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_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
31 static inline void set_max_mapnr(unsigned long limit)
36 static inline void set_max_mapnr(unsigned long limit) { }
39 extern unsigned long totalram_pages;
40 extern void * high_memory;
41 extern int page_cluster;
44 extern int sysctl_legacy_va_layout;
46 #define sysctl_legacy_va_layout 0
50 #include <asm/pgtable.h>
51 #include <asm/processor.h>
53 extern unsigned long sysctl_user_reserve_kbytes;
54 extern unsigned long sysctl_admin_reserve_kbytes;
56 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
58 /* to align the pointer to the (next) page boundary */
59 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
61 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
62 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
65 * Linux kernel virtual memory manager primitives.
66 * The idea being to have a "virtual" mm in the same way
67 * we have a virtual fs - giving a cleaner interface to the
68 * mm details, and allowing different kinds of memory mappings
69 * (from shared memory to executable loading to arbitrary
73 extern struct kmem_cache *vm_area_cachep;
76 extern struct rb_root nommu_region_tree;
77 extern struct rw_semaphore nommu_region_sem;
79 extern unsigned int kobjsize(const void *objp);
83 * vm_flags in vm_area_struct, see mm_types.h.
85 #define VM_NONE 0x00000000
87 #define VM_READ 0x00000001 /* currently active flags */
88 #define VM_WRITE 0x00000002
89 #define VM_EXEC 0x00000004
90 #define VM_SHARED 0x00000008
92 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
93 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
94 #define VM_MAYWRITE 0x00000020
95 #define VM_MAYEXEC 0x00000040
96 #define VM_MAYSHARE 0x00000080
98 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
99 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
100 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
102 #define VM_LOCKED 0x00002000
103 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
105 /* Used by sys_madvise() */
106 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
107 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
109 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
110 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
111 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
112 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
113 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
114 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
115 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
116 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
118 #ifdef CONFIG_MEM_SOFT_DIRTY
119 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
121 # define VM_SOFTDIRTY 0
124 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
125 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
126 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
127 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
129 #if defined(CONFIG_X86)
130 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
131 #elif defined(CONFIG_PPC)
132 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
133 #elif defined(CONFIG_PARISC)
134 # define VM_GROWSUP VM_ARCH_1
135 #elif defined(CONFIG_METAG)
136 # define VM_GROWSUP VM_ARCH_1
137 #elif defined(CONFIG_IA64)
138 # define VM_GROWSUP VM_ARCH_1
139 #elif !defined(CONFIG_MMU)
140 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
144 # define VM_GROWSUP VM_NONE
147 /* Bits set in the VMA until the stack is in its final location */
148 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
150 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
151 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
154 #ifdef CONFIG_STACK_GROWSUP
155 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
157 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
161 * Special vmas that are non-mergable, non-mlock()able.
162 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
164 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
167 * mapping from the currently active vm_flags protection bits (the
168 * low four bits) to a page protection mask..
170 extern pgprot_t protection_map[16];
172 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
173 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
174 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
175 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
176 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
177 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
178 #define FAULT_FLAG_TRIED 0x40 /* second try */
179 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
182 * vm_fault is filled by the the pagefault handler and passed to the vma's
183 * ->fault function. The vma's ->fault is responsible for returning a bitmask
184 * of VM_FAULT_xxx flags that give details about how the fault was handled.
186 * pgoff should be used in favour of virtual_address, if possible. If pgoff
187 * is used, one may implement ->remap_pages to get nonlinear mapping support.
190 unsigned int flags; /* FAULT_FLAG_xxx flags */
191 pgoff_t pgoff; /* Logical page offset based on vma */
192 void __user *virtual_address; /* Faulting virtual address */
194 struct page *page; /* ->fault handlers should return a
195 * page here, unless VM_FAULT_NOPAGE
196 * is set (which is also implied by
202 * These are the virtual MM functions - opening of an area, closing and
203 * unmapping it (needed to keep files on disk up-to-date etc), pointer
204 * to the functions called when a no-page or a wp-page exception occurs.
206 struct vm_operations_struct {
207 void (*open)(struct vm_area_struct * area);
208 void (*close)(struct vm_area_struct * area);
209 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
211 /* notification that a previously read-only page is about to become
212 * writable, if an error is returned it will cause a SIGBUS */
213 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
215 /* called by access_process_vm when get_user_pages() fails, typically
216 * for use by special VMAs that can switch between memory and hardware
218 int (*access)(struct vm_area_struct *vma, unsigned long addr,
219 void *buf, int len, int write);
222 * set_policy() op must add a reference to any non-NULL @new mempolicy
223 * to hold the policy upon return. Caller should pass NULL @new to
224 * remove a policy and fall back to surrounding context--i.e. do not
225 * install a MPOL_DEFAULT policy, nor the task or system default
228 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
231 * get_policy() op must add reference [mpol_get()] to any policy at
232 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
233 * in mm/mempolicy.c will do this automatically.
234 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
235 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
236 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
237 * must return NULL--i.e., do not "fallback" to task or system default
240 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
242 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
243 const nodemask_t *to, unsigned long flags);
245 /* called by sys_remap_file_pages() to populate non-linear mapping */
246 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
247 unsigned long size, pgoff_t pgoff);
253 #define page_private(page) ((page)->private)
254 #define set_page_private(page, v) ((page)->private = (v))
256 /* It's valid only if the page is free path or free_list */
257 static inline void set_freepage_migratetype(struct page *page, int migratetype)
259 page->index = migratetype;
262 /* It's valid only if the page is free path or free_list */
263 static inline int get_freepage_migratetype(struct page *page)
269 * FIXME: take this include out, include page-flags.h in
270 * files which need it (119 of them)
272 #include <linux/page-flags.h>
273 #include <linux/huge_mm.h>
276 * Methods to modify the page usage count.
278 * What counts for a page usage:
279 * - cache mapping (page->mapping)
280 * - private data (page->private)
281 * - page mapped in a task's page tables, each mapping
282 * is counted separately
284 * Also, many kernel routines increase the page count before a critical
285 * routine so they can be sure the page doesn't go away from under them.
289 * Drop a ref, return true if the refcount fell to zero (the page has no users)
291 static inline int put_page_testzero(struct page *page)
293 VM_BUG_ON(atomic_read(&page->_count) == 0);
294 return atomic_dec_and_test(&page->_count);
298 * Try to grab a ref unless the page has a refcount of zero, return false if
301 static inline int get_page_unless_zero(struct page *page)
303 return atomic_inc_not_zero(&page->_count);
306 extern int page_is_ram(unsigned long pfn);
308 /* Support for virtually mapped pages */
309 struct page *vmalloc_to_page(const void *addr);
310 unsigned long vmalloc_to_pfn(const void *addr);
313 * Determine if an address is within the vmalloc range
315 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
316 * is no special casing required.
318 static inline int is_vmalloc_addr(const void *x)
321 unsigned long addr = (unsigned long)x;
323 return addr >= VMALLOC_START && addr < VMALLOC_END;
329 extern int is_vmalloc_or_module_addr(const void *x);
331 static inline int is_vmalloc_or_module_addr(const void *x)
337 static inline void compound_lock(struct page *page)
339 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
340 VM_BUG_ON(PageSlab(page));
341 bit_spin_lock(PG_compound_lock, &page->flags);
345 static inline void compound_unlock(struct page *page)
347 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
348 VM_BUG_ON(PageSlab(page));
349 bit_spin_unlock(PG_compound_lock, &page->flags);
353 static inline unsigned long compound_lock_irqsave(struct page *page)
355 unsigned long uninitialized_var(flags);
356 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
357 local_irq_save(flags);
363 static inline void compound_unlock_irqrestore(struct page *page,
366 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
367 compound_unlock(page);
368 local_irq_restore(flags);
372 static inline struct page *compound_head(struct page *page)
374 if (unlikely(PageTail(page)))
375 return page->first_page;
380 * The atomic page->_mapcount, starts from -1: so that transitions
381 * both from it and to it can be tracked, using atomic_inc_and_test
382 * and atomic_add_negative(-1).
384 static inline void page_mapcount_reset(struct page *page)
386 atomic_set(&(page)->_mapcount, -1);
389 static inline int page_mapcount(struct page *page)
391 return atomic_read(&(page)->_mapcount) + 1;
394 static inline int page_count(struct page *page)
396 return atomic_read(&compound_head(page)->_count);
399 static inline void get_huge_page_tail(struct page *page)
402 * __split_huge_page_refcount() cannot run
405 VM_BUG_ON(page_mapcount(page) < 0);
406 VM_BUG_ON(atomic_read(&page->_count) != 0);
407 atomic_inc(&page->_mapcount);
410 extern bool __get_page_tail(struct page *page);
412 static inline void get_page(struct page *page)
414 if (unlikely(PageTail(page)))
415 if (likely(__get_page_tail(page)))
418 * Getting a normal page or the head of a compound page
419 * requires to already have an elevated page->_count.
421 VM_BUG_ON(atomic_read(&page->_count) <= 0);
422 atomic_inc(&page->_count);
425 static inline struct page *virt_to_head_page(const void *x)
427 struct page *page = virt_to_page(x);
428 return compound_head(page);
432 * Setup the page count before being freed into the page allocator for
433 * the first time (boot or memory hotplug)
435 static inline void init_page_count(struct page *page)
437 atomic_set(&page->_count, 1);
441 * PageBuddy() indicate that the page is free and in the buddy system
442 * (see mm/page_alloc.c).
444 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
445 * -2 so that an underflow of the page_mapcount() won't be mistaken
446 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
447 * efficiently by most CPU architectures.
449 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
451 static inline int PageBuddy(struct page *page)
453 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
456 static inline void __SetPageBuddy(struct page *page)
458 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
459 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
462 static inline void __ClearPageBuddy(struct page *page)
464 VM_BUG_ON(!PageBuddy(page));
465 atomic_set(&page->_mapcount, -1);
468 void put_page(struct page *page);
469 void put_pages_list(struct list_head *pages);
471 void split_page(struct page *page, unsigned int order);
472 int split_free_page(struct page *page);
475 * Compound pages have a destructor function. Provide a
476 * prototype for that function and accessor functions.
477 * These are _only_ valid on the head of a PG_compound page.
479 typedef void compound_page_dtor(struct page *);
481 static inline void set_compound_page_dtor(struct page *page,
482 compound_page_dtor *dtor)
484 page[1].lru.next = (void *)dtor;
487 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
489 return (compound_page_dtor *)page[1].lru.next;
492 static inline int compound_order(struct page *page)
496 return (unsigned long)page[1].lru.prev;
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_CPUPID] | ... | 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_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_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_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_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_CPUPID_MASK ((1UL << LAST_CPUPID_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 mainly used by the buddy allocator for
634 * determining if two pages could be buddies. We are not really identifying
635 * the zone since we could be using the section number id if we do not have
636 * node id available in page flags.
637 * We only guarantee that it will return the same value for two combinable
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 static inline int cpu_pid_to_cpupid(int cpu, int pid)
666 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
669 static inline int cpupid_to_pid(int cpupid)
671 return cpupid & LAST__PID_MASK;
674 static inline int cpupid_to_cpu(int cpupid)
676 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
679 static inline int cpupid_to_nid(int cpupid)
681 return cpu_to_node(cpupid_to_cpu(cpupid));
684 static inline bool cpupid_pid_unset(int cpupid)
686 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
689 static inline bool cpupid_cpu_unset(int cpupid)
691 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
694 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
695 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
697 return xchg(&page->_last_cpupid, cpupid);
700 static inline int page_cpupid_last(struct page *page)
702 return page->_last_cpupid;
704 static inline void page_cpupid_reset_last(struct page *page)
706 page->_last_cpupid = -1;
709 static inline int page_cpupid_last(struct page *page)
711 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
714 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
716 static inline void page_cpupid_reset_last(struct page *page)
718 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
720 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
721 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
723 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
724 #else /* !CONFIG_NUMA_BALANCING */
725 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
727 return page_to_nid(page); /* XXX */
730 static inline int page_cpupid_last(struct page *page)
732 return page_to_nid(page); /* XXX */
735 static inline int cpupid_to_nid(int cpupid)
740 static inline int cpupid_to_pid(int cpupid)
745 static inline int cpupid_to_cpu(int cpupid)
750 static inline int cpu_pid_to_cpupid(int nid, int pid)
755 static inline bool cpupid_pid_unset(int cpupid)
760 static inline void page_cpupid_reset_last(struct page *page)
763 #endif /* CONFIG_NUMA_BALANCING */
765 static inline struct zone *page_zone(const struct page *page)
767 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
770 #ifdef SECTION_IN_PAGE_FLAGS
771 static inline void set_page_section(struct page *page, unsigned long section)
773 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
774 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
777 static inline unsigned long page_to_section(const struct page *page)
779 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
783 static inline void set_page_zone(struct page *page, enum zone_type zone)
785 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
786 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
789 static inline void set_page_node(struct page *page, unsigned long node)
791 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
792 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
795 static inline void set_page_links(struct page *page, enum zone_type zone,
796 unsigned long node, unsigned long pfn)
798 set_page_zone(page, zone);
799 set_page_node(page, node);
800 #ifdef SECTION_IN_PAGE_FLAGS
801 set_page_section(page, pfn_to_section_nr(pfn));
806 * Some inline functions in vmstat.h depend on page_zone()
808 #include <linux/vmstat.h>
810 static __always_inline void *lowmem_page_address(const struct page *page)
812 return __va(PFN_PHYS(page_to_pfn(page)));
815 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
816 #define HASHED_PAGE_VIRTUAL
819 #if defined(WANT_PAGE_VIRTUAL)
820 #define page_address(page) ((page)->virtual)
821 #define set_page_address(page, address) \
823 (page)->virtual = (address); \
825 #define page_address_init() do { } while(0)
828 #if defined(HASHED_PAGE_VIRTUAL)
829 void *page_address(const struct page *page);
830 void set_page_address(struct page *page, void *virtual);
831 void page_address_init(void);
834 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
835 #define page_address(page) lowmem_page_address(page)
836 #define set_page_address(page, address) do { } while(0)
837 #define page_address_init() do { } while(0)
841 * On an anonymous page mapped into a user virtual memory area,
842 * page->mapping points to its anon_vma, not to a struct address_space;
843 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
845 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
846 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
847 * and then page->mapping points, not to an anon_vma, but to a private
848 * structure which KSM associates with that merged page. See ksm.h.
850 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
852 * Please note that, confusingly, "page_mapping" refers to the inode
853 * address_space which maps the page from disk; whereas "page_mapped"
854 * refers to user virtual address space into which the page is mapped.
856 #define PAGE_MAPPING_ANON 1
857 #define PAGE_MAPPING_KSM 2
858 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
860 extern struct address_space *page_mapping(struct page *page);
862 /* Neutral page->mapping pointer to address_space or anon_vma or other */
863 static inline void *page_rmapping(struct page *page)
865 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
868 extern struct address_space *__page_file_mapping(struct page *);
871 struct address_space *page_file_mapping(struct page *page)
873 if (unlikely(PageSwapCache(page)))
874 return __page_file_mapping(page);
876 return page->mapping;
879 static inline int PageAnon(struct page *page)
881 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
885 * Return the pagecache index of the passed page. Regular pagecache pages
886 * use ->index whereas swapcache pages use ->private
888 static inline pgoff_t page_index(struct page *page)
890 if (unlikely(PageSwapCache(page)))
891 return page_private(page);
895 extern pgoff_t __page_file_index(struct page *page);
898 * Return the file index of the page. Regular pagecache pages use ->index
899 * whereas swapcache pages use swp_offset(->private)
901 static inline pgoff_t page_file_index(struct page *page)
903 if (unlikely(PageSwapCache(page)))
904 return __page_file_index(page);
910 * Return true if this page is mapped into pagetables.
912 static inline int page_mapped(struct page *page)
914 return atomic_read(&(page)->_mapcount) >= 0;
918 * Different kinds of faults, as returned by handle_mm_fault().
919 * Used to decide whether a process gets delivered SIGBUS or
920 * just gets major/minor fault counters bumped up.
923 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
925 #define VM_FAULT_OOM 0x0001
926 #define VM_FAULT_SIGBUS 0x0002
927 #define VM_FAULT_MAJOR 0x0004
928 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
929 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
930 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
932 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
933 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
934 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
935 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
937 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
939 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
940 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
942 /* Encode hstate index for a hwpoisoned large page */
943 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
944 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
947 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
949 extern void pagefault_out_of_memory(void);
951 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
954 * Flags passed to show_mem() and show_free_areas() to suppress output in
957 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
958 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
960 extern void show_free_areas(unsigned int flags);
961 extern bool skip_free_areas_node(unsigned int flags, int nid);
963 int shmem_zero_setup(struct vm_area_struct *);
965 extern int can_do_mlock(void);
966 extern int user_shm_lock(size_t, struct user_struct *);
967 extern void user_shm_unlock(size_t, struct user_struct *);
970 * Parameter block passed down to zap_pte_range in exceptional cases.
973 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
974 struct address_space *check_mapping; /* Check page->mapping if set */
975 pgoff_t first_index; /* Lowest page->index to unmap */
976 pgoff_t last_index; /* Highest page->index to unmap */
979 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
982 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
984 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
985 unsigned long size, struct zap_details *);
986 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
987 unsigned long start, unsigned long end);
990 * mm_walk - callbacks for walk_page_range
991 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
992 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
993 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
994 * this handler is required to be able to handle
995 * pmd_trans_huge() pmds. They may simply choose to
996 * split_huge_page() instead of handling it explicitly.
997 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
998 * @pte_hole: if set, called for each hole at all levels
999 * @hugetlb_entry: if set, called for each hugetlb entry
1000 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1003 * (see walk_page_range for more details)
1006 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1007 unsigned long next, struct mm_walk *walk);
1008 int (*pud_entry)(pud_t *pud, unsigned long addr,
1009 unsigned long next, struct mm_walk *walk);
1010 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1011 unsigned long next, struct mm_walk *walk);
1012 int (*pte_entry)(pte_t *pte, unsigned long addr,
1013 unsigned long next, struct mm_walk *walk);
1014 int (*pte_hole)(unsigned long addr, unsigned long next,
1015 struct mm_walk *walk);
1016 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1017 unsigned long addr, unsigned long next,
1018 struct mm_walk *walk);
1019 struct mm_struct *mm;
1023 int walk_page_range(unsigned long addr, unsigned long end,
1024 struct mm_walk *walk);
1025 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1026 unsigned long end, unsigned long floor, unsigned long ceiling);
1027 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1028 struct vm_area_struct *vma);
1029 void unmap_mapping_range(struct address_space *mapping,
1030 loff_t const holebegin, loff_t const holelen, int even_cows);
1031 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1032 unsigned long *pfn);
1033 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1034 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1035 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1036 void *buf, int len, int write);
1038 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1039 loff_t const holebegin, loff_t const holelen)
1041 unmap_mapping_range(mapping, holebegin, holelen, 0);
1044 extern void truncate_pagecache(struct inode *inode, loff_t new);
1045 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1046 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1047 int truncate_inode_page(struct address_space *mapping, struct page *page);
1048 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1049 int invalidate_inode_page(struct page *page);
1052 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1053 unsigned long address, unsigned int flags);
1054 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1055 unsigned long address, unsigned int fault_flags);
1057 static inline int handle_mm_fault(struct mm_struct *mm,
1058 struct vm_area_struct *vma, unsigned long address,
1061 /* should never happen if there's no MMU */
1063 return VM_FAULT_SIGBUS;
1065 static inline int fixup_user_fault(struct task_struct *tsk,
1066 struct mm_struct *mm, unsigned long address,
1067 unsigned int fault_flags)
1069 /* should never happen if there's no MMU */
1075 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1076 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1077 void *buf, int len, int write);
1079 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1080 unsigned long start, unsigned long nr_pages,
1081 unsigned int foll_flags, struct page **pages,
1082 struct vm_area_struct **vmas, int *nonblocking);
1083 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1084 unsigned long start, unsigned long nr_pages,
1085 int write, int force, struct page **pages,
1086 struct vm_area_struct **vmas);
1087 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1088 struct page **pages);
1090 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1091 struct page **pages);
1092 int get_kernel_page(unsigned long start, int write, struct page **pages);
1093 struct page *get_dump_page(unsigned long addr);
1095 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1096 extern void do_invalidatepage(struct page *page, unsigned int offset,
1097 unsigned int length);
1099 int __set_page_dirty_nobuffers(struct page *page);
1100 int __set_page_dirty_no_writeback(struct page *page);
1101 int redirty_page_for_writepage(struct writeback_control *wbc,
1103 void account_page_dirtied(struct page *page, struct address_space *mapping);
1104 void account_page_writeback(struct page *page);
1105 int set_page_dirty(struct page *page);
1106 int set_page_dirty_lock(struct page *page);
1107 int clear_page_dirty_for_io(struct page *page);
1109 /* Is the vma a continuation of the stack vma above it? */
1110 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1112 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1115 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1118 return (vma->vm_flags & VM_GROWSDOWN) &&
1119 (vma->vm_start == addr) &&
1120 !vma_growsdown(vma->vm_prev, addr);
1123 /* Is the vma a continuation of the stack vma below it? */
1124 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1126 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1129 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1132 return (vma->vm_flags & VM_GROWSUP) &&
1133 (vma->vm_end == addr) &&
1134 !vma_growsup(vma->vm_next, addr);
1138 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1140 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1141 unsigned long old_addr, struct vm_area_struct *new_vma,
1142 unsigned long new_addr, unsigned long len,
1143 bool need_rmap_locks);
1144 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1145 unsigned long end, pgprot_t newprot,
1146 int dirty_accountable, int prot_numa);
1147 extern int mprotect_fixup(struct vm_area_struct *vma,
1148 struct vm_area_struct **pprev, unsigned long start,
1149 unsigned long end, unsigned long newflags);
1152 * doesn't attempt to fault and will return short.
1154 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1155 struct page **pages);
1157 * per-process(per-mm_struct) statistics.
1159 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1161 long val = atomic_long_read(&mm->rss_stat.count[member]);
1163 #ifdef SPLIT_RSS_COUNTING
1165 * counter is updated in asynchronous manner and may go to minus.
1166 * But it's never be expected number for users.
1171 return (unsigned long)val;
1174 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1176 atomic_long_add(value, &mm->rss_stat.count[member]);
1179 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1181 atomic_long_inc(&mm->rss_stat.count[member]);
1184 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1186 atomic_long_dec(&mm->rss_stat.count[member]);
1189 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1191 return get_mm_counter(mm, MM_FILEPAGES) +
1192 get_mm_counter(mm, MM_ANONPAGES);
1195 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1197 return max(mm->hiwater_rss, get_mm_rss(mm));
1200 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1202 return max(mm->hiwater_vm, mm->total_vm);
1205 static inline void update_hiwater_rss(struct mm_struct *mm)
1207 unsigned long _rss = get_mm_rss(mm);
1209 if ((mm)->hiwater_rss < _rss)
1210 (mm)->hiwater_rss = _rss;
1213 static inline void update_hiwater_vm(struct mm_struct *mm)
1215 if (mm->hiwater_vm < mm->total_vm)
1216 mm->hiwater_vm = mm->total_vm;
1219 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1220 struct mm_struct *mm)
1222 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1224 if (*maxrss < hiwater_rss)
1225 *maxrss = hiwater_rss;
1228 #if defined(SPLIT_RSS_COUNTING)
1229 void sync_mm_rss(struct mm_struct *mm);
1231 static inline void sync_mm_rss(struct mm_struct *mm)
1236 int vma_wants_writenotify(struct vm_area_struct *vma);
1238 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1240 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1244 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1248 #ifdef __PAGETABLE_PUD_FOLDED
1249 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1250 unsigned long address)
1255 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1258 #ifdef __PAGETABLE_PMD_FOLDED
1259 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1260 unsigned long address)
1265 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1268 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1269 pmd_t *pmd, unsigned long address);
1270 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1273 * The following ifdef needed to get the 4level-fixup.h header to work.
1274 * Remove it when 4level-fixup.h has been removed.
1276 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1277 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1279 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1280 NULL: pud_offset(pgd, address);
1283 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1285 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1286 NULL: pmd_offset(pud, address);
1288 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1290 #if USE_SPLIT_PTLOCKS
1292 * We tuck a spinlock to guard each pagetable page into its struct page,
1293 * at page->private, with BUILD_BUG_ON to make sure that this will not
1294 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1295 * When freeing, reset page->mapping so free_pages_check won't complain.
1297 #define __pte_lockptr(page) &((page)->ptl)
1298 #define pte_lock_init(_page) do { \
1299 spin_lock_init(__pte_lockptr(_page)); \
1301 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1302 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1303 #else /* !USE_SPLIT_PTLOCKS */
1305 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1307 #define pte_lock_init(page) do {} while (0)
1308 #define pte_lock_deinit(page) do {} while (0)
1309 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1310 #endif /* USE_SPLIT_PTLOCKS */
1312 static inline void pgtable_page_ctor(struct page *page)
1314 pte_lock_init(page);
1315 inc_zone_page_state(page, NR_PAGETABLE);
1318 static inline void pgtable_page_dtor(struct page *page)
1320 pte_lock_deinit(page);
1321 dec_zone_page_state(page, NR_PAGETABLE);
1324 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1326 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1327 pte_t *__pte = pte_offset_map(pmd, address); \
1333 #define pte_unmap_unlock(pte, ptl) do { \
1338 #define pte_alloc_map(mm, vma, pmd, address) \
1339 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1341 NULL: pte_offset_map(pmd, address))
1343 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1344 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1346 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1348 #define pte_alloc_kernel(pmd, address) \
1349 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1350 NULL: pte_offset_kernel(pmd, address))
1352 extern void free_area_init(unsigned long * zones_size);
1353 extern void free_area_init_node(int nid, unsigned long * zones_size,
1354 unsigned long zone_start_pfn, unsigned long *zholes_size);
1355 extern void free_initmem(void);
1358 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1359 * into the buddy system. The freed pages will be poisoned with pattern
1360 * "poison" if it's within range [0, UCHAR_MAX].
1361 * Return pages freed into the buddy system.
1363 extern unsigned long free_reserved_area(void *start, void *end,
1364 int poison, char *s);
1366 #ifdef CONFIG_HIGHMEM
1368 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1369 * and totalram_pages.
1371 extern void free_highmem_page(struct page *page);
1374 extern void adjust_managed_page_count(struct page *page, long count);
1375 extern void mem_init_print_info(const char *str);
1377 /* Free the reserved page into the buddy system, so it gets managed. */
1378 static inline void __free_reserved_page(struct page *page)
1380 ClearPageReserved(page);
1381 init_page_count(page);
1385 static inline void free_reserved_page(struct page *page)
1387 __free_reserved_page(page);
1388 adjust_managed_page_count(page, 1);
1391 static inline void mark_page_reserved(struct page *page)
1393 SetPageReserved(page);
1394 adjust_managed_page_count(page, -1);
1398 * Default method to free all the __init memory into the buddy system.
1399 * The freed pages will be poisoned with pattern "poison" if it's within
1400 * range [0, UCHAR_MAX].
1401 * Return pages freed into the buddy system.
1403 static inline unsigned long free_initmem_default(int poison)
1405 extern char __init_begin[], __init_end[];
1407 return free_reserved_area(&__init_begin, &__init_end,
1408 poison, "unused kernel");
1411 static inline unsigned long get_num_physpages(void)
1414 unsigned long phys_pages = 0;
1416 for_each_online_node(nid)
1417 phys_pages += node_present_pages(nid);
1422 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1424 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1425 * zones, allocate the backing mem_map and account for memory holes in a more
1426 * architecture independent manner. This is a substitute for creating the
1427 * zone_sizes[] and zholes_size[] arrays and passing them to
1428 * free_area_init_node()
1430 * An architecture is expected to register range of page frames backed by
1431 * physical memory with memblock_add[_node]() before calling
1432 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1433 * usage, an architecture is expected to do something like
1435 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1437 * for_each_valid_physical_page_range()
1438 * memblock_add_node(base, size, nid)
1439 * free_area_init_nodes(max_zone_pfns);
1441 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1442 * registered physical page range. Similarly
1443 * sparse_memory_present_with_active_regions() calls memory_present() for
1444 * each range when SPARSEMEM is enabled.
1446 * See mm/page_alloc.c for more information on each function exposed by
1447 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1449 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1450 unsigned long node_map_pfn_alignment(void);
1451 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1452 unsigned long end_pfn);
1453 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1454 unsigned long end_pfn);
1455 extern void get_pfn_range_for_nid(unsigned int nid,
1456 unsigned long *start_pfn, unsigned long *end_pfn);
1457 extern unsigned long find_min_pfn_with_active_regions(void);
1458 extern void free_bootmem_with_active_regions(int nid,
1459 unsigned long max_low_pfn);
1460 extern void sparse_memory_present_with_active_regions(int nid);
1462 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1464 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1465 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1466 static inline int __early_pfn_to_nid(unsigned long pfn)
1471 /* please see mm/page_alloc.c */
1472 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1473 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1474 /* there is a per-arch backend function. */
1475 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1476 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1479 extern void set_dma_reserve(unsigned long new_dma_reserve);
1480 extern void memmap_init_zone(unsigned long, int, unsigned long,
1481 unsigned long, enum memmap_context);
1482 extern void setup_per_zone_wmarks(void);
1483 extern int __meminit init_per_zone_wmark_min(void);
1484 extern void mem_init(void);
1485 extern void __init mmap_init(void);
1486 extern void show_mem(unsigned int flags);
1487 extern void si_meminfo(struct sysinfo * val);
1488 extern void si_meminfo_node(struct sysinfo *val, int nid);
1490 extern __printf(3, 4)
1491 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1493 extern void setup_per_cpu_pageset(void);
1495 extern void zone_pcp_update(struct zone *zone);
1496 extern void zone_pcp_reset(struct zone *zone);
1499 extern int min_free_kbytes;
1502 extern atomic_long_t mmap_pages_allocated;
1503 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1505 /* interval_tree.c */
1506 void vma_interval_tree_insert(struct vm_area_struct *node,
1507 struct rb_root *root);
1508 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1509 struct vm_area_struct *prev,
1510 struct rb_root *root);
1511 void vma_interval_tree_remove(struct vm_area_struct *node,
1512 struct rb_root *root);
1513 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1514 unsigned long start, unsigned long last);
1515 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1516 unsigned long start, unsigned long last);
1518 #define vma_interval_tree_foreach(vma, root, start, last) \
1519 for (vma = vma_interval_tree_iter_first(root, start, last); \
1520 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1522 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1523 struct list_head *list)
1525 list_add_tail(&vma->shared.nonlinear, list);
1528 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1529 struct rb_root *root);
1530 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1531 struct rb_root *root);
1532 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1533 struct rb_root *root, unsigned long start, unsigned long last);
1534 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1535 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1536 #ifdef CONFIG_DEBUG_VM_RB
1537 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1540 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1541 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1542 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1545 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1546 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1547 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1548 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1549 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1550 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1551 struct mempolicy *);
1552 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1553 extern int split_vma(struct mm_struct *,
1554 struct vm_area_struct *, unsigned long addr, int new_below);
1555 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1556 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1557 struct rb_node **, struct rb_node *);
1558 extern void unlink_file_vma(struct vm_area_struct *);
1559 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1560 unsigned long addr, unsigned long len, pgoff_t pgoff,
1561 bool *need_rmap_locks);
1562 extern void exit_mmap(struct mm_struct *);
1564 extern int mm_take_all_locks(struct mm_struct *mm);
1565 extern void mm_drop_all_locks(struct mm_struct *mm);
1567 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1568 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1570 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1571 extern int install_special_mapping(struct mm_struct *mm,
1572 unsigned long addr, unsigned long len,
1573 unsigned long flags, struct page **pages);
1575 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1577 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1578 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1579 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1580 unsigned long len, unsigned long prot, unsigned long flags,
1581 unsigned long pgoff, unsigned long *populate);
1582 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1585 extern int __mm_populate(unsigned long addr, unsigned long len,
1587 static inline void mm_populate(unsigned long addr, unsigned long len)
1590 (void) __mm_populate(addr, len, 1);
1593 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1596 /* These take the mm semaphore themselves */
1597 extern unsigned long vm_brk(unsigned long, unsigned long);
1598 extern int vm_munmap(unsigned long, size_t);
1599 extern unsigned long vm_mmap(struct file *, unsigned long,
1600 unsigned long, unsigned long,
1601 unsigned long, unsigned long);
1603 struct vm_unmapped_area_info {
1604 #define VM_UNMAPPED_AREA_TOPDOWN 1
1605 unsigned long flags;
1606 unsigned long length;
1607 unsigned long low_limit;
1608 unsigned long high_limit;
1609 unsigned long align_mask;
1610 unsigned long align_offset;
1613 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1614 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1617 * Search for an unmapped address range.
1619 * We are looking for a range that:
1620 * - does not intersect with any VMA;
1621 * - is contained within the [low_limit, high_limit) interval;
1622 * - is at least the desired size.
1623 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1625 static inline unsigned long
1626 vm_unmapped_area(struct vm_unmapped_area_info *info)
1628 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1629 return unmapped_area(info);
1631 return unmapped_area_topdown(info);
1635 extern void truncate_inode_pages(struct address_space *, loff_t);
1636 extern void truncate_inode_pages_range(struct address_space *,
1637 loff_t lstart, loff_t lend);
1639 /* generic vm_area_ops exported for stackable file systems */
1640 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1641 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1643 /* mm/page-writeback.c */
1644 int write_one_page(struct page *page, int wait);
1645 void task_dirty_inc(struct task_struct *tsk);
1648 #define VM_MAX_READAHEAD 128 /* kbytes */
1649 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1651 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1652 pgoff_t offset, unsigned long nr_to_read);
1654 void page_cache_sync_readahead(struct address_space *mapping,
1655 struct file_ra_state *ra,
1658 unsigned long size);
1660 void page_cache_async_readahead(struct address_space *mapping,
1661 struct file_ra_state *ra,
1665 unsigned long size);
1667 unsigned long max_sane_readahead(unsigned long nr);
1668 unsigned long ra_submit(struct file_ra_state *ra,
1669 struct address_space *mapping,
1672 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1673 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1675 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1676 extern int expand_downwards(struct vm_area_struct *vma,
1677 unsigned long address);
1679 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1681 #define expand_upwards(vma, address) do { } while (0)
1684 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1685 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1686 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1687 struct vm_area_struct **pprev);
1689 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1690 NULL if none. Assume start_addr < end_addr. */
1691 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1693 struct vm_area_struct * vma = find_vma(mm,start_addr);
1695 if (vma && end_addr <= vma->vm_start)
1700 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1702 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1705 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1706 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1707 unsigned long vm_start, unsigned long vm_end)
1709 struct vm_area_struct *vma = find_vma(mm, vm_start);
1711 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1718 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1720 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1726 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1727 unsigned long change_prot_numa(struct vm_area_struct *vma,
1728 unsigned long start, unsigned long end);
1731 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1732 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1733 unsigned long pfn, unsigned long size, pgprot_t);
1734 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1735 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1737 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1739 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1742 struct page *follow_page_mask(struct vm_area_struct *vma,
1743 unsigned long address, unsigned int foll_flags,
1744 unsigned int *page_mask);
1746 static inline struct page *follow_page(struct vm_area_struct *vma,
1747 unsigned long address, unsigned int foll_flags)
1749 unsigned int unused_page_mask;
1750 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1753 #define FOLL_WRITE 0x01 /* check pte is writable */
1754 #define FOLL_TOUCH 0x02 /* mark page accessed */
1755 #define FOLL_GET 0x04 /* do get_page on page */
1756 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1757 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1758 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1759 * and return without waiting upon it */
1760 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1761 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1762 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1763 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1764 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1766 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1768 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1769 unsigned long size, pte_fn_t fn, void *data);
1771 #ifdef CONFIG_PROC_FS
1772 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1774 static inline void vm_stat_account(struct mm_struct *mm,
1775 unsigned long flags, struct file *file, long pages)
1777 mm->total_vm += pages;
1779 #endif /* CONFIG_PROC_FS */
1781 #ifdef CONFIG_DEBUG_PAGEALLOC
1782 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1783 #ifdef CONFIG_HIBERNATION
1784 extern bool kernel_page_present(struct page *page);
1785 #endif /* CONFIG_HIBERNATION */
1788 kernel_map_pages(struct page *page, int numpages, int enable) {}
1789 #ifdef CONFIG_HIBERNATION
1790 static inline bool kernel_page_present(struct page *page) { return true; }
1791 #endif /* CONFIG_HIBERNATION */
1794 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1795 #ifdef __HAVE_ARCH_GATE_AREA
1796 int in_gate_area_no_mm(unsigned long addr);
1797 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1799 int in_gate_area_no_mm(unsigned long addr);
1800 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1801 #endif /* __HAVE_ARCH_GATE_AREA */
1803 #ifdef CONFIG_SYSCTL
1804 extern int sysctl_drop_caches;
1805 int drop_caches_sysctl_handler(struct ctl_table *, int,
1806 void __user *, size_t *, loff_t *);
1809 unsigned long shrink_slab(struct shrink_control *shrink,
1810 unsigned long nr_pages_scanned,
1811 unsigned long lru_pages);
1814 #define randomize_va_space 0
1816 extern int randomize_va_space;
1819 const char * arch_vma_name(struct vm_area_struct *vma);
1820 void print_vma_addr(char *prefix, unsigned long rip);
1822 void sparse_mem_maps_populate_node(struct page **map_map,
1823 unsigned long pnum_begin,
1824 unsigned long pnum_end,
1825 unsigned long map_count,
1828 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1829 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1830 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1831 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1832 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1833 void *vmemmap_alloc_block(unsigned long size, int node);
1834 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1835 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1836 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1838 int vmemmap_populate(unsigned long start, unsigned long end, int node);
1839 void vmemmap_populate_print_last(void);
1840 #ifdef CONFIG_MEMORY_HOTPLUG
1841 void vmemmap_free(unsigned long start, unsigned long end);
1843 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1844 unsigned long size);
1847 MF_COUNT_INCREASED = 1 << 0,
1848 MF_ACTION_REQUIRED = 1 << 1,
1849 MF_MUST_KILL = 1 << 2,
1850 MF_SOFT_OFFLINE = 1 << 3,
1852 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1853 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1854 extern int unpoison_memory(unsigned long pfn);
1855 extern int sysctl_memory_failure_early_kill;
1856 extern int sysctl_memory_failure_recovery;
1857 extern void shake_page(struct page *p, int access);
1858 extern atomic_long_t num_poisoned_pages;
1859 extern int soft_offline_page(struct page *page, int flags);
1861 extern void dump_page(struct page *page);
1863 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1864 extern void clear_huge_page(struct page *page,
1866 unsigned int pages_per_huge_page);
1867 extern void copy_user_huge_page(struct page *dst, struct page *src,
1868 unsigned long addr, struct vm_area_struct *vma,
1869 unsigned int pages_per_huge_page);
1870 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1872 #ifdef CONFIG_DEBUG_PAGEALLOC
1873 extern unsigned int _debug_guardpage_minorder;
1875 static inline unsigned int debug_guardpage_minorder(void)
1877 return _debug_guardpage_minorder;
1880 static inline bool page_is_guard(struct page *page)
1882 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1885 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1886 static inline bool page_is_guard(struct page *page) { return false; }
1887 #endif /* CONFIG_DEBUG_PAGEALLOC */
1889 #if MAX_NUMNODES > 1
1890 void __init setup_nr_node_ids(void);
1892 static inline void setup_nr_node_ids(void) {}
1895 #endif /* __KERNEL__ */
1896 #endif /* _LINUX_MM_H */