4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/prio_tree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
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_READ 0x00000001 /* currently active flags */
74 #define VM_WRITE 0x00000002
75 #define VM_EXEC 0x00000004
76 #define VM_SHARED 0x00000008
78 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
79 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
80 #define VM_MAYWRITE 0x00000020
81 #define VM_MAYEXEC 0x00000040
82 #define VM_MAYSHARE 0x00000080
84 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
85 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
86 #define VM_GROWSUP 0x00000200
88 #define VM_GROWSUP 0x00000000
89 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
91 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
92 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
94 #define VM_EXECUTABLE 0x00001000
95 #define VM_LOCKED 0x00002000
96 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
98 /* Used by sys_madvise() */
99 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
100 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
102 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
103 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
104 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
105 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
106 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
107 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
108 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
109 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
110 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
112 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
114 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
115 #define VM_NODUMP 0x04000000 /* Do not include in the core dump */
117 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
118 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
119 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
120 #define VM_PAT 0x40000000 /* PAT reserves whole VMA at once (x86) */
121 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
123 /* Bits set in the VMA until the stack is in its final location */
124 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
126 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
127 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
130 #ifdef CONFIG_STACK_GROWSUP
131 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
133 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
136 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
137 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
138 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
139 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
140 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
143 * Special vmas that are non-mergable, non-mlock()able.
144 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
146 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
149 * mapping from the currently active vm_flags protection bits (the
150 * low four bits) to a page protection mask..
152 extern pgprot_t protection_map[16];
154 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
155 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
156 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
157 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
158 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
159 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
160 #define FAULT_FLAG_TRIED 0x40 /* second try */
163 * Some architectures (such as x86) may need to preserve certain pgprot
164 * bits, without complicating generic pgprot code.
166 * Most architectures don't care:
168 #ifndef pgprot_modify
169 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
176 * vm_fault is filled by the the pagefault handler and passed to the vma's
177 * ->fault function. The vma's ->fault is responsible for returning a bitmask
178 * of VM_FAULT_xxx flags that give details about how the fault was handled.
180 * pgoff should be used in favour of virtual_address, if possible. If pgoff
181 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
185 unsigned int flags; /* FAULT_FLAG_xxx flags */
186 pgoff_t pgoff; /* Logical page offset based on vma */
187 void __user *virtual_address; /* Faulting virtual address */
189 struct page *page; /* ->fault handlers should return a
190 * page here, unless VM_FAULT_NOPAGE
191 * is set (which is also implied by
197 * These are the virtual MM functions - opening of an area, closing and
198 * unmapping it (needed to keep files on disk up-to-date etc), pointer
199 * to the functions called when a no-page or a wp-page exception occurs.
201 struct vm_operations_struct {
202 void (*open)(struct vm_area_struct * area);
203 void (*close)(struct vm_area_struct * area);
204 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
206 /* notification that a previously read-only page is about to become
207 * writable, if an error is returned it will cause a SIGBUS */
208 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
210 /* called by access_process_vm when get_user_pages() fails, typically
211 * for use by special VMAs that can switch between memory and hardware
213 int (*access)(struct vm_area_struct *vma, unsigned long addr,
214 void *buf, int len, int write);
217 * set_policy() op must add a reference to any non-NULL @new mempolicy
218 * to hold the policy upon return. Caller should pass NULL @new to
219 * remove a policy and fall back to surrounding context--i.e. do not
220 * install a MPOL_DEFAULT policy, nor the task or system default
223 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
226 * get_policy() op must add reference [mpol_get()] to any policy at
227 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
228 * in mm/mempolicy.c will do this automatically.
229 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
230 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
231 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
232 * must return NULL--i.e., do not "fallback" to task or system default
235 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
237 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
238 const nodemask_t *to, unsigned long flags);
245 #define page_private(page) ((page)->private)
246 #define set_page_private(page, v) ((page)->private = (v))
249 * FIXME: take this include out, include page-flags.h in
250 * files which need it (119 of them)
252 #include <linux/page-flags.h>
253 #include <linux/huge_mm.h>
256 * Methods to modify the page usage count.
258 * What counts for a page usage:
259 * - cache mapping (page->mapping)
260 * - private data (page->private)
261 * - page mapped in a task's page tables, each mapping
262 * is counted separately
264 * Also, many kernel routines increase the page count before a critical
265 * routine so they can be sure the page doesn't go away from under them.
269 * Drop a ref, return true if the refcount fell to zero (the page has no users)
271 static inline int put_page_testzero(struct page *page)
273 VM_BUG_ON(atomic_read(&page->_count) == 0);
274 return atomic_dec_and_test(&page->_count);
278 * Try to grab a ref unless the page has a refcount of zero, return false if
281 static inline int get_page_unless_zero(struct page *page)
283 return atomic_inc_not_zero(&page->_count);
286 extern int page_is_ram(unsigned long pfn);
288 /* Support for virtually mapped pages */
289 struct page *vmalloc_to_page(const void *addr);
290 unsigned long vmalloc_to_pfn(const void *addr);
293 * Determine if an address is within the vmalloc range
295 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
296 * is no special casing required.
298 static inline int is_vmalloc_addr(const void *x)
301 unsigned long addr = (unsigned long)x;
303 return addr >= VMALLOC_START && addr < VMALLOC_END;
309 extern int is_vmalloc_or_module_addr(const void *x);
311 static inline int is_vmalloc_or_module_addr(const void *x)
317 static inline void compound_lock(struct page *page)
319 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
320 VM_BUG_ON(PageSlab(page));
321 bit_spin_lock(PG_compound_lock, &page->flags);
325 static inline void compound_unlock(struct page *page)
327 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
328 VM_BUG_ON(PageSlab(page));
329 bit_spin_unlock(PG_compound_lock, &page->flags);
333 static inline unsigned long compound_lock_irqsave(struct page *page)
335 unsigned long uninitialized_var(flags);
336 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
337 local_irq_save(flags);
343 static inline void compound_unlock_irqrestore(struct page *page,
346 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
347 compound_unlock(page);
348 local_irq_restore(flags);
352 static inline struct page *compound_head(struct page *page)
354 if (unlikely(PageTail(page)))
355 return page->first_page;
360 * The atomic page->_mapcount, starts from -1: so that transitions
361 * both from it and to it can be tracked, using atomic_inc_and_test
362 * and atomic_add_negative(-1).
364 static inline void reset_page_mapcount(struct page *page)
366 atomic_set(&(page)->_mapcount, -1);
369 static inline int page_mapcount(struct page *page)
371 return atomic_read(&(page)->_mapcount) + 1;
374 static inline int page_count(struct page *page)
376 return atomic_read(&compound_head(page)->_count);
379 static inline void get_huge_page_tail(struct page *page)
382 * __split_huge_page_refcount() cannot run
385 VM_BUG_ON(page_mapcount(page) < 0);
386 VM_BUG_ON(atomic_read(&page->_count) != 0);
387 atomic_inc(&page->_mapcount);
390 extern bool __get_page_tail(struct page *page);
392 static inline void get_page(struct page *page)
394 if (unlikely(PageTail(page)))
395 if (likely(__get_page_tail(page)))
398 * Getting a normal page or the head of a compound page
399 * requires to already have an elevated page->_count.
401 VM_BUG_ON(atomic_read(&page->_count) <= 0);
402 atomic_inc(&page->_count);
405 static inline struct page *virt_to_head_page(const void *x)
407 struct page *page = virt_to_page(x);
408 return compound_head(page);
412 * Setup the page count before being freed into the page allocator for
413 * the first time (boot or memory hotplug)
415 static inline void init_page_count(struct page *page)
417 atomic_set(&page->_count, 1);
421 * PageBuddy() indicate that the page is free and in the buddy system
422 * (see mm/page_alloc.c).
424 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
425 * -2 so that an underflow of the page_mapcount() won't be mistaken
426 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
427 * efficiently by most CPU architectures.
429 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
431 static inline int PageBuddy(struct page *page)
433 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
436 static inline void __SetPageBuddy(struct page *page)
438 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
439 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
442 static inline void __ClearPageBuddy(struct page *page)
444 VM_BUG_ON(!PageBuddy(page));
445 atomic_set(&page->_mapcount, -1);
448 void put_page(struct page *page);
449 void put_pages_list(struct list_head *pages);
451 void split_page(struct page *page, unsigned int order);
452 int split_free_page(struct page *page);
455 * Compound pages have a destructor function. Provide a
456 * prototype for that function and accessor functions.
457 * These are _only_ valid on the head of a PG_compound page.
459 typedef void compound_page_dtor(struct page *);
461 static inline void set_compound_page_dtor(struct page *page,
462 compound_page_dtor *dtor)
464 page[1].lru.next = (void *)dtor;
467 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
469 return (compound_page_dtor *)page[1].lru.next;
472 static inline int compound_order(struct page *page)
476 return (unsigned long)page[1].lru.prev;
479 static inline int compound_trans_order(struct page *page)
487 flags = compound_lock_irqsave(page);
488 order = compound_order(page);
489 compound_unlock_irqrestore(page, flags);
493 static inline void set_compound_order(struct page *page, unsigned long order)
495 page[1].lru.prev = (void *)order;
500 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
501 * servicing faults for write access. In the normal case, do always want
502 * pte_mkwrite. But get_user_pages can cause write faults for mappings
503 * that do not have writing enabled, when used by access_process_vm.
505 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
507 if (likely(vma->vm_flags & VM_WRITE))
508 pte = pte_mkwrite(pte);
514 * Multiple processes may "see" the same page. E.g. for untouched
515 * mappings of /dev/null, all processes see the same page full of
516 * zeroes, and text pages of executables and shared libraries have
517 * only one copy in memory, at most, normally.
519 * For the non-reserved pages, page_count(page) denotes a reference count.
520 * page_count() == 0 means the page is free. page->lru is then used for
521 * freelist management in the buddy allocator.
522 * page_count() > 0 means the page has been allocated.
524 * Pages are allocated by the slab allocator in order to provide memory
525 * to kmalloc and kmem_cache_alloc. In this case, the management of the
526 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
527 * unless a particular usage is carefully commented. (the responsibility of
528 * freeing the kmalloc memory is the caller's, of course).
530 * A page may be used by anyone else who does a __get_free_page().
531 * In this case, page_count still tracks the references, and should only
532 * be used through the normal accessor functions. The top bits of page->flags
533 * and page->virtual store page management information, but all other fields
534 * are unused and could be used privately, carefully. The management of this
535 * page is the responsibility of the one who allocated it, and those who have
536 * subsequently been given references to it.
538 * The other pages (we may call them "pagecache pages") are completely
539 * managed by the Linux memory manager: I/O, buffers, swapping etc.
540 * The following discussion applies only to them.
542 * A pagecache page contains an opaque `private' member, which belongs to the
543 * page's address_space. Usually, this is the address of a circular list of
544 * the page's disk buffers. PG_private must be set to tell the VM to call
545 * into the filesystem to release these pages.
547 * A page may belong to an inode's memory mapping. In this case, page->mapping
548 * is the pointer to the inode, and page->index is the file offset of the page,
549 * in units of PAGE_CACHE_SIZE.
551 * If pagecache pages are not associated with an inode, they are said to be
552 * anonymous pages. These may become associated with the swapcache, and in that
553 * case PG_swapcache is set, and page->private is an offset into the swapcache.
555 * In either case (swapcache or inode backed), the pagecache itself holds one
556 * reference to the page. Setting PG_private should also increment the
557 * refcount. The each user mapping also has a reference to the page.
559 * The pagecache pages are stored in a per-mapping radix tree, which is
560 * rooted at mapping->page_tree, and indexed by offset.
561 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
562 * lists, we instead now tag pages as dirty/writeback in the radix tree.
564 * All pagecache pages may be subject to I/O:
565 * - inode pages may need to be read from disk,
566 * - inode pages which have been modified and are MAP_SHARED may need
567 * to be written back to the inode on disk,
568 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
569 * modified may need to be swapped out to swap space and (later) to be read
574 * The zone field is never updated after free_area_init_core()
575 * sets it, so none of the operations on it need to be atomic.
578 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
579 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
580 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
581 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
582 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
585 * Define the bit shifts to access each section. For non-existent
586 * sections we define the shift as 0; that plus a 0 mask ensures
587 * the compiler will optimise away reference to them.
589 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
590 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
591 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
592 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
594 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
595 #ifdef NODE_NOT_IN_PAGE_FLAGS
596 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
597 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
598 SECTIONS_PGOFF : ZONES_PGOFF)
600 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
601 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
602 NODES_PGOFF : ZONES_PGOFF)
605 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
607 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
608 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
611 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
612 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
613 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
614 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
615 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
617 static inline enum zone_type page_zonenum(const struct page *page)
619 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
623 * The identification function is only used by the buddy allocator for
624 * determining if two pages could be buddies. We are not really
625 * identifying a zone since we could be using a the section number
626 * id if we have not node id available in page flags.
627 * We guarantee only that it will return the same value for two
628 * combinable pages in a zone.
630 static inline int page_zone_id(struct page *page)
632 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
635 static inline int zone_to_nid(struct zone *zone)
644 #ifdef NODE_NOT_IN_PAGE_FLAGS
645 extern int page_to_nid(const struct page *page);
647 static inline int page_to_nid(const struct page *page)
649 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
653 #ifdef CONFIG_SCHED_NUMA
654 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
655 static inline int page_xchg_last_nid(struct page *page, int nid)
657 return xchg(&page->_last_nid, nid);
660 static inline int page_last_nid(struct page *page)
662 return page->_last_nid;
665 static inline int page_xchg_last_nid(struct page *page, int nid)
667 unsigned long old_flags, flags;
671 old_flags = flags = page->flags;
672 last_nid = (flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
674 flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
675 flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
676 } while (unlikely(cmpxchg(&page->flags, old_flags, flags) != old_flags));
681 static inline int page_last_nid(struct page *page)
683 return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
685 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
686 #else /* CONFIG_SCHED_NUMA */
687 static inline int page_xchg_last_nid(struct page *page, int nid)
689 return page_to_nid(page);
692 static inline int page_last_nid(struct page *page)
694 return page_to_nid(page);
696 #endif /* CONFIG_SCHED_NUMA */
698 static inline struct zone *page_zone(const struct page *page)
700 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
703 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
704 static inline void set_page_section(struct page *page, unsigned long section)
706 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
707 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
710 static inline unsigned long page_to_section(const struct page *page)
712 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
716 static inline void set_page_zone(struct page *page, enum zone_type zone)
718 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
719 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
722 static inline void set_page_node(struct page *page, unsigned long node)
724 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
725 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
728 static inline void set_page_links(struct page *page, enum zone_type zone,
729 unsigned long node, unsigned long pfn)
731 set_page_zone(page, zone);
732 set_page_node(page, node);
733 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
734 set_page_section(page, pfn_to_section_nr(pfn));
739 * Some inline functions in vmstat.h depend on page_zone()
741 #include <linux/vmstat.h>
743 static __always_inline void *lowmem_page_address(const struct page *page)
745 return __va(PFN_PHYS(page_to_pfn(page)));
748 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
749 #define HASHED_PAGE_VIRTUAL
752 #if defined(WANT_PAGE_VIRTUAL)
753 #define page_address(page) ((page)->virtual)
754 #define set_page_address(page, address) \
756 (page)->virtual = (address); \
758 #define page_address_init() do { } while(0)
761 #if defined(HASHED_PAGE_VIRTUAL)
762 void *page_address(const struct page *page);
763 void set_page_address(struct page *page, void *virtual);
764 void page_address_init(void);
767 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
768 #define page_address(page) lowmem_page_address(page)
769 #define set_page_address(page, address) do { } while(0)
770 #define page_address_init() do { } while(0)
774 * On an anonymous page mapped into a user virtual memory area,
775 * page->mapping points to its anon_vma, not to a struct address_space;
776 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
778 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
779 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
780 * and then page->mapping points, not to an anon_vma, but to a private
781 * structure which KSM associates with that merged page. See ksm.h.
783 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
785 * Please note that, confusingly, "page_mapping" refers to the inode
786 * address_space which maps the page from disk; whereas "page_mapped"
787 * refers to user virtual address space into which the page is mapped.
789 #define PAGE_MAPPING_ANON 1
790 #define PAGE_MAPPING_KSM 2
791 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
793 extern struct address_space swapper_space;
794 static inline struct address_space *page_mapping(struct page *page)
796 struct address_space *mapping = page->mapping;
798 VM_BUG_ON(PageSlab(page));
799 if (unlikely(PageSwapCache(page)))
800 mapping = &swapper_space;
801 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
806 /* Neutral page->mapping pointer to address_space or anon_vma or other */
807 static inline void *page_rmapping(struct page *page)
809 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
812 extern struct address_space *__page_file_mapping(struct page *);
815 struct address_space *page_file_mapping(struct page *page)
817 if (unlikely(PageSwapCache(page)))
818 return __page_file_mapping(page);
820 return page->mapping;
823 static inline int PageAnon(struct page *page)
825 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
829 * Return the pagecache index of the passed page. Regular pagecache pages
830 * use ->index whereas swapcache pages use ->private
832 static inline pgoff_t page_index(struct page *page)
834 if (unlikely(PageSwapCache(page)))
835 return page_private(page);
839 extern pgoff_t __page_file_index(struct page *page);
842 * Return the file index of the page. Regular pagecache pages use ->index
843 * whereas swapcache pages use swp_offset(->private)
845 static inline pgoff_t page_file_index(struct page *page)
847 if (unlikely(PageSwapCache(page)))
848 return __page_file_index(page);
854 * Return true if this page is mapped into pagetables.
856 static inline int page_mapped(struct page *page)
858 return atomic_read(&(page)->_mapcount) >= 0;
862 * Different kinds of faults, as returned by handle_mm_fault().
863 * Used to decide whether a process gets delivered SIGBUS or
864 * just gets major/minor fault counters bumped up.
867 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
869 #define VM_FAULT_OOM 0x0001
870 #define VM_FAULT_SIGBUS 0x0002
871 #define VM_FAULT_MAJOR 0x0004
872 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
873 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
874 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
876 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
877 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
878 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
880 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
882 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
883 VM_FAULT_HWPOISON_LARGE)
885 /* Encode hstate index for a hwpoisoned large page */
886 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
887 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
890 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
892 extern void pagefault_out_of_memory(void);
894 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
897 * Flags passed to show_mem() and show_free_areas() to suppress output in
900 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
902 extern void show_free_areas(unsigned int flags);
903 extern bool skip_free_areas_node(unsigned int flags, int nid);
905 int shmem_zero_setup(struct vm_area_struct *);
907 extern int can_do_mlock(void);
908 extern int user_shm_lock(size_t, struct user_struct *);
909 extern void user_shm_unlock(size_t, struct user_struct *);
912 * Parameter block passed down to zap_pte_range in exceptional cases.
915 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
916 struct address_space *check_mapping; /* Check page->mapping if set */
917 pgoff_t first_index; /* Lowest page->index to unmap */
918 pgoff_t last_index; /* Highest page->index to unmap */
921 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
924 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
926 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
927 unsigned long size, struct zap_details *);
928 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
929 unsigned long start, unsigned long end);
932 * mm_walk - callbacks for walk_page_range
933 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
934 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
935 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
936 * this handler is required to be able to handle
937 * pmd_trans_huge() pmds. They may simply choose to
938 * split_huge_page() instead of handling it explicitly.
939 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
940 * @pte_hole: if set, called for each hole at all levels
941 * @hugetlb_entry: if set, called for each hugetlb entry
942 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
945 * (see walk_page_range for more details)
948 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
949 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
950 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
951 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
952 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
953 int (*hugetlb_entry)(pte_t *, unsigned long,
954 unsigned long, unsigned long, struct mm_walk *);
955 struct mm_struct *mm;
959 int walk_page_range(unsigned long addr, unsigned long end,
960 struct mm_walk *walk);
961 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
962 unsigned long end, unsigned long floor, unsigned long ceiling);
963 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
964 struct vm_area_struct *vma);
965 void unmap_mapping_range(struct address_space *mapping,
966 loff_t const holebegin, loff_t const holelen, int even_cows);
967 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
969 int follow_phys(struct vm_area_struct *vma, unsigned long address,
970 unsigned int flags, unsigned long *prot, resource_size_t *phys);
971 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
972 void *buf, int len, int write);
974 static inline void unmap_shared_mapping_range(struct address_space *mapping,
975 loff_t const holebegin, loff_t const holelen)
977 unmap_mapping_range(mapping, holebegin, holelen, 0);
980 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
981 extern void truncate_setsize(struct inode *inode, loff_t newsize);
982 extern int vmtruncate(struct inode *inode, loff_t offset);
983 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
984 int truncate_inode_page(struct address_space *mapping, struct page *page);
985 int generic_error_remove_page(struct address_space *mapping, struct page *page);
986 int invalidate_inode_page(struct page *page);
989 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
990 unsigned long address, unsigned int flags);
991 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
992 unsigned long address, unsigned int fault_flags);
994 static inline int handle_mm_fault(struct mm_struct *mm,
995 struct vm_area_struct *vma, unsigned long address,
998 /* should never happen if there's no MMU */
1000 return VM_FAULT_SIGBUS;
1002 static inline int fixup_user_fault(struct task_struct *tsk,
1003 struct mm_struct *mm, unsigned long address,
1004 unsigned int fault_flags)
1006 /* should never happen if there's no MMU */
1012 extern int make_pages_present(unsigned long addr, unsigned long end);
1013 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1014 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1015 void *buf, int len, int write);
1017 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1018 unsigned long start, int len, unsigned int foll_flags,
1019 struct page **pages, struct vm_area_struct **vmas,
1021 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1022 unsigned long start, int nr_pages, int write, int force,
1023 struct page **pages, struct vm_area_struct **vmas);
1024 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1025 struct page **pages);
1027 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1028 struct page **pages);
1029 int get_kernel_page(unsigned long start, int write, struct page **pages);
1030 struct page *get_dump_page(unsigned long addr);
1032 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1033 extern void do_invalidatepage(struct page *page, unsigned long offset);
1035 int __set_page_dirty_nobuffers(struct page *page);
1036 int __set_page_dirty_no_writeback(struct page *page);
1037 int redirty_page_for_writepage(struct writeback_control *wbc,
1039 void account_page_dirtied(struct page *page, struct address_space *mapping);
1040 void account_page_writeback(struct page *page);
1041 int set_page_dirty(struct page *page);
1042 int set_page_dirty_lock(struct page *page);
1043 int clear_page_dirty_for_io(struct page *page);
1045 /* Is the vma a continuation of the stack vma above it? */
1046 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1048 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1051 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1054 return (vma->vm_flags & VM_GROWSDOWN) &&
1055 (vma->vm_start == addr) &&
1056 !vma_growsdown(vma->vm_prev, addr);
1059 /* Is the vma a continuation of the stack vma below it? */
1060 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1062 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1065 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1068 return (vma->vm_flags & VM_GROWSUP) &&
1069 (vma->vm_end == addr) &&
1070 !vma_growsup(vma->vm_next, addr);
1074 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1076 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1077 unsigned long old_addr, struct vm_area_struct *new_vma,
1078 unsigned long new_addr, unsigned long len);
1079 extern unsigned long do_mremap(unsigned long addr,
1080 unsigned long old_len, unsigned long new_len,
1081 unsigned long flags, unsigned long new_addr);
1082 extern void change_protection(struct vm_area_struct *vma, unsigned long start,
1083 unsigned long end, pgprot_t newprot,
1084 int dirty_accountable);
1085 extern int mprotect_fixup(struct vm_area_struct *vma,
1086 struct vm_area_struct **pprev, unsigned long start,
1087 unsigned long end, unsigned long newflags);
1090 * doesn't attempt to fault and will return short.
1092 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1093 struct page **pages);
1095 * per-process(per-mm_struct) statistics.
1097 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1099 long val = atomic_long_read(&mm->rss_stat.count[member]);
1101 #ifdef SPLIT_RSS_COUNTING
1103 * counter is updated in asynchronous manner and may go to minus.
1104 * But it's never be expected number for users.
1109 return (unsigned long)val;
1112 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1114 atomic_long_add(value, &mm->rss_stat.count[member]);
1117 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1119 atomic_long_inc(&mm->rss_stat.count[member]);
1122 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1124 atomic_long_dec(&mm->rss_stat.count[member]);
1127 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1129 return get_mm_counter(mm, MM_FILEPAGES) +
1130 get_mm_counter(mm, MM_ANONPAGES);
1133 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1135 return max(mm->hiwater_rss, get_mm_rss(mm));
1138 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1140 return max(mm->hiwater_vm, mm->total_vm);
1143 static inline void update_hiwater_rss(struct mm_struct *mm)
1145 unsigned long _rss = get_mm_rss(mm);
1147 if ((mm)->hiwater_rss < _rss)
1148 (mm)->hiwater_rss = _rss;
1151 static inline void update_hiwater_vm(struct mm_struct *mm)
1153 if (mm->hiwater_vm < mm->total_vm)
1154 mm->hiwater_vm = mm->total_vm;
1157 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1158 struct mm_struct *mm)
1160 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1162 if (*maxrss < hiwater_rss)
1163 *maxrss = hiwater_rss;
1166 #if defined(SPLIT_RSS_COUNTING)
1167 void sync_mm_rss(struct mm_struct *mm);
1169 static inline void sync_mm_rss(struct mm_struct *mm)
1174 int vma_wants_writenotify(struct vm_area_struct *vma);
1176 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1178 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1182 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1186 #ifdef __PAGETABLE_PUD_FOLDED
1187 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1188 unsigned long address)
1193 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1196 #ifdef __PAGETABLE_PMD_FOLDED
1197 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1198 unsigned long address)
1203 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1206 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1207 pmd_t *pmd, unsigned long address);
1208 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1211 * The following ifdef needed to get the 4level-fixup.h header to work.
1212 * Remove it when 4level-fixup.h has been removed.
1214 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1215 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1217 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1218 NULL: pud_offset(pgd, address);
1221 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1223 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1224 NULL: pmd_offset(pud, address);
1226 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1228 #if USE_SPLIT_PTLOCKS
1230 * We tuck a spinlock to guard each pagetable page into its struct page,
1231 * at page->private, with BUILD_BUG_ON to make sure that this will not
1232 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1233 * When freeing, reset page->mapping so free_pages_check won't complain.
1235 #define __pte_lockptr(page) &((page)->ptl)
1236 #define pte_lock_init(_page) do { \
1237 spin_lock_init(__pte_lockptr(_page)); \
1239 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1240 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1241 #else /* !USE_SPLIT_PTLOCKS */
1243 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1245 #define pte_lock_init(page) do {} while (0)
1246 #define pte_lock_deinit(page) do {} while (0)
1247 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1248 #endif /* USE_SPLIT_PTLOCKS */
1250 static inline void pgtable_page_ctor(struct page *page)
1252 pte_lock_init(page);
1253 inc_zone_page_state(page, NR_PAGETABLE);
1256 static inline void pgtable_page_dtor(struct page *page)
1258 pte_lock_deinit(page);
1259 dec_zone_page_state(page, NR_PAGETABLE);
1262 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1264 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1265 pte_t *__pte = pte_offset_map(pmd, address); \
1271 #define pte_unmap_unlock(pte, ptl) do { \
1276 #define pte_alloc_map(mm, vma, pmd, address) \
1277 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1279 NULL: pte_offset_map(pmd, address))
1281 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1282 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1284 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1286 #define pte_alloc_kernel(pmd, address) \
1287 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1288 NULL: pte_offset_kernel(pmd, address))
1290 extern void free_area_init(unsigned long * zones_size);
1291 extern void free_area_init_node(int nid, unsigned long * zones_size,
1292 unsigned long zone_start_pfn, unsigned long *zholes_size);
1293 extern void free_initmem(void);
1295 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1297 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1298 * zones, allocate the backing mem_map and account for memory holes in a more
1299 * architecture independent manner. This is a substitute for creating the
1300 * zone_sizes[] and zholes_size[] arrays and passing them to
1301 * free_area_init_node()
1303 * An architecture is expected to register range of page frames backed by
1304 * physical memory with memblock_add[_node]() before calling
1305 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1306 * usage, an architecture is expected to do something like
1308 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1310 * for_each_valid_physical_page_range()
1311 * memblock_add_node(base, size, nid)
1312 * free_area_init_nodes(max_zone_pfns);
1314 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1315 * registered physical page range. Similarly
1316 * sparse_memory_present_with_active_regions() calls memory_present() for
1317 * each range when SPARSEMEM is enabled.
1319 * See mm/page_alloc.c for more information on each function exposed by
1320 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1322 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1323 unsigned long node_map_pfn_alignment(void);
1324 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1325 unsigned long end_pfn);
1326 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1327 unsigned long end_pfn);
1328 extern void get_pfn_range_for_nid(unsigned int nid,
1329 unsigned long *start_pfn, unsigned long *end_pfn);
1330 extern unsigned long find_min_pfn_with_active_regions(void);
1331 extern void free_bootmem_with_active_regions(int nid,
1332 unsigned long max_low_pfn);
1333 extern void sparse_memory_present_with_active_regions(int nid);
1335 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1337 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1338 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1339 static inline int __early_pfn_to_nid(unsigned long pfn)
1344 /* please see mm/page_alloc.c */
1345 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1346 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1347 /* there is a per-arch backend function. */
1348 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1349 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1352 extern void set_dma_reserve(unsigned long new_dma_reserve);
1353 extern void memmap_init_zone(unsigned long, int, unsigned long,
1354 unsigned long, enum memmap_context);
1355 extern void setup_per_zone_wmarks(void);
1356 extern int __meminit init_per_zone_wmark_min(void);
1357 extern void mem_init(void);
1358 extern void __init mmap_init(void);
1359 extern void show_mem(unsigned int flags);
1360 extern void si_meminfo(struct sysinfo * val);
1361 extern void si_meminfo_node(struct sysinfo *val, int nid);
1362 extern int after_bootmem;
1364 extern __printf(3, 4)
1365 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1367 extern void setup_per_cpu_pageset(void);
1369 extern void zone_pcp_update(struct zone *zone);
1370 extern void zone_pcp_reset(struct zone *zone);
1373 extern atomic_long_t mmap_pages_allocated;
1374 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1377 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1378 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1379 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1380 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1381 struct prio_tree_iter *iter);
1383 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1384 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1385 (vma = vma_prio_tree_next(vma, iter)); )
1387 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1388 struct list_head *list)
1390 vma->shared.vm_set.parent = NULL;
1391 list_add_tail(&vma->shared.vm_set.list, list);
1395 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1396 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1397 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1398 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1399 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1400 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1401 struct mempolicy *);
1402 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1403 extern int split_vma(struct mm_struct *,
1404 struct vm_area_struct *, unsigned long addr, int new_below);
1405 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1406 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1407 struct rb_node **, struct rb_node *);
1408 extern void unlink_file_vma(struct vm_area_struct *);
1409 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1410 unsigned long addr, unsigned long len, pgoff_t pgoff);
1411 extern void exit_mmap(struct mm_struct *);
1413 extern int mm_take_all_locks(struct mm_struct *mm);
1414 extern void mm_drop_all_locks(struct mm_struct *mm);
1416 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1417 extern void added_exe_file_vma(struct mm_struct *mm);
1418 extern void removed_exe_file_vma(struct mm_struct *mm);
1419 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1420 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1422 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1423 extern int install_special_mapping(struct mm_struct *mm,
1424 unsigned long addr, unsigned long len,
1425 unsigned long flags, struct page **pages);
1427 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1429 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1430 unsigned long len, unsigned long flags,
1431 vm_flags_t vm_flags, unsigned long pgoff);
1432 extern unsigned long do_mmap_pgoff(struct file *, unsigned long,
1433 unsigned long, unsigned long,
1434 unsigned long, unsigned long);
1435 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1437 /* These take the mm semaphore themselves */
1438 extern unsigned long vm_brk(unsigned long, unsigned long);
1439 extern int vm_munmap(unsigned long, size_t);
1440 extern unsigned long vm_mmap(struct file *, unsigned long,
1441 unsigned long, unsigned long,
1442 unsigned long, unsigned long);
1445 extern void truncate_inode_pages(struct address_space *, loff_t);
1446 extern void truncate_inode_pages_range(struct address_space *,
1447 loff_t lstart, loff_t lend);
1449 /* generic vm_area_ops exported for stackable file systems */
1450 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1451 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1453 /* mm/page-writeback.c */
1454 int write_one_page(struct page *page, int wait);
1455 void task_dirty_inc(struct task_struct *tsk);
1458 #define VM_MAX_READAHEAD 128 /* kbytes */
1459 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1461 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1462 pgoff_t offset, unsigned long nr_to_read);
1464 void page_cache_sync_readahead(struct address_space *mapping,
1465 struct file_ra_state *ra,
1468 unsigned long size);
1470 void page_cache_async_readahead(struct address_space *mapping,
1471 struct file_ra_state *ra,
1475 unsigned long size);
1477 unsigned long max_sane_readahead(unsigned long nr);
1478 unsigned long ra_submit(struct file_ra_state *ra,
1479 struct address_space *mapping,
1482 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1483 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1485 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1486 extern int expand_downwards(struct vm_area_struct *vma,
1487 unsigned long address);
1489 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1491 #define expand_upwards(vma, address) do { } while (0)
1494 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1495 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1496 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1497 struct vm_area_struct **pprev);
1499 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1500 NULL if none. Assume start_addr < end_addr. */
1501 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1503 struct vm_area_struct * vma = find_vma(mm,start_addr);
1505 if (vma && end_addr <= vma->vm_start)
1510 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1512 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1515 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1516 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1517 unsigned long vm_start, unsigned long vm_end)
1519 struct vm_area_struct *vma = find_vma(mm, vm_start);
1521 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1528 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1530 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1536 static inline pgprot_t vma_prot_none(struct vm_area_struct *vma)
1539 * obtain PROT_NONE by removing READ|WRITE|EXEC privs
1541 vm_flags_t vmflags = vma->vm_flags & ~(VM_READ|VM_WRITE|VM_EXEC);
1542 return pgprot_modify(vma->vm_page_prot, vm_get_page_prot(vmflags));
1545 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1546 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1547 unsigned long pfn, unsigned long size, pgprot_t);
1548 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1549 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1551 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1554 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1555 unsigned int foll_flags);
1556 #define FOLL_WRITE 0x01 /* check pte is writable */
1557 #define FOLL_TOUCH 0x02 /* mark page accessed */
1558 #define FOLL_GET 0x04 /* do get_page on page */
1559 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1560 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1561 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1562 * and return without waiting upon it */
1563 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1564 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1565 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1567 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1569 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1570 unsigned long size, pte_fn_t fn, void *data);
1572 #ifdef CONFIG_PROC_FS
1573 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1575 static inline void vm_stat_account(struct mm_struct *mm,
1576 unsigned long flags, struct file *file, long pages)
1578 mm->total_vm += pages;
1580 #endif /* CONFIG_PROC_FS */
1582 #ifdef CONFIG_DEBUG_PAGEALLOC
1583 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1584 #ifdef CONFIG_HIBERNATION
1585 extern bool kernel_page_present(struct page *page);
1586 #endif /* CONFIG_HIBERNATION */
1589 kernel_map_pages(struct page *page, int numpages, int enable) {}
1590 #ifdef CONFIG_HIBERNATION
1591 static inline bool kernel_page_present(struct page *page) { return true; }
1592 #endif /* CONFIG_HIBERNATION */
1595 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1596 #ifdef __HAVE_ARCH_GATE_AREA
1597 int in_gate_area_no_mm(unsigned long addr);
1598 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1600 int in_gate_area_no_mm(unsigned long addr);
1601 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1602 #endif /* __HAVE_ARCH_GATE_AREA */
1604 int drop_caches_sysctl_handler(struct ctl_table *, int,
1605 void __user *, size_t *, loff_t *);
1606 unsigned long shrink_slab(struct shrink_control *shrink,
1607 unsigned long nr_pages_scanned,
1608 unsigned long lru_pages);
1611 #define randomize_va_space 0
1613 extern int randomize_va_space;
1616 const char * arch_vma_name(struct vm_area_struct *vma);
1617 void print_vma_addr(char *prefix, unsigned long rip);
1619 void sparse_mem_maps_populate_node(struct page **map_map,
1620 unsigned long pnum_begin,
1621 unsigned long pnum_end,
1622 unsigned long map_count,
1625 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1626 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1627 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1628 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1629 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1630 void *vmemmap_alloc_block(unsigned long size, int node);
1631 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1632 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1633 int vmemmap_populate_basepages(struct page *start_page,
1634 unsigned long pages, int node);
1635 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1636 void vmemmap_populate_print_last(void);
1640 MF_COUNT_INCREASED = 1 << 0,
1641 MF_ACTION_REQUIRED = 1 << 1,
1642 MF_MUST_KILL = 1 << 2,
1644 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1645 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1646 extern int unpoison_memory(unsigned long pfn);
1647 extern int sysctl_memory_failure_early_kill;
1648 extern int sysctl_memory_failure_recovery;
1649 extern void shake_page(struct page *p, int access);
1650 extern atomic_long_t mce_bad_pages;
1651 extern int soft_offline_page(struct page *page, int flags);
1653 extern void dump_page(struct page *page);
1655 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1656 extern void clear_huge_page(struct page *page,
1658 unsigned int pages_per_huge_page);
1659 extern void copy_user_huge_page(struct page *dst, struct page *src,
1660 unsigned long addr, struct vm_area_struct *vma,
1661 unsigned int pages_per_huge_page);
1662 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1664 #ifdef CONFIG_DEBUG_PAGEALLOC
1665 extern unsigned int _debug_guardpage_minorder;
1667 static inline unsigned int debug_guardpage_minorder(void)
1669 return _debug_guardpage_minorder;
1672 static inline bool page_is_guard(struct page *page)
1674 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1677 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1678 static inline bool page_is_guard(struct page *page) { return false; }
1679 #endif /* CONFIG_DEBUG_PAGEALLOC */
1681 #endif /* __KERNEL__ */
1682 #endif /* _LINUX_MM_H */