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_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
142 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
143 #define VM_ARCH_2 0x02000000
144 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
146 #ifdef CONFIG_MEM_SOFT_DIRTY
147 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
149 # define VM_SOFTDIRTY 0
152 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
153 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
154 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
155 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
157 #if defined(CONFIG_X86)
158 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
159 #elif defined(CONFIG_PPC)
160 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
161 #elif defined(CONFIG_PARISC)
162 # define VM_GROWSUP VM_ARCH_1
163 #elif defined(CONFIG_METAG)
164 # define VM_GROWSUP VM_ARCH_1
165 #elif defined(CONFIG_IA64)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif !defined(CONFIG_MMU)
168 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
171 #if defined(CONFIG_X86)
172 /* MPX specific bounds table or bounds directory */
173 # define VM_MPX VM_ARCH_2
177 # define VM_GROWSUP VM_NONE
180 /* Bits set in the VMA until the stack is in its final location */
181 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
183 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
184 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
187 #ifdef CONFIG_STACK_GROWSUP
188 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
190 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
194 * Special vmas that are non-mergable, non-mlock()able.
195 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
197 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
199 /* This mask defines which mm->def_flags a process can inherit its parent */
200 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
203 * mapping from the currently active vm_flags protection bits (the
204 * low four bits) to a page protection mask..
206 extern pgprot_t protection_map
[16];
208 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
209 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
210 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
211 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
212 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
213 #define FAULT_FLAG_TRIED 0x20 /* Second try */
214 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
217 * vm_fault is filled by the the pagefault handler and passed to the vma's
218 * ->fault function. The vma's ->fault is responsible for returning a bitmask
219 * of VM_FAULT_xxx flags that give details about how the fault was handled.
221 * pgoff should be used in favour of virtual_address, if possible.
224 unsigned int flags
; /* FAULT_FLAG_xxx flags */
225 pgoff_t pgoff
; /* Logical page offset based on vma */
226 void __user
*virtual_address
; /* Faulting virtual address */
228 struct page
*page
; /* ->fault handlers should return a
229 * page here, unless VM_FAULT_NOPAGE
230 * is set (which is also implied by
233 /* for ->map_pages() only */
234 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
235 * max_pgoff inclusive */
236 pte_t
*pte
; /* pte entry associated with ->pgoff */
240 * These are the virtual MM functions - opening of an area, closing and
241 * unmapping it (needed to keep files on disk up-to-date etc), pointer
242 * to the functions called when a no-page or a wp-page exception occurs.
244 struct vm_operations_struct
{
245 void (*open
)(struct vm_area_struct
* area
);
246 void (*close
)(struct vm_area_struct
* area
);
247 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
248 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
250 /* notification that a previously read-only page is about to become
251 * writable, if an error is returned it will cause a SIGBUS */
252 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
254 /* called by access_process_vm when get_user_pages() fails, typically
255 * for use by special VMAs that can switch between memory and hardware
257 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
258 void *buf
, int len
, int write
);
260 /* Called by the /proc/PID/maps code to ask the vma whether it
261 * has a special name. Returning non-NULL will also cause this
262 * vma to be dumped unconditionally. */
263 const char *(*name
)(struct vm_area_struct
*vma
);
267 * set_policy() op must add a reference to any non-NULL @new mempolicy
268 * to hold the policy upon return. Caller should pass NULL @new to
269 * remove a policy and fall back to surrounding context--i.e. do not
270 * install a MPOL_DEFAULT policy, nor the task or system default
273 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
276 * get_policy() op must add reference [mpol_get()] to any policy at
277 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
278 * in mm/mempolicy.c will do this automatically.
279 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
280 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
281 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
282 * must return NULL--i.e., do not "fallback" to task or system default
285 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
288 /* called by sys_remap_file_pages() to populate non-linear mapping */
289 int (*remap_pages
)(struct vm_area_struct
*vma
, unsigned long addr
,
290 unsigned long size
, pgoff_t pgoff
);
296 #define page_private(page) ((page)->private)
297 #define set_page_private(page, v) ((page)->private = (v))
299 /* It's valid only if the page is free path or free_list */
300 static inline void set_freepage_migratetype(struct page
*page
, int migratetype
)
302 page
->index
= migratetype
;
305 /* It's valid only if the page is free path or free_list */
306 static inline int get_freepage_migratetype(struct page
*page
)
312 * FIXME: take this include out, include page-flags.h in
313 * files which need it (119 of them)
315 #include <linux/page-flags.h>
316 #include <linux/huge_mm.h>
319 * Methods to modify the page usage count.
321 * What counts for a page usage:
322 * - cache mapping (page->mapping)
323 * - private data (page->private)
324 * - page mapped in a task's page tables, each mapping
325 * is counted separately
327 * Also, many kernel routines increase the page count before a critical
328 * routine so they can be sure the page doesn't go away from under them.
332 * Drop a ref, return true if the refcount fell to zero (the page has no users)
334 static inline int put_page_testzero(struct page
*page
)
336 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) == 0, page
);
337 return atomic_dec_and_test(&page
->_count
);
341 * Try to grab a ref unless the page has a refcount of zero, return false if
343 * This can be called when MMU is off so it must not access
344 * any of the virtual mappings.
346 static inline int get_page_unless_zero(struct page
*page
)
348 return atomic_inc_not_zero(&page
->_count
);
352 * Try to drop a ref unless the page has a refcount of one, return false if
354 * This is to make sure that the refcount won't become zero after this drop.
355 * This can be called when MMU is off so it must not access
356 * any of the virtual mappings.
358 static inline int put_page_unless_one(struct page
*page
)
360 return atomic_add_unless(&page
->_count
, -1, 1);
363 extern int page_is_ram(unsigned long pfn
);
364 extern int region_is_ram(resource_size_t phys_addr
, unsigned long size
);
366 /* Support for virtually mapped pages */
367 struct page
*vmalloc_to_page(const void *addr
);
368 unsigned long vmalloc_to_pfn(const void *addr
);
371 * Determine if an address is within the vmalloc range
373 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
374 * is no special casing required.
376 static inline int is_vmalloc_addr(const void *x
)
379 unsigned long addr
= (unsigned long)x
;
381 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
387 extern int is_vmalloc_or_module_addr(const void *x
);
389 static inline int is_vmalloc_or_module_addr(const void *x
)
395 extern void kvfree(const void *addr
);
397 static inline void compound_lock(struct page
*page
)
399 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
400 VM_BUG_ON_PAGE(PageSlab(page
), page
);
401 bit_spin_lock(PG_compound_lock
, &page
->flags
);
405 static inline void compound_unlock(struct page
*page
)
407 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
408 VM_BUG_ON_PAGE(PageSlab(page
), page
);
409 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
413 static inline unsigned long compound_lock_irqsave(struct page
*page
)
415 unsigned long uninitialized_var(flags
);
416 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
417 local_irq_save(flags
);
423 static inline void compound_unlock_irqrestore(struct page
*page
,
426 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
427 compound_unlock(page
);
428 local_irq_restore(flags
);
432 static inline struct page
*compound_head_by_tail(struct page
*tail
)
434 struct page
*head
= tail
->first_page
;
437 * page->first_page may be a dangling pointer to an old
438 * compound page, so recheck that it is still a tail
439 * page before returning.
442 if (likely(PageTail(tail
)))
448 * Since either compound page could be dismantled asynchronously in THP
449 * or we access asynchronously arbitrary positioned struct page, there
450 * would be tail flag race. To handle this race, we should call
451 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
453 static inline struct page
*compound_head(struct page
*page
)
455 if (unlikely(PageTail(page
)))
456 return compound_head_by_tail(page
);
461 * If we access compound page synchronously such as access to
462 * allocated page, there is no need to handle tail flag race, so we can
463 * check tail flag directly without any synchronization primitive.
465 static inline struct page
*compound_head_fast(struct page
*page
)
467 if (unlikely(PageTail(page
)))
468 return page
->first_page
;
473 * The atomic page->_mapcount, starts from -1: so that transitions
474 * both from it and to it can be tracked, using atomic_inc_and_test
475 * and atomic_add_negative(-1).
477 static inline void page_mapcount_reset(struct page
*page
)
479 atomic_set(&(page
)->_mapcount
, -1);
482 static inline int page_mapcount(struct page
*page
)
484 return atomic_read(&(page
)->_mapcount
) + 1;
487 static inline int page_count(struct page
*page
)
489 return atomic_read(&compound_head(page
)->_count
);
492 #ifdef CONFIG_HUGETLB_PAGE
493 extern int PageHeadHuge(struct page
*page_head
);
494 #else /* CONFIG_HUGETLB_PAGE */
495 static inline int PageHeadHuge(struct page
*page_head
)
499 #endif /* CONFIG_HUGETLB_PAGE */
501 static inline bool __compound_tail_refcounted(struct page
*page
)
503 return !PageSlab(page
) && !PageHeadHuge(page
);
507 * This takes a head page as parameter and tells if the
508 * tail page reference counting can be skipped.
510 * For this to be safe, PageSlab and PageHeadHuge must remain true on
511 * any given page where they return true here, until all tail pins
512 * have been released.
514 static inline bool compound_tail_refcounted(struct page
*page
)
516 VM_BUG_ON_PAGE(!PageHead(page
), page
);
517 return __compound_tail_refcounted(page
);
520 static inline void get_huge_page_tail(struct page
*page
)
523 * __split_huge_page_refcount() cannot run from under us.
525 VM_BUG_ON_PAGE(!PageTail(page
), page
);
526 VM_BUG_ON_PAGE(page_mapcount(page
) < 0, page
);
527 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) != 0, page
);
528 if (compound_tail_refcounted(page
->first_page
))
529 atomic_inc(&page
->_mapcount
);
532 extern bool __get_page_tail(struct page
*page
);
534 static inline void get_page(struct page
*page
)
536 if (unlikely(PageTail(page
)))
537 if (likely(__get_page_tail(page
)))
540 * Getting a normal page or the head of a compound page
541 * requires to already have an elevated page->_count.
543 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
544 atomic_inc(&page
->_count
);
547 static inline struct page
*virt_to_head_page(const void *x
)
549 struct page
*page
= virt_to_page(x
);
552 * We don't need to worry about synchronization of tail flag
553 * when we call virt_to_head_page() since it is only called for
554 * already allocated page and this page won't be freed until
555 * this virt_to_head_page() is finished. So use _fast variant.
557 return compound_head_fast(page
);
561 * Setup the page count before being freed into the page allocator for
562 * the first time (boot or memory hotplug)
564 static inline void init_page_count(struct page
*page
)
566 atomic_set(&page
->_count
, 1);
570 * PageBuddy() indicate that the page is free and in the buddy system
571 * (see mm/page_alloc.c).
573 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
574 * -2 so that an underflow of the page_mapcount() won't be mistaken
575 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
576 * efficiently by most CPU architectures.
578 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
580 static inline int PageBuddy(struct page
*page
)
582 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
585 static inline void __SetPageBuddy(struct page
*page
)
587 VM_BUG_ON_PAGE(atomic_read(&page
->_mapcount
) != -1, page
);
588 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
591 static inline void __ClearPageBuddy(struct page
*page
)
593 VM_BUG_ON_PAGE(!PageBuddy(page
), page
);
594 atomic_set(&page
->_mapcount
, -1);
597 #define PAGE_BALLOON_MAPCOUNT_VALUE (-256)
599 static inline int PageBalloon(struct page
*page
)
601 return atomic_read(&page
->_mapcount
) == PAGE_BALLOON_MAPCOUNT_VALUE
;
604 static inline void __SetPageBalloon(struct page
*page
)
606 VM_BUG_ON_PAGE(atomic_read(&page
->_mapcount
) != -1, page
);
607 atomic_set(&page
->_mapcount
, PAGE_BALLOON_MAPCOUNT_VALUE
);
610 static inline void __ClearPageBalloon(struct page
*page
)
612 VM_BUG_ON_PAGE(!PageBalloon(page
), page
);
613 atomic_set(&page
->_mapcount
, -1);
616 void put_page(struct page
*page
);
617 void put_pages_list(struct list_head
*pages
);
619 void split_page(struct page
*page
, unsigned int order
);
620 int split_free_page(struct page
*page
);
623 * Compound pages have a destructor function. Provide a
624 * prototype for that function and accessor functions.
625 * These are _only_ valid on the head of a PG_compound page.
627 typedef void compound_page_dtor(struct page
*);
629 static inline void set_compound_page_dtor(struct page
*page
,
630 compound_page_dtor
*dtor
)
632 page
[1].lru
.next
= (void *)dtor
;
635 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
637 return (compound_page_dtor
*)page
[1].lru
.next
;
640 static inline int compound_order(struct page
*page
)
644 return (unsigned long)page
[1].lru
.prev
;
647 static inline void set_compound_order(struct page
*page
, unsigned long order
)
649 page
[1].lru
.prev
= (void *)order
;
654 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
655 * servicing faults for write access. In the normal case, do always want
656 * pte_mkwrite. But get_user_pages can cause write faults for mappings
657 * that do not have writing enabled, when used by access_process_vm.
659 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
661 if (likely(vma
->vm_flags
& VM_WRITE
))
662 pte
= pte_mkwrite(pte
);
666 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
667 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
671 * Multiple processes may "see" the same page. E.g. for untouched
672 * mappings of /dev/null, all processes see the same page full of
673 * zeroes, and text pages of executables and shared libraries have
674 * only one copy in memory, at most, normally.
676 * For the non-reserved pages, page_count(page) denotes a reference count.
677 * page_count() == 0 means the page is free. page->lru is then used for
678 * freelist management in the buddy allocator.
679 * page_count() > 0 means the page has been allocated.
681 * Pages are allocated by the slab allocator in order to provide memory
682 * to kmalloc and kmem_cache_alloc. In this case, the management of the
683 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
684 * unless a particular usage is carefully commented. (the responsibility of
685 * freeing the kmalloc memory is the caller's, of course).
687 * A page may be used by anyone else who does a __get_free_page().
688 * In this case, page_count still tracks the references, and should only
689 * be used through the normal accessor functions. The top bits of page->flags
690 * and page->virtual store page management information, but all other fields
691 * are unused and could be used privately, carefully. The management of this
692 * page is the responsibility of the one who allocated it, and those who have
693 * subsequently been given references to it.
695 * The other pages (we may call them "pagecache pages") are completely
696 * managed by the Linux memory manager: I/O, buffers, swapping etc.
697 * The following discussion applies only to them.
699 * A pagecache page contains an opaque `private' member, which belongs to the
700 * page's address_space. Usually, this is the address of a circular list of
701 * the page's disk buffers. PG_private must be set to tell the VM to call
702 * into the filesystem to release these pages.
704 * A page may belong to an inode's memory mapping. In this case, page->mapping
705 * is the pointer to the inode, and page->index is the file offset of the page,
706 * in units of PAGE_CACHE_SIZE.
708 * If pagecache pages are not associated with an inode, they are said to be
709 * anonymous pages. These may become associated with the swapcache, and in that
710 * case PG_swapcache is set, and page->private is an offset into the swapcache.
712 * In either case (swapcache or inode backed), the pagecache itself holds one
713 * reference to the page. Setting PG_private should also increment the
714 * refcount. The each user mapping also has a reference to the page.
716 * The pagecache pages are stored in a per-mapping radix tree, which is
717 * rooted at mapping->page_tree, and indexed by offset.
718 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
719 * lists, we instead now tag pages as dirty/writeback in the radix tree.
721 * All pagecache pages may be subject to I/O:
722 * - inode pages may need to be read from disk,
723 * - inode pages which have been modified and are MAP_SHARED may need
724 * to be written back to the inode on disk,
725 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
726 * modified may need to be swapped out to swap space and (later) to be read
731 * The zone field is never updated after free_area_init_core()
732 * sets it, so none of the operations on it need to be atomic.
735 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
736 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
737 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
738 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
739 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
742 * Define the bit shifts to access each section. For non-existent
743 * sections we define the shift as 0; that plus a 0 mask ensures
744 * the compiler will optimise away reference to them.
746 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
747 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
748 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
749 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
751 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
752 #ifdef NODE_NOT_IN_PAGE_FLAGS
753 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
754 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
755 SECTIONS_PGOFF : ZONES_PGOFF)
757 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
758 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
759 NODES_PGOFF : ZONES_PGOFF)
762 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
764 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
765 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
768 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
769 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
770 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
771 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
772 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
774 static inline enum zone_type
page_zonenum(const struct page
*page
)
776 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
779 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
780 #define SECTION_IN_PAGE_FLAGS
784 * The identification function is mainly used by the buddy allocator for
785 * determining if two pages could be buddies. We are not really identifying
786 * the zone since we could be using the section number id if we do not have
787 * node id available in page flags.
788 * We only guarantee that it will return the same value for two combinable
791 static inline int page_zone_id(struct page
*page
)
793 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
796 static inline int zone_to_nid(struct zone
*zone
)
805 #ifdef NODE_NOT_IN_PAGE_FLAGS
806 extern int page_to_nid(const struct page
*page
);
808 static inline int page_to_nid(const struct page
*page
)
810 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
814 #ifdef CONFIG_NUMA_BALANCING
815 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
817 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
820 static inline int cpupid_to_pid(int cpupid
)
822 return cpupid
& LAST__PID_MASK
;
825 static inline int cpupid_to_cpu(int cpupid
)
827 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
830 static inline int cpupid_to_nid(int cpupid
)
832 return cpu_to_node(cpupid_to_cpu(cpupid
));
835 static inline bool cpupid_pid_unset(int cpupid
)
837 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
840 static inline bool cpupid_cpu_unset(int cpupid
)
842 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
845 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
847 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
850 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
851 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
852 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
854 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
857 static inline int page_cpupid_last(struct page
*page
)
859 return page
->_last_cpupid
;
861 static inline void page_cpupid_reset_last(struct page
*page
)
863 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
866 static inline int page_cpupid_last(struct page
*page
)
868 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
871 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
873 static inline void page_cpupid_reset_last(struct page
*page
)
875 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
877 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
878 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
880 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
881 #else /* !CONFIG_NUMA_BALANCING */
882 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
884 return page_to_nid(page
); /* XXX */
887 static inline int page_cpupid_last(struct page
*page
)
889 return page_to_nid(page
); /* XXX */
892 static inline int cpupid_to_nid(int cpupid
)
897 static inline int cpupid_to_pid(int cpupid
)
902 static inline int cpupid_to_cpu(int cpupid
)
907 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
912 static inline bool cpupid_pid_unset(int cpupid
)
917 static inline void page_cpupid_reset_last(struct page
*page
)
921 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
925 #endif /* CONFIG_NUMA_BALANCING */
927 static inline struct zone
*page_zone(const struct page
*page
)
929 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
932 #ifdef SECTION_IN_PAGE_FLAGS
933 static inline void set_page_section(struct page
*page
, unsigned long section
)
935 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
936 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
939 static inline unsigned long page_to_section(const struct page
*page
)
941 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
945 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
947 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
948 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
951 static inline void set_page_node(struct page
*page
, unsigned long node
)
953 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
954 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
957 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
958 unsigned long node
, unsigned long pfn
)
960 set_page_zone(page
, zone
);
961 set_page_node(page
, node
);
962 #ifdef SECTION_IN_PAGE_FLAGS
963 set_page_section(page
, pfn_to_section_nr(pfn
));
968 * Some inline functions in vmstat.h depend on page_zone()
970 #include <linux/vmstat.h>
972 static __always_inline
void *lowmem_page_address(const struct page
*page
)
974 return __va(PFN_PHYS(page_to_pfn(page
)));
977 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
978 #define HASHED_PAGE_VIRTUAL
981 #if defined(WANT_PAGE_VIRTUAL)
982 static inline void *page_address(const struct page
*page
)
984 return page
->virtual;
986 static inline void set_page_address(struct page
*page
, void *address
)
988 page
->virtual = address
;
990 #define page_address_init() do { } while(0)
993 #if defined(HASHED_PAGE_VIRTUAL)
994 void *page_address(const struct page
*page
);
995 void set_page_address(struct page
*page
, void *virtual);
996 void page_address_init(void);
999 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1000 #define page_address(page) lowmem_page_address(page)
1001 #define set_page_address(page, address) do { } while(0)
1002 #define page_address_init() do { } while(0)
1006 * On an anonymous page mapped into a user virtual memory area,
1007 * page->mapping points to its anon_vma, not to a struct address_space;
1008 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
1010 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
1011 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
1012 * and then page->mapping points, not to an anon_vma, but to a private
1013 * structure which KSM associates with that merged page. See ksm.h.
1015 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
1017 * Please note that, confusingly, "page_mapping" refers to the inode
1018 * address_space which maps the page from disk; whereas "page_mapped"
1019 * refers to user virtual address space into which the page is mapped.
1021 #define PAGE_MAPPING_ANON 1
1022 #define PAGE_MAPPING_KSM 2
1023 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
1025 extern struct address_space
*page_mapping(struct page
*page
);
1027 /* Neutral page->mapping pointer to address_space or anon_vma or other */
1028 static inline void *page_rmapping(struct page
*page
)
1030 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
1033 extern struct address_space
*__page_file_mapping(struct page
*);
1036 struct address_space
*page_file_mapping(struct page
*page
)
1038 if (unlikely(PageSwapCache(page
)))
1039 return __page_file_mapping(page
);
1041 return page
->mapping
;
1044 static inline int PageAnon(struct page
*page
)
1046 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
1050 * Return the pagecache index of the passed page. Regular pagecache pages
1051 * use ->index whereas swapcache pages use ->private
1053 static inline pgoff_t
page_index(struct page
*page
)
1055 if (unlikely(PageSwapCache(page
)))
1056 return page_private(page
);
1060 extern pgoff_t
__page_file_index(struct page
*page
);
1063 * Return the file index of the page. Regular pagecache pages use ->index
1064 * whereas swapcache pages use swp_offset(->private)
1066 static inline pgoff_t
page_file_index(struct page
*page
)
1068 if (unlikely(PageSwapCache(page
)))
1069 return __page_file_index(page
);
1075 * Return true if this page is mapped into pagetables.
1077 static inline int page_mapped(struct page
*page
)
1079 return atomic_read(&(page
)->_mapcount
) >= 0;
1083 * Different kinds of faults, as returned by handle_mm_fault().
1084 * Used to decide whether a process gets delivered SIGBUS or
1085 * just gets major/minor fault counters bumped up.
1088 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1090 #define VM_FAULT_OOM 0x0001
1091 #define VM_FAULT_SIGBUS 0x0002
1092 #define VM_FAULT_MAJOR 0x0004
1093 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1094 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1095 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1096 #define VM_FAULT_SIGSEGV 0x0040
1098 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1099 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1100 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1101 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1103 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1105 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1106 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1109 /* Encode hstate index for a hwpoisoned large page */
1110 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1111 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1114 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1116 extern void pagefault_out_of_memory(void);
1118 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1121 * Flags passed to show_mem() and show_free_areas() to suppress output in
1124 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1126 extern void show_free_areas(unsigned int flags
);
1127 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1129 int shmem_zero_setup(struct vm_area_struct
*);
1131 bool shmem_mapping(struct address_space
*mapping
);
1133 static inline bool shmem_mapping(struct address_space
*mapping
)
1139 extern int can_do_mlock(void);
1140 extern int user_shm_lock(size_t, struct user_struct
*);
1141 extern void user_shm_unlock(size_t, struct user_struct
*);
1144 * Parameter block passed down to zap_pte_range in exceptional cases.
1146 struct zap_details
{
1147 struct address_space
*check_mapping
; /* Check page->mapping if set */
1148 pgoff_t first_index
; /* Lowest page->index to unmap */
1149 pgoff_t last_index
; /* Highest page->index to unmap */
1152 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1155 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1156 unsigned long size
);
1157 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1158 unsigned long size
, struct zap_details
*);
1159 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1160 unsigned long start
, unsigned long end
);
1163 * mm_walk - callbacks for walk_page_range
1164 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1165 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1166 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1167 * this handler is required to be able to handle
1168 * pmd_trans_huge() pmds. They may simply choose to
1169 * split_huge_page() instead of handling it explicitly.
1170 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1171 * @pte_hole: if set, called for each hole at all levels
1172 * @hugetlb_entry: if set, called for each hugetlb entry
1173 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1176 * (see walk_page_range for more details)
1179 int (*pgd_entry
)(pgd_t
*pgd
, unsigned long addr
,
1180 unsigned long next
, struct mm_walk
*walk
);
1181 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1182 unsigned long next
, struct mm_walk
*walk
);
1183 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1184 unsigned long next
, struct mm_walk
*walk
);
1185 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1186 unsigned long next
, struct mm_walk
*walk
);
1187 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1188 struct mm_walk
*walk
);
1189 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1190 unsigned long addr
, unsigned long next
,
1191 struct mm_walk
*walk
);
1192 struct mm_struct
*mm
;
1196 int walk_page_range(unsigned long addr
, unsigned long end
,
1197 struct mm_walk
*walk
);
1198 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1199 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1200 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1201 struct vm_area_struct
*vma
);
1202 void unmap_mapping_range(struct address_space
*mapping
,
1203 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1204 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1205 unsigned long *pfn
);
1206 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1207 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1208 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1209 void *buf
, int len
, int write
);
1211 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1212 loff_t
const holebegin
, loff_t
const holelen
)
1214 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1217 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1218 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1219 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1220 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1221 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1222 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1223 int invalidate_inode_page(struct page
*page
);
1226 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1227 unsigned long address
, unsigned int flags
);
1228 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1229 unsigned long address
, unsigned int fault_flags
);
1231 static inline int handle_mm_fault(struct mm_struct
*mm
,
1232 struct vm_area_struct
*vma
, unsigned long address
,
1235 /* should never happen if there's no MMU */
1237 return VM_FAULT_SIGBUS
;
1239 static inline int fixup_user_fault(struct task_struct
*tsk
,
1240 struct mm_struct
*mm
, unsigned long address
,
1241 unsigned int fault_flags
)
1243 /* should never happen if there's no MMU */
1249 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1250 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1251 void *buf
, int len
, int write
);
1253 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1254 unsigned long start
, unsigned long nr_pages
,
1255 unsigned int foll_flags
, struct page
**pages
,
1256 struct vm_area_struct
**vmas
, int *nonblocking
);
1257 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1258 unsigned long start
, unsigned long nr_pages
,
1259 int write
, int force
, struct page
**pages
,
1260 struct vm_area_struct
**vmas
);
1261 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1262 struct page
**pages
);
1264 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1265 struct page
**pages
);
1266 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1267 struct page
*get_dump_page(unsigned long addr
);
1269 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1270 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1271 unsigned int length
);
1273 int __set_page_dirty_nobuffers(struct page
*page
);
1274 int __set_page_dirty_no_writeback(struct page
*page
);
1275 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1277 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1278 int set_page_dirty(struct page
*page
);
1279 int set_page_dirty_lock(struct page
*page
);
1280 int clear_page_dirty_for_io(struct page
*page
);
1281 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1283 /* Is the vma a continuation of the stack vma above it? */
1284 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1286 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1289 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1292 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1293 (vma
->vm_start
== addr
) &&
1294 !vma_growsdown(vma
->vm_prev
, addr
);
1297 /* Is the vma a continuation of the stack vma below it? */
1298 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1300 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1303 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1306 return (vma
->vm_flags
& VM_GROWSUP
) &&
1307 (vma
->vm_end
== addr
) &&
1308 !vma_growsup(vma
->vm_next
, addr
);
1311 extern struct task_struct
*task_of_stack(struct task_struct
*task
,
1312 struct vm_area_struct
*vma
, bool in_group
);
1314 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1315 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1316 unsigned long new_addr
, unsigned long len
,
1317 bool need_rmap_locks
);
1318 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1319 unsigned long end
, pgprot_t newprot
,
1320 int dirty_accountable
, int prot_numa
);
1321 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1322 struct vm_area_struct
**pprev
, unsigned long start
,
1323 unsigned long end
, unsigned long newflags
);
1326 * doesn't attempt to fault and will return short.
1328 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1329 struct page
**pages
);
1331 * per-process(per-mm_struct) statistics.
1333 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1335 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1337 #ifdef SPLIT_RSS_COUNTING
1339 * counter is updated in asynchronous manner and may go to minus.
1340 * But it's never be expected number for users.
1345 return (unsigned long)val
;
1348 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1350 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1353 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1355 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1358 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1360 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1363 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1365 return get_mm_counter(mm
, MM_FILEPAGES
) +
1366 get_mm_counter(mm
, MM_ANONPAGES
);
1369 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1371 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1374 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1376 return max(mm
->hiwater_vm
, mm
->total_vm
);
1379 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1381 unsigned long _rss
= get_mm_rss(mm
);
1383 if ((mm
)->hiwater_rss
< _rss
)
1384 (mm
)->hiwater_rss
= _rss
;
1387 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1389 if (mm
->hiwater_vm
< mm
->total_vm
)
1390 mm
->hiwater_vm
= mm
->total_vm
;
1393 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1394 struct mm_struct
*mm
)
1396 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1398 if (*maxrss
< hiwater_rss
)
1399 *maxrss
= hiwater_rss
;
1402 #if defined(SPLIT_RSS_COUNTING)
1403 void sync_mm_rss(struct mm_struct
*mm
);
1405 static inline void sync_mm_rss(struct mm_struct
*mm
)
1410 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1412 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1414 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1418 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1422 #ifdef __PAGETABLE_PUD_FOLDED
1423 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1424 unsigned long address
)
1429 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1432 #ifdef __PAGETABLE_PMD_FOLDED
1433 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1434 unsigned long address
)
1439 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1442 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1443 pmd_t
*pmd
, unsigned long address
);
1444 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1447 * The following ifdef needed to get the 4level-fixup.h header to work.
1448 * Remove it when 4level-fixup.h has been removed.
1450 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1451 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1453 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1454 NULL
: pud_offset(pgd
, address
);
1457 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1459 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1460 NULL
: pmd_offset(pud
, address
);
1462 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1464 #if USE_SPLIT_PTE_PTLOCKS
1465 #if ALLOC_SPLIT_PTLOCKS
1466 void __init
ptlock_cache_init(void);
1467 extern bool ptlock_alloc(struct page
*page
);
1468 extern void ptlock_free(struct page
*page
);
1470 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1474 #else /* ALLOC_SPLIT_PTLOCKS */
1475 static inline void ptlock_cache_init(void)
1479 static inline bool ptlock_alloc(struct page
*page
)
1484 static inline void ptlock_free(struct page
*page
)
1488 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1492 #endif /* ALLOC_SPLIT_PTLOCKS */
1494 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1496 return ptlock_ptr(pmd_page(*pmd
));
1499 static inline bool ptlock_init(struct page
*page
)
1502 * prep_new_page() initialize page->private (and therefore page->ptl)
1503 * with 0. Make sure nobody took it in use in between.
1505 * It can happen if arch try to use slab for page table allocation:
1506 * slab code uses page->slab_cache and page->first_page (for tail
1507 * pages), which share storage with page->ptl.
1509 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1510 if (!ptlock_alloc(page
))
1512 spin_lock_init(ptlock_ptr(page
));
1516 /* Reset page->mapping so free_pages_check won't complain. */
1517 static inline void pte_lock_deinit(struct page
*page
)
1519 page
->mapping
= NULL
;
1523 #else /* !USE_SPLIT_PTE_PTLOCKS */
1525 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1527 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1529 return &mm
->page_table_lock
;
1531 static inline void ptlock_cache_init(void) {}
1532 static inline bool ptlock_init(struct page
*page
) { return true; }
1533 static inline void pte_lock_deinit(struct page
*page
) {}
1534 #endif /* USE_SPLIT_PTE_PTLOCKS */
1536 static inline void pgtable_init(void)
1538 ptlock_cache_init();
1539 pgtable_cache_init();
1542 static inline bool pgtable_page_ctor(struct page
*page
)
1544 inc_zone_page_state(page
, NR_PAGETABLE
);
1545 return ptlock_init(page
);
1548 static inline void pgtable_page_dtor(struct page
*page
)
1550 pte_lock_deinit(page
);
1551 dec_zone_page_state(page
, NR_PAGETABLE
);
1554 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1556 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1557 pte_t *__pte = pte_offset_map(pmd, address); \
1563 #define pte_unmap_unlock(pte, ptl) do { \
1568 #define pte_alloc_map(mm, vma, pmd, address) \
1569 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1571 NULL: pte_offset_map(pmd, address))
1573 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1574 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1576 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1578 #define pte_alloc_kernel(pmd, address) \
1579 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1580 NULL: pte_offset_kernel(pmd, address))
1582 #if USE_SPLIT_PMD_PTLOCKS
1584 static struct page
*pmd_to_page(pmd_t
*pmd
)
1586 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1587 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1590 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1592 return ptlock_ptr(pmd_to_page(pmd
));
1595 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1597 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1598 page
->pmd_huge_pte
= NULL
;
1600 return ptlock_init(page
);
1603 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1605 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1606 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1611 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1615 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1617 return &mm
->page_table_lock
;
1620 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1621 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1623 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1627 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1629 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1634 extern void free_area_init(unsigned long * zones_size
);
1635 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1636 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1637 extern void free_initmem(void);
1640 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1641 * into the buddy system. The freed pages will be poisoned with pattern
1642 * "poison" if it's within range [0, UCHAR_MAX].
1643 * Return pages freed into the buddy system.
1645 extern unsigned long free_reserved_area(void *start
, void *end
,
1646 int poison
, char *s
);
1648 #ifdef CONFIG_HIGHMEM
1650 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1651 * and totalram_pages.
1653 extern void free_highmem_page(struct page
*page
);
1656 extern void adjust_managed_page_count(struct page
*page
, long count
);
1657 extern void mem_init_print_info(const char *str
);
1659 /* Free the reserved page into the buddy system, so it gets managed. */
1660 static inline void __free_reserved_page(struct page
*page
)
1662 ClearPageReserved(page
);
1663 init_page_count(page
);
1667 static inline void free_reserved_page(struct page
*page
)
1669 __free_reserved_page(page
);
1670 adjust_managed_page_count(page
, 1);
1673 static inline void mark_page_reserved(struct page
*page
)
1675 SetPageReserved(page
);
1676 adjust_managed_page_count(page
, -1);
1680 * Default method to free all the __init memory into the buddy system.
1681 * The freed pages will be poisoned with pattern "poison" if it's within
1682 * range [0, UCHAR_MAX].
1683 * Return pages freed into the buddy system.
1685 static inline unsigned long free_initmem_default(int poison
)
1687 extern char __init_begin
[], __init_end
[];
1689 return free_reserved_area(&__init_begin
, &__init_end
,
1690 poison
, "unused kernel");
1693 static inline unsigned long get_num_physpages(void)
1696 unsigned long phys_pages
= 0;
1698 for_each_online_node(nid
)
1699 phys_pages
+= node_present_pages(nid
);
1704 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1706 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1707 * zones, allocate the backing mem_map and account for memory holes in a more
1708 * architecture independent manner. This is a substitute for creating the
1709 * zone_sizes[] and zholes_size[] arrays and passing them to
1710 * free_area_init_node()
1712 * An architecture is expected to register range of page frames backed by
1713 * physical memory with memblock_add[_node]() before calling
1714 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1715 * usage, an architecture is expected to do something like
1717 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1719 * for_each_valid_physical_page_range()
1720 * memblock_add_node(base, size, nid)
1721 * free_area_init_nodes(max_zone_pfns);
1723 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1724 * registered physical page range. Similarly
1725 * sparse_memory_present_with_active_regions() calls memory_present() for
1726 * each range when SPARSEMEM is enabled.
1728 * See mm/page_alloc.c for more information on each function exposed by
1729 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1731 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1732 unsigned long node_map_pfn_alignment(void);
1733 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1734 unsigned long end_pfn
);
1735 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1736 unsigned long end_pfn
);
1737 extern void get_pfn_range_for_nid(unsigned int nid
,
1738 unsigned long *start_pfn
, unsigned long *end_pfn
);
1739 extern unsigned long find_min_pfn_with_active_regions(void);
1740 extern void free_bootmem_with_active_regions(int nid
,
1741 unsigned long max_low_pfn
);
1742 extern void sparse_memory_present_with_active_regions(int nid
);
1744 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1746 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1747 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1748 static inline int __early_pfn_to_nid(unsigned long pfn
)
1753 /* please see mm/page_alloc.c */
1754 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1755 /* there is a per-arch backend function. */
1756 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1759 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1760 extern void memmap_init_zone(unsigned long, int, unsigned long,
1761 unsigned long, enum memmap_context
);
1762 extern void setup_per_zone_wmarks(void);
1763 extern int __meminit
init_per_zone_wmark_min(void);
1764 extern void mem_init(void);
1765 extern void __init
mmap_init(void);
1766 extern void show_mem(unsigned int flags
);
1767 extern void si_meminfo(struct sysinfo
* val
);
1768 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1770 extern __printf(3, 4)
1771 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1773 extern void setup_per_cpu_pageset(void);
1775 extern void zone_pcp_update(struct zone
*zone
);
1776 extern void zone_pcp_reset(struct zone
*zone
);
1779 extern int min_free_kbytes
;
1782 extern atomic_long_t mmap_pages_allocated
;
1783 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1785 /* interval_tree.c */
1786 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1787 struct rb_root
*root
);
1788 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1789 struct vm_area_struct
*prev
,
1790 struct rb_root
*root
);
1791 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1792 struct rb_root
*root
);
1793 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1794 unsigned long start
, unsigned long last
);
1795 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1796 unsigned long start
, unsigned long last
);
1798 #define vma_interval_tree_foreach(vma, root, start, last) \
1799 for (vma = vma_interval_tree_iter_first(root, start, last); \
1800 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1802 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1803 struct list_head
*list
)
1805 list_add_tail(&vma
->shared
.nonlinear
, list
);
1808 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1809 struct rb_root
*root
);
1810 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1811 struct rb_root
*root
);
1812 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1813 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1814 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1815 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1816 #ifdef CONFIG_DEBUG_VM_RB
1817 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1820 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1821 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1822 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1825 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1826 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1827 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1828 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1829 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1830 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1831 struct mempolicy
*);
1832 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1833 extern int split_vma(struct mm_struct
*,
1834 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1835 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1836 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1837 struct rb_node
**, struct rb_node
*);
1838 extern void unlink_file_vma(struct vm_area_struct
*);
1839 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1840 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1841 bool *need_rmap_locks
);
1842 extern void exit_mmap(struct mm_struct
*);
1844 static inline int check_data_rlimit(unsigned long rlim
,
1846 unsigned long start
,
1847 unsigned long end_data
,
1848 unsigned long start_data
)
1850 if (rlim
< RLIM_INFINITY
) {
1851 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1858 extern int mm_take_all_locks(struct mm_struct
*mm
);
1859 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1861 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1862 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1864 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1865 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
1866 unsigned long addr
, unsigned long len
,
1867 unsigned long flags
,
1868 const struct vm_special_mapping
*spec
);
1869 /* This is an obsolete alternative to _install_special_mapping. */
1870 extern int install_special_mapping(struct mm_struct
*mm
,
1871 unsigned long addr
, unsigned long len
,
1872 unsigned long flags
, struct page
**pages
);
1874 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1876 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1877 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1878 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1879 unsigned long len
, unsigned long prot
, unsigned long flags
,
1880 unsigned long pgoff
, unsigned long *populate
);
1881 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1884 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1886 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1889 (void) __mm_populate(addr
, len
, 1);
1892 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1895 /* These take the mm semaphore themselves */
1896 extern unsigned long vm_brk(unsigned long, unsigned long);
1897 extern int vm_munmap(unsigned long, size_t);
1898 extern unsigned long vm_mmap(struct file
*, unsigned long,
1899 unsigned long, unsigned long,
1900 unsigned long, unsigned long);
1902 struct vm_unmapped_area_info
{
1903 #define VM_UNMAPPED_AREA_TOPDOWN 1
1904 unsigned long flags
;
1905 unsigned long length
;
1906 unsigned long low_limit
;
1907 unsigned long high_limit
;
1908 unsigned long align_mask
;
1909 unsigned long align_offset
;
1912 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1913 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1916 * Search for an unmapped address range.
1918 * We are looking for a range that:
1919 * - does not intersect with any VMA;
1920 * - is contained within the [low_limit, high_limit) interval;
1921 * - is at least the desired size.
1922 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1924 static inline unsigned long
1925 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1927 if (!(info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
))
1928 return unmapped_area(info
);
1930 return unmapped_area_topdown(info
);
1934 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1935 extern void truncate_inode_pages_range(struct address_space
*,
1936 loff_t lstart
, loff_t lend
);
1937 extern void truncate_inode_pages_final(struct address_space
*);
1939 /* generic vm_area_ops exported for stackable file systems */
1940 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1941 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1942 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1944 /* mm/page-writeback.c */
1945 int write_one_page(struct page
*page
, int wait
);
1946 void task_dirty_inc(struct task_struct
*tsk
);
1949 #define VM_MAX_READAHEAD 128 /* kbytes */
1950 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1952 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1953 pgoff_t offset
, unsigned long nr_to_read
);
1955 void page_cache_sync_readahead(struct address_space
*mapping
,
1956 struct file_ra_state
*ra
,
1959 unsigned long size
);
1961 void page_cache_async_readahead(struct address_space
*mapping
,
1962 struct file_ra_state
*ra
,
1966 unsigned long size
);
1968 unsigned long max_sane_readahead(unsigned long nr
);
1970 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1971 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1973 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1974 extern int expand_downwards(struct vm_area_struct
*vma
,
1975 unsigned long address
);
1977 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1979 #define expand_upwards(vma, address) (0)
1982 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1983 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1984 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1985 struct vm_area_struct
**pprev
);
1987 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1988 NULL if none. Assume start_addr < end_addr. */
1989 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1991 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1993 if (vma
&& end_addr
<= vma
->vm_start
)
1998 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2000 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2003 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2004 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2005 unsigned long vm_start
, unsigned long vm_end
)
2007 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2009 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2016 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2017 void vma_set_page_prot(struct vm_area_struct
*vma
);
2019 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2023 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2025 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2029 #ifdef CONFIG_NUMA_BALANCING
2030 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2031 unsigned long start
, unsigned long end
);
2034 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2035 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2036 unsigned long pfn
, unsigned long size
, pgprot_t
);
2037 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2038 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2040 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2042 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2045 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2046 unsigned long address
, unsigned int foll_flags
,
2047 unsigned int *page_mask
);
2049 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2050 unsigned long address
, unsigned int foll_flags
)
2052 unsigned int unused_page_mask
;
2053 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2056 #define FOLL_WRITE 0x01 /* check pte is writable */
2057 #define FOLL_TOUCH 0x02 /* mark page accessed */
2058 #define FOLL_GET 0x04 /* do get_page on page */
2059 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2060 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2061 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2062 * and return without waiting upon it */
2063 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
2064 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2065 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2066 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2067 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2068 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2070 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2072 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2073 unsigned long size
, pte_fn_t fn
, void *data
);
2075 #ifdef CONFIG_PROC_FS
2076 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
2078 static inline void vm_stat_account(struct mm_struct
*mm
,
2079 unsigned long flags
, struct file
*file
, long pages
)
2081 mm
->total_vm
+= pages
;
2083 #endif /* CONFIG_PROC_FS */
2085 #ifdef CONFIG_DEBUG_PAGEALLOC
2086 extern bool _debug_pagealloc_enabled
;
2087 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2089 static inline bool debug_pagealloc_enabled(void)
2091 return _debug_pagealloc_enabled
;
2095 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2097 if (!debug_pagealloc_enabled())
2100 __kernel_map_pages(page
, numpages
, enable
);
2102 #ifdef CONFIG_HIBERNATION
2103 extern bool kernel_page_present(struct page
*page
);
2104 #endif /* CONFIG_HIBERNATION */
2107 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2108 #ifdef CONFIG_HIBERNATION
2109 static inline bool kernel_page_present(struct page
*page
) { return true; }
2110 #endif /* CONFIG_HIBERNATION */
2113 #ifdef __HAVE_ARCH_GATE_AREA
2114 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2115 extern int in_gate_area_no_mm(unsigned long addr
);
2116 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2118 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2122 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2123 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2127 #endif /* __HAVE_ARCH_GATE_AREA */
2129 #ifdef CONFIG_SYSCTL
2130 extern int sysctl_drop_caches
;
2131 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2132 void __user
*, size_t *, loff_t
*);
2135 unsigned long shrink_node_slabs(gfp_t gfp_mask
, int nid
,
2136 unsigned long nr_scanned
,
2137 unsigned long nr_eligible
);
2140 #define randomize_va_space 0
2142 extern int randomize_va_space
;
2145 const char * arch_vma_name(struct vm_area_struct
*vma
);
2146 void print_vma_addr(char *prefix
, unsigned long rip
);
2148 void sparse_mem_maps_populate_node(struct page
**map_map
,
2149 unsigned long pnum_begin
,
2150 unsigned long pnum_end
,
2151 unsigned long map_count
,
2154 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2155 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2156 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2157 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2158 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2159 void *vmemmap_alloc_block(unsigned long size
, int node
);
2160 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2161 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2162 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2164 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2165 void vmemmap_populate_print_last(void);
2166 #ifdef CONFIG_MEMORY_HOTPLUG
2167 void vmemmap_free(unsigned long start
, unsigned long end
);
2169 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2170 unsigned long size
);
2173 MF_COUNT_INCREASED
= 1 << 0,
2174 MF_ACTION_REQUIRED
= 1 << 1,
2175 MF_MUST_KILL
= 1 << 2,
2176 MF_SOFT_OFFLINE
= 1 << 3,
2178 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2179 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2180 extern int unpoison_memory(unsigned long pfn
);
2181 extern int sysctl_memory_failure_early_kill
;
2182 extern int sysctl_memory_failure_recovery
;
2183 extern void shake_page(struct page
*p
, int access
);
2184 extern atomic_long_t num_poisoned_pages
;
2185 extern int soft_offline_page(struct page
*page
, int flags
);
2187 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2188 extern void clear_huge_page(struct page
*page
,
2190 unsigned int pages_per_huge_page
);
2191 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2192 unsigned long addr
, struct vm_area_struct
*vma
,
2193 unsigned int pages_per_huge_page
);
2194 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2196 extern struct page_ext_operations debug_guardpage_ops
;
2197 extern struct page_ext_operations page_poisoning_ops
;
2199 #ifdef CONFIG_DEBUG_PAGEALLOC
2200 extern unsigned int _debug_guardpage_minorder
;
2201 extern bool _debug_guardpage_enabled
;
2203 static inline unsigned int debug_guardpage_minorder(void)
2205 return _debug_guardpage_minorder
;
2208 static inline bool debug_guardpage_enabled(void)
2210 return _debug_guardpage_enabled
;
2213 static inline bool page_is_guard(struct page
*page
)
2215 struct page_ext
*page_ext
;
2217 if (!debug_guardpage_enabled())
2220 page_ext
= lookup_page_ext(page
);
2221 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2224 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2225 static inline bool debug_guardpage_enabled(void) { return false; }
2226 static inline bool page_is_guard(struct page
*page
) { return false; }
2227 #endif /* CONFIG_DEBUG_PAGEALLOC */
2229 #if MAX_NUMNODES > 1
2230 void __init
setup_nr_node_ids(void);
2232 static inline void setup_nr_node_ids(void) {}
2235 #endif /* __KERNEL__ */
2236 #endif /* _LINUX_MM_H */