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 */
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 */
170 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
171 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
174 #ifdef CONFIG_STACK_GROWSUP
175 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
177 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
181 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
182 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
183 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
184 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
187 * mapping from the currently active vm_flags protection bits (the
188 * low four bits) to a page protection mask..
190 extern pgprot_t protection_map
[16];
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
198 struct vm_operations_struct
{
199 void (*open
)(struct vm_area_struct
* area
);
200 void (*close
)(struct vm_area_struct
* area
);
201 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
202 unsigned long (*nopfn
)(struct vm_area_struct
* area
, unsigned long address
);
203 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
205 /* notification that a previously read-only page is about to become
206 * writable, if an error is returned it will cause a SIGBUS */
207 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
209 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
210 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
212 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
213 const nodemask_t
*to
, unsigned long flags
);
220 #if 0 // mask by Victor Yu. 02-12-2007
221 #define page_private(page) ((page)->private)
222 #define set_page_private(page, v) ((page)->private = (v))
224 #define page_private(page) ((page)->u.xx.private)
225 #define set_page_private(page, v) ((page)->u.xx.private = (v))
229 * FIXME: take this include out, include page-flags.h in
230 * files which need it (119 of them)
232 #include <linux/page-flags.h>
234 #ifdef CONFIG_DEBUG_VM
235 #define VM_BUG_ON(cond) BUG_ON(cond)
237 #define VM_BUG_ON(condition) do { } while(0)
241 * Methods to modify the page usage count.
243 * What counts for a page usage:
244 * - cache mapping (page->mapping)
245 * - private data (page->private)
246 * - page mapped in a task's page tables, each mapping
247 * is counted separately
249 * Also, many kernel routines increase the page count before a critical
250 * routine so they can be sure the page doesn't go away from under them.
254 * Drop a ref, return true if the refcount fell to zero (the page has no users)
256 static inline int put_page_testzero(struct page
*page
)
258 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
259 return atomic_dec_and_test(&page
->_count
);
263 * Try to grab a ref unless the page has a refcount of zero, return false if
266 static inline int get_page_unless_zero(struct page
*page
)
268 VM_BUG_ON(PageCompound(page
));
269 return atomic_inc_not_zero(&page
->_count
);
272 static inline int page_count(struct page
*page
)
274 if (unlikely(PageCompound(page
)))
275 page
= (struct page
*)page_private(page
);
276 return atomic_read(&page
->_count
);
279 static inline void get_page(struct page
*page
)
281 if (unlikely(PageCompound(page
)))
282 page
= (struct page
*)page_private(page
);
283 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
284 atomic_inc(&page
->_count
);
288 * Setup the page count before being freed into the page allocator for
289 * the first time (boot or memory hotplug)
291 static inline void init_page_count(struct page
*page
)
293 atomic_set(&page
->_count
, 1);
296 void put_page(struct page
*page
);
297 void put_pages_list(struct list_head
*pages
);
299 void split_page(struct page
*page
, unsigned int order
);
302 * Multiple processes may "see" the same page. E.g. for untouched
303 * mappings of /dev/null, all processes see the same page full of
304 * zeroes, and text pages of executables and shared libraries have
305 * only one copy in memory, at most, normally.
307 * For the non-reserved pages, page_count(page) denotes a reference count.
308 * page_count() == 0 means the page is free. page->lru is then used for
309 * freelist management in the buddy allocator.
310 * page_count() > 0 means the page has been allocated.
312 * Pages are allocated by the slab allocator in order to provide memory
313 * to kmalloc and kmem_cache_alloc. In this case, the management of the
314 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
315 * unless a particular usage is carefully commented. (the responsibility of
316 * freeing the kmalloc memory is the caller's, of course).
318 * A page may be used by anyone else who does a __get_free_page().
319 * In this case, page_count still tracks the references, and should only
320 * be used through the normal accessor functions. The top bits of page->flags
321 * and page->virtual store page management information, but all other fields
322 * are unused and could be used privately, carefully. The management of this
323 * page is the responsibility of the one who allocated it, and those who have
324 * subsequently been given references to it.
326 * The other pages (we may call them "pagecache pages") are completely
327 * managed by the Linux memory manager: I/O, buffers, swapping etc.
328 * The following discussion applies only to them.
330 * A pagecache page contains an opaque `private' member, which belongs to the
331 * page's address_space. Usually, this is the address of a circular list of
332 * the page's disk buffers. PG_private must be set to tell the VM to call
333 * into the filesystem to release these pages.
335 * A page may belong to an inode's memory mapping. In this case, page->mapping
336 * is the pointer to the inode, and page->index is the file offset of the page,
337 * in units of PAGE_CACHE_SIZE.
339 * If pagecache pages are not associated with an inode, they are said to be
340 * anonymous pages. These may become associated with the swapcache, and in that
341 * case PG_swapcache is set, and page->private is an offset into the swapcache.
343 * In either case (swapcache or inode backed), the pagecache itself holds one
344 * reference to the page. Setting PG_private should also increment the
345 * refcount. The each user mapping also has a reference to the page.
347 * The pagecache pages are stored in a per-mapping radix tree, which is
348 * rooted at mapping->page_tree, and indexed by offset.
349 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
350 * lists, we instead now tag pages as dirty/writeback in the radix tree.
352 * All pagecache pages may be subject to I/O:
353 * - inode pages may need to be read from disk,
354 * - inode pages which have been modified and are MAP_SHARED may need
355 * to be written back to the inode on disk,
356 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
357 * modified may need to be swapped out to swap space and (later) to be read
362 * The zone field is never updated after free_area_init_core()
363 * sets it, so none of the operations on it need to be atomic.
368 * page->flags layout:
370 * There are three possibilities for how page->flags get
371 * laid out. The first is for the normal case, without
372 * sparsemem. The second is for sparsemem when there is
373 * plenty of space for node and section. The last is when
374 * we have run out of space and have to fall back to an
375 * alternate (slower) way of determining the node.
377 * No sparsemem: | NODE | ZONE | ... | FLAGS |
378 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
379 * no space for node: | SECTION | ZONE | ... | FLAGS |
381 #ifdef CONFIG_SPARSEMEM
382 #define SECTIONS_WIDTH SECTIONS_SHIFT
384 #define SECTIONS_WIDTH 0
387 #define ZONES_WIDTH ZONES_SHIFT
389 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
390 #define NODES_WIDTH NODES_SHIFT
392 #define NODES_WIDTH 0
395 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
396 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
397 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
398 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
401 * We are going to use the flags for the page to node mapping if its in
402 * there. This includes the case where there is no node, so it is implicit.
404 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
406 #ifndef PFN_SECTION_SHIFT
407 #define PFN_SECTION_SHIFT 0
411 * Define the bit shifts to access each section. For non-existant
412 * sections we define the shift as 0; that plus a 0 mask ensures
413 * the compiler will optimise away reference to them.
415 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
416 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
417 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
419 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
421 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
423 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
425 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
427 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
428 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
431 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
432 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
433 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
434 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
436 static inline enum zone_type
page_zonenum(struct page
*page
)
438 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
442 extern struct zone
*zone_table
[];
444 static inline int page_zone_id(struct page
*page
)
446 return (page
->flags
>> ZONETABLE_PGSHIFT
) & ZONETABLE_MASK
;
448 static inline struct zone
*page_zone(struct page
*page
)
450 return zone_table
[page_zone_id(page
)];
453 static inline unsigned long zone_to_nid(struct zone
*zone
)
462 static inline unsigned long page_to_nid(struct page
*page
)
465 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
467 return zone_to_nid(page_zone(page
));
469 static inline unsigned long page_to_section(struct page
*page
)
471 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
474 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
476 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
477 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
480 static inline void set_page_node(struct page
*page
, unsigned long node
)
482 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
483 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
485 static inline void set_page_section(struct page
*page
, unsigned long section
)
487 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
488 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
491 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
492 unsigned long node
, unsigned long pfn
)
494 set_page_zone(page
, zone
);
495 set_page_node(page
, node
);
496 set_page_section(page
, pfn_to_section_nr(pfn
));
500 * Some inline functions in vmstat.h depend on page_zone()
502 #include <linux/vmstat.h>
504 static __always_inline
void *lowmem_page_address(struct page
*page
)
506 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
509 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
510 #define HASHED_PAGE_VIRTUAL
513 #if defined(WANT_PAGE_VIRTUAL)
514 #define page_address(page) ((page)->virtual)
515 #define set_page_address(page, address) \
517 (page)->virtual = (address); \
519 #define page_address_init() do { } while(0)
522 #if defined(HASHED_PAGE_VIRTUAL)
523 void *page_address(struct page
*page
);
524 void set_page_address(struct page
*page
, void *virtual);
525 void page_address_init(void);
528 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
529 #define page_address(page) lowmem_page_address(page)
530 #define set_page_address(page, address) do { } while(0)
531 #define page_address_init() do { } while(0)
535 * On an anonymous page mapped into a user virtual memory area,
536 * page->mapping points to its anon_vma, not to a struct address_space;
537 * with the PAGE_MAPPING_ANON bit set to distinguish it.
539 * Please note that, confusingly, "page_mapping" refers to the inode
540 * address_space which maps the page from disk; whereas "page_mapped"
541 * refers to user virtual address space into which the page is mapped.
543 #define PAGE_MAPPING_ANON 1
545 extern struct address_space swapper_space
;
546 static inline struct address_space
*page_mapping(struct page
*page
)
548 #if 0 // mask by Victor Yu. 02-12-2007
549 struct address_space
*mapping
= page
->mapping
;
551 struct address_space
*mapping
= page
->u
.xx
.mapping
;
554 if (unlikely(PageSwapCache(page
)))
555 mapping
= &swapper_space
;
556 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
561 static inline int PageAnon(struct page
*page
)
563 #if 0 // mask by Victor Yu. 02-12-2007
564 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
566 return ((unsigned long)page
->u
.xx
.mapping
& PAGE_MAPPING_ANON
) != 0;
571 * Return the pagecache index of the passed page. Regular pagecache pages
572 * use ->index whereas swapcache pages use ->private
574 static inline pgoff_t
page_index(struct page
*page
)
576 if (unlikely(PageSwapCache(page
)))
577 return page_private(page
);
582 * The atomic page->_mapcount, like _count, starts from -1:
583 * so that transitions both from it and to it can be tracked,
584 * using atomic_inc_and_test and atomic_add_negative(-1).
586 static inline void reset_page_mapcount(struct page
*page
)
588 atomic_set(&(page
)->_mapcount
, -1);
591 static inline int page_mapcount(struct page
*page
)
593 return atomic_read(&(page
)->_mapcount
) + 1;
597 * Return true if this page is mapped into pagetables.
599 static inline int page_mapped(struct page
*page
)
601 return atomic_read(&(page
)->_mapcount
) >= 0;
605 * Error return values for the *_nopage functions
607 #define NOPAGE_SIGBUS (NULL)
608 #define NOPAGE_OOM ((struct page *) (-1))
609 #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */
612 * Error return values for the *_nopfn functions
614 #define NOPFN_SIGBUS ((unsigned long) -1)
615 #define NOPFN_OOM ((unsigned long) -2)
618 * Different kinds of faults, as returned by handle_mm_fault().
619 * Used to decide whether a process gets delivered SIGBUS or
620 * just gets major/minor fault counters bumped up.
622 #define VM_FAULT_OOM 0x00
623 #define VM_FAULT_SIGBUS 0x01
624 #define VM_FAULT_MINOR 0x02
625 #define VM_FAULT_MAJOR 0x03
628 * Special case for get_user_pages.
629 * Must be in a distinct bit from the above VM_FAULT_ flags.
631 #define VM_FAULT_WRITE 0x10
633 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
635 extern void show_free_areas(void);
638 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
639 unsigned long address
, int *type
);
640 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
641 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
643 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
645 #define shmem_nopage filemap_nopage
647 static inline int shmem_lock(struct file
*file
, int lock
,
648 struct user_struct
*user
)
653 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
654 struct mempolicy
*new)
659 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
665 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
666 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
668 int shmem_zero_setup(struct vm_area_struct
*);
671 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
675 unsigned long flags
);
678 static inline int can_do_mlock(void)
680 if (capable(CAP_IPC_LOCK
))
682 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
686 extern int user_shm_lock(size_t, struct user_struct
*);
687 extern void user_shm_unlock(size_t, struct user_struct
*);
690 * Parameter block passed down to zap_pte_range in exceptional cases.
693 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
694 struct address_space
*check_mapping
; /* Check page->mapping if set */
695 pgoff_t first_index
; /* Lowest page->index to unmap */
696 pgoff_t last_index
; /* Highest page->index to unmap */
697 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
698 unsigned long truncate_count
; /* Compare vm_truncate_count */
701 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
702 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
703 unsigned long size
, struct zap_details
*);
704 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
705 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
706 unsigned long end_addr
, unsigned long *nr_accounted
,
707 struct zap_details
*);
708 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
709 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
710 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
711 unsigned long floor
, unsigned long ceiling
);
712 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
713 struct vm_area_struct
*vma
);
714 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
715 unsigned long size
, pgprot_t prot
);
716 void unmap_mapping_range(struct address_space
*mapping
,
717 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
719 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
720 loff_t
const holebegin
, loff_t
const holelen
)
722 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
725 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
726 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
727 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
728 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
731 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
732 unsigned long address
, int write_access
);
734 static inline int handle_mm_fault(struct mm_struct
*mm
,
735 struct vm_area_struct
*vma
, unsigned long address
,
738 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
742 static inline int handle_mm_fault(struct mm_struct
*mm
,
743 struct vm_area_struct
*vma
, unsigned long address
,
746 /* should never happen if there's no MMU */
748 return VM_FAULT_SIGBUS
;
752 extern int make_pages_present(unsigned long addr
, unsigned long end
);
753 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
754 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
756 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
757 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
758 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
760 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
761 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
763 int __set_page_dirty_nobuffers(struct page
*page
);
764 int redirty_page_for_writepage(struct writeback_control
*wbc
,
766 int FASTCALL(set_page_dirty(struct page
*page
));
767 int set_page_dirty_lock(struct page
*page
);
768 int clear_page_dirty_for_io(struct page
*page
);
770 extern unsigned long do_mremap(unsigned long addr
,
771 unsigned long old_len
, unsigned long new_len
,
772 unsigned long flags
, unsigned long new_addr
);
775 * Prototype to add a shrinker callback for ageable caches.
777 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
778 * scan `nr_to_scan' objects, attempting to free them.
780 * The callback must return the number of objects which remain in the cache.
782 * The callback will be passed nr_to_scan == 0 when the VM is querying the
783 * cache size, so a fastpath for that case is appropriate.
785 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
788 * Add an aging callback. The int is the number of 'seeks' it takes
789 * to recreate one of the objects that these functions age.
792 #define DEFAULT_SEEKS 2
794 extern struct shrinker
*set_shrinker(int, shrinker_t
);
795 extern void remove_shrinker(struct shrinker
*shrinker
);
798 * Some shared mappigns will want the pages marked read-only
799 * to track write events. If so, we'll downgrade vm_page_prot
800 * to the private version (using protection_map[] without the
803 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
805 unsigned int vm_flags
= vma
->vm_flags
;
807 /* If it was private or non-writable, the write bit is already clear */
808 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
811 /* The backer wishes to know when pages are first written to? */
812 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
815 /* The open routine did something to the protections already? */
816 if (pgprot_val(vma
->vm_page_prot
) !=
817 pgprot_val(protection_map
[vm_flags
&
818 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
821 /* Specialty mapping? */
822 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
825 /* Can the mapping track the dirty pages? */
826 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
827 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
830 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
832 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
833 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
834 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
835 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
838 * The following ifdef needed to get the 4level-fixup.h header to work.
839 * Remove it when 4level-fixup.h has been removed.
841 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
842 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
844 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
845 NULL
: pud_offset(pgd
, address
);
848 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
850 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
851 NULL
: pmd_offset(pud
, address
);
853 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
855 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
857 * We tuck a spinlock to guard each pagetable page into its struct page,
858 * at page->private, with BUILD_BUG_ON to make sure that this will not
859 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
860 * When freeing, reset page->mapping so free_pages_check won't complain.
862 #define __pte_lockptr(page) &((page)->ptl)
863 #define pte_lock_init(_page) do { \
864 spin_lock_init(__pte_lockptr(_page)); \
866 #define pte_lock_deinit(page) ((page)->mapping = NULL)
867 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
870 * We use mm->page_table_lock to guard all pagetable pages of the mm.
872 #define pte_lock_init(page) do {} while (0)
873 #define pte_lock_deinit(page) do {} while (0)
874 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
875 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
877 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
879 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
880 pte_t *__pte = pte_offset_map(pmd, address); \
886 #define pte_unmap_unlock(pte, ptl) do { \
891 #define pte_alloc_map(mm, pmd, address) \
892 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
893 NULL: pte_offset_map(pmd, address))
895 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
896 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
897 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
899 #define pte_alloc_kernel(pmd, address) \
900 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
901 NULL: pte_offset_kernel(pmd, address))
903 extern void free_area_init(unsigned long * zones_size
);
904 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
905 unsigned long * zones_size
, unsigned long zone_start_pfn
,
906 unsigned long *zholes_size
);
907 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
909 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
910 * zones, allocate the backing mem_map and account for memory holes in a more
911 * architecture independent manner. This is a substitute for creating the
912 * zone_sizes[] and zholes_size[] arrays and passing them to
913 * free_area_init_node()
915 * An architecture is expected to register range of page frames backed by
916 * physical memory with add_active_range() before calling
917 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
918 * usage, an architecture is expected to do something like
920 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
922 * for_each_valid_physical_page_range()
923 * add_active_range(node_id, start_pfn, end_pfn)
924 * free_area_init_nodes(max_zone_pfns);
926 * If the architecture guarantees that there are no holes in the ranges
927 * registered with add_active_range(), free_bootmem_active_regions()
928 * will call free_bootmem_node() for each registered physical page range.
929 * Similarly sparse_memory_present_with_active_regions() calls
930 * memory_present() for each range when SPARSEMEM is enabled.
932 * See mm/page_alloc.c for more information on each function exposed by
933 * CONFIG_ARCH_POPULATES_NODE_MAP
935 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
936 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
937 unsigned long end_pfn
);
938 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
939 unsigned long new_end_pfn
);
940 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
941 unsigned long end_pfn
);
942 extern void remove_all_active_ranges(void);
943 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
944 unsigned long end_pfn
);
945 extern void get_pfn_range_for_nid(unsigned int nid
,
946 unsigned long *start_pfn
, unsigned long *end_pfn
);
947 extern unsigned long find_min_pfn_with_active_regions(void);
948 extern unsigned long find_max_pfn_with_active_regions(void);
949 extern void free_bootmem_with_active_regions(int nid
,
950 unsigned long max_low_pfn
);
951 extern void sparse_memory_present_with_active_regions(int nid
);
952 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
953 extern int early_pfn_to_nid(unsigned long pfn
);
954 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
955 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
956 extern void set_dma_reserve(unsigned long new_dma_reserve
);
957 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
958 extern void setup_per_zone_pages_min(void);
959 extern void mem_init(void);
960 extern void show_mem(void);
961 extern void si_meminfo(struct sysinfo
* val
);
962 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
963 extern void zonetable_add(struct zone
*zone
, int nid
, enum zone_type zid
,
964 unsigned long pfn
, unsigned long size
);
967 extern void setup_per_cpu_pageset(void);
969 static inline void setup_per_cpu_pageset(void) {}
973 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
974 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
975 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
976 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
977 struct prio_tree_iter
*iter
);
979 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
980 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
981 (vma = vma_prio_tree_next(vma, iter)); )
983 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
984 struct list_head
*list
)
986 vma
->shared
.vm_set
.parent
= NULL
;
987 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
991 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
992 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
993 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
994 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
995 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
996 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
998 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
999 extern int split_vma(struct mm_struct
*,
1000 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1001 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1002 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1003 struct rb_node
**, struct rb_node
*);
1004 extern void unlink_file_vma(struct vm_area_struct
*);
1005 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1006 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1007 extern void exit_mmap(struct mm_struct
*);
1008 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1010 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1012 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1013 unsigned long len
, unsigned long prot
,
1014 unsigned long flag
, unsigned long pgoff
);
1016 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1017 unsigned long len
, unsigned long prot
,
1018 unsigned long flag
, unsigned long offset
)
1020 unsigned long ret
= -EINVAL
;
1021 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1023 if (!(offset
& ~PAGE_MASK
)) {
1024 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1030 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1032 extern unsigned long do_brk(unsigned long, unsigned long);
1035 extern unsigned long page_unuse(struct page
*);
1036 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1037 extern void truncate_inode_pages_range(struct address_space
*,
1038 loff_t lstart
, loff_t lend
);
1040 /* generic vm_area_ops exported for stackable file systems */
1041 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
1042 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
1043 unsigned long, pgprot_t
, unsigned long, int);
1045 /* mm/page-writeback.c */
1046 int write_one_page(struct page
*page
, int wait
);
1049 #define VM_MAX_READAHEAD 128 /* kbytes */
1050 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1051 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1052 * turning readahead off */
1054 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1055 pgoff_t offset
, unsigned long nr_to_read
);
1056 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1057 pgoff_t offset
, unsigned long nr_to_read
);
1058 unsigned long page_cache_readahead(struct address_space
*mapping
,
1059 struct file_ra_state
*ra
,
1062 unsigned long size
);
1063 void handle_ra_miss(struct address_space
*mapping
,
1064 struct file_ra_state
*ra
, pgoff_t offset
);
1065 unsigned long max_sane_readahead(unsigned long nr
);
1067 /* Do stack extension */
1068 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1070 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1073 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1074 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1075 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1076 struct vm_area_struct
**pprev
);
1078 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1079 NULL if none. Assume start_addr < end_addr. */
1080 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1082 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1084 if (vma
&& end_addr
<= vma
->vm_start
)
1089 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1091 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1094 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1095 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1096 struct page
*vmalloc_to_page(void *addr
);
1097 unsigned long vmalloc_to_pfn(void *addr
);
1098 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1099 unsigned long pfn
, unsigned long size
, pgprot_t
);
1100 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1102 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1103 unsigned int foll_flags
);
1104 #define FOLL_WRITE 0x01 /* check pte is writable */
1105 #define FOLL_TOUCH 0x02 /* mark page accessed */
1106 #define FOLL_GET 0x04 /* do get_page on page */
1107 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1109 #ifdef CONFIG_PROC_FS
1110 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1112 static inline void vm_stat_account(struct mm_struct
*mm
,
1113 unsigned long flags
, struct file
*file
, long pages
)
1116 #endif /* CONFIG_PROC_FS */
1118 #ifndef CONFIG_DEBUG_PAGEALLOC
1120 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1123 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1124 #ifdef __HAVE_ARCH_GATE_AREA
1125 int in_gate_area_no_task(unsigned long addr
);
1126 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1128 int in_gate_area_no_task(unsigned long addr
);
1129 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1130 #endif /* __HAVE_ARCH_GATE_AREA */
1132 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1133 void __user
*, size_t *, loff_t
*);
1134 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1135 unsigned long lru_pages
);
1136 void drop_pagecache(void);
1137 void drop_slab(void);
1140 #define randomize_va_space 0
1142 extern int randomize_va_space
;
1145 __attribute__((weak
)) const char *arch_vma_name(struct vm_area_struct
*vma
);
1147 #endif /* __KERNEL__ */
1148 #endif /* _LINUX_MM_H */