873 * Return true if this page is mapped into pagetables.
875 static inline int page_mapped(struct page *page)
877 return atomic_read(&(page)->_mapcount) >= 0;
881 * Different kinds of faults, as returned by handle_mm_fault().
882 * Used to decide whether a process gets delivered SIGBUS or
883 * just gets major/minor fault counters bumped up.
886 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
888 #define VM_FAULT_OOM 0x0001
889 #define VM_FAULT_SIGBUS 0x0002
890 #define VM_FAULT_MAJOR 0x0004
891 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
892 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
893 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
895 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
896 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
897 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
899 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
901 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
902 VM_FAULT_HWPOISON_LARGE)
904 /* Encode hstate index for a hwpoisoned large page */
905 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
906 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
909 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
911 extern void pagefault_out_of_memory(void);
913 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
916 * Flags passed to show_mem() and show_free_areas() to suppress output in
919 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
921 extern void show_free_areas(unsigned int flags);
922 extern bool skip_free_areas_node(unsigned int flags, int nid);
924 int shmem_zero_setup(struct vm_area_struct *);
926 extern int can_do_mlock(void);
927 extern int user_shm_lock(size_t, struct user_struct *);
928 extern void user_shm_unlock(size_t, struct user_struct *);
931 * Parameter block passed down to zap_pte_range in exceptional cases.
934 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
935 struct address_space *check_mapping; /* Check page->mapping if set */
936 pgoff_t first_index; /* Lowest page->index to unmap */
937 pgoff_t last_index; /* Highest page->index to unmap */
940 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
943 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
945 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
946 unsigned long size, struct zap_details *);
947 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
948 unsigned long start, unsigned long end);
951 * mm_walk - callbacks for walk_page_range
952 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
953 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
954 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
955 * this handler is required to be able to handle
956 * pmd_trans_huge() pmds. They may simply choose to
957 * split_huge_page() instead of handling it explicitly.
958 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
959 * @pte_hole: if set, called for each hole at all levels
960 * @hugetlb_entry: if set, called for each hugetlb entry
961 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
964 * (see walk_page_range for more details)
967 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
968 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
969 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
970 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
971 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
972 int (*hugetlb_entry)(pte_t *, unsigned long,
973 unsigned long, unsigned long, struct mm_walk *);
974 struct mm_struct *mm;
978 int walk_page_range(unsigned long addr, unsigned long end,
979 struct mm_walk *walk);
980 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
981 unsigned long end, unsigned long floor, unsigned long ceiling);
982 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
983 struct vm_area_struct *vma);
984 void unmap_mapping_range(struct address_space *mapping,
985 loff_t const holebegin, loff_t const holelen, int even_cows);
986 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
988 int follow_phys(struct vm_area_struct *vma, unsigned long address,
989 unsigned int flags, unsigned long *prot, resource_size_t *phys);
990 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
991 void *buf, int len, int write);
993 static inline void unmap_shared_mapping_range(struct address_space *mapping,
994 loff_t const holebegin, loff_t const holelen)
996 unmap_mapping_range(mapping, holebegin, holelen, 0);
999 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
1000 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1001 extern int vmtruncate(struct inode *inode, loff_t offset);
1002 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1003 int truncate_inode_page(struct address_space *mapping, struct page *page);
1004 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1005 int invalidate_inode_page(struct page *page);
1008 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1009 unsigned long address, unsigned int flags);
1010 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1011 unsigned long address, unsigned int fault_flags);
1013 static inline int handle_mm_fault(struct mm_struct *mm,
1014 struct vm_area_struct *vma, unsigned long address,
1017 /* should never happen if there's no MMU */
1019 return VM_FAULT_SIGBUS;
1021 static inline int fixup_user_fault(struct task_struct *tsk,
1022 struct mm_struct *mm, unsigned long address,
1023 unsigned int fault_flags)
1025 /* should never happen if there's no MMU */
1031 extern int make_pages_present(unsigned long addr, unsigned long end);
1032 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1033 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1034 void *buf, int len, int write);
1036 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1037 unsigned long start, int len, unsigned int foll_flags,
1038 struct page **pages, struct vm_area_struct **vmas,
1040 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1041 unsigned long start, int nr_pages, int write, int force,
1042 struct page **pages, struct vm_area_struct **vmas);
1043 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1044 struct page **pages);
1046 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1047 struct page **pages);
1048 int get_kernel_page(unsigned long start, int write, struct page **pages);
1049 struct page *get_dump_page(unsigned long addr);
1051 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1052 extern void do_invalidatepage(struct page *page, unsigned long offset);
1054 int __set_page_dirty_nobuffers(struct page *page);
1055 int __set_page_dirty_no_writeback(struct page *page);
1056 int redirty_page_for_writepage(struct writeback_control *wbc,
1058 void account_page_dirtied(struct page *page, struct address_space *mapping);
1059 void account_page_writeback(struct page *page);
1060 int set_page_dirty(struct page *page);
1061 int set_page_dirty_lock(struct page *page);
1062 int clear_page_dirty_for_io(struct page *page);
1064 /* Is the vma a continuation of the stack vma above it? */
1065 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1067 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1070 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1073 return (vma->vm_flags & VM_GROWSDOWN) &&
1074 (vma->vm_start == addr) &&
1075 !vma_growsdown(vma->vm_prev, addr);
1078 /* Is the vma a continuation of the stack vma below it? */
1079 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1081 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1084 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1087 return (vma->vm_flags & VM_GROWSUP) &&
1088 (vma->vm_end == addr) &&
1089 !vma_growsup(vma->vm_next, addr);
1093 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1095 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1096 unsigned long old_addr, struct vm_area_struct *new_vma,
1097 unsigned long new_addr, unsigned long len,
1098 bool need_rmap_locks);
1099 extern unsigned long do_mremap(unsigned long addr,
1100 unsigned long old_len, unsigned long new_len,
1101 unsigned long flags, unsigned long new_addr);
1102 extern void change_protection(struct vm_area_struct *vma, unsigned long start,
1103 unsigned long end, pgprot_t newprot,
1104 int dirty_accountable);
1105 extern int mprotect_fixup(struct vm_area_struct *vma,
1106 struct vm_area_struct **pprev, unsigned long start,
1107 unsigned long end, unsigned long newflags);
1110 * doesn't attempt to fault and will return short.
1112 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1113 struct page **pages);
1115 * per-process(per-mm_struct) statistics.
1117 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1119 long val = atomic_long_read(&mm->rss_stat.count[member]);
1121 #ifdef SPLIT_RSS_COUNTING
1123 * counter is updated in asynchronous manner and may go to minus.
1124 * But it's never be expected number for users.
1129 return (unsigned long)val;
1132 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1134 atomic_long_add(value, &mm->rss_stat.count[member]);
1137 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1139 atomic_long_inc(&mm->rss_stat.count[member]);
1142 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1144 atomic_long_dec(&mm->rss_stat.count[member]);
1147 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1149 return get_mm_counter(mm, MM_FILEPAGES) +
1150 get_mm_counter(mm, MM_ANONPAGES);
1153 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1155 return max(mm->hiwater_rss, get_mm_rss(mm));
1158 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1160 return max(mm->hiwater_vm, mm->total_vm);
1163 static inline void update_hiwater_rss(struct mm_struct *mm)
1165 unsigned long _rss = get_mm_rss(mm);
1167 if ((mm)->hiwater_rss < _rss)
1168 (mm)->hiwater_rss = _rss;
1171 static inline void update_hiwater_vm(struct mm_struct *mm)
1173 if (mm->hiwater_vm < mm->total_vm)
1174 mm->hiwater_vm = mm->total_vm;
1177 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1178 struct mm_struct *mm)
1180 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1182 if (*maxrss < hiwater_rss)
1183 *maxrss = hiwater_rss;
1186 #if defined(SPLIT_RSS_COUNTING)
1187 void sync_mm_rss(struct mm_struct *mm);
1189 static inline void sync_mm_rss(struct mm_struct *mm)
1194 int vma_wants_writenotify(struct vm_area_struct *vma);
1196 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1198 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1202 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1206 #ifdef __PAGETABLE_PUD_FOLDED
1207 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1208 unsigned long address)
1213 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1216 #ifdef __PAGETABLE_PMD_FOLDED
1217 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1218 unsigned long address)
1223 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1226 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1227 pmd_t *pmd, unsigned long address);
1228 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1231 * The following ifdef needed to get the 4level-fixup.h header to work.
1232 * Remove it when 4level-fixup.h has been removed.
1234 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1235 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1237 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1238 NULL: pud_offset(pgd, address);
1241 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1243 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1244 NULL: pmd_offset(pud, address);
1246 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1248 #if USE_SPLIT_PTLOCKS
1250 * We tuck a spinlock to guard each pagetable page into its struct page,
1251 * at page->private, with BUILD_BUG_ON to make sure that this will not
1252 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1253 * When freeing, reset page->mapping so free_pages_check won't complain.
1255 #define __pte_lockptr(page) &((page)->ptl)
1256 #define pte_lock_init(_page) do { \
1257 spin_lock_init(__pte_lockptr(_page)); \
1259 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1260 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1261 #else /* !USE_SPLIT_PTLOCKS */
1263 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1265 #define pte_lock_init(page) do {} while (0)
1266 #define pte_lock_deinit(page) do {} while (0)
1267 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1268 #endif /* USE_SPLIT_PTLOCKS */
1270 static inline void pgtable_page_ctor(struct page *page)
1272 pte_lock_init(page);
1273 inc_zone_page_state(page, NR_PAGETABLE);
1276 static inline void pgtable_page_dtor(struct page *page)
1278 pte_lock_deinit(page);
1279 dec_zone_page_state(page, NR_PAGETABLE);
1282 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1284 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1285 pte_t *__pte = pte_offset_map(pmd, address); \
1291 #define pte_unmap_unlock(pte, ptl) do { \
1296 #define pte_alloc_map(mm, vma, pmd, address) \
1297 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1299 NULL: pte_offset_map(pmd, address))
1301 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1302 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1304 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1306 #define pte_alloc_kernel(pmd, address) \
1307 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1308 NULL: pte_offset_kernel(pmd, address))
1310 extern void free_area_init(unsigned long * zones_size);
1311 extern void free_area_init_node(int nid, unsigned long * zones_size,
1312 unsigned long zone_start_pfn, unsigned long *zholes_size);
1313 extern void free_initmem(void);
1315 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1317 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1318 * zones, allocate the backing mem_map and account for memory holes in a more
1319 * architecture independent manner. This is a substitute for creating the
1320 * zone_sizes[] and zholes_size[] arrays and passing them to
1321 * free_area_init_node()
1323 * An architecture is expected to register range of page frames backed by
1324 * physical memory with memblock_add[_node]() before calling
1325 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1326 * usage, an architecture is expected to do something like
1328 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1330 * for_each_valid_physical_page_range()
1331 * memblock_add_node(base, size, nid)
1332 * free_area_init_nodes(max_zone_pfns);
1334 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1335 * registered physical page range. Similarly
1336 * sparse_memory_present_with_active_regions() calls memory_present() for
1337 * each range when SPARSEMEM is enabled.
1339 * See mm/page_alloc.c for more information on each function exposed by
1340 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1342 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1343 unsigned long node_map_pfn_alignment(void);
1344 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1345 unsigned long end_pfn);
1346 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1347 unsigned long end_pfn);
1348 extern void get_pfn_range_for_nid(unsigned int nid,
1349 unsigned long *start_pfn, unsigned long *end_pfn);
1350 extern unsigned long find_min_pfn_with_active_regions(void);
1351 extern void free_bootmem_with_active_regions(int nid,
1352 unsigned long max_low_pfn);
1353 extern void sparse_memory_present_with_active_regions(int nid);
1355 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1357 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1358 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1359 static inline int __early_pfn_to_nid(unsigned long pfn)
1364 /* please see mm/page_alloc.c */
1365 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1366 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1367 /* there is a per-arch backend function. */
1368 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1369 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1372 extern void set_dma_reserve(unsigned long new_dma_reserve);
1373 extern void memmap_init_zone(unsigned long, int, unsigned long,
1374 unsigned long, enum memmap_context);
1375 extern void setup_per_zone_wmarks(void);
1376 extern int __meminit init_per_zone_wmark_min(void);
1377 extern void mem_init(void);
1378 extern void __init mmap_init(void);
1379 extern void show_mem(unsigned int flags);
1380 extern void si_meminfo(struct sysinfo * val);
1381 extern void si_meminfo_node(struct sysinfo *val, int nid);
1382 extern int after_bootmem;
1384 extern __printf(3, 4)
1385 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1387 extern void setup_per_cpu_pageset(void);
1389 extern void zone_pcp_update(struct zone *zone);
1390 extern void zone_pcp_reset(struct zone *zone);
1393 extern atomic_long_t mmap_pages_allocated;
1394 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1396 /* interval_tree.c */
1397 void vma_interval_tree_insert(struct vm_area_struct *node,
1398 struct rb_root *root);
1399 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1400 struct vm_area_struct *prev,
1401 struct rb_root *root);
1402 void vma_interval_tree_remove(struct vm_area_struct *node,
1403 struct rb_root *root);
1404 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1405 unsigned long start, unsigned long last);
1406 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1407 unsigned long start, unsigned long last);
1409 #define vma_interval_tree_foreach(vma, root, start, last) \
1410 for (vma = vma_interval_tree_iter_first(root, start, last); \
1411 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1413 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1414 struct list_head *list)
1416 list_add_tail(&vma->shared.nonlinear, list);
1419 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1420 struct rb_root *root);
1421 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1422 struct rb_root *root);
1423 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1424 struct rb_root *root, unsigned long start, unsigned long last);
1425 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1426 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1427 #ifdef CONFIG_DEBUG_VM_RB
1428 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1431 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1432 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1433 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1436 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1437 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1438 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1439 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1440 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1441 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1442 struct mempolicy *);
1443 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1444 extern int split_vma(struct mm_struct *,
1445 struct vm_area_struct *, unsigned long addr, int new_below);
1446 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1447 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1448 struct rb_node **, struct rb_node *);
1449 extern void unlink_file_vma(struct vm_area_struct *);
1450 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1451 unsigned long addr, unsigned long len, pgoff_t pgoff,
1452 bool *need_rmap_locks);
1453 extern void exit_mmap(struct mm_struct *);
1455 extern int mm_take_all_locks(struct mm_struct *mm);
1456 extern void mm_drop_all_locks(struct mm_struct *mm);
1458 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1459 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1461 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1462 extern int install_special_mapping(struct mm_struct *mm,
1463 unsigned long addr, unsigned long len,
1464 unsigned long flags, struct page **pages);
1466 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1468 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1469 unsigned long len, unsigned long flags,
1470 vm_flags_t vm_flags, unsigned long pgoff);
1471 extern unsigned long do_mmap_pgoff(struct file *, unsigned long,
1472 unsigned long, unsigned long,
1473 unsigned long, unsigned long);
1474 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1476 /* These take the mm semaphore themselves */
1477 extern unsigned long vm_brk(unsigned long, unsigned long);
1478 extern int vm_munmap(unsigned long, size_t);
1479 extern unsigned long vm_mmap(struct file *, unsigned long,
1480 unsigned long, unsigned long,
1481 unsigned long, unsigned long);
1484 extern void truncate_inode_pages(struct address_space *, loff_t);
1485 extern void truncate_inode_pages_range(struct address_space *,
1486 loff_t lstart, loff_t lend);
1488 /* generic vm_area_ops exported for stackable file systems */
1489 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1490 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1492 /* mm/page-writeback.c */
1493 int write_one_page(struct page *page, int wait);
1494 void task_dirty_inc(struct task_struct *tsk);
1497 #define VM_MAX_READAHEAD 128 /* kbytes */
1498 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1500 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1501 pgoff_t offset, unsigned long nr_to_read);
1503 void page_cache_sync_readahead(struct address_space *mapping,
1504 struct file_ra_state *ra,
1507 unsigned long size);
1509 void page_cache_async_readahead(struct address_space *mapping,
1510 struct file_ra_state *ra,
1514 unsigned long size);
1516 unsigned long max_sane_readahead(unsigned long nr);
1517 unsigned long ra_submit(struct file_ra_state *ra,
1518 struct address_space *mapping,
1521 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1522 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1524 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1525 extern int expand_downwards(struct vm_area_struct *vma,
1526 unsigned long address);
1528 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1530 #define expand_upwards(vma, address) do { } while (0)
1533 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1534 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1535 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1536 struct vm_area_struct **pprev);
1538 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1539 NULL if none. Assume start_addr < end_addr. */
1540 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1542 struct vm_area_struct * vma = find_vma(mm,start_addr);
1544 if (vma && end_addr <= vma->vm_start)
1549 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1551 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1554 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1555 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1556 unsigned long vm_start, unsigned long vm_end)
1558 struct vm_area_struct *vma = find_vma(mm, vm_start);
1560 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1567 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1569 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1575 static inline pgprot_t vma_prot_none(struct vm_area_struct *vma)
1578 * obtain PROT_NONE by removing READ|WRITE|EXEC privs
1580 vm_flags_t vmflags = vma->vm_flags & ~(VM_READ|VM_WRITE|VM_EXEC);
1581 return pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vmflags));
1584 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1585 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1586 unsigned long pfn, unsigned long size, pgprot_t);
1587 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1588 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1590 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1593 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1594 unsigned int foll_flags);
1595 #define FOLL_WRITE 0x01 /* check pte is writable */
1596 #define FOLL_TOUCH 0x02 /* mark page accessed */
1597 #define FOLL_GET 0x04 /* do get_page on page */
1598 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1599 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1600 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1601 * and return without waiting upon it */
1602 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1603 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1604 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1606 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1608 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1609 unsigned long size, pte_fn_t fn, void *data);
1611 #ifdef CONFIG_PROC_FS
1612 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1614 static inline void vm_stat_account(struct mm_struct *mm,
1615 unsigned long flags, struct file *file, long pages)
1617 mm->total_vm += pages;
1619 #endif /* CONFIG_PROC_FS */
1621 #ifdef CONFIG_DEBUG_PAGEALLOC
1622 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1623 #ifdef CONFIG_HIBERNATION
1624 extern bool kernel_page_present(struct page *page);
1625 #endif /* CONFIG_HIBERNATION */
1628 kernel_map_pages(struct page *page, int numpages, int enable) {}
1629 #ifdef CONFIG_HIBERNATION
1630 static inline bool kernel_page_present(struct page *page) { return true; }
1631 #endif /* CONFIG_HIBERNATION */
1634 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1635 #ifdef __HAVE_ARCH_GATE_AREA
1636 int in_gate_area_no_mm(unsigned long addr);
1637 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1639 int in_gate_area_no_mm(unsigned long addr);
1640 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1641 #endif /* __HAVE_ARCH_GATE_AREA */
1643 int drop_caches_sysctl_handler(struct ctl_table *, int,
1644 void __user *, size_t *, loff_t *);
1645 unsigned long shrink_slab(struct shrink_control *shrink,
1646 unsigned long nr_pages_scanned,
1647 unsigned long lru_pages);
1650 #define randomize_va_space 0
1652 extern int randomize_va_space;
1655 const char * arch_vma_name(struct vm_area_struct *vma);
1656 void print_vma_addr(char *prefix, unsigned long rip);
1658 void sparse_mem_maps_populate_node(struct page **map_map,
1659 unsigned long pnum_begin,
1660 unsigned long pnum_end,
1661 unsigned long map_count,
1664 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1665 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1666 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1667 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1668 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1669 void *vmemmap_alloc_block(unsigned long size, int node);
1670 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1671 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1672 int vmemmap_populate_basepages(struct page *start_page,
1673 unsigned long pages, int node);
1674 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1675 void vmemmap_populate_print_last(void);
1679 MF_COUNT_INCREASED = 1 << 0,
1680 MF_ACTION_REQUIRED = 1 << 1,
1681 MF_MUST_KILL = 1 << 2,
1683 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1684 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1685 extern int unpoison_memory(unsigned long pfn);
1686 extern int sysctl_memory_failure_early_kill;
1687 extern int sysctl_memory_failure_recovery;
1688 extern void shake_page(struct page *p, int access);
1689 extern atomic_long_t mce_bad_pages;
1690 extern int soft_offline_page(struct page *page, int flags);
1692 extern void dump_page(struct page *page);
1694 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1695 extern void clear_huge_page(struct page *page,
1697 unsigned int pages_per_huge_page);
1698 extern void copy_user_huge_page(struct page *dst, struct page *src,
1699 unsigned long addr, struct vm_area_struct *vma,
1700 unsigned int pages_per_huge_page);
1701 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1703 #ifdef CONFIG_DEBUG_PAGEALLOC
1704 extern unsigned int _debug_guardpage_minorder;
1706 static inline unsigned int debug_guardpage_minorder(void)
1708 return _debug_guardpage_minorder;
1711 static inline bool page_is_guard(struct page *page)
1713 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1716 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1717 static inline bool page_is_guard(struct page *page) { return false; }
1718 #endif /* CONFIG_DEBUG_PAGEALLOC */
1720 #endif /* __KERNEL__ */
1721 #endif /* _LINUX_MM_H */