4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/bit_spinlock.h>
19 #include <linux/shrinker.h>
23 struct anon_vma_chain;
26 struct writeback_control;
28 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
32 extern unsigned long num_physpages;
33 extern unsigned long totalram_pages;
34 extern void * high_memory;
35 extern int page_cluster;
38 extern int sysctl_legacy_va_layout;
40 #define sysctl_legacy_va_layout 0
44 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 extern unsigned long sysctl_user_reserve_kbytes;
48 extern unsigned long sysctl_admin_reserve_kbytes;
50 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
52 /* to align the pointer to the (next) page boundary */
53 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
55 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
56 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
59 * Linux kernel virtual memory manager primitives.
60 * The idea being to have a "virtual" mm in the same way
61 * we have a virtual fs - giving a cleaner interface to the
62 * mm details, and allowing different kinds of memory mappings
63 * (from shared memory to executable loading to arbitrary
67 extern struct kmem_cache *vm_area_cachep;
70 extern struct rb_root nommu_region_tree;
71 extern struct rw_semaphore nommu_region_sem;
73 extern unsigned int kobjsize(const void *objp);
77 * vm_flags in vm_area_struct, see mm_types.h.
79 #define VM_NONE 0x00000000
81 #define VM_READ 0x00000001 /* currently active flags */
82 #define VM_WRITE 0x00000002
83 #define VM_EXEC 0x00000004
84 #define VM_SHARED 0x00000008
86 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
87 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
88 #define VM_MAYWRITE 0x00000020
89 #define VM_MAYEXEC 0x00000040
90 #define VM_MAYSHARE 0x00000080
92 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
93 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
94 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
96 #define VM_LOCKED 0x00002000
97 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
99 /* Used by sys_madvise() */
100 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
101 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
103 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
104 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
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 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
110 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
112 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
113 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
114 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
115 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
117 #if defined(CONFIG_X86)
118 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
119 #elif defined(CONFIG_PPC)
120 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
121 #elif defined(CONFIG_PARISC)
122 # define VM_GROWSUP VM_ARCH_1
123 #elif defined(CONFIG_METAG)
124 # define VM_GROWSUP VM_ARCH_1
125 #elif defined(CONFIG_IA64)
126 # define VM_GROWSUP VM_ARCH_1
127 #elif !defined(CONFIG_MMU)
128 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
132 # define VM_GROWSUP VM_NONE
135 /* Bits set in the VMA until the stack is in its final location */
136 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
138 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
139 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
142 #ifdef CONFIG_STACK_GROWSUP
143 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
145 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
148 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
149 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
150 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
151 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
152 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
155 * Special vmas that are non-mergable, non-mlock()able.
156 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
158 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
161 * mapping from the currently active vm_flags protection bits (the
162 * low four bits) to a page protection mask..
164 extern pgprot_t protection_map[16];
166 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
167 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
168 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
169 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
170 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
171 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
172 #define FAULT_FLAG_TRIED 0x40 /* second try */
173 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
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 implement ->remap_pages to get nonlinear mapping support.
184 unsigned int flags; /* FAULT_FLAG_xxx flags */
185 pgoff_t pgoff; /* Logical page offset based on vma */
186 void __user *virtual_address; /* Faulting virtual address */
188 struct page *page; /* ->fault handlers should return a
189 * page here, unless VM_FAULT_NOPAGE
190 * is set (which is also implied by
196 * These are the virtual MM functions - opening of an area, closing and
197 * unmapping it (needed to keep files on disk up-to-date etc), pointer
198 * to the functions called when a no-page or a wp-page exception occurs.
200 struct vm_operations_struct {
201 void (*open)(struct vm_area_struct * area);
202 void (*close)(struct vm_area_struct * area);
203 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
205 /* notification that a previously read-only page is about to become
206 * writable, if an error is returned it will cause a SIGBUS */
207 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
209 /* called by access_process_vm when get_user_pages() fails, typically
210 * for use by special VMAs that can switch between memory and hardware
212 int (*access)(struct vm_area_struct *vma, unsigned long addr,
213 void *buf, int len, int write);
216 * set_policy() op must add a reference to any non-NULL @new mempolicy
217 * to hold the policy upon return. Caller should pass NULL @new to
218 * remove a policy and fall back to surrounding context--i.e. do not
219 * install a MPOL_DEFAULT policy, nor the task or system default
222 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
225 * get_policy() op must add reference [mpol_get()] to any policy at
226 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
227 * in mm/mempolicy.c will do this automatically.
228 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
229 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
230 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
231 * must return NULL--i.e., do not "fallback" to task or system default
234 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
236 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
237 const nodemask_t *to, unsigned long flags);
239 /* called by sys_remap_file_pages() to populate non-linear mapping */
240 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
241 unsigned long size, pgoff_t pgoff);
247 #define page_private(page) ((page)->private)
248 #define set_page_private(page, v) ((page)->private = (v))
250 /* It's valid only if the page is free path or free_list */
251 static inline void set_freepage_migratetype(struct page *page, int migratetype)
253 page->index = migratetype;
256 /* It's valid only if the page is free path or free_list */
257 static inline int get_freepage_migratetype(struct page *page)
263 * FIXME: take this include out, include page-flags.h in
264 * files which need it (119 of them)
266 #include <linux/page-flags.h>
267 #include <linux/huge_mm.h>
270 * Methods to modify the page usage count.
272 * What counts for a page usage:
273 * - cache mapping (page->mapping)
274 * - private data (page->private)
275 * - page mapped in a task's page tables, each mapping
276 * is counted separately
278 * Also, many kernel routines increase the page count before a critical
279 * routine so they can be sure the page doesn't go away from under them.
283 * Drop a ref, return true if the refcount fell to zero (the page has no users)
285 static inline int put_page_testzero(struct page *page)
287 VM_BUG_ON(atomic_read(&page->_count) == 0);
288 return atomic_dec_and_test(&page->_count);
292 * Try to grab a ref unless the page has a refcount of zero, return false if
295 static inline int get_page_unless_zero(struct page *page)
297 return atomic_inc_not_zero(&page->_count);
300 extern int page_is_ram(unsigned long pfn);
302 /* Support for virtually mapped pages */
303 struct page *vmalloc_to_page(const void *addr);
304 unsigned long vmalloc_to_pfn(const void *addr);
307 * Determine if an address is within the vmalloc range
309 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
310 * is no special casing required.
312 static inline int is_vmalloc_addr(const void *x)
315 unsigned long addr = (unsigned long)x;
317 return addr >= VMALLOC_START && addr < VMALLOC_END;
323 extern int is_vmalloc_or_module_addr(const void *x);
325 static inline int is_vmalloc_or_module_addr(const void *x)
331 extern void kvfree(const void *addr);
333 static inline void compound_lock(struct page *page)
335 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
336 VM_BUG_ON(PageSlab(page));
337 bit_spin_lock(PG_compound_lock, &page->flags);
341 static inline void compound_unlock(struct page *page)
343 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
344 VM_BUG_ON(PageSlab(page));
345 bit_spin_unlock(PG_compound_lock, &page->flags);
349 static inline unsigned long compound_lock_irqsave(struct page *page)
351 unsigned long uninitialized_var(flags);
352 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
353 local_irq_save(flags);
359 static inline void compound_unlock_irqrestore(struct page *page,
362 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
363 compound_unlock(page);
364 local_irq_restore(flags);
368 static inline struct page *compound_head(struct page *page)
370 if (unlikely(PageTail(page))) {
371 struct page *head = page->first_page;
374 * page->first_page may be a dangling pointer to an old
375 * compound page, so recheck that it is still a tail
376 * page before returning.
379 if (likely(PageTail(page)))
386 * The atomic page->_mapcount, starts from -1: so that transitions
387 * both from it and to it can be tracked, using atomic_inc_and_test
388 * and atomic_add_negative(-1).
390 static inline void page_mapcount_reset(struct page *page)
392 atomic_set(&(page)->_mapcount, -1);
395 static inline int page_mapcount(struct page *page)
397 return atomic_read(&(page)->_mapcount) + 1;
400 static inline int page_count(struct page *page)
402 return atomic_read(&compound_head(page)->_count);
405 static inline void get_huge_page_tail(struct page *page)
408 * __split_huge_page_refcount() cannot run
411 VM_BUG_ON(page_mapcount(page) < 0);
412 VM_BUG_ON(atomic_read(&page->_count) != 0);
413 atomic_inc(&page->_mapcount);
416 extern bool __get_page_tail(struct page *page);
418 static inline void get_page(struct page *page)
420 if (unlikely(PageTail(page)))
421 if (likely(__get_page_tail(page)))
424 * Getting a normal page or the head of a compound page
425 * requires to already have an elevated page->_count.
427 VM_BUG_ON(atomic_read(&page->_count) <= 0);
428 atomic_inc(&page->_count);
431 static inline struct page *virt_to_head_page(const void *x)
433 struct page *page = virt_to_page(x);
434 return compound_head(page);
438 * Setup the page count before being freed into the page allocator for
439 * the first time (boot or memory hotplug)
441 static inline void init_page_count(struct page *page)
443 atomic_set(&page->_count, 1);
447 * PageBuddy() indicate that the page is free and in the buddy system
448 * (see mm/page_alloc.c).
450 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
451 * -2 so that an underflow of the page_mapcount() won't be mistaken
452 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
453 * efficiently by most CPU architectures.
455 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
457 static inline int PageBuddy(struct page *page)
459 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
462 static inline void __SetPageBuddy(struct page *page)
464 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
465 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
468 static inline void __ClearPageBuddy(struct page *page)
470 VM_BUG_ON(!PageBuddy(page));
471 atomic_set(&page->_mapcount, -1);
474 void put_page(struct page *page);
475 void put_pages_list(struct list_head *pages);
477 void split_page(struct page *page, unsigned int order);
478 int split_free_page(struct page *page);
481 * Compound pages have a destructor function. Provide a
482 * prototype for that function and accessor functions.
483 * These are _only_ valid on the head of a PG_compound page.
485 typedef void compound_page_dtor(struct page *);
487 static inline void set_compound_page_dtor(struct page *page,
488 compound_page_dtor *dtor)
490 page[1].lru.next = (void *)dtor;
493 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
495 return (compound_page_dtor *)page[1].lru.next;
498 static inline int compound_order(struct page *page)
502 return (unsigned long)page[1].lru.prev;
505 static inline int compound_trans_order(struct page *page)
513 flags = compound_lock_irqsave(page);
514 order = compound_order(page);
515 compound_unlock_irqrestore(page, flags);
519 static inline void set_compound_order(struct page *page, unsigned long order)
521 page[1].lru.prev = (void *)order;
526 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
527 * servicing faults for write access. In the normal case, do always want
528 * pte_mkwrite. But get_user_pages can cause write faults for mappings
529 * that do not have writing enabled, when used by access_process_vm.
531 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
533 if (likely(vma->vm_flags & VM_WRITE))
534 pte = pte_mkwrite(pte);
540 * Multiple processes may "see" the same page. E.g. for untouched
541 * mappings of /dev/null, all processes see the same page full of
542 * zeroes, and text pages of executables and shared libraries have
543 * only one copy in memory, at most, normally.
545 * For the non-reserved pages, page_count(page) denotes a reference count.
546 * page_count() == 0 means the page is free. page->lru is then used for
547 * freelist management in the buddy allocator.
548 * page_count() > 0 means the page has been allocated.
550 * Pages are allocated by the slab allocator in order to provide memory
551 * to kmalloc and kmem_cache_alloc. In this case, the management of the
552 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
553 * unless a particular usage is carefully commented. (the responsibility of
554 * freeing the kmalloc memory is the caller's, of course).
556 * A page may be used by anyone else who does a __get_free_page().
557 * In this case, page_count still tracks the references, and should only
558 * be used through the normal accessor functions. The top bits of page->flags
559 * and page->virtual store page management information, but all other fields
560 * are unused and could be used privately, carefully. The management of this
561 * page is the responsibility of the one who allocated it, and those who have
562 * subsequently been given references to it.
564 * The other pages (we may call them "pagecache pages") are completely
565 * managed by the Linux memory manager: I/O, buffers, swapping etc.
566 * The following discussion applies only to them.
568 * A pagecache page contains an opaque `private' member, which belongs to the
569 * page's address_space. Usually, this is the address of a circular list of
570 * the page's disk buffers. PG_private must be set to tell the VM to call
571 * into the filesystem to release these pages.
573 * A page may belong to an inode's memory mapping. In this case, page->mapping
574 * is the pointer to the inode, and page->index is the file offset of the page,
575 * in units of PAGE_CACHE_SIZE.
577 * If pagecache pages are not associated with an inode, they are said to be
578 * anonymous pages. These may become associated with the swapcache, and in that
579 * case PG_swapcache is set, and page->private is an offset into the swapcache.
581 * In either case (swapcache or inode backed), the pagecache itself holds one
582 * reference to the page. Setting PG_private should also increment the
583 * refcount. The each user mapping also has a reference to the page.
585 * The pagecache pages are stored in a per-mapping radix tree, which is
586 * rooted at mapping->page_tree, and indexed by offset.
587 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
588 * lists, we instead now tag pages as dirty/writeback in the radix tree.
590 * All pagecache pages may be subject to I/O:
591 * - inode pages may need to be read from disk,
592 * - inode pages which have been modified and are MAP_SHARED may need
593 * to be written back to the inode on disk,
594 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
595 * modified may need to be swapped out to swap space and (later) to be read
600 * The zone field is never updated after free_area_init_core()
601 * sets it, so none of the operations on it need to be atomic.
604 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
605 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
606 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
607 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
608 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
611 * Define the bit shifts to access each section. For non-existent
612 * sections we define the shift as 0; that plus a 0 mask ensures
613 * the compiler will optimise away reference to them.
615 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
616 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
617 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
618 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
620 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
621 #ifdef NODE_NOT_IN_PAGE_FLAGS
622 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
623 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
624 SECTIONS_PGOFF : ZONES_PGOFF)
626 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
627 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
628 NODES_PGOFF : ZONES_PGOFF)
631 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
633 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
634 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
637 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
638 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
639 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
640 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
641 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
643 static inline enum zone_type page_zonenum(const struct page *page)
645 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
648 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
649 #define SECTION_IN_PAGE_FLAGS
653 * The identification function is only used by the buddy allocator for
654 * determining if two pages could be buddies. We are not really
655 * identifying a zone since we could be using a the section number
656 * id if we have not node id available in page flags.
657 * We guarantee only that it will return the same value for two
658 * combinable pages in a zone.
660 static inline int page_zone_id(struct page *page)
662 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
665 static inline int zone_to_nid(struct zone *zone)
674 #ifdef NODE_NOT_IN_PAGE_FLAGS
675 extern int page_to_nid(const struct page *page);
677 static inline int page_to_nid(const struct page *page)
679 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
683 #ifdef CONFIG_NUMA_BALANCING
684 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
685 static inline int page_nid_xchg_last(struct page *page, int nid)
687 return xchg(&page->_last_nid, nid);
690 static inline int page_nid_last(struct page *page)
692 return page->_last_nid;
694 static inline void page_nid_reset_last(struct page *page)
696 page->_last_nid = -1;
699 static inline int page_nid_last(struct page *page)
701 return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
704 extern int page_nid_xchg_last(struct page *page, int nid);
706 static inline void page_nid_reset_last(struct page *page)
708 int nid = (1 << LAST_NID_SHIFT) - 1;
710 page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
711 page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
713 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
715 static inline int page_nid_xchg_last(struct page *page, int nid)
717 return page_to_nid(page);
720 static inline int page_nid_last(struct page *page)
722 return page_to_nid(page);
725 static inline void page_nid_reset_last(struct page *page)
730 static inline struct zone *page_zone(const struct page *page)
732 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
735 #ifdef SECTION_IN_PAGE_FLAGS
736 static inline void set_page_section(struct page *page, unsigned long section)
738 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
739 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
742 static inline unsigned long page_to_section(const struct page *page)
744 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
748 static inline void set_page_zone(struct page *page, enum zone_type zone)
750 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
751 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
754 static inline void set_page_node(struct page *page, unsigned long node)
756 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
757 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
760 static inline void set_page_links(struct page *page, enum zone_type zone,
761 unsigned long node, unsigned long pfn)
763 set_page_zone(page, zone);
764 set_page_node(page, node);
765 #ifdef SECTION_IN_PAGE_FLAGS
766 set_page_section(page, pfn_to_section_nr(pfn));
771 * Some inline functions in vmstat.h depend on page_zone()
773 #include <linux/vmstat.h>
775 static __always_inline void *lowmem_page_address(const struct page *page)
777 return __va(PFN_PHYS(page_to_pfn(page)));
780 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
781 #define HASHED_PAGE_VIRTUAL
784 #if defined(WANT_PAGE_VIRTUAL)
785 static inline void *page_address(const struct page *page)
787 return page->virtual;
789 static inline void set_page_address(struct page *page, void *address)
791 page->virtual = address;
793 #define page_address_init() do { } while(0)
796 #if defined(HASHED_PAGE_VIRTUAL)
797 void *page_address(const struct page *page);
798 void set_page_address(struct page *page, void *virtual);
799 void page_address_init(void);
802 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
803 #define page_address(page) lowmem_page_address(page)
804 #define set_page_address(page, address) do { } while(0)
805 #define page_address_init() do { } while(0)
809 * On an anonymous page mapped into a user virtual memory area,
810 * page->mapping points to its anon_vma, not to a struct address_space;
811 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
813 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
814 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
815 * and then page->mapping points, not to an anon_vma, but to a private
816 * structure which KSM associates with that merged page. See ksm.h.
818 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
820 * Please note that, confusingly, "page_mapping" refers to the inode
821 * address_space which maps the page from disk; whereas "page_mapped"
822 * refers to user virtual address space into which the page is mapped.
824 #define PAGE_MAPPING_ANON 1
825 #define PAGE_MAPPING_KSM 2
826 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
828 extern struct address_space *page_mapping(struct page *page);
830 /* Neutral page->mapping pointer to address_space or anon_vma or other */
831 static inline void *page_rmapping(struct page *page)
833 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
836 extern struct address_space *__page_file_mapping(struct page *);
839 struct address_space *page_file_mapping(struct page *page)
841 if (unlikely(PageSwapCache(page)))
842 return __page_file_mapping(page);
844 return page->mapping;
847 static inline int PageAnon(struct page *page)
849 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
853 * Return the pagecache index of the passed page. Regular pagecache pages
854 * use ->index whereas swapcache pages use ->private
856 static inline pgoff_t page_index(struct page *page)
858 if (unlikely(PageSwapCache(page)))
859 return page_private(page);
863 extern pgoff_t __page_file_index(struct page *page);
866 * Return the file index of the page. Regular pagecache pages use ->index
867 * whereas swapcache pages use swp_offset(->private)
869 static inline pgoff_t page_file_index(struct page *page)
871 if (unlikely(PageSwapCache(page)))
872 return __page_file_index(page);
878 * Return true if this page is mapped into pagetables.
880 static inline int page_mapped(struct page *page)
882 return atomic_read(&(page)->_mapcount) >= 0;
886 * Different kinds of faults, as returned by handle_mm_fault().
887 * Used to decide whether a process gets delivered SIGBUS or
888 * just gets major/minor fault counters bumped up.
891 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
893 #define VM_FAULT_OOM 0x0001
894 #define VM_FAULT_SIGBUS 0x0002
895 #define VM_FAULT_MAJOR 0x0004
896 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
897 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
898 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
900 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
901 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
902 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
904 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
906 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
907 VM_FAULT_HWPOISON_LARGE)
909 /* Encode hstate index for a hwpoisoned large page */
910 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
911 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
914 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
916 extern void pagefault_out_of_memory(void);
918 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
921 * Flags passed to show_mem() and show_free_areas() to suppress output in
924 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
925 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
927 extern void show_free_areas(unsigned int flags);
928 extern bool skip_free_areas_node(unsigned int flags, int nid);
930 void shmem_set_file(struct vm_area_struct *vma, struct file *file);
931 int shmem_zero_setup(struct vm_area_struct *);
933 extern int can_do_mlock(void);
934 extern int user_shm_lock(size_t, struct user_struct *);
935 extern void user_shm_unlock(size_t, struct user_struct *);
938 * Parameter block passed down to zap_pte_range in exceptional cases.
941 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
942 struct address_space *check_mapping; /* Check page->mapping if set */
943 pgoff_t first_index; /* Lowest page->index to unmap */
944 pgoff_t last_index; /* Highest page->index to unmap */
947 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
950 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
952 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
953 unsigned long size, struct zap_details *);
954 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
955 unsigned long start, unsigned long end);
958 * mm_walk - callbacks for walk_page_range
959 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
960 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
961 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
962 * this handler is required to be able to handle
963 * pmd_trans_huge() pmds. They may simply choose to
964 * split_huge_page() instead of handling it explicitly.
965 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
966 * @pte_hole: if set, called for each hole at all levels
967 * @hugetlb_entry: if set, called for each hugetlb entry
968 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
971 * (see walk_page_range for more details)
974 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
975 unsigned long next, struct mm_walk *walk);
976 int (*pud_entry)(pud_t *pud, unsigned long addr,
977 unsigned long next, struct mm_walk *walk);
978 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
979 unsigned long next, struct mm_walk *walk);
980 int (*pte_entry)(pte_t *pte, unsigned long addr,
981 unsigned long next, struct mm_walk *walk);
982 int (*pte_hole)(unsigned long addr, unsigned long next,
983 struct mm_walk *walk);
984 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
985 unsigned long addr, unsigned long next,
986 struct mm_walk *walk);
987 struct mm_struct *mm;
991 int walk_page_range(unsigned long addr, unsigned long end,
992 struct mm_walk *walk);
993 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
994 unsigned long end, unsigned long floor, unsigned long ceiling);
995 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
996 struct vm_area_struct *vma);
997 void unmap_mapping_range(struct address_space *mapping,
998 loff_t const holebegin, loff_t const holelen, int even_cows);
999 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1000 unsigned long *pfn);
1001 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1002 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1003 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1004 void *buf, int len, int write);
1006 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1007 loff_t const holebegin, loff_t const holelen)
1009 unmap_mapping_range(mapping, holebegin, holelen, 0);
1012 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
1013 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1014 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1015 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1016 int truncate_inode_page(struct address_space *mapping, struct page *page);
1017 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1018 int invalidate_inode_page(struct page *page);
1021 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1022 unsigned long address, unsigned int flags);
1023 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1024 unsigned long address, unsigned int fault_flags);
1026 static inline int handle_mm_fault(struct mm_struct *mm,
1027 struct vm_area_struct *vma, unsigned long address,
1030 /* should never happen if there's no MMU */
1032 return VM_FAULT_SIGBUS;
1034 static inline int fixup_user_fault(struct task_struct *tsk,
1035 struct mm_struct *mm, unsigned long address,
1036 unsigned int fault_flags)
1038 /* should never happen if there's no MMU */
1044 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1045 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1046 void *buf, int len, int write);
1048 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1049 unsigned long start, unsigned long nr_pages,
1050 unsigned int foll_flags, struct page **pages,
1051 struct vm_area_struct **vmas, int *nonblocking);
1052 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1053 unsigned long start, unsigned long nr_pages,
1054 int write, int force, struct page **pages,
1055 struct vm_area_struct **vmas);
1056 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1057 struct page **pages);
1059 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1060 struct page **pages);
1061 int get_kernel_page(unsigned long start, int write, struct page **pages);
1062 struct page *get_dump_page(unsigned long addr);
1064 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1065 extern void do_invalidatepage(struct page *page, unsigned long offset);
1067 int __set_page_dirty_nobuffers(struct page *page);
1068 int __set_page_dirty_no_writeback(struct page *page);
1069 int redirty_page_for_writepage(struct writeback_control *wbc,
1071 void account_page_dirtied(struct page *page, struct address_space *mapping);
1072 void account_page_writeback(struct page *page);
1073 int set_page_dirty(struct page *page);
1074 int set_page_dirty_lock(struct page *page);
1075 int clear_page_dirty_for_io(struct page *page);
1077 /* Is the vma a continuation of the stack vma above it? */
1078 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1080 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1083 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1086 return (vma->vm_flags & VM_GROWSDOWN) &&
1087 (vma->vm_start == addr) &&
1088 !vma_growsdown(vma->vm_prev, addr);
1091 /* Is the vma a continuation of the stack vma below it? */
1092 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1094 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1097 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1100 return (vma->vm_flags & VM_GROWSUP) &&
1101 (vma->vm_end == addr) &&
1102 !vma_growsup(vma->vm_next, addr);
1106 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1108 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1109 unsigned long old_addr, struct vm_area_struct *new_vma,
1110 unsigned long new_addr, unsigned long len,
1111 bool need_rmap_locks);
1112 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1113 unsigned long end, pgprot_t newprot,
1114 int dirty_accountable, int prot_numa);
1115 extern int mprotect_fixup(struct vm_area_struct *vma,
1116 struct vm_area_struct **pprev, unsigned long start,
1117 unsigned long end, unsigned long newflags);
1120 * doesn't attempt to fault and will return short.
1122 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1123 struct page **pages);
1125 * per-process(per-mm_struct) statistics.
1127 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1129 long val = atomic_long_read(&mm->rss_stat.count[member]);
1131 #ifdef SPLIT_RSS_COUNTING
1133 * counter is updated in asynchronous manner and may go to minus.
1134 * But it's never be expected number for users.
1139 return (unsigned long)val;
1142 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1144 atomic_long_add(value, &mm->rss_stat.count[member]);
1147 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1149 atomic_long_inc(&mm->rss_stat.count[member]);
1152 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1154 atomic_long_dec(&mm->rss_stat.count[member]);
1157 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1159 return get_mm_counter(mm, MM_FILEPAGES) +
1160 get_mm_counter(mm, MM_ANONPAGES);
1163 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1165 return max(mm->hiwater_rss, get_mm_rss(mm));
1168 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1170 return max(mm->hiwater_vm, mm->total_vm);
1173 static inline void update_hiwater_rss(struct mm_struct *mm)
1175 unsigned long _rss = get_mm_rss(mm);
1177 if ((mm)->hiwater_rss < _rss)
1178 (mm)->hiwater_rss = _rss;
1181 static inline void update_hiwater_vm(struct mm_struct *mm)
1183 if (mm->hiwater_vm < mm->total_vm)
1184 mm->hiwater_vm = mm->total_vm;
1187 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1189 mm->hiwater_rss = get_mm_rss(mm);
1192 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1193 struct mm_struct *mm)
1195 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1197 if (*maxrss < hiwater_rss)
1198 *maxrss = hiwater_rss;
1201 #if defined(SPLIT_RSS_COUNTING)
1202 void sync_mm_rss(struct mm_struct *mm);
1204 static inline void sync_mm_rss(struct mm_struct *mm)
1209 int vma_wants_writenotify(struct vm_area_struct *vma);
1211 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1213 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1217 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1221 #ifdef __PAGETABLE_PUD_FOLDED
1222 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1223 unsigned long address)
1228 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1231 #ifdef __PAGETABLE_PMD_FOLDED
1232 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1233 unsigned long address)
1238 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1241 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1242 pmd_t *pmd, unsigned long address);
1243 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1246 * The following ifdef needed to get the 4level-fixup.h header to work.
1247 * Remove it when 4level-fixup.h has been removed.
1249 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1250 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1252 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1253 NULL: pud_offset(pgd, address);
1256 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1258 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1259 NULL: pmd_offset(pud, address);
1261 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1263 #if USE_SPLIT_PTLOCKS
1265 * We tuck a spinlock to guard each pagetable page into its struct page,
1266 * at page->private, with BUILD_BUG_ON to make sure that this will not
1267 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1268 * When freeing, reset page->mapping so free_pages_check won't complain.
1270 #define __pte_lockptr(page) &((page)->ptl)
1271 #define pte_lock_init(_page) do { \
1272 spin_lock_init(__pte_lockptr(_page)); \
1274 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1275 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1276 #else /* !USE_SPLIT_PTLOCKS */
1278 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1280 #define pte_lock_init(page) do {} while (0)
1281 #define pte_lock_deinit(page) do {} while (0)
1282 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1283 #endif /* USE_SPLIT_PTLOCKS */
1285 static inline void pgtable_page_ctor(struct page *page)
1287 pte_lock_init(page);
1288 inc_zone_page_state(page, NR_PAGETABLE);
1291 static inline void pgtable_page_dtor(struct page *page)
1293 pte_lock_deinit(page);
1294 dec_zone_page_state(page, NR_PAGETABLE);
1297 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1299 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1300 pte_t *__pte = pte_offset_map(pmd, address); \
1306 #define pte_unmap_unlock(pte, ptl) do { \
1311 #define pte_alloc_map(mm, vma, pmd, address) \
1312 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1314 NULL: pte_offset_map(pmd, address))
1316 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1317 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1319 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1321 #define pte_alloc_kernel(pmd, address) \
1322 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1323 NULL: pte_offset_kernel(pmd, address))
1325 extern void free_area_init(unsigned long * zones_size);
1326 extern void free_area_init_node(int nid, unsigned long * zones_size,
1327 unsigned long zone_start_pfn, unsigned long *zholes_size);
1328 extern void free_initmem(void);
1331 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1332 * into the buddy system. The freed pages will be poisoned with pattern
1333 * "poison" if it's non-zero.
1334 * Return pages freed into the buddy system.
1336 extern unsigned long free_reserved_area(unsigned long start, unsigned long end,
1337 int poison, char *s);
1338 #ifdef CONFIG_HIGHMEM
1340 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1341 * and totalram_pages.
1343 extern void free_highmem_page(struct page *page);
1346 static inline void adjust_managed_page_count(struct page *page, long count)
1348 totalram_pages += count;
1351 /* Free the reserved page into the buddy system, so it gets managed. */
1352 static inline void __free_reserved_page(struct page *page)
1354 ClearPageReserved(page);
1355 init_page_count(page);
1359 static inline void free_reserved_page(struct page *page)
1361 __free_reserved_page(page);
1362 adjust_managed_page_count(page, 1);
1365 static inline void mark_page_reserved(struct page *page)
1367 SetPageReserved(page);
1368 adjust_managed_page_count(page, -1);
1372 * Default method to free all the __init memory into the buddy system.
1373 * The freed pages will be poisoned with pattern "poison" if it is
1374 * non-zero. Return pages freed into the buddy system.
1376 static inline unsigned long free_initmem_default(int poison)
1378 extern char __init_begin[], __init_end[];
1380 return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin) ,
1381 ((unsigned long)&__init_end) & PAGE_MASK,
1382 poison, "unused kernel");
1385 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1387 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1388 * zones, allocate the backing mem_map and account for memory holes in a more
1389 * architecture independent manner. This is a substitute for creating the
1390 * zone_sizes[] and zholes_size[] arrays and passing them to
1391 * free_area_init_node()
1393 * An architecture is expected to register range of page frames backed by
1394 * physical memory with memblock_add[_node]() before calling
1395 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1396 * usage, an architecture is expected to do something like
1398 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1400 * for_each_valid_physical_page_range()
1401 * memblock_add_node(base, size, nid)
1402 * free_area_init_nodes(max_zone_pfns);
1404 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1405 * registered physical page range. Similarly
1406 * sparse_memory_present_with_active_regions() calls memory_present() for
1407 * each range when SPARSEMEM is enabled.
1409 * See mm/page_alloc.c for more information on each function exposed by
1410 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1412 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1413 unsigned long node_map_pfn_alignment(void);
1414 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1415 unsigned long end_pfn);
1416 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1417 unsigned long end_pfn);
1418 extern void get_pfn_range_for_nid(unsigned int nid,
1419 unsigned long *start_pfn, unsigned long *end_pfn);
1420 extern unsigned long find_min_pfn_with_active_regions(void);
1421 extern void free_bootmem_with_active_regions(int nid,
1422 unsigned long max_low_pfn);
1423 extern void sparse_memory_present_with_active_regions(int nid);
1425 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1427 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1428 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1429 static inline int __early_pfn_to_nid(unsigned long pfn)
1434 /* please see mm/page_alloc.c */
1435 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1436 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1437 /* there is a per-arch backend function. */
1438 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1439 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1442 extern void set_dma_reserve(unsigned long new_dma_reserve);
1443 extern void memmap_init_zone(unsigned long, int, unsigned long,
1444 unsigned long, enum memmap_context);
1445 extern void setup_per_zone_wmarks(void);
1446 extern int __meminit init_per_zone_wmark_min(void);
1447 extern void mem_init(void);
1448 extern void __init mmap_init(void);
1449 extern void show_mem(unsigned int flags);
1450 extern void si_meminfo(struct sysinfo * val);
1451 extern void si_meminfo_node(struct sysinfo *val, int nid);
1453 extern __printf(3, 4)
1454 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1456 extern void setup_per_cpu_pageset(void);
1458 extern void zone_pcp_update(struct zone *zone);
1459 extern void zone_pcp_reset(struct zone *zone);
1462 extern int min_free_kbytes;
1465 extern atomic_long_t mmap_pages_allocated;
1466 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1468 /* interval_tree.c */
1469 void vma_interval_tree_insert(struct vm_area_struct *node,
1470 struct rb_root *root);
1471 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1472 struct vm_area_struct *prev,
1473 struct rb_root *root);
1474 void vma_interval_tree_remove(struct vm_area_struct *node,
1475 struct rb_root *root);
1476 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1477 unsigned long start, unsigned long last);
1478 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1479 unsigned long start, unsigned long last);
1481 #define vma_interval_tree_foreach(vma, root, start, last) \
1482 for (vma = vma_interval_tree_iter_first(root, start, last); \
1483 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1485 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1486 struct list_head *list)
1488 list_add_tail(&vma->shared.nonlinear, list);
1491 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1492 struct rb_root *root);
1493 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1494 struct rb_root *root);
1495 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1496 struct rb_root *root, unsigned long start, unsigned long last);
1497 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1498 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1499 #ifdef CONFIG_DEBUG_VM_RB
1500 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1503 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1504 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1505 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1508 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1509 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1510 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1511 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1512 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1513 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1514 struct mempolicy *, const char __user *);
1515 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1516 extern int split_vma(struct mm_struct *,
1517 struct vm_area_struct *, unsigned long addr, int new_below);
1518 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1519 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1520 struct rb_node **, struct rb_node *);
1521 extern void unlink_file_vma(struct vm_area_struct *);
1522 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1523 unsigned long addr, unsigned long len, pgoff_t pgoff,
1524 bool *need_rmap_locks);
1525 extern void exit_mmap(struct mm_struct *);
1527 extern int mm_take_all_locks(struct mm_struct *mm);
1528 extern void mm_drop_all_locks(struct mm_struct *mm);
1530 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1531 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1533 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1534 extern int install_special_mapping(struct mm_struct *mm,
1535 unsigned long addr, unsigned long len,
1536 unsigned long flags, struct page **pages);
1538 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1540 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1541 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1542 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1543 unsigned long len, unsigned long prot, unsigned long flags,
1544 unsigned long pgoff, unsigned long *populate);
1545 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1548 extern int __mm_populate(unsigned long addr, unsigned long len,
1550 static inline void mm_populate(unsigned long addr, unsigned long len)
1553 (void) __mm_populate(addr, len, 1);
1556 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1559 /* These take the mm semaphore themselves */
1560 extern unsigned long vm_brk(unsigned long, unsigned long);
1561 extern int vm_munmap(unsigned long, size_t);
1562 extern unsigned long vm_mmap(struct file *, unsigned long,
1563 unsigned long, unsigned long,
1564 unsigned long, unsigned long);
1566 struct vm_unmapped_area_info {
1567 #define VM_UNMAPPED_AREA_TOPDOWN 1
1568 unsigned long flags;
1569 unsigned long length;
1570 unsigned long low_limit;
1571 unsigned long high_limit;
1572 unsigned long align_mask;
1573 unsigned long align_offset;
1576 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1577 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1580 * Search for an unmapped address range.
1582 * We are looking for a range that:
1583 * - does not intersect with any VMA;
1584 * - is contained within the [low_limit, high_limit) interval;
1585 * - is at least the desired size.
1586 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1588 static inline unsigned long
1589 vm_unmapped_area(struct vm_unmapped_area_info *info)
1591 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1592 return unmapped_area(info);
1594 return unmapped_area_topdown(info);
1598 extern void truncate_inode_pages(struct address_space *, loff_t);
1599 extern void truncate_inode_pages_range(struct address_space *,
1600 loff_t lstart, loff_t lend);
1602 /* generic vm_area_ops exported for stackable file systems */
1603 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1604 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1606 /* mm/page-writeback.c */
1607 int write_one_page(struct page *page, int wait);
1608 void task_dirty_inc(struct task_struct *tsk);
1611 #define VM_MAX_READAHEAD 128 /* kbytes */
1612 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1614 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1615 pgoff_t offset, unsigned long nr_to_read);
1617 void page_cache_sync_readahead(struct address_space *mapping,
1618 struct file_ra_state *ra,
1621 unsigned long size);
1623 void page_cache_async_readahead(struct address_space *mapping,
1624 struct file_ra_state *ra,
1628 unsigned long size);
1630 unsigned long max_sane_readahead(unsigned long nr);
1631 unsigned long ra_submit(struct file_ra_state *ra,
1632 struct address_space *mapping,
1635 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1636 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1638 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1639 extern int expand_downwards(struct vm_area_struct *vma,
1640 unsigned long address);
1642 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1644 #define expand_upwards(vma, address) (0)
1647 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1648 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1649 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1650 struct vm_area_struct **pprev);
1652 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1653 NULL if none. Assume start_addr < end_addr. */
1654 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1656 struct vm_area_struct * vma = find_vma(mm,start_addr);
1658 if (vma && end_addr <= vma->vm_start)
1663 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1665 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1668 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1669 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1670 unsigned long vm_start, unsigned long vm_end)
1672 struct vm_area_struct *vma = find_vma(mm, vm_start);
1674 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1681 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1683 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1689 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1690 unsigned long change_prot_numa(struct vm_area_struct *vma,
1691 unsigned long start, unsigned long end);
1694 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1695 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1696 unsigned long pfn, unsigned long size, pgprot_t);
1697 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1698 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1700 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1702 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1705 struct page *follow_page_mask(struct vm_area_struct *vma,
1706 unsigned long address, unsigned int foll_flags,
1707 unsigned int *page_mask);
1709 static inline struct page *follow_page(struct vm_area_struct *vma,
1710 unsigned long address, unsigned int foll_flags)
1712 unsigned int unused_page_mask;
1713 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1716 #define FOLL_WRITE 0x01 /* check pte is writable */
1717 #define FOLL_TOUCH 0x02 /* mark page accessed */
1718 #define FOLL_GET 0x04 /* do get_page on page */
1719 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1720 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1721 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1722 * and return without waiting upon it */
1723 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1724 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1725 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1726 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1727 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1729 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1731 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1732 unsigned long size, pte_fn_t fn, void *data);
1734 #ifdef CONFIG_PROC_FS
1735 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1737 static inline void vm_stat_account(struct mm_struct *mm,
1738 unsigned long flags, struct file *file, long pages)
1740 mm->total_vm += pages;
1742 #endif /* CONFIG_PROC_FS */
1744 #ifdef CONFIG_DEBUG_PAGEALLOC
1745 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1746 #ifdef CONFIG_HIBERNATION
1747 extern bool kernel_page_present(struct page *page);
1748 #endif /* CONFIG_HIBERNATION */
1751 kernel_map_pages(struct page *page, int numpages, int enable) {}
1752 #ifdef CONFIG_HIBERNATION
1753 static inline bool kernel_page_present(struct page *page) { return true; }
1754 #endif /* CONFIG_HIBERNATION */
1757 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1758 #ifdef __HAVE_ARCH_GATE_AREA
1759 int in_gate_area_no_mm(unsigned long addr);
1760 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1762 int in_gate_area_no_mm(unsigned long addr);
1763 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1764 #endif /* __HAVE_ARCH_GATE_AREA */
1766 #ifdef CONFIG_SYSCTL
1767 extern int sysctl_drop_caches;
1768 int drop_caches_sysctl_handler(struct ctl_table *, int,
1769 void __user *, size_t *, loff_t *);
1772 unsigned long shrink_slab(struct shrink_control *shrink,
1773 unsigned long nr_pages_scanned,
1774 unsigned long lru_pages);
1777 #define randomize_va_space 0
1779 extern int randomize_va_space;
1782 const char * arch_vma_name(struct vm_area_struct *vma);
1783 void print_vma_addr(char *prefix, unsigned long rip);
1785 void sparse_mem_maps_populate_node(struct page **map_map,
1786 unsigned long pnum_begin,
1787 unsigned long pnum_end,
1788 unsigned long map_count,
1791 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1792 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1793 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1794 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1795 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1796 void *vmemmap_alloc_block(unsigned long size, int node);
1797 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1798 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1799 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1801 int vmemmap_populate(unsigned long start, unsigned long end, int node);
1802 void vmemmap_populate_print_last(void);
1803 #ifdef CONFIG_MEMORY_HOTPLUG
1804 void vmemmap_free(unsigned long start, unsigned long end);
1806 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1807 unsigned long size);
1810 MF_COUNT_INCREASED = 1 << 0,
1811 MF_ACTION_REQUIRED = 1 << 1,
1812 MF_MUST_KILL = 1 << 2,
1814 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1815 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1816 extern int unpoison_memory(unsigned long pfn);
1817 extern int sysctl_memory_failure_early_kill;
1818 extern int sysctl_memory_failure_recovery;
1819 extern void shake_page(struct page *p, int access);
1820 extern atomic_long_t num_poisoned_pages;
1821 extern int soft_offline_page(struct page *page, int flags);
1823 extern void dump_page(struct page *page);
1825 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1826 extern void clear_huge_page(struct page *page,
1828 unsigned int pages_per_huge_page);
1829 extern void copy_user_huge_page(struct page *dst, struct page *src,
1830 unsigned long addr, struct vm_area_struct *vma,
1831 unsigned int pages_per_huge_page);
1832 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1834 #ifdef CONFIG_DEBUG_PAGEALLOC
1835 extern unsigned int _debug_guardpage_minorder;
1837 static inline unsigned int debug_guardpage_minorder(void)
1839 return _debug_guardpage_minorder;
1842 static inline bool page_is_guard(struct page *page)
1844 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1847 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1848 static inline bool page_is_guard(struct page *page) { return false; }
1849 #endif /* CONFIG_DEBUG_PAGEALLOC */
1851 #if MAX_NUMNODES > 1
1852 void __init setup_nr_node_ids(void);
1854 static inline void setup_nr_node_ids(void) {}
1857 #endif /* __KERNEL__ */
1858 #endif /* _LINUX_MM_H */