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>
19 #include <linux/mm_types.h>
24 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
25 extern unsigned long max_mapnr
;
28 extern unsigned long num_physpages
;
29 extern void * high_memory
;
30 extern unsigned long vmalloc_earlyreserve
;
31 extern int page_cluster
;
34 extern int sysctl_legacy_va_layout
;
36 #define sysctl_legacy_va_layout 0
40 #include <asm/pgtable.h>
41 #include <asm/processor.h>
43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
46 * Linux kernel virtual memory manager primitives.
47 * The idea being to have a "virtual" mm in the same way
48 * we have a virtual fs - giving a cleaner interface to the
49 * mm details, and allowing different kinds of memory mappings
50 * (from shared memory to executable loading to arbitrary
55 * This struct defines a memory VMM memory area. There is one of these
56 * per VM-area/task. A VM area is any part of the process virtual memory
57 * space that has a special rule for the page-fault handlers (ie a shared
58 * library, the executable area etc).
60 struct vm_area_struct
{
61 struct mm_struct
* vm_mm
; /* The address space we belong to. */
62 unsigned long vm_start
; /* Our start address within vm_mm. */
63 unsigned long vm_end
; /* The first byte after our end address
66 /* linked list of VM areas per task, sorted by address */
67 struct vm_area_struct
*vm_next
;
69 pgprot_t vm_page_prot
; /* Access permissions of this VMA. */
70 unsigned long vm_flags
; /* Flags, listed below. */
75 * For areas with an address space and backing store,
76 * linkage into the address_space->i_mmap prio tree, or
77 * linkage to the list of like vmas hanging off its node, or
78 * linkage of vma in the address_space->i_mmap_nonlinear list.
82 struct list_head list
;
83 void *parent
; /* aligns with prio_tree_node parent */
84 struct vm_area_struct
*head
;
87 struct raw_prio_tree_node prio_tree_node
;
91 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
92 * list, after a COW of one of the file pages. A MAP_SHARED vma
93 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
94 * or brk vma (with NULL file) can only be in an anon_vma list.
96 struct list_head anon_vma_node
; /* Serialized by anon_vma->lock */
97 struct anon_vma
*anon_vma
; /* Serialized by page_table_lock */
99 /* Function pointers to deal with this struct. */
100 struct vm_operations_struct
* vm_ops
;
102 /* Information about our backing store: */
103 unsigned long vm_pgoff
; /* Offset (within vm_file) in PAGE_SIZE
104 units, *not* PAGE_CACHE_SIZE */
105 struct file
* vm_file
; /* File we map to (can be NULL). */
106 void * vm_private_data
; /* was vm_pte (shared mem) */
107 unsigned long vm_truncate_count
;/* truncate_count or restart_addr */
110 atomic_t vm_usage
; /* refcount (VMAs shared if !MMU) */
113 struct mempolicy
*vm_policy
; /* NUMA policy for the VMA */
117 extern struct kmem_cache
*vm_area_cachep
;
120 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
121 * disabled, then there's a single shared list of VMAs maintained by the
122 * system, and mm's subscribe to these individually
124 struct vm_list_struct
{
125 struct vm_list_struct
*next
;
126 struct vm_area_struct
*vma
;
130 extern struct rb_root nommu_vma_tree
;
131 extern struct rw_semaphore nommu_vma_sem
;
133 extern unsigned int kobjsize(const void *objp
);
139 #define VM_READ 0x00000001 /* currently active flags */
140 #define VM_WRITE 0x00000002
141 #define VM_EXEC 0x00000004
142 #define VM_SHARED 0x00000008
144 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
145 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
146 #define VM_MAYWRITE 0x00000020
147 #define VM_MAYEXEC 0x00000040
148 #define VM_MAYSHARE 0x00000080
150 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
151 #define VM_GROWSUP 0x00000200
152 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
153 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
155 #define VM_EXECUTABLE 0x00001000
156 #define VM_LOCKED 0x00002000
157 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
159 /* Used by sys_madvise() */
160 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
161 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
163 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
164 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
165 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
166 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
167 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
168 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
169 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
170 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
172 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
173 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
176 #ifdef CONFIG_STACK_GROWSUP
177 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
179 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
182 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
183 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
184 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
185 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
186 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
189 * mapping from the currently active vm_flags protection bits (the
190 * low four bits) to a page protection mask..
192 extern pgprot_t protection_map
[16];
196 * These are the virtual MM functions - opening of an area, closing and
197 * unmapping it (needed to keep files on disk up-to-date etc), pointer
198 * to the functions called when a no-page or a wp-page exception occurs.
200 struct vm_operations_struct
{
201 void (*open
)(struct vm_area_struct
* area
);
202 void (*close
)(struct vm_area_struct
* area
);
203 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
204 unsigned long (*nopfn
)(struct vm_area_struct
* area
, unsigned long address
);
205 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
207 /* notification that a previously read-only page is about to become
208 * writable, if an error is returned it will cause a SIGBUS */
209 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
211 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
212 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
214 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
215 const nodemask_t
*to
, unsigned long flags
);
222 #define page_private(page) ((page)->private)
223 #define set_page_private(page, v) ((page)->private = (v))
226 * FIXME: take this include out, include page-flags.h in
227 * files which need it (119 of them)
229 #include <linux/page-flags.h>
231 #ifdef CONFIG_DEBUG_VM
232 #define VM_BUG_ON(cond) BUG_ON(cond)
234 #define VM_BUG_ON(condition) do { } while(0)
238 * Methods to modify the page usage count.
240 * What counts for a page usage:
241 * - cache mapping (page->mapping)
242 * - private data (page->private)
243 * - page mapped in a task's page tables, each mapping
244 * is counted separately
246 * Also, many kernel routines increase the page count before a critical
247 * routine so they can be sure the page doesn't go away from under them.
251 * Drop a ref, return true if the refcount fell to zero (the page has no users)
253 static inline int put_page_testzero(struct page
*page
)
255 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
256 return atomic_dec_and_test(&page
->_count
);
260 * Try to grab a ref unless the page has a refcount of zero, return false if
263 static inline int get_page_unless_zero(struct page
*page
)
265 VM_BUG_ON(PageCompound(page
));
266 return atomic_inc_not_zero(&page
->_count
);
269 static inline int page_count(struct page
*page
)
271 if (unlikely(PageCompound(page
)))
272 page
= (struct page
*)page_private(page
);
273 return atomic_read(&page
->_count
);
276 static inline void get_page(struct page
*page
)
278 if (unlikely(PageCompound(page
)))
279 page
= (struct page
*)page_private(page
);
280 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
281 atomic_inc(&page
->_count
);
285 * Setup the page count before being freed into the page allocator for
286 * the first time (boot or memory hotplug)
288 static inline void init_page_count(struct page
*page
)
290 atomic_set(&page
->_count
, 1);
293 void put_page(struct page
*page
);
294 void put_pages_list(struct list_head
*pages
);
296 void split_page(struct page
*page
, unsigned int order
);
299 * Compound pages have a destructor function. Provide a
300 * prototype for that function and accessor functions.
301 * These are _only_ valid on the head of a PG_compound page.
303 typedef void compound_page_dtor(struct page
*);
305 static inline void set_compound_page_dtor(struct page
*page
,
306 compound_page_dtor
*dtor
)
308 page
[1].lru
.next
= (void *)dtor
;
311 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
313 return (compound_page_dtor
*)page
[1].lru
.next
;
317 * Multiple processes may "see" the same page. E.g. for untouched
318 * mappings of /dev/null, all processes see the same page full of
319 * zeroes, and text pages of executables and shared libraries have
320 * only one copy in memory, at most, normally.
322 * For the non-reserved pages, page_count(page) denotes a reference count.
323 * page_count() == 0 means the page is free. page->lru is then used for
324 * freelist management in the buddy allocator.
325 * page_count() > 0 means the page has been allocated.
327 * Pages are allocated by the slab allocator in order to provide memory
328 * to kmalloc and kmem_cache_alloc. In this case, the management of the
329 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
330 * unless a particular usage is carefully commented. (the responsibility of
331 * freeing the kmalloc memory is the caller's, of course).
333 * A page may be used by anyone else who does a __get_free_page().
334 * In this case, page_count still tracks the references, and should only
335 * be used through the normal accessor functions. The top bits of page->flags
336 * and page->virtual store page management information, but all other fields
337 * are unused and could be used privately, carefully. The management of this
338 * page is the responsibility of the one who allocated it, and those who have
339 * subsequently been given references to it.
341 * The other pages (we may call them "pagecache pages") are completely
342 * managed by the Linux memory manager: I/O, buffers, swapping etc.
343 * The following discussion applies only to them.
345 * A pagecache page contains an opaque `private' member, which belongs to the
346 * page's address_space. Usually, this is the address of a circular list of
347 * the page's disk buffers. PG_private must be set to tell the VM to call
348 * into the filesystem to release these pages.
350 * A page may belong to an inode's memory mapping. In this case, page->mapping
351 * is the pointer to the inode, and page->index is the file offset of the page,
352 * in units of PAGE_CACHE_SIZE.
354 * If pagecache pages are not associated with an inode, they are said to be
355 * anonymous pages. These may become associated with the swapcache, and in that
356 * case PG_swapcache is set, and page->private is an offset into the swapcache.
358 * In either case (swapcache or inode backed), the pagecache itself holds one
359 * reference to the page. Setting PG_private should also increment the
360 * refcount. The each user mapping also has a reference to the page.
362 * The pagecache pages are stored in a per-mapping radix tree, which is
363 * rooted at mapping->page_tree, and indexed by offset.
364 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
365 * lists, we instead now tag pages as dirty/writeback in the radix tree.
367 * All pagecache pages may be subject to I/O:
368 * - inode pages may need to be read from disk,
369 * - inode pages which have been modified and are MAP_SHARED may need
370 * to be written back to the inode on disk,
371 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
372 * modified may need to be swapped out to swap space and (later) to be read
377 * The zone field is never updated after free_area_init_core()
378 * sets it, so none of the operations on it need to be atomic.
383 * page->flags layout:
385 * There are three possibilities for how page->flags get
386 * laid out. The first is for the normal case, without
387 * sparsemem. The second is for sparsemem when there is
388 * plenty of space for node and section. The last is when
389 * we have run out of space and have to fall back to an
390 * alternate (slower) way of determining the node.
392 * No sparsemem: | NODE | ZONE | ... | FLAGS |
393 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
394 * no space for node: | SECTION | ZONE | ... | FLAGS |
396 #ifdef CONFIG_SPARSEMEM
397 #define SECTIONS_WIDTH SECTIONS_SHIFT
399 #define SECTIONS_WIDTH 0
402 #define ZONES_WIDTH ZONES_SHIFT
404 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
405 #define NODES_WIDTH NODES_SHIFT
407 #define NODES_WIDTH 0
410 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
411 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
412 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
413 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
416 * We are going to use the flags for the page to node mapping if its in
417 * there. This includes the case where there is no node, so it is implicit.
419 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
420 #define NODE_NOT_IN_PAGE_FLAGS
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 ID a zone for the buddy allcator */
437 #ifdef NODE_NOT_IN_PAGEFLAGS
438 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
440 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
444 #define ZONEID_PGSHIFT ZONES_PGSHIFT
446 #define ZONEID_PGSHIFT NODES_PGOFF
449 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
450 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
453 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
454 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
455 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
456 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
458 static inline enum zone_type
page_zonenum(struct page
*page
)
460 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
464 * The identification function is only used by the buddy allocator for
465 * determining if two pages could be buddies. We are not really
466 * identifying a zone since we could be using a the section number
467 * id if we have not node id available in page flags.
468 * We guarantee only that it will return the same value for two
469 * combinable pages in a zone.
471 static inline int page_zone_id(struct page
*page
)
473 BUILD_BUG_ON(ZONEID_PGSHIFT
== 0 && ZONEID_MASK
);
474 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
477 static inline int zone_to_nid(struct zone
*zone
)
486 #ifdef NODE_NOT_IN_PAGE_FLAGS
487 extern int page_to_nid(struct page
*page
);
489 static inline int page_to_nid(struct page
*page
)
491 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
495 static inline struct zone
*page_zone(struct page
*page
)
497 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
500 static inline unsigned long page_to_section(struct page
*page
)
502 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
505 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
507 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
508 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
511 static inline void set_page_node(struct page
*page
, unsigned long node
)
513 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
514 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
517 static inline void set_page_section(struct page
*page
, unsigned long section
)
519 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
520 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
523 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
524 unsigned long node
, unsigned long pfn
)
526 set_page_zone(page
, zone
);
527 set_page_node(page
, node
);
528 set_page_section(page
, pfn_to_section_nr(pfn
));
532 * Some inline functions in vmstat.h depend on page_zone()
534 #include <linux/vmstat.h>
536 static __always_inline
void *lowmem_page_address(struct page
*page
)
538 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
541 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
542 #define HASHED_PAGE_VIRTUAL
545 #if defined(WANT_PAGE_VIRTUAL)
546 #define page_address(page) ((page)->virtual)
547 #define set_page_address(page, address) \
549 (page)->virtual = (address); \
551 #define page_address_init() do { } while(0)
554 #if defined(HASHED_PAGE_VIRTUAL)
555 void *page_address(struct page
*page
);
556 void set_page_address(struct page
*page
, void *virtual);
557 void page_address_init(void);
560 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
561 #define page_address(page) lowmem_page_address(page)
562 #define set_page_address(page, address) do { } while(0)
563 #define page_address_init() do { } while(0)
567 * On an anonymous page mapped into a user virtual memory area,
568 * page->mapping points to its anon_vma, not to a struct address_space;
569 * with the PAGE_MAPPING_ANON bit set to distinguish it.
571 * Please note that, confusingly, "page_mapping" refers to the inode
572 * address_space which maps the page from disk; whereas "page_mapped"
573 * refers to user virtual address space into which the page is mapped.
575 #define PAGE_MAPPING_ANON 1
577 extern struct address_space swapper_space
;
578 static inline struct address_space
*page_mapping(struct page
*page
)
580 struct address_space
*mapping
= page
->mapping
;
582 if (unlikely(PageSwapCache(page
)))
583 mapping
= &swapper_space
;
584 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
589 static inline int PageAnon(struct page
*page
)
591 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
595 * Return the pagecache index of the passed page. Regular pagecache pages
596 * use ->index whereas swapcache pages use ->private
598 static inline pgoff_t
page_index(struct page
*page
)
600 if (unlikely(PageSwapCache(page
)))
601 return page_private(page
);
606 * The atomic page->_mapcount, like _count, starts from -1:
607 * so that transitions both from it and to it can be tracked,
608 * using atomic_inc_and_test and atomic_add_negative(-1).
610 static inline void reset_page_mapcount(struct page
*page
)
612 atomic_set(&(page
)->_mapcount
, -1);
615 static inline int page_mapcount(struct page
*page
)
617 return atomic_read(&(page
)->_mapcount
) + 1;
621 * Return true if this page is mapped into pagetables.
623 static inline int page_mapped(struct page
*page
)
625 return atomic_read(&(page
)->_mapcount
) >= 0;
629 * Error return values for the *_nopage functions
631 #define NOPAGE_SIGBUS (NULL)
632 #define NOPAGE_OOM ((struct page *) (-1))
633 #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
636 * Error return values for the *_nopfn functions
638 #define NOPFN_SIGBUS ((unsigned long) -1)
639 #define NOPFN_OOM ((unsigned long) -2)
642 * Different kinds of faults, as returned by handle_mm_fault().
643 * Used to decide whether a process gets delivered SIGBUS or
644 * just gets major/minor fault counters bumped up.
646 #define VM_FAULT_OOM 0x00
647 #define VM_FAULT_SIGBUS 0x01
648 #define VM_FAULT_MINOR 0x02
649 #define VM_FAULT_MAJOR 0x03
652 * Special case for get_user_pages.
653 * Must be in a distinct bit from the above VM_FAULT_ flags.
655 #define VM_FAULT_WRITE 0x10
657 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
659 extern void show_free_areas(void);
662 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
663 unsigned long address
, int *type
);
664 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
665 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
667 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
669 #define shmem_nopage filemap_nopage
671 static inline int shmem_lock(struct file
*file
, int lock
,
672 struct user_struct
*user
)
677 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
678 struct mempolicy
*new)
683 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
689 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
690 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
692 int shmem_zero_setup(struct vm_area_struct
*);
695 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
699 unsigned long flags
);
702 static inline int can_do_mlock(void)
704 if (capable(CAP_IPC_LOCK
))
706 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
710 extern int user_shm_lock(size_t, struct user_struct
*);
711 extern void user_shm_unlock(size_t, struct user_struct
*);
714 * Parameter block passed down to zap_pte_range in exceptional cases.
717 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
718 struct address_space
*check_mapping
; /* Check page->mapping if set */
719 pgoff_t first_index
; /* Lowest page->index to unmap */
720 pgoff_t last_index
; /* Highest page->index to unmap */
721 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
722 unsigned long truncate_count
; /* Compare vm_truncate_count */
725 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
726 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
727 unsigned long size
, struct zap_details
*);
728 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
729 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
730 unsigned long end_addr
, unsigned long *nr_accounted
,
731 struct zap_details
*);
732 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
733 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
734 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
735 unsigned long floor
, unsigned long ceiling
);
736 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
737 struct vm_area_struct
*vma
);
738 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
739 unsigned long size
, pgprot_t prot
);
740 void unmap_mapping_range(struct address_space
*mapping
,
741 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
743 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
744 loff_t
const holebegin
, loff_t
const holelen
)
746 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
749 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
750 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
751 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
752 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
755 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
756 unsigned long address
, int write_access
);
758 static inline int handle_mm_fault(struct mm_struct
*mm
,
759 struct vm_area_struct
*vma
, unsigned long address
,
762 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
766 static inline int handle_mm_fault(struct mm_struct
*mm
,
767 struct vm_area_struct
*vma
, unsigned long address
,
770 /* should never happen if there's no MMU */
772 return VM_FAULT_SIGBUS
;
776 extern int make_pages_present(unsigned long addr
, unsigned long end
);
777 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
778 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
780 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
781 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
782 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
784 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
785 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
787 int __set_page_dirty_nobuffers(struct page
*page
);
788 int redirty_page_for_writepage(struct writeback_control
*wbc
,
790 int FASTCALL(set_page_dirty(struct page
*page
));
791 int set_page_dirty_lock(struct page
*page
);
792 int clear_page_dirty_for_io(struct page
*page
);
794 extern unsigned long do_mremap(unsigned long addr
,
795 unsigned long old_len
, unsigned long new_len
,
796 unsigned long flags
, unsigned long new_addr
);
799 * Prototype to add a shrinker callback for ageable caches.
801 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
802 * scan `nr_to_scan' objects, attempting to free them.
804 * The callback must return the number of objects which remain in the cache.
806 * The callback will be passed nr_to_scan == 0 when the VM is querying the
807 * cache size, so a fastpath for that case is appropriate.
809 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
812 * Add an aging callback. The int is the number of 'seeks' it takes
813 * to recreate one of the objects that these functions age.
816 #define DEFAULT_SEEKS 2
818 extern struct shrinker
*set_shrinker(int, shrinker_t
);
819 extern void remove_shrinker(struct shrinker
*shrinker
);
822 * Some shared mappigns will want the pages marked read-only
823 * to track write events. If so, we'll downgrade vm_page_prot
824 * to the private version (using protection_map[] without the
827 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
829 unsigned int vm_flags
= vma
->vm_flags
;
831 /* If it was private or non-writable, the write bit is already clear */
832 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
835 /* The backer wishes to know when pages are first written to? */
836 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
839 /* The open routine did something to the protections already? */
840 if (pgprot_val(vma
->vm_page_prot
) !=
841 pgprot_val(protection_map
[vm_flags
&
842 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
845 /* Specialty mapping? */
846 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
849 /* Can the mapping track the dirty pages? */
850 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
851 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
854 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
856 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
857 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
858 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
859 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
862 * The following ifdef needed to get the 4level-fixup.h header to work.
863 * Remove it when 4level-fixup.h has been removed.
865 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
866 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
868 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
869 NULL
: pud_offset(pgd
, address
);
872 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
874 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
875 NULL
: pmd_offset(pud
, address
);
877 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
879 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
881 * We tuck a spinlock to guard each pagetable page into its struct page,
882 * at page->private, with BUILD_BUG_ON to make sure that this will not
883 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
884 * When freeing, reset page->mapping so free_pages_check won't complain.
886 #define __pte_lockptr(page) &((page)->ptl)
887 #define pte_lock_init(_page) do { \
888 spin_lock_init(__pte_lockptr(_page)); \
890 #define pte_lock_deinit(page) ((page)->mapping = NULL)
891 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
894 * We use mm->page_table_lock to guard all pagetable pages of the mm.
896 #define pte_lock_init(page) do {} while (0)
897 #define pte_lock_deinit(page) do {} while (0)
898 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
899 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
901 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
903 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
904 pte_t *__pte = pte_offset_map(pmd, address); \
910 #define pte_unmap_unlock(pte, ptl) do { \
915 #define pte_alloc_map(mm, pmd, address) \
916 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
917 NULL: pte_offset_map(pmd, address))
919 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
920 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
921 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
923 #define pte_alloc_kernel(pmd, address) \
924 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
925 NULL: pte_offset_kernel(pmd, address))
927 extern void free_area_init(unsigned long * zones_size
);
928 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
929 unsigned long * zones_size
, unsigned long zone_start_pfn
,
930 unsigned long *zholes_size
);
931 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
933 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
934 * zones, allocate the backing mem_map and account for memory holes in a more
935 * architecture independent manner. This is a substitute for creating the
936 * zone_sizes[] and zholes_size[] arrays and passing them to
937 * free_area_init_node()
939 * An architecture is expected to register range of page frames backed by
940 * physical memory with add_active_range() before calling
941 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
942 * usage, an architecture is expected to do something like
944 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
946 * for_each_valid_physical_page_range()
947 * add_active_range(node_id, start_pfn, end_pfn)
948 * free_area_init_nodes(max_zone_pfns);
950 * If the architecture guarantees that there are no holes in the ranges
951 * registered with add_active_range(), free_bootmem_active_regions()
952 * will call free_bootmem_node() for each registered physical page range.
953 * Similarly sparse_memory_present_with_active_regions() calls
954 * memory_present() for each range when SPARSEMEM is enabled.
956 * See mm/page_alloc.c for more information on each function exposed by
957 * CONFIG_ARCH_POPULATES_NODE_MAP
959 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
960 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
961 unsigned long end_pfn
);
962 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
963 unsigned long new_end_pfn
);
964 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
965 unsigned long end_pfn
);
966 extern void remove_all_active_ranges(void);
967 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
968 unsigned long end_pfn
);
969 extern void get_pfn_range_for_nid(unsigned int nid
,
970 unsigned long *start_pfn
, unsigned long *end_pfn
);
971 extern unsigned long find_min_pfn_with_active_regions(void);
972 extern unsigned long find_max_pfn_with_active_regions(void);
973 extern void free_bootmem_with_active_regions(int nid
,
974 unsigned long max_low_pfn
);
975 extern void sparse_memory_present_with_active_regions(int nid
);
976 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
977 extern int early_pfn_to_nid(unsigned long pfn
);
978 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
979 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
980 extern void set_dma_reserve(unsigned long new_dma_reserve
);
981 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
982 extern void setup_per_zone_pages_min(void);
983 extern void mem_init(void);
984 extern void show_mem(void);
985 extern void si_meminfo(struct sysinfo
* val
);
986 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
989 extern void setup_per_cpu_pageset(void);
991 static inline void setup_per_cpu_pageset(void) {}
995 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
996 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
997 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
998 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
999 struct prio_tree_iter
*iter
);
1001 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1002 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1003 (vma = vma_prio_tree_next(vma, iter)); )
1005 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1006 struct list_head
*list
)
1008 vma
->shared
.vm_set
.parent
= NULL
;
1009 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1013 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
1014 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1015 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1016 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1017 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1018 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1019 struct mempolicy
*);
1020 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1021 extern int split_vma(struct mm_struct
*,
1022 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1023 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1024 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1025 struct rb_node
**, struct rb_node
*);
1026 extern void unlink_file_vma(struct vm_area_struct
*);
1027 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1028 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1029 extern void exit_mmap(struct mm_struct
*);
1030 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1032 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1034 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1035 unsigned long len
, unsigned long prot
,
1036 unsigned long flag
, unsigned long pgoff
);
1038 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1039 unsigned long len
, unsigned long prot
,
1040 unsigned long flag
, unsigned long offset
)
1042 unsigned long ret
= -EINVAL
;
1043 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1045 if (!(offset
& ~PAGE_MASK
))
1046 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1051 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1053 extern unsigned long do_brk(unsigned long, unsigned long);
1056 extern unsigned long page_unuse(struct page
*);
1057 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1058 extern void truncate_inode_pages_range(struct address_space
*,
1059 loff_t lstart
, loff_t lend
);
1061 /* generic vm_area_ops exported for stackable file systems */
1062 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
1063 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
1064 unsigned long, pgprot_t
, unsigned long, int);
1066 /* mm/page-writeback.c */
1067 int write_one_page(struct page
*page
, int wait
);
1070 #define VM_MAX_READAHEAD 128 /* kbytes */
1071 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1072 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1073 * turning readahead off */
1075 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1076 pgoff_t offset
, unsigned long nr_to_read
);
1077 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1078 pgoff_t offset
, unsigned long nr_to_read
);
1079 unsigned long page_cache_readahead(struct address_space
*mapping
,
1080 struct file_ra_state
*ra
,
1083 unsigned long size
);
1084 void handle_ra_miss(struct address_space
*mapping
,
1085 struct file_ra_state
*ra
, pgoff_t offset
);
1086 unsigned long max_sane_readahead(unsigned long nr
);
1088 /* Do stack extension */
1089 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1091 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1094 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1095 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1096 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1097 struct vm_area_struct
**pprev
);
1099 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1100 NULL if none. Assume start_addr < end_addr. */
1101 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1103 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1105 if (vma
&& end_addr
<= vma
->vm_start
)
1110 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1112 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1115 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1116 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1117 struct page
*vmalloc_to_page(void *addr
);
1118 unsigned long vmalloc_to_pfn(void *addr
);
1119 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1120 unsigned long pfn
, unsigned long size
, pgprot_t
);
1121 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1123 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1124 unsigned int foll_flags
);
1125 #define FOLL_WRITE 0x01 /* check pte is writable */
1126 #define FOLL_TOUCH 0x02 /* mark page accessed */
1127 #define FOLL_GET 0x04 /* do get_page on page */
1128 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1130 #ifdef CONFIG_PROC_FS
1131 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1133 static inline void vm_stat_account(struct mm_struct
*mm
,
1134 unsigned long flags
, struct file
*file
, long pages
)
1137 #endif /* CONFIG_PROC_FS */
1139 #ifndef CONFIG_DEBUG_PAGEALLOC
1141 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1144 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1145 #ifdef __HAVE_ARCH_GATE_AREA
1146 int in_gate_area_no_task(unsigned long addr
);
1147 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1149 int in_gate_area_no_task(unsigned long addr
);
1150 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1151 #endif /* __HAVE_ARCH_GATE_AREA */
1153 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1154 void __user
*, size_t *, loff_t
*);
1155 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1156 unsigned long lru_pages
);
1157 void drop_pagecache(void);
1158 void drop_slab(void);
1161 #define randomize_va_space 0
1163 extern int randomize_va_space
;
1166 __attribute__((weak
)) const char *arch_vma_name(struct vm_area_struct
*vma
);
1168 #endif /* __KERNEL__ */
1169 #endif /* _LINUX_MM_H */