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>
22 #include <linux/page_ext.h>
26 struct anon_vma_chain
;
29 struct writeback_control
;
31 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
32 extern unsigned long max_mapnr
;
34 static inline void set_max_mapnr(unsigned long limit
)
39 static inline void set_max_mapnr(unsigned long limit
) { }
42 extern unsigned long totalram_pages
;
43 extern void * high_memory
;
44 extern int page_cluster
;
47 extern int sysctl_legacy_va_layout
;
49 #define sysctl_legacy_va_layout 0
53 #include <asm/pgtable.h>
54 #include <asm/processor.h>
57 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
61 * To prevent common memory management code establishing
62 * a zero page mapping on a read fault.
63 * This macro should be defined within <asm/pgtable.h>.
64 * s390 does this to prevent multiplexing of hardware bits
65 * related to the physical page in case of virtualization.
67 #ifndef mm_forbids_zeropage
68 #define mm_forbids_zeropage(X) (0)
71 extern unsigned long sysctl_user_reserve_kbytes
;
72 extern unsigned long sysctl_admin_reserve_kbytes
;
74 extern int sysctl_overcommit_memory
;
75 extern int sysctl_overcommit_ratio
;
76 extern unsigned long sysctl_overcommit_kbytes
;
78 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
80 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
83 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
85 /* to align the pointer to the (next) page boundary */
86 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
88 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
89 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
92 * Linux kernel virtual memory manager primitives.
93 * The idea being to have a "virtual" mm in the same way
94 * we have a virtual fs - giving a cleaner interface to the
95 * mm details, and allowing different kinds of memory mappings
96 * (from shared memory to executable loading to arbitrary
100 extern struct kmem_cache
*vm_area_cachep
;
103 extern struct rb_root nommu_region_tree
;
104 extern struct rw_semaphore nommu_region_sem
;
106 extern unsigned int kobjsize(const void *objp
);
110 * vm_flags in vm_area_struct, see mm_types.h.
112 #define VM_NONE 0x00000000
114 #define VM_READ 0x00000001 /* currently active flags */
115 #define VM_WRITE 0x00000002
116 #define VM_EXEC 0x00000004
117 #define VM_SHARED 0x00000008
119 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
120 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
121 #define VM_MAYWRITE 0x00000020
122 #define VM_MAYEXEC 0x00000040
123 #define VM_MAYSHARE 0x00000080
125 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
126 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
127 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
129 #define VM_LOCKED 0x00002000
130 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
132 /* Used by sys_madvise() */
133 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
134 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
136 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
137 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
138 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
139 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
140 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
141 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
142 #define VM_ARCH_2 0x02000000
143 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
145 #ifdef CONFIG_MEM_SOFT_DIRTY
146 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
148 # define VM_SOFTDIRTY 0
151 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
152 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
153 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
154 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
156 #if defined(CONFIG_X86)
157 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
158 #elif defined(CONFIG_PPC)
159 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
160 #elif defined(CONFIG_PARISC)
161 # define VM_GROWSUP VM_ARCH_1
162 #elif defined(CONFIG_METAG)
163 # define VM_GROWSUP VM_ARCH_1
164 #elif defined(CONFIG_IA64)
165 # define VM_GROWSUP VM_ARCH_1
166 #elif !defined(CONFIG_MMU)
167 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
170 #if defined(CONFIG_X86)
171 /* MPX specific bounds table or bounds directory */
172 # define VM_MPX VM_ARCH_2
176 # define VM_GROWSUP VM_NONE
179 /* Bits set in the VMA until the stack is in its final location */
180 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
182 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
183 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
186 #ifdef CONFIG_STACK_GROWSUP
187 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
189 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
193 * Special vmas that are non-mergable, non-mlock()able.
194 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
196 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
198 /* This mask defines which mm->def_flags a process can inherit its parent */
199 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
202 * mapping from the currently active vm_flags protection bits (the
203 * low four bits) to a page protection mask..
205 extern pgprot_t protection_map
[16];
207 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
208 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
209 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
210 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
211 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
212 #define FAULT_FLAG_TRIED 0x20 /* Second try */
213 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
216 * vm_fault is filled by the the pagefault handler and passed to the vma's
217 * ->fault function. The vma's ->fault is responsible for returning a bitmask
218 * of VM_FAULT_xxx flags that give details about how the fault was handled.
220 * pgoff should be used in favour of virtual_address, if possible.
223 unsigned int flags
; /* FAULT_FLAG_xxx flags */
224 pgoff_t pgoff
; /* Logical page offset based on vma */
225 void __user
*virtual_address
; /* Faulting virtual address */
227 struct page
*page
; /* ->fault handlers should return a
228 * page here, unless VM_FAULT_NOPAGE
229 * is set (which is also implied by
232 /* for ->map_pages() only */
233 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
234 * max_pgoff inclusive */
235 pte_t
*pte
; /* pte entry associated with ->pgoff */
239 * These are the virtual MM functions - opening of an area, closing and
240 * unmapping it (needed to keep files on disk up-to-date etc), pointer
241 * to the functions called when a no-page or a wp-page exception occurs.
243 struct vm_operations_struct
{
244 void (*open
)(struct vm_area_struct
* area
);
245 void (*close
)(struct vm_area_struct
* area
);
246 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
247 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
249 /* notification that a previously read-only page is about to become
250 * writable, if an error is returned it will cause a SIGBUS */
251 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
253 /* called by access_process_vm when get_user_pages() fails, typically
254 * for use by special VMAs that can switch between memory and hardware
256 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
257 void *buf
, int len
, int write
);
259 /* Called by the /proc/PID/maps code to ask the vma whether it
260 * has a special name. Returning non-NULL will also cause this
261 * vma to be dumped unconditionally. */
262 const char *(*name
)(struct vm_area_struct
*vma
);
266 * set_policy() op must add a reference to any non-NULL @new mempolicy
267 * to hold the policy upon return. Caller should pass NULL @new to
268 * remove a policy and fall back to surrounding context--i.e. do not
269 * install a MPOL_DEFAULT policy, nor the task or system default
272 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
275 * get_policy() op must add reference [mpol_get()] to any policy at
276 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
277 * in mm/mempolicy.c will do this automatically.
278 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
279 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
280 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
281 * must return NULL--i.e., do not "fallback" to task or system default
284 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
288 * Called by vm_normal_page() for special PTEs to find the
289 * page for @addr. This is useful if the default behavior
290 * (using pte_page()) would not find the correct page.
292 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
299 #define page_private(page) ((page)->private)
300 #define set_page_private(page, v) ((page)->private = (v))
302 /* It's valid only if the page is free path or free_list */
303 static inline void set_freepage_migratetype(struct page
*page
, int migratetype
)
305 page
->index
= migratetype
;
308 /* It's valid only if the page is free path or free_list */
309 static inline int get_freepage_migratetype(struct page
*page
)
315 * FIXME: take this include out, include page-flags.h in
316 * files which need it (119 of them)
318 #include <linux/page-flags.h>
319 #include <linux/huge_mm.h>
322 * Methods to modify the page usage count.
324 * What counts for a page usage:
325 * - cache mapping (page->mapping)
326 * - private data (page->private)
327 * - page mapped in a task's page tables, each mapping
328 * is counted separately
330 * Also, many kernel routines increase the page count before a critical
331 * routine so they can be sure the page doesn't go away from under them.
335 * Drop a ref, return true if the refcount fell to zero (the page has no users)
337 static inline int put_page_testzero(struct page
*page
)
339 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) == 0, page
);
340 return atomic_dec_and_test(&page
->_count
);
344 * Try to grab a ref unless the page has a refcount of zero, return false if
346 * This can be called when MMU is off so it must not access
347 * any of the virtual mappings.
349 static inline int get_page_unless_zero(struct page
*page
)
351 return atomic_inc_not_zero(&page
->_count
);
355 * Try to drop a ref unless the page has a refcount of one, return false if
357 * This is to make sure that the refcount won't become zero after this drop.
358 * This can be called when MMU is off so it must not access
359 * any of the virtual mappings.
361 static inline int put_page_unless_one(struct page
*page
)
363 return atomic_add_unless(&page
->_count
, -1, 1);
366 extern int page_is_ram(unsigned long pfn
);
367 extern int region_is_ram(resource_size_t phys_addr
, unsigned long size
);
369 /* Support for virtually mapped pages */
370 struct page
*vmalloc_to_page(const void *addr
);
371 unsigned long vmalloc_to_pfn(const void *addr
);
374 * Determine if an address is within the vmalloc range
376 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
377 * is no special casing required.
379 static inline int is_vmalloc_addr(const void *x
)
382 unsigned long addr
= (unsigned long)x
;
384 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
390 extern int is_vmalloc_or_module_addr(const void *x
);
392 static inline int is_vmalloc_or_module_addr(const void *x
)
398 extern void kvfree(const void *addr
);
400 static inline void compound_lock(struct page
*page
)
402 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
403 VM_BUG_ON_PAGE(PageSlab(page
), page
);
404 bit_spin_lock(PG_compound_lock
, &page
->flags
);
408 static inline void compound_unlock(struct page
*page
)
410 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
411 VM_BUG_ON_PAGE(PageSlab(page
), page
);
412 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
416 static inline unsigned long compound_lock_irqsave(struct page
*page
)
418 unsigned long uninitialized_var(flags
);
419 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
420 local_irq_save(flags
);
426 static inline void compound_unlock_irqrestore(struct page
*page
,
429 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
430 compound_unlock(page
);
431 local_irq_restore(flags
);
435 static inline struct page
*compound_head_by_tail(struct page
*tail
)
437 struct page
*head
= tail
->first_page
;
440 * page->first_page may be a dangling pointer to an old
441 * compound page, so recheck that it is still a tail
442 * page before returning.
445 if (likely(PageTail(tail
)))
451 * Since either compound page could be dismantled asynchronously in THP
452 * or we access asynchronously arbitrary positioned struct page, there
453 * would be tail flag race. To handle this race, we should call
454 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
456 static inline struct page
*compound_head(struct page
*page
)
458 if (unlikely(PageTail(page
)))
459 return compound_head_by_tail(page
);
464 * If we access compound page synchronously such as access to
465 * allocated page, there is no need to handle tail flag race, so we can
466 * check tail flag directly without any synchronization primitive.
468 static inline struct page
*compound_head_fast(struct page
*page
)
470 if (unlikely(PageTail(page
)))
471 return page
->first_page
;
476 * The atomic page->_mapcount, starts from -1: so that transitions
477 * both from it and to it can be tracked, using atomic_inc_and_test
478 * and atomic_add_negative(-1).
480 static inline void page_mapcount_reset(struct page
*page
)
482 atomic_set(&(page
)->_mapcount
, -1);
485 static inline int page_mapcount(struct page
*page
)
487 VM_BUG_ON_PAGE(PageSlab(page
), page
);
488 return atomic_read(&page
->_mapcount
) + 1;
491 static inline int page_count(struct page
*page
)
493 return atomic_read(&compound_head(page
)->_count
);
496 #ifdef CONFIG_HUGETLB_PAGE
497 extern int PageHeadHuge(struct page
*page_head
);
498 #else /* CONFIG_HUGETLB_PAGE */
499 static inline int PageHeadHuge(struct page
*page_head
)
503 #endif /* CONFIG_HUGETLB_PAGE */
505 static inline bool __compound_tail_refcounted(struct page
*page
)
507 return !PageSlab(page
) && !PageHeadHuge(page
);
511 * This takes a head page as parameter and tells if the
512 * tail page reference counting can be skipped.
514 * For this to be safe, PageSlab and PageHeadHuge must remain true on
515 * any given page where they return true here, until all tail pins
516 * have been released.
518 static inline bool compound_tail_refcounted(struct page
*page
)
520 VM_BUG_ON_PAGE(!PageHead(page
), page
);
521 return __compound_tail_refcounted(page
);
524 static inline void get_huge_page_tail(struct page
*page
)
527 * __split_huge_page_refcount() cannot run from under us.
529 VM_BUG_ON_PAGE(!PageTail(page
), page
);
530 VM_BUG_ON_PAGE(page_mapcount(page
) < 0, page
);
531 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) != 0, page
);
532 if (compound_tail_refcounted(page
->first_page
))
533 atomic_inc(&page
->_mapcount
);
536 extern bool __get_page_tail(struct page
*page
);
538 static inline void get_page(struct page
*page
)
540 if (unlikely(PageTail(page
)))
541 if (likely(__get_page_tail(page
)))
544 * Getting a normal page or the head of a compound page
545 * requires to already have an elevated page->_count.
547 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
548 atomic_inc(&page
->_count
);
551 static inline struct page
*virt_to_head_page(const void *x
)
553 struct page
*page
= virt_to_page(x
);
556 * We don't need to worry about synchronization of tail flag
557 * when we call virt_to_head_page() since it is only called for
558 * already allocated page and this page won't be freed until
559 * this virt_to_head_page() is finished. So use _fast variant.
561 return compound_head_fast(page
);
565 * Setup the page count before being freed into the page allocator for
566 * the first time (boot or memory hotplug)
568 static inline void init_page_count(struct page
*page
)
570 atomic_set(&page
->_count
, 1);
574 * PageBuddy() indicate that the page is free and in the buddy system
575 * (see mm/page_alloc.c).
577 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
578 * -2 so that an underflow of the page_mapcount() won't be mistaken
579 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
580 * efficiently by most CPU architectures.
582 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
584 static inline int PageBuddy(struct page
*page
)
586 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
589 static inline void __SetPageBuddy(struct page
*page
)
591 VM_BUG_ON_PAGE(atomic_read(&page
->_mapcount
) != -1, page
);
592 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
595 static inline void __ClearPageBuddy(struct page
*page
)
597 VM_BUG_ON_PAGE(!PageBuddy(page
), page
);
598 atomic_set(&page
->_mapcount
, -1);
601 #define PAGE_BALLOON_MAPCOUNT_VALUE (-256)
603 static inline int PageBalloon(struct page
*page
)
605 return atomic_read(&page
->_mapcount
) == PAGE_BALLOON_MAPCOUNT_VALUE
;
608 static inline void __SetPageBalloon(struct page
*page
)
610 VM_BUG_ON_PAGE(atomic_read(&page
->_mapcount
) != -1, page
);
611 atomic_set(&page
->_mapcount
, PAGE_BALLOON_MAPCOUNT_VALUE
);
614 static inline void __ClearPageBalloon(struct page
*page
)
616 VM_BUG_ON_PAGE(!PageBalloon(page
), page
);
617 atomic_set(&page
->_mapcount
, -1);
620 void put_page(struct page
*page
);
621 void put_pages_list(struct list_head
*pages
);
623 void split_page(struct page
*page
, unsigned int order
);
624 int split_free_page(struct page
*page
);
627 * Compound pages have a destructor function. Provide a
628 * prototype for that function and accessor functions.
629 * These are _only_ valid on the head of a PG_compound page.
632 static inline void set_compound_page_dtor(struct page
*page
,
633 compound_page_dtor
*dtor
)
635 page
[1].compound_dtor
= dtor
;
638 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
640 return page
[1].compound_dtor
;
643 static inline int compound_order(struct page
*page
)
647 return page
[1].compound_order
;
650 static inline void set_compound_order(struct page
*page
, unsigned long order
)
652 page
[1].compound_order
= order
;
657 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
658 * servicing faults for write access. In the normal case, do always want
659 * pte_mkwrite. But get_user_pages can cause write faults for mappings
660 * that do not have writing enabled, when used by access_process_vm.
662 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
664 if (likely(vma
->vm_flags
& VM_WRITE
))
665 pte
= pte_mkwrite(pte
);
669 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
670 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
674 * Multiple processes may "see" the same page. E.g. for untouched
675 * mappings of /dev/null, all processes see the same page full of
676 * zeroes, and text pages of executables and shared libraries have
677 * only one copy in memory, at most, normally.
679 * For the non-reserved pages, page_count(page) denotes a reference count.
680 * page_count() == 0 means the page is free. page->lru is then used for
681 * freelist management in the buddy allocator.
682 * page_count() > 0 means the page has been allocated.
684 * Pages are allocated by the slab allocator in order to provide memory
685 * to kmalloc and kmem_cache_alloc. In this case, the management of the
686 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
687 * unless a particular usage is carefully commented. (the responsibility of
688 * freeing the kmalloc memory is the caller's, of course).
690 * A page may be used by anyone else who does a __get_free_page().
691 * In this case, page_count still tracks the references, and should only
692 * be used through the normal accessor functions. The top bits of page->flags
693 * and page->virtual store page management information, but all other fields
694 * are unused and could be used privately, carefully. The management of this
695 * page is the responsibility of the one who allocated it, and those who have
696 * subsequently been given references to it.
698 * The other pages (we may call them "pagecache pages") are completely
699 * managed by the Linux memory manager: I/O, buffers, swapping etc.
700 * The following discussion applies only to them.
702 * A pagecache page contains an opaque `private' member, which belongs to the
703 * page's address_space. Usually, this is the address of a circular list of
704 * the page's disk buffers. PG_private must be set to tell the VM to call
705 * into the filesystem to release these pages.
707 * A page may belong to an inode's memory mapping. In this case, page->mapping
708 * is the pointer to the inode, and page->index is the file offset of the page,
709 * in units of PAGE_CACHE_SIZE.
711 * If pagecache pages are not associated with an inode, they are said to be
712 * anonymous pages. These may become associated with the swapcache, and in that
713 * case PG_swapcache is set, and page->private is an offset into the swapcache.
715 * In either case (swapcache or inode backed), the pagecache itself holds one
716 * reference to the page. Setting PG_private should also increment the
717 * refcount. The each user mapping also has a reference to the page.
719 * The pagecache pages are stored in a per-mapping radix tree, which is
720 * rooted at mapping->page_tree, and indexed by offset.
721 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
722 * lists, we instead now tag pages as dirty/writeback in the radix tree.
724 * All pagecache pages may be subject to I/O:
725 * - inode pages may need to be read from disk,
726 * - inode pages which have been modified and are MAP_SHARED may need
727 * to be written back to the inode on disk,
728 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
729 * modified may need to be swapped out to swap space and (later) to be read
734 * The zone field is never updated after free_area_init_core()
735 * sets it, so none of the operations on it need to be atomic.
738 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
739 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
740 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
741 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
742 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
745 * Define the bit shifts to access each section. For non-existent
746 * sections we define the shift as 0; that plus a 0 mask ensures
747 * the compiler will optimise away reference to them.
749 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
750 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
751 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
752 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
754 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
755 #ifdef NODE_NOT_IN_PAGE_FLAGS
756 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
757 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
758 SECTIONS_PGOFF : ZONES_PGOFF)
760 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
761 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
762 NODES_PGOFF : ZONES_PGOFF)
765 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
767 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
768 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
771 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
772 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
773 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
774 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
775 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
777 static inline enum zone_type
page_zonenum(const struct page
*page
)
779 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
782 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
783 #define SECTION_IN_PAGE_FLAGS
787 * The identification function is mainly used by the buddy allocator for
788 * determining if two pages could be buddies. We are not really identifying
789 * the zone since we could be using the section number id if we do not have
790 * node id available in page flags.
791 * We only guarantee that it will return the same value for two combinable
794 static inline int page_zone_id(struct page
*page
)
796 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
799 static inline int zone_to_nid(struct zone
*zone
)
808 #ifdef NODE_NOT_IN_PAGE_FLAGS
809 extern int page_to_nid(const struct page
*page
);
811 static inline int page_to_nid(const struct page
*page
)
813 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
817 #ifdef CONFIG_NUMA_BALANCING
818 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
820 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
823 static inline int cpupid_to_pid(int cpupid
)
825 return cpupid
& LAST__PID_MASK
;
828 static inline int cpupid_to_cpu(int cpupid
)
830 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
833 static inline int cpupid_to_nid(int cpupid
)
835 return cpu_to_node(cpupid_to_cpu(cpupid
));
838 static inline bool cpupid_pid_unset(int cpupid
)
840 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
843 static inline bool cpupid_cpu_unset(int cpupid
)
845 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
848 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
850 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
853 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
854 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
855 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
857 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
860 static inline int page_cpupid_last(struct page
*page
)
862 return page
->_last_cpupid
;
864 static inline void page_cpupid_reset_last(struct page
*page
)
866 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
869 static inline int page_cpupid_last(struct page
*page
)
871 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
874 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
876 static inline void page_cpupid_reset_last(struct page
*page
)
878 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
880 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
881 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
883 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
884 #else /* !CONFIG_NUMA_BALANCING */
885 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
887 return page_to_nid(page
); /* XXX */
890 static inline int page_cpupid_last(struct page
*page
)
892 return page_to_nid(page
); /* XXX */
895 static inline int cpupid_to_nid(int cpupid
)
900 static inline int cpupid_to_pid(int cpupid
)
905 static inline int cpupid_to_cpu(int cpupid
)
910 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
915 static inline bool cpupid_pid_unset(int cpupid
)
920 static inline void page_cpupid_reset_last(struct page
*page
)
924 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
928 #endif /* CONFIG_NUMA_BALANCING */
930 static inline struct zone
*page_zone(const struct page
*page
)
932 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
935 #ifdef SECTION_IN_PAGE_FLAGS
936 static inline void set_page_section(struct page
*page
, unsigned long section
)
938 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
939 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
942 static inline unsigned long page_to_section(const struct page
*page
)
944 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
948 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
950 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
951 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
954 static inline void set_page_node(struct page
*page
, unsigned long node
)
956 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
957 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
960 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
961 unsigned long node
, unsigned long pfn
)
963 set_page_zone(page
, zone
);
964 set_page_node(page
, node
);
965 #ifdef SECTION_IN_PAGE_FLAGS
966 set_page_section(page
, pfn_to_section_nr(pfn
));
971 * Some inline functions in vmstat.h depend on page_zone()
973 #include <linux/vmstat.h>
975 static __always_inline
void *lowmem_page_address(const struct page
*page
)
977 return __va(PFN_PHYS(page_to_pfn(page
)));
980 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
981 #define HASHED_PAGE_VIRTUAL
984 #if defined(WANT_PAGE_VIRTUAL)
985 static inline void *page_address(const struct page
*page
)
987 return page
->virtual;
989 static inline void set_page_address(struct page
*page
, void *address
)
991 page
->virtual = address
;
993 #define page_address_init() do { } while(0)
996 #if defined(HASHED_PAGE_VIRTUAL)
997 void *page_address(const struct page
*page
);
998 void set_page_address(struct page
*page
, void *virtual);
999 void page_address_init(void);
1002 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1003 #define page_address(page) lowmem_page_address(page)
1004 #define set_page_address(page, address) do { } while(0)
1005 #define page_address_init() do { } while(0)
1009 * On an anonymous page mapped into a user virtual memory area,
1010 * page->mapping points to its anon_vma, not to a struct address_space;
1011 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
1013 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
1014 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
1015 * and then page->mapping points, not to an anon_vma, but to a private
1016 * structure which KSM associates with that merged page. See ksm.h.
1018 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
1020 * Please note that, confusingly, "page_mapping" refers to the inode
1021 * address_space which maps the page from disk; whereas "page_mapped"
1022 * refers to user virtual address space into which the page is mapped.
1024 #define PAGE_MAPPING_ANON 1
1025 #define PAGE_MAPPING_KSM 2
1026 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
1028 extern struct address_space
*page_mapping(struct page
*page
);
1030 /* Neutral page->mapping pointer to address_space or anon_vma or other */
1031 static inline void *page_rmapping(struct page
*page
)
1033 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
1036 extern struct address_space
*__page_file_mapping(struct page
*);
1039 struct address_space
*page_file_mapping(struct page
*page
)
1041 if (unlikely(PageSwapCache(page
)))
1042 return __page_file_mapping(page
);
1044 return page
->mapping
;
1047 static inline int PageAnon(struct page
*page
)
1049 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
1053 * Return the pagecache index of the passed page. Regular pagecache pages
1054 * use ->index whereas swapcache pages use ->private
1056 static inline pgoff_t
page_index(struct page
*page
)
1058 if (unlikely(PageSwapCache(page
)))
1059 return page_private(page
);
1063 extern pgoff_t
__page_file_index(struct page
*page
);
1066 * Return the file index of the page. Regular pagecache pages use ->index
1067 * whereas swapcache pages use swp_offset(->private)
1069 static inline pgoff_t
page_file_index(struct page
*page
)
1071 if (unlikely(PageSwapCache(page
)))
1072 return __page_file_index(page
);
1078 * Return true if this page is mapped into pagetables.
1080 static inline int page_mapped(struct page
*page
)
1082 return atomic_read(&(page
)->_mapcount
) >= 0;
1086 * Different kinds of faults, as returned by handle_mm_fault().
1087 * Used to decide whether a process gets delivered SIGBUS or
1088 * just gets major/minor fault counters bumped up.
1091 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1093 #define VM_FAULT_OOM 0x0001
1094 #define VM_FAULT_SIGBUS 0x0002
1095 #define VM_FAULT_MAJOR 0x0004
1096 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1097 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1098 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1099 #define VM_FAULT_SIGSEGV 0x0040
1101 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1102 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1103 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1104 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1106 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1108 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1109 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1112 /* Encode hstate index for a hwpoisoned large page */
1113 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1114 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1117 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1119 extern void pagefault_out_of_memory(void);
1121 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1124 * Flags passed to show_mem() and show_free_areas() to suppress output in
1127 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1129 extern void show_free_areas(unsigned int flags
);
1130 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1132 int shmem_zero_setup(struct vm_area_struct
*);
1134 bool shmem_mapping(struct address_space
*mapping
);
1136 static inline bool shmem_mapping(struct address_space
*mapping
)
1142 extern int can_do_mlock(void);
1143 extern int user_shm_lock(size_t, struct user_struct
*);
1144 extern void user_shm_unlock(size_t, struct user_struct
*);
1147 * Parameter block passed down to zap_pte_range in exceptional cases.
1149 struct zap_details
{
1150 struct address_space
*check_mapping
; /* Check page->mapping if set */
1151 pgoff_t first_index
; /* Lowest page->index to unmap */
1152 pgoff_t last_index
; /* Highest page->index to unmap */
1155 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1158 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1159 unsigned long size
);
1160 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1161 unsigned long size
, struct zap_details
*);
1162 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1163 unsigned long start
, unsigned long end
);
1166 * mm_walk - callbacks for walk_page_range
1167 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1168 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1169 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1170 * this handler is required to be able to handle
1171 * pmd_trans_huge() pmds. They may simply choose to
1172 * split_huge_page() instead of handling it explicitly.
1173 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1174 * @pte_hole: if set, called for each hole at all levels
1175 * @hugetlb_entry: if set, called for each hugetlb entry
1176 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1179 * (see walk_page_range for more details)
1182 int (*pgd_entry
)(pgd_t
*pgd
, unsigned long addr
,
1183 unsigned long next
, struct mm_walk
*walk
);
1184 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1185 unsigned long next
, struct mm_walk
*walk
);
1186 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1187 unsigned long next
, struct mm_walk
*walk
);
1188 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1189 unsigned long next
, struct mm_walk
*walk
);
1190 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1191 struct mm_walk
*walk
);
1192 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1193 unsigned long addr
, unsigned long next
,
1194 struct mm_walk
*walk
);
1195 struct mm_struct
*mm
;
1199 int walk_page_range(unsigned long addr
, unsigned long end
,
1200 struct mm_walk
*walk
);
1201 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1202 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1203 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1204 struct vm_area_struct
*vma
);
1205 void unmap_mapping_range(struct address_space
*mapping
,
1206 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1207 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1208 unsigned long *pfn
);
1209 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1210 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1211 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1212 void *buf
, int len
, int write
);
1214 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1215 loff_t
const holebegin
, loff_t
const holelen
)
1217 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1220 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1221 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1222 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1223 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1224 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1225 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1226 int invalidate_inode_page(struct page
*page
);
1229 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1230 unsigned long address
, unsigned int flags
);
1231 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1232 unsigned long address
, unsigned int fault_flags
);
1234 static inline int handle_mm_fault(struct mm_struct
*mm
,
1235 struct vm_area_struct
*vma
, unsigned long address
,
1238 /* should never happen if there's no MMU */
1240 return VM_FAULT_SIGBUS
;
1242 static inline int fixup_user_fault(struct task_struct
*tsk
,
1243 struct mm_struct
*mm
, unsigned long address
,
1244 unsigned int fault_flags
)
1246 /* should never happen if there's no MMU */
1252 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1253 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1254 void *buf
, int len
, int write
);
1256 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1257 unsigned long start
, unsigned long nr_pages
,
1258 unsigned int foll_flags
, struct page
**pages
,
1259 struct vm_area_struct
**vmas
, int *nonblocking
);
1260 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1261 unsigned long start
, unsigned long nr_pages
,
1262 int write
, int force
, struct page
**pages
,
1263 struct vm_area_struct
**vmas
);
1264 long get_user_pages_locked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1265 unsigned long start
, unsigned long nr_pages
,
1266 int write
, int force
, struct page
**pages
,
1268 long __get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1269 unsigned long start
, unsigned long nr_pages
,
1270 int write
, int force
, struct page
**pages
,
1271 unsigned int gup_flags
);
1272 long get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1273 unsigned long start
, unsigned long nr_pages
,
1274 int write
, int force
, struct page
**pages
);
1275 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1276 struct page
**pages
);
1278 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1279 struct page
**pages
);
1280 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1281 struct page
*get_dump_page(unsigned long addr
);
1283 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1284 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1285 unsigned int length
);
1287 int __set_page_dirty_nobuffers(struct page
*page
);
1288 int __set_page_dirty_no_writeback(struct page
*page
);
1289 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1291 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1292 int set_page_dirty(struct page
*page
);
1293 int set_page_dirty_lock(struct page
*page
);
1294 int clear_page_dirty_for_io(struct page
*page
);
1295 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1297 /* Is the vma a continuation of the stack vma above it? */
1298 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1300 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1303 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1306 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1307 (vma
->vm_start
== addr
) &&
1308 !vma_growsdown(vma
->vm_prev
, addr
);
1311 /* Is the vma a continuation of the stack vma below it? */
1312 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1314 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1317 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1320 return (vma
->vm_flags
& VM_GROWSUP
) &&
1321 (vma
->vm_end
== addr
) &&
1322 !vma_growsup(vma
->vm_next
, addr
);
1325 extern struct task_struct
*task_of_stack(struct task_struct
*task
,
1326 struct vm_area_struct
*vma
, bool in_group
);
1328 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1329 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1330 unsigned long new_addr
, unsigned long len
,
1331 bool need_rmap_locks
);
1332 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1333 unsigned long end
, pgprot_t newprot
,
1334 int dirty_accountable
, int prot_numa
);
1335 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1336 struct vm_area_struct
**pprev
, unsigned long start
,
1337 unsigned long end
, unsigned long newflags
);
1340 * doesn't attempt to fault and will return short.
1342 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1343 struct page
**pages
);
1345 * per-process(per-mm_struct) statistics.
1347 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1349 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1351 #ifdef SPLIT_RSS_COUNTING
1353 * counter is updated in asynchronous manner and may go to minus.
1354 * But it's never be expected number for users.
1359 return (unsigned long)val
;
1362 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1364 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1367 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1369 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1372 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1374 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1377 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1379 return get_mm_counter(mm
, MM_FILEPAGES
) +
1380 get_mm_counter(mm
, MM_ANONPAGES
);
1383 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1385 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1388 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1390 return max(mm
->hiwater_vm
, mm
->total_vm
);
1393 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1395 unsigned long _rss
= get_mm_rss(mm
);
1397 if ((mm
)->hiwater_rss
< _rss
)
1398 (mm
)->hiwater_rss
= _rss
;
1401 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1403 if (mm
->hiwater_vm
< mm
->total_vm
)
1404 mm
->hiwater_vm
= mm
->total_vm
;
1407 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1408 struct mm_struct
*mm
)
1410 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1412 if (*maxrss
< hiwater_rss
)
1413 *maxrss
= hiwater_rss
;
1416 #if defined(SPLIT_RSS_COUNTING)
1417 void sync_mm_rss(struct mm_struct
*mm
);
1419 static inline void sync_mm_rss(struct mm_struct
*mm
)
1424 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1426 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1428 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1432 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1436 #ifdef __PAGETABLE_PUD_FOLDED
1437 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1438 unsigned long address
)
1443 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1446 #ifdef __PAGETABLE_PMD_FOLDED
1447 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1448 unsigned long address
)
1453 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1458 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1459 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1462 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1464 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1466 return atomic_long_read(&mm
->nr_pmds
);
1469 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1471 atomic_long_inc(&mm
->nr_pmds
);
1474 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1476 atomic_long_dec(&mm
->nr_pmds
);
1480 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1481 pmd_t
*pmd
, unsigned long address
);
1482 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1485 * The following ifdef needed to get the 4level-fixup.h header to work.
1486 * Remove it when 4level-fixup.h has been removed.
1488 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1489 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1491 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1492 NULL
: pud_offset(pgd
, address
);
1495 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1497 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1498 NULL
: pmd_offset(pud
, address
);
1500 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1502 #if USE_SPLIT_PTE_PTLOCKS
1503 #if ALLOC_SPLIT_PTLOCKS
1504 void __init
ptlock_cache_init(void);
1505 extern bool ptlock_alloc(struct page
*page
);
1506 extern void ptlock_free(struct page
*page
);
1508 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1512 #else /* ALLOC_SPLIT_PTLOCKS */
1513 static inline void ptlock_cache_init(void)
1517 static inline bool ptlock_alloc(struct page
*page
)
1522 static inline void ptlock_free(struct page
*page
)
1526 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1530 #endif /* ALLOC_SPLIT_PTLOCKS */
1532 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1534 return ptlock_ptr(pmd_page(*pmd
));
1537 static inline bool ptlock_init(struct page
*page
)
1540 * prep_new_page() initialize page->private (and therefore page->ptl)
1541 * with 0. Make sure nobody took it in use in between.
1543 * It can happen if arch try to use slab for page table allocation:
1544 * slab code uses page->slab_cache and page->first_page (for tail
1545 * pages), which share storage with page->ptl.
1547 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1548 if (!ptlock_alloc(page
))
1550 spin_lock_init(ptlock_ptr(page
));
1554 /* Reset page->mapping so free_pages_check won't complain. */
1555 static inline void pte_lock_deinit(struct page
*page
)
1557 page
->mapping
= NULL
;
1561 #else /* !USE_SPLIT_PTE_PTLOCKS */
1563 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1565 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1567 return &mm
->page_table_lock
;
1569 static inline void ptlock_cache_init(void) {}
1570 static inline bool ptlock_init(struct page
*page
) { return true; }
1571 static inline void pte_lock_deinit(struct page
*page
) {}
1572 #endif /* USE_SPLIT_PTE_PTLOCKS */
1574 static inline void pgtable_init(void)
1576 ptlock_cache_init();
1577 pgtable_cache_init();
1580 static inline bool pgtable_page_ctor(struct page
*page
)
1582 inc_zone_page_state(page
, NR_PAGETABLE
);
1583 return ptlock_init(page
);
1586 static inline void pgtable_page_dtor(struct page
*page
)
1588 pte_lock_deinit(page
);
1589 dec_zone_page_state(page
, NR_PAGETABLE
);
1592 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1594 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1595 pte_t *__pte = pte_offset_map(pmd, address); \
1601 #define pte_unmap_unlock(pte, ptl) do { \
1606 #define pte_alloc_map(mm, vma, pmd, address) \
1607 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1609 NULL: pte_offset_map(pmd, address))
1611 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1612 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1614 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1616 #define pte_alloc_kernel(pmd, address) \
1617 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1618 NULL: pte_offset_kernel(pmd, address))
1620 #if USE_SPLIT_PMD_PTLOCKS
1622 static struct page
*pmd_to_page(pmd_t
*pmd
)
1624 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1625 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1628 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1630 return ptlock_ptr(pmd_to_page(pmd
));
1633 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1635 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1636 page
->pmd_huge_pte
= NULL
;
1638 return ptlock_init(page
);
1641 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1643 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1644 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1649 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1653 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1655 return &mm
->page_table_lock
;
1658 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1659 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1661 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1665 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1667 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1672 extern void free_area_init(unsigned long * zones_size
);
1673 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1674 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1675 extern void free_initmem(void);
1678 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1679 * into the buddy system. The freed pages will be poisoned with pattern
1680 * "poison" if it's within range [0, UCHAR_MAX].
1681 * Return pages freed into the buddy system.
1683 extern unsigned long free_reserved_area(void *start
, void *end
,
1684 int poison
, char *s
);
1686 #ifdef CONFIG_HIGHMEM
1688 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1689 * and totalram_pages.
1691 extern void free_highmem_page(struct page
*page
);
1694 extern void adjust_managed_page_count(struct page
*page
, long count
);
1695 extern void mem_init_print_info(const char *str
);
1697 /* Free the reserved page into the buddy system, so it gets managed. */
1698 static inline void __free_reserved_page(struct page
*page
)
1700 ClearPageReserved(page
);
1701 init_page_count(page
);
1705 static inline void free_reserved_page(struct page
*page
)
1707 __free_reserved_page(page
);
1708 adjust_managed_page_count(page
, 1);
1711 static inline void mark_page_reserved(struct page
*page
)
1713 SetPageReserved(page
);
1714 adjust_managed_page_count(page
, -1);
1718 * Default method to free all the __init memory into the buddy system.
1719 * The freed pages will be poisoned with pattern "poison" if it's within
1720 * range [0, UCHAR_MAX].
1721 * Return pages freed into the buddy system.
1723 static inline unsigned long free_initmem_default(int poison
)
1725 extern char __init_begin
[], __init_end
[];
1727 return free_reserved_area(&__init_begin
, &__init_end
,
1728 poison
, "unused kernel");
1731 static inline unsigned long get_num_physpages(void)
1734 unsigned long phys_pages
= 0;
1736 for_each_online_node(nid
)
1737 phys_pages
+= node_present_pages(nid
);
1742 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1744 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1745 * zones, allocate the backing mem_map and account for memory holes in a more
1746 * architecture independent manner. This is a substitute for creating the
1747 * zone_sizes[] and zholes_size[] arrays and passing them to
1748 * free_area_init_node()
1750 * An architecture is expected to register range of page frames backed by
1751 * physical memory with memblock_add[_node]() before calling
1752 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1753 * usage, an architecture is expected to do something like
1755 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1757 * for_each_valid_physical_page_range()
1758 * memblock_add_node(base, size, nid)
1759 * free_area_init_nodes(max_zone_pfns);
1761 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1762 * registered physical page range. Similarly
1763 * sparse_memory_present_with_active_regions() calls memory_present() for
1764 * each range when SPARSEMEM is enabled.
1766 * See mm/page_alloc.c for more information on each function exposed by
1767 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1769 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1770 unsigned long node_map_pfn_alignment(void);
1771 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1772 unsigned long end_pfn
);
1773 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1774 unsigned long end_pfn
);
1775 extern void get_pfn_range_for_nid(unsigned int nid
,
1776 unsigned long *start_pfn
, unsigned long *end_pfn
);
1777 extern unsigned long find_min_pfn_with_active_regions(void);
1778 extern void free_bootmem_with_active_regions(int nid
,
1779 unsigned long max_low_pfn
);
1780 extern void sparse_memory_present_with_active_regions(int nid
);
1782 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1784 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1785 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1786 static inline int __early_pfn_to_nid(unsigned long pfn
)
1791 /* please see mm/page_alloc.c */
1792 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1793 /* there is a per-arch backend function. */
1794 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1797 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1798 extern void memmap_init_zone(unsigned long, int, unsigned long,
1799 unsigned long, enum memmap_context
);
1800 extern void setup_per_zone_wmarks(void);
1801 extern int __meminit
init_per_zone_wmark_min(void);
1802 extern void mem_init(void);
1803 extern void __init
mmap_init(void);
1804 extern void show_mem(unsigned int flags
);
1805 extern void si_meminfo(struct sysinfo
* val
);
1806 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1808 extern __printf(3, 4)
1809 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1811 extern void setup_per_cpu_pageset(void);
1813 extern void zone_pcp_update(struct zone
*zone
);
1814 extern void zone_pcp_reset(struct zone
*zone
);
1817 extern int min_free_kbytes
;
1820 extern atomic_long_t mmap_pages_allocated
;
1821 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1823 /* interval_tree.c */
1824 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1825 struct rb_root
*root
);
1826 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1827 struct vm_area_struct
*prev
,
1828 struct rb_root
*root
);
1829 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1830 struct rb_root
*root
);
1831 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1832 unsigned long start
, unsigned long last
);
1833 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1834 unsigned long start
, unsigned long last
);
1836 #define vma_interval_tree_foreach(vma, root, start, last) \
1837 for (vma = vma_interval_tree_iter_first(root, start, last); \
1838 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1840 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1841 struct rb_root
*root
);
1842 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1843 struct rb_root
*root
);
1844 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1845 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1846 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1847 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1848 #ifdef CONFIG_DEBUG_VM_RB
1849 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1852 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1853 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1854 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1857 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1858 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1859 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1860 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1861 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1862 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1863 struct mempolicy
*);
1864 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1865 extern int split_vma(struct mm_struct
*,
1866 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1867 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1868 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1869 struct rb_node
**, struct rb_node
*);
1870 extern void unlink_file_vma(struct vm_area_struct
*);
1871 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1872 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1873 bool *need_rmap_locks
);
1874 extern void exit_mmap(struct mm_struct
*);
1876 static inline int check_data_rlimit(unsigned long rlim
,
1878 unsigned long start
,
1879 unsigned long end_data
,
1880 unsigned long start_data
)
1882 if (rlim
< RLIM_INFINITY
) {
1883 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1890 extern int mm_take_all_locks(struct mm_struct
*mm
);
1891 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1893 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1894 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1896 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1897 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
1898 unsigned long addr
, unsigned long len
,
1899 unsigned long flags
,
1900 const struct vm_special_mapping
*spec
);
1901 /* This is an obsolete alternative to _install_special_mapping. */
1902 extern int install_special_mapping(struct mm_struct
*mm
,
1903 unsigned long addr
, unsigned long len
,
1904 unsigned long flags
, struct page
**pages
);
1906 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1908 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1909 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1910 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1911 unsigned long len
, unsigned long prot
, unsigned long flags
,
1912 unsigned long pgoff
, unsigned long *populate
);
1913 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1916 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1918 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1921 (void) __mm_populate(addr
, len
, 1);
1924 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1927 /* These take the mm semaphore themselves */
1928 extern unsigned long vm_brk(unsigned long, unsigned long);
1929 extern int vm_munmap(unsigned long, size_t);
1930 extern unsigned long vm_mmap(struct file
*, unsigned long,
1931 unsigned long, unsigned long,
1932 unsigned long, unsigned long);
1934 struct vm_unmapped_area_info
{
1935 #define VM_UNMAPPED_AREA_TOPDOWN 1
1936 unsigned long flags
;
1937 unsigned long length
;
1938 unsigned long low_limit
;
1939 unsigned long high_limit
;
1940 unsigned long align_mask
;
1941 unsigned long align_offset
;
1944 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1945 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1948 * Search for an unmapped address range.
1950 * We are looking for a range that:
1951 * - does not intersect with any VMA;
1952 * - is contained within the [low_limit, high_limit) interval;
1953 * - is at least the desired size.
1954 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1956 static inline unsigned long
1957 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1959 if (!(info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
))
1960 return unmapped_area(info
);
1962 return unmapped_area_topdown(info
);
1966 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1967 extern void truncate_inode_pages_range(struct address_space
*,
1968 loff_t lstart
, loff_t lend
);
1969 extern void truncate_inode_pages_final(struct address_space
*);
1971 /* generic vm_area_ops exported for stackable file systems */
1972 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1973 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1974 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1976 /* mm/page-writeback.c */
1977 int write_one_page(struct page
*page
, int wait
);
1978 void task_dirty_inc(struct task_struct
*tsk
);
1981 #define VM_MAX_READAHEAD 128 /* kbytes */
1982 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1984 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1985 pgoff_t offset
, unsigned long nr_to_read
);
1987 void page_cache_sync_readahead(struct address_space
*mapping
,
1988 struct file_ra_state
*ra
,
1991 unsigned long size
);
1993 void page_cache_async_readahead(struct address_space
*mapping
,
1994 struct file_ra_state
*ra
,
1998 unsigned long size
);
2000 unsigned long max_sane_readahead(unsigned long nr
);
2002 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2003 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2005 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2006 extern int expand_downwards(struct vm_area_struct
*vma
,
2007 unsigned long address
);
2009 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2011 #define expand_upwards(vma, address) (0)
2014 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2015 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2016 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2017 struct vm_area_struct
**pprev
);
2019 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2020 NULL if none. Assume start_addr < end_addr. */
2021 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2023 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2025 if (vma
&& end_addr
<= vma
->vm_start
)
2030 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2032 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2035 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2036 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2037 unsigned long vm_start
, unsigned long vm_end
)
2039 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2041 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2048 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2049 void vma_set_page_prot(struct vm_area_struct
*vma
);
2051 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2055 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2057 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2061 #ifdef CONFIG_NUMA_BALANCING
2062 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2063 unsigned long start
, unsigned long end
);
2066 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2067 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2068 unsigned long pfn
, unsigned long size
, pgprot_t
);
2069 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2070 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2072 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2074 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2077 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2078 unsigned long address
, unsigned int foll_flags
,
2079 unsigned int *page_mask
);
2081 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2082 unsigned long address
, unsigned int foll_flags
)
2084 unsigned int unused_page_mask
;
2085 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2088 #define FOLL_WRITE 0x01 /* check pte is writable */
2089 #define FOLL_TOUCH 0x02 /* mark page accessed */
2090 #define FOLL_GET 0x04 /* do get_page on page */
2091 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2092 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2093 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2094 * and return without waiting upon it */
2095 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
2096 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2097 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2098 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2099 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2100 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2102 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2104 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2105 unsigned long size
, pte_fn_t fn
, void *data
);
2107 #ifdef CONFIG_PROC_FS
2108 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
2110 static inline void vm_stat_account(struct mm_struct
*mm
,
2111 unsigned long flags
, struct file
*file
, long pages
)
2113 mm
->total_vm
+= pages
;
2115 #endif /* CONFIG_PROC_FS */
2117 #ifdef CONFIG_DEBUG_PAGEALLOC
2118 extern bool _debug_pagealloc_enabled
;
2119 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2121 static inline bool debug_pagealloc_enabled(void)
2123 return _debug_pagealloc_enabled
;
2127 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2129 if (!debug_pagealloc_enabled())
2132 __kernel_map_pages(page
, numpages
, enable
);
2134 #ifdef CONFIG_HIBERNATION
2135 extern bool kernel_page_present(struct page
*page
);
2136 #endif /* CONFIG_HIBERNATION */
2139 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2140 #ifdef CONFIG_HIBERNATION
2141 static inline bool kernel_page_present(struct page
*page
) { return true; }
2142 #endif /* CONFIG_HIBERNATION */
2145 #ifdef __HAVE_ARCH_GATE_AREA
2146 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2147 extern int in_gate_area_no_mm(unsigned long addr
);
2148 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2150 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2154 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2155 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2159 #endif /* __HAVE_ARCH_GATE_AREA */
2161 #ifdef CONFIG_SYSCTL
2162 extern int sysctl_drop_caches
;
2163 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2164 void __user
*, size_t *, loff_t
*);
2167 unsigned long shrink_node_slabs(gfp_t gfp_mask
, int nid
,
2168 unsigned long nr_scanned
,
2169 unsigned long nr_eligible
);
2172 #define randomize_va_space 0
2174 extern int randomize_va_space
;
2177 const char * arch_vma_name(struct vm_area_struct
*vma
);
2178 void print_vma_addr(char *prefix
, unsigned long rip
);
2180 void sparse_mem_maps_populate_node(struct page
**map_map
,
2181 unsigned long pnum_begin
,
2182 unsigned long pnum_end
,
2183 unsigned long map_count
,
2186 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2187 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2188 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2189 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2190 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2191 void *vmemmap_alloc_block(unsigned long size
, int node
);
2192 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2193 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2194 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2196 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2197 void vmemmap_populate_print_last(void);
2198 #ifdef CONFIG_MEMORY_HOTPLUG
2199 void vmemmap_free(unsigned long start
, unsigned long end
);
2201 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2202 unsigned long size
);
2205 MF_COUNT_INCREASED
= 1 << 0,
2206 MF_ACTION_REQUIRED
= 1 << 1,
2207 MF_MUST_KILL
= 1 << 2,
2208 MF_SOFT_OFFLINE
= 1 << 3,
2210 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2211 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2212 extern int unpoison_memory(unsigned long pfn
);
2213 extern int sysctl_memory_failure_early_kill
;
2214 extern int sysctl_memory_failure_recovery
;
2215 extern void shake_page(struct page
*p
, int access
);
2216 extern atomic_long_t num_poisoned_pages
;
2217 extern int soft_offline_page(struct page
*page
, int flags
);
2219 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2220 extern void clear_huge_page(struct page
*page
,
2222 unsigned int pages_per_huge_page
);
2223 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2224 unsigned long addr
, struct vm_area_struct
*vma
,
2225 unsigned int pages_per_huge_page
);
2226 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2228 extern struct page_ext_operations debug_guardpage_ops
;
2229 extern struct page_ext_operations page_poisoning_ops
;
2231 #ifdef CONFIG_DEBUG_PAGEALLOC
2232 extern unsigned int _debug_guardpage_minorder
;
2233 extern bool _debug_guardpage_enabled
;
2235 static inline unsigned int debug_guardpage_minorder(void)
2237 return _debug_guardpage_minorder
;
2240 static inline bool debug_guardpage_enabled(void)
2242 return _debug_guardpage_enabled
;
2245 static inline bool page_is_guard(struct page
*page
)
2247 struct page_ext
*page_ext
;
2249 if (!debug_guardpage_enabled())
2252 page_ext
= lookup_page_ext(page
);
2253 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2256 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2257 static inline bool debug_guardpage_enabled(void) { return false; }
2258 static inline bool page_is_guard(struct page
*page
) { return false; }
2259 #endif /* CONFIG_DEBUG_PAGEALLOC */
2261 #if MAX_NUMNODES > 1
2262 void __init
setup_nr_node_ids(void);
2264 static inline void setup_nr_node_ids(void) {}
2267 #endif /* __KERNEL__ */
2268 #endif /* _LINUX_MM_H */