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_POPULATE 0x00001000
91 #define VM_LOCKED 0x00002000
92 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
94 /* Used by sys_madvise() */
95 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
96 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
98 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
99 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
100 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
101 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
102 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
103 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
104 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
105 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
107 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
108 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
109 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
110 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
112 #if defined(CONFIG_X86)
113 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
114 #elif defined(CONFIG_PPC)
115 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
116 #elif defined(CONFIG_PARISC)
117 # define VM_GROWSUP VM_ARCH_1
118 #elif defined(CONFIG_IA64)
119 # define VM_GROWSUP VM_ARCH_1
120 #elif !defined(CONFIG_MMU)
121 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
125 # define VM_GROWSUP VM_NONE
128 /* Bits set in the VMA until the stack is in its final location */
129 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
131 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
132 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
135 #ifdef CONFIG_STACK_GROWSUP
136 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
138 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
141 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
142 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
143 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
144 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
145 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
148 * Special vmas that are non-mergable, non-mlock()able.
149 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
151 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
154 * mapping from the currently active vm_flags protection bits (the
155 * low four bits) to a page protection mask..
157 extern pgprot_t protection_map[16];
159 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
160 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
161 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
162 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
163 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
164 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
165 #define FAULT_FLAG_TRIED 0x40 /* second try */
168 * vm_fault is filled by the the pagefault handler and passed to the vma's
169 * ->fault function. The vma's ->fault is responsible for returning a bitmask
170 * of VM_FAULT_xxx flags that give details about how the fault was handled.
172 * pgoff should be used in favour of virtual_address, if possible. If pgoff
173 * is used, one may implement ->remap_pages to get nonlinear mapping support.
176 unsigned int flags; /* FAULT_FLAG_xxx flags */
177 pgoff_t pgoff; /* Logical page offset based on vma */
178 void __user *virtual_address; /* Faulting virtual address */
180 struct page *page; /* ->fault handlers should return a
181 * page here, unless VM_FAULT_NOPAGE
182 * is set (which is also implied by
188 * These are the virtual MM functions - opening of an area, closing and
189 * unmapping it (needed to keep files on disk up-to-date etc), pointer
190 * to the functions called when a no-page or a wp-page exception occurs.
192 struct vm_operations_struct {
193 void (*open)(struct vm_area_struct * area);
194 void (*close)(struct vm_area_struct * area);
195 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
197 /* notification that a previously read-only page is about to become
198 * writable, if an error is returned it will cause a SIGBUS */
199 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
201 /* called by access_process_vm when get_user_pages() fails, typically
202 * for use by special VMAs that can switch between memory and hardware
204 int (*access)(struct vm_area_struct *vma, unsigned long addr,
205 void *buf, int len, int write);
208 * set_policy() op must add a reference to any non-NULL @new mempolicy
209 * to hold the policy upon return. Caller should pass NULL @new to
210 * remove a policy and fall back to surrounding context--i.e. do not
211 * install a MPOL_DEFAULT policy, nor the task or system default
214 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
217 * get_policy() op must add reference [mpol_get()] to any policy at
218 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
219 * in mm/mempolicy.c will do this automatically.
220 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
221 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
222 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
223 * must return NULL--i.e., do not "fallback" to task or system default
226 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
228 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
229 const nodemask_t *to, unsigned long flags);
231 /* called by sys_remap_file_pages() to populate non-linear mapping */
232 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
233 unsigned long size, pgoff_t pgoff);
239 #define page_private(page) ((page)->private)
240 #define set_page_private(page, v) ((page)->private = (v))
242 /* It's valid only if the page is free path or free_list */
243 static inline void set_freepage_migratetype(struct page *page, int migratetype)
245 page->index = migratetype;
248 /* It's valid only if the page is free path or free_list */
249 static inline int get_freepage_migratetype(struct page *page)
255 * FIXME: take this include out, include page-flags.h in
256 * files which need it (119 of them)
258 #include <linux/page-flags.h>
259 #include <linux/huge_mm.h>
262 * Methods to modify the page usage count.
264 * What counts for a page usage:
265 * - cache mapping (page->mapping)
266 * - private data (page->private)
267 * - page mapped in a task's page tables, each mapping
268 * is counted separately
270 * Also, many kernel routines increase the page count before a critical
271 * routine so they can be sure the page doesn't go away from under them.
275 * Drop a ref, return true if the refcount fell to zero (the page has no users)
277 static inline int put_page_testzero(struct page *page)
279 VM_BUG_ON(atomic_read(&page->_count) == 0);
280 return atomic_dec_and_test(&page->_count);
284 * Try to grab a ref unless the page has a refcount of zero, return false if
287 static inline int get_page_unless_zero(struct page *page)
289 return atomic_inc_not_zero(&page->_count);
292 extern int page_is_ram(unsigned long pfn);
294 /* Support for virtually mapped pages */
295 struct page *vmalloc_to_page(const void *addr);
296 unsigned long vmalloc_to_pfn(const void *addr);
299 * Determine if an address is within the vmalloc range
301 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
302 * is no special casing required.
304 static inline int is_vmalloc_addr(const void *x)
307 unsigned long addr = (unsigned long)x;
309 return addr >= VMALLOC_START && addr < VMALLOC_END;
315 extern int is_vmalloc_or_module_addr(const void *x);
317 static inline int is_vmalloc_or_module_addr(const void *x)
323 static inline void compound_lock(struct page *page)
325 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
326 VM_BUG_ON(PageSlab(page));
327 bit_spin_lock(PG_compound_lock, &page->flags);
331 static inline void compound_unlock(struct page *page)
333 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
334 VM_BUG_ON(PageSlab(page));
335 bit_spin_unlock(PG_compound_lock, &page->flags);
339 static inline unsigned long compound_lock_irqsave(struct page *page)
341 unsigned long uninitialized_var(flags);
342 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
343 local_irq_save(flags);
349 static inline void compound_unlock_irqrestore(struct page *page,
352 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
353 compound_unlock(page);
354 local_irq_restore(flags);
358 static inline struct page *compound_head(struct page *page)
360 if (unlikely(PageTail(page)))
361 return page->first_page;
366 * The atomic page->_mapcount, starts from -1: so that transitions
367 * both from it and to it can be tracked, using atomic_inc_and_test
368 * and atomic_add_negative(-1).
370 static inline void page_mapcount_reset(struct page *page)
372 atomic_set(&(page)->_mapcount, -1);
375 static inline int page_mapcount(struct page *page)
377 return atomic_read(&(page)->_mapcount) + 1;
380 static inline int page_count(struct page *page)
382 return atomic_read(&compound_head(page)->_count);
385 static inline void get_huge_page_tail(struct page *page)
388 * __split_huge_page_refcount() cannot run
391 VM_BUG_ON(page_mapcount(page) < 0);
392 VM_BUG_ON(atomic_read(&page->_count) != 0);
393 atomic_inc(&page->_mapcount);
396 extern bool __get_page_tail(struct page *page);
398 static inline void get_page(struct page *page)
400 if (unlikely(PageTail(page)))
401 if (likely(__get_page_tail(page)))
404 * Getting a normal page or the head of a compound page
405 * requires to already have an elevated page->_count.
407 VM_BUG_ON(atomic_read(&page->_count) <= 0);
408 atomic_inc(&page->_count);
411 static inline struct page *virt_to_head_page(const void *x)
413 struct page *page = virt_to_page(x);
414 return compound_head(page);
418 * Setup the page count before being freed into the page allocator for
419 * the first time (boot or memory hotplug)
421 static inline void init_page_count(struct page *page)
423 atomic_set(&page->_count, 1);
427 * PageBuddy() indicate that the page is free and in the buddy system
428 * (see mm/page_alloc.c).
430 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
431 * -2 so that an underflow of the page_mapcount() won't be mistaken
432 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
433 * efficiently by most CPU architectures.
435 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
437 static inline int PageBuddy(struct page *page)
439 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
442 static inline void __SetPageBuddy(struct page *page)
444 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
445 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
448 static inline void __ClearPageBuddy(struct page *page)
450 VM_BUG_ON(!PageBuddy(page));
451 atomic_set(&page->_mapcount, -1);
454 void put_page(struct page *page);
455 void put_pages_list(struct list_head *pages);
457 void split_page(struct page *page, unsigned int order);
458 int split_free_page(struct page *page);
461 * Compound pages have a destructor function. Provide a
462 * prototype for that function and accessor functions.
463 * These are _only_ valid on the head of a PG_compound page.
465 typedef void compound_page_dtor(struct page *);
467 static inline void set_compound_page_dtor(struct page *page,
468 compound_page_dtor *dtor)
470 page[1].lru.next = (void *)dtor;
473 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
475 return (compound_page_dtor *)page[1].lru.next;
478 static inline int compound_order(struct page *page)
482 return (unsigned long)page[1].lru.prev;
485 static inline int compound_trans_order(struct page *page)
493 flags = compound_lock_irqsave(page);
494 order = compound_order(page);
495 compound_unlock_irqrestore(page, flags);
499 static inline void set_compound_order(struct page *page, unsigned long order)
501 page[1].lru.prev = (void *)order;
506 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
507 * servicing faults for write access. In the normal case, do always want
508 * pte_mkwrite. But get_user_pages can cause write faults for mappings
509 * that do not have writing enabled, when used by access_process_vm.
511 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
513 if (likely(vma->vm_flags & VM_WRITE))
514 pte = pte_mkwrite(pte);
520 * Multiple processes may "see" the same page. E.g. for untouched
521 * mappings of /dev/null, all processes see the same page full of
522 * zeroes, and text pages of executables and shared libraries have
523 * only one copy in memory, at most, normally.
525 * For the non-reserved pages, page_count(page) denotes a reference count.
526 * page_count() == 0 means the page is free. page->lru is then used for
527 * freelist management in the buddy allocator.
528 * page_count() > 0 means the page has been allocated.
530 * Pages are allocated by the slab allocator in order to provide memory
531 * to kmalloc and kmem_cache_alloc. In this case, the management of the
532 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
533 * unless a particular usage is carefully commented. (the responsibility of
534 * freeing the kmalloc memory is the caller's, of course).
536 * A page may be used by anyone else who does a __get_free_page().
537 * In this case, page_count still tracks the references, and should only
538 * be used through the normal accessor functions. The top bits of page->flags
539 * and page->virtual store page management information, but all other fields
540 * are unused and could be used privately, carefully. The management of this
541 * page is the responsibility of the one who allocated it, and those who have
542 * subsequently been given references to it.
544 * The other pages (we may call them "pagecache pages") are completely
545 * managed by the Linux memory manager: I/O, buffers, swapping etc.
546 * The following discussion applies only to them.
548 * A pagecache page contains an opaque `private' member, which belongs to the
549 * page's address_space. Usually, this is the address of a circular list of
550 * the page's disk buffers. PG_private must be set to tell the VM to call
551 * into the filesystem to release these pages.
553 * A page may belong to an inode's memory mapping. In this case, page->mapping
554 * is the pointer to the inode, and page->index is the file offset of the page,
555 * in units of PAGE_CACHE_SIZE.
557 * If pagecache pages are not associated with an inode, they are said to be
558 * anonymous pages. These may become associated with the swapcache, and in that
559 * case PG_swapcache is set, and page->private is an offset into the swapcache.
561 * In either case (swapcache or inode backed), the pagecache itself holds one
562 * reference to the page. Setting PG_private should also increment the
563 * refcount. The each user mapping also has a reference to the page.
565 * The pagecache pages are stored in a per-mapping radix tree, which is
566 * rooted at mapping->page_tree, and indexed by offset.
567 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
568 * lists, we instead now tag pages as dirty/writeback in the radix tree.
570 * All pagecache pages may be subject to I/O:
571 * - inode pages may need to be read from disk,
572 * - inode pages which have been modified and are MAP_SHARED may need
573 * to be written back to the inode on disk,
574 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
575 * modified may need to be swapped out to swap space and (later) to be read
580 * The zone field is never updated after free_area_init_core()
581 * sets it, so none of the operations on it need to be atomic.
584 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
585 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
586 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
587 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
588 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
591 * Define the bit shifts to access each section. For non-existent
592 * sections we define the shift as 0; that plus a 0 mask ensures
593 * the compiler will optimise away reference to them.
595 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
596 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
597 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
598 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
600 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
601 #ifdef NODE_NOT_IN_PAGE_FLAGS
602 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
603 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
604 SECTIONS_PGOFF : ZONES_PGOFF)
606 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
607 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
608 NODES_PGOFF : ZONES_PGOFF)
611 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
613 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
614 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
617 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
618 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
619 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
620 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
621 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
623 static inline enum zone_type page_zonenum(const struct page *page)
625 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
628 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
629 #define SECTION_IN_PAGE_FLAGS
633 * The identification function is only used by the buddy allocator for
634 * determining if two pages could be buddies. We are not really
635 * identifying a zone since we could be using a the section number
636 * id if we have not node id available in page flags.
637 * We guarantee only that it will return the same value for two
638 * combinable pages in a zone.
640 static inline int page_zone_id(struct page *page)
642 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
645 static inline int zone_to_nid(struct zone *zone)
654 #ifdef NODE_NOT_IN_PAGE_FLAGS
655 extern int page_to_nid(const struct page *page);
657 static inline int page_to_nid(const struct page *page)
659 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
663 #ifdef CONFIG_NUMA_BALANCING
664 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
665 static inline int page_nid_xchg_last(struct page *page, int nid)
667 return xchg(&page->_last_nid, nid);
670 static inline int page_nid_last(struct page *page)
672 return page->_last_nid;
674 static inline void page_nid_reset_last(struct page *page)
676 page->_last_nid = -1;
679 static inline int page_nid_last(struct page *page)
681 return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
684 extern int page_nid_xchg_last(struct page *page, int nid);
686 static inline void page_nid_reset_last(struct page *page)
688 int nid = (1 << LAST_NID_SHIFT) - 1;
690 page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
691 page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
693 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
695 static inline int page_nid_xchg_last(struct page *page, int nid)
697 return page_to_nid(page);
700 static inline int page_nid_last(struct page *page)
702 return page_to_nid(page);
705 static inline void page_nid_reset_last(struct page *page)
710 static inline struct zone *page_zone(const struct page *page)
712 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
715 #ifdef SECTION_IN_PAGE_FLAGS
716 static inline void set_page_section(struct page *page, unsigned long section)
718 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
719 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
722 static inline unsigned long page_to_section(const struct page *page)
724 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
728 static inline void set_page_zone(struct page *page, enum zone_type zone)
730 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
731 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
734 static inline void set_page_node(struct page *page, unsigned long node)
736 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
737 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
740 static inline void set_page_links(struct page *page, enum zone_type zone,
741 unsigned long node, unsigned long pfn)
743 set_page_zone(page, zone);
744 set_page_node(page, node);
745 #ifdef SECTION_IN_PAGE_FLAGS
746 set_page_section(page, pfn_to_section_nr(pfn));
751 * Some inline functions in vmstat.h depend on page_zone()
753 #include <linux/vmstat.h>
755 static __always_inline void *lowmem_page_address(const struct page *page)
757 return __va(PFN_PHYS(page_to_pfn(page)));
760 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
761 #define HASHED_PAGE_VIRTUAL
764 #if defined(WANT_PAGE_VIRTUAL)
765 #define page_address(page) ((page)->virtual)
766 #define set_page_address(page, address) \
768 (page)->virtual = (address); \
770 #define page_address_init() do { } while(0)
773 #if defined(HASHED_PAGE_VIRTUAL)
774 void *page_address(const struct page *page);
775 void set_page_address(struct page *page, void *virtual);
776 void page_address_init(void);
779 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
780 #define page_address(page) lowmem_page_address(page)
781 #define set_page_address(page, address) do { } while(0)
782 #define page_address_init() do { } while(0)
786 * On an anonymous page mapped into a user virtual memory area,
787 * page->mapping points to its anon_vma, not to a struct address_space;
788 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
790 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
791 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
792 * and then page->mapping points, not to an anon_vma, but to a private
793 * structure which KSM associates with that merged page. See ksm.h.
795 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
797 * Please note that, confusingly, "page_mapping" refers to the inode
798 * address_space which maps the page from disk; whereas "page_mapped"
799 * refers to user virtual address space into which the page is mapped.
801 #define PAGE_MAPPING_ANON 1
802 #define PAGE_MAPPING_KSM 2
803 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
805 extern struct address_space *page_mapping(struct page *page);
807 /* Neutral page->mapping pointer to address_space or anon_vma or other */
808 static inline void *page_rmapping(struct page *page)
810 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
813 extern struct address_space *__page_file_mapping(struct page *);
816 struct address_space *page_file_mapping(struct page *page)
818 if (unlikely(PageSwapCache(page)))
819 return __page_file_mapping(page);
821 return page->mapping;
824 static inline int PageAnon(struct page *page)
826 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
830 * Return the pagecache index of the passed page. Regular pagecache pages
831 * use ->index whereas swapcache pages use ->private
833 static inline pgoff_t page_index(struct page *page)
835 if (unlikely(PageSwapCache(page)))
836 return page_private(page);
840 extern pgoff_t __page_file_index(struct page *page);
843 * Return the file index of the page. Regular pagecache pages use ->index
844 * whereas swapcache pages use swp_offset(->private)
846 static inline pgoff_t page_file_index(struct page *page)
848 if (unlikely(PageSwapCache(page)))
849 return __page_file_index(page);
855 * Return true if this page is mapped into pagetables.
857 static inline int page_mapped(struct page *page)
859 return atomic_read(&(page)->_mapcount) >= 0;
863 * Different kinds of faults, as returned by handle_mm_fault().
864 * Used to decide whether a process gets delivered SIGBUS or
865 * just gets major/minor fault counters bumped up.
868 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
870 #define VM_FAULT_OOM 0x0001
871 #define VM_FAULT_SIGBUS 0x0002
872 #define VM_FAULT_MAJOR 0x0004
873 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
874 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
875 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
877 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
878 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
879 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
881 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
883 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
884 VM_FAULT_HWPOISON_LARGE)
886 /* Encode hstate index for a hwpoisoned large page */
887 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
888 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
891 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
893 extern void pagefault_out_of_memory(void);
895 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
898 * Flags passed to show_mem() and show_free_areas() to suppress output in
901 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
903 extern void show_free_areas(unsigned int flags);
904 extern bool skip_free_areas_node(unsigned int flags, int nid);
906 int shmem_zero_setup(struct vm_area_struct *);
908 extern int can_do_mlock(void);
909 extern int user_shm_lock(size_t, struct user_struct *);
910 extern void user_shm_unlock(size_t, struct user_struct *);
913 * Parameter block passed down to zap_pte_range in exceptional cases.
916 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
917 struct address_space *check_mapping; /* Check page->mapping if set */
918 pgoff_t first_index; /* Lowest page->index to unmap */
919 pgoff_t last_index; /* Highest page->index to unmap */
922 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
925 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
927 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
928 unsigned long size, struct zap_details *);
929 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
930 unsigned long start, unsigned long end);
933 * mm_walk - callbacks for walk_page_range
934 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
935 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
936 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
937 * this handler is required to be able to handle
938 * pmd_trans_huge() pmds. They may simply choose to
939 * split_huge_page() instead of handling it explicitly.
940 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
941 * @pte_hole: if set, called for each hole at all levels
942 * @hugetlb_entry: if set, called for each hugetlb entry
943 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
946 * (see walk_page_range for more details)
949 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
950 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
951 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
952 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
953 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
954 int (*hugetlb_entry)(pte_t *, unsigned long,
955 unsigned long, unsigned long, struct mm_walk *);
956 struct mm_struct *mm;
960 int walk_page_range(unsigned long addr, unsigned long end,
961 struct mm_walk *walk);
962 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
963 unsigned long end, unsigned long floor, unsigned long ceiling);
964 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
965 struct vm_area_struct *vma);
966 void unmap_mapping_range(struct address_space *mapping,
967 loff_t const holebegin, loff_t const holelen, int even_cows);
968 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
970 int follow_phys(struct vm_area_struct *vma, unsigned long address,
971 unsigned int flags, unsigned long *prot, resource_size_t *phys);
972 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
973 void *buf, int len, int write);
975 static inline void unmap_shared_mapping_range(struct address_space *mapping,
976 loff_t const holebegin, loff_t const holelen)
978 unmap_mapping_range(mapping, holebegin, holelen, 0);
981 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
982 extern void truncate_setsize(struct inode *inode, loff_t newsize);
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 access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1013 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1014 void *buf, int len, int write);
1016 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1017 unsigned long start, unsigned long nr_pages,
1018 unsigned int foll_flags, struct page **pages,
1019 struct vm_area_struct **vmas, int *nonblocking);
1020 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1021 unsigned long start, unsigned long nr_pages,
1022 int write, int force, struct page **pages,
1023 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 bool need_rmap_locks);
1080 extern unsigned long do_mremap(unsigned long addr,
1081 unsigned long old_len, unsigned long new_len,
1082 unsigned long flags, unsigned long new_addr);
1083 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1084 unsigned long end, pgprot_t newprot,
1085 int dirty_accountable, int prot_numa);
1086 extern int mprotect_fixup(struct vm_area_struct *vma,
1087 struct vm_area_struct **pprev, unsigned long start,
1088 unsigned long end, unsigned long newflags);
1091 * doesn't attempt to fault and will return short.
1093 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1094 struct page **pages);
1096 * per-process(per-mm_struct) statistics.
1098 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1100 long val = atomic_long_read(&mm->rss_stat.count[member]);
1102 #ifdef SPLIT_RSS_COUNTING
1104 * counter is updated in asynchronous manner and may go to minus.
1105 * But it's never be expected number for users.
1110 return (unsigned long)val;
1113 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1115 atomic_long_add(value, &mm->rss_stat.count[member]);
1118 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1120 atomic_long_inc(&mm->rss_stat.count[member]);
1123 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1125 atomic_long_dec(&mm->rss_stat.count[member]);
1128 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1130 return get_mm_counter(mm, MM_FILEPAGES) +
1131 get_mm_counter(mm, MM_ANONPAGES);
1134 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1136 return max(mm->hiwater_rss, get_mm_rss(mm));
1139 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1141 return max(mm->hiwater_vm, mm->total_vm);
1144 static inline void update_hiwater_rss(struct mm_struct *mm)
1146 unsigned long _rss = get_mm_rss(mm);
1148 if ((mm)->hiwater_rss < _rss)
1149 (mm)->hiwater_rss = _rss;
1152 static inline void update_hiwater_vm(struct mm_struct *mm)
1154 if (mm->hiwater_vm < mm->total_vm)
1155 mm->hiwater_vm = mm->total_vm;
1158 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1159 struct mm_struct *mm)
1161 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1163 if (*maxrss < hiwater_rss)
1164 *maxrss = hiwater_rss;
1167 #if defined(SPLIT_RSS_COUNTING)
1168 void sync_mm_rss(struct mm_struct *mm);
1170 static inline void sync_mm_rss(struct mm_struct *mm)
1175 int vma_wants_writenotify(struct vm_area_struct *vma);
1177 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1179 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1183 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1187 #ifdef __PAGETABLE_PUD_FOLDED
1188 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1189 unsigned long address)
1194 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1197 #ifdef __PAGETABLE_PMD_FOLDED
1198 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1199 unsigned long address)
1204 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1207 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1208 pmd_t *pmd, unsigned long address);
1209 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1212 * The following ifdef needed to get the 4level-fixup.h header to work.
1213 * Remove it when 4level-fixup.h has been removed.
1215 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1216 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1218 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1219 NULL: pud_offset(pgd, address);
1222 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1224 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1225 NULL: pmd_offset(pud, address);
1227 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1229 #if USE_SPLIT_PTLOCKS
1231 * We tuck a spinlock to guard each pagetable page into its struct page,
1232 * at page->private, with BUILD_BUG_ON to make sure that this will not
1233 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1234 * When freeing, reset page->mapping so free_pages_check won't complain.
1236 #define __pte_lockptr(page) &((page)->ptl)
1237 #define pte_lock_init(_page) do { \
1238 spin_lock_init(__pte_lockptr(_page)); \
1240 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1241 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1242 #else /* !USE_SPLIT_PTLOCKS */
1244 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1246 #define pte_lock_init(page) do {} while (0)
1247 #define pte_lock_deinit(page) do {} while (0)
1248 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1249 #endif /* USE_SPLIT_PTLOCKS */
1251 static inline void pgtable_page_ctor(struct page *page)
1253 pte_lock_init(page);
1254 inc_zone_page_state(page, NR_PAGETABLE);
1257 static inline void pgtable_page_dtor(struct page *page)
1259 pte_lock_deinit(page);
1260 dec_zone_page_state(page, NR_PAGETABLE);
1263 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1265 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1266 pte_t *__pte = pte_offset_map(pmd, address); \
1272 #define pte_unmap_unlock(pte, ptl) do { \
1277 #define pte_alloc_map(mm, vma, pmd, address) \
1278 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1280 NULL: pte_offset_map(pmd, address))
1282 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1283 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1285 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1287 #define pte_alloc_kernel(pmd, address) \
1288 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1289 NULL: pte_offset_kernel(pmd, address))
1291 extern void free_area_init(unsigned long * zones_size);
1292 extern void free_area_init_node(int nid, unsigned long * zones_size,
1293 unsigned long zone_start_pfn, unsigned long *zholes_size);
1294 extern void free_initmem(void);
1296 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1298 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1299 * zones, allocate the backing mem_map and account for memory holes in a more
1300 * architecture independent manner. This is a substitute for creating the
1301 * zone_sizes[] and zholes_size[] arrays and passing them to
1302 * free_area_init_node()
1304 * An architecture is expected to register range of page frames backed by
1305 * physical memory with memblock_add[_node]() before calling
1306 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1307 * usage, an architecture is expected to do something like
1309 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1311 * for_each_valid_physical_page_range()
1312 * memblock_add_node(base, size, nid)
1313 * free_area_init_nodes(max_zone_pfns);
1315 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1316 * registered physical page range. Similarly
1317 * sparse_memory_present_with_active_regions() calls memory_present() for
1318 * each range when SPARSEMEM is enabled.
1320 * See mm/page_alloc.c for more information on each function exposed by
1321 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1323 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1324 unsigned long node_map_pfn_alignment(void);
1325 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1326 unsigned long end_pfn);
1327 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1328 unsigned long end_pfn);
1329 extern void get_pfn_range_for_nid(unsigned int nid,
1330 unsigned long *start_pfn, unsigned long *end_pfn);
1331 extern unsigned long find_min_pfn_with_active_regions(void);
1332 extern void free_bootmem_with_active_regions(int nid,
1333 unsigned long max_low_pfn);
1334 extern void sparse_memory_present_with_active_regions(int nid);
1336 #define MOVABLEMEM_MAP_MAX MAX_NUMNODES
1337 struct movablemem_entry {
1338 unsigned long start_pfn; /* start pfn of memory segment */
1339 unsigned long end_pfn; /* end pfn of memory segment (exclusive) */
1342 struct movablemem_map {
1343 bool acpi; /* true if using SRAT info */
1345 struct movablemem_entry map[MOVABLEMEM_MAP_MAX];
1346 nodemask_t numa_nodes_hotplug; /* on which nodes we specify memory */
1347 nodemask_t numa_nodes_kernel; /* on which nodes kernel resides in */
1350 extern void __init insert_movablemem_map(unsigned long start_pfn,
1351 unsigned long end_pfn);
1352 extern int __init movablemem_map_overlap(unsigned long start_pfn,
1353 unsigned long end_pfn);
1354 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1356 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1357 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1358 static inline int __early_pfn_to_nid(unsigned long pfn)
1363 /* please see mm/page_alloc.c */
1364 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1365 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1366 /* there is a per-arch backend function. */
1367 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1368 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1371 extern void set_dma_reserve(unsigned long new_dma_reserve);
1372 extern void memmap_init_zone(unsigned long, int, unsigned long,
1373 unsigned long, enum memmap_context);
1374 extern void setup_per_zone_wmarks(void);
1375 extern int __meminit init_per_zone_wmark_min(void);
1376 extern void mem_init(void);
1377 extern void __init mmap_init(void);
1378 extern void show_mem(unsigned int flags);
1379 extern void si_meminfo(struct sysinfo * val);
1380 extern void si_meminfo_node(struct sysinfo *val, int nid);
1382 extern __printf(3, 4)
1383 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1385 extern void setup_per_cpu_pageset(void);
1387 extern void zone_pcp_update(struct zone *zone);
1388 extern void zone_pcp_reset(struct zone *zone);
1391 extern int min_free_kbytes;
1394 extern atomic_long_t mmap_pages_allocated;
1395 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1397 /* interval_tree.c */
1398 void vma_interval_tree_insert(struct vm_area_struct *node,
1399 struct rb_root *root);
1400 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1401 struct vm_area_struct *prev,
1402 struct rb_root *root);
1403 void vma_interval_tree_remove(struct vm_area_struct *node,
1404 struct rb_root *root);
1405 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1406 unsigned long start, unsigned long last);
1407 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1408 unsigned long start, unsigned long last);
1410 #define vma_interval_tree_foreach(vma, root, start, last) \
1411 for (vma = vma_interval_tree_iter_first(root, start, last); \
1412 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1414 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1415 struct list_head *list)
1417 list_add_tail(&vma->shared.nonlinear, list);
1420 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1421 struct rb_root *root);
1422 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1423 struct rb_root *root);
1424 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1425 struct rb_root *root, unsigned long start, unsigned long last);
1426 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1427 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1428 #ifdef CONFIG_DEBUG_VM_RB
1429 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1432 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1433 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1434 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1437 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1438 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1439 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1440 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1441 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1442 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1443 struct mempolicy *);
1444 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1445 extern int split_vma(struct mm_struct *,
1446 struct vm_area_struct *, unsigned long addr, int new_below);
1447 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1448 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1449 struct rb_node **, struct rb_node *);
1450 extern void unlink_file_vma(struct vm_area_struct *);
1451 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1452 unsigned long addr, unsigned long len, pgoff_t pgoff,
1453 bool *need_rmap_locks);
1454 extern void exit_mmap(struct mm_struct *);
1456 extern int mm_take_all_locks(struct mm_struct *mm);
1457 extern void mm_drop_all_locks(struct mm_struct *mm);
1459 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1460 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1462 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1463 extern int install_special_mapping(struct mm_struct *mm,
1464 unsigned long addr, unsigned long len,
1465 unsigned long flags, struct page **pages);
1467 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1469 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1470 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1471 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1472 unsigned long len, unsigned long prot, unsigned long flags,
1473 unsigned long pgoff, unsigned long *populate);
1474 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1477 extern int __mm_populate(unsigned long addr, unsigned long len,
1479 static inline void mm_populate(unsigned long addr, unsigned long len)
1482 (void) __mm_populate(addr, len, 1);
1485 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1488 /* These take the mm semaphore themselves */
1489 extern unsigned long vm_brk(unsigned long, unsigned long);
1490 extern int vm_munmap(unsigned long, size_t);
1491 extern unsigned long vm_mmap(struct file *, unsigned long,
1492 unsigned long, unsigned long,
1493 unsigned long, unsigned long);
1495 struct vm_unmapped_area_info {
1496 #define VM_UNMAPPED_AREA_TOPDOWN 1
1497 unsigned long flags;
1498 unsigned long length;
1499 unsigned long low_limit;
1500 unsigned long high_limit;
1501 unsigned long align_mask;
1502 unsigned long align_offset;
1505 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1506 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1509 * Search for an unmapped address range.
1511 * We are looking for a range that:
1512 * - does not intersect with any VMA;
1513 * - is contained within the [low_limit, high_limit) interval;
1514 * - is at least the desired size.
1515 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1517 static inline unsigned long
1518 vm_unmapped_area(struct vm_unmapped_area_info *info)
1520 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1521 return unmapped_area(info);
1523 return unmapped_area_topdown(info);
1527 extern void truncate_inode_pages(struct address_space *, loff_t);
1528 extern void truncate_inode_pages_range(struct address_space *,
1529 loff_t lstart, loff_t lend);
1531 /* generic vm_area_ops exported for stackable file systems */
1532 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1533 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1535 /* mm/page-writeback.c */
1536 int write_one_page(struct page *page, int wait);
1537 void task_dirty_inc(struct task_struct *tsk);
1540 #define VM_MAX_READAHEAD 128 /* kbytes */
1541 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1543 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1544 pgoff_t offset, unsigned long nr_to_read);
1546 void page_cache_sync_readahead(struct address_space *mapping,
1547 struct file_ra_state *ra,
1550 unsigned long size);
1552 void page_cache_async_readahead(struct address_space *mapping,
1553 struct file_ra_state *ra,
1557 unsigned long size);
1559 unsigned long max_sane_readahead(unsigned long nr);
1560 unsigned long ra_submit(struct file_ra_state *ra,
1561 struct address_space *mapping,
1564 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1565 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1567 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1568 extern int expand_downwards(struct vm_area_struct *vma,
1569 unsigned long address);
1571 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1573 #define expand_upwards(vma, address) do { } while (0)
1576 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1577 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1578 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1579 struct vm_area_struct **pprev);
1581 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1582 NULL if none. Assume start_addr < end_addr. */
1583 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1585 struct vm_area_struct * vma = find_vma(mm,start_addr);
1587 if (vma && end_addr <= vma->vm_start)
1592 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1594 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1597 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1598 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1599 unsigned long vm_start, unsigned long vm_end)
1601 struct vm_area_struct *vma = find_vma(mm, vm_start);
1603 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1610 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1612 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1618 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1619 unsigned long change_prot_numa(struct vm_area_struct *vma,
1620 unsigned long start, unsigned long end);
1623 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1624 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1625 unsigned long pfn, unsigned long size, pgprot_t);
1626 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1627 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1629 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1632 struct page *follow_page_mask(struct vm_area_struct *vma,
1633 unsigned long address, unsigned int foll_flags,
1634 unsigned int *page_mask);
1636 static inline struct page *follow_page(struct vm_area_struct *vma,
1637 unsigned long address, unsigned int foll_flags)
1639 unsigned int unused_page_mask;
1640 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1643 #define FOLL_WRITE 0x01 /* check pte is writable */
1644 #define FOLL_TOUCH 0x02 /* mark page accessed */
1645 #define FOLL_GET 0x04 /* do get_page on page */
1646 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1647 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1648 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1649 * and return without waiting upon it */
1650 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1651 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1652 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1653 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1654 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1656 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1658 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1659 unsigned long size, pte_fn_t fn, void *data);
1661 #ifdef CONFIG_PROC_FS
1662 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1664 static inline void vm_stat_account(struct mm_struct *mm,
1665 unsigned long flags, struct file *file, long pages)
1667 mm->total_vm += pages;
1669 #endif /* CONFIG_PROC_FS */
1671 #ifdef CONFIG_DEBUG_PAGEALLOC
1672 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1673 #ifdef CONFIG_HIBERNATION
1674 extern bool kernel_page_present(struct page *page);
1675 #endif /* CONFIG_HIBERNATION */
1678 kernel_map_pages(struct page *page, int numpages, int enable) {}
1679 #ifdef CONFIG_HIBERNATION
1680 static inline bool kernel_page_present(struct page *page) { return true; }
1681 #endif /* CONFIG_HIBERNATION */
1684 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1685 #ifdef __HAVE_ARCH_GATE_AREA
1686 int in_gate_area_no_mm(unsigned long addr);
1687 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1689 int in_gate_area_no_mm(unsigned long addr);
1690 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1691 #endif /* __HAVE_ARCH_GATE_AREA */
1693 int drop_caches_sysctl_handler(struct ctl_table *, int,
1694 void __user *, size_t *, loff_t *);
1695 unsigned long shrink_slab(struct shrink_control *shrink,
1696 unsigned long nr_pages_scanned,
1697 unsigned long lru_pages);
1700 #define randomize_va_space 0
1702 extern int randomize_va_space;
1705 const char * arch_vma_name(struct vm_area_struct *vma);
1706 void print_vma_addr(char *prefix, unsigned long rip);
1708 void sparse_mem_maps_populate_node(struct page **map_map,
1709 unsigned long pnum_begin,
1710 unsigned long pnum_end,
1711 unsigned long map_count,
1714 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1715 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1716 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1717 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1718 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1719 void *vmemmap_alloc_block(unsigned long size, int node);
1720 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1721 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1722 int vmemmap_populate_basepages(struct page *start_page,
1723 unsigned long pages, int node);
1724 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1725 void vmemmap_populate_print_last(void);
1726 #ifdef CONFIG_MEMORY_HOTPLUG
1727 void vmemmap_free(struct page *memmap, unsigned long nr_pages);
1729 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1730 unsigned long size);
1733 MF_COUNT_INCREASED = 1 << 0,
1734 MF_ACTION_REQUIRED = 1 << 1,
1735 MF_MUST_KILL = 1 << 2,
1737 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1738 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1739 extern int unpoison_memory(unsigned long pfn);
1740 extern int sysctl_memory_failure_early_kill;
1741 extern int sysctl_memory_failure_recovery;
1742 extern void shake_page(struct page *p, int access);
1743 extern atomic_long_t num_poisoned_pages;
1744 extern int soft_offline_page(struct page *page, int flags);
1746 extern void dump_page(struct page *page);
1748 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1749 extern void clear_huge_page(struct page *page,
1751 unsigned int pages_per_huge_page);
1752 extern void copy_user_huge_page(struct page *dst, struct page *src,
1753 unsigned long addr, struct vm_area_struct *vma,
1754 unsigned int pages_per_huge_page);
1755 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1757 #ifdef CONFIG_DEBUG_PAGEALLOC
1758 extern unsigned int _debug_guardpage_minorder;
1760 static inline unsigned int debug_guardpage_minorder(void)
1762 return _debug_guardpage_minorder;
1765 static inline bool page_is_guard(struct page *page)
1767 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1770 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1771 static inline bool page_is_guard(struct page *page) { return false; }
1772 #endif /* CONFIG_DEBUG_PAGEALLOC */
1774 #endif /* __KERNEL__ */
1775 #endif /* _LINUX_MM_H */