4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/prio_tree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
26 struct writeback_control;
28 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
32 extern unsigned long num_physpages;
33 extern unsigned long totalram_pages;
34 extern void * high_memory;
35 extern int page_cluster;
38 extern int sysctl_legacy_va_layout;
40 #define sysctl_legacy_va_layout 0
44 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
49 /* to align the pointer to the (next) page boundary */
50 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
53 * Linux kernel virtual memory manager primitives.
54 * The idea being to have a "virtual" mm in the same way
55 * we have a virtual fs - giving a cleaner interface to the
56 * mm details, and allowing different kinds of memory mappings
57 * (from shared memory to executable loading to arbitrary
61 extern struct kmem_cache *vm_area_cachep;
64 extern struct rb_root nommu_region_tree;
65 extern struct rw_semaphore nommu_region_sem;
67 extern unsigned int kobjsize(const void *objp);
71 * vm_flags in vm_area_struct, see mm_types.h.
73 #define VM_READ 0x00000001 /* currently active flags */
74 #define VM_WRITE 0x00000002
75 #define VM_EXEC 0x00000004
76 #define VM_SHARED 0x00000008
78 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
79 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
80 #define VM_MAYWRITE 0x00000020
81 #define VM_MAYEXEC 0x00000040
82 #define VM_MAYSHARE 0x00000080
84 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
85 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
86 #define VM_GROWSUP 0x00000200
88 #define VM_GROWSUP 0x00000000
89 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
91 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
92 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
94 #define VM_EXECUTABLE 0x00001000
95 #define VM_LOCKED 0x00002000
96 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
98 /* Used by sys_madvise() */
99 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
100 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
102 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
103 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
104 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
105 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
106 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
107 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
108 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
109 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
110 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
112 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
114 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
115 #define VM_NODUMP 0x04000000 /* Do not include in the core dump */
117 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
118 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
120 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
123 #define VM_FOREIGN 0x20000000 /* Has pages belonging to another VM */
125 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
126 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
128 /* Bits set in the VMA until the stack is in its final location */
129 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
131 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
132 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
135 #ifdef CONFIG_STACK_GROWSUP
136 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
138 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
141 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
142 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
143 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
144 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
145 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
148 * Special vmas that are non-mergable, non-mlock()able.
149 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
151 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
154 struct vm_foreign_map {
160 * mapping from the currently active vm_flags protection bits (the
161 * low four bits) to a page protection mask..
163 extern pgprot_t protection_map[16];
165 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
166 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
167 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
168 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
169 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
170 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
173 * This interface is used by x86 PAT code to identify a pfn mapping that is
174 * linear over entire vma. This is to optimize PAT code that deals with
175 * marking the physical region with a particular prot. This is not for generic
176 * mm use. Note also that this check will not work if the pfn mapping is
177 * linear for a vma starting at physical address 0. In which case PAT code
178 * falls back to slow path of reserving physical range page by page.
180 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
182 return !!(vma->vm_flags & VM_PFN_AT_MMAP);
185 static inline int is_pfn_mapping(struct vm_area_struct *vma)
187 return !!(vma->vm_flags & VM_PFNMAP);
191 * vm_fault is filled by the the pagefault handler and passed to the vma's
192 * ->fault function. The vma's ->fault is responsible for returning a bitmask
193 * of VM_FAULT_xxx flags that give details about how the fault was handled.
195 * pgoff should be used in favour of virtual_address, if possible. If pgoff
196 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
200 unsigned int flags; /* FAULT_FLAG_xxx flags */
201 pgoff_t pgoff; /* Logical page offset based on vma */
202 void __user *virtual_address; /* Faulting virtual address */
204 struct page *page; /* ->fault handlers should return a
205 * page here, unless VM_FAULT_NOPAGE
206 * is set (which is also implied by
212 * These are the virtual MM functions - opening of an area, closing and
213 * unmapping it (needed to keep files on disk up-to-date etc), pointer
214 * to the functions called when a no-page or a wp-page exception occurs.
216 struct vm_operations_struct {
217 void (*open)(struct vm_area_struct * area);
218 void (*close)(struct vm_area_struct * area);
219 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
221 /* notification that a previously read-only page is about to become
222 * writable, if an error is returned it will cause a SIGBUS */
223 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
225 /* called by access_process_vm when get_user_pages() fails, typically
226 * for use by special VMAs that can switch between memory and hardware
228 int (*access)(struct vm_area_struct *vma, unsigned long addr,
229 void *buf, int len, int write);
232 /* Area-specific function for clearing the PTE at @ptep. Returns the
233 * original value of @ptep. */
234 pte_t (*zap_pte)(struct vm_area_struct *vma,
235 unsigned long addr, pte_t *ptep, int is_fullmm);
237 /* called before close() to indicate no more pages should be mapped */
238 void (*unmap)(struct vm_area_struct *area);
243 * set_policy() op must add a reference to any non-NULL @new mempolicy
244 * to hold the policy upon return. Caller should pass NULL @new to
245 * remove a policy and fall back to surrounding context--i.e. do not
246 * install a MPOL_DEFAULT policy, nor the task or system default
249 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
252 * get_policy() op must add reference [mpol_get()] to any policy at
253 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
254 * in mm/mempolicy.c will do this automatically.
255 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
256 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
257 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
258 * must return NULL--i.e., do not "fallback" to task or system default
261 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
263 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
264 const nodemask_t *to, unsigned long flags);
271 #define page_private(page) ((page)->private)
272 #define set_page_private(page, v) ((page)->private = (v))
275 * FIXME: take this include out, include page-flags.h in
276 * files which need it (119 of them)
278 #include <linux/page-flags.h>
279 #include <linux/huge_mm.h>
282 * Methods to modify the page usage count.
284 * What counts for a page usage:
285 * - cache mapping (page->mapping)
286 * - private data (page->private)
287 * - page mapped in a task's page tables, each mapping
288 * is counted separately
290 * Also, many kernel routines increase the page count before a critical
291 * routine so they can be sure the page doesn't go away from under them.
295 * Drop a ref, return true if the refcount fell to zero (the page has no users)
297 static inline int put_page_testzero(struct page *page)
299 VM_BUG_ON(atomic_read(&page->_count) == 0);
300 return atomic_dec_and_test(&page->_count);
304 * Try to grab a ref unless the page has a refcount of zero, return false if
307 static inline int get_page_unless_zero(struct page *page)
309 return atomic_inc_not_zero(&page->_count);
312 extern int page_is_ram(unsigned long pfn);
314 /* Support for virtually mapped pages */
315 struct page *vmalloc_to_page(const void *addr);
316 unsigned long vmalloc_to_pfn(const void *addr);
319 * Determine if an address is within the vmalloc range
321 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
322 * is no special casing required.
324 static inline int is_vmalloc_addr(const void *x)
327 unsigned long addr = (unsigned long)x;
329 return addr >= VMALLOC_START && addr < VMALLOC_END;
335 extern int is_vmalloc_or_module_addr(const void *x);
337 static inline int is_vmalloc_or_module_addr(const void *x)
343 static inline void compound_lock(struct page *page)
345 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
346 bit_spin_lock(PG_compound_lock, &page->flags);
350 static inline void compound_unlock(struct page *page)
352 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
353 bit_spin_unlock(PG_compound_lock, &page->flags);
357 static inline unsigned long compound_lock_irqsave(struct page *page)
359 unsigned long uninitialized_var(flags);
360 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
361 local_irq_save(flags);
367 static inline void compound_unlock_irqrestore(struct page *page,
370 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
371 compound_unlock(page);
372 local_irq_restore(flags);
376 static inline struct page *compound_head(struct page *page)
378 if (unlikely(PageTail(page)))
379 return page->first_page;
384 * The atomic page->_mapcount, starts from -1: so that transitions
385 * both from it and to it can be tracked, using atomic_inc_and_test
386 * and atomic_add_negative(-1).
388 static inline void reset_page_mapcount(struct page *page)
390 atomic_set(&(page)->_mapcount, -1);
393 static inline int page_mapcount(struct page *page)
395 return atomic_read(&(page)->_mapcount) + 1;
398 static inline int page_count(struct page *page)
400 return atomic_read(&compound_head(page)->_count);
403 static inline void get_huge_page_tail(struct page *page)
406 * __split_huge_page_refcount() cannot run
409 VM_BUG_ON(page_mapcount(page) < 0);
410 VM_BUG_ON(atomic_read(&page->_count) != 0);
411 atomic_inc(&page->_mapcount);
414 extern bool __get_page_tail(struct page *page);
416 static inline void get_page(struct page *page)
418 if (unlikely(PageTail(page)))
419 if (likely(__get_page_tail(page)))
422 * Getting a normal page or the head of a compound page
423 * requires to already have an elevated page->_count.
425 VM_BUG_ON(atomic_read(&page->_count) <= 0);
426 atomic_inc(&page->_count);
429 static inline struct page *virt_to_head_page(const void *x)
431 struct page *page = virt_to_page(x);
432 return compound_head(page);
436 * Setup the page count before being freed into the page allocator for
437 * the first time (boot or memory hotplug)
439 static inline void init_page_count(struct page *page)
441 atomic_set(&page->_count, 1);
445 * PageBuddy() indicate that the page is free and in the buddy system
446 * (see mm/page_alloc.c).
448 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
449 * -2 so that an underflow of the page_mapcount() won't be mistaken
450 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
451 * efficiently by most CPU architectures.
453 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
455 static inline int PageBuddy(struct page *page)
457 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
460 static inline void __SetPageBuddy(struct page *page)
462 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
463 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
466 static inline void __ClearPageBuddy(struct page *page)
468 VM_BUG_ON(!PageBuddy(page));
469 atomic_set(&page->_mapcount, -1);
472 void put_page(struct page *page);
473 void put_pages_list(struct list_head *pages);
475 void split_page(struct page *page, unsigned int order);
476 int split_free_page(struct page *page);
479 * Compound pages have a destructor function. Provide a
480 * prototype for that function and accessor functions.
481 * These are _only_ valid on the head of a PG_compound page.
483 typedef void compound_page_dtor(struct page *);
485 static inline void set_compound_page_dtor(struct page *page,
486 compound_page_dtor *dtor)
488 page[1].lru.next = (void *)dtor;
491 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
493 return (compound_page_dtor *)page[1].lru.next;
496 static inline int compound_order(struct page *page)
500 return (unsigned long)page[1].lru.prev;
503 static inline int compound_trans_order(struct page *page)
511 flags = compound_lock_irqsave(page);
512 order = compound_order(page);
513 compound_unlock_irqrestore(page, flags);
517 static inline void set_compound_order(struct page *page, unsigned long order)
519 page[1].lru.prev = (void *)order;
524 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
525 * servicing faults for write access. In the normal case, do always want
526 * pte_mkwrite. But get_user_pages can cause write faults for mappings
527 * that do not have writing enabled, when used by access_process_vm.
529 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
531 if (likely(vma->vm_flags & VM_WRITE))
532 pte = pte_mkwrite(pte);
538 * Multiple processes may "see" the same page. E.g. for untouched
539 * mappings of /dev/null, all processes see the same page full of
540 * zeroes, and text pages of executables and shared libraries have
541 * only one copy in memory, at most, normally.
543 * For the non-reserved pages, page_count(page) denotes a reference count.
544 * page_count() == 0 means the page is free. page->lru is then used for
545 * freelist management in the buddy allocator.
546 * page_count() > 0 means the page has been allocated.
548 * Pages are allocated by the slab allocator in order to provide memory
549 * to kmalloc and kmem_cache_alloc. In this case, the management of the
550 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
551 * unless a particular usage is carefully commented. (the responsibility of
552 * freeing the kmalloc memory is the caller's, of course).
554 * A page may be used by anyone else who does a __get_free_page().
555 * In this case, page_count still tracks the references, and should only
556 * be used through the normal accessor functions. The top bits of page->flags
557 * and page->virtual store page management information, but all other fields
558 * are unused and could be used privately, carefully. The management of this
559 * page is the responsibility of the one who allocated it, and those who have
560 * subsequently been given references to it.
562 * The other pages (we may call them "pagecache pages") are completely
563 * managed by the Linux memory manager: I/O, buffers, swapping etc.
564 * The following discussion applies only to them.
566 * A pagecache page contains an opaque `private' member, which belongs to the
567 * page's address_space. Usually, this is the address of a circular list of
568 * the page's disk buffers. PG_private must be set to tell the VM to call
569 * into the filesystem to release these pages.
571 * A page may belong to an inode's memory mapping. In this case, page->mapping
572 * is the pointer to the inode, and page->index is the file offset of the page,
573 * in units of PAGE_CACHE_SIZE.
575 * If pagecache pages are not associated with an inode, they are said to be
576 * anonymous pages. These may become associated with the swapcache, and in that
577 * case PG_swapcache is set, and page->private is an offset into the swapcache.
579 * In either case (swapcache or inode backed), the pagecache itself holds one
580 * reference to the page. Setting PG_private should also increment the
581 * refcount. The each user mapping also has a reference to the page.
583 * The pagecache pages are stored in a per-mapping radix tree, which is
584 * rooted at mapping->page_tree, and indexed by offset.
585 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
586 * lists, we instead now tag pages as dirty/writeback in the radix tree.
588 * All pagecache pages may be subject to I/O:
589 * - inode pages may need to be read from disk,
590 * - inode pages which have been modified and are MAP_SHARED may need
591 * to be written back to the inode on disk,
592 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
593 * modified may need to be swapped out to swap space and (later) to be read
598 * The zone field is never updated after free_area_init_core()
599 * sets it, so none of the operations on it need to be atomic.
604 * page->flags layout:
606 * There are three possibilities for how page->flags get
607 * laid out. The first is for the normal case, without
608 * sparsemem. The second is for sparsemem when there is
609 * plenty of space for node and section. The last is when
610 * we have run out of space and have to fall back to an
611 * alternate (slower) way of determining the node.
613 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
614 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
615 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
617 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
618 #define SECTIONS_WIDTH SECTIONS_SHIFT
620 #define SECTIONS_WIDTH 0
623 #define ZONES_WIDTH ZONES_SHIFT
625 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
626 #define NODES_WIDTH NODES_SHIFT
628 #ifdef CONFIG_SPARSEMEM_VMEMMAP
629 #error "Vmemmap: No space for nodes field in page flags"
631 #define NODES_WIDTH 0
634 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
635 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
636 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
637 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
640 * We are going to use the flags for the page to node mapping if its in
641 * there. This includes the case where there is no node, so it is implicit.
643 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
644 #define NODE_NOT_IN_PAGE_FLAGS
648 * Define the bit shifts to access each section. For non-existent
649 * sections we define the shift as 0; that plus a 0 mask ensures
650 * the compiler will optimise away reference to them.
652 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
653 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
654 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
656 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
657 #ifdef NODE_NOT_IN_PAGE_FLAGS
658 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
659 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
660 SECTIONS_PGOFF : ZONES_PGOFF)
662 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
663 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
664 NODES_PGOFF : ZONES_PGOFF)
667 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
669 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
670 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
673 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
674 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
675 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
676 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
678 static inline enum zone_type page_zonenum(const struct page *page)
680 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
684 * The identification function is only used by the buddy allocator for
685 * determining if two pages could be buddies. We are not really
686 * identifying a zone since we could be using a the section number
687 * id if we have not node id available in page flags.
688 * We guarantee only that it will return the same value for two
689 * combinable pages in a zone.
691 static inline int page_zone_id(struct page *page)
693 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
696 static inline int zone_to_nid(struct zone *zone)
705 #ifdef NODE_NOT_IN_PAGE_FLAGS
706 extern int page_to_nid(const struct page *page);
708 static inline int page_to_nid(const struct page *page)
710 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
714 static inline struct zone *page_zone(const struct page *page)
716 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
719 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
720 static inline void set_page_section(struct page *page, unsigned long section)
722 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
723 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
726 static inline unsigned long page_to_section(const struct page *page)
728 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
732 static inline void set_page_zone(struct page *page, enum zone_type zone)
734 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
735 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
738 static inline void set_page_node(struct page *page, unsigned long node)
740 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
741 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
744 static inline void set_page_links(struct page *page, enum zone_type zone,
745 unsigned long node, unsigned long pfn)
747 set_page_zone(page, zone);
748 set_page_node(page, node);
749 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
750 set_page_section(page, pfn_to_section_nr(pfn));
755 * Some inline functions in vmstat.h depend on page_zone()
757 #include <linux/vmstat.h>
759 static __always_inline void *lowmem_page_address(const struct page *page)
761 return __va(PFN_PHYS(page_to_pfn(page)));
764 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
765 #define HASHED_PAGE_VIRTUAL
768 #if defined(WANT_PAGE_VIRTUAL)
769 #define page_address(page) ((page)->virtual)
770 #define set_page_address(page, address) \
772 (page)->virtual = (address); \
774 #define page_address_init() do { } while(0)
777 #if defined(HASHED_PAGE_VIRTUAL)
778 void *page_address(const struct page *page);
779 void set_page_address(struct page *page, void *virtual);
780 void page_address_init(void);
783 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
784 #define page_address(page) lowmem_page_address(page)
785 #define set_page_address(page, address) do { } while(0)
786 #define page_address_init() do { } while(0)
790 * On an anonymous page mapped into a user virtual memory area,
791 * page->mapping points to its anon_vma, not to a struct address_space;
792 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
794 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
795 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
796 * and then page->mapping points, not to an anon_vma, but to a private
797 * structure which KSM associates with that merged page. See ksm.h.
799 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
801 * Please note that, confusingly, "page_mapping" refers to the inode
802 * address_space which maps the page from disk; whereas "page_mapped"
803 * refers to user virtual address space into which the page is mapped.
805 #define PAGE_MAPPING_ANON 1
806 #define PAGE_MAPPING_KSM 2
807 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
809 extern struct address_space swapper_space;
810 static inline struct address_space *page_mapping(struct page *page)
812 struct address_space *mapping = page->mapping;
814 VM_BUG_ON(PageSlab(page));
815 if (unlikely(PageSwapCache(page)))
816 mapping = &swapper_space;
817 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
822 /* Neutral page->mapping pointer to address_space or anon_vma or other */
823 static inline void *page_rmapping(struct page *page)
825 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
828 static inline int PageAnon(struct page *page)
830 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
834 * Return the pagecache index of the passed page. Regular pagecache pages
835 * use ->index whereas swapcache pages use ->private
837 static inline pgoff_t page_index(struct page *page)
839 if (unlikely(PageSwapCache(page)))
840 return page_private(page);
845 * Return true if this page is mapped into pagetables.
847 static inline int page_mapped(struct page *page)
849 return atomic_read(&(page)->_mapcount) >= 0;
853 * Different kinds of faults, as returned by handle_mm_fault().
854 * Used to decide whether a process gets delivered SIGBUS or
855 * just gets major/minor fault counters bumped up.
858 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
860 #define VM_FAULT_OOM 0x0001
861 #define VM_FAULT_SIGBUS 0x0002
862 #define VM_FAULT_MAJOR 0x0004
863 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
864 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
865 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
867 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
868 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
869 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
871 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
873 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
874 VM_FAULT_HWPOISON_LARGE)
876 /* Encode hstate index for a hwpoisoned large page */
877 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
878 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
881 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
883 extern void pagefault_out_of_memory(void);
885 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
888 * Flags passed to show_mem() and show_free_areas() to suppress output in
891 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
893 extern void show_free_areas(unsigned int flags);
894 extern bool skip_free_areas_node(unsigned int flags, int nid);
896 int shmem_lock(struct file *file, int lock, struct user_struct *user);
897 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
898 int shmem_zero_setup(struct vm_area_struct *);
900 extern int can_do_mlock(void);
901 extern int user_shm_lock(size_t, struct user_struct *);
902 extern void user_shm_unlock(size_t, struct user_struct *);
905 * Parameter block passed down to zap_pte_range in exceptional cases.
908 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
909 struct address_space *check_mapping; /* Check page->mapping if set */
910 pgoff_t first_index; /* Lowest page->index to unmap */
911 pgoff_t last_index; /* Highest page->index to unmap */
914 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
917 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
919 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
920 unsigned long size, struct zap_details *);
921 void unmap_vmas(struct mmu_gather *tlb,
922 struct vm_area_struct *start_vma, unsigned long start_addr,
923 unsigned long end_addr, unsigned long *nr_accounted,
924 struct zap_details *);
927 * mm_walk - callbacks for walk_page_range
928 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
929 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
930 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
931 * this handler is required to be able to handle
932 * pmd_trans_huge() pmds. They may simply choose to
933 * split_huge_page() instead of handling it explicitly.
934 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
935 * @pte_hole: if set, called for each hole at all levels
936 * @hugetlb_entry: if set, called for each hugetlb entry
937 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
940 * (see walk_page_range for more details)
943 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
944 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
945 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
946 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
947 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
948 int (*hugetlb_entry)(pte_t *, unsigned long,
949 unsigned long, unsigned long, struct mm_walk *);
950 struct mm_struct *mm;
954 int walk_page_range(unsigned long addr, unsigned long end,
955 struct mm_walk *walk);
956 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
957 unsigned long end, unsigned long floor, unsigned long ceiling);
958 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
959 struct vm_area_struct *vma);
960 void unmap_mapping_range(struct address_space *mapping,
961 loff_t const holebegin, loff_t const holelen, int even_cows);
962 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
964 int follow_phys(struct vm_area_struct *vma, unsigned long address,
965 unsigned int flags, unsigned long *prot, resource_size_t *phys);
966 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
967 void *buf, int len, int write);
969 static inline void unmap_shared_mapping_range(struct address_space *mapping,
970 loff_t const holebegin, loff_t const holelen)
972 unmap_mapping_range(mapping, holebegin, holelen, 0);
975 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
976 extern void truncate_setsize(struct inode *inode, loff_t newsize);
977 extern int vmtruncate(struct inode *inode, loff_t offset);
978 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
979 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
980 int truncate_inode_page(struct address_space *mapping, struct page *page);
981 int generic_error_remove_page(struct address_space *mapping, struct page *page);
983 int invalidate_inode_page(struct page *page);
986 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
987 unsigned long address, unsigned int flags);
988 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
989 unsigned long address, unsigned int fault_flags);
991 static inline int handle_mm_fault(struct mm_struct *mm,
992 struct vm_area_struct *vma, unsigned long address,
995 /* should never happen if there's no MMU */
997 return VM_FAULT_SIGBUS;
999 static inline int fixup_user_fault(struct task_struct *tsk,
1000 struct mm_struct *mm, unsigned long address,
1001 unsigned int fault_flags)
1003 /* should never happen if there's no MMU */
1009 extern int make_pages_present(unsigned long addr, unsigned long end);
1010 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1011 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1012 void *buf, int len, int write);
1014 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1015 unsigned long start, int len, unsigned int foll_flags,
1016 struct page **pages, struct vm_area_struct **vmas,
1018 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1019 unsigned long start, int nr_pages, int write, int force,
1020 struct page **pages, struct vm_area_struct **vmas);
1021 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1022 struct page **pages);
1023 struct page *get_dump_page(unsigned long addr);
1025 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1026 extern void do_invalidatepage(struct page *page, unsigned long offset);
1028 int __set_page_dirty_nobuffers(struct page *page);
1029 int __set_page_dirty_no_writeback(struct page *page);
1030 int redirty_page_for_writepage(struct writeback_control *wbc,
1032 void account_page_dirtied(struct page *page, struct address_space *mapping);
1033 void account_page_writeback(struct page *page);
1034 int set_page_dirty(struct page *page);
1035 int set_page_dirty_lock(struct page *page);
1036 int clear_page_dirty_for_io(struct page *page);
1038 /* Is the vma a continuation of the stack vma above it? */
1039 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1041 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1044 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1047 return (vma->vm_flags & VM_GROWSDOWN) &&
1048 (vma->vm_start == addr) &&
1049 !vma_growsdown(vma->vm_prev, addr);
1052 /* Is the vma a continuation of the stack vma below it? */
1053 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1055 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1058 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1061 return (vma->vm_flags & VM_GROWSUP) &&
1062 (vma->vm_end == addr) &&
1063 !vma_growsup(vma->vm_next, addr);
1067 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1069 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1070 unsigned long old_addr, struct vm_area_struct *new_vma,
1071 unsigned long new_addr, unsigned long len);
1072 extern unsigned long do_mremap(unsigned long addr,
1073 unsigned long old_len, unsigned long new_len,
1074 unsigned long flags, unsigned long new_addr);
1075 extern int mprotect_fixup(struct vm_area_struct *vma,
1076 struct vm_area_struct **pprev, unsigned long start,
1077 unsigned long end, unsigned long newflags);
1080 * doesn't attempt to fault and will return short.
1082 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1083 struct page **pages);
1085 * per-process(per-mm_struct) statistics.
1087 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1089 long val = atomic_long_read(&mm->rss_stat.count[member]);
1091 #ifdef SPLIT_RSS_COUNTING
1093 * counter is updated in asynchronous manner and may go to minus.
1094 * But it's never be expected number for users.
1099 return (unsigned long)val;
1102 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1104 atomic_long_add(value, &mm->rss_stat.count[member]);
1107 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1109 atomic_long_inc(&mm->rss_stat.count[member]);
1112 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1114 atomic_long_dec(&mm->rss_stat.count[member]);
1117 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1119 return get_mm_counter(mm, MM_FILEPAGES) +
1120 get_mm_counter(mm, MM_ANONPAGES);
1123 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1125 return max(mm->hiwater_rss, get_mm_rss(mm));
1128 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1130 return max(mm->hiwater_vm, mm->total_vm);
1133 static inline void update_hiwater_rss(struct mm_struct *mm)
1135 unsigned long _rss = get_mm_rss(mm);
1137 if ((mm)->hiwater_rss < _rss)
1138 (mm)->hiwater_rss = _rss;
1141 static inline void update_hiwater_vm(struct mm_struct *mm)
1143 if (mm->hiwater_vm < mm->total_vm)
1144 mm->hiwater_vm = mm->total_vm;
1147 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1148 struct mm_struct *mm)
1150 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1152 if (*maxrss < hiwater_rss)
1153 *maxrss = hiwater_rss;
1156 #if defined(SPLIT_RSS_COUNTING)
1157 void sync_mm_rss(struct mm_struct *mm);
1159 static inline void sync_mm_rss(struct mm_struct *mm)
1164 int vma_wants_writenotify(struct vm_area_struct *vma);
1166 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1168 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1172 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1176 #ifdef __PAGETABLE_PUD_FOLDED
1177 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1178 unsigned long address)
1183 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1186 #ifdef __PAGETABLE_PMD_FOLDED
1187 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1188 unsigned long address)
1193 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1196 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1197 pmd_t *pmd, unsigned long address);
1198 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1201 * The following ifdef needed to get the 4level-fixup.h header to work.
1202 * Remove it when 4level-fixup.h has been removed.
1204 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1205 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1207 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1208 NULL: pud_offset(pgd, address);
1211 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1213 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1214 NULL: pmd_offset(pud, address);
1216 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1218 #if USE_SPLIT_PTLOCKS
1220 * We tuck a spinlock to guard each pagetable page into its struct page,
1221 * at page->private, with BUILD_BUG_ON to make sure that this will not
1222 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1223 * When freeing, reset page->mapping so free_pages_check won't complain.
1225 #define __pte_lockptr(page) &((page)->ptl)
1226 #define pte_lock_init(_page) do { \
1227 spin_lock_init(__pte_lockptr(_page)); \
1229 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1230 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1231 #else /* !USE_SPLIT_PTLOCKS */
1233 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1235 #define pte_lock_init(page) do {} while (0)
1236 #define pte_lock_deinit(page) do {} while (0)
1237 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1238 #endif /* USE_SPLIT_PTLOCKS */
1240 static inline void pgtable_page_ctor(struct page *page)
1242 pte_lock_init(page);
1243 inc_zone_page_state(page, NR_PAGETABLE);
1246 static inline void pgtable_page_dtor(struct page *page)
1248 pte_lock_deinit(page);
1249 dec_zone_page_state(page, NR_PAGETABLE);
1252 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1254 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1255 pte_t *__pte = pte_offset_map(pmd, address); \
1261 #define pte_unmap_unlock(pte, ptl) do { \
1266 #define pte_alloc_map(mm, vma, pmd, address) \
1267 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1269 NULL: pte_offset_map(pmd, address))
1271 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1272 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1274 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1276 #define pte_alloc_kernel(pmd, address) \
1277 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1278 NULL: pte_offset_kernel(pmd, address))
1280 extern void free_area_init(unsigned long * zones_size);
1281 extern void free_area_init_node(int nid, unsigned long * zones_size,
1282 unsigned long zone_start_pfn, unsigned long *zholes_size);
1283 extern void free_initmem(void);
1285 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1287 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1288 * zones, allocate the backing mem_map and account for memory holes in a more
1289 * architecture independent manner. This is a substitute for creating the
1290 * zone_sizes[] and zholes_size[] arrays and passing them to
1291 * free_area_init_node()
1293 * An architecture is expected to register range of page frames backed by
1294 * physical memory with memblock_add[_node]() before calling
1295 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1296 * usage, an architecture is expected to do something like
1298 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1300 * for_each_valid_physical_page_range()
1301 * memblock_add_node(base, size, nid)
1302 * free_area_init_nodes(max_zone_pfns);
1304 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1305 * registered physical page range. Similarly
1306 * sparse_memory_present_with_active_regions() calls memory_present() for
1307 * each range when SPARSEMEM is enabled.
1309 * See mm/page_alloc.c for more information on each function exposed by
1310 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1312 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1313 unsigned long node_map_pfn_alignment(void);
1314 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1315 unsigned long end_pfn);
1316 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1317 unsigned long end_pfn);
1318 extern void get_pfn_range_for_nid(unsigned int nid,
1319 unsigned long *start_pfn, unsigned long *end_pfn);
1320 extern unsigned long find_min_pfn_with_active_regions(void);
1321 extern void free_bootmem_with_active_regions(int nid,
1322 unsigned long max_low_pfn);
1323 extern void sparse_memory_present_with_active_regions(int nid);
1325 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1327 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1328 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1329 static inline int __early_pfn_to_nid(unsigned long pfn)
1334 /* please see mm/page_alloc.c */
1335 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1336 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1337 /* there is a per-arch backend function. */
1338 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1339 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1342 extern void set_dma_reserve(unsigned long new_dma_reserve);
1343 extern void memmap_init_zone(unsigned long, int, unsigned long,
1344 unsigned long, enum memmap_context);
1345 extern void setup_per_zone_wmarks(void);
1346 extern int __meminit init_per_zone_wmark_min(void);
1347 extern void mem_init(void);
1348 extern void __init mmap_init(void);
1349 extern void show_mem(unsigned int flags);
1350 extern void si_meminfo(struct sysinfo * val);
1351 extern void si_meminfo_node(struct sysinfo *val, int nid);
1352 extern int after_bootmem;
1354 extern __printf(3, 4)
1355 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1357 extern void setup_per_cpu_pageset(void);
1359 extern void zone_pcp_update(struct zone *zone);
1362 extern atomic_long_t mmap_pages_allocated;
1363 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1366 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1367 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1368 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1369 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1370 struct prio_tree_iter *iter);
1372 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1373 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1374 (vma = vma_prio_tree_next(vma, iter)); )
1376 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1377 struct list_head *list)
1379 vma->shared.vm_set.parent = NULL;
1380 list_add_tail(&vma->shared.vm_set.list, list);
1384 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1385 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1386 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1387 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1388 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1389 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1390 struct mempolicy *);
1391 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1392 extern int split_vma(struct mm_struct *,
1393 struct vm_area_struct *, unsigned long addr, int new_below);
1394 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1395 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1396 struct rb_node **, struct rb_node *);
1397 extern void unlink_file_vma(struct vm_area_struct *);
1398 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1399 unsigned long addr, unsigned long len, pgoff_t pgoff);
1400 extern void exit_mmap(struct mm_struct *);
1402 extern int mm_take_all_locks(struct mm_struct *mm);
1403 extern void mm_drop_all_locks(struct mm_struct *mm);
1405 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1406 extern void added_exe_file_vma(struct mm_struct *mm);
1407 extern void removed_exe_file_vma(struct mm_struct *mm);
1408 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1409 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1411 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1412 extern int install_special_mapping(struct mm_struct *mm,
1413 unsigned long addr, unsigned long len,
1414 unsigned long flags, struct page **pages);
1416 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1418 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1419 unsigned long len, unsigned long flags,
1420 vm_flags_t vm_flags, unsigned long pgoff);
1421 extern unsigned long do_mmap(struct file *, unsigned long,
1422 unsigned long, unsigned long,
1423 unsigned long, unsigned long);
1424 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1426 /* These take the mm semaphore themselves */
1427 extern unsigned long vm_brk(unsigned long, unsigned long);
1428 extern int vm_munmap(unsigned long, size_t);
1429 extern unsigned long vm_mmap(struct file *, unsigned long,
1430 unsigned long, unsigned long,
1431 unsigned long, unsigned long);
1434 extern void truncate_inode_pages(struct address_space *, loff_t);
1435 extern void truncate_inode_pages_range(struct address_space *,
1436 loff_t lstart, loff_t lend);
1438 /* generic vm_area_ops exported for stackable file systems */
1439 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1441 /* mm/page-writeback.c */
1442 int write_one_page(struct page *page, int wait);
1443 void task_dirty_inc(struct task_struct *tsk);
1446 #ifndef CONFIG_KERNEL_DESKTOP
1447 #define VM_MAX_READAHEAD 512 /* kbytes */
1449 #define VM_MAX_READAHEAD 128 /* kbytes */
1451 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1453 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1454 pgoff_t offset, unsigned long nr_to_read);
1456 void page_cache_sync_readahead(struct address_space *mapping,
1457 struct file_ra_state *ra,
1460 unsigned long size);
1462 void page_cache_async_readahead(struct address_space *mapping,
1463 struct file_ra_state *ra,
1467 unsigned long size);
1469 unsigned long max_sane_readahead(unsigned long nr);
1470 unsigned long ra_submit(struct file_ra_state *ra,
1471 struct address_space *mapping,
1474 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1475 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1477 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1478 extern int expand_downwards(struct vm_area_struct *vma,
1479 unsigned long address);
1481 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1483 #define expand_upwards(vma, address) do { } while (0)
1486 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1487 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1488 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1489 struct vm_area_struct **pprev);
1491 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1492 NULL if none. Assume start_addr < end_addr. */
1493 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1495 struct vm_area_struct * vma = find_vma(mm,start_addr);
1497 if (vma && end_addr <= vma->vm_start)
1502 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1504 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1507 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1508 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1509 unsigned long vm_start, unsigned long vm_end)
1511 struct vm_area_struct *vma = find_vma(mm, vm_start);
1513 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1520 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1522 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1528 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1529 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1530 unsigned long pfn, unsigned long size, pgprot_t);
1531 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1532 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1534 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1537 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1538 unsigned int foll_flags);
1539 #define FOLL_WRITE 0x01 /* check pte is writable */
1540 #define FOLL_TOUCH 0x02 /* mark page accessed */
1541 #define FOLL_GET 0x04 /* do get_page on page */
1542 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1543 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1544 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1545 * and return without waiting upon it */
1546 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1547 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1548 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1550 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1552 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1553 unsigned long size, pte_fn_t fn, void *data);
1555 #ifdef CONFIG_PROC_FS
1556 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1558 static inline void vm_stat_account(struct mm_struct *mm,
1559 unsigned long flags, struct file *file, long pages)
1562 #endif /* CONFIG_PROC_FS */
1564 #ifdef CONFIG_DEBUG_PAGEALLOC
1565 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1566 #ifdef CONFIG_HIBERNATION
1567 extern bool kernel_page_present(struct page *page);
1568 #endif /* CONFIG_HIBERNATION */
1571 kernel_map_pages(struct page *page, int numpages, int enable) {}
1572 #ifdef CONFIG_HIBERNATION
1573 static inline bool kernel_page_present(struct page *page) { return true; }
1574 #endif /* CONFIG_HIBERNATION */
1577 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1578 #ifdef __HAVE_ARCH_GATE_AREA
1579 int in_gate_area_no_mm(unsigned long addr);
1580 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1582 int in_gate_area_no_mm(unsigned long addr);
1583 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1584 #endif /* __HAVE_ARCH_GATE_AREA */
1586 int drop_caches_sysctl_handler(struct ctl_table *, int,
1587 void __user *, size_t *, loff_t *);
1588 unsigned long shrink_slab(struct shrink_control *shrink,
1589 unsigned long nr_pages_scanned,
1590 unsigned long lru_pages);
1593 #define randomize_va_space 0
1595 extern int randomize_va_space;
1598 const char * arch_vma_name(struct vm_area_struct *vma);
1599 void print_vma_addr(char *prefix, unsigned long rip);
1601 void sparse_mem_maps_populate_node(struct page **map_map,
1602 unsigned long pnum_begin,
1603 unsigned long pnum_end,
1604 unsigned long map_count,
1607 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1608 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1609 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1610 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1611 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1612 void *vmemmap_alloc_block(unsigned long size, int node);
1613 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1614 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1615 int vmemmap_populate_basepages(struct page *start_page,
1616 unsigned long pages, int node);
1617 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1618 void vmemmap_populate_print_last(void);
1622 MF_COUNT_INCREASED = 1 << 0,
1623 MF_ACTION_REQUIRED = 1 << 1,
1625 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1626 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1627 extern int unpoison_memory(unsigned long pfn);
1628 extern int sysctl_memory_failure_early_kill;
1629 extern int sysctl_memory_failure_recovery;
1630 extern void shake_page(struct page *p, int access);
1631 extern atomic_long_t mce_bad_pages;
1632 extern int soft_offline_page(struct page *page, int flags);
1634 extern void dump_page(struct page *page);
1636 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1637 extern void clear_huge_page(struct page *page,
1639 unsigned int pages_per_huge_page);
1640 extern void copy_user_huge_page(struct page *dst, struct page *src,
1641 unsigned long addr, struct vm_area_struct *vma,
1642 unsigned int pages_per_huge_page);
1643 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1645 #ifdef CONFIG_DEBUG_PAGEALLOC
1646 extern unsigned int _debug_guardpage_minorder;
1648 static inline unsigned int debug_guardpage_minorder(void)
1650 return _debug_guardpage_minorder;
1653 static inline bool page_is_guard(struct page *page)
1655 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1658 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1659 static inline bool page_is_guard(struct page *page) { return false; }
1660 #endif /* CONFIG_DEBUG_PAGEALLOC */
1662 #endif /* __KERNEL__ */
1663 #endif /* _LINUX_MM_H */