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>
22 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
23 extern unsigned long max_mapnr
;
26 extern unsigned long num_physpages
;
27 extern void * high_memory
;
28 extern unsigned long vmalloc_earlyreserve
;
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 */
116 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
117 * disabled, then there's a single shared list of VMAs maintained by the
118 * system, and mm's subscribe to these individually
120 struct vm_list_struct
{
121 struct vm_list_struct
*next
;
122 struct vm_area_struct
*vma
;
126 extern struct rb_root nommu_vma_tree
;
127 extern struct rw_semaphore nommu_vma_sem
;
129 extern unsigned int kobjsize(const void *objp
);
135 #define VM_READ 0x00000001 /* currently active flags */
136 #define VM_WRITE 0x00000002
137 #define VM_EXEC 0x00000004
138 #define VM_SHARED 0x00000008
140 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
141 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
142 #define VM_MAYWRITE 0x00000020
143 #define VM_MAYEXEC 0x00000040
144 #define VM_MAYSHARE 0x00000080
146 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
147 #define VM_GROWSUP 0x00000200
148 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
149 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
151 #define VM_EXECUTABLE 0x00001000
152 #define VM_LOCKED 0x00002000
153 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
155 /* Used by sys_madvise() */
156 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
157 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
159 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
160 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
161 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
162 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
163 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
164 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
165 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
166 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
168 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
169 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
172 #ifdef CONFIG_STACK_GROWSUP
173 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
175 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
178 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
179 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
180 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
181 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
182 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
185 * mapping from the currently active vm_flags protection bits (the
186 * low four bits) to a page protection mask..
188 extern pgprot_t protection_map
[16];
192 * These are the virtual MM functions - opening of an area, closing and
193 * unmapping it (needed to keep files on disk up-to-date etc), pointer
194 * to the functions called when a no-page or a wp-page exception occurs.
196 struct vm_operations_struct
{
197 void (*open
)(struct vm_area_struct
* area
);
198 void (*close
)(struct vm_area_struct
* area
);
199 struct page
* (*nopage
)(struct vm_area_struct
* area
, unsigned long address
, int *type
);
200 int (*populate
)(struct vm_area_struct
* area
, unsigned long address
, unsigned long len
, pgprot_t prot
, unsigned long pgoff
, int nonblock
);
202 /* notification that a previously read-only page is about to become
203 * writable, if an error is returned it will cause a SIGBUS */
204 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
206 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
207 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
209 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
210 const nodemask_t
*to
, unsigned long flags
);
218 * Each physical page in the system has a struct page associated with
219 * it to keep track of whatever it is we are using the page for at the
220 * moment. Note that we have no way to track which tasks are using
224 unsigned long flags
; /* Atomic flags, some possibly
225 * updated asynchronously */
226 atomic_t _count
; /* Usage count, see below. */
227 atomic_t _mapcount
; /* Count of ptes mapped in mms,
228 * to show when page is mapped
229 * & limit reverse map searches.
233 unsigned long private; /* Mapping-private opaque data:
234 * usually used for buffer_heads
235 * if PagePrivate set; used for
236 * swp_entry_t if PageSwapCache;
237 * indicates order in the buddy
238 * system if PG_buddy is set.
240 struct address_space
*mapping
; /* If low bit clear, points to
241 * inode address_space, or NULL.
242 * If page mapped as anonymous
243 * memory, low bit is set, and
244 * it points to anon_vma object:
245 * see PAGE_MAPPING_ANON below.
248 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
252 pgoff_t index
; /* Our offset within mapping. */
253 struct list_head lru
; /* Pageout list, eg. active_list
254 * protected by zone->lru_lock !
257 * On machines where all RAM is mapped into kernel address space,
258 * we can simply calculate the virtual address. On machines with
259 * highmem some memory is mapped into kernel virtual memory
260 * dynamically, so we need a place to store that address.
261 * Note that this field could be 16 bits on x86 ... ;)
263 * Architectures with slow multiplication can define
264 * WANT_PAGE_VIRTUAL in asm/page.h
266 #if defined(WANT_PAGE_VIRTUAL)
267 void *virtual; /* Kernel virtual address (NULL if
268 not kmapped, ie. highmem) */
269 #endif /* WANT_PAGE_VIRTUAL */
272 #define page_private(page) ((page)->private)
273 #define set_page_private(page, v) ((page)->private = (v))
276 * FIXME: take this include out, include page-flags.h in
277 * files which need it (119 of them)
279 #include <linux/page-flags.h>
282 * Methods to modify the page usage count.
284 * What counts for a page usage:
285 * - cache mapping (page->mapping)
286 * - private data (page->private)
287 * - page mapped in a task's page tables, each mapping
288 * is counted separately
290 * Also, many kernel routines increase the page count before a critical
291 * routine so they can be sure the page doesn't go away from under them.
295 * Drop a ref, return true if the logical refcount fell to zero (the page has
298 static inline int put_page_testzero(struct page
*page
)
300 BUG_ON(atomic_read(&page
->_count
) == 0);
301 return atomic_dec_and_test(&page
->_count
);
305 * Try to grab a ref unless the page has a refcount of zero, return false if
308 static inline int get_page_unless_zero(struct page
*page
)
310 return atomic_inc_not_zero(&page
->_count
);
313 extern void FASTCALL(__page_cache_release(struct page
*));
315 static inline int page_count(struct page
*page
)
317 if (unlikely(PageCompound(page
)))
318 page
= (struct page
*)page_private(page
);
319 return atomic_read(&page
->_count
);
322 static inline void get_page(struct page
*page
)
324 if (unlikely(PageCompound(page
)))
325 page
= (struct page
*)page_private(page
);
326 atomic_inc(&page
->_count
);
330 * Setup the page count before being freed into the page allocator for
331 * the first time (boot or memory hotplug)
333 static inline void init_page_count(struct page
*page
)
335 atomic_set(&page
->_count
, 1);
338 void put_page(struct page
*page
);
339 void put_pages_list(struct list_head
*pages
);
341 void split_page(struct page
*page
, unsigned int order
);
344 * Multiple processes may "see" the same page. E.g. for untouched
345 * mappings of /dev/null, all processes see the same page full of
346 * zeroes, and text pages of executables and shared libraries have
347 * only one copy in memory, at most, normally.
349 * For the non-reserved pages, page_count(page) denotes a reference count.
350 * page_count() == 0 means the page is free. page->lru is then used for
351 * freelist management in the buddy allocator.
352 * page_count() == 1 means the page is used for exactly one purpose
353 * (e.g. a private data page of one process).
355 * A page may be used for kmalloc() or anyone else who does a
356 * __get_free_page(). In this case the page_count() is at least 1, and
357 * all other fields are unused but should be 0 or NULL. The
358 * management of this page is the responsibility of the one who uses
361 * The other pages (we may call them "process pages") are completely
362 * managed by the Linux memory manager: I/O, buffers, swapping etc.
363 * The following discussion applies only to them.
365 * A page may belong to an inode's memory mapping. In this case,
366 * page->mapping is the pointer to the inode, and page->index is the
367 * file offset of the page, in units of PAGE_CACHE_SIZE.
369 * A page contains an opaque `private' member, which belongs to the
370 * page's address_space. Usually, this is the address of a circular
371 * list of the page's disk buffers.
373 * For pages belonging to inodes, the page_count() is the number of
374 * attaches, plus 1 if `private' contains something, plus one for
375 * the page cache itself.
377 * Instead of keeping dirty/clean pages in per address-space lists, we instead
378 * now tag pages as dirty/under writeback in the radix tree.
380 * There is also a per-mapping radix tree mapping index to the page
381 * in memory if present. The tree is rooted at mapping->root.
383 * All process pages can do I/O:
384 * - inode pages may need to be read from disk,
385 * - inode pages which have been modified and are MAP_SHARED may need
386 * to be written to disk,
387 * - private pages which have been modified may need to be swapped out
388 * to swap space and (later) to be read back into memory.
392 * The zone field is never updated after free_area_init_core()
393 * sets it, so none of the operations on it need to be atomic.
398 * page->flags layout:
400 * There are three possibilities for how page->flags get
401 * laid out. The first is for the normal case, without
402 * sparsemem. The second is for sparsemem when there is
403 * plenty of space for node and section. The last is when
404 * we have run out of space and have to fall back to an
405 * alternate (slower) way of determining the node.
407 * No sparsemem: | NODE | ZONE | ... | FLAGS |
408 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
409 * no space for node: | SECTION | ZONE | ... | FLAGS |
411 #ifdef CONFIG_SPARSEMEM
412 #define SECTIONS_WIDTH SECTIONS_SHIFT
414 #define SECTIONS_WIDTH 0
417 #define ZONES_WIDTH ZONES_SHIFT
419 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
420 #define NODES_WIDTH NODES_SHIFT
422 #define NODES_WIDTH 0
425 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
426 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
427 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
428 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
431 * We are going to use the flags for the page to node mapping if its in
432 * there. This includes the case where there is no node, so it is implicit.
434 #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
436 #ifndef PFN_SECTION_SHIFT
437 #define PFN_SECTION_SHIFT 0
441 * Define the bit shifts to access each section. For non-existant
442 * sections we define the shift as 0; that plus a 0 mask ensures
443 * the compiler will optimise away reference to them.
445 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
446 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
447 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
449 /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
451 #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
453 #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
455 #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
457 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
458 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
461 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
462 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
463 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
464 #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
466 static inline unsigned long page_zonenum(struct page
*page
)
468 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
472 extern struct zone
*zone_table
[];
474 static inline int page_zone_id(struct page
*page
)
476 return (page
->flags
>> ZONETABLE_PGSHIFT
) & ZONETABLE_MASK
;
478 static inline struct zone
*page_zone(struct page
*page
)
480 return zone_table
[page_zone_id(page
)];
483 static inline unsigned long page_to_nid(struct page
*page
)
486 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
488 return page_zone(page
)->zone_pgdat
->node_id
;
490 static inline unsigned long page_to_section(struct page
*page
)
492 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
495 static inline void set_page_zone(struct page
*page
, unsigned long zone
)
497 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
498 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
500 static inline void set_page_node(struct page
*page
, unsigned long node
)
502 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
503 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
505 static inline void set_page_section(struct page
*page
, unsigned long section
)
507 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
508 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
511 static inline void set_page_links(struct page
*page
, unsigned long zone
,
512 unsigned long node
, unsigned long pfn
)
514 set_page_zone(page
, zone
);
515 set_page_node(page
, node
);
516 set_page_section(page
, pfn_to_section_nr(pfn
));
520 * Some inline functions in vmstat.h depend on page_zone()
522 #include <linux/vmstat.h>
524 #ifndef CONFIG_DISCONTIGMEM
525 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
526 extern struct page
*mem_map
;
529 static __always_inline
void *lowmem_page_address(struct page
*page
)
531 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
534 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
535 #define HASHED_PAGE_VIRTUAL
538 #if defined(WANT_PAGE_VIRTUAL)
539 #define page_address(page) ((page)->virtual)
540 #define set_page_address(page, address) \
542 (page)->virtual = (address); \
544 #define page_address_init() do { } while(0)
547 #if defined(HASHED_PAGE_VIRTUAL)
548 void *page_address(struct page
*page
);
549 void set_page_address(struct page
*page
, void *virtual);
550 void page_address_init(void);
553 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
554 #define page_address(page) lowmem_page_address(page)
555 #define set_page_address(page, address) do { } while(0)
556 #define page_address_init() do { } while(0)
560 * On an anonymous page mapped into a user virtual memory area,
561 * page->mapping points to its anon_vma, not to a struct address_space;
562 * with the PAGE_MAPPING_ANON bit set to distinguish it.
564 * Please note that, confusingly, "page_mapping" refers to the inode
565 * address_space which maps the page from disk; whereas "page_mapped"
566 * refers to user virtual address space into which the page is mapped.
568 #define PAGE_MAPPING_ANON 1
570 extern struct address_space swapper_space
;
571 static inline struct address_space
*page_mapping(struct page
*page
)
573 struct address_space
*mapping
= page
->mapping
;
575 if (unlikely(PageSwapCache(page
)))
576 mapping
= &swapper_space
;
577 else if (unlikely((unsigned long)mapping
& PAGE_MAPPING_ANON
))
582 static inline int PageAnon(struct page
*page
)
584 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
588 * Return the pagecache index of the passed page. Regular pagecache pages
589 * use ->index whereas swapcache pages use ->private
591 static inline pgoff_t
page_index(struct page
*page
)
593 if (unlikely(PageSwapCache(page
)))
594 return page_private(page
);
599 * The atomic page->_mapcount, like _count, starts from -1:
600 * so that transitions both from it and to it can be tracked,
601 * using atomic_inc_and_test and atomic_add_negative(-1).
603 static inline void reset_page_mapcount(struct page
*page
)
605 atomic_set(&(page
)->_mapcount
, -1);
608 static inline int page_mapcount(struct page
*page
)
610 return atomic_read(&(page
)->_mapcount
) + 1;
614 * Return true if this page is mapped into pagetables.
616 static inline int page_mapped(struct page
*page
)
618 return atomic_read(&(page
)->_mapcount
) >= 0;
622 * Error return values for the *_nopage functions
624 #define NOPAGE_SIGBUS (NULL)
625 #define NOPAGE_OOM ((struct page *) (-1))
628 * Different kinds of faults, as returned by handle_mm_fault().
629 * Used to decide whether a process gets delivered SIGBUS or
630 * just gets major/minor fault counters bumped up.
632 #define VM_FAULT_OOM 0x00
633 #define VM_FAULT_SIGBUS 0x01
634 #define VM_FAULT_MINOR 0x02
635 #define VM_FAULT_MAJOR 0x03
638 * Special case for get_user_pages.
639 * Must be in a distinct bit from the above VM_FAULT_ flags.
641 #define VM_FAULT_WRITE 0x10
643 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
645 extern void show_free_areas(void);
648 struct page
*shmem_nopage(struct vm_area_struct
*vma
,
649 unsigned long address
, int *type
);
650 int shmem_set_policy(struct vm_area_struct
*vma
, struct mempolicy
*new);
651 struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
653 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
655 #define shmem_nopage filemap_nopage
657 static inline int shmem_lock(struct file
*file
, int lock
,
658 struct user_struct
*user
)
663 static inline int shmem_set_policy(struct vm_area_struct
*vma
,
664 struct mempolicy
*new)
669 static inline struct mempolicy
*shmem_get_policy(struct vm_area_struct
*vma
,
675 struct file
*shmem_file_setup(char *name
, loff_t size
, unsigned long flags
);
676 extern int shmem_mmap(struct file
*file
, struct vm_area_struct
*vma
);
678 int shmem_zero_setup(struct vm_area_struct
*);
681 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
685 unsigned long flags
);
688 static inline int can_do_mlock(void)
690 if (capable(CAP_IPC_LOCK
))
692 if (current
->signal
->rlim
[RLIMIT_MEMLOCK
].rlim_cur
!= 0)
696 extern int user_shm_lock(size_t, struct user_struct
*);
697 extern void user_shm_unlock(size_t, struct user_struct
*);
700 * Parameter block passed down to zap_pte_range in exceptional cases.
703 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
704 struct address_space
*check_mapping
; /* Check page->mapping if set */
705 pgoff_t first_index
; /* Lowest page->index to unmap */
706 pgoff_t last_index
; /* Highest page->index to unmap */
707 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
708 unsigned long truncate_count
; /* Compare vm_truncate_count */
711 struct page
*vm_normal_page(struct vm_area_struct
*, unsigned long, pte_t
);
712 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
713 unsigned long size
, struct zap_details
*);
714 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
715 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
716 unsigned long end_addr
, unsigned long *nr_accounted
,
717 struct zap_details
*);
718 void free_pgd_range(struct mmu_gather
**tlb
, unsigned long addr
,
719 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
720 void free_pgtables(struct mmu_gather
**tlb
, struct vm_area_struct
*start_vma
,
721 unsigned long floor
, unsigned long ceiling
);
722 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
723 struct vm_area_struct
*vma
);
724 int zeromap_page_range(struct vm_area_struct
*vma
, unsigned long from
,
725 unsigned long size
, pgprot_t prot
);
726 void unmap_mapping_range(struct address_space
*mapping
,
727 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
729 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
730 loff_t
const holebegin
, loff_t
const holelen
)
732 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
735 extern int vmtruncate(struct inode
* inode
, loff_t offset
);
736 extern int vmtruncate_range(struct inode
* inode
, loff_t offset
, loff_t end
);
737 extern int install_page(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, struct page
*page
, pgprot_t prot
);
738 extern int install_file_pte(struct mm_struct
*mm
, struct vm_area_struct
*vma
, unsigned long addr
, unsigned long pgoff
, pgprot_t prot
);
741 extern int __handle_mm_fault(struct mm_struct
*mm
,struct vm_area_struct
*vma
,
742 unsigned long address
, int write_access
);
744 static inline int handle_mm_fault(struct mm_struct
*mm
,
745 struct vm_area_struct
*vma
, unsigned long address
,
748 return __handle_mm_fault(mm
, vma
, address
, write_access
) &
752 static inline int handle_mm_fault(struct mm_struct
*mm
,
753 struct vm_area_struct
*vma
, unsigned long address
,
756 /* should never happen if there's no MMU */
758 return VM_FAULT_SIGBUS
;
762 extern int make_pages_present(unsigned long addr
, unsigned long end
);
763 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
764 void install_arg_page(struct vm_area_struct
*, struct page
*, unsigned long);
766 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
, unsigned long start
,
767 int len
, int write
, int force
, struct page
**pages
, struct vm_area_struct
**vmas
);
768 void print_bad_pte(struct vm_area_struct
*, pte_t
, unsigned long);
770 int __set_page_dirty_buffers(struct page
*page
);
771 int __set_page_dirty_nobuffers(struct page
*page
);
772 int redirty_page_for_writepage(struct writeback_control
*wbc
,
774 int FASTCALL(set_page_dirty(struct page
*page
));
775 int set_page_dirty_lock(struct page
*page
);
776 int clear_page_dirty_for_io(struct page
*page
);
778 extern unsigned long do_mremap(unsigned long addr
,
779 unsigned long old_len
, unsigned long new_len
,
780 unsigned long flags
, unsigned long new_addr
);
783 * Prototype to add a shrinker callback for ageable caches.
785 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
786 * scan `nr_to_scan' objects, attempting to free them.
788 * The callback must return the number of objects which remain in the cache.
790 * The callback will be passed nr_to_scan == 0 when the VM is querying the
791 * cache size, so a fastpath for that case is appropriate.
793 typedef int (*shrinker_t
)(int nr_to_scan
, gfp_t gfp_mask
);
796 * Add an aging callback. The int is the number of 'seeks' it takes
797 * to recreate one of the objects that these functions age.
800 #define DEFAULT_SEEKS 2
802 extern struct shrinker
*set_shrinker(int, shrinker_t
);
803 extern void remove_shrinker(struct shrinker
*shrinker
);
805 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
807 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
808 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
809 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
810 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
813 * The following ifdef needed to get the 4level-fixup.h header to work.
814 * Remove it when 4level-fixup.h has been removed.
816 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
817 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
819 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
820 NULL
: pud_offset(pgd
, address
);
823 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
825 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
826 NULL
: pmd_offset(pud
, address
);
828 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
830 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
832 * We tuck a spinlock to guard each pagetable page into its struct page,
833 * at page->private, with BUILD_BUG_ON to make sure that this will not
834 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
835 * When freeing, reset page->mapping so free_pages_check won't complain.
837 #define __pte_lockptr(page) &((page)->ptl)
838 #define pte_lock_init(_page) do { \
839 spin_lock_init(__pte_lockptr(_page)); \
841 #define pte_lock_deinit(page) ((page)->mapping = NULL)
842 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
845 * We use mm->page_table_lock to guard all pagetable pages of the mm.
847 #define pte_lock_init(page) do {} while (0)
848 #define pte_lock_deinit(page) do {} while (0)
849 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
850 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
852 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
854 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
855 pte_t *__pte = pte_offset_map(pmd, address); \
861 #define pte_unmap_unlock(pte, ptl) do { \
866 #define pte_alloc_map(mm, pmd, address) \
867 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
868 NULL: pte_offset_map(pmd, address))
870 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
871 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
872 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
874 #define pte_alloc_kernel(pmd, address) \
875 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
876 NULL: pte_offset_kernel(pmd, address))
878 extern void free_area_init(unsigned long * zones_size
);
879 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
880 unsigned long * zones_size
, unsigned long zone_start_pfn
,
881 unsigned long *zholes_size
);
882 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
883 extern void setup_per_zone_pages_min(void);
884 extern void mem_init(void);
885 extern void show_mem(void);
886 extern void si_meminfo(struct sysinfo
* val
);
887 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
890 extern void setup_per_cpu_pageset(void);
892 static inline void setup_per_cpu_pageset(void) {}
896 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
897 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
898 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
899 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
900 struct prio_tree_iter
*iter
);
902 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
903 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
904 (vma = vma_prio_tree_next(vma, iter)); )
906 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
907 struct list_head
*list
)
909 vma
->shared
.vm_set
.parent
= NULL
;
910 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
914 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
915 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
916 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
917 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
918 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
919 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
921 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
922 extern int split_vma(struct mm_struct
*,
923 struct vm_area_struct
*, unsigned long addr
, int new_below
);
924 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
925 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
926 struct rb_node
**, struct rb_node
*);
927 extern void unlink_file_vma(struct vm_area_struct
*);
928 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
929 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
930 extern void exit_mmap(struct mm_struct
*);
931 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
933 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
935 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
936 unsigned long len
, unsigned long prot
,
937 unsigned long flag
, unsigned long pgoff
);
939 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
940 unsigned long len
, unsigned long prot
,
941 unsigned long flag
, unsigned long offset
)
943 unsigned long ret
= -EINVAL
;
944 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
946 if (!(offset
& ~PAGE_MASK
))
947 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
952 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
954 extern unsigned long do_brk(unsigned long, unsigned long);
957 extern unsigned long page_unuse(struct page
*);
958 extern void truncate_inode_pages(struct address_space
*, loff_t
);
959 extern void truncate_inode_pages_range(struct address_space
*,
960 loff_t lstart
, loff_t lend
);
962 /* generic vm_area_ops exported for stackable file systems */
963 extern struct page
*filemap_nopage(struct vm_area_struct
*, unsigned long, int *);
964 extern int filemap_populate(struct vm_area_struct
*, unsigned long,
965 unsigned long, pgprot_t
, unsigned long, int);
967 /* mm/page-writeback.c */
968 int write_one_page(struct page
*page
, int wait
);
971 #define VM_MAX_READAHEAD 128 /* kbytes */
972 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
973 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
974 * turning readahead off */
976 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
977 pgoff_t offset
, unsigned long nr_to_read
);
978 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
979 pgoff_t offset
, unsigned long nr_to_read
);
980 unsigned long page_cache_readahead(struct address_space
*mapping
,
981 struct file_ra_state
*ra
,
985 void handle_ra_miss(struct address_space
*mapping
,
986 struct file_ra_state
*ra
, pgoff_t offset
);
987 unsigned long max_sane_readahead(unsigned long nr
);
989 /* Do stack extension */
990 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
992 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
995 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
996 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
997 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
998 struct vm_area_struct
**pprev
);
1000 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1001 NULL if none. Assume start_addr < end_addr. */
1002 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1004 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1006 if (vma
&& end_addr
<= vma
->vm_start
)
1011 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1013 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1016 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1017 struct page
*vmalloc_to_page(void *addr
);
1018 unsigned long vmalloc_to_pfn(void *addr
);
1019 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1020 unsigned long pfn
, unsigned long size
, pgprot_t
);
1021 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1023 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1024 unsigned int foll_flags
);
1025 #define FOLL_WRITE 0x01 /* check pte is writable */
1026 #define FOLL_TOUCH 0x02 /* mark page accessed */
1027 #define FOLL_GET 0x04 /* do get_page on page */
1028 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1030 #ifdef CONFIG_PROC_FS
1031 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1033 static inline void vm_stat_account(struct mm_struct
*mm
,
1034 unsigned long flags
, struct file
*file
, long pages
)
1037 #endif /* CONFIG_PROC_FS */
1039 #ifndef CONFIG_DEBUG_PAGEALLOC
1041 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
1043 if (!PageHighMem(page
) && !enable
)
1044 debug_check_no_locks_freed(page_address(page
),
1045 numpages
* PAGE_SIZE
);
1049 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1050 #ifdef __HAVE_ARCH_GATE_AREA
1051 int in_gate_area_no_task(unsigned long addr
);
1052 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1054 int in_gate_area_no_task(unsigned long addr
);
1055 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1056 #endif /* __HAVE_ARCH_GATE_AREA */
1058 /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
1059 #define OOM_DISABLE -17
1061 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1062 void __user
*, size_t *, loff_t
*);
1063 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1064 unsigned long lru_pages
);
1065 void drop_pagecache(void);
1066 void drop_slab(void);
1069 #define randomize_va_space 0
1071 extern int randomize_va_space
;
1074 const char *arch_vma_name(struct vm_area_struct
*vma
);
1076 #endif /* __KERNEL__ */
1077 #endif /* _LINUX_MM_H */