4 #include <linux/sched.h>
5 #include <linux/errno.h>
6 #include <linux/capability.h>
10 #include <linux/config.h>
11 #include <linux/gfp.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/prio_tree.h>
17 #include <linux/mutex.h>
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 unsigned long vmalloc_earlyreserve
;
29 extern int page_cluster
;
32 extern int sysctl_legacy_va_layout
;
34 #define sysctl_legacy_va_layout 0
38 #include <asm/pgtable.h>
39 #include <asm/processor.h>
40 #include <asm/atomic.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
54 * This struct defines a memory VMM memory area. There is one of these
55 * per VM-area/task. A VM area is any part of the process virtual memory
56 * space that has a special rule for the page-fault handlers (ie a shared
57 * library, the executable area etc).
59 struct vm_area_struct
{
60 struct mm_struct
* vm_mm
; /* The address space we belong to. */
61 unsigned long vm_start
; /* Our start address within vm_mm. */
62 unsigned long vm_end
; /* The first byte after our end address
65 /* linked list of VM areas per task, sorted by address */
66 struct vm_area_struct
*vm_next
;
68 pgprot_t vm_page_prot
; /* Access permissions of this VMA. */
69 unsigned long vm_flags
; /* Flags, listed below. */
74 * For areas with an address space and backing store,
75 * linkage into the address_space->i_mmap prio tree, or
76 * linkage to the list of like vmas hanging off its node, or
77 * linkage of vma in the address_space->i_mmap_nonlinear list.
81 struct list_head list
;
82 void *parent
; /* aligns with prio_tree_node parent */
83 struct vm_area_struct
*head
;
86 struct raw_prio_tree_node prio_tree_node
;
90 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
91 * list, after a COW of one of the file pages. A MAP_SHARED vma
92 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
93 * or brk vma (with NULL file) can only be in an anon_vma list.
95 struct list_head anon_vma_node
; /* Serialized by anon_vma->lock */
96 struct anon_vma
*anon_vma
; /* Serialized by page_table_lock */
98 /* Function pointers to deal with this struct. */
99 struct vm_operations_struct
* vm_ops
;
101 /* Information about our backing store: */
102 unsigned long vm_pgoff
; /* Offset (within vm_file) in PAGE_SIZE
103 units, *not* PAGE_CACHE_SIZE */
104 struct file
* vm_file
; /* File we map to (can be NULL). */
105 void * vm_private_data
; /* was vm_pte (shared mem) */
106 unsigned long vm_truncate_count
;/* truncate_count or restart_addr */
109 atomic_t vm_usage
; /* refcount (VMAs shared if !MMU) */
112 struct mempolicy
*vm_policy
; /* NUMA policy for the VMA */
117 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
118 * disabled, then there's a single shared list of VMAs maintained by the
119 * system, and mm's subscribe to these individually
121 struct vm_list_struct
{
122 struct vm_list_struct
*next
;
123 struct vm_area_struct
*vma
;
127 extern struct rb_root nommu_vma_tree
;
128 extern struct rw_semaphore nommu_vma_sem
;
130 extern unsigned int kobjsize(const void *objp
);
136 #define VM_READ 0x00000001 /* currently active flags */
137 #define VM_WRITE 0x00000002
138 #define VM_EXEC 0x00000004
139 #define VM_SHARED 0x00000008
141 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
142 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
143 #define VM_MAYWRITE 0x00000020
144 #define VM_MAYEXEC 0x00000040
145 #define VM_MAYSHARE 0x00000080
147 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
148 #define VM_GROWSUP 0x00000200
149 #define VM_SHM 0x00000000 /* Means nothing: delete it later */
150 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
151 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
153 #define VM_EXECUTABLE 0x00001000
154 #define VM_LOCKED 0x00002000
155 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
157 /* Used by sys_madvise() */
158 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
161 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
163 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
164 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
168 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
170 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
171 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
174 #ifdef CONFIG_STACK_GROWSUP
175 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
177 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
181 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
182 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
183 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
184 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
187 * mapping from the currently active vm_flags protection bits (the
188 * low four bits) to a page protection mask..
190 extern pgprot_t protection_map
[16];
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
198 struct vm_operations_struct
{
199 void (*open
)(struct vm_area_struct
* area
);
200 void (*close
)(struct vm_area_struct
* area
);
201 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
202 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
204 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
205 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
214 * Each physical page in the system has a struct page associated with
215 * it to keep track of whatever it is we are using the page for at the
216 * moment. Note that we have no way to track which tasks are using
220 unsigned long flags
; /* Atomic flags, some possibly
221 * updated asynchronously */
222 atomic_t _count
; /* Usage count, see below. */
223 atomic_t _mapcount
; /* Count of ptes mapped in mms,
224 * to show when page is mapped
225 * & limit reverse map searches.
229 unsigned long private; /* Mapping-private opaque data:
230 * usually used for buffer_heads
231 * if PagePrivate set; used for
232 * swp_entry_t if PageSwapCache.
233 * When page is free, this
234 * indicates order in the buddy
237 struct address_space
*mapping
; /* If low bit clear, points to
238 * inode address_space, or NULL.
239 * If page mapped as anonymous
240 * memory, low bit is set, and
241 * it points to anon_vma object:
242 * see PAGE_MAPPING_ANON below.
245 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
249 pgoff_t index
; /* Our offset within mapping. */
250 struct list_head lru
; /* Pageout list, eg. active_list
251 * protected by zone->lru_lock !
254 * On machines where all RAM is mapped into kernel address space,
255 * we can simply calculate the virtual address. On machines with
256 * highmem some memory is mapped into kernel virtual memory
257 * dynamically, so we need a place to store that address.
258 * Note that this field could be 16 bits on x86 ... ;)
260 * Architectures with slow multiplication can define
261 * WANT_PAGE_VIRTUAL in asm/page.h
263 #if defined(WANT_PAGE_VIRTUAL)
264 void *virtual; /* Kernel virtual address (NULL if
265 not kmapped, ie. highmem) */
266 #endif /* WANT_PAGE_VIRTUAL */
269 #define page_private(page) ((page)->private)
270 #define set_page_private(page, v) ((page)->private = (v))
273 * FIXME: take this include out, include page-flags.h in
274 * files which need it (119 of them)
276 #include <linux/page-flags.h>
279 * Methods to modify the page usage count.
281 * What counts for a page usage:
282 * - cache mapping (page->mapping)
283 * - private data (page->private)
284 * - page mapped in a task's page tables, each mapping
285 * is counted separately
287 * Also, many kernel routines increase the page count before a critical
288 * routine so they can be sure the page doesn't go away from under them.
292 * Drop a ref, return true if the logical refcount fell to zero (the page has
295 static inline int put_page_testzero(struct page
*page
)
297 BUG_ON(atomic_read(&page
->_count
) == 0);
298 return atomic_dec_and_test(&page
->_count
);
302 * Try to grab a ref unless the page has a refcount of zero, return false if
305 static inline int get_page_unless_zero(struct page
*page
)
307 return atomic_inc_not_zero(&page
->_count
);
310 extern void FASTCALL(__page_cache_release(struct page
*));
312 static inline int page_count(struct page
*page
)
314 if (PageCompound(page
))
315 page
= (struct page
*)page_private(page
);
316 return atomic_read(&page
->_count
);
319 static inline void get_page(struct page
*page
)
321 if (unlikely(PageCompound(page
)))
322 page
= (struct page
*)page_private(page
);
323 atomic_inc(&page
->_count
);
327 * Setup the page count before being freed into the page allocator for
328 * the first time (boot or memory hotplug)
330 static inline void init_page_count(struct page
*page
)
332 atomic_set(&page
->_count
, 1);
335 void put_page(struct page
*page
);
337 void split_page(struct page
*page
, unsigned int order
);
340 * Multiple processes may "see" the same page. E.g. for untouched
341 * mappings of /dev/null, all processes see the same page full of
342 * zeroes, and text pages of executables and shared libraries have
343 * only one copy in memory, at most, normally.
345 * For the non-reserved pages, page_count(page) denotes a reference count.
346 * page_count() == 0 means the page is free. page->lru is then used for
347 * freelist management in the buddy allocator.
348 * page_count() == 1 means the page is used for exactly one purpose
349 * (e.g. a private data page of one process).
351 * A page may be used for kmalloc() or anyone else who does a
352 * __get_free_page(). In this case the page_count() is at least 1, and
353 * all other fields are unused but should be 0 or NULL. The
354 * management of this page is the responsibility of the one who uses
357 * The other pages (we may call them "process pages") are completely
358 * managed by the Linux memory manager: I/O, buffers, swapping etc.
359 * The following discussion applies only to them.
361 * A page may belong to an inode's memory mapping. In this case,
362 * page->mapping is the pointer to the inode, and page->index is the
363 * file offset of the page, in units of PAGE_CACHE_SIZE.
365 * A page contains an opaque `private' member, which belongs to the
366 * page's address_space. Usually, this is the address of a circular
367 * list of the page's disk buffers.
369 * For pages belonging to inodes, the page_count() is the number of
370 * attaches, plus 1 if `private' contains something, plus one for
371 * the page cache itself.
373 * Instead of keeping dirty/clean pages in per address-space lists, we instead
374 * now tag pages as dirty/under writeback in the radix tree.
376 * There is also a per-mapping radix tree mapping index to the page
377 * in memory if present. The tree is rooted at mapping->root.
379 * All process pages can do I/O:
380 * - inode pages may need to be read from disk,
381 * - inode pages which have been modified and are MAP_SHARED may need
382 * to be written to disk,
383 * - private pages which have been modified may need to be swapped out
384 * to swap space and (later) to be read back into memory.
388 * The zone field is never updated after free_area_init_core()
389 * sets it, so none of the operations on it need to be atomic.
394 * page->flags layout:
396 * There are three possibilities for how page->flags get
397 * laid out. The first is for the normal case, without
398 * sparsemem. The second is for sparsemem when there is
399 * plenty of space for node and section. The last is when
400 * we have run out of space and have to fall back to an
401 * alternate (slower) way of determining the node.
403 * No sparsemem: | NODE | ZONE | ... | FLAGS |
404 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
405 * no space for node: | SECTION | ZONE | ... | FLAGS |
407 #ifdef CONFIG_SPARSEMEM
408 #define SECTIONS_WIDTH SECTIONS_SHIFT
410 #define SECTIONS_WIDTH 0
413 #define ZONES_WIDTH ZONES_SHIFT
415 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
416 #define NODES_WIDTH NODES_SHIFT
418 #define NODES_WIDTH 0
421 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
422 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
423 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
424 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
427 * We are going to use the flags for the page to node mapping if its in
428 * there. This includes the case where there is no node, so it is implicit.
430 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
432 #ifndef PFN_SECTION_SHIFT
433 #define PFN_SECTION_SHIFT 0
437 * Define the bit shifts to access each section. For non-existant
438 * sections we define the shift as 0; that plus a 0 mask ensures
439 * the compiler will optimise away reference to them.
441 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
442 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
443 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
445 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
447 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
449 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
451 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
453 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
454 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
457 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
458 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
459 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
460 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
462 static inline unsigned long page_zonenum(struct page
*page
)
464 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
468 extern struct zone
*zone_table
[];
470 static inline struct zone
*page_zone(struct page
*page
)
472 return zone_table
[(page
->flags
>> ZONETABLE_PGSHIFT
) &
476 static inline unsigned long page_to_nid(struct page
*page
)
479 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
481 return page_zone(page
)->zone_pgdat
->node_id
;
483 static inline unsigned long page_to_section(struct page
*page
)
485 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
488 static inline void set_page_zone(struct page
*page
, unsigned long zone
)
490 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
491 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
493 static inline void set_page_node(struct page
*page
, unsigned long node
)
495 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
496 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
498 static inline void set_page_section(struct page
*page
, unsigned long section
)
500 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
501 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
504 static inline void set_page_links(struct page
*page
, unsigned long zone
,
505 unsigned long node
, unsigned long pfn
)
507 set_page_zone(page
, zone
);
508 set_page_node(page
, node
);
509 set_page_section(page
, pfn_to_section_nr(pfn
));
512 #ifndef CONFIG_DISCONTIGMEM
513 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
514 extern struct page
*mem_map
;
517 static __always_inline
void *lowmem_page_address(struct page
*page
)
519 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
522 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
523 #define HASHED_PAGE_VIRTUAL
526 #if defined(WANT_PAGE_VIRTUAL)
527 #define page_address(page) ((page)->virtual)
528 #define set_page_address(page, address) \
530 (page)->virtual = (address); \
532 #define page_address_init() do { } while(0)
535 #if defined(HASHED_PAGE_VIRTUAL)
536 void *page_address(struct page
*page
);
537 void set_page_address(struct page
*page
, void *virtual);
538 void page_address_init(void);
541 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
542 #define page_address(page) lowmem_page_address(page)
543 #define set_page_address(page, address) do { } while(0)
544 #define page_address_init() do { } while(0)
548 * On an anonymous page mapped into a user virtual memory area,
549 * page->mapping points to its anon_vma, not to a struct address_space;
550 * with the PAGE_MAPPING_ANON bit set to distinguish it.
552 * Please note that, confusingly, "page_mapping" refers to the inode
553 * address_space which maps the page from disk; whereas "page_mapped"
554 * refers to user virtual address space into which the page is mapped.
556 #define PAGE_MAPPING_ANON 1
558 extern struct address_space swapper_space
;
559 static inline struct address_space
*page_mapping(struct page
*page
)
561 struct address_space
*mapping
= page
->mapping
;
563 if (unlikely(PageSwapCache(page
)))
564 mapping
= &swapper_space
;
565 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
570 static inline int PageAnon(struct page
*page
)
572 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
576 * Return the pagecache index of the passed page. Regular pagecache pages
577 * use ->index whereas swapcache pages use ->private
579 static inline pgoff_t
page_index(struct page
*page
)
581 if (unlikely(PageSwapCache(page
)))
582 return page_private(page
);
587 * The atomic page->_mapcount, like _count, starts from -1:
588 * so that transitions both from it and to it can be tracked,
589 * using atomic_inc_and_test and atomic_add_negative(-1).
591 static inline void reset_page_mapcount(struct page
*page
)
593 atomic_set(&(page
)->_mapcount
, -1);
596 static inline int page_mapcount(struct page
*page
)
598 return atomic_read(&(page
)->_mapcount
) + 1;
602 * Return true if this page is mapped into pagetables.
604 static inline int page_mapped(struct page
*page
)
606 return atomic_read(&(page
)->_mapcount
) >= 0;
610 * Error return values for the *_nopage functions
612 #define NOPAGE_SIGBUS (NULL)
613 #define NOPAGE_OOM ((struct page *) (-1))
616 * Different kinds of faults, as returned by handle_mm_fault().
617 * Used to decide whether a process gets delivered SIGBUS or
618 * just gets major/minor fault counters bumped up.
620 #define VM_FAULT_OOM 0x00
621 #define VM_FAULT_SIGBUS 0x01
622 #define VM_FAULT_MINOR 0x02
623 #define VM_FAULT_MAJOR 0x03
626 * Special case for get_user_pages.
627 * Must be in a distinct bit from the above VM_FAULT_ flags.
629 #define VM_FAULT_WRITE 0x10
631 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
633 extern void show_free_areas(void);
636 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
637 unsigned long address
, int *type
);
638 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
639 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
641 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
643 #define shmem_nopage filemap_nopage
645 static inline int shmem_lock(struct file
*file
, int lock
,
646 struct user_struct
*user
)
651 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
652 struct mempolicy
*new)
657 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
663 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
664 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
666 int shmem_zero_setup(struct vm_area_struct
*);
669 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
673 unsigned long flags
);
676 static inline int can_do_mlock(void)
678 if (capable(CAP_IPC_LOCK
))
680 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
684 extern int user_shm_lock(size_t, struct user_struct
*);
685 extern void user_shm_unlock(size_t, struct user_struct
*);
688 * Parameter block passed down to zap_pte_range in exceptional cases.
691 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
692 struct address_space
*check_mapping
; /* Check page->mapping if set */
693 pgoff_t first_index
; /* Lowest page->index to unmap */
694 pgoff_t last_index
; /* Highest page->index to unmap */
695 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
696 unsigned long truncate_count
; /* Compare vm_truncate_count */
699 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
700 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
701 unsigned long size
, struct zap_details
*);
702 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
703 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
704 unsigned long end_addr
, unsigned long *nr_accounted
,
705 struct zap_details
*);
706 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
707 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
708 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
709 unsigned long floor
, unsigned long ceiling
);
710 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
711 struct vm_area_struct
*vma
);
712 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
713 unsigned long size
, pgprot_t prot
);
714 void unmap_mapping_range(struct address_space
*mapping
,
715 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
717 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
718 loff_t
const holebegin
, loff_t
const holelen
)
720 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
723 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
724 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
725 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
726 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
729 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
730 unsigned long address
, int write_access
);
732 static inline int handle_mm_fault(struct mm_struct
*mm
,
733 struct vm_area_struct
*vma
, unsigned long address
,
736 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
740 static inline int handle_mm_fault(struct mm_struct
*mm
,
741 struct vm_area_struct
*vma
, unsigned long address
,
744 /* should never happen if there's no MMU */
746 return VM_FAULT_SIGBUS
;
750 extern int make_pages_present(unsigned long addr
, unsigned long end
);
751 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
752 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
754 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
755 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
756 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
758 int __set_page_dirty_buffers(struct page
*page
);
759 int __set_page_dirty_nobuffers(struct page
*page
);
760 int redirty_page_for_writepage(struct writeback_control
*wbc
,
762 int FASTCALL(set_page_dirty(struct page
*page
));
763 int set_page_dirty_lock(struct page
*page
);
764 int clear_page_dirty_for_io(struct page
*page
);
766 extern unsigned long do_mremap(unsigned long addr
,
767 unsigned long old_len
, unsigned long new_len
,
768 unsigned long flags
, unsigned long new_addr
);
771 * Prototype to add a shrinker callback for ageable caches.
773 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
774 * scan `nr_to_scan' objects, attempting to free them.
776 * The callback must return the number of objects which remain in the cache.
778 * The callback will be passed nr_to_scan == 0 when the VM is querying the
779 * cache size, so a fastpath for that case is appropriate.
781 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
784 * Add an aging callback. The int is the number of 'seeks' it takes
785 * to recreate one of the objects that these functions age.
788 #define DEFAULT_SEEKS 2
790 extern struct shrinker
*set_shrinker(int, shrinker_t
);
791 extern void remove_shrinker(struct shrinker
*shrinker
);
793 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
795 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
796 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
797 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
798 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
801 * The following ifdef needed to get the 4level-fixup.h header to work.
802 * Remove it when 4level-fixup.h has been removed.
804 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
805 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
807 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
808 NULL
: pud_offset(pgd
, address
);
811 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
813 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
814 NULL
: pmd_offset(pud
, address
);
816 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
818 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
820 * We tuck a spinlock to guard each pagetable page into its struct page,
821 * at page->private, with BUILD_BUG_ON to make sure that this will not
822 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
823 * When freeing, reset page->mapping so free_pages_check won't complain.
825 #define __pte_lockptr(page) &((page)->ptl)
826 #define pte_lock_init(_page) do { \
827 spin_lock_init(__pte_lockptr(_page)); \
829 #define pte_lock_deinit(page) ((page)->mapping = NULL)
830 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
833 * We use mm->page_table_lock to guard all pagetable pages of the mm.
835 #define pte_lock_init(page) do {} while (0)
836 #define pte_lock_deinit(page) do {} while (0)
837 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
838 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
840 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
842 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
843 pte_t *__pte = pte_offset_map(pmd, address); \
849 #define pte_unmap_unlock(pte, ptl) do { \
854 #define pte_alloc_map(mm, pmd, address) \
855 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
856 NULL: pte_offset_map(pmd, address))
858 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
859 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
860 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
862 #define pte_alloc_kernel(pmd, address) \
863 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
864 NULL: pte_offset_kernel(pmd, address))
866 extern void free_area_init(unsigned long * zones_size
);
867 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
868 unsigned long * zones_size
, unsigned long zone_start_pfn
,
869 unsigned long *zholes_size
);
870 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
871 extern void setup_per_zone_pages_min(void);
872 extern void mem_init(void);
873 extern void show_mem(void);
874 extern void si_meminfo(struct sysinfo
* val
);
875 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
878 extern void setup_per_cpu_pageset(void);
880 static inline void setup_per_cpu_pageset(void) {}
884 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
885 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
886 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
887 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
888 struct prio_tree_iter
*iter
);
890 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
891 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
892 (vma = vma_prio_tree_next(vma, iter)); )
894 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
895 struct list_head
*list
)
897 vma
->shared
.vm_set
.parent
= NULL
;
898 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
902 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
903 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
904 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
905 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
906 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
907 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
909 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
910 extern int split_vma(struct mm_struct
*,
911 struct vm_area_struct
*, unsigned long addr
, int new_below
);
912 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
913 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
914 struct rb_node
**, struct rb_node
*);
915 extern void unlink_file_vma(struct vm_area_struct
*);
916 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
917 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
918 extern void exit_mmap(struct mm_struct
*);
919 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
921 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
923 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
924 unsigned long len
, unsigned long prot
,
925 unsigned long flag
, unsigned long pgoff
);
927 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
928 unsigned long len
, unsigned long prot
,
929 unsigned long flag
, unsigned long offset
)
931 unsigned long ret
= -EINVAL
;
932 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
934 if (!(offset
& ~PAGE_MASK
))
935 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
940 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
942 extern unsigned long do_brk(unsigned long, unsigned long);
945 extern unsigned long page_unuse(struct page
*);
946 extern void truncate_inode_pages(struct address_space
*, loff_t
);
947 extern void truncate_inode_pages_range(struct address_space
*,
948 loff_t lstart
, loff_t lend
);
950 /* generic vm_area_ops exported for stackable file systems */
951 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
952 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
953 unsigned long, pgprot_t
, unsigned long, int);
955 /* mm/page-writeback.c */
956 int write_one_page(struct page
*page
, int wait
);
959 #define VM_MAX_READAHEAD 128 /* kbytes */
960 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
961 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
962 * turning readahead off */
964 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
965 pgoff_t offset
, unsigned long nr_to_read
);
966 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
967 pgoff_t offset
, unsigned long nr_to_read
);
968 unsigned long page_cache_readahead(struct address_space
*mapping
,
969 struct file_ra_state
*ra
,
973 void handle_ra_miss(struct address_space
*mapping
,
974 struct file_ra_state
*ra
, pgoff_t offset
);
975 unsigned long max_sane_readahead(unsigned long nr
);
977 /* Do stack extension */
978 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
980 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
983 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
984 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
985 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
986 struct vm_area_struct
**pprev
);
988 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
989 NULL if none. Assume start_addr < end_addr. */
990 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
992 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
994 if (vma
&& end_addr
<= vma
->vm_start
)
999 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1001 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1004 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1005 struct page
*vmalloc_to_page(void *addr
);
1006 unsigned long vmalloc_to_pfn(void *addr
);
1007 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1008 unsigned long pfn
, unsigned long size
, pgprot_t
);
1009 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1011 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1012 unsigned int foll_flags
);
1013 #define FOLL_WRITE 0x01 /* check pte is writable */
1014 #define FOLL_TOUCH 0x02 /* mark page accessed */
1015 #define FOLL_GET 0x04 /* do get_page on page */
1016 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1018 #ifdef CONFIG_PROC_FS
1019 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1021 static inline void vm_stat_account(struct mm_struct
*mm
,
1022 unsigned long flags
, struct file
*file
, long pages
)
1025 #endif /* CONFIG_PROC_FS */
1027 #ifndef CONFIG_DEBUG_PAGEALLOC
1029 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
1031 if (!PageHighMem(page
) && !enable
)
1032 mutex_debug_check_no_locks_freed(page_address(page
),
1033 numpages
* PAGE_SIZE
);
1037 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1038 #ifdef __HAVE_ARCH_GATE_AREA
1039 int in_gate_area_no_task(unsigned long addr
);
1040 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1042 int in_gate_area_no_task(unsigned long addr
);
1043 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1044 #endif /* __HAVE_ARCH_GATE_AREA */
1046 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1047 #define OOM_DISABLE -17
1049 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1050 void __user
*, size_t *, loff_t
*);
1051 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1052 unsigned long lru_pages
);
1053 void drop_pagecache(void);
1054 void drop_slab(void);
1057 #define randomize_va_space 0
1059 extern int randomize_va_space
;
1062 #endif /* __KERNEL__ */
1063 #endif /* _LINUX_MM_H */