4 #include <linux/errno.h>
5 #include <linux/capability.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/prio_tree.h>
15 #include <linux/mutex.h>
16 #include <linux/debug_locks.h>
17 #include <linux/backing-dev.h>
18 #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 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 */
116 extern struct kmem_cache
*vm_area_cachep
;
119 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
120 * disabled, then there's a single shared list of VMAs maintained by the
121 * system, and mm's subscribe to these individually
123 struct vm_list_struct
{
124 struct vm_list_struct
*next
;
125 struct vm_area_struct
*vma
;
129 extern struct rb_root nommu_vma_tree
;
130 extern struct rw_semaphore nommu_vma_sem
;
132 extern unsigned int kobjsize(const void *objp
);
138 #define VM_READ 0x00000001 /* currently active flags */
139 #define VM_WRITE 0x00000002
140 #define VM_EXEC 0x00000004
141 #define VM_SHARED 0x00000008
143 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
144 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
145 #define VM_MAYWRITE 0x00000020
146 #define VM_MAYEXEC 0x00000040
147 #define VM_MAYSHARE 0x00000080
149 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
150 #define VM_GROWSUP 0x00000200
151 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
152 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
154 #define VM_EXECUTABLE 0x00001000
155 #define VM_LOCKED 0x00002000
156 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
158 /* Used by sys_madvise() */
159 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
160 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
162 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
163 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
164 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
165 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
166 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
167 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
168 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
169 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
170 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
172 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
174 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
175 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
178 #ifdef CONFIG_STACK_GROWSUP
179 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
181 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
184 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
185 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
186 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
187 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
188 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
191 * mapping from the currently active vm_flags protection bits (the
192 * low four bits) to a page protection mask..
194 extern pgprot_t protection_map
[16];
196 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
197 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
201 * vm_fault is filled by the the pagefault handler and passed to the vma's
202 * ->fault function. The vma's ->fault is responsible for returning a bitmask
203 * of VM_FAULT_xxx flags that give details about how the fault was handled.
205 * pgoff should be used in favour of virtual_address, if possible. If pgoff
206 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
210 unsigned int flags
; /* FAULT_FLAG_xxx flags */
211 pgoff_t pgoff
; /* Logical page offset based on vma */
212 void __user
*virtual_address
; /* Faulting virtual address */
214 struct page
*page
; /* ->fault handlers should return a
215 * page here, unless VM_FAULT_NOPAGE
216 * is set (which is also implied by
222 * These are the virtual MM functions - opening of an area, closing and
223 * unmapping it (needed to keep files on disk up-to-date etc), pointer
224 * to the functions called when a no-page or a wp-page exception occurs.
226 struct vm_operations_struct
{
227 void (*open
)(struct vm_area_struct
* area
);
228 void (*close
)(struct vm_area_struct
* area
);
229 struct page
*(*nopage
)(struct vm_area_struct
*area
,
230 unsigned long address
, int *type
);
231 unsigned long (*nopfn
)(struct vm_area_struct
*area
,
232 unsigned long address
);
233 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
235 /* notification that a previously read-only page is about to become
236 * writable, if an error is returned it will cause a SIGBUS */
237 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
239 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
240 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
242 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
243 const nodemask_t
*to
, unsigned long flags
);
250 #define page_private(page) ((page)->private)
251 #define set_page_private(page, v) ((page)->private = (v))
254 * FIXME: take this include out, include page-flags.h in
255 * files which need it (119 of them)
257 #include <linux/page-flags.h>
259 #ifdef CONFIG_DEBUG_VM
260 #define VM_BUG_ON(cond) BUG_ON(cond)
262 #define VM_BUG_ON(condition) do { } while(0)
266 * Methods to modify the page usage count.
268 * What counts for a page usage:
269 * - cache mapping (page->mapping)
270 * - private data (page->private)
271 * - page mapped in a task's page tables, each mapping
272 * is counted separately
274 * Also, many kernel routines increase the page count before a critical
275 * routine so they can be sure the page doesn't go away from under them.
279 * Drop a ref, return true if the refcount fell to zero (the page has no users)
281 static inline int put_page_testzero(struct page
*page
)
283 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
284 return atomic_dec_and_test(&page
->_count
);
288 * Try to grab a ref unless the page has a refcount of zero, return false if
291 static inline int get_page_unless_zero(struct page
*page
)
293 VM_BUG_ON(PageTail(page
));
294 return atomic_inc_not_zero(&page
->_count
);
297 static inline struct page
*compound_head(struct page
*page
)
299 if (unlikely(PageTail(page
)))
300 return page
->first_page
;
304 static inline int page_count(struct page
*page
)
306 return atomic_read(&compound_head(page
)->_count
);
309 static inline void get_page(struct page
*page
)
311 page
= compound_head(page
);
312 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
313 atomic_inc(&page
->_count
);
316 static inline struct page
*virt_to_head_page(const void *x
)
318 struct page
*page
= virt_to_page(x
);
319 return compound_head(page
);
323 * Setup the page count before being freed into the page allocator for
324 * the first time (boot or memory hotplug)
326 static inline void init_page_count(struct page
*page
)
328 atomic_set(&page
->_count
, 1);
331 void put_page(struct page
*page
);
332 void put_pages_list(struct list_head
*pages
);
334 void split_page(struct page
*page
, unsigned int order
);
337 * Compound pages have a destructor function. Provide a
338 * prototype for that function and accessor functions.
339 * These are _only_ valid on the head of a PG_compound page.
341 typedef void compound_page_dtor(struct page
*);
343 static inline void set_compound_page_dtor(struct page
*page
,
344 compound_page_dtor
*dtor
)
346 page
[1].lru
.next
= (void *)dtor
;
349 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
351 return (compound_page_dtor
*)page
[1].lru
.next
;
354 static inline int compound_order(struct page
*page
)
358 return (unsigned long)page
[1].lru
.prev
;
361 static inline void set_compound_order(struct page
*page
, unsigned long order
)
363 page
[1].lru
.prev
= (void *)order
;
367 * Multiple processes may "see" the same page. E.g. for untouched
368 * mappings of /dev/null, all processes see the same page full of
369 * zeroes, and text pages of executables and shared libraries have
370 * only one copy in memory, at most, normally.
372 * For the non-reserved pages, page_count(page) denotes a reference count.
373 * page_count() == 0 means the page is free. page->lru is then used for
374 * freelist management in the buddy allocator.
375 * page_count() > 0 means the page has been allocated.
377 * Pages are allocated by the slab allocator in order to provide memory
378 * to kmalloc and kmem_cache_alloc. In this case, the management of the
379 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
380 * unless a particular usage is carefully commented. (the responsibility of
381 * freeing the kmalloc memory is the caller's, of course).
383 * A page may be used by anyone else who does a __get_free_page().
384 * In this case, page_count still tracks the references, and should only
385 * be used through the normal accessor functions. The top bits of page->flags
386 * and page->virtual store page management information, but all other fields
387 * are unused and could be used privately, carefully. The management of this
388 * page is the responsibility of the one who allocated it, and those who have
389 * subsequently been given references to it.
391 * The other pages (we may call them "pagecache pages") are completely
392 * managed by the Linux memory manager: I/O, buffers, swapping etc.
393 * The following discussion applies only to them.
395 * A pagecache page contains an opaque `private' member, which belongs to the
396 * page's address_space. Usually, this is the address of a circular list of
397 * the page's disk buffers. PG_private must be set to tell the VM to call
398 * into the filesystem to release these pages.
400 * A page may belong to an inode's memory mapping. In this case, page->mapping
401 * is the pointer to the inode, and page->index is the file offset of the page,
402 * in units of PAGE_CACHE_SIZE.
404 * If pagecache pages are not associated with an inode, they are said to be
405 * anonymous pages. These may become associated with the swapcache, and in that
406 * case PG_swapcache is set, and page->private is an offset into the swapcache.
408 * In either case (swapcache or inode backed), the pagecache itself holds one
409 * reference to the page. Setting PG_private should also increment the
410 * refcount. The each user mapping also has a reference to the page.
412 * The pagecache pages are stored in a per-mapping radix tree, which is
413 * rooted at mapping->page_tree, and indexed by offset.
414 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
415 * lists, we instead now tag pages as dirty/writeback in the radix tree.
417 * All pagecache pages may be subject to I/O:
418 * - inode pages may need to be read from disk,
419 * - inode pages which have been modified and are MAP_SHARED may need
420 * to be written back to the inode on disk,
421 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
422 * modified may need to be swapped out to swap space and (later) to be read
427 * The zone field is never updated after free_area_init_core()
428 * sets it, so none of the operations on it need to be atomic.
433 * page->flags layout:
435 * There are three possibilities for how page->flags get
436 * laid out. The first is for the normal case, without
437 * sparsemem. The second is for sparsemem when there is
438 * plenty of space for node and section. The last is when
439 * we have run out of space and have to fall back to an
440 * alternate (slower) way of determining the node.
442 * No sparsemem: | NODE | ZONE | ... | FLAGS |
443 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
444 * no space for node: | SECTION | ZONE | ... | FLAGS |
446 #ifdef CONFIG_SPARSEMEM
447 #define SECTIONS_WIDTH SECTIONS_SHIFT
449 #define SECTIONS_WIDTH 0
452 #define ZONES_WIDTH ZONES_SHIFT
454 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
455 #define NODES_WIDTH NODES_SHIFT
457 #define NODES_WIDTH 0
460 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
461 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
462 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
463 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
466 * We are going to use the flags for the page to node mapping if its in
467 * there. This includes the case where there is no node, so it is implicit.
469 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
470 #define NODE_NOT_IN_PAGE_FLAGS
473 #ifndef PFN_SECTION_SHIFT
474 #define PFN_SECTION_SHIFT 0
478 * Define the bit shifts to access each section. For non-existant
479 * sections we define the shift as 0; that plus a 0 mask ensures
480 * the compiler will optimise away reference to them.
482 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
483 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
484 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
486 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
487 #ifdef NODE_NOT_IN_PAGEFLAGS
488 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
489 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
490 SECTIONS_PGOFF : ZONES_PGOFF)
492 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
493 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
494 NODES_PGOFF : ZONES_PGOFF)
497 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
499 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
500 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
503 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
504 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
505 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
506 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
508 static inline enum zone_type
page_zonenum(struct page
*page
)
510 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
514 * The identification function is only used by the buddy allocator for
515 * determining if two pages could be buddies. We are not really
516 * identifying a zone since we could be using a the section number
517 * id if we have not node id available in page flags.
518 * We guarantee only that it will return the same value for two
519 * combinable pages in a zone.
521 static inline int page_zone_id(struct page
*page
)
523 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
526 static inline int zone_to_nid(struct zone
*zone
)
535 #ifdef NODE_NOT_IN_PAGE_FLAGS
536 extern int page_to_nid(struct page
*page
);
538 static inline int page_to_nid(struct page
*page
)
540 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
544 static inline struct zone
*page_zone(struct page
*page
)
546 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
549 static inline unsigned long page_to_section(struct page
*page
)
551 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
554 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
556 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
557 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
560 static inline void set_page_node(struct page
*page
, unsigned long node
)
562 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
563 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
566 static inline void set_page_section(struct page
*page
, unsigned long section
)
568 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
569 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
572 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
573 unsigned long node
, unsigned long pfn
)
575 set_page_zone(page
, zone
);
576 set_page_node(page
, node
);
577 set_page_section(page
, pfn_to_section_nr(pfn
));
581 * Some inline functions in vmstat.h depend on page_zone()
583 #include <linux/vmstat.h>
585 static __always_inline
void *lowmem_page_address(struct page
*page
)
587 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
590 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
591 #define HASHED_PAGE_VIRTUAL
594 #if defined(WANT_PAGE_VIRTUAL)
595 #define page_address(page) ((page)->virtual)
596 #define set_page_address(page, address) \
598 (page)->virtual = (address); \
600 #define page_address_init() do { } while(0)
603 #if defined(HASHED_PAGE_VIRTUAL)
604 void *page_address(struct page
*page
);
605 void set_page_address(struct page
*page
, void *virtual);
606 void page_address_init(void);
609 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
610 #define page_address(page) lowmem_page_address(page)
611 #define set_page_address(page, address) do { } while(0)
612 #define page_address_init() do { } while(0)
616 * On an anonymous page mapped into a user virtual memory area,
617 * page->mapping points to its anon_vma, not to a struct address_space;
618 * with the PAGE_MAPPING_ANON bit set to distinguish it.
620 * Please note that, confusingly, "page_mapping" refers to the inode
621 * address_space which maps the page from disk; whereas "page_mapped"
622 * refers to user virtual address space into which the page is mapped.
624 #define PAGE_MAPPING_ANON 1
626 extern struct address_space swapper_space
;
627 static inline struct address_space
*page_mapping(struct page
*page
)
629 struct address_space
*mapping
= page
->mapping
;
631 if (unlikely(PageSwapCache(page
)))
632 mapping
= &swapper_space
;
634 else if (unlikely(PageSlab(page
)))
637 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
642 static inline int PageAnon(struct page
*page
)
644 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
648 * Return the pagecache index of the passed page. Regular pagecache pages
649 * use ->index whereas swapcache pages use ->private
651 static inline pgoff_t
page_index(struct page
*page
)
653 if (unlikely(PageSwapCache(page
)))
654 return page_private(page
);
659 * The atomic page->_mapcount, like _count, starts from -1:
660 * so that transitions both from it and to it can be tracked,
661 * using atomic_inc_and_test and atomic_add_negative(-1).
663 static inline void reset_page_mapcount(struct page
*page
)
665 atomic_set(&(page
)->_mapcount
, -1);
668 static inline int page_mapcount(struct page
*page
)
670 return atomic_read(&(page
)->_mapcount
) + 1;
674 * Return true if this page is mapped into pagetables.
676 static inline int page_mapped(struct page
*page
)
678 return atomic_read(&(page
)->_mapcount
) >= 0;
682 * Error return values for the *_nopage functions
684 #define NOPAGE_SIGBUS (NULL)
685 #define NOPAGE_OOM ((struct page *) (-1))
688 * Error return values for the *_nopfn functions
690 #define NOPFN_SIGBUS ((unsigned long) -1)
691 #define NOPFN_OOM ((unsigned long) -2)
692 #define NOPFN_REFAULT ((unsigned long) -3)
695 * Different kinds of faults, as returned by handle_mm_fault().
696 * Used to decide whether a process gets delivered SIGBUS or
697 * just gets major/minor fault counters bumped up.
700 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
702 #define VM_FAULT_OOM 0x0001
703 #define VM_FAULT_SIGBUS 0x0002
704 #define VM_FAULT_MAJOR 0x0004
705 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
707 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
708 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
710 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS)
712 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
714 extern void show_free_areas(void);
717 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
718 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
720 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
722 static inline int shmem_lock(struct file
*file
, int lock
,
723 struct user_struct
*user
)
728 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
729 struct mempolicy
*new)
734 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
740 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
742 int shmem_zero_setup(struct vm_area_struct
*);
745 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
749 unsigned long flags
);
752 extern int can_do_mlock(void);
753 extern int user_shm_lock(size_t, struct user_struct
*);
754 extern void user_shm_unlock(size_t, struct user_struct
*);
757 * Parameter block passed down to zap_pte_range in exceptional cases.
760 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
761 struct address_space
*check_mapping
; /* Check page->mapping if set */
762 pgoff_t first_index
; /* Lowest page->index to unmap */
763 pgoff_t last_index
; /* Highest page->index to unmap */
764 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
765 unsigned long truncate_count
; /* Compare vm_truncate_count */
768 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
769 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
770 unsigned long size
, struct zap_details
*);
771 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
772 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
773 unsigned long end_addr
, unsigned long *nr_accounted
,
774 struct zap_details
*);
775 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
776 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
777 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
778 unsigned long floor
, unsigned long ceiling
);
779 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
780 struct vm_area_struct
*vma
);
781 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
782 unsigned long size
, pgprot_t prot
);
783 void unmap_mapping_range(struct address_space
*mapping
,
784 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
786 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
787 loff_t
const holebegin
, loff_t
const holelen
)
789 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
792 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
793 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
796 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
797 unsigned long address
, int write_access
);
799 static inline int handle_mm_fault(struct mm_struct
*mm
,
800 struct vm_area_struct
*vma
, unsigned long address
,
803 /* should never happen if there's no MMU */
805 return VM_FAULT_SIGBUS
;
809 extern int make_pages_present(unsigned long addr
, unsigned long end
);
810 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
811 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
813 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
814 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
815 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
817 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
818 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
820 int __set_page_dirty_nobuffers(struct page
*page
);
821 int __set_page_dirty_no_writeback(struct page
*page
);
822 int redirty_page_for_writepage(struct writeback_control
*wbc
,
824 int FASTCALL(set_page_dirty(struct page
*page
));
825 int set_page_dirty_lock(struct page
*page
);
826 int clear_page_dirty_for_io(struct page
*page
);
828 extern unsigned long do_mremap(unsigned long addr
,
829 unsigned long old_len
, unsigned long new_len
,
830 unsigned long flags
, unsigned long new_addr
);
833 * Prototype to add a shrinker callback for ageable caches.
835 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
836 * scan `nr_to_scan' objects, attempting to free them.
838 * The callback must return the number of objects which remain in the cache.
840 * The callback will be passed nr_to_scan == 0 when the VM is querying the
841 * cache size, so a fastpath for that case is appropriate.
843 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
846 * Add an aging callback. The int is the number of 'seeks' it takes
847 * to recreate one of the objects that these functions age.
850 #define DEFAULT_SEEKS 2
852 extern struct shrinker
*set_shrinker(int, shrinker_t
);
853 extern void remove_shrinker(struct shrinker
*shrinker
);
856 * Some shared mappigns will want the pages marked read-only
857 * to track write events. If so, we'll downgrade vm_page_prot
858 * to the private version (using protection_map[] without the
861 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
863 unsigned int vm_flags
= vma
->vm_flags
;
865 /* If it was private or non-writable, the write bit is already clear */
866 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
869 /* The backer wishes to know when pages are first written to? */
870 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
873 /* The open routine did something to the protections already? */
874 if (pgprot_val(vma
->vm_page_prot
) !=
875 pgprot_val(protection_map
[vm_flags
&
876 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
879 /* Specialty mapping? */
880 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
883 /* Can the mapping track the dirty pages? */
884 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
885 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
888 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
890 #ifdef __PAGETABLE_PUD_FOLDED
891 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
892 unsigned long address
)
897 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
900 #ifdef __PAGETABLE_PMD_FOLDED
901 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
902 unsigned long address
)
907 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
910 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
911 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
914 * The following ifdef needed to get the 4level-fixup.h header to work.
915 * Remove it when 4level-fixup.h has been removed.
917 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
918 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
920 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
921 NULL
: pud_offset(pgd
, address
);
924 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
926 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
927 NULL
: pmd_offset(pud
, address
);
929 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
931 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
933 * We tuck a spinlock to guard each pagetable page into its struct page,
934 * at page->private, with BUILD_BUG_ON to make sure that this will not
935 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
936 * When freeing, reset page->mapping so free_pages_check won't complain.
938 #define __pte_lockptr(page) &((page)->ptl)
939 #define pte_lock_init(_page) do { \
940 spin_lock_init(__pte_lockptr(_page)); \
942 #define pte_lock_deinit(page) ((page)->mapping = NULL)
943 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
946 * We use mm->page_table_lock to guard all pagetable pages of the mm.
948 #define pte_lock_init(page) do {} while (0)
949 #define pte_lock_deinit(page) do {} while (0)
950 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
951 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
953 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
955 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
956 pte_t *__pte = pte_offset_map(pmd, address); \
962 #define pte_unmap_unlock(pte, ptl) do { \
967 #define pte_alloc_map(mm, pmd, address) \
968 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
969 NULL: pte_offset_map(pmd, address))
971 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
972 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
973 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
975 #define pte_alloc_kernel(pmd, address) \
976 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
977 NULL: pte_offset_kernel(pmd, address))
979 extern void free_area_init(unsigned long * zones_size
);
980 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
981 unsigned long * zones_size
, unsigned long zone_start_pfn
,
982 unsigned long *zholes_size
);
983 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
985 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
986 * zones, allocate the backing mem_map and account for memory holes in a more
987 * architecture independent manner. This is a substitute for creating the
988 * zone_sizes[] and zholes_size[] arrays and passing them to
989 * free_area_init_node()
991 * An architecture is expected to register range of page frames backed by
992 * physical memory with add_active_range() before calling
993 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
994 * usage, an architecture is expected to do something like
996 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
998 * for_each_valid_physical_page_range()
999 * add_active_range(node_id, start_pfn, end_pfn)
1000 * free_area_init_nodes(max_zone_pfns);
1002 * If the architecture guarantees that there are no holes in the ranges
1003 * registered with add_active_range(), free_bootmem_active_regions()
1004 * will call free_bootmem_node() for each registered physical page range.
1005 * Similarly sparse_memory_present_with_active_regions() calls
1006 * memory_present() for each range when SPARSEMEM is enabled.
1008 * See mm/page_alloc.c for more information on each function exposed by
1009 * CONFIG_ARCH_POPULATES_NODE_MAP
1011 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1012 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
1013 unsigned long end_pfn
);
1014 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
1015 unsigned long new_end_pfn
);
1016 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
1017 unsigned long end_pfn
);
1018 extern void remove_all_active_ranges(void);
1019 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1020 unsigned long end_pfn
);
1021 extern void get_pfn_range_for_nid(unsigned int nid
,
1022 unsigned long *start_pfn
, unsigned long *end_pfn
);
1023 extern unsigned long find_min_pfn_with_active_regions(void);
1024 extern unsigned long find_max_pfn_with_active_regions(void);
1025 extern void free_bootmem_with_active_regions(int nid
,
1026 unsigned long max_low_pfn
);
1027 extern void sparse_memory_present_with_active_regions(int nid
);
1028 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1029 extern int early_pfn_to_nid(unsigned long pfn
);
1030 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1031 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1032 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1033 extern void memmap_init_zone(unsigned long, int, unsigned long,
1034 unsigned long, enum memmap_context
);
1035 extern void setup_per_zone_pages_min(void);
1036 extern void mem_init(void);
1037 extern void show_mem(void);
1038 extern void si_meminfo(struct sysinfo
* val
);
1039 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1042 extern void setup_per_cpu_pageset(void);
1044 static inline void setup_per_cpu_pageset(void) {}
1048 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1049 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1050 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1051 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1052 struct prio_tree_iter
*iter
);
1054 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1055 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1056 (vma = vma_prio_tree_next(vma, iter)); )
1058 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1059 struct list_head
*list
)
1061 vma
->shared
.vm_set
.parent
= NULL
;
1062 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1066 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1067 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1068 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1069 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1070 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1071 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1072 struct mempolicy
*);
1073 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1074 extern int split_vma(struct mm_struct
*,
1075 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1076 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1077 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1078 struct rb_node
**, struct rb_node
*);
1079 extern void unlink_file_vma(struct vm_area_struct
*);
1080 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1081 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1082 extern void exit_mmap(struct mm_struct
*);
1083 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1084 extern int install_special_mapping(struct mm_struct
*mm
,
1085 unsigned long addr
, unsigned long len
,
1086 unsigned long flags
, struct page
**pages
);
1088 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1090 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1091 unsigned long len
, unsigned long prot
,
1092 unsigned long flag
, unsigned long pgoff
);
1094 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1095 unsigned long len
, unsigned long prot
,
1096 unsigned long flag
, unsigned long offset
)
1098 unsigned long ret
= -EINVAL
;
1099 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1101 if (!(offset
& ~PAGE_MASK
))
1102 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1107 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1109 extern unsigned long do_brk(unsigned long, unsigned long);
1112 extern unsigned long page_unuse(struct page
*);
1113 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1114 extern void truncate_inode_pages_range(struct address_space
*,
1115 loff_t lstart
, loff_t lend
);
1117 /* generic vm_area_ops exported for stackable file systems */
1118 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1120 /* mm/page-writeback.c */
1121 int write_one_page(struct page
*page
, int wait
);
1124 #define VM_MAX_READAHEAD 128 /* kbytes */
1125 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1126 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1127 * turning readahead off */
1129 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1130 pgoff_t offset
, unsigned long nr_to_read
);
1131 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1132 pgoff_t offset
, unsigned long nr_to_read
);
1133 unsigned long page_cache_readahead(struct address_space
*mapping
,
1134 struct file_ra_state
*ra
,
1137 unsigned long size
);
1138 void handle_ra_miss(struct address_space
*mapping
,
1139 struct file_ra_state
*ra
, pgoff_t offset
);
1140 unsigned long max_sane_readahead(unsigned long nr
);
1142 /* Do stack extension */
1143 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1145 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1148 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1149 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1150 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1151 struct vm_area_struct
**pprev
);
1153 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1154 NULL if none. Assume start_addr < end_addr. */
1155 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1157 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1159 if (vma
&& end_addr
<= vma
->vm_start
)
1164 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1166 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1169 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1170 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1171 struct page
*vmalloc_to_page(void *addr
);
1172 unsigned long vmalloc_to_pfn(void *addr
);
1173 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1174 unsigned long pfn
, unsigned long size
, pgprot_t
);
1175 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1176 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1179 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1180 unsigned int foll_flags
);
1181 #define FOLL_WRITE 0x01 /* check pte is writable */
1182 #define FOLL_TOUCH 0x02 /* mark page accessed */
1183 #define FOLL_GET 0x04 /* do get_page on page */
1184 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1186 typedef int (*pte_fn_t
)(pte_t
*pte
, struct page
*pmd_page
, unsigned long addr
,
1188 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1189 unsigned long size
, pte_fn_t fn
, void *data
);
1191 #ifdef CONFIG_PROC_FS
1192 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1194 static inline void vm_stat_account(struct mm_struct
*mm
,
1195 unsigned long flags
, struct file
*file
, long pages
)
1198 #endif /* CONFIG_PROC_FS */
1200 #ifndef CONFIG_DEBUG_PAGEALLOC
1202 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1205 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1206 #ifdef __HAVE_ARCH_GATE_AREA
1207 int in_gate_area_no_task(unsigned long addr
);
1208 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1210 int in_gate_area_no_task(unsigned long addr
);
1211 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1212 #endif /* __HAVE_ARCH_GATE_AREA */
1214 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1215 void __user
*, size_t *, loff_t
*);
1216 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1217 unsigned long lru_pages
);
1218 void drop_pagecache(void);
1219 void drop_slab(void);
1222 #define randomize_va_space 0
1224 extern int randomize_va_space
;
1227 __attribute__((weak
)) const char *arch_vma_name(struct vm_area_struct
*vma
);
1229 #endif /* __KERNEL__ */
1230 #endif /* _LINUX_MM_H */