4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
14 #include <linux/mutex.h>
15 #include <linux/debug_locks.h>
16 #include <linux/backing-dev.h>
17 #include <linux/mm_types.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 int page_cluster
;
32 extern int sysctl_legacy_va_layout
;
34 #define sysctl_legacy_va_layout 0
38 #include <asm/pgtable.h>
39 #include <asm/processor.h>
41 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44 * Linux kernel virtual memory manager primitives.
45 * The idea being to have a "virtual" mm in the same way
46 * we have a virtual fs - giving a cleaner interface to the
47 * mm details, and allowing different kinds of memory mappings
48 * (from shared memory to executable loading to arbitrary
53 * This struct defines a memory VMM memory area. There is one of these
54 * per VM-area/task. A VM area is any part of the process virtual memory
55 * space that has a special rule for the page-fault handlers (ie a shared
56 * library, the executable area etc).
58 struct vm_area_struct
{
59 struct mm_struct
* vm_mm
; /* The address space we belong to. */
60 unsigned long vm_start
; /* Our start address within vm_mm. */
61 unsigned long vm_end
; /* The first byte after our end address
64 /* linked list of VM areas per task, sorted by address */
65 struct vm_area_struct
*vm_next
;
67 pgprot_t vm_page_prot
; /* Access permissions of this VMA. */
68 unsigned long vm_flags
; /* Flags, listed below. */
73 * For areas with an address space and backing store,
74 * linkage into the address_space->i_mmap prio tree, or
75 * linkage to the list of like vmas hanging off its node, or
76 * linkage of vma in the address_space->i_mmap_nonlinear list.
80 struct list_head list
;
81 void *parent
; /* aligns with prio_tree_node parent */
82 struct vm_area_struct
*head
;
85 struct raw_prio_tree_node prio_tree_node
;
89 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
90 * list, after a COW of one of the file pages. A MAP_SHARED vma
91 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
92 * or brk vma (with NULL file) can only be in an anon_vma list.
94 struct list_head anon_vma_node
; /* Serialized by anon_vma->lock */
95 struct anon_vma
*anon_vma
; /* Serialized by page_table_lock */
97 /* Function pointers to deal with this struct. */
98 struct vm_operations_struct
* vm_ops
;
100 /* Information about our backing store: */
101 unsigned long vm_pgoff
; /* Offset (within vm_file) in PAGE_SIZE
102 units, *not* PAGE_CACHE_SIZE */
103 struct file
* vm_file
; /* File we map to (can be NULL). */
104 void * vm_private_data
; /* was vm_pte (shared mem) */
105 unsigned long vm_truncate_count
;/* truncate_count or restart_addr */
108 atomic_t vm_usage
; /* refcount (VMAs shared if !MMU) */
111 struct mempolicy
*vm_policy
; /* NUMA policy for the VMA */
115 extern struct kmem_cache
*vm_area_cachep
;
118 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
119 * disabled, then there's a single shared list of VMAs maintained by the
120 * system, and mm's subscribe to these individually
122 struct vm_list_struct
{
123 struct vm_list_struct
*next
;
124 struct vm_area_struct
*vma
;
128 extern struct rb_root nommu_vma_tree
;
129 extern struct rw_semaphore nommu_vma_sem
;
131 extern unsigned int kobjsize(const void *objp
);
137 #define VM_READ 0x00000001 /* currently active flags */
138 #define VM_WRITE 0x00000002
139 #define VM_EXEC 0x00000004
140 #define VM_SHARED 0x00000008
142 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
143 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
144 #define VM_MAYWRITE 0x00000020
145 #define VM_MAYEXEC 0x00000040
146 #define VM_MAYSHARE 0x00000080
148 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
149 #define VM_GROWSUP 0x00000200
150 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
151 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
153 #define VM_EXECUTABLE 0x00001000
154 #define VM_LOCKED 0x00002000
155 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
157 /* Used by sys_madvise() */
158 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
161 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
163 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
164 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
168 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
169 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
171 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
172 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
175 #ifdef CONFIG_STACK_GROWSUP
176 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
178 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
181 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
182 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
183 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
184 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
185 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
188 * mapping from the currently active vm_flags protection bits (the
189 * low four bits) to a page protection mask..
191 extern pgprot_t protection_map
[16];
195 * These are the virtual MM functions - opening of an area, closing and
196 * unmapping it (needed to keep files on disk up-to-date etc), pointer
197 * to the functions called when a no-page or a wp-page exception occurs.
199 struct vm_operations_struct
{
200 void (*open
)(struct vm_area_struct
* area
);
201 void (*close
)(struct vm_area_struct
* area
);
202 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
203 unsigned long (*nopfn
)(struct vm_area_struct
* area
, unsigned long address
);
204 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
206 /* notification that a previously read-only page is about to become
207 * writable, if an error is returned it will cause a SIGBUS */
208 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
210 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
211 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
213 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
214 const nodemask_t
*to
, unsigned long flags
);
221 #define page_private(page) ((page)->private)
222 #define set_page_private(page, v) ((page)->private = (v))
225 * FIXME: take this include out, include page-flags.h in
226 * files which need it (119 of them)
228 #include <linux/page-flags.h>
230 #ifdef CONFIG_DEBUG_VM
231 #define VM_BUG_ON(cond) BUG_ON(cond)
233 #define VM_BUG_ON(condition) do { } while(0)
237 * Methods to modify the page usage count.
239 * What counts for a page usage:
240 * - cache mapping (page->mapping)
241 * - private data (page->private)
242 * - page mapped in a task's page tables, each mapping
243 * is counted separately
245 * Also, many kernel routines increase the page count before a critical
246 * routine so they can be sure the page doesn't go away from under them.
250 * Drop a ref, return true if the refcount fell to zero (the page has no users)
252 static inline int put_page_testzero(struct page
*page
)
254 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
255 return atomic_dec_and_test(&page
->_count
);
259 * Try to grab a ref unless the page has a refcount of zero, return false if
262 static inline int get_page_unless_zero(struct page
*page
)
264 VM_BUG_ON(PageCompound(page
));
265 return atomic_inc_not_zero(&page
->_count
);
268 static inline struct page
*compound_head(struct page
*page
)
270 if (unlikely(PageTail(page
)))
271 return page
->first_page
;
275 static inline int page_count(struct page
*page
)
277 return atomic_read(&compound_head(page
)->_count
);
280 static inline void get_page(struct page
*page
)
282 page
= compound_head(page
);
283 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
284 atomic_inc(&page
->_count
);
287 static inline struct page
*virt_to_head_page(const void *x
)
289 struct page
*page
= virt_to_page(x
);
290 return compound_head(page
);
294 * Setup the page count before being freed into the page allocator for
295 * the first time (boot or memory hotplug)
297 static inline void init_page_count(struct page
*page
)
299 atomic_set(&page
->_count
, 1);
302 void put_page(struct page
*page
);
303 void put_pages_list(struct list_head
*pages
);
305 void split_page(struct page
*page
, unsigned int order
);
308 * Compound pages have a destructor function. Provide a
309 * prototype for that function and accessor functions.
310 * These are _only_ valid on the head of a PG_compound page.
312 typedef void compound_page_dtor(struct page
*);
314 static inline void set_compound_page_dtor(struct page
*page
,
315 compound_page_dtor
*dtor
)
317 page
[1].lru
.next
= (void *)dtor
;
320 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
322 return (compound_page_dtor
*)page
[1].lru
.next
;
325 static inline int compound_order(struct page
*page
)
329 return (unsigned long)page
[1].lru
.prev
;
332 static inline void set_compound_order(struct page
*page
, unsigned long order
)
334 page
[1].lru
.prev
= (void *)order
;
338 * Multiple processes may "see" the same page. E.g. for untouched
339 * mappings of /dev/null, all processes see the same page full of
340 * zeroes, and text pages of executables and shared libraries have
341 * only one copy in memory, at most, normally.
343 * For the non-reserved pages, page_count(page) denotes a reference count.
344 * page_count() == 0 means the page is free. page->lru is then used for
345 * freelist management in the buddy allocator.
346 * page_count() > 0 means the page has been allocated.
348 * Pages are allocated by the slab allocator in order to provide memory
349 * to kmalloc and kmem_cache_alloc. In this case, the management of the
350 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
351 * unless a particular usage is carefully commented. (the responsibility of
352 * freeing the kmalloc memory is the caller's, of course).
354 * A page may be used by anyone else who does a __get_free_page().
355 * In this case, page_count still tracks the references, and should only
356 * be used through the normal accessor functions. The top bits of page->flags
357 * and page->virtual store page management information, but all other fields
358 * are unused and could be used privately, carefully. The management of this
359 * page is the responsibility of the one who allocated it, and those who have
360 * subsequently been given references to it.
362 * The other pages (we may call them "pagecache pages") are completely
363 * managed by the Linux memory manager: I/O, buffers, swapping etc.
364 * The following discussion applies only to them.
366 * A pagecache page contains an opaque `private' member, which belongs to the
367 * page's address_space. Usually, this is the address of a circular list of
368 * the page's disk buffers. PG_private must be set to tell the VM to call
369 * into the filesystem to release these pages.
371 * A page may belong to an inode's memory mapping. In this case, page->mapping
372 * is the pointer to the inode, and page->index is the file offset of the page,
373 * in units of PAGE_CACHE_SIZE.
375 * If pagecache pages are not associated with an inode, they are said to be
376 * anonymous pages. These may become associated with the swapcache, and in that
377 * case PG_swapcache is set, and page->private is an offset into the swapcache.
379 * In either case (swapcache or inode backed), the pagecache itself holds one
380 * reference to the page. Setting PG_private should also increment the
381 * refcount. The each user mapping also has a reference to the page.
383 * The pagecache pages are stored in a per-mapping radix tree, which is
384 * rooted at mapping->page_tree, and indexed by offset.
385 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
386 * lists, we instead now tag pages as dirty/writeback in the radix tree.
388 * All pagecache pages may be subject to I/O:
389 * - inode pages may need to be read from disk,
390 * - inode pages which have been modified and are MAP_SHARED may need
391 * to be written back to the inode on disk,
392 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
393 * modified may need to be swapped out to swap space and (later) to be read
398 * The zone field is never updated after free_area_init_core()
399 * sets it, so none of the operations on it need to be atomic.
404 * page->flags layout:
406 * There are three possibilities for how page->flags get
407 * laid out. The first is for the normal case, without
408 * sparsemem. The second is for sparsemem when there is
409 * plenty of space for node and section. The last is when
410 * we have run out of space and have to fall back to an
411 * alternate (slower) way of determining the node.
413 * No sparsemem: | NODE | ZONE | ... | FLAGS |
414 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
415 * no space for node: | SECTION | ZONE | ... | FLAGS |
417 #ifdef CONFIG_SPARSEMEM
418 #define SECTIONS_WIDTH SECTIONS_SHIFT
420 #define SECTIONS_WIDTH 0
423 #define ZONES_WIDTH ZONES_SHIFT
425 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
426 #define NODES_WIDTH NODES_SHIFT
428 #define NODES_WIDTH 0
431 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
432 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
433 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
434 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
437 * We are going to use the flags for the page to node mapping if its in
438 * there. This includes the case where there is no node, so it is implicit.
440 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
441 #define NODE_NOT_IN_PAGE_FLAGS
444 #ifndef PFN_SECTION_SHIFT
445 #define PFN_SECTION_SHIFT 0
449 * Define the bit shifts to access each section. For non-existant
450 * sections we define the shift as 0; that plus a 0 mask ensures
451 * the compiler will optimise away reference to them.
453 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
454 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
455 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
457 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
458 #ifdef NODE_NOT_IN_PAGEFLAGS
459 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
460 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
461 SECTIONS_PGOFF : ZONES_PGOFF)
463 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
464 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
465 NODES_PGOFF : ZONES_PGOFF)
468 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
470 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
471 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
474 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
475 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
476 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
477 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
479 static inline enum zone_type
page_zonenum(struct page
*page
)
481 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
485 * The identification function is only used by the buddy allocator for
486 * determining if two pages could be buddies. We are not really
487 * identifying a zone since we could be using a the section number
488 * id if we have not node id available in page flags.
489 * We guarantee only that it will return the same value for two
490 * combinable pages in a zone.
492 static inline int page_zone_id(struct page
*page
)
494 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
497 static inline int zone_to_nid(struct zone
*zone
)
506 #ifdef NODE_NOT_IN_PAGE_FLAGS
507 extern int page_to_nid(struct page
*page
);
509 static inline int page_to_nid(struct page
*page
)
511 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
515 static inline struct zone
*page_zone(struct page
*page
)
517 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
520 static inline unsigned long page_to_section(struct page
*page
)
522 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
525 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
527 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
528 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
531 static inline void set_page_node(struct page
*page
, unsigned long node
)
533 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
534 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
537 static inline void set_page_section(struct page
*page
, unsigned long section
)
539 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
540 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
543 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
544 unsigned long node
, unsigned long pfn
)
546 set_page_zone(page
, zone
);
547 set_page_node(page
, node
);
548 set_page_section(page
, pfn_to_section_nr(pfn
));
552 * Some inline functions in vmstat.h depend on page_zone()
554 #include <linux/vmstat.h>
556 static __always_inline
void *lowmem_page_address(struct page
*page
)
558 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
561 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
562 #define HASHED_PAGE_VIRTUAL
565 #if defined(WANT_PAGE_VIRTUAL)
566 #define page_address(page) ((page)->virtual)
567 #define set_page_address(page, address) \
569 (page)->virtual = (address); \
571 #define page_address_init() do { } while(0)
574 #if defined(HASHED_PAGE_VIRTUAL)
575 void *page_address(struct page
*page
);
576 void set_page_address(struct page
*page
, void *virtual);
577 void page_address_init(void);
580 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
581 #define page_address(page) lowmem_page_address(page)
582 #define set_page_address(page, address) do { } while(0)
583 #define page_address_init() do { } while(0)
587 * On an anonymous page mapped into a user virtual memory area,
588 * page->mapping points to its anon_vma, not to a struct address_space;
589 * with the PAGE_MAPPING_ANON bit set to distinguish it.
591 * Please note that, confusingly, "page_mapping" refers to the inode
592 * address_space which maps the page from disk; whereas "page_mapped"
593 * refers to user virtual address space into which the page is mapped.
595 #define PAGE_MAPPING_ANON 1
597 extern struct address_space swapper_space
;
598 static inline struct address_space
*page_mapping(struct page
*page
)
600 struct address_space
*mapping
= page
->mapping
;
602 if (unlikely(PageSwapCache(page
)))
603 mapping
= &swapper_space
;
605 else if (unlikely(PageSlab(page
)))
608 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
613 static inline int PageAnon(struct page
*page
)
615 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
619 * Return the pagecache index of the passed page. Regular pagecache pages
620 * use ->index whereas swapcache pages use ->private
622 static inline pgoff_t
page_index(struct page
*page
)
624 if (unlikely(PageSwapCache(page
)))
625 return page_private(page
);
630 * The atomic page->_mapcount, like _count, starts from -1:
631 * so that transitions both from it and to it can be tracked,
632 * using atomic_inc_and_test and atomic_add_negative(-1).
634 static inline void reset_page_mapcount(struct page
*page
)
636 atomic_set(&(page
)->_mapcount
, -1);
639 static inline int page_mapcount(struct page
*page
)
641 return atomic_read(&(page
)->_mapcount
) + 1;
645 * Return true if this page is mapped into pagetables.
647 static inline int page_mapped(struct page
*page
)
649 return atomic_read(&(page
)->_mapcount
) >= 0;
653 * Error return values for the *_nopage functions
655 #define NOPAGE_SIGBUS (NULL)
656 #define NOPAGE_OOM ((struct page *) (-1))
657 #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
660 * Error return values for the *_nopfn functions
662 #define NOPFN_SIGBUS ((unsigned long) -1)
663 #define NOPFN_OOM ((unsigned long) -2)
664 #define NOPFN_REFAULT ((unsigned long) -3)
667 * Different kinds of faults, as returned by handle_mm_fault().
668 * Used to decide whether a process gets delivered SIGBUS or
669 * just gets major/minor fault counters bumped up.
671 #define VM_FAULT_OOM 0x00
672 #define VM_FAULT_SIGBUS 0x01
673 #define VM_FAULT_MINOR 0x02
674 #define VM_FAULT_MAJOR 0x03
677 * Special case for get_user_pages.
678 * Must be in a distinct bit from the above VM_FAULT_ flags.
680 #define VM_FAULT_WRITE 0x10
682 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
684 extern void show_free_areas(void);
687 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
688 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
690 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
692 static inline int shmem_lock(struct file
*file
, int lock
,
693 struct user_struct
*user
)
698 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
699 struct mempolicy
*new)
704 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
710 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
712 int shmem_zero_setup(struct vm_area_struct
*);
715 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
719 unsigned long flags
);
722 extern int can_do_mlock(void);
723 extern int user_shm_lock(size_t, struct user_struct
*);
724 extern void user_shm_unlock(size_t, struct user_struct
*);
727 * Parameter block passed down to zap_pte_range in exceptional cases.
730 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
731 struct address_space
*check_mapping
; /* Check page->mapping if set */
732 pgoff_t first_index
; /* Lowest page->index to unmap */
733 pgoff_t last_index
; /* Highest page->index to unmap */
734 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
735 unsigned long truncate_count
; /* Compare vm_truncate_count */
738 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
739 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
740 unsigned long size
, struct zap_details
*);
741 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
742 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
743 unsigned long end_addr
, unsigned long *nr_accounted
,
744 struct zap_details
*);
745 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
746 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
747 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
748 unsigned long floor
, unsigned long ceiling
);
749 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
750 struct vm_area_struct
*vma
);
751 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
752 unsigned long size
, pgprot_t prot
);
753 void unmap_mapping_range(struct address_space
*mapping
,
754 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
756 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
757 loff_t
const holebegin
, loff_t
const holelen
)
759 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
762 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
763 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
764 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
765 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
768 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
769 unsigned long address
, int write_access
);
771 static inline int handle_mm_fault(struct mm_struct
*mm
,
772 struct vm_area_struct
*vma
, unsigned long address
,
775 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
779 static inline int handle_mm_fault(struct mm_struct
*mm
,
780 struct vm_area_struct
*vma
, unsigned long address
,
783 /* should never happen if there's no MMU */
785 return VM_FAULT_SIGBUS
;
789 extern int make_pages_present(unsigned long addr
, unsigned long end
);
790 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
791 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
793 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
794 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
795 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
797 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
798 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
800 int __set_page_dirty_nobuffers(struct page
*page
);
801 int __set_page_dirty_no_writeback(struct page
*page
);
802 int redirty_page_for_writepage(struct writeback_control
*wbc
,
804 int FASTCALL(set_page_dirty(struct page
*page
));
805 int set_page_dirty_lock(struct page
*page
);
806 int clear_page_dirty_for_io(struct page
*page
);
808 extern unsigned long do_mremap(unsigned long addr
,
809 unsigned long old_len
, unsigned long new_len
,
810 unsigned long flags
, unsigned long new_addr
);
813 * Prototype to add a shrinker callback for ageable caches.
815 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
816 * scan `nr_to_scan' objects, attempting to free them.
818 * The callback must return the number of objects which remain in the cache.
820 * The callback will be passed nr_to_scan == 0 when the VM is querying the
821 * cache size, so a fastpath for that case is appropriate.
823 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
826 * Add an aging callback. The int is the number of 'seeks' it takes
827 * to recreate one of the objects that these functions age.
830 #define DEFAULT_SEEKS 2
832 extern struct shrinker
*set_shrinker(int, shrinker_t
);
833 extern void remove_shrinker(struct shrinker
*shrinker
);
836 * Some shared mappigns will want the pages marked read-only
837 * to track write events. If so, we'll downgrade vm_page_prot
838 * to the private version (using protection_map[] without the
841 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
843 unsigned int vm_flags
= vma
->vm_flags
;
845 /* If it was private or non-writable, the write bit is already clear */
846 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
849 /* The backer wishes to know when pages are first written to? */
850 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
853 /* The open routine did something to the protections already? */
854 if (pgprot_val(vma
->vm_page_prot
) !=
855 pgprot_val(protection_map
[vm_flags
&
856 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
859 /* Specialty mapping? */
860 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
863 /* Can the mapping track the dirty pages? */
864 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
865 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
868 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
870 #ifdef __PAGETABLE_PUD_FOLDED
871 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
872 unsigned long address
)
877 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
880 #ifdef __PAGETABLE_PMD_FOLDED
881 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
882 unsigned long address
)
887 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
890 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
891 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
894 * The following ifdef needed to get the 4level-fixup.h header to work.
895 * Remove it when 4level-fixup.h has been removed.
897 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
898 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
900 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
901 NULL
: pud_offset(pgd
, address
);
904 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
906 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
907 NULL
: pmd_offset(pud
, address
);
909 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
911 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
913 * We tuck a spinlock to guard each pagetable page into its struct page,
914 * at page->private, with BUILD_BUG_ON to make sure that this will not
915 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
916 * When freeing, reset page->mapping so free_pages_check won't complain.
918 #define __pte_lockptr(page) &((page)->ptl)
919 #define pte_lock_init(_page) do { \
920 spin_lock_init(__pte_lockptr(_page)); \
922 #define pte_lock_deinit(page) ((page)->mapping = NULL)
923 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
926 * We use mm->page_table_lock to guard all pagetable pages of the mm.
928 #define pte_lock_init(page) do {} while (0)
929 #define pte_lock_deinit(page) do {} while (0)
930 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
931 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
933 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
935 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
936 pte_t *__pte = pte_offset_map(pmd, address); \
942 #define pte_unmap_unlock(pte, ptl) do { \
947 #define pte_alloc_map(mm, pmd, address) \
948 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
949 NULL: pte_offset_map(pmd, address))
951 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
952 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
953 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
955 #define pte_alloc_kernel(pmd, address) \
956 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
957 NULL: pte_offset_kernel(pmd, address))
959 extern void free_area_init(unsigned long * zones_size
);
960 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
961 unsigned long * zones_size
, unsigned long zone_start_pfn
,
962 unsigned long *zholes_size
);
963 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
965 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
966 * zones, allocate the backing mem_map and account for memory holes in a more
967 * architecture independent manner. This is a substitute for creating the
968 * zone_sizes[] and zholes_size[] arrays and passing them to
969 * free_area_init_node()
971 * An architecture is expected to register range of page frames backed by
972 * physical memory with add_active_range() before calling
973 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
974 * usage, an architecture is expected to do something like
976 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
978 * for_each_valid_physical_page_range()
979 * add_active_range(node_id, start_pfn, end_pfn)
980 * free_area_init_nodes(max_zone_pfns);
982 * If the architecture guarantees that there are no holes in the ranges
983 * registered with add_active_range(), free_bootmem_active_regions()
984 * will call free_bootmem_node() for each registered physical page range.
985 * Similarly sparse_memory_present_with_active_regions() calls
986 * memory_present() for each range when SPARSEMEM is enabled.
988 * See mm/page_alloc.c for more information on each function exposed by
989 * CONFIG_ARCH_POPULATES_NODE_MAP
991 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
992 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
993 unsigned long end_pfn
);
994 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
995 unsigned long new_end_pfn
);
996 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
997 unsigned long end_pfn
);
998 extern void remove_all_active_ranges(void);
999 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1000 unsigned long end_pfn
);
1001 extern void get_pfn_range_for_nid(unsigned int nid
,
1002 unsigned long *start_pfn
, unsigned long *end_pfn
);
1003 extern unsigned long find_min_pfn_with_active_regions(void);
1004 extern unsigned long find_max_pfn_with_active_regions(void);
1005 extern void free_bootmem_with_active_regions(int nid
,
1006 unsigned long max_low_pfn
);
1007 extern void sparse_memory_present_with_active_regions(int nid
);
1008 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1009 extern int early_pfn_to_nid(unsigned long pfn
);
1010 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1011 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1012 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1013 extern void memmap_init_zone(unsigned long, int, unsigned long,
1014 unsigned long, enum memmap_context
);
1015 extern void setup_per_zone_pages_min(void);
1016 extern void mem_init(void);
1017 extern void show_mem(void);
1018 extern void si_meminfo(struct sysinfo
* val
);
1019 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1022 extern void setup_per_cpu_pageset(void);
1024 static inline void setup_per_cpu_pageset(void) {}
1028 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1029 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1030 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1031 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1032 struct prio_tree_iter
*iter
);
1034 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1035 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1036 (vma = vma_prio_tree_next(vma, iter)); )
1038 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1039 struct list_head
*list
)
1041 vma
->shared
.vm_set
.parent
= NULL
;
1042 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1046 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
1047 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1048 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1049 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1050 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1051 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1052 struct mempolicy
*);
1053 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1054 extern int split_vma(struct mm_struct
*,
1055 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1056 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1057 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1058 struct rb_node
**, struct rb_node
*);
1059 extern void unlink_file_vma(struct vm_area_struct
*);
1060 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1061 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1062 extern void exit_mmap(struct mm_struct
*);
1063 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1064 extern int install_special_mapping(struct mm_struct
*mm
,
1065 unsigned long addr
, unsigned long len
,
1066 unsigned long flags
, struct page
**pages
);
1068 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1070 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1071 unsigned long len
, unsigned long prot
,
1072 unsigned long flag
, unsigned long pgoff
);
1073 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1074 unsigned long len
, unsigned long flags
,
1075 unsigned int vm_flags
, unsigned long pgoff
,
1078 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1079 unsigned long len
, unsigned long prot
,
1080 unsigned long flag
, unsigned long offset
)
1082 unsigned long ret
= -EINVAL
;
1083 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1085 if (!(offset
& ~PAGE_MASK
))
1086 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1091 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1093 extern unsigned long do_brk(unsigned long, unsigned long);
1096 extern unsigned long page_unuse(struct page
*);
1097 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1098 extern void truncate_inode_pages_range(struct address_space
*,
1099 loff_t lstart
, loff_t lend
);
1101 /* generic vm_area_ops exported for stackable file systems */
1102 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
1103 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
1104 unsigned long, pgprot_t
, unsigned long, int);
1106 /* mm/page-writeback.c */
1107 int write_one_page(struct page
*page
, int wait
);
1110 #define VM_MAX_READAHEAD 128 /* kbytes */
1111 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1112 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1113 * turning readahead off */
1115 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1116 pgoff_t offset
, unsigned long nr_to_read
);
1117 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1118 pgoff_t offset
, unsigned long nr_to_read
);
1119 unsigned long page_cache_readahead(struct address_space
*mapping
,
1120 struct file_ra_state
*ra
,
1123 unsigned long size
);
1124 void handle_ra_miss(struct address_space
*mapping
,
1125 struct file_ra_state
*ra
, pgoff_t offset
);
1126 unsigned long max_sane_readahead(unsigned long nr
);
1128 /* Do stack extension */
1129 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1131 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1134 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1135 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1136 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1137 struct vm_area_struct
**pprev
);
1139 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1140 NULL if none. Assume start_addr < end_addr. */
1141 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1143 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1145 if (vma
&& end_addr
<= vma
->vm_start
)
1150 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1152 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1155 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1156 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1157 struct page
*vmalloc_to_page(void *addr
);
1158 unsigned long vmalloc_to_pfn(void *addr
);
1159 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1160 unsigned long pfn
, unsigned long size
, pgprot_t
);
1161 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1162 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1165 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1166 unsigned int foll_flags
);
1167 #define FOLL_WRITE 0x01 /* check pte is writable */
1168 #define FOLL_TOUCH 0x02 /* mark page accessed */
1169 #define FOLL_GET 0x04 /* do get_page on page */
1170 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1172 typedef int (*pte_fn_t
)(pte_t
*pte
, struct page
*pmd_page
, unsigned long addr
,
1174 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1175 unsigned long size
, pte_fn_t fn
, void *data
);
1177 #ifdef CONFIG_PROC_FS
1178 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1180 static inline void vm_stat_account(struct mm_struct
*mm
,
1181 unsigned long flags
, struct file
*file
, long pages
)
1184 #endif /* CONFIG_PROC_FS */
1186 #ifndef CONFIG_DEBUG_PAGEALLOC
1188 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1191 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1192 #ifdef __HAVE_ARCH_GATE_AREA
1193 int in_gate_area_no_task(unsigned long addr
);
1194 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1196 int in_gate_area_no_task(unsigned long addr
);
1197 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1198 #endif /* __HAVE_ARCH_GATE_AREA */
1200 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1201 void __user
*, size_t *, loff_t
*);
1202 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1203 unsigned long lru_pages
);
1204 void drop_pagecache(void);
1205 void drop_slab(void);
1208 #define randomize_va_space 0
1210 extern int randomize_va_space
;
1213 __attribute__((weak
)) const char *arch_vma_name(struct vm_area_struct
*vma
);
1215 #endif /* __KERNEL__ */
1216 #endif /* _LINUX_MM_H */