4 #include <linux/sched.h>
5 #include <linux/errno.h>
9 #include <linux/config.h>
10 #include <linux/gfp.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/prio_tree.h>
20 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
21 extern unsigned long max_mapnr
;
24 extern unsigned long num_physpages
;
25 extern void * high_memory
;
26 extern unsigned long vmalloc_earlyreserve
;
27 extern int page_cluster
;
30 extern int sysctl_legacy_va_layout
;
32 #define sysctl_legacy_va_layout 0
36 #include <asm/pgtable.h>
37 #include <asm/processor.h>
38 #include <asm/atomic.h>
40 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
43 * Linux kernel virtual memory manager primitives.
44 * The idea being to have a "virtual" mm in the same way
45 * we have a virtual fs - giving a cleaner interface to the
46 * mm details, and allowing different kinds of memory mappings
47 * (from shared memory to executable loading to arbitrary
52 * This struct defines a memory VMM memory area. There is one of these
53 * per VM-area/task. A VM area is any part of the process virtual memory
54 * space that has a special rule for the page-fault handlers (ie a shared
55 * library, the executable area etc).
57 struct vm_area_struct
{
58 struct mm_struct
* vm_mm
; /* The address space we belong to. */
59 unsigned long vm_start
; /* Our start address within vm_mm. */
60 unsigned long vm_end
; /* The first byte after our end address
63 /* linked list of VM areas per task, sorted by address */
64 struct vm_area_struct
*vm_next
;
66 pgprot_t vm_page_prot
; /* Access permissions of this VMA. */
67 unsigned long vm_flags
; /* Flags, listed below. */
72 * For areas with an address space and backing store,
73 * linkage into the address_space->i_mmap prio tree, or
74 * linkage to the list of like vmas hanging off its node, or
75 * linkage of vma in the address_space->i_mmap_nonlinear list.
79 struct list_head list
;
80 void *parent
; /* aligns with prio_tree_node parent */
81 struct vm_area_struct
*head
;
84 struct raw_prio_tree_node prio_tree_node
;
88 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
89 * list, after a COW of one of the file pages. A MAP_SHARED vma
90 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
91 * or brk vma (with NULL file) can only be in an anon_vma list.
93 struct list_head anon_vma_node
; /* Serialized by anon_vma->lock */
94 struct anon_vma
*anon_vma
; /* Serialized by page_table_lock */
96 /* Function pointers to deal with this struct. */
97 struct vm_operations_struct
* vm_ops
;
99 /* Information about our backing store: */
100 unsigned long vm_pgoff
; /* Offset (within vm_file) in PAGE_SIZE
101 units, *not* PAGE_CACHE_SIZE */
102 struct file
* vm_file
; /* File we map to (can be NULL). */
103 void * vm_private_data
; /* was vm_pte (shared mem) */
104 unsigned long vm_truncate_count
;/* truncate_count or restart_addr */
107 atomic_t vm_usage
; /* refcount (VMAs shared if !MMU) */
110 struct mempolicy
*vm_policy
; /* NUMA policy for the VMA */
115 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
116 * disabled, then there's a single shared list of VMAs maintained by the
117 * system, and mm's subscribe to these individually
119 struct vm_list_struct
{
120 struct vm_list_struct
*next
;
121 struct vm_area_struct
*vma
;
125 extern struct rb_root nommu_vma_tree
;
126 extern struct rw_semaphore nommu_vma_sem
;
128 extern unsigned int kobjsize(const void *objp
);
134 #define VM_READ 0x00000001 /* currently active flags */
135 #define VM_WRITE 0x00000002
136 #define VM_EXEC 0x00000004
137 #define VM_SHARED 0x00000008
139 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
140 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
141 #define VM_MAYWRITE 0x00000020
142 #define VM_MAYEXEC 0x00000040
143 #define VM_MAYSHARE 0x00000080
145 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
146 #define VM_GROWSUP 0x00000200
147 #define VM_SHM 0x00000400 /* shared memory area, don't swap out */
148 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
150 #define VM_EXECUTABLE 0x00001000
151 #define VM_LOCKED 0x00002000
152 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
154 /* Used by sys_madvise() */
155 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
156 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
158 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
159 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
160 #define VM_RESERVED 0x00080000 /* Pages managed in a special way */
161 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
162 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
163 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
164 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
166 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
167 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
170 #ifdef CONFIG_STACK_GROWSUP
171 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
173 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
177 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
178 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
179 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
180 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
183 * mapping from the currently active vm_flags protection bits (the
184 * low four bits) to a page protection mask..
186 extern pgprot_t protection_map
[16];
190 * These are the virtual MM functions - opening of an area, closing and
191 * unmapping it (needed to keep files on disk up-to-date etc), pointer
192 * to the functions called when a no-page or a wp-page exception occurs.
194 struct vm_operations_struct
{
195 void (*open
)(struct vm_area_struct
* area
);
196 void (*close
)(struct vm_area_struct
* area
);
197 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
198 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
200 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
201 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
210 * Each physical page in the system has a struct page associated with
211 * it to keep track of whatever it is we are using the page for at the
212 * moment. Note that we have no way to track which tasks are using
216 unsigned long flags
; /* Atomic flags, some possibly
217 * updated asynchronously */
218 atomic_t _count
; /* Usage count, see below. */
219 atomic_t _mapcount
; /* Count of ptes mapped in mms,
220 * to show when page is mapped
221 * & limit reverse map searches.
224 unsigned long private; /* Mapping-private opaque data:
225 * usually used for buffer_heads
226 * if PagePrivate set; used for
227 * swp_entry_t if PageSwapCache
228 * When page is free, this indicates
229 * order in the buddy system.
231 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
235 struct address_space
*mapping
; /* If low bit clear, points to
236 * inode address_space, or NULL.
237 * If page mapped as anonymous
238 * memory, low bit is set, and
239 * it points to anon_vma object:
240 * see PAGE_MAPPING_ANON below.
242 pgoff_t index
; /* Our offset within mapping. */
243 struct list_head lru
; /* Pageout list, eg. active_list
244 * protected by zone->lru_lock !
247 * On machines where all RAM is mapped into kernel address space,
248 * we can simply calculate the virtual address. On machines with
249 * highmem some memory is mapped into kernel virtual memory
250 * dynamically, so we need a place to store that address.
251 * Note that this field could be 16 bits on x86 ... ;)
253 * Architectures with slow multiplication can define
254 * WANT_PAGE_VIRTUAL in asm/page.h
256 #if defined(WANT_PAGE_VIRTUAL)
257 void *virtual; /* Kernel virtual address (NULL if
258 not kmapped, ie. highmem) */
259 #endif /* WANT_PAGE_VIRTUAL */
262 #define page_private(page) ((page)->u.private)
263 #define set_page_private(page, v) ((page)->u.private = (v))
266 * FIXME: take this include out, include page-flags.h in
267 * files which need it (119 of them)
269 #include <linux/page-flags.h>
272 * Methods to modify the page usage count.
274 * What counts for a page usage:
275 * - cache mapping (page->mapping)
276 * - private data (page->private)
277 * - page mapped in a task's page tables, each mapping
278 * is counted separately
280 * Also, many kernel routines increase the page count before a critical
281 * routine so they can be sure the page doesn't go away from under them.
283 * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we
284 * can use atomic_add_negative(-1, page->_count) to detect when the page
285 * becomes free and so that we can also use atomic_inc_and_test to atomically
286 * detect when we just tried to grab a ref on a page which some other CPU has
287 * already deemed to be freeable.
289 * NO code should make assumptions about this internal detail! Use the provided
290 * macros which retain the old rules: page_count(page) == 0 is a free page.
294 * Drop a ref, return true if the logical refcount fell to zero (the page has
297 #define put_page_testzero(p) \
299 BUG_ON(page_count(p) == 0); \
300 atomic_add_negative(-1, &(p)->_count); \
304 * Grab a ref, return true if the page previously had a logical refcount of
305 * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page
307 #define get_page_testone(p) atomic_inc_and_test(&(p)->_count)
309 #define set_page_count(p,v) atomic_set(&(p)->_count, v - 1)
310 #define __put_page(p) atomic_dec(&(p)->_count)
312 extern void FASTCALL(__page_cache_release(struct page
*));
314 static inline int page_count(struct page
*page
)
316 if (PageCompound(page
))
317 page
= (struct page
*)page_private(page
);
318 return atomic_read(&page
->_count
) + 1;
321 static inline void get_page(struct page
*page
)
323 if (unlikely(PageCompound(page
)))
324 page
= (struct page
*)page_private(page
);
325 atomic_inc(&page
->_count
);
328 void put_page(struct page
*page
);
331 * Multiple processes may "see" the same page. E.g. for untouched
332 * mappings of /dev/null, all processes see the same page full of
333 * zeroes, and text pages of executables and shared libraries have
334 * only one copy in memory, at most, normally.
336 * For the non-reserved pages, page_count(page) denotes a reference count.
337 * page_count() == 0 means the page is free. page->lru is then used for
338 * freelist management in the buddy allocator.
339 * page_count() == 1 means the page is used for exactly one purpose
340 * (e.g. a private data page of one process).
342 * A page may be used for kmalloc() or anyone else who does a
343 * __get_free_page(). In this case the page_count() is at least 1, and
344 * all other fields are unused but should be 0 or NULL. The
345 * management of this page is the responsibility of the one who uses
348 * The other pages (we may call them "process pages") are completely
349 * managed by the Linux memory manager: I/O, buffers, swapping etc.
350 * The following discussion applies only to them.
352 * A page may belong to an inode's memory mapping. In this case,
353 * page->mapping is the pointer to the inode, and page->index is the
354 * file offset of the page, in units of PAGE_CACHE_SIZE.
356 * A page contains an opaque `private' member, which belongs to the
357 * page's address_space. Usually, this is the address of a circular
358 * list of the page's disk buffers.
360 * For pages belonging to inodes, the page_count() is the number of
361 * attaches, plus 1 if `private' contains something, plus one for
362 * the page cache itself.
364 * Instead of keeping dirty/clean pages in per address-space lists, we instead
365 * now tag pages as dirty/under writeback in the radix tree.
367 * There is also a per-mapping radix tree mapping index to the page
368 * in memory if present. The tree is rooted at mapping->root.
370 * All process pages can do I/O:
371 * - inode pages may need to be read from disk,
372 * - inode pages which have been modified and are MAP_SHARED may need
373 * to be written to disk,
374 * - private pages which have been modified may need to be swapped out
375 * to swap space and (later) to be read back into memory.
379 * The zone field is never updated after free_area_init_core()
380 * sets it, so none of the operations on it need to be atomic.
385 * page->flags layout:
387 * There are three possibilities for how page->flags get
388 * laid out. The first is for the normal case, without
389 * sparsemem. The second is for sparsemem when there is
390 * plenty of space for node and section. The last is when
391 * we have run out of space and have to fall back to an
392 * alternate (slower) way of determining the node.
394 * No sparsemem: | NODE | ZONE | ... | FLAGS |
395 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
396 * no space for node: | SECTION | ZONE | ... | FLAGS |
398 #ifdef CONFIG_SPARSEMEM
399 #define SECTIONS_WIDTH SECTIONS_SHIFT
401 #define SECTIONS_WIDTH 0
404 #define ZONES_WIDTH ZONES_SHIFT
406 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
407 #define NODES_WIDTH NODES_SHIFT
409 #define NODES_WIDTH 0
412 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
413 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
414 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
415 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
418 * We are going to use the flags for the page to node mapping if its in
419 * there. This includes the case where there is no node, so it is implicit.
421 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
423 #ifndef PFN_SECTION_SHIFT
424 #define PFN_SECTION_SHIFT 0
428 * Define the bit shifts to access each section. For non-existant
429 * sections we define the shift as 0; that plus a 0 mask ensures
430 * the compiler will optimise away reference to them.
432 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
433 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
434 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
436 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
438 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
440 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
442 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
444 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
445 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
448 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
449 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
450 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
451 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
453 static inline unsigned long page_zonenum(struct page
*page
)
455 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
459 extern struct zone
*zone_table
[];
461 static inline struct zone
*page_zone(struct page
*page
)
463 return zone_table
[(page
->flags
>> ZONETABLE_PGSHIFT
) &
467 static inline unsigned long page_to_nid(struct page
*page
)
470 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
472 return page_zone(page
)->zone_pgdat
->node_id
;
474 static inline unsigned long page_to_section(struct page
*page
)
476 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
479 static inline void set_page_zone(struct page
*page
, unsigned long zone
)
481 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
482 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
484 static inline void set_page_node(struct page
*page
, unsigned long node
)
486 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
487 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
489 static inline void set_page_section(struct page
*page
, unsigned long section
)
491 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
492 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
495 static inline void set_page_links(struct page
*page
, unsigned long zone
,
496 unsigned long node
, unsigned long pfn
)
498 set_page_zone(page
, zone
);
499 set_page_node(page
, node
);
500 set_page_section(page
, pfn_to_section_nr(pfn
));
503 #ifndef CONFIG_DISCONTIGMEM
504 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
505 extern struct page
*mem_map
;
508 static inline void *lowmem_page_address(struct page
*page
)
510 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
513 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
514 #define HASHED_PAGE_VIRTUAL
517 #if defined(WANT_PAGE_VIRTUAL)
518 #define page_address(page) ((page)->virtual)
519 #define set_page_address(page, address) \
521 (page)->virtual = (address); \
523 #define page_address_init() do { } while(0)
526 #if defined(HASHED_PAGE_VIRTUAL)
527 void *page_address(struct page
*page
);
528 void set_page_address(struct page
*page
, void *virtual);
529 void page_address_init(void);
532 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
533 #define page_address(page) lowmem_page_address(page)
534 #define set_page_address(page, address) do { } while(0)
535 #define page_address_init() do { } while(0)
539 * On an anonymous page mapped into a user virtual memory area,
540 * page->mapping points to its anon_vma, not to a struct address_space;
541 * with the PAGE_MAPPING_ANON bit set to distinguish it.
543 * Please note that, confusingly, "page_mapping" refers to the inode
544 * address_space which maps the page from disk; whereas "page_mapped"
545 * refers to user virtual address space into which the page is mapped.
547 #define PAGE_MAPPING_ANON 1
549 extern struct address_space swapper_space
;
550 static inline struct address_space
*page_mapping(struct page
*page
)
552 struct address_space
*mapping
= page
->mapping
;
554 if (unlikely(PageSwapCache(page
)))
555 mapping
= &swapper_space
;
556 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
561 static inline int PageAnon(struct page
*page
)
563 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
567 * Return the pagecache index of the passed page. Regular pagecache pages
568 * use ->index whereas swapcache pages use ->private
570 static inline pgoff_t
page_index(struct page
*page
)
572 if (unlikely(PageSwapCache(page
)))
573 return page_private(page
);
578 * The atomic page->_mapcount, like _count, starts from -1:
579 * so that transitions both from it and to it can be tracked,
580 * using atomic_inc_and_test and atomic_add_negative(-1).
582 static inline void reset_page_mapcount(struct page
*page
)
584 atomic_set(&(page
)->_mapcount
, -1);
587 static inline int page_mapcount(struct page
*page
)
589 return atomic_read(&(page
)->_mapcount
) + 1;
593 * Return true if this page is mapped into pagetables.
595 static inline int page_mapped(struct page
*page
)
597 return atomic_read(&(page
)->_mapcount
) >= 0;
601 * Error return values for the *_nopage functions
603 #define NOPAGE_SIGBUS (NULL)
604 #define NOPAGE_OOM ((struct page *) (-1))
607 * Different kinds of faults, as returned by handle_mm_fault().
608 * Used to decide whether a process gets delivered SIGBUS or
609 * just gets major/minor fault counters bumped up.
611 #define VM_FAULT_OOM 0x00
612 #define VM_FAULT_SIGBUS 0x01
613 #define VM_FAULT_MINOR 0x02
614 #define VM_FAULT_MAJOR 0x03
617 * Special case for get_user_pages.
618 * Must be in a distinct bit from the above VM_FAULT_ flags.
620 #define VM_FAULT_WRITE 0x10
622 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
624 extern void show_free_areas(void);
627 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
628 unsigned long address
, int *type
);
629 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
630 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
632 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
634 #define shmem_nopage filemap_nopage
635 #define shmem_lock(a, b, c) ({0;}) /* always in memory, no need to lock */
636 #define shmem_set_policy(a, b) (0)
637 #define shmem_get_policy(a, b) (NULL)
639 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
641 int shmem_zero_setup(struct vm_area_struct
*);
643 static inline int can_do_mlock(void)
645 if (capable(CAP_IPC_LOCK
))
647 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
651 extern int user_shm_lock(size_t, struct user_struct
*);
652 extern void user_shm_unlock(size_t, struct user_struct
*);
655 * Parameter block passed down to zap_pte_range in exceptional cases.
658 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
659 struct address_space
*check_mapping
; /* Check page->mapping if set */
660 pgoff_t first_index
; /* Lowest page->index to unmap */
661 pgoff_t last_index
; /* Highest page->index to unmap */
662 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
663 unsigned long truncate_count
; /* Compare vm_truncate_count */
666 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
667 unsigned long size
, struct zap_details
*);
668 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
669 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
670 unsigned long end_addr
, unsigned long *nr_accounted
,
671 struct zap_details
*);
672 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
673 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
674 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
675 unsigned long floor
, unsigned long ceiling
);
676 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
677 struct vm_area_struct
*vma
);
678 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
679 unsigned long size
, pgprot_t prot
);
680 void unmap_mapping_range(struct address_space
*mapping
,
681 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
683 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
684 loff_t
const holebegin
, loff_t
const holelen
)
686 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
689 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
690 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
691 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
692 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
, unsigned long address
, int write_access
);
694 static inline int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long address
, int write_access
)
696 return __handle_mm_fault(mm
, vma
, address
, write_access
) & (~VM_FAULT_WRITE
);
699 extern int make_pages_present(unsigned long addr
, unsigned long end
);
700 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
701 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
703 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
704 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
705 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
707 int __set_page_dirty_buffers(struct page
*page
);
708 int __set_page_dirty_nobuffers(struct page
*page
);
709 int redirty_page_for_writepage(struct writeback_control
*wbc
,
711 int FASTCALL(set_page_dirty(struct page
*page
));
712 int set_page_dirty_lock(struct page
*page
);
713 int clear_page_dirty_for_io(struct page
*page
);
715 extern unsigned long do_mremap(unsigned long addr
,
716 unsigned long old_len
, unsigned long new_len
,
717 unsigned long flags
, unsigned long new_addr
);
720 * Prototype to add a shrinker callback for ageable caches.
722 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
723 * scan `nr_to_scan' objects, attempting to free them.
725 * The callback must return the number of objects which remain in the cache.
727 * The callback will be passed nr_to_scan == 0 when the VM is querying the
728 * cache size, so a fastpath for that case is appropriate.
730 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
733 * Add an aging callback. The int is the number of 'seeks' it takes
734 * to recreate one of the objects that these functions age.
737 #define DEFAULT_SEEKS 2
739 extern struct shrinker
*set_shrinker(int, shrinker_t
);
740 extern void remove_shrinker(struct shrinker
*shrinker
);
742 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
743 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
744 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
745 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
748 * The following ifdef needed to get the 4level-fixup.h header to work.
749 * Remove it when 4level-fixup.h has been removed.
751 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
752 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
754 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
755 NULL
: pud_offset(pgd
, address
);
758 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
760 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
761 NULL
: pmd_offset(pud
, address
);
763 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
765 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
767 * We tuck a spinlock to guard each pagetable page into its struct page,
768 * at page->private, with BUILD_BUG_ON to make sure that this will not
769 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
770 * When freeing, reset page->mapping so free_pages_check won't complain.
772 #define __pte_lockptr(page) &((page)->u.ptl)
773 #define pte_lock_init(_page) do { \
774 spin_lock_init(__pte_lockptr(_page)); \
776 #define pte_lock_deinit(page) ((page)->mapping = NULL)
777 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
780 * We use mm->page_table_lock to guard all pagetable pages of the mm.
782 #define pte_lock_init(page) do {} while (0)
783 #define pte_lock_deinit(page) do {} while (0)
784 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
785 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
787 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
789 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
790 pte_t *__pte = pte_offset_map(pmd, address); \
796 #define pte_unmap_unlock(pte, ptl) do { \
801 #define pte_alloc_map(mm, pmd, address) \
802 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
803 NULL: pte_offset_map(pmd, address))
805 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
806 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
807 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
809 #define pte_alloc_kernel(pmd, address) \
810 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
811 NULL: pte_offset_kernel(pmd, address))
813 extern void free_area_init(unsigned long * zones_size
);
814 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
815 unsigned long * zones_size
, unsigned long zone_start_pfn
,
816 unsigned long *zholes_size
);
817 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
818 extern void setup_per_zone_pages_min(void);
819 extern void mem_init(void);
820 extern void show_mem(void);
821 extern void si_meminfo(struct sysinfo
* val
);
822 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
825 extern void setup_per_cpu_pageset(void);
827 static inline void setup_per_cpu_pageset(void) {}
831 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
832 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
833 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
834 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
835 struct prio_tree_iter
*iter
);
837 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
838 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
839 (vma = vma_prio_tree_next(vma, iter)); )
841 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
842 struct list_head
*list
)
844 vma
->shared
.vm_set
.parent
= NULL
;
845 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
849 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
850 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
851 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
852 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
853 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
854 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
856 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
857 extern int split_vma(struct mm_struct
*,
858 struct vm_area_struct
*, unsigned long addr
, int new_below
);
859 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
860 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
861 struct rb_node
**, struct rb_node
*);
862 extern void unlink_file_vma(struct vm_area_struct
*);
863 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
864 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
865 extern void exit_mmap(struct mm_struct
*);
866 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
868 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
870 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
871 unsigned long len
, unsigned long prot
,
872 unsigned long flag
, unsigned long pgoff
);
874 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
875 unsigned long len
, unsigned long prot
,
876 unsigned long flag
, unsigned long offset
)
878 unsigned long ret
= -EINVAL
;
879 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
881 if (!(offset
& ~PAGE_MASK
))
882 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
887 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
889 extern unsigned long do_brk(unsigned long, unsigned long);
892 extern unsigned long page_unuse(struct page
*);
893 extern void truncate_inode_pages(struct address_space
*, loff_t
);
895 /* generic vm_area_ops exported for stackable file systems */
896 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
897 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
898 unsigned long, pgprot_t
, unsigned long, int);
900 /* mm/page-writeback.c */
901 int write_one_page(struct page
*page
, int wait
);
904 #define VM_MAX_READAHEAD 128 /* kbytes */
905 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
906 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
907 * turning readahead off */
909 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
910 pgoff_t offset
, unsigned long nr_to_read
);
911 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
912 pgoff_t offset
, unsigned long nr_to_read
);
913 unsigned long page_cache_readahead(struct address_space
*mapping
,
914 struct file_ra_state
*ra
,
918 void handle_ra_miss(struct address_space
*mapping
,
919 struct file_ra_state
*ra
, pgoff_t offset
);
920 unsigned long max_sane_readahead(unsigned long nr
);
922 /* Do stack extension */
923 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
925 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
928 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
929 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
930 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
931 struct vm_area_struct
**pprev
);
933 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
934 NULL if none. Assume start_addr < end_addr. */
935 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
937 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
939 if (vma
&& end_addr
<= vma
->vm_start
)
944 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
946 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
949 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
950 struct page
*vmalloc_to_page(void *addr
);
951 unsigned long vmalloc_to_pfn(void *addr
);
952 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
953 unsigned long pfn
, unsigned long size
, pgprot_t
);
955 struct page
*follow_page(struct mm_struct
*, unsigned long address
,
956 unsigned int foll_flags
);
957 #define FOLL_WRITE 0x01 /* check pte is writable */
958 #define FOLL_TOUCH 0x02 /* mark page accessed */
959 #define FOLL_GET 0x04 /* do get_page on page */
960 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
962 #ifdef CONFIG_PROC_FS
963 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
965 static inline void vm_stat_account(struct mm_struct
*mm
,
966 unsigned long flags
, struct file
*file
, long pages
)
969 #endif /* CONFIG_PROC_FS */
971 #ifndef CONFIG_DEBUG_PAGEALLOC
973 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
978 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
979 #ifdef __HAVE_ARCH_GATE_AREA
980 int in_gate_area_no_task(unsigned long addr
);
981 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
983 int in_gate_area_no_task(unsigned long addr
);
984 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
985 #endif /* __HAVE_ARCH_GATE_AREA */
987 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
988 #define OOM_DISABLE -17
990 #endif /* __KERNEL__ */
991 #endif /* _LINUX_MM_H */