4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.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>
24 struct anon_vma_chain;
27 struct writeback_control;
29 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
30 extern unsigned long max_mapnr;
32 static inline void set_max_mapnr(unsigned long limit)
37 static inline void set_max_mapnr(unsigned long limit) { }
40 extern unsigned long totalram_pages;
41 extern void * high_memory;
42 extern int page_cluster;
45 extern int sysctl_legacy_va_layout;
47 #define sysctl_legacy_va_layout 0
51 #include <asm/pgtable.h>
52 #include <asm/processor.h>
55 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
58 extern unsigned long sysctl_user_reserve_kbytes;
59 extern unsigned long sysctl_admin_reserve_kbytes;
61 extern int sysctl_overcommit_memory;
62 extern int sysctl_overcommit_ratio;
63 extern unsigned long sysctl_overcommit_kbytes;
65 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
67 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
70 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
72 /* to align the pointer to the (next) page boundary */
73 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
75 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
76 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
79 * Linux kernel virtual memory manager primitives.
80 * The idea being to have a "virtual" mm in the same way
81 * we have a virtual fs - giving a cleaner interface to the
82 * mm details, and allowing different kinds of memory mappings
83 * (from shared memory to executable loading to arbitrary
87 extern struct kmem_cache *vm_area_cachep;
90 extern struct rb_root nommu_region_tree;
91 extern struct rw_semaphore nommu_region_sem;
93 extern unsigned int kobjsize(const void *objp);
97 * vm_flags in vm_area_struct, see mm_types.h.
99 #define VM_NONE 0x00000000
101 #define VM_READ 0x00000001 /* currently active flags */
102 #define VM_WRITE 0x00000002
103 #define VM_EXEC 0x00000004
104 #define VM_SHARED 0x00000008
106 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
107 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
108 #define VM_MAYWRITE 0x00000020
109 #define VM_MAYEXEC 0x00000040
110 #define VM_MAYSHARE 0x00000080
112 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
113 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
114 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
116 #define VM_LOCKED 0x00002000
117 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
119 /* Used by sys_madvise() */
120 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
121 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
123 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
124 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
125 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
126 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
127 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
128 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
129 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
130 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
132 #ifdef CONFIG_MEM_SOFT_DIRTY
133 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
135 # define VM_SOFTDIRTY 0
138 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
139 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
140 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
141 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
143 #if defined(CONFIG_X86)
144 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
145 #elif defined(CONFIG_PPC)
146 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
147 #elif defined(CONFIG_PARISC)
148 # define VM_GROWSUP VM_ARCH_1
149 #elif defined(CONFIG_METAG)
150 # define VM_GROWSUP VM_ARCH_1
151 #elif defined(CONFIG_IA64)
152 # define VM_GROWSUP VM_ARCH_1
153 #elif !defined(CONFIG_MMU)
154 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
158 # define VM_GROWSUP VM_NONE
161 /* Bits set in the VMA until the stack is in its final location */
162 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
164 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
165 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
168 #ifdef CONFIG_STACK_GROWSUP
169 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
171 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
175 * Special vmas that are non-mergable, non-mlock()able.
176 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
178 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
180 /* This mask defines which mm->def_flags a process can inherit its parent */
181 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
184 * mapping from the currently active vm_flags protection bits (the
185 * low four bits) to a page protection mask..
187 extern pgprot_t protection_map[16];
189 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
190 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
191 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
192 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
193 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
194 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
195 #define FAULT_FLAG_TRIED 0x40 /* second try */
196 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
199 * vm_fault is filled by the the pagefault handler and passed to the vma's
200 * ->fault function. The vma's ->fault is responsible for returning a bitmask
201 * of VM_FAULT_xxx flags that give details about how the fault was handled.
203 * pgoff should be used in favour of virtual_address, if possible. If pgoff
204 * is used, one may implement ->remap_pages to get nonlinear mapping support.
207 unsigned int flags; /* FAULT_FLAG_xxx flags */
208 pgoff_t pgoff; /* Logical page offset based on vma */
209 void __user *virtual_address; /* Faulting virtual address */
211 struct page *page; /* ->fault handlers should return a
212 * page here, unless VM_FAULT_NOPAGE
213 * is set (which is also implied by
216 /* for ->map_pages() only */
217 pgoff_t max_pgoff; /* map pages for offset from pgoff till
218 * max_pgoff inclusive */
219 pte_t *pte; /* pte entry associated with ->pgoff */
223 * These are the virtual MM functions - opening of an area, closing and
224 * unmapping it (needed to keep files on disk up-to-date etc), pointer
225 * to the functions called when a no-page or a wp-page exception occurs.
227 struct vm_operations_struct {
228 void (*open)(struct vm_area_struct * area);
229 void (*close)(struct vm_area_struct * area);
230 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
231 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
233 /* notification that a previously read-only page is about to become
234 * writable, if an error is returned it will cause a SIGBUS */
235 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
237 /* called by access_process_vm when get_user_pages() fails, typically
238 * for use by special VMAs that can switch between memory and hardware
240 int (*access)(struct vm_area_struct *vma, unsigned long addr,
241 void *buf, int len, int write);
243 /* Called by the /proc/PID/maps code to ask the vma whether it
244 * has a special name. Returning non-NULL will also cause this
245 * vma to be dumped unconditionally. */
246 const char *(*name)(struct vm_area_struct *vma);
250 * set_policy() op must add a reference to any non-NULL @new mempolicy
251 * to hold the policy upon return. Caller should pass NULL @new to
252 * remove a policy and fall back to surrounding context--i.e. do not
253 * install a MPOL_DEFAULT policy, nor the task or system default
256 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
259 * get_policy() op must add reference [mpol_get()] to any policy at
260 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
261 * in mm/mempolicy.c will do this automatically.
262 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
263 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
264 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
265 * must return NULL--i.e., do not "fallback" to task or system default
268 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
270 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
271 const nodemask_t *to, unsigned long flags);
273 /* called by sys_remap_file_pages() to populate non-linear mapping */
274 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
275 unsigned long size, pgoff_t pgoff);
281 #define page_private(page) ((page)->private)
282 #define set_page_private(page, v) ((page)->private = (v))
284 /* It's valid only if the page is free path or free_list */
285 static inline void set_freepage_migratetype(struct page *page, int migratetype)
287 page->index = migratetype;
290 /* It's valid only if the page is free path or free_list */
291 static inline int get_freepage_migratetype(struct page *page)
297 * FIXME: take this include out, include page-flags.h in
298 * files which need it (119 of them)
300 #include <linux/page-flags.h>
301 #include <linux/huge_mm.h>
304 * Methods to modify the page usage count.
306 * What counts for a page usage:
307 * - cache mapping (page->mapping)
308 * - private data (page->private)
309 * - page mapped in a task's page tables, each mapping
310 * is counted separately
312 * Also, many kernel routines increase the page count before a critical
313 * routine so they can be sure the page doesn't go away from under them.
317 * Drop a ref, return true if the refcount fell to zero (the page has no users)
319 static inline int put_page_testzero(struct page *page)
321 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
322 return atomic_dec_and_test(&page->_count);
326 * Try to grab a ref unless the page has a refcount of zero, return false if
328 * This can be called when MMU is off so it must not access
329 * any of the virtual mappings.
331 static inline int get_page_unless_zero(struct page *page)
333 return atomic_inc_not_zero(&page->_count);
337 * Try to drop a ref unless the page has a refcount of one, return false if
339 * This is to make sure that the refcount won't become zero after this drop.
340 * This can be called when MMU is off so it must not access
341 * any of the virtual mappings.
343 static inline int put_page_unless_one(struct page *page)
345 return atomic_add_unless(&page->_count, -1, 1);
348 extern int page_is_ram(unsigned long pfn);
350 /* Support for virtually mapped pages */
351 struct page *vmalloc_to_page(const void *addr);
352 unsigned long vmalloc_to_pfn(const void *addr);
355 * Determine if an address is within the vmalloc range
357 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
358 * is no special casing required.
360 static inline int is_vmalloc_addr(const void *x)
363 unsigned long addr = (unsigned long)x;
365 return addr >= VMALLOC_START && addr < VMALLOC_END;
371 extern int is_vmalloc_or_module_addr(const void *x);
373 static inline int is_vmalloc_or_module_addr(const void *x)
379 static inline void compound_lock(struct page *page)
381 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
382 VM_BUG_ON_PAGE(PageSlab(page), page);
383 bit_spin_lock(PG_compound_lock, &page->flags);
387 static inline void compound_unlock(struct page *page)
389 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
390 VM_BUG_ON_PAGE(PageSlab(page), page);
391 bit_spin_unlock(PG_compound_lock, &page->flags);
395 static inline unsigned long compound_lock_irqsave(struct page *page)
397 unsigned long uninitialized_var(flags);
398 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
399 local_irq_save(flags);
405 static inline void compound_unlock_irqrestore(struct page *page,
408 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
409 compound_unlock(page);
410 local_irq_restore(flags);
414 static inline struct page *compound_head(struct page *page)
416 if (unlikely(PageTail(page))) {
417 struct page *head = page->first_page;
420 * page->first_page may be a dangling pointer to an old
421 * compound page, so recheck that it is still a tail
422 * page before returning.
425 if (likely(PageTail(page)))
432 * The atomic page->_mapcount, starts from -1: so that transitions
433 * both from it and to it can be tracked, using atomic_inc_and_test
434 * and atomic_add_negative(-1).
436 static inline void page_mapcount_reset(struct page *page)
438 atomic_set(&(page)->_mapcount, -1);
441 static inline int page_mapcount(struct page *page)
443 return atomic_read(&(page)->_mapcount) + 1;
446 static inline int page_count(struct page *page)
448 return atomic_read(&compound_head(page)->_count);
451 #ifdef CONFIG_HUGETLB_PAGE
452 extern int PageHeadHuge(struct page *page_head);
453 #else /* CONFIG_HUGETLB_PAGE */
454 static inline int PageHeadHuge(struct page *page_head)
458 #endif /* CONFIG_HUGETLB_PAGE */
460 static inline bool __compound_tail_refcounted(struct page *page)
462 return !PageSlab(page) && !PageHeadHuge(page);
466 * This takes a head page as parameter and tells if the
467 * tail page reference counting can be skipped.
469 * For this to be safe, PageSlab and PageHeadHuge must remain true on
470 * any given page where they return true here, until all tail pins
471 * have been released.
473 static inline bool compound_tail_refcounted(struct page *page)
475 VM_BUG_ON_PAGE(!PageHead(page), page);
476 return __compound_tail_refcounted(page);
479 static inline void get_huge_page_tail(struct page *page)
482 * __split_huge_page_refcount() cannot run from under us.
484 VM_BUG_ON_PAGE(!PageTail(page), page);
485 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
486 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
487 if (compound_tail_refcounted(page->first_page))
488 atomic_inc(&page->_mapcount);
491 extern bool __get_page_tail(struct page *page);
493 static inline void get_page(struct page *page)
495 if (unlikely(PageTail(page)))
496 if (likely(__get_page_tail(page)))
499 * Getting a normal page or the head of a compound page
500 * requires to already have an elevated page->_count.
502 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
503 atomic_inc(&page->_count);
506 static inline struct page *virt_to_head_page(const void *x)
508 struct page *page = virt_to_page(x);
509 return compound_head(page);
513 * Setup the page count before being freed into the page allocator for
514 * the first time (boot or memory hotplug)
516 static inline void init_page_count(struct page *page)
518 atomic_set(&page->_count, 1);
522 * PageBuddy() indicate that the page is free and in the buddy system
523 * (see mm/page_alloc.c).
525 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
526 * -2 so that an underflow of the page_mapcount() won't be mistaken
527 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
528 * efficiently by most CPU architectures.
530 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
532 static inline int PageBuddy(struct page *page)
534 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
537 static inline void __SetPageBuddy(struct page *page)
539 VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
540 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
543 static inline void __ClearPageBuddy(struct page *page)
545 VM_BUG_ON_PAGE(!PageBuddy(page), page);
546 atomic_set(&page->_mapcount, -1);
549 void put_page(struct page *page);
550 void put_pages_list(struct list_head *pages);
552 void split_page(struct page *page, unsigned int order);
553 int split_free_page(struct page *page);
556 * Compound pages have a destructor function. Provide a
557 * prototype for that function and accessor functions.
558 * These are _only_ valid on the head of a PG_compound page.
560 typedef void compound_page_dtor(struct page *);
562 static inline void set_compound_page_dtor(struct page *page,
563 compound_page_dtor *dtor)
565 page[1].lru.next = (void *)dtor;
568 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
570 return (compound_page_dtor *)page[1].lru.next;
573 static inline int compound_order(struct page *page)
577 return (unsigned long)page[1].lru.prev;
580 static inline void set_compound_order(struct page *page, unsigned long order)
582 page[1].lru.prev = (void *)order;
587 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
588 * servicing faults for write access. In the normal case, do always want
589 * pte_mkwrite. But get_user_pages can cause write faults for mappings
590 * that do not have writing enabled, when used by access_process_vm.
592 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
594 if (likely(vma->vm_flags & VM_WRITE))
595 pte = pte_mkwrite(pte);
599 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
600 struct page *page, pte_t *pte, bool write, bool anon);
604 * Multiple processes may "see" the same page. E.g. for untouched
605 * mappings of /dev/null, all processes see the same page full of
606 * zeroes, and text pages of executables and shared libraries have
607 * only one copy in memory, at most, normally.
609 * For the non-reserved pages, page_count(page) denotes a reference count.
610 * page_count() == 0 means the page is free. page->lru is then used for
611 * freelist management in the buddy allocator.
612 * page_count() > 0 means the page has been allocated.
614 * Pages are allocated by the slab allocator in order to provide memory
615 * to kmalloc and kmem_cache_alloc. In this case, the management of the
616 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
617 * unless a particular usage is carefully commented. (the responsibility of
618 * freeing the kmalloc memory is the caller's, of course).
620 * A page may be used by anyone else who does a __get_free_page().
621 * In this case, page_count still tracks the references, and should only
622 * be used through the normal accessor functions. The top bits of page->flags
623 * and page->virtual store page management information, but all other fields
624 * are unused and could be used privately, carefully. The management of this
625 * page is the responsibility of the one who allocated it, and those who have
626 * subsequently been given references to it.
628 * The other pages (we may call them "pagecache pages") are completely
629 * managed by the Linux memory manager: I/O, buffers, swapping etc.
630 * The following discussion applies only to them.
632 * A pagecache page contains an opaque `private' member, which belongs to the
633 * page's address_space. Usually, this is the address of a circular list of
634 * the page's disk buffers. PG_private must be set to tell the VM to call
635 * into the filesystem to release these pages.
637 * A page may belong to an inode's memory mapping. In this case, page->mapping
638 * is the pointer to the inode, and page->index is the file offset of the page,
639 * in units of PAGE_CACHE_SIZE.
641 * If pagecache pages are not associated with an inode, they are said to be
642 * anonymous pages. These may become associated with the swapcache, and in that
643 * case PG_swapcache is set, and page->private is an offset into the swapcache.
645 * In either case (swapcache or inode backed), the pagecache itself holds one
646 * reference to the page. Setting PG_private should also increment the
647 * refcount. The each user mapping also has a reference to the page.
649 * The pagecache pages are stored in a per-mapping radix tree, which is
650 * rooted at mapping->page_tree, and indexed by offset.
651 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
652 * lists, we instead now tag pages as dirty/writeback in the radix tree.
654 * All pagecache pages may be subject to I/O:
655 * - inode pages may need to be read from disk,
656 * - inode pages which have been modified and are MAP_SHARED may need
657 * to be written back to the inode on disk,
658 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
659 * modified may need to be swapped out to swap space and (later) to be read
664 * The zone field is never updated after free_area_init_core()
665 * sets it, so none of the operations on it need to be atomic.
668 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
669 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
670 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
671 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
672 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
675 * Define the bit shifts to access each section. For non-existent
676 * sections we define the shift as 0; that plus a 0 mask ensures
677 * the compiler will optimise away reference to them.
679 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
680 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
681 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
682 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
684 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
685 #ifdef NODE_NOT_IN_PAGE_FLAGS
686 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
687 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
688 SECTIONS_PGOFF : ZONES_PGOFF)
690 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
691 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
692 NODES_PGOFF : ZONES_PGOFF)
695 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
697 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
698 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
701 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
702 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
703 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
704 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
705 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
707 static inline enum zone_type page_zonenum(const struct page *page)
709 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
712 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
713 #define SECTION_IN_PAGE_FLAGS
717 * The identification function is mainly used by the buddy allocator for
718 * determining if two pages could be buddies. We are not really identifying
719 * the zone since we could be using the section number id if we do not have
720 * node id available in page flags.
721 * We only guarantee that it will return the same value for two combinable
724 static inline int page_zone_id(struct page *page)
726 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
729 static inline int zone_to_nid(struct zone *zone)
738 #ifdef NODE_NOT_IN_PAGE_FLAGS
739 extern int page_to_nid(const struct page *page);
741 static inline int page_to_nid(const struct page *page)
743 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
747 #ifdef CONFIG_NUMA_BALANCING
748 static inline int cpu_pid_to_cpupid(int cpu, int pid)
750 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
753 static inline int cpupid_to_pid(int cpupid)
755 return cpupid & LAST__PID_MASK;
758 static inline int cpupid_to_cpu(int cpupid)
760 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
763 static inline int cpupid_to_nid(int cpupid)
765 return cpu_to_node(cpupid_to_cpu(cpupid));
768 static inline bool cpupid_pid_unset(int cpupid)
770 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
773 static inline bool cpupid_cpu_unset(int cpupid)
775 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
778 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
780 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
783 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
784 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
785 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
787 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
790 static inline int page_cpupid_last(struct page *page)
792 return page->_last_cpupid;
794 static inline void page_cpupid_reset_last(struct page *page)
796 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
799 static inline int page_cpupid_last(struct page *page)
801 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
804 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
806 static inline void page_cpupid_reset_last(struct page *page)
808 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
810 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
811 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
813 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
814 #else /* !CONFIG_NUMA_BALANCING */
815 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
817 return page_to_nid(page); /* XXX */
820 static inline int page_cpupid_last(struct page *page)
822 return page_to_nid(page); /* XXX */
825 static inline int cpupid_to_nid(int cpupid)
830 static inline int cpupid_to_pid(int cpupid)
835 static inline int cpupid_to_cpu(int cpupid)
840 static inline int cpu_pid_to_cpupid(int nid, int pid)
845 static inline bool cpupid_pid_unset(int cpupid)
850 static inline void page_cpupid_reset_last(struct page *page)
854 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
858 #endif /* CONFIG_NUMA_BALANCING */
860 static inline struct zone *page_zone(const struct page *page)
862 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
865 #ifdef SECTION_IN_PAGE_FLAGS
866 static inline void set_page_section(struct page *page, unsigned long section)
868 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
869 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
872 static inline unsigned long page_to_section(const struct page *page)
874 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
878 static inline void set_page_zone(struct page *page, enum zone_type zone)
880 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
881 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
884 static inline void set_page_node(struct page *page, unsigned long node)
886 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
887 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
890 static inline void set_page_links(struct page *page, enum zone_type zone,
891 unsigned long node, unsigned long pfn)
893 set_page_zone(page, zone);
894 set_page_node(page, node);
895 #ifdef SECTION_IN_PAGE_FLAGS
896 set_page_section(page, pfn_to_section_nr(pfn));
901 * Some inline functions in vmstat.h depend on page_zone()
903 #include <linux/vmstat.h>
905 static __always_inline void *lowmem_page_address(const struct page *page)
907 return __va(PFN_PHYS(page_to_pfn(page)));
910 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
911 #define HASHED_PAGE_VIRTUAL
914 #if defined(WANT_PAGE_VIRTUAL)
915 static inline void *page_address(const struct page *page)
917 return page->virtual;
919 static inline void set_page_address(struct page *page, void *address)
921 page->virtual = address;
923 #define page_address_init() do { } while(0)
926 #if defined(HASHED_PAGE_VIRTUAL)
927 void *page_address(const struct page *page);
928 void set_page_address(struct page *page, void *virtual);
929 void page_address_init(void);
932 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
933 #define page_address(page) lowmem_page_address(page)
934 #define set_page_address(page, address) do { } while(0)
935 #define page_address_init() do { } while(0)
939 * On an anonymous page mapped into a user virtual memory area,
940 * page->mapping points to its anon_vma, not to a struct address_space;
941 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
943 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
944 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
945 * and then page->mapping points, not to an anon_vma, but to a private
946 * structure which KSM associates with that merged page. See ksm.h.
948 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
950 * Please note that, confusingly, "page_mapping" refers to the inode
951 * address_space which maps the page from disk; whereas "page_mapped"
952 * refers to user virtual address space into which the page is mapped.
954 #define PAGE_MAPPING_ANON 1
955 #define PAGE_MAPPING_KSM 2
956 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
958 extern struct address_space *page_mapping(struct page *page);
960 /* Neutral page->mapping pointer to address_space or anon_vma or other */
961 static inline void *page_rmapping(struct page *page)
963 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
966 extern struct address_space *__page_file_mapping(struct page *);
969 struct address_space *page_file_mapping(struct page *page)
971 if (unlikely(PageSwapCache(page)))
972 return __page_file_mapping(page);
974 return page->mapping;
977 static inline int PageAnon(struct page *page)
979 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
983 * Return the pagecache index of the passed page. Regular pagecache pages
984 * use ->index whereas swapcache pages use ->private
986 static inline pgoff_t page_index(struct page *page)
988 if (unlikely(PageSwapCache(page)))
989 return page_private(page);
993 extern pgoff_t __page_file_index(struct page *page);
996 * Return the file index of the page. Regular pagecache pages use ->index
997 * whereas swapcache pages use swp_offset(->private)
999 static inline pgoff_t page_file_index(struct page *page)
1001 if (unlikely(PageSwapCache(page)))
1002 return __page_file_index(page);
1008 * Return true if this page is mapped into pagetables.
1010 static inline int page_mapped(struct page *page)
1012 return atomic_read(&(page)->_mapcount) >= 0;
1016 * Different kinds of faults, as returned by handle_mm_fault().
1017 * Used to decide whether a process gets delivered SIGBUS or
1018 * just gets major/minor fault counters bumped up.
1021 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1023 #define VM_FAULT_OOM 0x0001
1024 #define VM_FAULT_SIGBUS 0x0002
1025 #define VM_FAULT_MAJOR 0x0004
1026 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1027 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1028 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1030 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1031 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1032 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1033 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1035 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1037 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
1038 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
1040 /* Encode hstate index for a hwpoisoned large page */
1041 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1042 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1045 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1047 extern void pagefault_out_of_memory(void);
1049 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1052 * Flags passed to show_mem() and show_free_areas() to suppress output in
1055 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1057 extern void show_free_areas(unsigned int flags);
1058 extern bool skip_free_areas_node(unsigned int flags, int nid);
1060 int shmem_zero_setup(struct vm_area_struct *);
1062 bool shmem_mapping(struct address_space *mapping);
1064 static inline bool shmem_mapping(struct address_space *mapping)
1070 extern int can_do_mlock(void);
1071 extern int user_shm_lock(size_t, struct user_struct *);
1072 extern void user_shm_unlock(size_t, struct user_struct *);
1075 * Parameter block passed down to zap_pte_range in exceptional cases.
1077 struct zap_details {
1078 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
1079 struct address_space *check_mapping; /* Check page->mapping if set */
1080 pgoff_t first_index; /* Lowest page->index to unmap */
1081 pgoff_t last_index; /* Highest page->index to unmap */
1084 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1087 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1088 unsigned long size);
1089 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1090 unsigned long size, struct zap_details *);
1091 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1092 unsigned long start, unsigned long end);
1095 * mm_walk - callbacks for walk_page_range
1096 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1097 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1098 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1099 * this handler is required to be able to handle
1100 * pmd_trans_huge() pmds. They may simply choose to
1101 * split_huge_page() instead of handling it explicitly.
1102 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1103 * @pte_hole: if set, called for each hole at all levels
1104 * @hugetlb_entry: if set, called for each hugetlb entry
1105 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1108 * (see walk_page_range for more details)
1111 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1112 unsigned long next, struct mm_walk *walk);
1113 int (*pud_entry)(pud_t *pud, unsigned long addr,
1114 unsigned long next, struct mm_walk *walk);
1115 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1116 unsigned long next, struct mm_walk *walk);
1117 int (*pte_entry)(pte_t *pte, unsigned long addr,
1118 unsigned long next, struct mm_walk *walk);
1119 int (*pte_hole)(unsigned long addr, unsigned long next,
1120 struct mm_walk *walk);
1121 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1122 unsigned long addr, unsigned long next,
1123 struct mm_walk *walk);
1124 struct mm_struct *mm;
1128 int walk_page_range(unsigned long addr, unsigned long end,
1129 struct mm_walk *walk);
1130 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1131 unsigned long end, unsigned long floor, unsigned long ceiling);
1132 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1133 struct vm_area_struct *vma);
1134 void unmap_mapping_range(struct address_space *mapping,
1135 loff_t const holebegin, loff_t const holelen, int even_cows);
1136 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1137 unsigned long *pfn);
1138 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1139 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1140 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1141 void *buf, int len, int write);
1143 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1144 loff_t const holebegin, loff_t const holelen)
1146 unmap_mapping_range(mapping, holebegin, holelen, 0);
1149 extern void truncate_pagecache(struct inode *inode, loff_t new);
1150 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1151 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1152 int truncate_inode_page(struct address_space *mapping, struct page *page);
1153 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1154 int invalidate_inode_page(struct page *page);
1157 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1158 unsigned long address, unsigned int flags);
1159 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1160 unsigned long address, unsigned int fault_flags);
1162 static inline int handle_mm_fault(struct mm_struct *mm,
1163 struct vm_area_struct *vma, unsigned long address,
1166 /* should never happen if there's no MMU */
1168 return VM_FAULT_SIGBUS;
1170 static inline int fixup_user_fault(struct task_struct *tsk,
1171 struct mm_struct *mm, unsigned long address,
1172 unsigned int fault_flags)
1174 /* should never happen if there's no MMU */
1180 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1181 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1182 void *buf, int len, int write);
1184 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1185 unsigned long start, unsigned long nr_pages,
1186 unsigned int foll_flags, struct page **pages,
1187 struct vm_area_struct **vmas, int *nonblocking);
1188 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1189 unsigned long start, unsigned long nr_pages,
1190 int write, int force, struct page **pages,
1191 struct vm_area_struct **vmas);
1192 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1193 struct page **pages);
1195 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1196 struct page **pages);
1197 int get_kernel_page(unsigned long start, int write, struct page **pages);
1198 struct page *get_dump_page(unsigned long addr);
1200 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1201 extern void do_invalidatepage(struct page *page, unsigned int offset,
1202 unsigned int length);
1204 int __set_page_dirty_nobuffers(struct page *page);
1205 int __set_page_dirty_no_writeback(struct page *page);
1206 int redirty_page_for_writepage(struct writeback_control *wbc,
1208 void account_page_dirtied(struct page *page, struct address_space *mapping);
1209 void account_page_writeback(struct page *page);
1210 int set_page_dirty(struct page *page);
1211 int set_page_dirty_lock(struct page *page);
1212 int clear_page_dirty_for_io(struct page *page);
1213 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1215 /* Is the vma a continuation of the stack vma above it? */
1216 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1218 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1221 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1224 return (vma->vm_flags & VM_GROWSDOWN) &&
1225 (vma->vm_start == addr) &&
1226 !vma_growsdown(vma->vm_prev, addr);
1229 /* Is the vma a continuation of the stack vma below it? */
1230 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1232 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1235 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1238 return (vma->vm_flags & VM_GROWSUP) &&
1239 (vma->vm_end == addr) &&
1240 !vma_growsup(vma->vm_next, addr);
1244 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1246 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1247 unsigned long old_addr, struct vm_area_struct *new_vma,
1248 unsigned long new_addr, unsigned long len,
1249 bool need_rmap_locks);
1250 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1251 unsigned long end, pgprot_t newprot,
1252 int dirty_accountable, int prot_numa);
1253 extern int mprotect_fixup(struct vm_area_struct *vma,
1254 struct vm_area_struct **pprev, unsigned long start,
1255 unsigned long end, unsigned long newflags);
1258 * doesn't attempt to fault and will return short.
1260 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1261 struct page **pages);
1263 * per-process(per-mm_struct) statistics.
1265 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1267 long val = atomic_long_read(&mm->rss_stat.count[member]);
1269 #ifdef SPLIT_RSS_COUNTING
1271 * counter is updated in asynchronous manner and may go to minus.
1272 * But it's never be expected number for users.
1277 return (unsigned long)val;
1280 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1282 atomic_long_add(value, &mm->rss_stat.count[member]);
1285 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1287 atomic_long_inc(&mm->rss_stat.count[member]);
1290 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1292 atomic_long_dec(&mm->rss_stat.count[member]);
1295 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1297 return get_mm_counter(mm, MM_FILEPAGES) +
1298 get_mm_counter(mm, MM_ANONPAGES);
1301 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1303 return max(mm->hiwater_rss, get_mm_rss(mm));
1306 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1308 return max(mm->hiwater_vm, mm->total_vm);
1311 static inline void update_hiwater_rss(struct mm_struct *mm)
1313 unsigned long _rss = get_mm_rss(mm);
1315 if ((mm)->hiwater_rss < _rss)
1316 (mm)->hiwater_rss = _rss;
1319 static inline void update_hiwater_vm(struct mm_struct *mm)
1321 if (mm->hiwater_vm < mm->total_vm)
1322 mm->hiwater_vm = mm->total_vm;
1325 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1326 struct mm_struct *mm)
1328 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1330 if (*maxrss < hiwater_rss)
1331 *maxrss = hiwater_rss;
1334 #if defined(SPLIT_RSS_COUNTING)
1335 void sync_mm_rss(struct mm_struct *mm);
1337 static inline void sync_mm_rss(struct mm_struct *mm)
1342 int vma_wants_writenotify(struct vm_area_struct *vma);
1344 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1346 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1350 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1354 #ifdef __PAGETABLE_PUD_FOLDED
1355 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1356 unsigned long address)
1361 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1364 #ifdef __PAGETABLE_PMD_FOLDED
1365 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1366 unsigned long address)
1371 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1374 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1375 pmd_t *pmd, unsigned long address);
1376 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1379 * The following ifdef needed to get the 4level-fixup.h header to work.
1380 * Remove it when 4level-fixup.h has been removed.
1382 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1383 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1385 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1386 NULL: pud_offset(pgd, address);
1389 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1391 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1392 NULL: pmd_offset(pud, address);
1394 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1396 #if USE_SPLIT_PTE_PTLOCKS
1397 #if ALLOC_SPLIT_PTLOCKS
1398 void __init ptlock_cache_init(void);
1399 extern bool ptlock_alloc(struct page *page);
1400 extern void ptlock_free(struct page *page);
1402 static inline spinlock_t *ptlock_ptr(struct page *page)
1406 #else /* ALLOC_SPLIT_PTLOCKS */
1407 static inline void ptlock_cache_init(void)
1411 static inline bool ptlock_alloc(struct page *page)
1416 static inline void ptlock_free(struct page *page)
1420 static inline spinlock_t *ptlock_ptr(struct page *page)
1424 #endif /* ALLOC_SPLIT_PTLOCKS */
1426 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1428 return ptlock_ptr(pmd_page(*pmd));
1431 static inline bool ptlock_init(struct page *page)
1434 * prep_new_page() initialize page->private (and therefore page->ptl)
1435 * with 0. Make sure nobody took it in use in between.
1437 * It can happen if arch try to use slab for page table allocation:
1438 * slab code uses page->slab_cache and page->first_page (for tail
1439 * pages), which share storage with page->ptl.
1441 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1442 if (!ptlock_alloc(page))
1444 spin_lock_init(ptlock_ptr(page));
1448 /* Reset page->mapping so free_pages_check won't complain. */
1449 static inline void pte_lock_deinit(struct page *page)
1451 page->mapping = NULL;
1455 #else /* !USE_SPLIT_PTE_PTLOCKS */
1457 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1459 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1461 return &mm->page_table_lock;
1463 static inline void ptlock_cache_init(void) {}
1464 static inline bool ptlock_init(struct page *page) { return true; }
1465 static inline void pte_lock_deinit(struct page *page) {}
1466 #endif /* USE_SPLIT_PTE_PTLOCKS */
1468 static inline void pgtable_init(void)
1470 ptlock_cache_init();
1471 pgtable_cache_init();
1474 static inline bool pgtable_page_ctor(struct page *page)
1476 inc_zone_page_state(page, NR_PAGETABLE);
1477 return ptlock_init(page);
1480 static inline void pgtable_page_dtor(struct page *page)
1482 pte_lock_deinit(page);
1483 dec_zone_page_state(page, NR_PAGETABLE);
1486 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1488 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1489 pte_t *__pte = pte_offset_map(pmd, address); \
1495 #define pte_unmap_unlock(pte, ptl) do { \
1500 #define pte_alloc_map(mm, vma, pmd, address) \
1501 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1503 NULL: pte_offset_map(pmd, address))
1505 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1506 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1508 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1510 #define pte_alloc_kernel(pmd, address) \
1511 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1512 NULL: pte_offset_kernel(pmd, address))
1514 #if USE_SPLIT_PMD_PTLOCKS
1516 static struct page *pmd_to_page(pmd_t *pmd)
1518 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1519 return virt_to_page((void *)((unsigned long) pmd & mask));
1522 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1524 return ptlock_ptr(pmd_to_page(pmd));
1527 static inline bool pgtable_pmd_page_ctor(struct page *page)
1529 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1530 page->pmd_huge_pte = NULL;
1532 return ptlock_init(page);
1535 static inline void pgtable_pmd_page_dtor(struct page *page)
1537 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1538 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1543 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1547 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1549 return &mm->page_table_lock;
1552 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1553 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1555 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1559 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1561 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1566 extern void free_area_init(unsigned long * zones_size);
1567 extern void free_area_init_node(int nid, unsigned long * zones_size,
1568 unsigned long zone_start_pfn, unsigned long *zholes_size);
1569 extern void free_initmem(void);
1572 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1573 * into the buddy system. The freed pages will be poisoned with pattern
1574 * "poison" if it's within range [0, UCHAR_MAX].
1575 * Return pages freed into the buddy system.
1577 extern unsigned long free_reserved_area(void *start, void *end,
1578 int poison, char *s);
1580 #ifdef CONFIG_HIGHMEM
1582 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1583 * and totalram_pages.
1585 extern void free_highmem_page(struct page *page);
1588 extern void adjust_managed_page_count(struct page *page, long count);
1589 extern void mem_init_print_info(const char *str);
1591 /* Free the reserved page into the buddy system, so it gets managed. */
1592 static inline void __free_reserved_page(struct page *page)
1594 ClearPageReserved(page);
1595 init_page_count(page);
1599 static inline void free_reserved_page(struct page *page)
1601 __free_reserved_page(page);
1602 adjust_managed_page_count(page, 1);
1605 static inline void mark_page_reserved(struct page *page)
1607 SetPageReserved(page);
1608 adjust_managed_page_count(page, -1);
1612 * Default method to free all the __init memory into the buddy system.
1613 * The freed pages will be poisoned with pattern "poison" if it's within
1614 * range [0, UCHAR_MAX].
1615 * Return pages freed into the buddy system.
1617 static inline unsigned long free_initmem_default(int poison)
1619 extern char __init_begin[], __init_end[];
1621 return free_reserved_area(&__init_begin, &__init_end,
1622 poison, "unused kernel");
1625 static inline unsigned long get_num_physpages(void)
1628 unsigned long phys_pages = 0;
1630 for_each_online_node(nid)
1631 phys_pages += node_present_pages(nid);
1636 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1638 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1639 * zones, allocate the backing mem_map and account for memory holes in a more
1640 * architecture independent manner. This is a substitute for creating the
1641 * zone_sizes[] and zholes_size[] arrays and passing them to
1642 * free_area_init_node()
1644 * An architecture is expected to register range of page frames backed by
1645 * physical memory with memblock_add[_node]() before calling
1646 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1647 * usage, an architecture is expected to do something like
1649 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1651 * for_each_valid_physical_page_range()
1652 * memblock_add_node(base, size, nid)
1653 * free_area_init_nodes(max_zone_pfns);
1655 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1656 * registered physical page range. Similarly
1657 * sparse_memory_present_with_active_regions() calls memory_present() for
1658 * each range when SPARSEMEM is enabled.
1660 * See mm/page_alloc.c for more information on each function exposed by
1661 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1663 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1664 unsigned long node_map_pfn_alignment(void);
1665 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1666 unsigned long end_pfn);
1667 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1668 unsigned long end_pfn);
1669 extern void get_pfn_range_for_nid(unsigned int nid,
1670 unsigned long *start_pfn, unsigned long *end_pfn);
1671 extern unsigned long find_min_pfn_with_active_regions(void);
1672 extern void free_bootmem_with_active_regions(int nid,
1673 unsigned long max_low_pfn);
1674 extern void sparse_memory_present_with_active_regions(int nid);
1676 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1678 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1679 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1680 static inline int __early_pfn_to_nid(unsigned long pfn)
1685 /* please see mm/page_alloc.c */
1686 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1687 /* there is a per-arch backend function. */
1688 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1691 extern void set_dma_reserve(unsigned long new_dma_reserve);
1692 extern void memmap_init_zone(unsigned long, int, unsigned long,
1693 unsigned long, enum memmap_context);
1694 extern void setup_per_zone_wmarks(void);
1695 extern int __meminit init_per_zone_wmark_min(void);
1696 extern void mem_init(void);
1697 extern void __init mmap_init(void);
1698 extern void show_mem(unsigned int flags);
1699 extern void si_meminfo(struct sysinfo * val);
1700 extern void si_meminfo_node(struct sysinfo *val, int nid);
1702 extern __printf(3, 4)
1703 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1705 extern void setup_per_cpu_pageset(void);
1707 extern void zone_pcp_update(struct zone *zone);
1708 extern void zone_pcp_reset(struct zone *zone);
1711 extern int min_free_kbytes;
1714 extern atomic_long_t mmap_pages_allocated;
1715 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1717 /* interval_tree.c */
1718 void vma_interval_tree_insert(struct vm_area_struct *node,
1719 struct rb_root *root);
1720 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1721 struct vm_area_struct *prev,
1722 struct rb_root *root);
1723 void vma_interval_tree_remove(struct vm_area_struct *node,
1724 struct rb_root *root);
1725 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1726 unsigned long start, unsigned long last);
1727 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1728 unsigned long start, unsigned long last);
1730 #define vma_interval_tree_foreach(vma, root, start, last) \
1731 for (vma = vma_interval_tree_iter_first(root, start, last); \
1732 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1734 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1735 struct list_head *list)
1737 list_add_tail(&vma->shared.nonlinear, list);
1740 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1741 struct rb_root *root);
1742 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1743 struct rb_root *root);
1744 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1745 struct rb_root *root, unsigned long start, unsigned long last);
1746 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1747 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1748 #ifdef CONFIG_DEBUG_VM_RB
1749 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1752 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1753 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1754 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1757 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1758 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1759 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1760 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1761 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1762 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1763 struct mempolicy *);
1764 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1765 extern int split_vma(struct mm_struct *,
1766 struct vm_area_struct *, unsigned long addr, int new_below);
1767 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1768 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1769 struct rb_node **, struct rb_node *);
1770 extern void unlink_file_vma(struct vm_area_struct *);
1771 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1772 unsigned long addr, unsigned long len, pgoff_t pgoff,
1773 bool *need_rmap_locks);
1774 extern void exit_mmap(struct mm_struct *);
1776 extern int mm_take_all_locks(struct mm_struct *mm);
1777 extern void mm_drop_all_locks(struct mm_struct *mm);
1779 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1780 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1782 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1783 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1784 unsigned long addr, unsigned long len,
1785 unsigned long flags,
1786 const struct vm_special_mapping *spec);
1787 /* This is an obsolete alternative to _install_special_mapping. */
1788 extern int install_special_mapping(struct mm_struct *mm,
1789 unsigned long addr, unsigned long len,
1790 unsigned long flags, struct page **pages);
1792 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1794 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1795 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1796 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1797 unsigned long len, unsigned long prot, unsigned long flags,
1798 unsigned long pgoff, unsigned long *populate);
1799 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1802 extern int __mm_populate(unsigned long addr, unsigned long len,
1804 static inline void mm_populate(unsigned long addr, unsigned long len)
1807 (void) __mm_populate(addr, len, 1);
1810 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1813 /* These take the mm semaphore themselves */
1814 extern unsigned long vm_brk(unsigned long, unsigned long);
1815 extern int vm_munmap(unsigned long, size_t);
1816 extern unsigned long vm_mmap(struct file *, unsigned long,
1817 unsigned long, unsigned long,
1818 unsigned long, unsigned long);
1820 struct vm_unmapped_area_info {
1821 #define VM_UNMAPPED_AREA_TOPDOWN 1
1822 unsigned long flags;
1823 unsigned long length;
1824 unsigned long low_limit;
1825 unsigned long high_limit;
1826 unsigned long align_mask;
1827 unsigned long align_offset;
1830 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1831 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1834 * Search for an unmapped address range.
1836 * We are looking for a range that:
1837 * - does not intersect with any VMA;
1838 * - is contained within the [low_limit, high_limit) interval;
1839 * - is at least the desired size.
1840 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1842 static inline unsigned long
1843 vm_unmapped_area(struct vm_unmapped_area_info *info)
1845 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1846 return unmapped_area(info);
1848 return unmapped_area_topdown(info);
1852 extern void truncate_inode_pages(struct address_space *, loff_t);
1853 extern void truncate_inode_pages_range(struct address_space *,
1854 loff_t lstart, loff_t lend);
1855 extern void truncate_inode_pages_final(struct address_space *);
1857 /* generic vm_area_ops exported for stackable file systems */
1858 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1859 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1860 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1862 /* mm/page-writeback.c */
1863 int write_one_page(struct page *page, int wait);
1864 void task_dirty_inc(struct task_struct *tsk);
1867 #define VM_MAX_READAHEAD 128 /* kbytes */
1868 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1870 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1871 pgoff_t offset, unsigned long nr_to_read);
1873 void page_cache_sync_readahead(struct address_space *mapping,
1874 struct file_ra_state *ra,
1877 unsigned long size);
1879 void page_cache_async_readahead(struct address_space *mapping,
1880 struct file_ra_state *ra,
1884 unsigned long size);
1886 unsigned long max_sane_readahead(unsigned long nr);
1888 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1889 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1891 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1892 extern int expand_downwards(struct vm_area_struct *vma,
1893 unsigned long address);
1895 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1897 #define expand_upwards(vma, address) do { } while (0)
1900 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1901 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1902 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1903 struct vm_area_struct **pprev);
1905 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1906 NULL if none. Assume start_addr < end_addr. */
1907 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1909 struct vm_area_struct * vma = find_vma(mm,start_addr);
1911 if (vma && end_addr <= vma->vm_start)
1916 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1918 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1921 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1922 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1923 unsigned long vm_start, unsigned long vm_end)
1925 struct vm_area_struct *vma = find_vma(mm, vm_start);
1927 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1934 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1936 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1942 #ifdef CONFIG_NUMA_BALANCING
1943 unsigned long change_prot_numa(struct vm_area_struct *vma,
1944 unsigned long start, unsigned long end);
1947 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1948 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1949 unsigned long pfn, unsigned long size, pgprot_t);
1950 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1951 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1953 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1955 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1958 struct page *follow_page_mask(struct vm_area_struct *vma,
1959 unsigned long address, unsigned int foll_flags,
1960 unsigned int *page_mask);
1962 static inline struct page *follow_page(struct vm_area_struct *vma,
1963 unsigned long address, unsigned int foll_flags)
1965 unsigned int unused_page_mask;
1966 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1969 #define FOLL_WRITE 0x01 /* check pte is writable */
1970 #define FOLL_TOUCH 0x02 /* mark page accessed */
1971 #define FOLL_GET 0x04 /* do get_page on page */
1972 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1973 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1974 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1975 * and return without waiting upon it */
1976 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1977 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1978 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1979 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1980 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1982 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1984 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1985 unsigned long size, pte_fn_t fn, void *data);
1987 #ifdef CONFIG_PROC_FS
1988 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1990 static inline void vm_stat_account(struct mm_struct *mm,
1991 unsigned long flags, struct file *file, long pages)
1993 mm->total_vm += pages;
1995 #endif /* CONFIG_PROC_FS */
1997 #ifdef CONFIG_DEBUG_PAGEALLOC
1998 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1999 #ifdef CONFIG_HIBERNATION
2000 extern bool kernel_page_present(struct page *page);
2001 #endif /* CONFIG_HIBERNATION */
2004 kernel_map_pages(struct page *page, int numpages, int enable) {}
2005 #ifdef CONFIG_HIBERNATION
2006 static inline bool kernel_page_present(struct page *page) { return true; }
2007 #endif /* CONFIG_HIBERNATION */
2010 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2011 #ifdef __HAVE_ARCH_GATE_AREA
2012 int in_gate_area_no_mm(unsigned long addr);
2013 int in_gate_area(struct mm_struct *mm, unsigned long addr);
2015 int in_gate_area_no_mm(unsigned long addr);
2016 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
2017 #endif /* __HAVE_ARCH_GATE_AREA */
2019 #ifdef CONFIG_SYSCTL
2020 extern int sysctl_drop_caches;
2021 int drop_caches_sysctl_handler(struct ctl_table *, int,
2022 void __user *, size_t *, loff_t *);
2025 unsigned long shrink_slab(struct shrink_control *shrink,
2026 unsigned long nr_pages_scanned,
2027 unsigned long lru_pages);
2030 #define randomize_va_space 0
2032 extern int randomize_va_space;
2035 const char * arch_vma_name(struct vm_area_struct *vma);
2036 void print_vma_addr(char *prefix, unsigned long rip);
2038 void sparse_mem_maps_populate_node(struct page **map_map,
2039 unsigned long pnum_begin,
2040 unsigned long pnum_end,
2041 unsigned long map_count,
2044 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2045 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2046 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2047 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2048 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2049 void *vmemmap_alloc_block(unsigned long size, int node);
2050 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2051 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2052 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2054 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2055 void vmemmap_populate_print_last(void);
2056 #ifdef CONFIG_MEMORY_HOTPLUG
2057 void vmemmap_free(unsigned long start, unsigned long end);
2059 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2060 unsigned long size);
2063 MF_COUNT_INCREASED = 1 << 0,
2064 MF_ACTION_REQUIRED = 1 << 1,
2065 MF_MUST_KILL = 1 << 2,
2066 MF_SOFT_OFFLINE = 1 << 3,
2068 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2069 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2070 extern int unpoison_memory(unsigned long pfn);
2071 extern int sysctl_memory_failure_early_kill;
2072 extern int sysctl_memory_failure_recovery;
2073 extern void shake_page(struct page *p, int access);
2074 extern atomic_long_t num_poisoned_pages;
2075 extern int soft_offline_page(struct page *page, int flags);
2077 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2078 extern void clear_huge_page(struct page *page,
2080 unsigned int pages_per_huge_page);
2081 extern void copy_user_huge_page(struct page *dst, struct page *src,
2082 unsigned long addr, struct vm_area_struct *vma,
2083 unsigned int pages_per_huge_page);
2084 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2086 #ifdef CONFIG_DEBUG_PAGEALLOC
2087 extern unsigned int _debug_guardpage_minorder;
2089 static inline unsigned int debug_guardpage_minorder(void)
2091 return _debug_guardpage_minorder;
2094 static inline bool page_is_guard(struct page *page)
2096 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2099 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2100 static inline bool page_is_guard(struct page *page) { return false; }
2101 #endif /* CONFIG_DEBUG_PAGEALLOC */
2103 #if MAX_NUMNODES > 1
2104 void __init setup_nr_node_ids(void);
2106 static inline void setup_nr_node_ids(void) {}
2109 #endif /* __KERNEL__ */
2110 #endif /* _LINUX_MM_H */