4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
20 struct writeback_control
;
22 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23 extern unsigned long max_mapnr
;
26 extern unsigned long num_physpages
;
27 extern void * high_memory
;
28 extern int page_cluster
;
31 extern int sysctl_legacy_va_layout
;
33 #define sysctl_legacy_va_layout 0
37 #include <asm/pgtable.h>
38 #include <asm/processor.h>
40 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
43 * Linux kernel virtual memory manager primitives.
44 * The idea being to have a "virtual" mm in the same way
45 * we have a virtual fs - giving a cleaner interface to the
46 * mm details, and allowing different kinds of memory mappings
47 * (from shared memory to executable loading to arbitrary
51 extern struct kmem_cache
*vm_area_cachep
;
54 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
55 * disabled, then there's a single shared list of VMAs maintained by the
56 * system, and mm's subscribe to these individually
58 struct vm_list_struct
{
59 struct vm_list_struct
*next
;
60 struct vm_area_struct
*vma
;
64 extern struct rb_root nommu_vma_tree
;
65 extern struct rw_semaphore nommu_vma_sem
;
67 extern unsigned int kobjsize(const void *objp
);
73 #define VM_READ 0x00000001 /* currently active flags */
74 #define VM_WRITE 0x00000002
75 #define VM_EXEC 0x00000004
76 #define VM_SHARED 0x00000008
78 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
79 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
80 #define VM_MAYWRITE 0x00000020
81 #define VM_MAYEXEC 0x00000040
82 #define VM_MAYSHARE 0x00000080
84 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
85 #define VM_GROWSUP 0x00000200
86 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
87 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
89 #define VM_EXECUTABLE 0x00001000
90 #define VM_LOCKED 0x00002000
91 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
93 /* Used by sys_madvise() */
94 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
95 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
97 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
98 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
99 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
100 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
101 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
102 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
103 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
104 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
105 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
107 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
109 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
110 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
113 #ifdef CONFIG_STACK_GROWSUP
114 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
116 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
119 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
120 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
121 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
122 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
123 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
126 * mapping from the currently active vm_flags protection bits (the
127 * low four bits) to a page protection mask..
129 extern pgprot_t protection_map
[16];
131 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
132 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
136 * vm_fault is filled by the the pagefault handler and passed to the vma's
137 * ->fault function. The vma's ->fault is responsible for returning a bitmask
138 * of VM_FAULT_xxx flags that give details about how the fault was handled.
140 * pgoff should be used in favour of virtual_address, if possible. If pgoff
141 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
145 unsigned int flags
; /* FAULT_FLAG_xxx flags */
146 pgoff_t pgoff
; /* Logical page offset based on vma */
147 void __user
*virtual_address
; /* Faulting virtual address */
149 struct page
*page
; /* ->fault handlers should return a
150 * page here, unless VM_FAULT_NOPAGE
151 * is set (which is also implied by
157 * These are the virtual MM functions - opening of an area, closing and
158 * unmapping it (needed to keep files on disk up-to-date etc), pointer
159 * to the functions called when a no-page or a wp-page exception occurs.
161 struct vm_operations_struct
{
162 void (*open
)(struct vm_area_struct
* area
);
163 void (*close
)(struct vm_area_struct
* area
);
164 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
165 struct page
*(*nopage
)(struct vm_area_struct
*area
,
166 unsigned long address
, int *type
);
167 unsigned long (*nopfn
)(struct vm_area_struct
*area
,
168 unsigned long address
);
170 /* notification that a previously read-only page is about to become
171 * writable, if an error is returned it will cause a SIGBUS */
172 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
174 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
175 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
177 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
178 const nodemask_t
*to
, unsigned long flags
);
185 #define page_private(page) ((page)->private)
186 #define set_page_private(page, v) ((page)->private = (v))
189 * FIXME: take this include out, include page-flags.h in
190 * files which need it (119 of them)
192 #include <linux/page-flags.h>
194 #ifdef CONFIG_DEBUG_VM
195 #define VM_BUG_ON(cond) BUG_ON(cond)
197 #define VM_BUG_ON(condition) do { } while(0)
201 * Methods to modify the page usage count.
203 * What counts for a page usage:
204 * - cache mapping (page->mapping)
205 * - private data (page->private)
206 * - page mapped in a task's page tables, each mapping
207 * is counted separately
209 * Also, many kernel routines increase the page count before a critical
210 * routine so they can be sure the page doesn't go away from under them.
214 * Drop a ref, return true if the refcount fell to zero (the page has no users)
216 static inline int put_page_testzero(struct page
*page
)
218 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
219 return atomic_dec_and_test(&page
->_count
);
223 * Try to grab a ref unless the page has a refcount of zero, return false if
226 static inline int get_page_unless_zero(struct page
*page
)
228 VM_BUG_ON(PageCompound(page
));
229 return atomic_inc_not_zero(&page
->_count
);
232 static inline struct page
*compound_head(struct page
*page
)
234 if (unlikely(PageTail(page
)))
235 return page
->first_page
;
239 static inline int page_count(struct page
*page
)
241 return atomic_read(&compound_head(page
)->_count
);
244 static inline void get_page(struct page
*page
)
246 page
= compound_head(page
);
247 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
248 atomic_inc(&page
->_count
);
251 static inline struct page
*virt_to_head_page(const void *x
)
253 struct page
*page
= virt_to_page(x
);
254 return compound_head(page
);
258 * Setup the page count before being freed into the page allocator for
259 * the first time (boot or memory hotplug)
261 static inline void init_page_count(struct page
*page
)
263 atomic_set(&page
->_count
, 1);
266 void put_page(struct page
*page
);
267 void put_pages_list(struct list_head
*pages
);
269 void split_page(struct page
*page
, unsigned int order
);
272 * Compound pages have a destructor function. Provide a
273 * prototype for that function and accessor functions.
274 * These are _only_ valid on the head of a PG_compound page.
276 typedef void compound_page_dtor(struct page
*);
278 static inline void set_compound_page_dtor(struct page
*page
,
279 compound_page_dtor
*dtor
)
281 page
[1].lru
.next
= (void *)dtor
;
284 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
286 return (compound_page_dtor
*)page
[1].lru
.next
;
289 static inline int compound_order(struct page
*page
)
293 return (unsigned long)page
[1].lru
.prev
;
296 static inline void set_compound_order(struct page
*page
, unsigned long order
)
298 page
[1].lru
.prev
= (void *)order
;
302 * Multiple processes may "see" the same page. E.g. for untouched
303 * mappings of /dev/null, all processes see the same page full of
304 * zeroes, and text pages of executables and shared libraries have
305 * only one copy in memory, at most, normally.
307 * For the non-reserved pages, page_count(page) denotes a reference count.
308 * page_count() == 0 means the page is free. page->lru is then used for
309 * freelist management in the buddy allocator.
310 * page_count() > 0 means the page has been allocated.
312 * Pages are allocated by the slab allocator in order to provide memory
313 * to kmalloc and kmem_cache_alloc. In this case, the management of the
314 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
315 * unless a particular usage is carefully commented. (the responsibility of
316 * freeing the kmalloc memory is the caller's, of course).
318 * A page may be used by anyone else who does a __get_free_page().
319 * In this case, page_count still tracks the references, and should only
320 * be used through the normal accessor functions. The top bits of page->flags
321 * and page->virtual store page management information, but all other fields
322 * are unused and could be used privately, carefully. The management of this
323 * page is the responsibility of the one who allocated it, and those who have
324 * subsequently been given references to it.
326 * The other pages (we may call them "pagecache pages") are completely
327 * managed by the Linux memory manager: I/O, buffers, swapping etc.
328 * The following discussion applies only to them.
330 * A pagecache page contains an opaque `private' member, which belongs to the
331 * page's address_space. Usually, this is the address of a circular list of
332 * the page's disk buffers. PG_private must be set to tell the VM to call
333 * into the filesystem to release these pages.
335 * A page may belong to an inode's memory mapping. In this case, page->mapping
336 * is the pointer to the inode, and page->index is the file offset of the page,
337 * in units of PAGE_CACHE_SIZE.
339 * If pagecache pages are not associated with an inode, they are said to be
340 * anonymous pages. These may become associated with the swapcache, and in that
341 * case PG_swapcache is set, and page->private is an offset into the swapcache.
343 * In either case (swapcache or inode backed), the pagecache itself holds one
344 * reference to the page. Setting PG_private should also increment the
345 * refcount. The each user mapping also has a reference to the page.
347 * The pagecache pages are stored in a per-mapping radix tree, which is
348 * rooted at mapping->page_tree, and indexed by offset.
349 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
350 * lists, we instead now tag pages as dirty/writeback in the radix tree.
352 * All pagecache pages may be subject to I/O:
353 * - inode pages may need to be read from disk,
354 * - inode pages which have been modified and are MAP_SHARED may need
355 * to be written back to the inode on disk,
356 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
357 * modified may need to be swapped out to swap space and (later) to be read
362 * The zone field is never updated after free_area_init_core()
363 * sets it, so none of the operations on it need to be atomic.
368 * page->flags layout:
370 * There are three possibilities for how page->flags get
371 * laid out. The first is for the normal case, without
372 * sparsemem. The second is for sparsemem when there is
373 * plenty of space for node and section. The last is when
374 * we have run out of space and have to fall back to an
375 * alternate (slower) way of determining the node.
377 * No sparsemem: | NODE | ZONE | ... | FLAGS |
378 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
379 * no space for node: | SECTION | ZONE | ... | FLAGS |
381 #ifdef CONFIG_SPARSEMEM
382 #define SECTIONS_WIDTH SECTIONS_SHIFT
384 #define SECTIONS_WIDTH 0
387 #define ZONES_WIDTH ZONES_SHIFT
389 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
390 #define NODES_WIDTH NODES_SHIFT
392 #define NODES_WIDTH 0
395 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
396 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
397 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
398 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
401 * We are going to use the flags for the page to node mapping if its in
402 * there. This includes the case where there is no node, so it is implicit.
404 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
405 #define NODE_NOT_IN_PAGE_FLAGS
408 #ifndef PFN_SECTION_SHIFT
409 #define PFN_SECTION_SHIFT 0
413 * Define the bit shifts to access each section. For non-existant
414 * sections we define the shift as 0; that plus a 0 mask ensures
415 * the compiler will optimise away reference to them.
417 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
418 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
419 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
421 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
422 #ifdef NODE_NOT_IN_PAGEFLAGS
423 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
424 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
425 SECTIONS_PGOFF : ZONES_PGOFF)
427 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
428 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
429 NODES_PGOFF : ZONES_PGOFF)
432 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
434 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
435 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
438 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
439 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
440 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
441 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
443 static inline enum zone_type
page_zonenum(struct page
*page
)
445 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
449 * The identification function is only used by the buddy allocator for
450 * determining if two pages could be buddies. We are not really
451 * identifying a zone since we could be using a the section number
452 * id if we have not node id available in page flags.
453 * We guarantee only that it will return the same value for two
454 * combinable pages in a zone.
456 static inline int page_zone_id(struct page
*page
)
458 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
461 static inline int zone_to_nid(struct zone
*zone
)
470 #ifdef NODE_NOT_IN_PAGE_FLAGS
471 extern int page_to_nid(struct page
*page
);
473 static inline int page_to_nid(struct page
*page
)
475 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
479 static inline struct zone
*page_zone(struct page
*page
)
481 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
484 static inline unsigned long page_to_section(struct page
*page
)
486 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
489 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
491 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
492 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
495 static inline void set_page_node(struct page
*page
, unsigned long node
)
497 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
498 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
501 static inline void set_page_section(struct page
*page
, unsigned long section
)
503 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
504 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
507 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
508 unsigned long node
, unsigned long pfn
)
510 set_page_zone(page
, zone
);
511 set_page_node(page
, node
);
512 set_page_section(page
, pfn_to_section_nr(pfn
));
516 * Some inline functions in vmstat.h depend on page_zone()
518 #include <linux/vmstat.h>
520 static __always_inline
void *lowmem_page_address(struct page
*page
)
522 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
525 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
526 #define HASHED_PAGE_VIRTUAL
529 #if defined(WANT_PAGE_VIRTUAL)
530 #define page_address(page) ((page)->virtual)
531 #define set_page_address(page, address) \
533 (page)->virtual = (address); \
535 #define page_address_init() do { } while(0)
538 #if defined(HASHED_PAGE_VIRTUAL)
539 void *page_address(struct page
*page
);
540 void set_page_address(struct page
*page
, void *virtual);
541 void page_address_init(void);
544 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
545 #define page_address(page) lowmem_page_address(page)
546 #define set_page_address(page, address) do { } while(0)
547 #define page_address_init() do { } while(0)
551 * On an anonymous page mapped into a user virtual memory area,
552 * page->mapping points to its anon_vma, not to a struct address_space;
553 * with the PAGE_MAPPING_ANON bit set to distinguish it.
555 * Please note that, confusingly, "page_mapping" refers to the inode
556 * address_space which maps the page from disk; whereas "page_mapped"
557 * refers to user virtual address space into which the page is mapped.
559 #define PAGE_MAPPING_ANON 1
561 extern struct address_space swapper_space
;
562 static inline struct address_space
*page_mapping(struct page
*page
)
564 struct address_space
*mapping
= page
->mapping
;
566 VM_BUG_ON(PageSlab(page
));
567 if (unlikely(PageSwapCache(page
)))
568 mapping
= &swapper_space
;
569 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
574 static inline int PageAnon(struct page
*page
)
576 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
580 * Return the pagecache index of the passed page. Regular pagecache pages
581 * use ->index whereas swapcache pages use ->private
583 static inline pgoff_t
page_index(struct page
*page
)
585 if (unlikely(PageSwapCache(page
)))
586 return page_private(page
);
591 * The atomic page->_mapcount, like _count, starts from -1:
592 * so that transitions both from it and to it can be tracked,
593 * using atomic_inc_and_test and atomic_add_negative(-1).
595 static inline void reset_page_mapcount(struct page
*page
)
597 atomic_set(&(page
)->_mapcount
, -1);
600 static inline int page_mapcount(struct page
*page
)
602 return atomic_read(&(page
)->_mapcount
) + 1;
606 * Return true if this page is mapped into pagetables.
608 static inline int page_mapped(struct page
*page
)
610 return atomic_read(&(page
)->_mapcount
) >= 0;
614 * Error return values for the *_nopage functions
616 #define NOPAGE_SIGBUS (NULL)
617 #define NOPAGE_OOM ((struct page *) (-1))
620 * Error return values for the *_nopfn functions
622 #define NOPFN_SIGBUS ((unsigned long) -1)
623 #define NOPFN_OOM ((unsigned long) -2)
624 #define NOPFN_REFAULT ((unsigned long) -3)
627 * Different kinds of faults, as returned by handle_mm_fault().
628 * Used to decide whether a process gets delivered SIGBUS or
629 * just gets major/minor fault counters bumped up.
632 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
634 #define VM_FAULT_OOM 0x0001
635 #define VM_FAULT_SIGBUS 0x0002
636 #define VM_FAULT_MAJOR 0x0004
637 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
639 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
640 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
642 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
644 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
646 extern void show_free_areas(void);
649 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
651 static inline int shmem_lock(struct file
*file
, int lock
,
652 struct user_struct
*user
)
657 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
659 int shmem_zero_setup(struct vm_area_struct
*);
662 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
666 unsigned long flags
);
669 extern int can_do_mlock(void);
670 extern int user_shm_lock(size_t, struct user_struct
*);
671 extern void user_shm_unlock(size_t, struct user_struct
*);
674 * Parameter block passed down to zap_pte_range in exceptional cases.
677 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
678 struct address_space
*check_mapping
; /* Check page->mapping if set */
679 pgoff_t first_index
; /* Lowest page->index to unmap */
680 pgoff_t last_index
; /* Highest page->index to unmap */
681 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
682 unsigned long truncate_count
; /* Compare vm_truncate_count */
685 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
686 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
687 unsigned long size
, struct zap_details
*);
688 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
689 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
690 unsigned long end_addr
, unsigned long *nr_accounted
,
691 struct zap_details
*);
692 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
693 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
694 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
695 unsigned long floor
, unsigned long ceiling
);
696 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
697 struct vm_area_struct
*vma
);
698 void unmap_mapping_range(struct address_space
*mapping
,
699 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
701 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
702 loff_t
const holebegin
, loff_t
const holelen
)
704 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
707 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
708 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
711 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
712 unsigned long address
, int write_access
);
714 static inline int handle_mm_fault(struct mm_struct
*mm
,
715 struct vm_area_struct
*vma
, unsigned long address
,
718 /* should never happen if there's no MMU */
720 return VM_FAULT_SIGBUS
;
724 extern int make_pages_present(unsigned long addr
, unsigned long end
);
725 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
727 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
728 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
729 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
731 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
732 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
734 int __set_page_dirty_nobuffers(struct page
*page
);
735 int __set_page_dirty_no_writeback(struct page
*page
);
736 int redirty_page_for_writepage(struct writeback_control
*wbc
,
738 int FASTCALL(set_page_dirty(struct page
*page
));
739 int set_page_dirty_lock(struct page
*page
);
740 int clear_page_dirty_for_io(struct page
*page
);
742 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
743 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
744 unsigned long new_addr
, unsigned long len
);
745 extern unsigned long do_mremap(unsigned long addr
,
746 unsigned long old_len
, unsigned long new_len
,
747 unsigned long flags
, unsigned long new_addr
);
748 extern int mprotect_fixup(struct vm_area_struct
*vma
,
749 struct vm_area_struct
**pprev
, unsigned long start
,
750 unsigned long end
, unsigned long newflags
);
753 * A callback you can register to apply pressure to ageable caches.
755 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
756 * look through the least-recently-used 'nr_to_scan' entries and
757 * attempt to free them up. It should return the number of objects
758 * which remain in the cache. If it returns -1, it means it cannot do
759 * any scanning at this time (eg. there is a risk of deadlock).
761 * The 'gfpmask' refers to the allocation we are currently trying to
764 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
765 * querying the cache size, so a fastpath for that case is appropriate.
768 int (*shrink
)(int nr_to_scan
, gfp_t gfp_mask
);
769 int seeks
; /* seeks to recreate an obj */
771 /* These are for internal use */
772 struct list_head list
;
773 long nr
; /* objs pending delete */
775 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
776 extern void register_shrinker(struct shrinker
*);
777 extern void unregister_shrinker(struct shrinker
*);
779 int vma_wants_writenotify(struct vm_area_struct
*vma
);
781 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
783 #ifdef __PAGETABLE_PUD_FOLDED
784 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
785 unsigned long address
)
790 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
793 #ifdef __PAGETABLE_PMD_FOLDED
794 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
795 unsigned long address
)
800 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
803 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
804 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
807 * The following ifdef needed to get the 4level-fixup.h header to work.
808 * Remove it when 4level-fixup.h has been removed.
810 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
811 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
813 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
814 NULL
: pud_offset(pgd
, address
);
817 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
819 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
820 NULL
: pmd_offset(pud
, address
);
822 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
824 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
826 * We tuck a spinlock to guard each pagetable page into its struct page,
827 * at page->private, with BUILD_BUG_ON to make sure that this will not
828 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
829 * When freeing, reset page->mapping so free_pages_check won't complain.
831 #define __pte_lockptr(page) &((page)->ptl)
832 #define pte_lock_init(_page) do { \
833 spin_lock_init(__pte_lockptr(_page)); \
835 #define pte_lock_deinit(page) ((page)->mapping = NULL)
836 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
839 * We use mm->page_table_lock to guard all pagetable pages of the mm.
841 #define pte_lock_init(page) do {} while (0)
842 #define pte_lock_deinit(page) do {} while (0)
843 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
844 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
846 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
848 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
849 pte_t *__pte = pte_offset_map(pmd, address); \
855 #define pte_unmap_unlock(pte, ptl) do { \
860 #define pte_alloc_map(mm, pmd, address) \
861 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
862 NULL: pte_offset_map(pmd, address))
864 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
865 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
866 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
868 #define pte_alloc_kernel(pmd, address) \
869 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
870 NULL: pte_offset_kernel(pmd, address))
872 extern void free_area_init(unsigned long * zones_size
);
873 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
874 unsigned long * zones_size
, unsigned long zone_start_pfn
,
875 unsigned long *zholes_size
);
876 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
878 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
879 * zones, allocate the backing mem_map and account for memory holes in a more
880 * architecture independent manner. This is a substitute for creating the
881 * zone_sizes[] and zholes_size[] arrays and passing them to
882 * free_area_init_node()
884 * An architecture is expected to register range of page frames backed by
885 * physical memory with add_active_range() before calling
886 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
887 * usage, an architecture is expected to do something like
889 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
891 * for_each_valid_physical_page_range()
892 * add_active_range(node_id, start_pfn, end_pfn)
893 * free_area_init_nodes(max_zone_pfns);
895 * If the architecture guarantees that there are no holes in the ranges
896 * registered with add_active_range(), free_bootmem_active_regions()
897 * will call free_bootmem_node() for each registered physical page range.
898 * Similarly sparse_memory_present_with_active_regions() calls
899 * memory_present() for each range when SPARSEMEM is enabled.
901 * See mm/page_alloc.c for more information on each function exposed by
902 * CONFIG_ARCH_POPULATES_NODE_MAP
904 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
905 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
906 unsigned long end_pfn
);
907 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
908 unsigned long new_end_pfn
);
909 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
910 unsigned long end_pfn
);
911 extern void remove_all_active_ranges(void);
912 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
913 unsigned long end_pfn
);
914 extern void get_pfn_range_for_nid(unsigned int nid
,
915 unsigned long *start_pfn
, unsigned long *end_pfn
);
916 extern unsigned long find_min_pfn_with_active_regions(void);
917 extern unsigned long find_max_pfn_with_active_regions(void);
918 extern void free_bootmem_with_active_regions(int nid
,
919 unsigned long max_low_pfn
);
920 extern void sparse_memory_present_with_active_regions(int nid
);
921 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
922 extern int early_pfn_to_nid(unsigned long pfn
);
923 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
924 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
925 extern void set_dma_reserve(unsigned long new_dma_reserve
);
926 extern void memmap_init_zone(unsigned long, int, unsigned long,
927 unsigned long, enum memmap_context
);
928 extern void setup_per_zone_pages_min(void);
929 extern void mem_init(void);
930 extern void show_mem(void);
931 extern void si_meminfo(struct sysinfo
* val
);
932 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
935 extern void setup_per_cpu_pageset(void);
937 static inline void setup_per_cpu_pageset(void) {}
941 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
942 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
943 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
944 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
945 struct prio_tree_iter
*iter
);
947 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
948 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
949 (vma = vma_prio_tree_next(vma, iter)); )
951 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
952 struct list_head
*list
)
954 vma
->shared
.vm_set
.parent
= NULL
;
955 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
959 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
960 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
961 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
962 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
963 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
964 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
966 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
967 extern int split_vma(struct mm_struct
*,
968 struct vm_area_struct
*, unsigned long addr
, int new_below
);
969 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
970 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
971 struct rb_node
**, struct rb_node
*);
972 extern void unlink_file_vma(struct vm_area_struct
*);
973 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
974 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
975 extern void exit_mmap(struct mm_struct
*);
976 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
977 extern int install_special_mapping(struct mm_struct
*mm
,
978 unsigned long addr
, unsigned long len
,
979 unsigned long flags
, struct page
**pages
);
981 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
983 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
984 unsigned long len
, unsigned long prot
,
985 unsigned long flag
, unsigned long pgoff
);
986 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
987 unsigned long len
, unsigned long flags
,
988 unsigned int vm_flags
, unsigned long pgoff
,
991 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
992 unsigned long len
, unsigned long prot
,
993 unsigned long flag
, unsigned long offset
)
995 unsigned long ret
= -EINVAL
;
996 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
998 if (!(offset
& ~PAGE_MASK
))
999 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1004 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1006 extern unsigned long do_brk(unsigned long, unsigned long);
1009 extern unsigned long page_unuse(struct page
*);
1010 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1011 extern void truncate_inode_pages_range(struct address_space
*,
1012 loff_t lstart
, loff_t lend
);
1014 /* generic vm_area_ops exported for stackable file systems */
1015 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1017 /* mm/page-writeback.c */
1018 int write_one_page(struct page
*page
, int wait
);
1021 #define VM_MAX_READAHEAD 128 /* kbytes */
1022 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1024 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1025 pgoff_t offset
, unsigned long nr_to_read
);
1026 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1027 pgoff_t offset
, unsigned long nr_to_read
);
1029 void page_cache_sync_readahead(struct address_space
*mapping
,
1030 struct file_ra_state
*ra
,
1033 unsigned long size
);
1035 void page_cache_async_readahead(struct address_space
*mapping
,
1036 struct file_ra_state
*ra
,
1040 unsigned long size
);
1042 unsigned long max_sane_readahead(unsigned long nr
);
1044 /* Do stack extension */
1045 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1047 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1049 extern int expand_stack_downwards(struct vm_area_struct
*vma
,
1050 unsigned long address
);
1052 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1053 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1054 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1055 struct vm_area_struct
**pprev
);
1057 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1058 NULL if none. Assume start_addr < end_addr. */
1059 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1061 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1063 if (vma
&& end_addr
<= vma
->vm_start
)
1068 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1070 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1073 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1074 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1075 struct page
*vmalloc_to_page(void *addr
);
1076 unsigned long vmalloc_to_pfn(void *addr
);
1077 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1078 unsigned long pfn
, unsigned long size
, pgprot_t
);
1079 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1080 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1083 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1084 unsigned int foll_flags
);
1085 #define FOLL_WRITE 0x01 /* check pte is writable */
1086 #define FOLL_TOUCH 0x02 /* mark page accessed */
1087 #define FOLL_GET 0x04 /* do get_page on page */
1088 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1090 typedef int (*pte_fn_t
)(pte_t
*pte
, struct page
*pmd_page
, unsigned long addr
,
1092 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1093 unsigned long size
, pte_fn_t fn
, void *data
);
1095 #ifdef CONFIG_PROC_FS
1096 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1098 static inline void vm_stat_account(struct mm_struct
*mm
,
1099 unsigned long flags
, struct file
*file
, long pages
)
1102 #endif /* CONFIG_PROC_FS */
1104 #ifndef CONFIG_DEBUG_PAGEALLOC
1106 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1109 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1110 #ifdef __HAVE_ARCH_GATE_AREA
1111 int in_gate_area_no_task(unsigned long addr
);
1112 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1114 int in_gate_area_no_task(unsigned long addr
);
1115 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1116 #endif /* __HAVE_ARCH_GATE_AREA */
1118 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1119 void __user
*, size_t *, loff_t
*);
1120 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1121 unsigned long lru_pages
);
1122 void drop_pagecache(void);
1123 void drop_slab(void);
1126 #define randomize_va_space 0
1128 extern int randomize_va_space
;
1131 const char * arch_vma_name(struct vm_area_struct
*vma
);
1133 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1134 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1135 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1136 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1137 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1138 void *vmemmap_alloc_block(unsigned long size
, int node
);
1139 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1140 int vmemmap_populate_basepages(struct page
*start_page
,
1141 unsigned long pages
, int node
);
1142 int vmemmap_populate(struct page
*start_page
, unsigned long pages
, int node
);
1144 #endif /* __KERNEL__ */
1145 #endif /* _LINUX_MM_H */