4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
14 #include <linux/mutex.h>
15 #include <linux/debug_locks.h>
16 #include <linux/backing-dev.h>
17 #include <linux/mm_types.h>
23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
24 extern unsigned long max_mapnr
;
27 extern unsigned long num_physpages
;
28 extern void * high_memory
;
29 extern int page_cluster
;
32 extern int sysctl_legacy_va_layout
;
34 #define sysctl_legacy_va_layout 0
38 #include <asm/pgtable.h>
39 #include <asm/processor.h>
41 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
44 * Linux kernel virtual memory manager primitives.
45 * The idea being to have a "virtual" mm in the same way
46 * we have a virtual fs - giving a cleaner interface to the
47 * mm details, and allowing different kinds of memory mappings
48 * (from shared memory to executable loading to arbitrary
53 * This struct defines a memory VMM memory area. There is one of these
54 * per VM-area/task. A VM area is any part of the process virtual memory
55 * space that has a special rule for the page-fault handlers (ie a shared
56 * library, the executable area etc).
58 struct vm_area_struct
{
59 struct mm_struct
* vm_mm
; /* The address space we belong to. */
60 unsigned long vm_start
; /* Our start address within vm_mm. */
61 unsigned long vm_end
; /* The first byte after our end address
64 /* linked list of VM areas per task, sorted by address */
65 struct vm_area_struct
*vm_next
;
67 pgprot_t vm_page_prot
; /* Access permissions of this VMA. */
68 unsigned long vm_flags
; /* Flags, listed below. */
73 * For areas with an address space and backing store,
74 * linkage into the address_space->i_mmap prio tree, or
75 * linkage to the list of like vmas hanging off its node, or
76 * linkage of vma in the address_space->i_mmap_nonlinear list.
80 struct list_head list
;
81 void *parent
; /* aligns with prio_tree_node parent */
82 struct vm_area_struct
*head
;
85 struct raw_prio_tree_node prio_tree_node
;
89 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
90 * list, after a COW of one of the file pages. A MAP_SHARED vma
91 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
92 * or brk vma (with NULL file) can only be in an anon_vma list.
94 struct list_head anon_vma_node
; /* Serialized by anon_vma->lock */
95 struct anon_vma
*anon_vma
; /* Serialized by page_table_lock */
97 /* Function pointers to deal with this struct. */
98 struct vm_operations_struct
* vm_ops
;
100 /* Information about our backing store: */
101 unsigned long vm_pgoff
; /* Offset (within vm_file) in PAGE_SIZE
102 units, *not* PAGE_CACHE_SIZE */
103 struct file
* vm_file
; /* File we map to (can be NULL). */
104 void * vm_private_data
; /* was vm_pte (shared mem) */
105 unsigned long vm_truncate_count
;/* truncate_count or restart_addr */
108 atomic_t vm_usage
; /* refcount (VMAs shared if !MMU) */
111 struct mempolicy
*vm_policy
; /* NUMA policy for the VMA */
115 extern struct kmem_cache
*vm_area_cachep
;
118 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
119 * disabled, then there's a single shared list of VMAs maintained by the
120 * system, and mm's subscribe to these individually
122 struct vm_list_struct
{
123 struct vm_list_struct
*next
;
124 struct vm_area_struct
*vma
;
128 extern struct rb_root nommu_vma_tree
;
129 extern struct rw_semaphore nommu_vma_sem
;
131 extern unsigned int kobjsize(const void *objp
);
137 #define VM_READ 0x00000001 /* currently active flags */
138 #define VM_WRITE 0x00000002
139 #define VM_EXEC 0x00000004
140 #define VM_SHARED 0x00000008
142 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
143 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
144 #define VM_MAYWRITE 0x00000020
145 #define VM_MAYEXEC 0x00000040
146 #define VM_MAYSHARE 0x00000080
148 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
149 #define VM_GROWSUP 0x00000200
150 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
151 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
153 #define VM_EXECUTABLE 0x00001000
154 #define VM_LOCKED 0x00002000
155 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
157 /* Used by sys_madvise() */
158 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
159 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
161 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
162 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
163 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
164 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
165 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
166 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
167 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
168 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
169 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
171 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
173 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
174 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
177 #ifdef CONFIG_STACK_GROWSUP
178 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
180 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
183 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
184 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
185 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
186 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
187 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
190 * mapping from the currently active vm_flags protection bits (the
191 * low four bits) to a page protection mask..
193 extern pgprot_t protection_map
[16];
195 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
196 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
200 * vm_fault is filled by the the pagefault handler and passed to the vma's
201 * ->fault function. The vma's ->fault is responsible for returning a bitmask
202 * of VM_FAULT_xxx flags that give details about how the fault was handled.
204 * pgoff should be used in favour of virtual_address, if possible. If pgoff
205 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
209 unsigned int flags
; /* FAULT_FLAG_xxx flags */
210 pgoff_t pgoff
; /* Logical page offset based on vma */
211 void __user
*virtual_address
; /* Faulting virtual address */
213 struct page
*page
; /* ->fault handlers should return a
214 * page here, unless VM_FAULT_NOPAGE
215 * is set (which is also implied by
221 * These are the virtual MM functions - opening of an area, closing and
222 * unmapping it (needed to keep files on disk up-to-date etc), pointer
223 * to the functions called when a no-page or a wp-page exception occurs.
225 struct vm_operations_struct
{
226 void (*open
)(struct vm_area_struct
* area
);
227 void (*close
)(struct vm_area_struct
* area
);
228 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
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
);
234 /* notification that a previously read-only page is about to become
235 * writable, if an error is returned it will cause a SIGBUS */
236 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct page
*page
);
238 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
239 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
241 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
242 const nodemask_t
*to
, unsigned long flags
);
249 #define page_private(page) ((page)->private)
250 #define set_page_private(page, v) ((page)->private = (v))
253 * FIXME: take this include out, include page-flags.h in
254 * files which need it (119 of them)
256 #include <linux/page-flags.h>
258 #ifdef CONFIG_DEBUG_VM
259 #define VM_BUG_ON(cond) BUG_ON(cond)
261 #define VM_BUG_ON(condition) do { } while(0)
265 * Methods to modify the page usage count.
267 * What counts for a page usage:
268 * - cache mapping (page->mapping)
269 * - private data (page->private)
270 * - page mapped in a task's page tables, each mapping
271 * is counted separately
273 * Also, many kernel routines increase the page count before a critical
274 * routine so they can be sure the page doesn't go away from under them.
278 * Drop a ref, return true if the refcount fell to zero (the page has no users)
280 static inline int put_page_testzero(struct page
*page
)
282 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
283 return atomic_dec_and_test(&page
->_count
);
287 * Try to grab a ref unless the page has a refcount of zero, return false if
290 static inline int get_page_unless_zero(struct page
*page
)
292 VM_BUG_ON(PageCompound(page
));
293 return atomic_inc_not_zero(&page
->_count
);
296 static inline struct page
*compound_head(struct page
*page
)
298 if (unlikely(PageTail(page
)))
299 return page
->first_page
;
303 static inline int page_count(struct page
*page
)
305 return atomic_read(&compound_head(page
)->_count
);
308 static inline void get_page(struct page
*page
)
310 page
= compound_head(page
);
311 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
312 atomic_inc(&page
->_count
);
315 static inline struct page
*virt_to_head_page(const void *x
)
317 struct page
*page
= virt_to_page(x
);
318 return compound_head(page
);
322 * Setup the page count before being freed into the page allocator for
323 * the first time (boot or memory hotplug)
325 static inline void init_page_count(struct page
*page
)
327 atomic_set(&page
->_count
, 1);
330 void put_page(struct page
*page
);
331 void put_pages_list(struct list_head
*pages
);
333 void split_page(struct page
*page
, unsigned int order
);
336 * Compound pages have a destructor function. Provide a
337 * prototype for that function and accessor functions.
338 * These are _only_ valid on the head of a PG_compound page.
340 typedef void compound_page_dtor(struct page
*);
342 static inline void set_compound_page_dtor(struct page
*page
,
343 compound_page_dtor
*dtor
)
345 page
[1].lru
.next
= (void *)dtor
;
348 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
350 return (compound_page_dtor
*)page
[1].lru
.next
;
353 static inline int compound_order(struct page
*page
)
357 return (unsigned long)page
[1].lru
.prev
;
360 static inline void set_compound_order(struct page
*page
, unsigned long order
)
362 page
[1].lru
.prev
= (void *)order
;
366 * Multiple processes may "see" the same page. E.g. for untouched
367 * mappings of /dev/null, all processes see the same page full of
368 * zeroes, and text pages of executables and shared libraries have
369 * only one copy in memory, at most, normally.
371 * For the non-reserved pages, page_count(page) denotes a reference count.
372 * page_count() == 0 means the page is free. page->lru is then used for
373 * freelist management in the buddy allocator.
374 * page_count() > 0 means the page has been allocated.
376 * Pages are allocated by the slab allocator in order to provide memory
377 * to kmalloc and kmem_cache_alloc. In this case, the management of the
378 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
379 * unless a particular usage is carefully commented. (the responsibility of
380 * freeing the kmalloc memory is the caller's, of course).
382 * A page may be used by anyone else who does a __get_free_page().
383 * In this case, page_count still tracks the references, and should only
384 * be used through the normal accessor functions. The top bits of page->flags
385 * and page->virtual store page management information, but all other fields
386 * are unused and could be used privately, carefully. The management of this
387 * page is the responsibility of the one who allocated it, and those who have
388 * subsequently been given references to it.
390 * The other pages (we may call them "pagecache pages") are completely
391 * managed by the Linux memory manager: I/O, buffers, swapping etc.
392 * The following discussion applies only to them.
394 * A pagecache page contains an opaque `private' member, which belongs to the
395 * page's address_space. Usually, this is the address of a circular list of
396 * the page's disk buffers. PG_private must be set to tell the VM to call
397 * into the filesystem to release these pages.
399 * A page may belong to an inode's memory mapping. In this case, page->mapping
400 * is the pointer to the inode, and page->index is the file offset of the page,
401 * in units of PAGE_CACHE_SIZE.
403 * If pagecache pages are not associated with an inode, they are said to be
404 * anonymous pages. These may become associated with the swapcache, and in that
405 * case PG_swapcache is set, and page->private is an offset into the swapcache.
407 * In either case (swapcache or inode backed), the pagecache itself holds one
408 * reference to the page. Setting PG_private should also increment the
409 * refcount. The each user mapping also has a reference to the page.
411 * The pagecache pages are stored in a per-mapping radix tree, which is
412 * rooted at mapping->page_tree, and indexed by offset.
413 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
414 * lists, we instead now tag pages as dirty/writeback in the radix tree.
416 * All pagecache pages may be subject to I/O:
417 * - inode pages may need to be read from disk,
418 * - inode pages which have been modified and are MAP_SHARED may need
419 * to be written back to the inode on disk,
420 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
421 * modified may need to be swapped out to swap space and (later) to be read
426 * The zone field is never updated after free_area_init_core()
427 * sets it, so none of the operations on it need to be atomic.
432 * page->flags layout:
434 * There are three possibilities for how page->flags get
435 * laid out. The first is for the normal case, without
436 * sparsemem. The second is for sparsemem when there is
437 * plenty of space for node and section. The last is when
438 * we have run out of space and have to fall back to an
439 * alternate (slower) way of determining the node.
441 * No sparsemem: | NODE | ZONE | ... | FLAGS |
442 * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
443 * no space for node: | SECTION | ZONE | ... | FLAGS |
445 #ifdef CONFIG_SPARSEMEM
446 #define SECTIONS_WIDTH SECTIONS_SHIFT
448 #define SECTIONS_WIDTH 0
451 #define ZONES_WIDTH ZONES_SHIFT
453 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
454 #define NODES_WIDTH NODES_SHIFT
456 #define NODES_WIDTH 0
459 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
460 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
461 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
462 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
465 * We are going to use the flags for the page to node mapping if its in
466 * there. This includes the case where there is no node, so it is implicit.
468 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
469 #define NODE_NOT_IN_PAGE_FLAGS
472 #ifndef PFN_SECTION_SHIFT
473 #define PFN_SECTION_SHIFT 0
477 * Define the bit shifts to access each section. For non-existant
478 * sections we define the shift as 0; that plus a 0 mask ensures
479 * the compiler will optimise away reference to them.
481 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
482 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
483 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
485 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
486 #ifdef NODE_NOT_IN_PAGEFLAGS
487 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
488 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
489 SECTIONS_PGOFF : ZONES_PGOFF)
491 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
492 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
493 NODES_PGOFF : ZONES_PGOFF)
496 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
498 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
499 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
502 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
503 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
504 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
505 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
507 static inline enum zone_type
page_zonenum(struct page
*page
)
509 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
513 * The identification function is only used by the buddy allocator for
514 * determining if two pages could be buddies. We are not really
515 * identifying a zone since we could be using a the section number
516 * id if we have not node id available in page flags.
517 * We guarantee only that it will return the same value for two
518 * combinable pages in a zone.
520 static inline int page_zone_id(struct page
*page
)
522 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
525 static inline int zone_to_nid(struct zone
*zone
)
534 #ifdef NODE_NOT_IN_PAGE_FLAGS
535 extern int page_to_nid(struct page
*page
);
537 static inline int page_to_nid(struct page
*page
)
539 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
543 static inline struct zone
*page_zone(struct page
*page
)
545 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
548 static inline unsigned long page_to_section(struct page
*page
)
550 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
553 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
555 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
556 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
559 static inline void set_page_node(struct page
*page
, unsigned long node
)
561 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
562 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
565 static inline void set_page_section(struct page
*page
, unsigned long section
)
567 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
568 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
571 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
572 unsigned long node
, unsigned long pfn
)
574 set_page_zone(page
, zone
);
575 set_page_node(page
, node
);
576 set_page_section(page
, pfn_to_section_nr(pfn
));
580 * Some inline functions in vmstat.h depend on page_zone()
582 #include <linux/vmstat.h>
584 static __always_inline
void *lowmem_page_address(struct page
*page
)
586 return __va(page_to_pfn(page
) << PAGE_SHIFT
);
589 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
590 #define HASHED_PAGE_VIRTUAL
593 #if defined(WANT_PAGE_VIRTUAL)
594 #define page_address(page) ((page)->virtual)
595 #define set_page_address(page, address) \
597 (page)->virtual = (address); \
599 #define page_address_init() do { } while(0)
602 #if defined(HASHED_PAGE_VIRTUAL)
603 void *page_address(struct page
*page
);
604 void set_page_address(struct page
*page
, void *virtual);
605 void page_address_init(void);
608 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
609 #define page_address(page) lowmem_page_address(page)
610 #define set_page_address(page, address) do { } while(0)
611 #define page_address_init() do { } while(0)
615 * On an anonymous page mapped into a user virtual memory area,
616 * page->mapping points to its anon_vma, not to a struct address_space;
617 * with the PAGE_MAPPING_ANON bit set to distinguish it.
619 * Please note that, confusingly, "page_mapping" refers to the inode
620 * address_space which maps the page from disk; whereas "page_mapped"
621 * refers to user virtual address space into which the page is mapped.
623 #define PAGE_MAPPING_ANON 1
625 extern struct address_space swapper_space
;
626 static inline struct address_space
*page_mapping(struct page
*page
)
628 struct address_space
*mapping
= page
->mapping
;
630 VM_BUG_ON(PageSlab(page
));
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 * A callback you can register to apply pressure to ageable caches.
835 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
836 * look through the least-recently-used 'nr_to_scan' entries and
837 * attempt to free them up. It should return the number of objects
838 * which remain in the cache. If it returns -1, it means it cannot do
839 * any scanning at this time (eg. there is a risk of deadlock).
841 * The 'gfpmask' refers to the allocation we are currently trying to
844 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
845 * querying the cache size, so a fastpath for that case is appropriate.
848 int (*shrink
)(int nr_to_scan
, gfp_t gfp_mask
);
849 int seeks
; /* seeks to recreate an obj */
851 /* These are for internal use */
852 struct list_head list
;
853 long nr
; /* objs pending delete */
855 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
856 extern void register_shrinker(struct shrinker
*);
857 extern void unregister_shrinker(struct shrinker
*);
860 * Some shared mappigns will want the pages marked read-only
861 * to track write events. If so, we'll downgrade vm_page_prot
862 * to the private version (using protection_map[] without the
865 static inline int vma_wants_writenotify(struct vm_area_struct
*vma
)
867 unsigned int vm_flags
= vma
->vm_flags
;
869 /* If it was private or non-writable, the write bit is already clear */
870 if ((vm_flags
& (VM_WRITE
|VM_SHARED
)) != ((VM_WRITE
|VM_SHARED
)))
873 /* The backer wishes to know when pages are first written to? */
874 if (vma
->vm_ops
&& vma
->vm_ops
->page_mkwrite
)
877 /* The open routine did something to the protections already? */
878 if (pgprot_val(vma
->vm_page_prot
) !=
879 pgprot_val(protection_map
[vm_flags
&
880 (VM_READ
|VM_WRITE
|VM_EXEC
|VM_SHARED
)]))
883 /* Specialty mapping? */
884 if (vm_flags
& (VM_PFNMAP
|VM_INSERTPAGE
))
887 /* Can the mapping track the dirty pages? */
888 return vma
->vm_file
&& vma
->vm_file
->f_mapping
&&
889 mapping_cap_account_dirty(vma
->vm_file
->f_mapping
);
892 extern pte_t
*FASTCALL(get_locked_pte(struct mm_struct
*mm
, unsigned long addr
, spinlock_t
**ptl
));
894 #ifdef __PAGETABLE_PUD_FOLDED
895 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
896 unsigned long address
)
901 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
904 #ifdef __PAGETABLE_PMD_FOLDED
905 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
906 unsigned long address
)
911 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
914 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
915 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
918 * The following ifdef needed to get the 4level-fixup.h header to work.
919 * Remove it when 4level-fixup.h has been removed.
921 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
922 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
924 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
925 NULL
: pud_offset(pgd
, address
);
928 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
930 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
931 NULL
: pmd_offset(pud
, address
);
933 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
935 #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
937 * We tuck a spinlock to guard each pagetable page into its struct page,
938 * at page->private, with BUILD_BUG_ON to make sure that this will not
939 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
940 * When freeing, reset page->mapping so free_pages_check won't complain.
942 #define __pte_lockptr(page) &((page)->ptl)
943 #define pte_lock_init(_page) do { \
944 spin_lock_init(__pte_lockptr(_page)); \
946 #define pte_lock_deinit(page) ((page)->mapping = NULL)
947 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
950 * We use mm->page_table_lock to guard all pagetable pages of the mm.
952 #define pte_lock_init(page) do {} while (0)
953 #define pte_lock_deinit(page) do {} while (0)
954 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
955 #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
957 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
959 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
960 pte_t *__pte = pte_offset_map(pmd, address); \
966 #define pte_unmap_unlock(pte, ptl) do { \
971 #define pte_alloc_map(mm, pmd, address) \
972 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
973 NULL: pte_offset_map(pmd, address))
975 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
976 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
977 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
979 #define pte_alloc_kernel(pmd, address) \
980 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
981 NULL: pte_offset_kernel(pmd, address))
983 extern void free_area_init(unsigned long * zones_size
);
984 extern void free_area_init_node(int nid
, pg_data_t
*pgdat
,
985 unsigned long * zones_size
, unsigned long zone_start_pfn
,
986 unsigned long *zholes_size
);
987 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
989 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
990 * zones, allocate the backing mem_map and account for memory holes in a more
991 * architecture independent manner. This is a substitute for creating the
992 * zone_sizes[] and zholes_size[] arrays and passing them to
993 * free_area_init_node()
995 * An architecture is expected to register range of page frames backed by
996 * physical memory with add_active_range() before calling
997 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
998 * usage, an architecture is expected to do something like
1000 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1002 * for_each_valid_physical_page_range()
1003 * add_active_range(node_id, start_pfn, end_pfn)
1004 * free_area_init_nodes(max_zone_pfns);
1006 * If the architecture guarantees that there are no holes in the ranges
1007 * registered with add_active_range(), free_bootmem_active_regions()
1008 * will call free_bootmem_node() for each registered physical page range.
1009 * Similarly sparse_memory_present_with_active_regions() calls
1010 * memory_present() for each range when SPARSEMEM is enabled.
1012 * See mm/page_alloc.c for more information on each function exposed by
1013 * CONFIG_ARCH_POPULATES_NODE_MAP
1015 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1016 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
1017 unsigned long end_pfn
);
1018 extern void shrink_active_range(unsigned int nid
, unsigned long old_end_pfn
,
1019 unsigned long new_end_pfn
);
1020 extern void push_node_boundaries(unsigned int nid
, unsigned long start_pfn
,
1021 unsigned long end_pfn
);
1022 extern void remove_all_active_ranges(void);
1023 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1024 unsigned long end_pfn
);
1025 extern void get_pfn_range_for_nid(unsigned int nid
,
1026 unsigned long *start_pfn
, unsigned long *end_pfn
);
1027 extern unsigned long find_min_pfn_with_active_regions(void);
1028 extern unsigned long find_max_pfn_with_active_regions(void);
1029 extern void free_bootmem_with_active_regions(int nid
,
1030 unsigned long max_low_pfn
);
1031 extern void sparse_memory_present_with_active_regions(int nid
);
1032 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1033 extern int early_pfn_to_nid(unsigned long pfn
);
1034 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1035 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1036 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1037 extern void memmap_init_zone(unsigned long, int, unsigned long,
1038 unsigned long, enum memmap_context
);
1039 extern void setup_per_zone_pages_min(void);
1040 extern void mem_init(void);
1041 extern void show_mem(void);
1042 extern void si_meminfo(struct sysinfo
* val
);
1043 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1046 extern void setup_per_cpu_pageset(void);
1048 static inline void setup_per_cpu_pageset(void) {}
1052 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1053 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1054 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1055 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1056 struct prio_tree_iter
*iter
);
1058 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1059 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1060 (vma = vma_prio_tree_next(vma, iter)); )
1062 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1063 struct list_head
*list
)
1065 vma
->shared
.vm_set
.parent
= NULL
;
1066 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1070 extern int __vm_enough_memory(long pages
, int cap_sys_admin
);
1071 extern void vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1072 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1073 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1074 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1075 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1076 struct mempolicy
*);
1077 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1078 extern int split_vma(struct mm_struct
*,
1079 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1080 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1081 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1082 struct rb_node
**, struct rb_node
*);
1083 extern void unlink_file_vma(struct vm_area_struct
*);
1084 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1085 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1086 extern void exit_mmap(struct mm_struct
*);
1087 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1088 extern int install_special_mapping(struct mm_struct
*mm
,
1089 unsigned long addr
, unsigned long len
,
1090 unsigned long flags
, struct page
**pages
);
1092 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1094 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1095 unsigned long len
, unsigned long prot
,
1096 unsigned long flag
, unsigned long pgoff
);
1097 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1098 unsigned long len
, unsigned long flags
,
1099 unsigned int vm_flags
, unsigned long pgoff
,
1102 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1103 unsigned long len
, unsigned long prot
,
1104 unsigned long flag
, unsigned long offset
)
1106 unsigned long ret
= -EINVAL
;
1107 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1109 if (!(offset
& ~PAGE_MASK
))
1110 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1115 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1117 extern unsigned long do_brk(unsigned long, unsigned long);
1120 extern unsigned long page_unuse(struct page
*);
1121 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1122 extern void truncate_inode_pages_range(struct address_space
*,
1123 loff_t lstart
, loff_t lend
);
1125 /* generic vm_area_ops exported for stackable file systems */
1126 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1128 /* mm/page-writeback.c */
1129 int write_one_page(struct page
*page
, int wait
);
1132 #define VM_MAX_READAHEAD 128 /* kbytes */
1133 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1134 #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
1135 * turning readahead off */
1137 int do_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1138 pgoff_t offset
, unsigned long nr_to_read
);
1139 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1140 pgoff_t offset
, unsigned long nr_to_read
);
1142 void page_cache_sync_readahead(struct address_space
*mapping
,
1143 struct file_ra_state
*ra
,
1146 unsigned long size
);
1148 void page_cache_async_readahead(struct address_space
*mapping
,
1149 struct file_ra_state
*ra
,
1153 unsigned long size
);
1155 unsigned long max_sane_readahead(unsigned long nr
);
1157 /* Do stack extension */
1158 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1160 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1163 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1164 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1165 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1166 struct vm_area_struct
**pprev
);
1168 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1169 NULL if none. Assume start_addr < end_addr. */
1170 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1172 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1174 if (vma
&& end_addr
<= vma
->vm_start
)
1179 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1181 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1184 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1185 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1186 struct page
*vmalloc_to_page(void *addr
);
1187 unsigned long vmalloc_to_pfn(void *addr
);
1188 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1189 unsigned long pfn
, unsigned long size
, pgprot_t
);
1190 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1191 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1194 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1195 unsigned int foll_flags
);
1196 #define FOLL_WRITE 0x01 /* check pte is writable */
1197 #define FOLL_TOUCH 0x02 /* mark page accessed */
1198 #define FOLL_GET 0x04 /* do get_page on page */
1199 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
1201 typedef int (*pte_fn_t
)(pte_t
*pte
, struct page
*pmd_page
, unsigned long addr
,
1203 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1204 unsigned long size
, pte_fn_t fn
, void *data
);
1206 #ifdef CONFIG_PROC_FS
1207 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1209 static inline void vm_stat_account(struct mm_struct
*mm
,
1210 unsigned long flags
, struct file
*file
, long pages
)
1213 #endif /* CONFIG_PROC_FS */
1215 #ifndef CONFIG_DEBUG_PAGEALLOC
1217 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1220 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1221 #ifdef __HAVE_ARCH_GATE_AREA
1222 int in_gate_area_no_task(unsigned long addr
);
1223 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1225 int in_gate_area_no_task(unsigned long addr
);
1226 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1227 #endif /* __HAVE_ARCH_GATE_AREA */
1229 int drop_caches_sysctl_handler(struct ctl_table
*, int, struct file
*,
1230 void __user
*, size_t *, loff_t
*);
1231 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1232 unsigned long lru_pages
);
1233 void drop_pagecache(void);
1234 void drop_slab(void);
1237 #define randomize_va_space 0
1239 extern int randomize_va_space
;
1242 __attribute__((weak
)) const char *arch_vma_name(struct vm_area_struct
*vma
);
1244 #endif /* __KERNEL__ */
1245 #endif /* _LINUX_MM_H */