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 * indicates order in the buddy
234 * system if PG_buddy is set.
236 struct address_space
*mapping
; /* If low bit clear, points to
237 * inode address_space, or NULL.
238 * If page mapped as anonymous
239 * memory, low bit is set, and
240 * it points to anon_vma object:
241 * see PAGE_MAPPING_ANON below.
244 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
248 pgoff_t index
; /* Our offset within mapping. */
249 struct list_head lru
; /* Pageout list, eg. active_list
250 * protected by zone->lru_lock !
253 * On machines where all RAM is mapped into kernel address space,
254 * we can simply calculate the virtual address. On machines with
255 * highmem some memory is mapped into kernel virtual memory
256 * dynamically, so we need a place to store that address.
257 * Note that this field could be 16 bits on x86 ... ;)
259 * Architectures with slow multiplication can define
260 * WANT_PAGE_VIRTUAL in asm/page.h
262 #if defined(WANT_PAGE_VIRTUAL)
263 void *virtual; /* Kernel virtual address (NULL if
264 not kmapped, ie. highmem) */
265 #endif /* WANT_PAGE_VIRTUAL */
268 #define page_private(page) ((page)->private)
269 #define set_page_private(page, v) ((page)->private = (v))
272 * FIXME: take this include out, include page-flags.h in
273 * files which need it (119 of them)
275 #include <linux/page-flags.h>
278 * Methods to modify the page usage count.
280 * What counts for a page usage:
281 * - cache mapping (page->mapping)
282 * - private data (page->private)
283 * - page mapped in a task's page tables, each mapping
284 * is counted separately
286 * Also, many kernel routines increase the page count before a critical
287 * routine so they can be sure the page doesn't go away from under them.
289 * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we
290 * can use atomic_add_negative(-1, page->_count) to detect when the page
291 * becomes free and so that we can also use atomic_inc_and_test to atomically
292 * detect when we just tried to grab a ref on a page which some other CPU has
293 * already deemed to be freeable.
295 * NO code should make assumptions about this internal detail! Use the provided
296 * macros which retain the old rules: page_count(page) == 0 is a free page.
300 * Drop a ref, return true if the logical refcount fell to zero (the page has
303 #define put_page_testzero(p) \
305 BUG_ON(atomic_read(&(p)->_count) == -1);\
306 atomic_add_negative(-1, &(p)->_count); \
310 * Grab a ref, return true if the page previously had a logical refcount of
311 * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page
313 #define get_page_testone(p) atomic_inc_and_test(&(p)->_count)
315 #define set_page_count(p,v) atomic_set(&(p)->_count, (v) - 1)
316 #define __put_page(p) atomic_dec(&(p)->_count)
318 extern void FASTCALL(__page_cache_release(struct page
*));
320 static inline int page_count(struct page
*page
)
322 if (PageCompound(page
))
323 page
= (struct page
*)page_private(page
);
324 return atomic_read(&page
->_count
) + 1;
327 static inline void get_page(struct page
*page
)
329 if (unlikely(PageCompound(page
)))
330 page
= (struct page
*)page_private(page
);
331 atomic_inc(&page
->_count
);
334 void put_page(struct page
*page
);
337 * Multiple processes may "see" the same page. E.g. for untouched
338 * mappings of /dev/null, all processes see the same page full of
339 * zeroes, and text pages of executables and shared libraries have
340 * only one copy in memory, at most, normally.
342 * For the non-reserved pages, page_count(page) denotes a reference count.
343 * page_count() == 0 means the page is free. page->lru is then used for
344 * freelist management in the buddy allocator.
345 * page_count() == 1 means the page is used for exactly one purpose
346 * (e.g. a private data page of one process).
348 * A page may be used for kmalloc() or anyone else who does a
349 * __get_free_page(). In this case the page_count() is at least 1, and
350 * all other fields are unused but should be 0 or NULL. The
351 * management of this page is the responsibility of the one who uses
354 * The other pages (we may call them "process pages") are completely
355 * managed by the Linux memory manager: I/O, buffers, swapping etc.
356 * The following discussion applies only to them.
358 * A page may belong to an inode's memory mapping. In this case,
359 * page->mapping is the pointer to the inode, and page->index is the
360 * file offset of the page, in units of PAGE_CACHE_SIZE.
362 * A page contains an opaque `private' member, which belongs to the
363 * page's address_space. Usually, this is the address of a circular
364 * list of the page's disk buffers.
366 * For pages belonging to inodes, the page_count() is the number of
367 * attaches, plus 1 if `private' contains something, plus one for
368 * the page cache itself.
370 * Instead of keeping dirty/clean pages in per address-space lists, we instead
371 * now tag pages as dirty/under writeback in the radix tree.
373 * There is also a per-mapping radix tree mapping index to the page
374 * in memory if present. The tree is rooted at mapping->root.
376 * All process pages can do I/O:
377 * - inode pages may need to be read from disk,
378 * - inode pages which have been modified and are MAP_SHARED may need
379 * to be written to disk,
380 * - private pages which have been modified may need to be swapped out
381 * to swap space and (later) to be read back into memory.
385 * The zone field is never updated after free_area_init_core()
386 * sets it, so none of the operations on it need to be atomic.
391 * page->flags layout:
393 * There are three possibilities for how page->flags get
394 * laid out. The first is for the normal case, without
395 * sparsemem. The second is for sparsemem when there is
396 * plenty of space for node and section. The last is when
397 * we have run out of space and have to fall back to an
398 * alternate (slower) way of determining the node.
400 * No sparsemem: | NODE | ZONE | ... | FLAGS |
401 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
402 * no space for node: | SECTION | ZONE | ... | FLAGS |
404 #ifdef CONFIG_SPARSEMEM
405 #define SECTIONS_WIDTH SECTIONS_SHIFT
407 #define SECTIONS_WIDTH 0
410 #define ZONES_WIDTH ZONES_SHIFT
412 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
413 #define NODES_WIDTH NODES_SHIFT
415 #define NODES_WIDTH 0
418 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
419 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
420 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
421 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
424 * We are going to use the flags for the page to node mapping if its in
425 * there. This includes the case where there is no node, so it is implicit.
427 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
429 #ifndef PFN_SECTION_SHIFT
430 #define PFN_SECTION_SHIFT 0
434 * Define the bit shifts to access each section. For non-existant
435 * sections we define the shift as 0; that plus a 0 mask ensures
436 * the compiler will optimise away reference to them.
438 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
439 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
440 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
442 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
444 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
446 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
448 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
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 ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
459 static inline unsigned long page_zonenum(struct page
*page
)
461 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
465 extern struct zone
*zone_table
[];
467 static inline struct zone
*page_zone(struct page
*page
)
469 return zone_table
[(page
->flags
>> ZONETABLE_PGSHIFT
) &
473 static inline unsigned long page_to_nid(struct page
*page
)
476 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
478 return page_zone(page
)->zone_pgdat
->node_id
;
480 static inline unsigned long page_to_section(struct page
*page
)
482 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
485 static inline void set_page_zone(struct page
*page
, unsigned long zone
)
487 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
488 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
490 static inline void set_page_node(struct page
*page
, unsigned long node
)
492 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
493 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
495 static inline void set_page_section(struct page
*page
, unsigned long section
)
497 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
498 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
501 static inline void set_page_links(struct page
*page
, unsigned long zone
,
502 unsigned long node
, unsigned long pfn
)
504 set_page_zone(page
, zone
);
505 set_page_node(page
, node
);
506 set_page_section(page
, pfn_to_section_nr(pfn
));
509 #ifndef CONFIG_DISCONTIGMEM
510 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
511 extern struct page
*mem_map
;
514 static __always_inline
void *lowmem_page_address(struct page
*page
)
516 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
519 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
520 #define HASHED_PAGE_VIRTUAL
523 #if defined(WANT_PAGE_VIRTUAL)
524 #define page_address(page) ((page)->virtual)
525 #define set_page_address(page, address) \
527 (page)->virtual = (address); \
529 #define page_address_init() do { } while(0)
532 #if defined(HASHED_PAGE_VIRTUAL)
533 void *page_address(struct page
*page
);
534 void set_page_address(struct page
*page
, void *virtual);
535 void page_address_init(void);
538 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
539 #define page_address(page) lowmem_page_address(page)
540 #define set_page_address(page, address) do { } while(0)
541 #define page_address_init() do { } while(0)
545 * On an anonymous page mapped into a user virtual memory area,
546 * page->mapping points to its anon_vma, not to a struct address_space;
547 * with the PAGE_MAPPING_ANON bit set to distinguish it.
549 * Please note that, confusingly, "page_mapping" refers to the inode
550 * address_space which maps the page from disk; whereas "page_mapped"
551 * refers to user virtual address space into which the page is mapped.
553 #define PAGE_MAPPING_ANON 1
555 extern struct address_space swapper_space
;
556 static inline struct address_space
*page_mapping(struct page
*page
)
558 struct address_space
*mapping
= page
->mapping
;
560 if (unlikely(PageSwapCache(page
)))
561 mapping
= &swapper_space
;
562 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
567 static inline int PageAnon(struct page
*page
)
569 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
573 * Return the pagecache index of the passed page. Regular pagecache pages
574 * use ->index whereas swapcache pages use ->private
576 static inline pgoff_t
page_index(struct page
*page
)
578 if (unlikely(PageSwapCache(page
)))
579 return page_private(page
);
584 * The atomic page->_mapcount, like _count, starts from -1:
585 * so that transitions both from it and to it can be tracked,
586 * using atomic_inc_and_test and atomic_add_negative(-1).
588 static inline void reset_page_mapcount(struct page
*page
)
590 atomic_set(&(page
)->_mapcount
, -1);
593 static inline int page_mapcount(struct page
*page
)
595 return atomic_read(&(page
)->_mapcount
) + 1;
599 * Return true if this page is mapped into pagetables.
601 static inline int page_mapped(struct page
*page
)
603 return atomic_read(&(page
)->_mapcount
) >= 0;
607 * Error return values for the *_nopage functions
609 #define NOPAGE_SIGBUS (NULL)
610 #define NOPAGE_OOM ((struct page *) (-1))
613 * Different kinds of faults, as returned by handle_mm_fault().
614 * Used to decide whether a process gets delivered SIGBUS or
615 * just gets major/minor fault counters bumped up.
617 #define VM_FAULT_OOM 0x00
618 #define VM_FAULT_SIGBUS 0x01
619 #define VM_FAULT_MINOR 0x02
620 #define VM_FAULT_MAJOR 0x03
623 * Special case for get_user_pages.
624 * Must be in a distinct bit from the above VM_FAULT_ flags.
626 #define VM_FAULT_WRITE 0x10
628 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
630 extern void show_free_areas(void);
633 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
634 unsigned long address
, int *type
);
635 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
636 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
638 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
640 #define shmem_nopage filemap_nopage
642 static inline int shmem_lock(struct file
*file
, int lock
,
643 struct user_struct
*user
)
648 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
649 struct mempolicy
*new)
654 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
660 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
661 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
663 int shmem_zero_setup(struct vm_area_struct
*);
666 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
670 unsigned long flags
);
673 static inline int can_do_mlock(void)
675 if (capable(CAP_IPC_LOCK
))
677 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
681 extern int user_shm_lock(size_t, struct user_struct
*);
682 extern void user_shm_unlock(size_t, struct user_struct
*);
685 * Parameter block passed down to zap_pte_range in exceptional cases.
688 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
689 struct address_space
*check_mapping
; /* Check page->mapping if set */
690 pgoff_t first_index
; /* Lowest page->index to unmap */
691 pgoff_t last_index
; /* Highest page->index to unmap */
692 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
693 unsigned long truncate_count
; /* Compare vm_truncate_count */
696 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
697 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
698 unsigned long size
, struct zap_details
*);
699 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
700 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
701 unsigned long end_addr
, unsigned long *nr_accounted
,
702 struct zap_details
*);
703 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
704 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
705 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
706 unsigned long floor
, unsigned long ceiling
);
707 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
708 struct vm_area_struct
*vma
);
709 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
710 unsigned long size
, pgprot_t prot
);
711 void unmap_mapping_range(struct address_space
*mapping
,
712 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
714 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
715 loff_t
const holebegin
, loff_t
const holelen
)
717 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
720 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
721 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
722 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
723 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
726 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
727 unsigned long address
, int write_access
);
729 static inline int handle_mm_fault(struct mm_struct
*mm
,
730 struct vm_area_struct
*vma
, unsigned long address
,
733 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
737 static inline int handle_mm_fault(struct mm_struct
*mm
,
738 struct vm_area_struct
*vma
, unsigned long address
,
741 /* should never happen if there's no MMU */
743 return VM_FAULT_SIGBUS
;
747 extern int make_pages_present(unsigned long addr
, unsigned long end
);
748 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
749 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
751 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
752 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
753 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
755 int __set_page_dirty_buffers(struct page
*page
);
756 int __set_page_dirty_nobuffers(struct page
*page
);
757 int redirty_page_for_writepage(struct writeback_control
*wbc
,
759 int FASTCALL(set_page_dirty(struct page
*page
));
760 int set_page_dirty_lock(struct page
*page
);
761 int clear_page_dirty_for_io(struct page
*page
);
763 extern unsigned long do_mremap(unsigned long addr
,
764 unsigned long old_len
, unsigned long new_len
,
765 unsigned long flags
, unsigned long new_addr
);
768 * Prototype to add a shrinker callback for ageable caches.
770 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
771 * scan `nr_to_scan' objects, attempting to free them.
773 * The callback must return the number of objects which remain in the cache.
775 * The callback will be passed nr_to_scan == 0 when the VM is querying the
776 * cache size, so a fastpath for that case is appropriate.
778 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
781 * Add an aging callback. The int is the number of 'seeks' it takes
782 * to recreate one of the objects that these functions age.
785 #define DEFAULT_SEEKS 2
787 extern struct shrinker
*set_shrinker(int, shrinker_t
);
788 extern void remove_shrinker(struct shrinker
*shrinker
);
790 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
792 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
793 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
794 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
795 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
798 * The following ifdef needed to get the 4level-fixup.h header to work.
799 * Remove it when 4level-fixup.h has been removed.
801 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
802 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
804 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
805 NULL
: pud_offset(pgd
, address
);
808 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
810 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
811 NULL
: pmd_offset(pud
, address
);
813 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
815 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
817 * We tuck a spinlock to guard each pagetable page into its struct page,
818 * at page->private, with BUILD_BUG_ON to make sure that this will not
819 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
820 * When freeing, reset page->mapping so free_pages_check won't complain.
822 #define __pte_lockptr(page) &((page)->ptl)
823 #define pte_lock_init(_page) do { \
824 spin_lock_init(__pte_lockptr(_page)); \
826 #define pte_lock_deinit(page) ((page)->mapping = NULL)
827 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
830 * We use mm->page_table_lock to guard all pagetable pages of the mm.
832 #define pte_lock_init(page) do {} while (0)
833 #define pte_lock_deinit(page) do {} while (0)
834 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
835 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
837 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
839 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
840 pte_t *__pte = pte_offset_map(pmd, address); \
846 #define pte_unmap_unlock(pte, ptl) do { \
851 #define pte_alloc_map(mm, pmd, address) \
852 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
853 NULL: pte_offset_map(pmd, address))
855 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
856 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
857 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
859 #define pte_alloc_kernel(pmd, address) \
860 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
861 NULL: pte_offset_kernel(pmd, address))
863 extern void free_area_init(unsigned long * zones_size
);
864 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
865 unsigned long * zones_size
, unsigned long zone_start_pfn
,
866 unsigned long *zholes_size
);
867 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
868 extern void setup_per_zone_pages_min(void);
869 extern void mem_init(void);
870 extern void show_mem(void);
871 extern void si_meminfo(struct sysinfo
* val
);
872 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
875 extern void setup_per_cpu_pageset(void);
877 static inline void setup_per_cpu_pageset(void) {}
881 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
882 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
883 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
884 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
885 struct prio_tree_iter
*iter
);
887 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
888 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
889 (vma = vma_prio_tree_next(vma, iter)); )
891 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
892 struct list_head
*list
)
894 vma
->shared
.vm_set
.parent
= NULL
;
895 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
899 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
900 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
901 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
902 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
903 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
904 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
906 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
907 extern int split_vma(struct mm_struct
*,
908 struct vm_area_struct
*, unsigned long addr
, int new_below
);
909 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
910 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
911 struct rb_node
**, struct rb_node
*);
912 extern void unlink_file_vma(struct vm_area_struct
*);
913 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
914 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
915 extern void exit_mmap(struct mm_struct
*);
916 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
918 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
920 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
921 unsigned long len
, unsigned long prot
,
922 unsigned long flag
, unsigned long pgoff
);
924 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
925 unsigned long len
, unsigned long prot
,
926 unsigned long flag
, unsigned long offset
)
928 unsigned long ret
= -EINVAL
;
929 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
931 if (!(offset
& ~PAGE_MASK
))
932 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
937 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
939 extern unsigned long do_brk(unsigned long, unsigned long);
942 extern unsigned long page_unuse(struct page
*);
943 extern void truncate_inode_pages(struct address_space
*, loff_t
);
944 extern void truncate_inode_pages_range(struct address_space
*,
945 loff_t lstart
, loff_t lend
);
947 /* generic vm_area_ops exported for stackable file systems */
948 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
949 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
950 unsigned long, pgprot_t
, unsigned long, int);
952 /* mm/page-writeback.c */
953 int write_one_page(struct page
*page
, int wait
);
956 #define VM_MAX_READAHEAD 128 /* kbytes */
957 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
958 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
959 * turning readahead off */
961 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
962 pgoff_t offset
, unsigned long nr_to_read
);
963 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
964 pgoff_t offset
, unsigned long nr_to_read
);
965 unsigned long page_cache_readahead(struct address_space
*mapping
,
966 struct file_ra_state
*ra
,
970 void handle_ra_miss(struct address_space
*mapping
,
971 struct file_ra_state
*ra
, pgoff_t offset
);
972 unsigned long max_sane_readahead(unsigned long nr
);
974 /* Do stack extension */
975 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
977 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
980 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
981 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
982 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
983 struct vm_area_struct
**pprev
);
985 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
986 NULL if none. Assume start_addr < end_addr. */
987 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
989 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
991 if (vma
&& end_addr
<= vma
->vm_start
)
996 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
998 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1001 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1002 struct page
*vmalloc_to_page(void *addr
);
1003 unsigned long vmalloc_to_pfn(void *addr
);
1004 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1005 unsigned long pfn
, unsigned long size
, pgprot_t
);
1006 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1008 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1009 unsigned int foll_flags
);
1010 #define FOLL_WRITE 0x01 /* check pte is writable */
1011 #define FOLL_TOUCH 0x02 /* mark page accessed */
1012 #define FOLL_GET 0x04 /* do get_page on page */
1013 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1015 #ifdef CONFIG_PROC_FS
1016 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1018 static inline void vm_stat_account(struct mm_struct
*mm
,
1019 unsigned long flags
, struct file
*file
, long pages
)
1022 #endif /* CONFIG_PROC_FS */
1024 #ifndef CONFIG_DEBUG_PAGEALLOC
1026 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
1028 if (!PageHighMem(page
) && !enable
)
1029 mutex_debug_check_no_locks_freed(page_address(page
),
1030 numpages
* PAGE_SIZE
);
1034 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1035 #ifdef __HAVE_ARCH_GATE_AREA
1036 int in_gate_area_no_task(unsigned long addr
);
1037 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1039 int in_gate_area_no_task(unsigned long addr
);
1040 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1041 #endif /* __HAVE_ARCH_GATE_AREA */
1043 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1044 #define OOM_DISABLE -17
1046 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1047 void __user
*, size_t *, loff_t
*);
1048 int shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1049 unsigned long lru_pages
);
1050 void drop_pagecache(void);
1051 void drop_slab(void);
1054 #define randomize_va_space 0
1056 extern int randomize_va_space
;
1059 #endif /* __KERNEL__ */
1060 #endif /* _LINUX_MM_H */