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
36 extern unsigned long mmap_min_addr
;
39 #include <asm/pgtable.h>
40 #include <asm/processor.h>
42 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
53 extern struct kmem_cache
*vm_area_cachep
;
56 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
57 * disabled, then there's a single shared list of VMAs maintained by the
58 * system, and mm's subscribe to these individually
60 struct vm_list_struct
{
61 struct vm_list_struct
*next
;
62 struct vm_area_struct
*vma
;
66 extern struct rb_root nommu_vma_tree
;
67 extern struct rw_semaphore nommu_vma_sem
;
69 extern unsigned int kobjsize(const void *objp
);
75 #define VM_READ 0x00000001 /* currently active flags */
76 #define VM_WRITE 0x00000002
77 #define VM_EXEC 0x00000004
78 #define VM_SHARED 0x00000008
80 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
81 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
82 #define VM_MAYWRITE 0x00000020
83 #define VM_MAYEXEC 0x00000040
84 #define VM_MAYSHARE 0x00000080
86 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
87 #define VM_GROWSUP 0x00000200
88 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
89 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
91 #define VM_EXECUTABLE 0x00001000
92 #define VM_LOCKED 0x00002000
93 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
95 /* Used by sys_madvise() */
96 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
99 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
101 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
102 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
104 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
105 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
106 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
107 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
109 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
111 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
112 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
115 #ifdef CONFIG_STACK_GROWSUP
116 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
118 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
121 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
122 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
123 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
124 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
125 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
128 * mapping from the currently active vm_flags protection bits (the
129 * low four bits) to a page protection mask..
131 extern pgprot_t protection_map
[16];
133 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
134 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
138 * vm_fault is filled by the the pagefault handler and passed to the vma's
139 * ->fault function. The vma's ->fault is responsible for returning a bitmask
140 * of VM_FAULT_xxx flags that give details about how the fault was handled.
142 * pgoff should be used in favour of virtual_address, if possible. If pgoff
143 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
147 unsigned int flags
; /* FAULT_FLAG_xxx flags */
148 pgoff_t pgoff
; /* Logical page offset based on vma */
149 void __user
*virtual_address
; /* Faulting virtual address */
151 struct page
*page
; /* ->fault handlers should return a
152 * page here, unless VM_FAULT_NOPAGE
153 * is set (which is also implied by
159 * These are the virtual MM functions - opening of an area, closing and
160 * unmapping it (needed to keep files on disk up-to-date etc), pointer
161 * to the functions called when a no-page or a wp-page exception occurs.
163 struct vm_operations_struct
{
164 void (*open
)(struct vm_area_struct
* area
);
165 void (*close
)(struct vm_area_struct
* area
);
166 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
167 struct page
*(*nopage
)(struct vm_area_struct
*area
,
168 unsigned long address
, int *type
);
169 unsigned long (*nopfn
)(struct vm_area_struct
*area
,
170 unsigned long address
);
172 /* notification that a previously read-only page is about to become
173 * writable, if an error is returned it will cause a SIGBUS */
174 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
176 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
177 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
179 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
180 const nodemask_t
*to
, unsigned long flags
);
187 #define page_private(page) ((page)->private)
188 #define set_page_private(page, v) ((page)->private = (v))
191 * FIXME: take this include out, include page-flags.h in
192 * files which need it (119 of them)
194 #include <linux/page-flags.h>
196 #ifdef CONFIG_DEBUG_VM
197 #define VM_BUG_ON(cond) BUG_ON(cond)
199 #define VM_BUG_ON(condition) do { } while(0)
203 * Methods to modify the page usage count.
205 * What counts for a page usage:
206 * - cache mapping (page->mapping)
207 * - private data (page->private)
208 * - page mapped in a task's page tables, each mapping
209 * is counted separately
211 * Also, many kernel routines increase the page count before a critical
212 * routine so they can be sure the page doesn't go away from under them.
216 * Drop a ref, return true if the refcount fell to zero (the page has no users)
218 static inline int put_page_testzero(struct page
*page
)
220 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
221 return atomic_dec_and_test(&page
->_count
);
225 * Try to grab a ref unless the page has a refcount of zero, return false if
228 static inline int get_page_unless_zero(struct page
*page
)
230 VM_BUG_ON(PageTail(page
));
231 return atomic_inc_not_zero(&page
->_count
);
234 /* Support for virtually mapped pages */
235 struct page
*vmalloc_to_page(const void *addr
);
236 unsigned long vmalloc_to_pfn(const void *addr
);
239 /* Determine if an address is within the vmalloc range */
240 static inline int is_vmalloc_addr(const void *x
)
242 unsigned long addr
= (unsigned long)x
;
244 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
248 static inline struct page
*compound_head(struct page
*page
)
250 if (unlikely(PageTail(page
)))
251 return page
->first_page
;
255 static inline int page_count(struct page
*page
)
257 return atomic_read(&compound_head(page
)->_count
);
260 static inline void get_page(struct page
*page
)
262 page
= compound_head(page
);
263 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
264 atomic_inc(&page
->_count
);
267 static inline struct page
*virt_to_head_page(const void *x
)
269 struct page
*page
= virt_to_page(x
);
270 return compound_head(page
);
274 * Setup the page count before being freed into the page allocator for
275 * the first time (boot or memory hotplug)
277 static inline void init_page_count(struct page
*page
)
279 atomic_set(&page
->_count
, 1);
282 void put_page(struct page
*page
);
283 void put_pages_list(struct list_head
*pages
);
285 void split_page(struct page
*page
, unsigned int order
);
288 * Compound pages have a destructor function. Provide a
289 * prototype for that function and accessor functions.
290 * These are _only_ valid on the head of a PG_compound page.
292 typedef void compound_page_dtor(struct page
*);
294 static inline void set_compound_page_dtor(struct page
*page
,
295 compound_page_dtor
*dtor
)
297 page
[1].lru
.next
= (void *)dtor
;
300 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
302 return (compound_page_dtor
*)page
[1].lru
.next
;
305 static inline int compound_order(struct page
*page
)
309 return (unsigned long)page
[1].lru
.prev
;
312 static inline void set_compound_order(struct page
*page
, unsigned long order
)
314 page
[1].lru
.prev
= (void *)order
;
318 * Multiple processes may "see" the same page. E.g. for untouched
319 * mappings of /dev/null, all processes see the same page full of
320 * zeroes, and text pages of executables and shared libraries have
321 * only one copy in memory, at most, normally.
323 * For the non-reserved pages, page_count(page) denotes a reference count.
324 * page_count() == 0 means the page is free. page->lru is then used for
325 * freelist management in the buddy allocator.
326 * page_count() > 0 means the page has been allocated.
328 * Pages are allocated by the slab allocator in order to provide memory
329 * to kmalloc and kmem_cache_alloc. In this case, the management of the
330 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
331 * unless a particular usage is carefully commented. (the responsibility of
332 * freeing the kmalloc memory is the caller's, of course).
334 * A page may be used by anyone else who does a __get_free_page().
335 * In this case, page_count still tracks the references, and should only
336 * be used through the normal accessor functions. The top bits of page->flags
337 * and page->virtual store page management information, but all other fields
338 * are unused and could be used privately, carefully. The management of this
339 * page is the responsibility of the one who allocated it, and those who have
340 * subsequently been given references to it.
342 * The other pages (we may call them "pagecache pages") are completely
343 * managed by the Linux memory manager: I/O, buffers, swapping etc.
344 * The following discussion applies only to them.
346 * A pagecache page contains an opaque `private' member, which belongs to the
347 * page's address_space. Usually, this is the address of a circular list of
348 * the page's disk buffers. PG_private must be set to tell the VM to call
349 * into the filesystem to release these pages.
351 * A page may belong to an inode's memory mapping. In this case, page->mapping
352 * is the pointer to the inode, and page->index is the file offset of the page,
353 * in units of PAGE_CACHE_SIZE.
355 * If pagecache pages are not associated with an inode, they are said to be
356 * anonymous pages. These may become associated with the swapcache, and in that
357 * case PG_swapcache is set, and page->private is an offset into the swapcache.
359 * In either case (swapcache or inode backed), the pagecache itself holds one
360 * reference to the page. Setting PG_private should also increment the
361 * refcount. The each user mapping also has a reference to the page.
363 * The pagecache pages are stored in a per-mapping radix tree, which is
364 * rooted at mapping->page_tree, and indexed by offset.
365 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
366 * lists, we instead now tag pages as dirty/writeback in the radix tree.
368 * All pagecache pages may be subject to I/O:
369 * - inode pages may need to be read from disk,
370 * - inode pages which have been modified and are MAP_SHARED may need
371 * to be written back to the inode on disk,
372 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
373 * modified may need to be swapped out to swap space and (later) to be read
378 * The zone field is never updated after free_area_init_core()
379 * sets it, so none of the operations on it need to be atomic.
384 * page->flags layout:
386 * There are three possibilities for how page->flags get
387 * laid out. The first is for the normal case, without
388 * sparsemem. The second is for sparsemem when there is
389 * plenty of space for node and section. The last is when
390 * we have run out of space and have to fall back to an
391 * alternate (slower) way of determining the node.
393 * No sparsemem: | NODE | ZONE | ... | FLAGS |
394 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
395 * no space for node: | SECTION | ZONE | ... | FLAGS |
397 #ifdef CONFIG_SPARSEMEM
398 #define SECTIONS_WIDTH SECTIONS_SHIFT
400 #define SECTIONS_WIDTH 0
403 #define ZONES_WIDTH ZONES_SHIFT
405 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
406 #define NODES_WIDTH NODES_SHIFT
408 #define NODES_WIDTH 0
411 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
412 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
413 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
414 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
417 * We are going to use the flags for the page to node mapping if its in
418 * there. This includes the case where there is no node, so it is implicit.
420 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
421 #define NODE_NOT_IN_PAGE_FLAGS
424 #ifndef PFN_SECTION_SHIFT
425 #define PFN_SECTION_SHIFT 0
429 * Define the bit shifts to access each section. For non-existant
430 * sections we define the shift as 0; that plus a 0 mask ensures
431 * the compiler will optimise away reference to them.
433 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
434 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
435 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
437 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
438 #ifdef NODE_NOT_IN_PAGEFLAGS
439 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
440 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
441 SECTIONS_PGOFF : ZONES_PGOFF)
443 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
444 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
445 NODES_PGOFF : ZONES_PGOFF)
448 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
450 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
451 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
454 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
455 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
456 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
457 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
459 static inline enum zone_type
page_zonenum(struct page
*page
)
461 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
465 * The identification function is only used by the buddy allocator for
466 * determining if two pages could be buddies. We are not really
467 * identifying a zone since we could be using a the section number
468 * id if we have not node id available in page flags.
469 * We guarantee only that it will return the same value for two
470 * combinable pages in a zone.
472 static inline int page_zone_id(struct page
*page
)
474 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
477 static inline int zone_to_nid(struct zone
*zone
)
486 #ifdef NODE_NOT_IN_PAGE_FLAGS
487 extern int page_to_nid(struct page
*page
);
489 static inline int page_to_nid(struct page
*page
)
491 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
495 static inline struct zone
*page_zone(struct page
*page
)
497 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
500 static inline unsigned long page_to_section(struct page
*page
)
502 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
505 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
507 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
508 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
511 static inline void set_page_node(struct page
*page
, unsigned long node
)
513 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
514 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
517 static inline void set_page_section(struct page
*page
, unsigned long section
)
519 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
520 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
523 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
524 unsigned long node
, unsigned long pfn
)
526 set_page_zone(page
, zone
);
527 set_page_node(page
, node
);
528 set_page_section(page
, pfn_to_section_nr(pfn
));
532 * If a hint addr is less than mmap_min_addr change hint to be as
533 * low as possible but still greater than mmap_min_addr
535 static inline unsigned long round_hint_to_min(unsigned long hint
)
537 #ifdef CONFIG_SECURITY
539 if (((void *)hint
!= NULL
) &&
540 (hint
< mmap_min_addr
))
541 return PAGE_ALIGN(mmap_min_addr
);
547 * Some inline functions in vmstat.h depend on page_zone()
549 #include <linux/vmstat.h>
551 static __always_inline
void *lowmem_page_address(struct page
*page
)
553 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
556 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
557 #define HASHED_PAGE_VIRTUAL
560 #if defined(WANT_PAGE_VIRTUAL)
561 #define page_address(page) ((page)->virtual)
562 #define set_page_address(page, address) \
564 (page)->virtual = (address); \
566 #define page_address_init() do { } while(0)
569 #if defined(HASHED_PAGE_VIRTUAL)
570 void *page_address(struct page
*page
);
571 void set_page_address(struct page
*page
, void *virtual);
572 void page_address_init(void);
575 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
576 #define page_address(page) lowmem_page_address(page)
577 #define set_page_address(page, address) do { } while(0)
578 #define page_address_init() do { } while(0)
582 * On an anonymous page mapped into a user virtual memory area,
583 * page->mapping points to its anon_vma, not to a struct address_space;
584 * with the PAGE_MAPPING_ANON bit set to distinguish it.
586 * Please note that, confusingly, "page_mapping" refers to the inode
587 * address_space which maps the page from disk; whereas "page_mapped"
588 * refers to user virtual address space into which the page is mapped.
590 #define PAGE_MAPPING_ANON 1
592 extern struct address_space swapper_space
;
593 static inline struct address_space
*page_mapping(struct page
*page
)
595 struct address_space
*mapping
= page
->mapping
;
597 VM_BUG_ON(PageSlab(page
));
598 if (unlikely(PageSwapCache(page
)))
599 mapping
= &swapper_space
;
600 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
605 static inline int PageAnon(struct page
*page
)
607 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
611 * Return the pagecache index of the passed page. Regular pagecache pages
612 * use ->index whereas swapcache pages use ->private
614 static inline pgoff_t
page_index(struct page
*page
)
616 if (unlikely(PageSwapCache(page
)))
617 return page_private(page
);
622 * The atomic page->_mapcount, like _count, starts from -1:
623 * so that transitions both from it and to it can be tracked,
624 * using atomic_inc_and_test and atomic_add_negative(-1).
626 static inline void reset_page_mapcount(struct page
*page
)
628 atomic_set(&(page
)->_mapcount
, -1);
631 static inline int page_mapcount(struct page
*page
)
633 return atomic_read(&(page
)->_mapcount
) + 1;
637 * Return true if this page is mapped into pagetables.
639 static inline int page_mapped(struct page
*page
)
641 return atomic_read(&(page
)->_mapcount
) >= 0;
645 * Error return values for the *_nopage functions
647 #define NOPAGE_SIGBUS (NULL)
648 #define NOPAGE_OOM ((struct page *) (-1))
651 * Error return values for the *_nopfn functions
653 #define NOPFN_SIGBUS ((unsigned long) -1)
654 #define NOPFN_OOM ((unsigned long) -2)
655 #define NOPFN_REFAULT ((unsigned long) -3)
658 * Different kinds of faults, as returned by handle_mm_fault().
659 * Used to decide whether a process gets delivered SIGBUS or
660 * just gets major/minor fault counters bumped up.
663 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
665 #define VM_FAULT_OOM 0x0001
666 #define VM_FAULT_SIGBUS 0x0002
667 #define VM_FAULT_MAJOR 0x0004
668 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
670 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
671 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
673 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
675 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
677 extern void show_free_areas(void);
680 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
682 static inline int shmem_lock(struct file
*file
, int lock
,
683 struct user_struct
*user
)
688 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
690 int shmem_zero_setup(struct vm_area_struct
*);
693 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
697 unsigned long flags
);
700 extern int can_do_mlock(void);
701 extern int user_shm_lock(size_t, struct user_struct
*);
702 extern void user_shm_unlock(size_t, struct user_struct
*);
705 * Parameter block passed down to zap_pte_range in exceptional cases.
708 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
709 struct address_space
*check_mapping
; /* Check page->mapping if set */
710 pgoff_t first_index
; /* Lowest page->index to unmap */
711 pgoff_t last_index
; /* Highest page->index to unmap */
712 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
713 unsigned long truncate_count
; /* Compare vm_truncate_count */
716 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
717 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
718 unsigned long size
, struct zap_details
*);
719 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
720 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
721 unsigned long end_addr
, unsigned long *nr_accounted
,
722 struct zap_details
*);
725 * mm_walk - callbacks for walk_page_range
726 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
727 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
728 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
729 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
730 * @pte_hole: if set, called for each hole at all levels
732 * (see walk_page_range for more details)
735 int (*pgd_entry
)(pgd_t
*, unsigned long, unsigned long, void *);
736 int (*pud_entry
)(pud_t
*, unsigned long, unsigned long, void *);
737 int (*pmd_entry
)(pmd_t
*, unsigned long, unsigned long, void *);
738 int (*pte_entry
)(pte_t
*, unsigned long, unsigned long, void *);
739 int (*pte_hole
)(unsigned long, unsigned long, void *);
742 int walk_page_range(const struct mm_struct
*, unsigned long addr
,
743 unsigned long end
, const struct mm_walk
*walk
,
745 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
746 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
747 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
748 unsigned long floor
, unsigned long ceiling
);
749 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
750 struct vm_area_struct
*vma
);
751 void unmap_mapping_range(struct address_space
*mapping
,
752 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
754 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
755 loff_t
const holebegin
, loff_t
const holelen
)
757 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
760 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
761 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
764 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
765 unsigned long address
, int write_access
);
767 static inline int handle_mm_fault(struct mm_struct
*mm
,
768 struct vm_area_struct
*vma
, unsigned long address
,
771 /* should never happen if there's no MMU */
773 return VM_FAULT_SIGBUS
;
777 extern int make_pages_present(unsigned long addr
, unsigned long end
);
778 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
780 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
781 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
782 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
784 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
785 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
787 int __set_page_dirty_nobuffers(struct page
*page
);
788 int __set_page_dirty_no_writeback(struct page
*page
);
789 int redirty_page_for_writepage(struct writeback_control
*wbc
,
791 int set_page_dirty(struct page
*page
);
792 int set_page_dirty_lock(struct page
*page
);
793 int clear_page_dirty_for_io(struct page
*page
);
795 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
796 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
797 unsigned long new_addr
, unsigned long len
);
798 extern unsigned long do_mremap(unsigned long addr
,
799 unsigned long old_len
, unsigned long new_len
,
800 unsigned long flags
, unsigned long new_addr
);
801 extern int mprotect_fixup(struct vm_area_struct
*vma
,
802 struct vm_area_struct
**pprev
, unsigned long start
,
803 unsigned long end
, unsigned long newflags
);
806 * A callback you can register to apply pressure to ageable caches.
808 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
809 * look through the least-recently-used 'nr_to_scan' entries and
810 * attempt to free them up. It should return the number of objects
811 * which remain in the cache. If it returns -1, it means it cannot do
812 * any scanning at this time (eg. there is a risk of deadlock).
814 * The 'gfpmask' refers to the allocation we are currently trying to
817 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
818 * querying the cache size, so a fastpath for that case is appropriate.
821 int (*shrink
)(int nr_to_scan
, gfp_t gfp_mask
);
822 int seeks
; /* seeks to recreate an obj */
824 /* These are for internal use */
825 struct list_head list
;
826 long nr
; /* objs pending delete */
828 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
829 extern void register_shrinker(struct shrinker
*);
830 extern void unregister_shrinker(struct shrinker
*);
832 int vma_wants_writenotify(struct vm_area_struct
*vma
);
834 extern pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
);
836 #ifdef __PAGETABLE_PUD_FOLDED
837 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
838 unsigned long address
)
843 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
846 #ifdef __PAGETABLE_PMD_FOLDED
847 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
848 unsigned long address
)
853 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
856 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
857 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
860 * The following ifdef needed to get the 4level-fixup.h header to work.
861 * Remove it when 4level-fixup.h has been removed.
863 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
864 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
866 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
867 NULL
: pud_offset(pgd
, address
);
870 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
872 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
873 NULL
: pmd_offset(pud
, address
);
875 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
877 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
879 * We tuck a spinlock to guard each pagetable page into its struct page,
880 * at page->private, with BUILD_BUG_ON to make sure that this will not
881 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
882 * When freeing, reset page->mapping so free_pages_check won't complain.
884 #define __pte_lockptr(page) &((page)->ptl)
885 #define pte_lock_init(_page) do { \
886 spin_lock_init(__pte_lockptr(_page)); \
888 #define pte_lock_deinit(page) ((page)->mapping = NULL)
889 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
892 * We use mm->page_table_lock to guard all pagetable pages of the mm.
894 #define pte_lock_init(page) do {} while (0)
895 #define pte_lock_deinit(page) do {} while (0)
896 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
897 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
899 static inline void pgtable_page_ctor(struct page
*page
)
902 inc_zone_page_state(page
, NR_PAGETABLE
);
905 static inline void pgtable_page_dtor(struct page
*page
)
907 pte_lock_deinit(page
);
908 dec_zone_page_state(page
, NR_PAGETABLE
);
911 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
913 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
914 pte_t *__pte = pte_offset_map(pmd, address); \
920 #define pte_unmap_unlock(pte, ptl) do { \
925 #define pte_alloc_map(mm, pmd, address) \
926 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
927 NULL: pte_offset_map(pmd, address))
929 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
930 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
931 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
933 #define pte_alloc_kernel(pmd, address) \
934 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
935 NULL: pte_offset_kernel(pmd, address))
937 extern void free_area_init(unsigned long * zones_size
);
938 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
939 unsigned long * zones_size
, unsigned long zone_start_pfn
,
940 unsigned long *zholes_size
);
941 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
943 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
944 * zones, allocate the backing mem_map and account for memory holes in a more
945 * architecture independent manner. This is a substitute for creating the
946 * zone_sizes[] and zholes_size[] arrays and passing them to
947 * free_area_init_node()
949 * An architecture is expected to register range of page frames backed by
950 * physical memory with add_active_range() before calling
951 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
952 * usage, an architecture is expected to do something like
954 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
956 * for_each_valid_physical_page_range()
957 * add_active_range(node_id, start_pfn, end_pfn)
958 * free_area_init_nodes(max_zone_pfns);
960 * If the architecture guarantees that there are no holes in the ranges
961 * registered with add_active_range(), free_bootmem_active_regions()
962 * will call free_bootmem_node() for each registered physical page range.
963 * Similarly sparse_memory_present_with_active_regions() calls
964 * memory_present() for each range when SPARSEMEM is enabled.
966 * See mm/page_alloc.c for more information on each function exposed by
967 * CONFIG_ARCH_POPULATES_NODE_MAP
969 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
970 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
971 unsigned long end_pfn
);
972 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
973 unsigned long new_end_pfn
);
974 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
975 unsigned long end_pfn
);
976 extern void remove_all_active_ranges(void);
977 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
978 unsigned long end_pfn
);
979 extern void get_pfn_range_for_nid(unsigned int nid
,
980 unsigned long *start_pfn
, unsigned long *end_pfn
);
981 extern unsigned long find_min_pfn_with_active_regions(void);
982 extern unsigned long find_max_pfn_with_active_regions(void);
983 extern void free_bootmem_with_active_regions(int nid
,
984 unsigned long max_low_pfn
);
985 extern void sparse_memory_present_with_active_regions(int nid
);
986 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
987 extern int early_pfn_to_nid(unsigned long pfn
);
988 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
989 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
990 extern void set_dma_reserve(unsigned long new_dma_reserve
);
991 extern void memmap_init_zone(unsigned long, int, unsigned long,
992 unsigned long, enum memmap_context
);
993 extern void setup_per_zone_pages_min(void);
994 extern void mem_init(void);
995 extern void show_mem(void);
996 extern void si_meminfo(struct sysinfo
* val
);
997 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1000 extern void setup_per_cpu_pageset(void);
1002 static inline void setup_per_cpu_pageset(void) {}
1006 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1007 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1008 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1009 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1010 struct prio_tree_iter
*iter
);
1012 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1013 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1014 (vma = vma_prio_tree_next(vma, iter)); )
1016 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1017 struct list_head
*list
)
1019 vma
->shared
.vm_set
.parent
= NULL
;
1020 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1024 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1025 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1026 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1027 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1028 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1029 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1030 struct mempolicy
*);
1031 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1032 extern int split_vma(struct mm_struct
*,
1033 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1034 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1035 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1036 struct rb_node
**, struct rb_node
*);
1037 extern void unlink_file_vma(struct vm_area_struct
*);
1038 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1039 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1040 extern void exit_mmap(struct mm_struct
*);
1041 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1042 extern int install_special_mapping(struct mm_struct
*mm
,
1043 unsigned long addr
, unsigned long len
,
1044 unsigned long flags
, struct page
**pages
);
1046 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1048 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1049 unsigned long len
, unsigned long prot
,
1050 unsigned long flag
, unsigned long pgoff
);
1051 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1052 unsigned long len
, unsigned long flags
,
1053 unsigned int vm_flags
, unsigned long pgoff
,
1056 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1057 unsigned long len
, unsigned long prot
,
1058 unsigned long flag
, unsigned long offset
)
1060 unsigned long ret
= -EINVAL
;
1061 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1063 if (!(offset
& ~PAGE_MASK
))
1064 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1069 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1071 extern unsigned long do_brk(unsigned long, unsigned long);
1074 extern unsigned long page_unuse(struct page
*);
1075 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1076 extern void truncate_inode_pages_range(struct address_space
*,
1077 loff_t lstart
, loff_t lend
);
1079 /* generic vm_area_ops exported for stackable file systems */
1080 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1082 /* mm/page-writeback.c */
1083 int write_one_page(struct page
*page
, int wait
);
1086 #define VM_MAX_READAHEAD 128 /* kbytes */
1087 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1089 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1090 pgoff_t offset
, unsigned long nr_to_read
);
1091 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1092 pgoff_t offset
, unsigned long nr_to_read
);
1094 void page_cache_sync_readahead(struct address_space
*mapping
,
1095 struct file_ra_state
*ra
,
1098 unsigned long size
);
1100 void page_cache_async_readahead(struct address_space
*mapping
,
1101 struct file_ra_state
*ra
,
1105 unsigned long size
);
1107 unsigned long max_sane_readahead(unsigned long nr
);
1109 /* Do stack extension */
1110 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1112 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1114 extern int expand_stack_downwards(struct vm_area_struct
*vma
,
1115 unsigned long address
);
1117 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1118 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1119 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1120 struct vm_area_struct
**pprev
);
1122 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1123 NULL if none. Assume start_addr < end_addr. */
1124 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1126 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1128 if (vma
&& end_addr
<= vma
->vm_start
)
1133 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1135 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1138 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1139 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1140 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1141 unsigned long pfn
, unsigned long size
, pgprot_t
);
1142 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1143 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1146 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1147 unsigned int foll_flags
);
1148 #define FOLL_WRITE 0x01 /* check pte is writable */
1149 #define FOLL_TOUCH 0x02 /* mark page accessed */
1150 #define FOLL_GET 0x04 /* do get_page on page */
1151 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1153 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1155 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1156 unsigned long size
, pte_fn_t fn
, void *data
);
1158 #ifdef CONFIG_PROC_FS
1159 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1161 static inline void vm_stat_account(struct mm_struct
*mm
,
1162 unsigned long flags
, struct file
*file
, long pages
)
1165 #endif /* CONFIG_PROC_FS */
1167 #ifdef CONFIG_DEBUG_PAGEALLOC
1168 extern int debug_pagealloc_enabled
;
1170 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1172 static inline void enable_debug_pagealloc(void)
1174 debug_pagealloc_enabled
= 1;
1176 #ifdef CONFIG_HIBERNATION
1177 extern bool kernel_page_present(struct page
*page
);
1178 #endif /* CONFIG_HIBERNATION */
1181 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1182 static inline void enable_debug_pagealloc(void)
1185 #ifdef CONFIG_HIBERNATION
1186 static inline bool kernel_page_present(struct page
*page
) { return true; }
1187 #endif /* CONFIG_HIBERNATION */
1190 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1191 #ifdef __HAVE_ARCH_GATE_AREA
1192 int in_gate_area_no_task(unsigned long addr
);
1193 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1195 int in_gate_area_no_task(unsigned long addr
);
1196 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1197 #endif /* __HAVE_ARCH_GATE_AREA */
1199 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1200 void __user
*, size_t *, loff_t
*);
1201 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1202 unsigned long lru_pages
);
1203 void drop_pagecache(void);
1204 void drop_slab(void);
1207 #define randomize_va_space 0
1209 extern int randomize_va_space
;
1212 const char * arch_vma_name(struct vm_area_struct
*vma
);
1213 void print_vma_addr(char *prefix
, unsigned long rip
);
1215 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1216 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1217 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1218 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1219 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1220 void *vmemmap_alloc_block(unsigned long size
, int node
);
1221 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1222 int vmemmap_populate_basepages(struct page
*start_page
,
1223 unsigned long pages
, int node
);
1224 int vmemmap_populate(struct page
*start_page
, unsigned long pages
, int node
);
1226 #endif /* __KERNEL__ */
1227 #endif /* _LINUX_MM_H */