4 #include <linux/sched.h>
5 #include <linux/errno.h>
6 #include <linux/capability.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>
16 #include <linux/mutex.h>
17 #include <linux/debug_locks.h>
18 #include <linux/backing-dev.h>
23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24 extern unsigned long max_mapnr
;
27 extern unsigned long num_physpages
;
28 extern void * high_memory
;
29 extern unsigned long vmalloc_earlyreserve
;
30 extern int page_cluster
;
33 extern int sysctl_legacy_va_layout
;
35 #define sysctl_legacy_va_layout 0
39 #include <asm/pgtable.h>
40 #include <asm/processor.h>
42 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45 * Linux kernel virtual memory manager primitives.
46 * The idea being to have a "virtual" mm in the same way
47 * we have a virtual fs - giving a cleaner interface to the
48 * mm details, and allowing different kinds of memory mappings
49 * (from shared memory to executable loading to arbitrary
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_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
150 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
152 #define VM_EXECUTABLE 0x00001000
153 #define VM_LOCKED 0x00002000
154 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
156 /* Used by sys_madvise() */
157 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
158 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
160 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
161 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
162 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
163 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
164 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
165 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
166 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
167 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
169 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
170 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
173 #ifdef CONFIG_STACK_GROWSUP
174 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
176 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
179 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
180 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
181 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
182 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
183 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
186 * mapping from the currently active vm_flags protection bits (the
187 * low four bits) to a page protection mask..
189 extern pgprot_t protection_map
[16];
193 * These are the virtual MM functions - opening of an area, closing and
194 * unmapping it (needed to keep files on disk up-to-date etc), pointer
195 * to the functions called when a no-page or a wp-page exception occurs.
197 struct vm_operations_struct
{
198 void (*open
)(struct vm_area_struct
* area
);
199 void (*close
)(struct vm_area_struct
* area
);
200 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
201 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
203 /* notification that a previously read-only page is about to become
204 * writable, if an error is returned it will cause a SIGBUS */
205 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
207 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
208 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
210 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
211 const nodemask_t
*to
, unsigned long flags
);
219 * Each physical page in the system has a struct page associated with
220 * it to keep track of whatever it is we are using the page for at the
221 * moment. Note that we have no way to track which tasks are using
222 * a page, though if it is a pagecache page, rmap structures can tell us
226 unsigned long flags
; /* Atomic flags, some possibly
227 * updated asynchronously */
228 atomic_t _count
; /* Usage count, see below. */
229 atomic_t _mapcount
; /* Count of ptes mapped in mms,
230 * to show when page is mapped
231 * & limit reverse map searches.
235 unsigned long private; /* Mapping-private opaque data:
236 * usually used for buffer_heads
237 * if PagePrivate set; used for
238 * swp_entry_t if PageSwapCache;
239 * indicates order in the buddy
240 * system if PG_buddy is set.
242 struct address_space
*mapping
; /* If low bit clear, points to
243 * inode address_space, or NULL.
244 * If page mapped as anonymous
245 * memory, low bit is set, and
246 * it points to anon_vma object:
247 * see PAGE_MAPPING_ANON below.
250 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
254 pgoff_t index
; /* Our offset within mapping. */
255 struct list_head lru
; /* Pageout list, eg. active_list
256 * protected by zone->lru_lock !
259 * On machines where all RAM is mapped into kernel address space,
260 * we can simply calculate the virtual address. On machines with
261 * highmem some memory is mapped into kernel virtual memory
262 * dynamically, so we need a place to store that address.
263 * Note that this field could be 16 bits on x86 ... ;)
265 * Architectures with slow multiplication can define
266 * WANT_PAGE_VIRTUAL in asm/page.h
268 #if defined(WANT_PAGE_VIRTUAL)
269 void *virtual; /* Kernel virtual address (NULL if
270 not kmapped, ie. highmem) */
271 #endif /* WANT_PAGE_VIRTUAL */
274 #define page_private(page) ((page)->private)
275 #define set_page_private(page, v) ((page)->private = (v))
278 * FIXME: take this include out, include page-flags.h in
279 * files which need it (119 of them)
281 #include <linux/page-flags.h>
283 #ifdef CONFIG_DEBUG_VM
284 #define VM_BUG_ON(cond) BUG_ON(cond)
286 #define VM_BUG_ON(condition) do { } while(0)
290 * Methods to modify the page usage count.
292 * What counts for a page usage:
293 * - cache mapping (page->mapping)
294 * - private data (page->private)
295 * - page mapped in a task's page tables, each mapping
296 * is counted separately
298 * Also, many kernel routines increase the page count before a critical
299 * routine so they can be sure the page doesn't go away from under them.
303 * Drop a ref, return true if the refcount fell to zero (the page has no users)
305 static inline int put_page_testzero(struct page
*page
)
307 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
308 return atomic_dec_and_test(&page
->_count
);
312 * Try to grab a ref unless the page has a refcount of zero, return false if
315 static inline int get_page_unless_zero(struct page
*page
)
317 VM_BUG_ON(PageCompound(page
));
318 return atomic_inc_not_zero(&page
->_count
);
321 static inline int page_count(struct page
*page
)
323 if (unlikely(PageCompound(page
)))
324 page
= (struct page
*)page_private(page
);
325 return atomic_read(&page
->_count
);
328 static inline void get_page(struct page
*page
)
330 if (unlikely(PageCompound(page
)))
331 page
= (struct page
*)page_private(page
);
332 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
333 atomic_inc(&page
->_count
);
337 * Setup the page count before being freed into the page allocator for
338 * the first time (boot or memory hotplug)
340 static inline void init_page_count(struct page
*page
)
342 atomic_set(&page
->_count
, 1);
345 void put_page(struct page
*page
);
346 void put_pages_list(struct list_head
*pages
);
348 void split_page(struct page
*page
, unsigned int order
);
351 * Multiple processes may "see" the same page. E.g. for untouched
352 * mappings of /dev/null, all processes see the same page full of
353 * zeroes, and text pages of executables and shared libraries have
354 * only one copy in memory, at most, normally.
356 * For the non-reserved pages, page_count(page) denotes a reference count.
357 * page_count() == 0 means the page is free. page->lru is then used for
358 * freelist management in the buddy allocator.
359 * page_count() > 0 means the page has been allocated.
361 * Pages are allocated by the slab allocator in order to provide memory
362 * to kmalloc and kmem_cache_alloc. In this case, the management of the
363 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
364 * unless a particular usage is carefully commented. (the responsibility of
365 * freeing the kmalloc memory is the caller's, of course).
367 * A page may be used by anyone else who does a __get_free_page().
368 * In this case, page_count still tracks the references, and should only
369 * be used through the normal accessor functions. The top bits of page->flags
370 * and page->virtual store page management information, but all other fields
371 * are unused and could be used privately, carefully. The management of this
372 * page is the responsibility of the one who allocated it, and those who have
373 * subsequently been given references to it.
375 * The other pages (we may call them "pagecache pages") are completely
376 * managed by the Linux memory manager: I/O, buffers, swapping etc.
377 * The following discussion applies only to them.
379 * A pagecache page contains an opaque `private' member, which belongs to the
380 * page's address_space. Usually, this is the address of a circular list of
381 * the page's disk buffers. PG_private must be set to tell the VM to call
382 * into the filesystem to release these pages.
384 * A page may belong to an inode's memory mapping. In this case, page->mapping
385 * is the pointer to the inode, and page->index is the file offset of the page,
386 * in units of PAGE_CACHE_SIZE.
388 * If pagecache pages are not associated with an inode, they are said to be
389 * anonymous pages. These may become associated with the swapcache, and in that
390 * case PG_swapcache is set, and page->private is an offset into the swapcache.
392 * In either case (swapcache or inode backed), the pagecache itself holds one
393 * reference to the page. Setting PG_private should also increment the
394 * refcount. The each user mapping also has a reference to the page.
396 * The pagecache pages are stored in a per-mapping radix tree, which is
397 * rooted at mapping->page_tree, and indexed by offset.
398 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
399 * lists, we instead now tag pages as dirty/writeback in the radix tree.
401 * All pagecache pages may be subject to I/O:
402 * - inode pages may need to be read from disk,
403 * - inode pages which have been modified and are MAP_SHARED may need
404 * to be written back to the inode on disk,
405 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
406 * modified may need to be swapped out to swap space and (later) to be read
411 * The zone field is never updated after free_area_init_core()
412 * sets it, so none of the operations on it need to be atomic.
417 * page->flags layout:
419 * There are three possibilities for how page->flags get
420 * laid out. The first is for the normal case, without
421 * sparsemem. The second is for sparsemem when there is
422 * plenty of space for node and section. The last is when
423 * we have run out of space and have to fall back to an
424 * alternate (slower) way of determining the node.
426 * No sparsemem: | NODE | ZONE | ... | FLAGS |
427 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
428 * no space for node: | SECTION | ZONE | ... | FLAGS |
430 #ifdef CONFIG_SPARSEMEM
431 #define SECTIONS_WIDTH SECTIONS_SHIFT
433 #define SECTIONS_WIDTH 0
436 #define ZONES_WIDTH ZONES_SHIFT
438 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
439 #define NODES_WIDTH NODES_SHIFT
441 #define NODES_WIDTH 0
444 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
445 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
446 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
447 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
450 * We are going to use the flags for the page to node mapping if its in
451 * there. This includes the case where there is no node, so it is implicit.
453 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
455 #ifndef PFN_SECTION_SHIFT
456 #define PFN_SECTION_SHIFT 0
460 * Define the bit shifts to access each section. For non-existant
461 * sections we define the shift as 0; that plus a 0 mask ensures
462 * the compiler will optimise away reference to them.
464 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
465 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
466 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
468 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
470 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
472 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
474 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
476 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
477 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
480 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
481 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
482 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
483 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
485 static inline enum zone_type
page_zonenum(struct page
*page
)
487 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
491 extern struct zone
*zone_table
[];
493 static inline int page_zone_id(struct page
*page
)
495 return (page
->flags
>> ZONETABLE_PGSHIFT
) & ZONETABLE_MASK
;
497 static inline struct zone
*page_zone(struct page
*page
)
499 return zone_table
[page_zone_id(page
)];
502 static inline unsigned long zone_to_nid(struct zone
*zone
)
504 return zone
->zone_pgdat
->node_id
;
507 static inline unsigned long page_to_nid(struct page
*page
)
510 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
512 return zone_to_nid(page_zone(page
));
514 static inline unsigned long page_to_section(struct page
*page
)
516 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
519 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
521 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
522 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
525 static inline void set_page_node(struct page
*page
, unsigned long node
)
527 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
528 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
530 static inline void set_page_section(struct page
*page
, unsigned long section
)
532 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
533 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
536 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
537 unsigned long node
, unsigned long pfn
)
539 set_page_zone(page
, zone
);
540 set_page_node(page
, node
);
541 set_page_section(page
, pfn_to_section_nr(pfn
));
545 * Some inline functions in vmstat.h depend on page_zone()
547 #include <linux/vmstat.h>
549 #ifndef CONFIG_DISCONTIGMEM
550 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
551 extern struct page
*mem_map
;
554 static __always_inline
void *lowmem_page_address(struct page
*page
)
556 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
559 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
560 #define HASHED_PAGE_VIRTUAL
563 #if defined(WANT_PAGE_VIRTUAL)
564 #define page_address(page) ((page)->virtual)
565 #define set_page_address(page, address) \
567 (page)->virtual = (address); \
569 #define page_address_init() do { } while(0)
572 #if defined(HASHED_PAGE_VIRTUAL)
573 void *page_address(struct page
*page
);
574 void set_page_address(struct page
*page
, void *virtual);
575 void page_address_init(void);
578 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
579 #define page_address(page) lowmem_page_address(page)
580 #define set_page_address(page, address) do { } while(0)
581 #define page_address_init() do { } while(0)
585 * On an anonymous page mapped into a user virtual memory area,
586 * page->mapping points to its anon_vma, not to a struct address_space;
587 * with the PAGE_MAPPING_ANON bit set to distinguish it.
589 * Please note that, confusingly, "page_mapping" refers to the inode
590 * address_space which maps the page from disk; whereas "page_mapped"
591 * refers to user virtual address space into which the page is mapped.
593 #define PAGE_MAPPING_ANON 1
595 extern struct address_space swapper_space
;
596 static inline struct address_space
*page_mapping(struct page
*page
)
598 struct address_space
*mapping
= page
->mapping
;
600 if (unlikely(PageSwapCache(page
)))
601 mapping
= &swapper_space
;
602 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
607 static inline int PageAnon(struct page
*page
)
609 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
613 * Return the pagecache index of the passed page. Regular pagecache pages
614 * use ->index whereas swapcache pages use ->private
616 static inline pgoff_t
page_index(struct page
*page
)
618 if (unlikely(PageSwapCache(page
)))
619 return page_private(page
);
624 * The atomic page->_mapcount, like _count, starts from -1:
625 * so that transitions both from it and to it can be tracked,
626 * using atomic_inc_and_test and atomic_add_negative(-1).
628 static inline void reset_page_mapcount(struct page
*page
)
630 atomic_set(&(page
)->_mapcount
, -1);
633 static inline int page_mapcount(struct page
*page
)
635 return atomic_read(&(page
)->_mapcount
) + 1;
639 * Return true if this page is mapped into pagetables.
641 static inline int page_mapped(struct page
*page
)
643 return atomic_read(&(page
)->_mapcount
) >= 0;
647 * Error return values for the *_nopage functions
649 #define NOPAGE_SIGBUS (NULL)
650 #define NOPAGE_OOM ((struct page *) (-1))
653 * Different kinds of faults, as returned by handle_mm_fault().
654 * Used to decide whether a process gets delivered SIGBUS or
655 * just gets major/minor fault counters bumped up.
657 #define VM_FAULT_OOM 0x00
658 #define VM_FAULT_SIGBUS 0x01
659 #define VM_FAULT_MINOR 0x02
660 #define VM_FAULT_MAJOR 0x03
663 * Special case for get_user_pages.
664 * Must be in a distinct bit from the above VM_FAULT_ flags.
666 #define VM_FAULT_WRITE 0x10
668 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
670 extern void show_free_areas(void);
673 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
674 unsigned long address
, int *type
);
675 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
676 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
678 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
680 #define shmem_nopage filemap_nopage
682 static inline int shmem_lock(struct file
*file
, int lock
,
683 struct user_struct
*user
)
688 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
689 struct mempolicy
*new)
694 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
700 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
701 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
703 int shmem_zero_setup(struct vm_area_struct
*);
706 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
710 unsigned long flags
);
713 static inline int can_do_mlock(void)
715 if (capable(CAP_IPC_LOCK
))
717 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
721 extern int user_shm_lock(size_t, struct user_struct
*);
722 extern void user_shm_unlock(size_t, struct user_struct
*);
725 * Parameter block passed down to zap_pte_range in exceptional cases.
728 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
729 struct address_space
*check_mapping
; /* Check page->mapping if set */
730 pgoff_t first_index
; /* Lowest page->index to unmap */
731 pgoff_t last_index
; /* Highest page->index to unmap */
732 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
733 unsigned long truncate_count
; /* Compare vm_truncate_count */
736 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
737 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
738 unsigned long size
, struct zap_details
*);
739 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
740 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
741 unsigned long end_addr
, unsigned long *nr_accounted
,
742 struct zap_details
*);
743 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
744 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
745 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
746 unsigned long floor
, unsigned long ceiling
);
747 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
748 struct vm_area_struct
*vma
);
749 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
750 unsigned long size
, pgprot_t prot
);
751 void unmap_mapping_range(struct address_space
*mapping
,
752 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
754 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
755 loff_t
const holebegin
, loff_t
const holelen
)
757 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
760 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
761 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
762 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
763 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
766 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
767 unsigned long address
, int write_access
);
769 static inline int handle_mm_fault(struct mm_struct
*mm
,
770 struct vm_area_struct
*vma
, unsigned long address
,
773 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
777 static inline int handle_mm_fault(struct mm_struct
*mm
,
778 struct vm_area_struct
*vma
, unsigned long address
,
781 /* should never happen if there's no MMU */
783 return VM_FAULT_SIGBUS
;
787 extern int make_pages_present(unsigned long addr
, unsigned long end
);
788 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
789 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
791 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
792 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
793 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
795 int __set_page_dirty_buffers(struct page
*page
);
796 int __set_page_dirty_nobuffers(struct page
*page
);
797 int redirty_page_for_writepage(struct writeback_control
*wbc
,
799 int FASTCALL(set_page_dirty(struct page
*page
));
800 int set_page_dirty_lock(struct page
*page
);
801 int clear_page_dirty_for_io(struct page
*page
);
803 extern unsigned long do_mremap(unsigned long addr
,
804 unsigned long old_len
, unsigned long new_len
,
805 unsigned long flags
, unsigned long new_addr
);
808 * Prototype to add a shrinker callback for ageable caches.
810 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
811 * scan `nr_to_scan' objects, attempting to free them.
813 * The callback must return the number of objects which remain in the cache.
815 * The callback will be passed nr_to_scan == 0 when the VM is querying the
816 * cache size, so a fastpath for that case is appropriate.
818 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
821 * Add an aging callback. The int is the number of 'seeks' it takes
822 * to recreate one of the objects that these functions age.
825 #define DEFAULT_SEEKS 2
827 extern struct shrinker
*set_shrinker(int, shrinker_t
);
828 extern void remove_shrinker(struct shrinker
*shrinker
);
831 * Some shared mappigns will want the pages marked read-only
832 * to track write events. If so, we'll downgrade vm_page_prot
833 * to the private version (using protection_map[] without the
836 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
838 unsigned int vm_flags
= vma
->vm_flags
;
840 /* If it was private or non-writable, the write bit is already clear */
841 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
844 /* The backer wishes to know when pages are first written to? */
845 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
848 /* The open routine did something to the protections already? */
849 if (pgprot_val(vma
->vm_page_prot
) !=
850 pgprot_val(protection_map
[vm_flags
&
851 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
854 /* Specialty mapping? */
855 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
858 /* Can the mapping track the dirty pages? */
859 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
860 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
863 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
865 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
866 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
867 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
868 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
871 * The following ifdef needed to get the 4level-fixup.h header to work.
872 * Remove it when 4level-fixup.h has been removed.
874 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
875 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
877 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
878 NULL
: pud_offset(pgd
, address
);
881 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
883 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
884 NULL
: pmd_offset(pud
, address
);
886 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
888 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
890 * We tuck a spinlock to guard each pagetable page into its struct page,
891 * at page->private, with BUILD_BUG_ON to make sure that this will not
892 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
893 * When freeing, reset page->mapping so free_pages_check won't complain.
895 #define __pte_lockptr(page) &((page)->ptl)
896 #define pte_lock_init(_page) do { \
897 spin_lock_init(__pte_lockptr(_page)); \
899 #define pte_lock_deinit(page) ((page)->mapping = NULL)
900 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
903 * We use mm->page_table_lock to guard all pagetable pages of the mm.
905 #define pte_lock_init(page) do {} while (0)
906 #define pte_lock_deinit(page) do {} while (0)
907 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
908 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
910 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
912 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
913 pte_t *__pte = pte_offset_map(pmd, address); \
919 #define pte_unmap_unlock(pte, ptl) do { \
924 #define pte_alloc_map(mm, pmd, address) \
925 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
926 NULL: pte_offset_map(pmd, address))
928 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
929 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
930 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
932 #define pte_alloc_kernel(pmd, address) \
933 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
934 NULL: pte_offset_kernel(pmd, address))
936 extern void free_area_init(unsigned long * zones_size
);
937 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
938 unsigned long * zones_size
, unsigned long zone_start_pfn
,
939 unsigned long *zholes_size
);
940 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
941 extern void setup_per_zone_pages_min(void);
942 extern void mem_init(void);
943 extern void show_mem(void);
944 extern void si_meminfo(struct sysinfo
* val
);
945 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
948 extern void setup_per_cpu_pageset(void);
950 static inline void setup_per_cpu_pageset(void) {}
954 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
955 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
956 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
957 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
958 struct prio_tree_iter
*iter
);
960 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
961 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
962 (vma = vma_prio_tree_next(vma, iter)); )
964 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
965 struct list_head
*list
)
967 vma
->shared
.vm_set
.parent
= NULL
;
968 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
972 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
973 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
974 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
975 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
976 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
977 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
979 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
980 extern int split_vma(struct mm_struct
*,
981 struct vm_area_struct
*, unsigned long addr
, int new_below
);
982 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
983 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
984 struct rb_node
**, struct rb_node
*);
985 extern void unlink_file_vma(struct vm_area_struct
*);
986 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
987 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
988 extern void exit_mmap(struct mm_struct
*);
989 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
991 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
993 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
994 unsigned long len
, unsigned long prot
,
995 unsigned long flag
, unsigned long pgoff
);
997 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
998 unsigned long len
, unsigned long prot
,
999 unsigned long flag
, unsigned long offset
)
1001 unsigned long ret
= -EINVAL
;
1002 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1004 if (!(offset
& ~PAGE_MASK
))
1005 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1010 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1012 extern unsigned long do_brk(unsigned long, unsigned long);
1015 extern unsigned long page_unuse(struct page
*);
1016 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1017 extern void truncate_inode_pages_range(struct address_space
*,
1018 loff_t lstart
, loff_t lend
);
1020 /* generic vm_area_ops exported for stackable file systems */
1021 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
1022 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
1023 unsigned long, pgprot_t
, unsigned long, int);
1025 /* mm/page-writeback.c */
1026 int write_one_page(struct page
*page
, int wait
);
1029 #define VM_MAX_READAHEAD 128 /* kbytes */
1030 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1031 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1032 * turning readahead off */
1034 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1035 pgoff_t offset
, unsigned long nr_to_read
);
1036 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1037 pgoff_t offset
, unsigned long nr_to_read
);
1038 unsigned long page_cache_readahead(struct address_space
*mapping
,
1039 struct file_ra_state
*ra
,
1042 unsigned long size
);
1043 void handle_ra_miss(struct address_space
*mapping
,
1044 struct file_ra_state
*ra
, pgoff_t offset
);
1045 unsigned long max_sane_readahead(unsigned long nr
);
1047 /* Do stack extension */
1048 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1050 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1053 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1054 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1055 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1056 struct vm_area_struct
**pprev
);
1058 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1059 NULL if none. Assume start_addr < end_addr. */
1060 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1062 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1064 if (vma
&& end_addr
<= vma
->vm_start
)
1069 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1071 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1074 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1075 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1076 struct page
*vmalloc_to_page(void *addr
);
1077 unsigned long vmalloc_to_pfn(void *addr
);
1078 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1079 unsigned long pfn
, unsigned long size
, pgprot_t
);
1080 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1082 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1083 unsigned int foll_flags
);
1084 #define FOLL_WRITE 0x01 /* check pte is writable */
1085 #define FOLL_TOUCH 0x02 /* mark page accessed */
1086 #define FOLL_GET 0x04 /* do get_page on page */
1087 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1089 #ifdef CONFIG_PROC_FS
1090 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1092 static inline void vm_stat_account(struct mm_struct
*mm
,
1093 unsigned long flags
, struct file
*file
, long pages
)
1096 #endif /* CONFIG_PROC_FS */
1098 #ifndef CONFIG_DEBUG_PAGEALLOC
1100 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
1102 if (!PageHighMem(page
) && !enable
)
1103 debug_check_no_locks_freed(page_address(page
),
1104 numpages
* PAGE_SIZE
);
1108 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1109 #ifdef __HAVE_ARCH_GATE_AREA
1110 int in_gate_area_no_task(unsigned long addr
);
1111 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1113 int in_gate_area_no_task(unsigned long addr
);
1114 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1115 #endif /* __HAVE_ARCH_GATE_AREA */
1117 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1118 #define OOM_DISABLE -17
1120 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1121 void __user
*, size_t *, loff_t
*);
1122 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1123 unsigned long lru_pages
);
1124 void drop_pagecache(void);
1125 void drop_slab(void);
1128 #define randomize_va_space 0
1130 extern int randomize_va_space
;
1133 const char *arch_vma_name(struct vm_area_struct
*vma
);
1135 #endif /* __KERNEL__ */
1136 #endif /* _LINUX_MM_H */