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>
21 #include <linux/resource.h>
25 struct anon_vma_chain;
28 struct writeback_control;
30 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
31 extern unsigned long max_mapnr;
33 static inline void set_max_mapnr(unsigned long limit)
38 static inline void set_max_mapnr(unsigned long limit) { }
41 extern unsigned long totalram_pages;
42 extern void * high_memory;
43 extern int page_cluster;
46 extern int sysctl_legacy_va_layout;
48 #define sysctl_legacy_va_layout 0
52 #include <asm/pgtable.h>
53 #include <asm/processor.h>
56 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
59 extern unsigned long sysctl_user_reserve_kbytes;
60 extern unsigned long sysctl_admin_reserve_kbytes;
62 extern int sysctl_overcommit_memory;
63 extern int sysctl_overcommit_ratio;
64 extern unsigned long sysctl_overcommit_kbytes;
66 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
68 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
71 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
73 /* to align the pointer to the (next) page boundary */
74 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
76 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
77 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
80 * Linux kernel virtual memory manager primitives.
81 * The idea being to have a "virtual" mm in the same way
82 * we have a virtual fs - giving a cleaner interface to the
83 * mm details, and allowing different kinds of memory mappings
84 * (from shared memory to executable loading to arbitrary
88 extern struct kmem_cache *vm_area_cachep;
91 extern struct rb_root nommu_region_tree;
92 extern struct rw_semaphore nommu_region_sem;
94 extern unsigned int kobjsize(const void *objp);
98 * vm_flags in vm_area_struct, see mm_types.h.
100 #define VM_NONE 0x00000000
102 #define VM_READ 0x00000001 /* currently active flags */
103 #define VM_WRITE 0x00000002
104 #define VM_EXEC 0x00000004
105 #define VM_SHARED 0x00000008
107 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
108 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
109 #define VM_MAYWRITE 0x00000020
110 #define VM_MAYEXEC 0x00000040
111 #define VM_MAYSHARE 0x00000080
113 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
114 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
115 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
117 #define VM_LOCKED 0x00002000
118 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
120 /* Used by sys_madvise() */
121 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
122 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
124 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
125 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
126 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
127 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
128 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
129 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
130 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
131 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
133 #ifdef CONFIG_MEM_SOFT_DIRTY
134 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
136 # define VM_SOFTDIRTY 0
139 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
140 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
141 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
142 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
144 #if defined(CONFIG_X86)
145 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
146 #elif defined(CONFIG_PPC)
147 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
148 #elif defined(CONFIG_PARISC)
149 # define VM_GROWSUP VM_ARCH_1
150 #elif defined(CONFIG_METAG)
151 # define VM_GROWSUP VM_ARCH_1
152 #elif defined(CONFIG_IA64)
153 # define VM_GROWSUP VM_ARCH_1
154 #elif !defined(CONFIG_MMU)
155 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
159 # define VM_GROWSUP VM_NONE
162 /* Bits set in the VMA until the stack is in its final location */
163 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
165 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
166 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
169 #ifdef CONFIG_STACK_GROWSUP
170 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
172 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176 * Special vmas that are non-mergable, non-mlock()able.
177 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
179 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
181 /* This mask defines which mm->def_flags a process can inherit its parent */
182 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
185 * mapping from the currently active vm_flags protection bits (the
186 * low four bits) to a page protection mask..
188 extern pgprot_t protection_map[16];
190 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
191 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
192 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
193 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
194 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
195 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
196 #define FAULT_FLAG_TRIED 0x40 /* second try */
197 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
200 * vm_fault is filled by the the pagefault handler and passed to the vma's
201 * ->fault function. The vma's ->fault is responsible for returning a bitmask
202 * of VM_FAULT_xxx flags that give details about how the fault was handled.
204 * pgoff should be used in favour of virtual_address, if possible. If pgoff
205 * is used, one may implement ->remap_pages to get nonlinear mapping support.
208 unsigned int flags; /* FAULT_FLAG_xxx flags */
209 pgoff_t pgoff; /* Logical page offset based on vma */
210 void __user *virtual_address; /* Faulting virtual address */
212 struct page *page; /* ->fault handlers should return a
213 * page here, unless VM_FAULT_NOPAGE
214 * is set (which is also implied by
217 /* for ->map_pages() only */
218 pgoff_t max_pgoff; /* map pages for offset from pgoff till
219 * max_pgoff inclusive */
220 pte_t *pte; /* pte entry associated with ->pgoff */
224 * These are the virtual MM functions - opening of an area, closing and
225 * unmapping it (needed to keep files on disk up-to-date etc), pointer
226 * to the functions called when a no-page or a wp-page exception occurs.
228 struct vm_operations_struct {
229 void (*open)(struct vm_area_struct * area);
230 void (*close)(struct vm_area_struct * area);
231 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
232 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
234 /* notification that a previously read-only page is about to become
235 * writable, if an error is returned it will cause a SIGBUS */
236 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
238 /* called by access_process_vm when get_user_pages() fails, typically
239 * for use by special VMAs that can switch between memory and hardware
241 int (*access)(struct vm_area_struct *vma, unsigned long addr,
242 void *buf, int len, int write);
244 /* Called by the /proc/PID/maps code to ask the vma whether it
245 * has a special name. Returning non-NULL will also cause this
246 * vma to be dumped unconditionally. */
247 const char *(*name)(struct vm_area_struct *vma);
251 * set_policy() op must add a reference to any non-NULL @new mempolicy
252 * to hold the policy upon return. Caller should pass NULL @new to
253 * remove a policy and fall back to surrounding context--i.e. do not
254 * install a MPOL_DEFAULT policy, nor the task or system default
257 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
260 * get_policy() op must add reference [mpol_get()] to any policy at
261 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
262 * in mm/mempolicy.c will do this automatically.
263 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
264 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
265 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
266 * must return NULL--i.e., do not "fallback" to task or system default
269 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
271 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
272 const nodemask_t *to, unsigned long flags);
274 /* called by sys_remap_file_pages() to populate non-linear mapping */
275 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
276 unsigned long size, pgoff_t pgoff);
282 #define page_private(page) ((page)->private)
283 #define set_page_private(page, v) ((page)->private = (v))
285 /* It's valid only if the page is free path or free_list */
286 static inline void set_freepage_migratetype(struct page *page, int migratetype)
288 page->index = migratetype;
291 /* It's valid only if the page is free path or free_list */
292 static inline int get_freepage_migratetype(struct page *page)
298 * FIXME: take this include out, include page-flags.h in
299 * files which need it (119 of them)
301 #include <linux/page-flags.h>
302 #include <linux/huge_mm.h>
305 * Methods to modify the page usage count.
307 * What counts for a page usage:
308 * - cache mapping (page->mapping)
309 * - private data (page->private)
310 * - page mapped in a task's page tables, each mapping
311 * is counted separately
313 * Also, many kernel routines increase the page count before a critical
314 * routine so they can be sure the page doesn't go away from under them.
318 * Drop a ref, return true if the refcount fell to zero (the page has no users)
320 static inline int put_page_testzero(struct page *page)
322 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
323 return atomic_dec_and_test(&page->_count);
327 * Try to grab a ref unless the page has a refcount of zero, return false if
329 * This can be called when MMU is off so it must not access
330 * any of the virtual mappings.
332 static inline int get_page_unless_zero(struct page *page)
334 return atomic_inc_not_zero(&page->_count);
338 * Try to drop a ref unless the page has a refcount of one, return false if
340 * This is to make sure that the refcount won't become zero after this drop.
341 * This can be called when MMU is off so it must not access
342 * any of the virtual mappings.
344 static inline int put_page_unless_one(struct page *page)
346 return atomic_add_unless(&page->_count, -1, 1);
349 extern int page_is_ram(unsigned long pfn);
351 /* Support for virtually mapped pages */
352 struct page *vmalloc_to_page(const void *addr);
353 unsigned long vmalloc_to_pfn(const void *addr);
356 * Determine if an address is within the vmalloc range
358 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
359 * is no special casing required.
361 static inline int is_vmalloc_addr(const void *x)
364 unsigned long addr = (unsigned long)x;
366 return addr >= VMALLOC_START && addr < VMALLOC_END;
372 extern int is_vmalloc_or_module_addr(const void *x);
374 static inline int is_vmalloc_or_module_addr(const void *x)
380 extern void kvfree(const void *addr);
382 static inline void compound_lock(struct page *page)
384 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
385 VM_BUG_ON_PAGE(PageSlab(page), page);
386 bit_spin_lock(PG_compound_lock, &page->flags);
390 static inline void compound_unlock(struct page *page)
392 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
393 VM_BUG_ON_PAGE(PageSlab(page), page);
394 bit_spin_unlock(PG_compound_lock, &page->flags);
398 static inline unsigned long compound_lock_irqsave(struct page *page)
400 unsigned long uninitialized_var(flags);
401 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
402 local_irq_save(flags);
408 static inline void compound_unlock_irqrestore(struct page *page,
411 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
412 compound_unlock(page);
413 local_irq_restore(flags);
417 static inline struct page *compound_head_by_tail(struct page *tail)
419 struct page *head = tail->first_page;
422 * page->first_page may be a dangling pointer to an old
423 * compound page, so recheck that it is still a tail
424 * page before returning.
427 if (likely(PageTail(tail)))
432 static inline struct page *compound_head(struct page *page)
434 if (unlikely(PageTail(page)))
435 return compound_head_by_tail(page);
440 * The atomic page->_mapcount, starts from -1: so that transitions
441 * both from it and to it can be tracked, using atomic_inc_and_test
442 * and atomic_add_negative(-1).
444 static inline void page_mapcount_reset(struct page *page)
446 atomic_set(&(page)->_mapcount, -1);
449 static inline int page_mapcount(struct page *page)
451 return atomic_read(&(page)->_mapcount) + 1;
454 static inline int page_count(struct page *page)
456 return atomic_read(&compound_head(page)->_count);
459 #ifdef CONFIG_HUGETLB_PAGE
460 extern int PageHeadHuge(struct page *page_head);
461 #else /* CONFIG_HUGETLB_PAGE */
462 static inline int PageHeadHuge(struct page *page_head)
466 #endif /* CONFIG_HUGETLB_PAGE */
468 static inline bool __compound_tail_refcounted(struct page *page)
470 return !PageSlab(page) && !PageHeadHuge(page);
474 * This takes a head page as parameter and tells if the
475 * tail page reference counting can be skipped.
477 * For this to be safe, PageSlab and PageHeadHuge must remain true on
478 * any given page where they return true here, until all tail pins
479 * have been released.
481 static inline bool compound_tail_refcounted(struct page *page)
483 VM_BUG_ON_PAGE(!PageHead(page), page);
484 return __compound_tail_refcounted(page);
487 static inline void get_huge_page_tail(struct page *page)
490 * __split_huge_page_refcount() cannot run from under us.
492 VM_BUG_ON_PAGE(!PageTail(page), page);
493 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
494 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
495 if (compound_tail_refcounted(page->first_page))
496 atomic_inc(&page->_mapcount);
499 extern bool __get_page_tail(struct page *page);
501 static inline void get_page(struct page *page)
503 if (unlikely(PageTail(page)))
504 if (likely(__get_page_tail(page)))
507 * Getting a normal page or the head of a compound page
508 * requires to already have an elevated page->_count.
510 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
511 atomic_inc(&page->_count);
514 static inline struct page *virt_to_head_page(const void *x)
516 struct page *page = virt_to_page(x);
517 return compound_head(page);
521 * Setup the page count before being freed into the page allocator for
522 * the first time (boot or memory hotplug)
524 static inline void init_page_count(struct page *page)
526 atomic_set(&page->_count, 1);
530 * PageBuddy() indicate that the page is free and in the buddy system
531 * (see mm/page_alloc.c).
533 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
534 * -2 so that an underflow of the page_mapcount() won't be mistaken
535 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
536 * efficiently by most CPU architectures.
538 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
540 static inline int PageBuddy(struct page *page)
542 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
545 static inline void __SetPageBuddy(struct page *page)
547 VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page);
548 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
551 static inline void __ClearPageBuddy(struct page *page)
553 VM_BUG_ON_PAGE(!PageBuddy(page), page);
554 atomic_set(&page->_mapcount, -1);
557 void put_page(struct page *page);
558 void put_pages_list(struct list_head *pages);
560 void split_page(struct page *page, unsigned int order);
561 int split_free_page(struct page *page);
564 * Compound pages have a destructor function. Provide a
565 * prototype for that function and accessor functions.
566 * These are _only_ valid on the head of a PG_compound page.
568 typedef void compound_page_dtor(struct page *);
570 static inline void set_compound_page_dtor(struct page *page,
571 compound_page_dtor *dtor)
573 page[1].lru.next = (void *)dtor;
576 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
578 return (compound_page_dtor *)page[1].lru.next;
581 static inline int compound_order(struct page *page)
585 return (unsigned long)page[1].lru.prev;
588 static inline void set_compound_order(struct page *page, unsigned long order)
590 page[1].lru.prev = (void *)order;
595 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
596 * servicing faults for write access. In the normal case, do always want
597 * pte_mkwrite. But get_user_pages can cause write faults for mappings
598 * that do not have writing enabled, when used by access_process_vm.
600 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
602 if (likely(vma->vm_flags & VM_WRITE))
603 pte = pte_mkwrite(pte);
607 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
608 struct page *page, pte_t *pte, bool write, bool anon);
612 * Multiple processes may "see" the same page. E.g. for untouched
613 * mappings of /dev/null, all processes see the same page full of
614 * zeroes, and text pages of executables and shared libraries have
615 * only one copy in memory, at most, normally.
617 * For the non-reserved pages, page_count(page) denotes a reference count.
618 * page_count() == 0 means the page is free. page->lru is then used for
619 * freelist management in the buddy allocator.
620 * page_count() > 0 means the page has been allocated.
622 * Pages are allocated by the slab allocator in order to provide memory
623 * to kmalloc and kmem_cache_alloc. In this case, the management of the
624 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
625 * unless a particular usage is carefully commented. (the responsibility of
626 * freeing the kmalloc memory is the caller's, of course).
628 * A page may be used by anyone else who does a __get_free_page().
629 * In this case, page_count still tracks the references, and should only
630 * be used through the normal accessor functions. The top bits of page->flags
631 * and page->virtual store page management information, but all other fields
632 * are unused and could be used privately, carefully. The management of this
633 * page is the responsibility of the one who allocated it, and those who have
634 * subsequently been given references to it.
636 * The other pages (we may call them "pagecache pages") are completely
637 * managed by the Linux memory manager: I/O, buffers, swapping etc.
638 * The following discussion applies only to them.
640 * A pagecache page contains an opaque `private' member, which belongs to the
641 * page's address_space. Usually, this is the address of a circular list of
642 * the page's disk buffers. PG_private must be set to tell the VM to call
643 * into the filesystem to release these pages.
645 * A page may belong to an inode's memory mapping. In this case, page->mapping
646 * is the pointer to the inode, and page->index is the file offset of the page,
647 * in units of PAGE_CACHE_SIZE.
649 * If pagecache pages are not associated with an inode, they are said to be
650 * anonymous pages. These may become associated with the swapcache, and in that
651 * case PG_swapcache is set, and page->private is an offset into the swapcache.
653 * In either case (swapcache or inode backed), the pagecache itself holds one
654 * reference to the page. Setting PG_private should also increment the
655 * refcount. The each user mapping also has a reference to the page.
657 * The pagecache pages are stored in a per-mapping radix tree, which is
658 * rooted at mapping->page_tree, and indexed by offset.
659 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
660 * lists, we instead now tag pages as dirty/writeback in the radix tree.
662 * All pagecache pages may be subject to I/O:
663 * - inode pages may need to be read from disk,
664 * - inode pages which have been modified and are MAP_SHARED may need
665 * to be written back to the inode on disk,
666 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
667 * modified may need to be swapped out to swap space and (later) to be read
672 * The zone field is never updated after free_area_init_core()
673 * sets it, so none of the operations on it need to be atomic.
676 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
677 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
678 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
679 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
680 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
683 * Define the bit shifts to access each section. For non-existent
684 * sections we define the shift as 0; that plus a 0 mask ensures
685 * the compiler will optimise away reference to them.
687 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
688 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
689 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
690 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
692 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
693 #ifdef NODE_NOT_IN_PAGE_FLAGS
694 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
695 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
696 SECTIONS_PGOFF : ZONES_PGOFF)
698 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
699 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
700 NODES_PGOFF : ZONES_PGOFF)
703 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
705 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
706 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
709 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
710 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
711 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
712 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
713 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
715 static inline enum zone_type page_zonenum(const struct page *page)
717 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
720 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
721 #define SECTION_IN_PAGE_FLAGS
725 * The identification function is mainly used by the buddy allocator for
726 * determining if two pages could be buddies. We are not really identifying
727 * the zone since we could be using the section number id if we do not have
728 * node id available in page flags.
729 * We only guarantee that it will return the same value for two combinable
732 static inline int page_zone_id(struct page *page)
734 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
737 static inline int zone_to_nid(struct zone *zone)
746 #ifdef NODE_NOT_IN_PAGE_FLAGS
747 extern int page_to_nid(const struct page *page);
749 static inline int page_to_nid(const struct page *page)
751 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
755 #ifdef CONFIG_NUMA_BALANCING
756 static inline int cpu_pid_to_cpupid(int cpu, int pid)
758 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
761 static inline int cpupid_to_pid(int cpupid)
763 return cpupid & LAST__PID_MASK;
766 static inline int cpupid_to_cpu(int cpupid)
768 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
771 static inline int cpupid_to_nid(int cpupid)
773 return cpu_to_node(cpupid_to_cpu(cpupid));
776 static inline bool cpupid_pid_unset(int cpupid)
778 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
781 static inline bool cpupid_cpu_unset(int cpupid)
783 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
786 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
788 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
791 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
792 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
793 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
795 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
798 static inline int page_cpupid_last(struct page *page)
800 return page->_last_cpupid;
802 static inline void page_cpupid_reset_last(struct page *page)
804 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
807 static inline int page_cpupid_last(struct page *page)
809 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
812 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
814 static inline void page_cpupid_reset_last(struct page *page)
816 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
818 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
819 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
821 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
822 #else /* !CONFIG_NUMA_BALANCING */
823 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
825 return page_to_nid(page); /* XXX */
828 static inline int page_cpupid_last(struct page *page)
830 return page_to_nid(page); /* XXX */
833 static inline int cpupid_to_nid(int cpupid)
838 static inline int cpupid_to_pid(int cpupid)
843 static inline int cpupid_to_cpu(int cpupid)
848 static inline int cpu_pid_to_cpupid(int nid, int pid)
853 static inline bool cpupid_pid_unset(int cpupid)
858 static inline void page_cpupid_reset_last(struct page *page)
862 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
866 #endif /* CONFIG_NUMA_BALANCING */
868 static inline struct zone *page_zone(const struct page *page)
870 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
873 #ifdef SECTION_IN_PAGE_FLAGS
874 static inline void set_page_section(struct page *page, unsigned long section)
876 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
877 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
880 static inline unsigned long page_to_section(const struct page *page)
882 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
886 static inline void set_page_zone(struct page *page, enum zone_type zone)
888 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
889 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
892 static inline void set_page_node(struct page *page, unsigned long node)
894 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
895 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
898 static inline void set_page_links(struct page *page, enum zone_type zone,
899 unsigned long node, unsigned long pfn)
901 set_page_zone(page, zone);
902 set_page_node(page, node);
903 #ifdef SECTION_IN_PAGE_FLAGS
904 set_page_section(page, pfn_to_section_nr(pfn));
909 * Some inline functions in vmstat.h depend on page_zone()
911 #include <linux/vmstat.h>
913 static __always_inline void *lowmem_page_address(const struct page *page)
915 return __va(PFN_PHYS(page_to_pfn(page)));
918 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
919 #define HASHED_PAGE_VIRTUAL
922 #if defined(WANT_PAGE_VIRTUAL)
923 static inline void *page_address(const struct page *page)
925 return page->virtual;
927 static inline void set_page_address(struct page *page, void *address)
929 page->virtual = address;
931 #define page_address_init() do { } while(0)
934 #if defined(HASHED_PAGE_VIRTUAL)
935 void *page_address(const struct page *page);
936 void set_page_address(struct page *page, void *virtual);
937 void page_address_init(void);
940 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
941 #define page_address(page) lowmem_page_address(page)
942 #define set_page_address(page, address) do { } while(0)
943 #define page_address_init() do { } while(0)
947 * On an anonymous page mapped into a user virtual memory area,
948 * page->mapping points to its anon_vma, not to a struct address_space;
949 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
951 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
952 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
953 * and then page->mapping points, not to an anon_vma, but to a private
954 * structure which KSM associates with that merged page. See ksm.h.
956 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
958 * Please note that, confusingly, "page_mapping" refers to the inode
959 * address_space which maps the page from disk; whereas "page_mapped"
960 * refers to user virtual address space into which the page is mapped.
962 #define PAGE_MAPPING_ANON 1
963 #define PAGE_MAPPING_KSM 2
964 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
966 extern struct address_space *page_mapping(struct page *page);
968 /* Neutral page->mapping pointer to address_space or anon_vma or other */
969 static inline void *page_rmapping(struct page *page)
971 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
974 extern struct address_space *__page_file_mapping(struct page *);
977 struct address_space *page_file_mapping(struct page *page)
979 if (unlikely(PageSwapCache(page)))
980 return __page_file_mapping(page);
982 return page->mapping;
985 static inline int PageAnon(struct page *page)
987 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
991 * Return the pagecache index of the passed page. Regular pagecache pages
992 * use ->index whereas swapcache pages use ->private
994 static inline pgoff_t page_index(struct page *page)
996 if (unlikely(PageSwapCache(page)))
997 return page_private(page);
1001 extern pgoff_t __page_file_index(struct page *page);
1004 * Return the file index of the page. Regular pagecache pages use ->index
1005 * whereas swapcache pages use swp_offset(->private)
1007 static inline pgoff_t page_file_index(struct page *page)
1009 if (unlikely(PageSwapCache(page)))
1010 return __page_file_index(page);
1016 * Return true if this page is mapped into pagetables.
1018 static inline int page_mapped(struct page *page)
1020 return atomic_read(&(page)->_mapcount) >= 0;
1024 * Different kinds of faults, as returned by handle_mm_fault().
1025 * Used to decide whether a process gets delivered SIGBUS or
1026 * just gets major/minor fault counters bumped up.
1029 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1031 #define VM_FAULT_OOM 0x0001
1032 #define VM_FAULT_SIGBUS 0x0002
1033 #define VM_FAULT_MAJOR 0x0004
1034 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1035 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1036 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1038 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1039 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1040 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1041 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1043 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1045 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
1046 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
1048 /* Encode hstate index for a hwpoisoned large page */
1049 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1050 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1053 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1055 extern void pagefault_out_of_memory(void);
1057 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1060 * Flags passed to show_mem() and show_free_areas() to suppress output in
1063 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1065 extern void show_free_areas(unsigned int flags);
1066 extern bool skip_free_areas_node(unsigned int flags, int nid);
1068 int shmem_zero_setup(struct vm_area_struct *);
1070 bool shmem_mapping(struct address_space *mapping);
1072 static inline bool shmem_mapping(struct address_space *mapping)
1078 extern int can_do_mlock(void);
1079 extern int user_shm_lock(size_t, struct user_struct *);
1080 extern void user_shm_unlock(size_t, struct user_struct *);
1083 * Parameter block passed down to zap_pte_range in exceptional cases.
1085 struct zap_details {
1086 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
1087 struct address_space *check_mapping; /* Check page->mapping if set */
1088 pgoff_t first_index; /* Lowest page->index to unmap */
1089 pgoff_t last_index; /* Highest page->index to unmap */
1092 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1095 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1096 unsigned long size);
1097 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1098 unsigned long size, struct zap_details *);
1099 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1100 unsigned long start, unsigned long end);
1103 * mm_walk - callbacks for walk_page_range
1104 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1105 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1106 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1107 * this handler is required to be able to handle
1108 * pmd_trans_huge() pmds. They may simply choose to
1109 * split_huge_page() instead of handling it explicitly.
1110 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1111 * @pte_hole: if set, called for each hole at all levels
1112 * @hugetlb_entry: if set, called for each hugetlb entry
1113 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1116 * (see walk_page_range for more details)
1119 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1120 unsigned long next, struct mm_walk *walk);
1121 int (*pud_entry)(pud_t *pud, unsigned long addr,
1122 unsigned long next, struct mm_walk *walk);
1123 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1124 unsigned long next, struct mm_walk *walk);
1125 int (*pte_entry)(pte_t *pte, unsigned long addr,
1126 unsigned long next, struct mm_walk *walk);
1127 int (*pte_hole)(unsigned long addr, unsigned long next,
1128 struct mm_walk *walk);
1129 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1130 unsigned long addr, unsigned long next,
1131 struct mm_walk *walk);
1132 struct mm_struct *mm;
1136 int walk_page_range(unsigned long addr, unsigned long end,
1137 struct mm_walk *walk);
1138 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1139 unsigned long end, unsigned long floor, unsigned long ceiling);
1140 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1141 struct vm_area_struct *vma);
1142 void unmap_mapping_range(struct address_space *mapping,
1143 loff_t const holebegin, loff_t const holelen, int even_cows);
1144 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1145 unsigned long *pfn);
1146 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1147 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1148 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1149 void *buf, int len, int write);
1151 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1152 loff_t const holebegin, loff_t const holelen)
1154 unmap_mapping_range(mapping, holebegin, holelen, 0);
1157 extern void truncate_pagecache(struct inode *inode, loff_t new);
1158 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1159 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1160 int truncate_inode_page(struct address_space *mapping, struct page *page);
1161 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1162 int invalidate_inode_page(struct page *page);
1165 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1166 unsigned long address, unsigned int flags);
1167 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1168 unsigned long address, unsigned int fault_flags);
1170 static inline int handle_mm_fault(struct mm_struct *mm,
1171 struct vm_area_struct *vma, unsigned long address,
1174 /* should never happen if there's no MMU */
1176 return VM_FAULT_SIGBUS;
1178 static inline int fixup_user_fault(struct task_struct *tsk,
1179 struct mm_struct *mm, unsigned long address,
1180 unsigned int fault_flags)
1182 /* should never happen if there's no MMU */
1188 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1189 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1190 void *buf, int len, int write);
1192 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1193 unsigned long start, unsigned long nr_pages,
1194 unsigned int foll_flags, struct page **pages,
1195 struct vm_area_struct **vmas, int *nonblocking);
1196 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1197 unsigned long start, unsigned long nr_pages,
1198 int write, int force, struct page **pages,
1199 struct vm_area_struct **vmas);
1200 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1201 struct page **pages);
1203 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1204 struct page **pages);
1205 int get_kernel_page(unsigned long start, int write, struct page **pages);
1206 struct page *get_dump_page(unsigned long addr);
1208 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1209 extern void do_invalidatepage(struct page *page, unsigned int offset,
1210 unsigned int length);
1212 int __set_page_dirty_nobuffers(struct page *page);
1213 int __set_page_dirty_no_writeback(struct page *page);
1214 int redirty_page_for_writepage(struct writeback_control *wbc,
1216 void account_page_dirtied(struct page *page, struct address_space *mapping);
1217 void account_page_writeback(struct page *page);
1218 int set_page_dirty(struct page *page);
1219 int set_page_dirty_lock(struct page *page);
1220 int clear_page_dirty_for_io(struct page *page);
1221 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1223 /* Is the vma a continuation of the stack vma above it? */
1224 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1226 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1229 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1232 return (vma->vm_flags & VM_GROWSDOWN) &&
1233 (vma->vm_start == addr) &&
1234 !vma_growsdown(vma->vm_prev, addr);
1237 /* Is the vma a continuation of the stack vma below it? */
1238 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1240 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1243 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1246 return (vma->vm_flags & VM_GROWSUP) &&
1247 (vma->vm_end == addr) &&
1248 !vma_growsup(vma->vm_next, addr);
1251 extern struct task_struct *task_of_stack(struct task_struct *task,
1252 struct vm_area_struct *vma, bool in_group);
1254 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1255 unsigned long old_addr, struct vm_area_struct *new_vma,
1256 unsigned long new_addr, unsigned long len,
1257 bool need_rmap_locks);
1258 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1259 unsigned long end, pgprot_t newprot,
1260 int dirty_accountable, int prot_numa);
1261 extern int mprotect_fixup(struct vm_area_struct *vma,
1262 struct vm_area_struct **pprev, unsigned long start,
1263 unsigned long end, unsigned long newflags);
1266 * doesn't attempt to fault and will return short.
1268 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1269 struct page **pages);
1271 * per-process(per-mm_struct) statistics.
1273 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1275 long val = atomic_long_read(&mm->rss_stat.count[member]);
1277 #ifdef SPLIT_RSS_COUNTING
1279 * counter is updated in asynchronous manner and may go to minus.
1280 * But it's never be expected number for users.
1285 return (unsigned long)val;
1288 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1290 atomic_long_add(value, &mm->rss_stat.count[member]);
1293 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1295 atomic_long_inc(&mm->rss_stat.count[member]);
1298 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1300 atomic_long_dec(&mm->rss_stat.count[member]);
1303 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1305 return get_mm_counter(mm, MM_FILEPAGES) +
1306 get_mm_counter(mm, MM_ANONPAGES);
1309 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1311 return max(mm->hiwater_rss, get_mm_rss(mm));
1314 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1316 return max(mm->hiwater_vm, mm->total_vm);
1319 static inline void update_hiwater_rss(struct mm_struct *mm)
1321 unsigned long _rss = get_mm_rss(mm);
1323 if ((mm)->hiwater_rss < _rss)
1324 (mm)->hiwater_rss = _rss;
1327 static inline void update_hiwater_vm(struct mm_struct *mm)
1329 if (mm->hiwater_vm < mm->total_vm)
1330 mm->hiwater_vm = mm->total_vm;
1333 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1334 struct mm_struct *mm)
1336 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1338 if (*maxrss < hiwater_rss)
1339 *maxrss = hiwater_rss;
1342 #if defined(SPLIT_RSS_COUNTING)
1343 void sync_mm_rss(struct mm_struct *mm);
1345 static inline void sync_mm_rss(struct mm_struct *mm)
1350 int vma_wants_writenotify(struct vm_area_struct *vma);
1352 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1354 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1358 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1362 #ifdef __PAGETABLE_PUD_FOLDED
1363 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1364 unsigned long address)
1369 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1372 #ifdef __PAGETABLE_PMD_FOLDED
1373 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1374 unsigned long address)
1379 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1382 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1383 pmd_t *pmd, unsigned long address);
1384 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1387 * The following ifdef needed to get the 4level-fixup.h header to work.
1388 * Remove it when 4level-fixup.h has been removed.
1390 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1391 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1393 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1394 NULL: pud_offset(pgd, address);
1397 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1399 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1400 NULL: pmd_offset(pud, address);
1402 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1404 #if USE_SPLIT_PTE_PTLOCKS
1405 #if ALLOC_SPLIT_PTLOCKS
1406 void __init ptlock_cache_init(void);
1407 extern bool ptlock_alloc(struct page *page);
1408 extern void ptlock_free(struct page *page);
1410 static inline spinlock_t *ptlock_ptr(struct page *page)
1414 #else /* ALLOC_SPLIT_PTLOCKS */
1415 static inline void ptlock_cache_init(void)
1419 static inline bool ptlock_alloc(struct page *page)
1424 static inline void ptlock_free(struct page *page)
1428 static inline spinlock_t *ptlock_ptr(struct page *page)
1432 #endif /* ALLOC_SPLIT_PTLOCKS */
1434 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1436 return ptlock_ptr(pmd_page(*pmd));
1439 static inline bool ptlock_init(struct page *page)
1442 * prep_new_page() initialize page->private (and therefore page->ptl)
1443 * with 0. Make sure nobody took it in use in between.
1445 * It can happen if arch try to use slab for page table allocation:
1446 * slab code uses page->slab_cache and page->first_page (for tail
1447 * pages), which share storage with page->ptl.
1449 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1450 if (!ptlock_alloc(page))
1452 spin_lock_init(ptlock_ptr(page));
1456 /* Reset page->mapping so free_pages_check won't complain. */
1457 static inline void pte_lock_deinit(struct page *page)
1459 page->mapping = NULL;
1463 #else /* !USE_SPLIT_PTE_PTLOCKS */
1465 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1467 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1469 return &mm->page_table_lock;
1471 static inline void ptlock_cache_init(void) {}
1472 static inline bool ptlock_init(struct page *page) { return true; }
1473 static inline void pte_lock_deinit(struct page *page) {}
1474 #endif /* USE_SPLIT_PTE_PTLOCKS */
1476 static inline void pgtable_init(void)
1478 ptlock_cache_init();
1479 pgtable_cache_init();
1482 static inline bool pgtable_page_ctor(struct page *page)
1484 inc_zone_page_state(page, NR_PAGETABLE);
1485 return ptlock_init(page);
1488 static inline void pgtable_page_dtor(struct page *page)
1490 pte_lock_deinit(page);
1491 dec_zone_page_state(page, NR_PAGETABLE);
1494 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1496 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1497 pte_t *__pte = pte_offset_map(pmd, address); \
1503 #define pte_unmap_unlock(pte, ptl) do { \
1508 #define pte_alloc_map(mm, vma, pmd, address) \
1509 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1511 NULL: pte_offset_map(pmd, address))
1513 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1514 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1516 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1518 #define pte_alloc_kernel(pmd, address) \
1519 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1520 NULL: pte_offset_kernel(pmd, address))
1522 #if USE_SPLIT_PMD_PTLOCKS
1524 static struct page *pmd_to_page(pmd_t *pmd)
1526 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1527 return virt_to_page((void *)((unsigned long) pmd & mask));
1530 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1532 return ptlock_ptr(pmd_to_page(pmd));
1535 static inline bool pgtable_pmd_page_ctor(struct page *page)
1537 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1538 page->pmd_huge_pte = NULL;
1540 return ptlock_init(page);
1543 static inline void pgtable_pmd_page_dtor(struct page *page)
1545 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1546 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1551 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1555 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1557 return &mm->page_table_lock;
1560 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1561 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1563 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1567 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1569 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1574 extern void free_area_init(unsigned long * zones_size);
1575 extern void free_area_init_node(int nid, unsigned long * zones_size,
1576 unsigned long zone_start_pfn, unsigned long *zholes_size);
1577 extern void free_initmem(void);
1580 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1581 * into the buddy system. The freed pages will be poisoned with pattern
1582 * "poison" if it's within range [0, UCHAR_MAX].
1583 * Return pages freed into the buddy system.
1585 extern unsigned long free_reserved_area(void *start, void *end,
1586 int poison, char *s);
1588 #ifdef CONFIG_HIGHMEM
1590 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1591 * and totalram_pages.
1593 extern void free_highmem_page(struct page *page);
1596 extern void adjust_managed_page_count(struct page *page, long count);
1597 extern void mem_init_print_info(const char *str);
1599 /* Free the reserved page into the buddy system, so it gets managed. */
1600 static inline void __free_reserved_page(struct page *page)
1602 ClearPageReserved(page);
1603 init_page_count(page);
1607 static inline void free_reserved_page(struct page *page)
1609 __free_reserved_page(page);
1610 adjust_managed_page_count(page, 1);
1613 static inline void mark_page_reserved(struct page *page)
1615 SetPageReserved(page);
1616 adjust_managed_page_count(page, -1);
1620 * Default method to free all the __init memory into the buddy system.
1621 * The freed pages will be poisoned with pattern "poison" if it's within
1622 * range [0, UCHAR_MAX].
1623 * Return pages freed into the buddy system.
1625 static inline unsigned long free_initmem_default(int poison)
1627 extern char __init_begin[], __init_end[];
1629 return free_reserved_area(&__init_begin, &__init_end,
1630 poison, "unused kernel");
1633 static inline unsigned long get_num_physpages(void)
1636 unsigned long phys_pages = 0;
1638 for_each_online_node(nid)
1639 phys_pages += node_present_pages(nid);
1644 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1646 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1647 * zones, allocate the backing mem_map and account for memory holes in a more
1648 * architecture independent manner. This is a substitute for creating the
1649 * zone_sizes[] and zholes_size[] arrays and passing them to
1650 * free_area_init_node()
1652 * An architecture is expected to register range of page frames backed by
1653 * physical memory with memblock_add[_node]() before calling
1654 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1655 * usage, an architecture is expected to do something like
1657 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1659 * for_each_valid_physical_page_range()
1660 * memblock_add_node(base, size, nid)
1661 * free_area_init_nodes(max_zone_pfns);
1663 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1664 * registered physical page range. Similarly
1665 * sparse_memory_present_with_active_regions() calls memory_present() for
1666 * each range when SPARSEMEM is enabled.
1668 * See mm/page_alloc.c for more information on each function exposed by
1669 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1671 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1672 unsigned long node_map_pfn_alignment(void);
1673 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1674 unsigned long end_pfn);
1675 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1676 unsigned long end_pfn);
1677 extern void get_pfn_range_for_nid(unsigned int nid,
1678 unsigned long *start_pfn, unsigned long *end_pfn);
1679 extern unsigned long find_min_pfn_with_active_regions(void);
1680 extern void free_bootmem_with_active_regions(int nid,
1681 unsigned long max_low_pfn);
1682 extern void sparse_memory_present_with_active_regions(int nid);
1684 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1686 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1687 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1688 static inline int __early_pfn_to_nid(unsigned long pfn)
1693 /* please see mm/page_alloc.c */
1694 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1695 /* there is a per-arch backend function. */
1696 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1699 extern void set_dma_reserve(unsigned long new_dma_reserve);
1700 extern void memmap_init_zone(unsigned long, int, unsigned long,
1701 unsigned long, enum memmap_context);
1702 extern void setup_per_zone_wmarks(void);
1703 extern int __meminit init_per_zone_wmark_min(void);
1704 extern void mem_init(void);
1705 extern void __init mmap_init(void);
1706 extern void show_mem(unsigned int flags);
1707 extern void si_meminfo(struct sysinfo * val);
1708 extern void si_meminfo_node(struct sysinfo *val, int nid);
1710 extern __printf(3, 4)
1711 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1713 extern void setup_per_cpu_pageset(void);
1715 extern void zone_pcp_update(struct zone *zone);
1716 extern void zone_pcp_reset(struct zone *zone);
1719 extern int min_free_kbytes;
1722 extern atomic_long_t mmap_pages_allocated;
1723 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1725 /* interval_tree.c */
1726 void vma_interval_tree_insert(struct vm_area_struct *node,
1727 struct rb_root *root);
1728 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1729 struct vm_area_struct *prev,
1730 struct rb_root *root);
1731 void vma_interval_tree_remove(struct vm_area_struct *node,
1732 struct rb_root *root);
1733 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1734 unsigned long start, unsigned long last);
1735 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1736 unsigned long start, unsigned long last);
1738 #define vma_interval_tree_foreach(vma, root, start, last) \
1739 for (vma = vma_interval_tree_iter_first(root, start, last); \
1740 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1742 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1743 struct list_head *list)
1745 list_add_tail(&vma->shared.nonlinear, list);
1748 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1749 struct rb_root *root);
1750 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1751 struct rb_root *root);
1752 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1753 struct rb_root *root, unsigned long start, unsigned long last);
1754 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1755 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1756 #ifdef CONFIG_DEBUG_VM_RB
1757 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1760 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1761 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1762 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1765 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1766 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1767 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1768 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1769 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1770 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1771 struct mempolicy *);
1772 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1773 extern int split_vma(struct mm_struct *,
1774 struct vm_area_struct *, unsigned long addr, int new_below);
1775 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1776 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1777 struct rb_node **, struct rb_node *);
1778 extern void unlink_file_vma(struct vm_area_struct *);
1779 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1780 unsigned long addr, unsigned long len, pgoff_t pgoff,
1781 bool *need_rmap_locks);
1782 extern void exit_mmap(struct mm_struct *);
1784 static inline int check_data_rlimit(unsigned long rlim,
1786 unsigned long start,
1787 unsigned long end_data,
1788 unsigned long start_data)
1790 if (rlim < RLIM_INFINITY) {
1791 if (((new - start) + (end_data - start_data)) > rlim)
1798 extern int mm_take_all_locks(struct mm_struct *mm);
1799 extern void mm_drop_all_locks(struct mm_struct *mm);
1801 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1802 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1804 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1805 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1806 unsigned long addr, unsigned long len,
1807 unsigned long flags,
1808 const struct vm_special_mapping *spec);
1809 /* This is an obsolete alternative to _install_special_mapping. */
1810 extern int install_special_mapping(struct mm_struct *mm,
1811 unsigned long addr, unsigned long len,
1812 unsigned long flags, struct page **pages);
1814 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1816 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1817 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1818 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1819 unsigned long len, unsigned long prot, unsigned long flags,
1820 unsigned long pgoff, unsigned long *populate);
1821 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1824 extern int __mm_populate(unsigned long addr, unsigned long len,
1826 static inline void mm_populate(unsigned long addr, unsigned long len)
1829 (void) __mm_populate(addr, len, 1);
1832 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1835 /* These take the mm semaphore themselves */
1836 extern unsigned long vm_brk(unsigned long, unsigned long);
1837 extern int vm_munmap(unsigned long, size_t);
1838 extern unsigned long vm_mmap(struct file *, unsigned long,
1839 unsigned long, unsigned long,
1840 unsigned long, unsigned long);
1842 struct vm_unmapped_area_info {
1843 #define VM_UNMAPPED_AREA_TOPDOWN 1
1844 unsigned long flags;
1845 unsigned long length;
1846 unsigned long low_limit;
1847 unsigned long high_limit;
1848 unsigned long align_mask;
1849 unsigned long align_offset;
1852 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1853 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1856 * Search for an unmapped address range.
1858 * We are looking for a range that:
1859 * - does not intersect with any VMA;
1860 * - is contained within the [low_limit, high_limit) interval;
1861 * - is at least the desired size.
1862 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1864 static inline unsigned long
1865 vm_unmapped_area(struct vm_unmapped_area_info *info)
1867 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1868 return unmapped_area(info);
1870 return unmapped_area_topdown(info);
1874 extern void truncate_inode_pages(struct address_space *, loff_t);
1875 extern void truncate_inode_pages_range(struct address_space *,
1876 loff_t lstart, loff_t lend);
1877 extern void truncate_inode_pages_final(struct address_space *);
1879 /* generic vm_area_ops exported for stackable file systems */
1880 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1881 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1882 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1884 /* mm/page-writeback.c */
1885 int write_one_page(struct page *page, int wait);
1886 void task_dirty_inc(struct task_struct *tsk);
1889 #define VM_MAX_READAHEAD 128 /* kbytes */
1890 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1892 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1893 pgoff_t offset, unsigned long nr_to_read);
1895 void page_cache_sync_readahead(struct address_space *mapping,
1896 struct file_ra_state *ra,
1899 unsigned long size);
1901 void page_cache_async_readahead(struct address_space *mapping,
1902 struct file_ra_state *ra,
1906 unsigned long size);
1908 unsigned long max_sane_readahead(unsigned long nr);
1910 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1911 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1913 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1914 extern int expand_downwards(struct vm_area_struct *vma,
1915 unsigned long address);
1917 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1919 #define expand_upwards(vma, address) do { } while (0)
1922 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1923 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1924 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1925 struct vm_area_struct **pprev);
1927 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1928 NULL if none. Assume start_addr < end_addr. */
1929 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1931 struct vm_area_struct * vma = find_vma(mm,start_addr);
1933 if (vma && end_addr <= vma->vm_start)
1938 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1940 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1943 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1944 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1945 unsigned long vm_start, unsigned long vm_end)
1947 struct vm_area_struct *vma = find_vma(mm, vm_start);
1949 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1956 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1958 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1964 #ifdef CONFIG_NUMA_BALANCING
1965 unsigned long change_prot_numa(struct vm_area_struct *vma,
1966 unsigned long start, unsigned long end);
1969 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1970 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1971 unsigned long pfn, unsigned long size, pgprot_t);
1972 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1973 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1975 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1977 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1980 struct page *follow_page_mask(struct vm_area_struct *vma,
1981 unsigned long address, unsigned int foll_flags,
1982 unsigned int *page_mask);
1984 static inline struct page *follow_page(struct vm_area_struct *vma,
1985 unsigned long address, unsigned int foll_flags)
1987 unsigned int unused_page_mask;
1988 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1991 #define FOLL_WRITE 0x01 /* check pte is writable */
1992 #define FOLL_TOUCH 0x02 /* mark page accessed */
1993 #define FOLL_GET 0x04 /* do get_page on page */
1994 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1995 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1996 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1997 * and return without waiting upon it */
1998 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1999 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2000 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2001 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2002 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2003 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2005 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2007 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2008 unsigned long size, pte_fn_t fn, void *data);
2010 #ifdef CONFIG_PROC_FS
2011 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2013 static inline void vm_stat_account(struct mm_struct *mm,
2014 unsigned long flags, struct file *file, long pages)
2016 mm->total_vm += pages;
2018 #endif /* CONFIG_PROC_FS */
2020 #ifdef CONFIG_DEBUG_PAGEALLOC
2021 extern void kernel_map_pages(struct page *page, int numpages, int enable);
2022 #ifdef CONFIG_HIBERNATION
2023 extern bool kernel_page_present(struct page *page);
2024 #endif /* CONFIG_HIBERNATION */
2027 kernel_map_pages(struct page *page, int numpages, int enable) {}
2028 #ifdef CONFIG_HIBERNATION
2029 static inline bool kernel_page_present(struct page *page) { return true; }
2030 #endif /* CONFIG_HIBERNATION */
2033 #ifdef __HAVE_ARCH_GATE_AREA
2034 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2035 extern int in_gate_area_no_mm(unsigned long addr);
2036 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2038 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2042 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2043 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2047 #endif /* __HAVE_ARCH_GATE_AREA */
2049 #ifdef CONFIG_SYSCTL
2050 extern int sysctl_drop_caches;
2051 int drop_caches_sysctl_handler(struct ctl_table *, int,
2052 void __user *, size_t *, loff_t *);
2055 unsigned long shrink_slab(struct shrink_control *shrink,
2056 unsigned long nr_pages_scanned,
2057 unsigned long lru_pages);
2060 #define randomize_va_space 0
2062 extern int randomize_va_space;
2065 const char * arch_vma_name(struct vm_area_struct *vma);
2066 void print_vma_addr(char *prefix, unsigned long rip);
2068 void sparse_mem_maps_populate_node(struct page **map_map,
2069 unsigned long pnum_begin,
2070 unsigned long pnum_end,
2071 unsigned long map_count,
2074 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2075 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2076 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2077 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2078 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2079 void *vmemmap_alloc_block(unsigned long size, int node);
2080 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2081 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2082 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2084 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2085 void vmemmap_populate_print_last(void);
2086 #ifdef CONFIG_MEMORY_HOTPLUG
2087 void vmemmap_free(unsigned long start, unsigned long end);
2089 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2090 unsigned long size);
2093 MF_COUNT_INCREASED = 1 << 0,
2094 MF_ACTION_REQUIRED = 1 << 1,
2095 MF_MUST_KILL = 1 << 2,
2096 MF_SOFT_OFFLINE = 1 << 3,
2098 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2099 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2100 extern int unpoison_memory(unsigned long pfn);
2101 extern int sysctl_memory_failure_early_kill;
2102 extern int sysctl_memory_failure_recovery;
2103 extern void shake_page(struct page *p, int access);
2104 extern atomic_long_t num_poisoned_pages;
2105 extern int soft_offline_page(struct page *page, int flags);
2107 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2108 extern void clear_huge_page(struct page *page,
2110 unsigned int pages_per_huge_page);
2111 extern void copy_user_huge_page(struct page *dst, struct page *src,
2112 unsigned long addr, struct vm_area_struct *vma,
2113 unsigned int pages_per_huge_page);
2114 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2116 #ifdef CONFIG_DEBUG_PAGEALLOC
2117 extern unsigned int _debug_guardpage_minorder;
2119 static inline unsigned int debug_guardpage_minorder(void)
2121 return _debug_guardpage_minorder;
2124 static inline bool page_is_guard(struct page *page)
2126 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2129 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2130 static inline bool page_is_guard(struct page *page) { return false; }
2131 #endif /* CONFIG_DEBUG_PAGEALLOC */
2133 #if MAX_NUMNODES > 1
2134 void __init setup_nr_node_ids(void);
2136 static inline void setup_nr_node_ids(void) {}
2139 #endif /* __KERNEL__ */
2140 #endif /* _LINUX_MM_H */