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>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/range.h>
16 #include <linux/pfn.h>
17 #include <linux/bit_spinlock.h>
23 struct writeback_control
;
25 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
26 extern unsigned long max_mapnr
;
29 extern unsigned long num_physpages
;
30 extern unsigned long totalram_pages
;
31 extern void * high_memory
;
32 extern int page_cluster
;
35 extern int sysctl_legacy_va_layout
;
37 #define sysctl_legacy_va_layout 0
41 #include <asm/pgtable.h>
42 #include <asm/processor.h>
44 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
46 /* to align the pointer to the (next) page boundary */
47 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
50 * Linux kernel virtual memory manager primitives.
51 * The idea being to have a "virtual" mm in the same way
52 * we have a virtual fs - giving a cleaner interface to the
53 * mm details, and allowing different kinds of memory mappings
54 * (from shared memory to executable loading to arbitrary
58 extern struct kmem_cache
*vm_area_cachep
;
61 extern struct rb_root nommu_region_tree
;
62 extern struct rw_semaphore nommu_region_sem
;
64 extern unsigned int kobjsize(const void *objp
);
68 * vm_flags in vm_area_struct, see mm_types.h.
70 #define VM_READ 0x00000001 /* currently active flags */
71 #define VM_WRITE 0x00000002
72 #define VM_EXEC 0x00000004
73 #define VM_SHARED 0x00000008
75 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
76 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
77 #define VM_MAYWRITE 0x00000020
78 #define VM_MAYEXEC 0x00000040
79 #define VM_MAYSHARE 0x00000080
81 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
82 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
83 #define VM_GROWSUP 0x00000200
85 #define VM_GROWSUP 0x00000000
86 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
88 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
89 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
91 #define VM_EXECUTABLE 0x00001000
92 #define VM_LOCKED 0x00002000
93 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
95 /* Used by sys_madvise() */
96 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
99 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
101 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
102 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
104 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
105 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
106 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
107 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
109 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
111 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
112 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
114 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
115 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
116 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
117 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
118 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
120 /* Bits set in the VMA until the stack is in its final location */
121 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
123 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
124 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
127 #ifdef CONFIG_STACK_GROWSUP
128 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
130 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
133 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
134 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
135 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
136 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
137 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
140 * special vmas that are non-mergable, non-mlock()able
142 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
145 * mapping from the currently active vm_flags protection bits (the
146 * low four bits) to a page protection mask..
148 extern pgprot_t protection_map
[16];
150 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
151 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
152 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
153 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
154 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
157 * This interface is used by x86 PAT code to identify a pfn mapping that is
158 * linear over entire vma. This is to optimize PAT code that deals with
159 * marking the physical region with a particular prot. This is not for generic
160 * mm use. Note also that this check will not work if the pfn mapping is
161 * linear for a vma starting at physical address 0. In which case PAT code
162 * falls back to slow path of reserving physical range page by page.
164 static inline int is_linear_pfn_mapping(struct vm_area_struct
*vma
)
166 return (vma
->vm_flags
& VM_PFN_AT_MMAP
);
169 static inline int is_pfn_mapping(struct vm_area_struct
*vma
)
171 return (vma
->vm_flags
& VM_PFNMAP
);
175 * vm_fault is filled by the the pagefault handler and passed to the vma's
176 * ->fault function. The vma's ->fault is responsible for returning a bitmask
177 * of VM_FAULT_xxx flags that give details about how the fault was handled.
179 * pgoff should be used in favour of virtual_address, if possible. If pgoff
180 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
184 unsigned int flags
; /* FAULT_FLAG_xxx flags */
185 pgoff_t pgoff
; /* Logical page offset based on vma */
186 void __user
*virtual_address
; /* Faulting virtual address */
188 struct page
*page
; /* ->fault handlers should return a
189 * page here, unless VM_FAULT_NOPAGE
190 * is set (which is also implied by
196 * These are the virtual MM functions - opening of an area, closing and
197 * unmapping it (needed to keep files on disk up-to-date etc), pointer
198 * to the functions called when a no-page or a wp-page exception occurs.
200 struct vm_operations_struct
{
201 void (*open
)(struct vm_area_struct
* area
);
202 void (*close
)(struct vm_area_struct
* area
);
203 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
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 vm_fault
*vmf
);
209 /* called by access_process_vm when get_user_pages() fails, typically
210 * for use by special VMAs that can switch between memory and hardware
212 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
213 void *buf
, int len
, int write
);
216 * set_policy() op must add a reference to any non-NULL @new mempolicy
217 * to hold the policy upon return. Caller should pass NULL @new to
218 * remove a policy and fall back to surrounding context--i.e. do not
219 * install a MPOL_DEFAULT policy, nor the task or system default
222 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
225 * get_policy() op must add reference [mpol_get()] to any policy at
226 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
227 * in mm/mempolicy.c will do this automatically.
228 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
229 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
230 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
231 * must return NULL--i.e., do not "fallback" to task or system default
234 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
236 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
237 const nodemask_t
*to
, unsigned long flags
);
244 #define page_private(page) ((page)->private)
245 #define set_page_private(page, v) ((page)->private = (v))
248 * FIXME: take this include out, include page-flags.h in
249 * files which need it (119 of them)
251 #include <linux/page-flags.h>
252 #include <linux/huge_mm.h>
255 * Methods to modify the page usage count.
257 * What counts for a page usage:
258 * - cache mapping (page->mapping)
259 * - private data (page->private)
260 * - page mapped in a task's page tables, each mapping
261 * is counted separately
263 * Also, many kernel routines increase the page count before a critical
264 * routine so they can be sure the page doesn't go away from under them.
268 * Drop a ref, return true if the refcount fell to zero (the page has no users)
270 static inline int put_page_testzero(struct page
*page
)
272 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
273 return atomic_dec_and_test(&page
->_count
);
277 * Try to grab a ref unless the page has a refcount of zero, return false if
280 static inline int get_page_unless_zero(struct page
*page
)
282 return atomic_inc_not_zero(&page
->_count
);
285 extern int page_is_ram(unsigned long pfn
);
287 /* Support for virtually mapped pages */
288 struct page
*vmalloc_to_page(const void *addr
);
289 unsigned long vmalloc_to_pfn(const void *addr
);
292 * Determine if an address is within the vmalloc range
294 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
295 * is no special casing required.
297 static inline int is_vmalloc_addr(const void *x
)
300 unsigned long addr
= (unsigned long)x
;
302 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
308 extern int is_vmalloc_or_module_addr(const void *x
);
310 static inline int is_vmalloc_or_module_addr(const void *x
)
316 static inline void compound_lock(struct page
*page
)
318 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
319 bit_spin_lock(PG_compound_lock
, &page
->flags
);
323 static inline void compound_unlock(struct page
*page
)
325 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
326 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
330 static inline unsigned long compound_lock_irqsave(struct page
*page
)
332 unsigned long uninitialized_var(flags
);
333 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
334 local_irq_save(flags
);
340 static inline void compound_unlock_irqrestore(struct page
*page
,
343 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
344 compound_unlock(page
);
345 local_irq_restore(flags
);
349 static inline struct page
*compound_head(struct page
*page
)
351 if (unlikely(PageTail(page
)))
352 return page
->first_page
;
356 static inline int page_count(struct page
*page
)
358 return atomic_read(&compound_head(page
)->_count
);
361 static inline void get_page(struct page
*page
)
364 * Getting a normal page or the head of a compound page
365 * requires to already have an elevated page->_count. Only if
366 * we're getting a tail page, the elevated page->_count is
367 * required only in the head page, so for tail pages the
368 * bugcheck only verifies that the page->_count isn't
371 VM_BUG_ON(atomic_read(&page
->_count
) < !PageTail(page
));
372 atomic_inc(&page
->_count
);
374 * Getting a tail page will elevate both the head and tail
377 if (unlikely(PageTail(page
))) {
379 * This is safe only because
380 * __split_huge_page_refcount can't run under
383 VM_BUG_ON(atomic_read(&page
->first_page
->_count
) <= 0);
384 atomic_inc(&page
->first_page
->_count
);
388 static inline struct page
*virt_to_head_page(const void *x
)
390 struct page
*page
= virt_to_page(x
);
391 return compound_head(page
);
395 * Setup the page count before being freed into the page allocator for
396 * the first time (boot or memory hotplug)
398 static inline void init_page_count(struct page
*page
)
400 atomic_set(&page
->_count
, 1);
404 * PageBuddy() indicate that the page is free and in the buddy system
405 * (see mm/page_alloc.c).
407 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
408 * -2 so that an underflow of the page_mapcount() won't be mistaken
409 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
410 * efficiently by most CPU architectures.
412 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
414 static inline int PageBuddy(struct page
*page
)
416 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
419 static inline void __SetPageBuddy(struct page
*page
)
421 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
422 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
425 static inline void __ClearPageBuddy(struct page
*page
)
427 VM_BUG_ON(!PageBuddy(page
));
428 atomic_set(&page
->_mapcount
, -1);
431 void put_page(struct page
*page
);
432 void put_pages_list(struct list_head
*pages
);
434 void split_page(struct page
*page
, unsigned int order
);
435 int split_free_page(struct page
*page
);
438 * Compound pages have a destructor function. Provide a
439 * prototype for that function and accessor functions.
440 * These are _only_ valid on the head of a PG_compound page.
442 typedef void compound_page_dtor(struct page
*);
444 static inline void set_compound_page_dtor(struct page
*page
,
445 compound_page_dtor
*dtor
)
447 page
[1].lru
.next
= (void *)dtor
;
450 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
452 return (compound_page_dtor
*)page
[1].lru
.next
;
455 static inline int compound_order(struct page
*page
)
459 return (unsigned long)page
[1].lru
.prev
;
462 static inline int compound_trans_order(struct page
*page
)
470 flags
= compound_lock_irqsave(page
);
471 order
= compound_order(page
);
472 compound_unlock_irqrestore(page
, flags
);
476 static inline void set_compound_order(struct page
*page
, unsigned long order
)
478 page
[1].lru
.prev
= (void *)order
;
483 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
484 * servicing faults for write access. In the normal case, do always want
485 * pte_mkwrite. But get_user_pages can cause write faults for mappings
486 * that do not have writing enabled, when used by access_process_vm.
488 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
490 if (likely(vma
->vm_flags
& VM_WRITE
))
491 pte
= pte_mkwrite(pte
);
497 * Multiple processes may "see" the same page. E.g. for untouched
498 * mappings of /dev/null, all processes see the same page full of
499 * zeroes, and text pages of executables and shared libraries have
500 * only one copy in memory, at most, normally.
502 * For the non-reserved pages, page_count(page) denotes a reference count.
503 * page_count() == 0 means the page is free. page->lru is then used for
504 * freelist management in the buddy allocator.
505 * page_count() > 0 means the page has been allocated.
507 * Pages are allocated by the slab allocator in order to provide memory
508 * to kmalloc and kmem_cache_alloc. In this case, the management of the
509 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
510 * unless a particular usage is carefully commented. (the responsibility of
511 * freeing the kmalloc memory is the caller's, of course).
513 * A page may be used by anyone else who does a __get_free_page().
514 * In this case, page_count still tracks the references, and should only
515 * be used through the normal accessor functions. The top bits of page->flags
516 * and page->virtual store page management information, but all other fields
517 * are unused and could be used privately, carefully. The management of this
518 * page is the responsibility of the one who allocated it, and those who have
519 * subsequently been given references to it.
521 * The other pages (we may call them "pagecache pages") are completely
522 * managed by the Linux memory manager: I/O, buffers, swapping etc.
523 * The following discussion applies only to them.
525 * A pagecache page contains an opaque `private' member, which belongs to the
526 * page's address_space. Usually, this is the address of a circular list of
527 * the page's disk buffers. PG_private must be set to tell the VM to call
528 * into the filesystem to release these pages.
530 * A page may belong to an inode's memory mapping. In this case, page->mapping
531 * is the pointer to the inode, and page->index is the file offset of the page,
532 * in units of PAGE_CACHE_SIZE.
534 * If pagecache pages are not associated with an inode, they are said to be
535 * anonymous pages. These may become associated with the swapcache, and in that
536 * case PG_swapcache is set, and page->private is an offset into the swapcache.
538 * In either case (swapcache or inode backed), the pagecache itself holds one
539 * reference to the page. Setting PG_private should also increment the
540 * refcount. The each user mapping also has a reference to the page.
542 * The pagecache pages are stored in a per-mapping radix tree, which is
543 * rooted at mapping->page_tree, and indexed by offset.
544 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
545 * lists, we instead now tag pages as dirty/writeback in the radix tree.
547 * All pagecache pages may be subject to I/O:
548 * - inode pages may need to be read from disk,
549 * - inode pages which have been modified and are MAP_SHARED may need
550 * to be written back to the inode on disk,
551 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
552 * modified may need to be swapped out to swap space and (later) to be read
557 * The zone field is never updated after free_area_init_core()
558 * sets it, so none of the operations on it need to be atomic.
563 * page->flags layout:
565 * There are three possibilities for how page->flags get
566 * laid out. The first is for the normal case, without
567 * sparsemem. The second is for sparsemem when there is
568 * plenty of space for node and section. The last is when
569 * we have run out of space and have to fall back to an
570 * alternate (slower) way of determining the node.
572 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
573 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
574 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
576 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
577 #define SECTIONS_WIDTH SECTIONS_SHIFT
579 #define SECTIONS_WIDTH 0
582 #define ZONES_WIDTH ZONES_SHIFT
584 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
585 #define NODES_WIDTH NODES_SHIFT
587 #ifdef CONFIG_SPARSEMEM_VMEMMAP
588 #error "Vmemmap: No space for nodes field in page flags"
590 #define NODES_WIDTH 0
593 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
594 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
595 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
596 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
599 * We are going to use the flags for the page to node mapping if its in
600 * there. This includes the case where there is no node, so it is implicit.
602 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
603 #define NODE_NOT_IN_PAGE_FLAGS
606 #ifndef PFN_SECTION_SHIFT
607 #define PFN_SECTION_SHIFT 0
611 * Define the bit shifts to access each section. For non-existant
612 * sections we define the shift as 0; that plus a 0 mask ensures
613 * the compiler will optimise away reference to them.
615 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
616 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
617 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
619 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
620 #ifdef NODE_NOT_IN_PAGE_FLAGS
621 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
622 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
623 SECTIONS_PGOFF : ZONES_PGOFF)
625 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
626 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
627 NODES_PGOFF : ZONES_PGOFF)
630 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
632 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
633 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
636 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
637 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
638 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
639 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
641 static inline enum zone_type
page_zonenum(struct page
*page
)
643 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
647 * The identification function is only used by the buddy allocator for
648 * determining if two pages could be buddies. We are not really
649 * identifying a zone since we could be using a the section number
650 * id if we have not node id available in page flags.
651 * We guarantee only that it will return the same value for two
652 * combinable pages in a zone.
654 static inline int page_zone_id(struct page
*page
)
656 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
659 static inline int zone_to_nid(struct zone
*zone
)
668 #ifdef NODE_NOT_IN_PAGE_FLAGS
669 extern int page_to_nid(struct page
*page
);
671 static inline int page_to_nid(struct page
*page
)
673 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
677 static inline struct zone
*page_zone(struct page
*page
)
679 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
682 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
683 static inline unsigned long page_to_section(struct page
*page
)
685 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
689 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
691 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
692 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
695 static inline void set_page_node(struct page
*page
, unsigned long node
)
697 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
698 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
701 static inline void set_page_section(struct page
*page
, unsigned long section
)
703 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
704 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
707 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
708 unsigned long node
, unsigned long pfn
)
710 set_page_zone(page
, zone
);
711 set_page_node(page
, node
);
712 set_page_section(page
, pfn_to_section_nr(pfn
));
716 * Some inline functions in vmstat.h depend on page_zone()
718 #include <linux/vmstat.h>
720 static __always_inline
void *lowmem_page_address(struct page
*page
)
722 return __va(PFN_PHYS(page_to_pfn(page
)));
725 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
726 #define HASHED_PAGE_VIRTUAL
729 #if defined(WANT_PAGE_VIRTUAL)
730 #define page_address(page) ((page)->virtual)
731 #define set_page_address(page, address) \
733 (page)->virtual = (address); \
735 #define page_address_init() do { } while(0)
738 #if defined(HASHED_PAGE_VIRTUAL)
739 void *page_address(struct page
*page
);
740 void set_page_address(struct page
*page
, void *virtual);
741 void page_address_init(void);
744 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
745 #define page_address(page) lowmem_page_address(page)
746 #define set_page_address(page, address) do { } while(0)
747 #define page_address_init() do { } while(0)
751 * On an anonymous page mapped into a user virtual memory area,
752 * page->mapping points to its anon_vma, not to a struct address_space;
753 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
755 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
756 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
757 * and then page->mapping points, not to an anon_vma, but to a private
758 * structure which KSM associates with that merged page. See ksm.h.
760 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
762 * Please note that, confusingly, "page_mapping" refers to the inode
763 * address_space which maps the page from disk; whereas "page_mapped"
764 * refers to user virtual address space into which the page is mapped.
766 #define PAGE_MAPPING_ANON 1
767 #define PAGE_MAPPING_KSM 2
768 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
770 extern struct address_space swapper_space
;
771 static inline struct address_space
*page_mapping(struct page
*page
)
773 struct address_space
*mapping
= page
->mapping
;
775 VM_BUG_ON(PageSlab(page
));
776 if (unlikely(PageSwapCache(page
)))
777 mapping
= &swapper_space
;
778 else if ((unsigned long)mapping
& PAGE_MAPPING_ANON
)
783 /* Neutral page->mapping pointer to address_space or anon_vma or other */
784 static inline void *page_rmapping(struct page
*page
)
786 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
789 static inline int PageAnon(struct page
*page
)
791 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
795 * Return the pagecache index of the passed page. Regular pagecache pages
796 * use ->index whereas swapcache pages use ->private
798 static inline pgoff_t
page_index(struct page
*page
)
800 if (unlikely(PageSwapCache(page
)))
801 return page_private(page
);
806 * The atomic page->_mapcount, like _count, starts from -1:
807 * so that transitions both from it and to it can be tracked,
808 * using atomic_inc_and_test and atomic_add_negative(-1).
810 static inline void reset_page_mapcount(struct page
*page
)
812 atomic_set(&(page
)->_mapcount
, -1);
815 static inline int page_mapcount(struct page
*page
)
817 return atomic_read(&(page
)->_mapcount
) + 1;
821 * Return true if this page is mapped into pagetables.
823 static inline int page_mapped(struct page
*page
)
825 return atomic_read(&(page
)->_mapcount
) >= 0;
829 * Different kinds of faults, as returned by handle_mm_fault().
830 * Used to decide whether a process gets delivered SIGBUS or
831 * just gets major/minor fault counters bumped up.
834 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
836 #define VM_FAULT_OOM 0x0001
837 #define VM_FAULT_SIGBUS 0x0002
838 #define VM_FAULT_MAJOR 0x0004
839 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
840 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
841 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
843 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
844 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
845 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
847 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
849 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
850 VM_FAULT_HWPOISON_LARGE)
852 /* Encode hstate index for a hwpoisoned large page */
853 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
854 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
857 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
859 extern void pagefault_out_of_memory(void);
861 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
864 * Flags passed to __show_mem() and __show_free_areas() to suppress output in
867 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
869 extern void show_free_areas(void);
870 extern void __show_free_areas(unsigned int flags
);
872 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
873 struct file
*shmem_file_setup(const char *name
, loff_t size
, unsigned long flags
);
874 int shmem_zero_setup(struct vm_area_struct
*);
877 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
881 unsigned long flags
);
884 extern int can_do_mlock(void);
885 extern int user_shm_lock(size_t, struct user_struct
*);
886 extern void user_shm_unlock(size_t, struct user_struct
*);
889 * Parameter block passed down to zap_pte_range in exceptional cases.
892 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
893 struct address_space
*check_mapping
; /* Check page->mapping if set */
894 pgoff_t first_index
; /* Lowest page->index to unmap */
895 pgoff_t last_index
; /* Highest page->index to unmap */
896 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
897 unsigned long truncate_count
; /* Compare vm_truncate_count */
900 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
903 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
905 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
906 unsigned long size
, struct zap_details
*);
907 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
908 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
909 unsigned long end_addr
, unsigned long *nr_accounted
,
910 struct zap_details
*);
913 * mm_walk - callbacks for walk_page_range
914 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
915 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
916 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
917 * this handler is required to be able to handle
918 * pmd_trans_huge() pmds. They may simply choose to
919 * split_huge_page() instead of handling it explicitly.
920 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
921 * @pte_hole: if set, called for each hole at all levels
922 * @hugetlb_entry: if set, called for each hugetlb entry
924 * (see walk_page_range for more details)
927 int (*pgd_entry
)(pgd_t
*, unsigned long, unsigned long, struct mm_walk
*);
928 int (*pud_entry
)(pud_t
*, unsigned long, unsigned long, struct mm_walk
*);
929 int (*pmd_entry
)(pmd_t
*, unsigned long, unsigned long, struct mm_walk
*);
930 int (*pte_entry
)(pte_t
*, unsigned long, unsigned long, struct mm_walk
*);
931 int (*pte_hole
)(unsigned long, unsigned long, struct mm_walk
*);
932 int (*hugetlb_entry
)(pte_t
*, unsigned long,
933 unsigned long, unsigned long, struct mm_walk
*);
934 struct mm_struct
*mm
;
938 int walk_page_range(unsigned long addr
, unsigned long end
,
939 struct mm_walk
*walk
);
940 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
941 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
942 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
943 struct vm_area_struct
*vma
);
944 void unmap_mapping_range(struct address_space
*mapping
,
945 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
946 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
948 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
949 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
950 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
951 void *buf
, int len
, int write
);
953 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
954 loff_t
const holebegin
, loff_t
const holelen
)
956 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
959 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
960 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
961 extern int vmtruncate(struct inode
*inode
, loff_t offset
);
962 extern int vmtruncate_range(struct inode
*inode
, loff_t offset
, loff_t end
);
964 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
965 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
967 int invalidate_inode_page(struct page
*page
);
970 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
971 unsigned long address
, unsigned int flags
);
973 static inline int handle_mm_fault(struct mm_struct
*mm
,
974 struct vm_area_struct
*vma
, unsigned long address
,
977 /* should never happen if there's no MMU */
979 return VM_FAULT_SIGBUS
;
983 extern int make_pages_present(unsigned long addr
, unsigned long end
);
984 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
986 int __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
987 unsigned long start
, int len
, unsigned int foll_flags
,
988 struct page
**pages
, struct vm_area_struct
**vmas
,
990 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
991 unsigned long start
, int nr_pages
, int write
, int force
,
992 struct page
**pages
, struct vm_area_struct
**vmas
);
993 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
994 struct page
**pages
);
995 struct page
*get_dump_page(unsigned long addr
);
997 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
998 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
1000 int __set_page_dirty_nobuffers(struct page
*page
);
1001 int __set_page_dirty_no_writeback(struct page
*page
);
1002 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1004 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1005 void account_page_writeback(struct page
*page
);
1006 int set_page_dirty(struct page
*page
);
1007 int set_page_dirty_lock(struct page
*page
);
1008 int clear_page_dirty_for_io(struct page
*page
);
1010 /* Is the vma a continuation of the stack vma above it? */
1011 static inline int vma_stack_continue(struct vm_area_struct
*vma
, unsigned long addr
)
1013 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1016 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1017 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1018 unsigned long new_addr
, unsigned long len
);
1019 extern unsigned long do_mremap(unsigned long addr
,
1020 unsigned long old_len
, unsigned long new_len
,
1021 unsigned long flags
, unsigned long new_addr
);
1022 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1023 struct vm_area_struct
**pprev
, unsigned long start
,
1024 unsigned long end
, unsigned long newflags
);
1027 * doesn't attempt to fault and will return short.
1029 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1030 struct page
**pages
);
1032 * per-process(per-mm_struct) statistics.
1034 #if defined(SPLIT_RSS_COUNTING)
1036 * The mm counters are not protected by its page_table_lock,
1037 * so must be incremented atomically.
1039 static inline void set_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1041 atomic_long_set(&mm
->rss_stat
.count
[member
], value
);
1044 unsigned long get_mm_counter(struct mm_struct
*mm
, int member
);
1046 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1048 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1051 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1053 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1056 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1058 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1061 #else /* !USE_SPLIT_PTLOCKS */
1063 * The mm counters are protected by its page_table_lock,
1064 * so can be incremented directly.
1066 static inline void set_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1068 mm
->rss_stat
.count
[member
] = value
;
1071 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1073 return mm
->rss_stat
.count
[member
];
1076 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1078 mm
->rss_stat
.count
[member
] += value
;
1081 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1083 mm
->rss_stat
.count
[member
]++;
1086 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1088 mm
->rss_stat
.count
[member
]--;
1091 #endif /* !USE_SPLIT_PTLOCKS */
1093 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1095 return get_mm_counter(mm
, MM_FILEPAGES
) +
1096 get_mm_counter(mm
, MM_ANONPAGES
);
1099 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1101 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1104 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1106 return max(mm
->hiwater_vm
, mm
->total_vm
);
1109 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1111 unsigned long _rss
= get_mm_rss(mm
);
1113 if ((mm
)->hiwater_rss
< _rss
)
1114 (mm
)->hiwater_rss
= _rss
;
1117 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1119 if (mm
->hiwater_vm
< mm
->total_vm
)
1120 mm
->hiwater_vm
= mm
->total_vm
;
1123 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1124 struct mm_struct
*mm
)
1126 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1128 if (*maxrss
< hiwater_rss
)
1129 *maxrss
= hiwater_rss
;
1132 #if defined(SPLIT_RSS_COUNTING)
1133 void sync_mm_rss(struct task_struct
*task
, struct mm_struct
*mm
);
1135 static inline void sync_mm_rss(struct task_struct
*task
, struct mm_struct
*mm
)
1141 * A callback you can register to apply pressure to ageable caches.
1143 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
1144 * look through the least-recently-used 'nr_to_scan' entries and
1145 * attempt to free them up. It should return the number of objects
1146 * which remain in the cache. If it returns -1, it means it cannot do
1147 * any scanning at this time (eg. there is a risk of deadlock).
1149 * The 'gfpmask' refers to the allocation we are currently trying to
1152 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1153 * querying the cache size, so a fastpath for that case is appropriate.
1156 int (*shrink
)(struct shrinker
*, int nr_to_scan
, gfp_t gfp_mask
);
1157 int seeks
; /* seeks to recreate an obj */
1159 /* These are for internal use */
1160 struct list_head list
;
1161 long nr
; /* objs pending delete */
1163 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1164 extern void register_shrinker(struct shrinker
*);
1165 extern void unregister_shrinker(struct shrinker
*);
1167 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1169 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1171 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1175 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1179 #ifdef __PAGETABLE_PUD_FOLDED
1180 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1181 unsigned long address
)
1186 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1189 #ifdef __PAGETABLE_PMD_FOLDED
1190 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1191 unsigned long address
)
1196 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1199 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1200 pmd_t
*pmd
, unsigned long address
);
1201 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1204 * The following ifdef needed to get the 4level-fixup.h header to work.
1205 * Remove it when 4level-fixup.h has been removed.
1207 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1208 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1210 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1211 NULL
: pud_offset(pgd
, address
);
1214 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1216 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1217 NULL
: pmd_offset(pud
, address
);
1219 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1221 #if USE_SPLIT_PTLOCKS
1223 * We tuck a spinlock to guard each pagetable page into its struct page,
1224 * at page->private, with BUILD_BUG_ON to make sure that this will not
1225 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1226 * When freeing, reset page->mapping so free_pages_check won't complain.
1228 #define __pte_lockptr(page) &((page)->ptl)
1229 #define pte_lock_init(_page) do { \
1230 spin_lock_init(__pte_lockptr(_page)); \
1232 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1233 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1234 #else /* !USE_SPLIT_PTLOCKS */
1236 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1238 #define pte_lock_init(page) do {} while (0)
1239 #define pte_lock_deinit(page) do {} while (0)
1240 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1241 #endif /* USE_SPLIT_PTLOCKS */
1243 static inline void pgtable_page_ctor(struct page
*page
)
1245 pte_lock_init(page
);
1246 inc_zone_page_state(page
, NR_PAGETABLE
);
1249 static inline void pgtable_page_dtor(struct page
*page
)
1251 pte_lock_deinit(page
);
1252 dec_zone_page_state(page
, NR_PAGETABLE
);
1255 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1257 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1258 pte_t *__pte = pte_offset_map(pmd, address); \
1264 #define pte_unmap_unlock(pte, ptl) do { \
1269 #define pte_alloc_map(mm, vma, pmd, address) \
1270 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1272 NULL: pte_offset_map(pmd, address))
1274 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1275 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1277 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1279 #define pte_alloc_kernel(pmd, address) \
1280 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1281 NULL: pte_offset_kernel(pmd, address))
1283 extern void free_area_init(unsigned long * zones_size
);
1284 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1285 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1286 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1288 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1289 * zones, allocate the backing mem_map and account for memory holes in a more
1290 * architecture independent manner. This is a substitute for creating the
1291 * zone_sizes[] and zholes_size[] arrays and passing them to
1292 * free_area_init_node()
1294 * An architecture is expected to register range of page frames backed by
1295 * physical memory with add_active_range() before calling
1296 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1297 * usage, an architecture is expected to do something like
1299 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1301 * for_each_valid_physical_page_range()
1302 * add_active_range(node_id, start_pfn, end_pfn)
1303 * free_area_init_nodes(max_zone_pfns);
1305 * If the architecture guarantees that there are no holes in the ranges
1306 * registered with add_active_range(), free_bootmem_active_regions()
1307 * will call free_bootmem_node() for each registered physical page range.
1308 * Similarly sparse_memory_present_with_active_regions() calls
1309 * memory_present() for each range when SPARSEMEM is enabled.
1311 * See mm/page_alloc.c for more information on each function exposed by
1312 * CONFIG_ARCH_POPULATES_NODE_MAP
1314 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1315 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
1316 unsigned long end_pfn
);
1317 extern void remove_active_range(unsigned int nid
, unsigned long start_pfn
,
1318 unsigned long end_pfn
);
1319 extern void remove_all_active_ranges(void);
1320 void sort_node_map(void);
1321 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1322 unsigned long end_pfn
);
1323 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1324 unsigned long end_pfn
);
1325 extern void get_pfn_range_for_nid(unsigned int nid
,
1326 unsigned long *start_pfn
, unsigned long *end_pfn
);
1327 extern unsigned long find_min_pfn_with_active_regions(void);
1328 extern void free_bootmem_with_active_regions(int nid
,
1329 unsigned long max_low_pfn
);
1330 int add_from_early_node_map(struct range
*range
, int az
,
1331 int nr_range
, int nid
);
1332 u64 __init
find_memory_core_early(int nid
, u64 size
, u64 align
,
1333 u64 goal
, u64 limit
);
1334 typedef int (*work_fn_t
)(unsigned long, unsigned long, void *);
1335 extern void work_with_active_regions(int nid
, work_fn_t work_fn
, void *data
);
1336 extern void sparse_memory_present_with_active_regions(int nid
);
1337 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1339 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1340 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1341 static inline int __early_pfn_to_nid(unsigned long pfn
)
1346 /* please see mm/page_alloc.c */
1347 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1348 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1349 /* there is a per-arch backend function. */
1350 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1351 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1354 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1355 extern void memmap_init_zone(unsigned long, int, unsigned long,
1356 unsigned long, enum memmap_context
);
1357 extern void setup_per_zone_wmarks(void);
1358 extern void calculate_zone_inactive_ratio(struct zone
*zone
);
1359 extern void mem_init(void);
1360 extern void __init
mmap_init(void);
1361 extern void show_mem(void);
1362 extern void __show_mem(unsigned int flags
);
1363 extern void si_meminfo(struct sysinfo
* val
);
1364 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1365 extern int after_bootmem
;
1367 extern void setup_per_cpu_pageset(void);
1369 extern void zone_pcp_update(struct zone
*zone
);
1372 extern atomic_long_t mmap_pages_allocated
;
1373 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1376 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1377 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1378 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1379 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1380 struct prio_tree_iter
*iter
);
1382 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1383 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1384 (vma = vma_prio_tree_next(vma, iter)); )
1386 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1387 struct list_head
*list
)
1389 vma
->shared
.vm_set
.parent
= NULL
;
1390 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1394 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1395 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1396 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1397 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1398 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1399 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1400 struct mempolicy
*);
1401 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1402 extern int split_vma(struct mm_struct
*,
1403 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1404 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1405 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1406 struct rb_node
**, struct rb_node
*);
1407 extern void unlink_file_vma(struct vm_area_struct
*);
1408 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1409 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1410 extern void exit_mmap(struct mm_struct
*);
1412 extern int mm_take_all_locks(struct mm_struct
*mm
);
1413 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1415 #ifdef CONFIG_PROC_FS
1416 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1417 extern void added_exe_file_vma(struct mm_struct
*mm
);
1418 extern void removed_exe_file_vma(struct mm_struct
*mm
);
1420 static inline void added_exe_file_vma(struct mm_struct
*mm
)
1423 static inline void removed_exe_file_vma(struct mm_struct
*mm
)
1425 #endif /* CONFIG_PROC_FS */
1427 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1428 extern int install_special_mapping(struct mm_struct
*mm
,
1429 unsigned long addr
, unsigned long len
,
1430 unsigned long flags
, struct page
**pages
);
1432 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1434 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1435 unsigned long len
, unsigned long prot
,
1436 unsigned long flag
, unsigned long pgoff
);
1437 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1438 unsigned long len
, unsigned long flags
,
1439 unsigned int vm_flags
, unsigned long pgoff
);
1441 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1442 unsigned long len
, unsigned long prot
,
1443 unsigned long flag
, unsigned long offset
)
1445 unsigned long ret
= -EINVAL
;
1446 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1448 if (!(offset
& ~PAGE_MASK
))
1449 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1454 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1456 extern unsigned long do_brk(unsigned long, unsigned long);
1459 extern unsigned long page_unuse(struct page
*);
1460 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1461 extern void truncate_inode_pages_range(struct address_space
*,
1462 loff_t lstart
, loff_t lend
);
1464 /* generic vm_area_ops exported for stackable file systems */
1465 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1467 /* mm/page-writeback.c */
1468 int write_one_page(struct page
*page
, int wait
);
1469 void task_dirty_inc(struct task_struct
*tsk
);
1472 #define VM_MAX_READAHEAD 128 /* kbytes */
1473 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1475 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1476 pgoff_t offset
, unsigned long nr_to_read
);
1478 void page_cache_sync_readahead(struct address_space
*mapping
,
1479 struct file_ra_state
*ra
,
1482 unsigned long size
);
1484 void page_cache_async_readahead(struct address_space
*mapping
,
1485 struct file_ra_state
*ra
,
1489 unsigned long size
);
1491 unsigned long max_sane_readahead(unsigned long nr
);
1492 unsigned long ra_submit(struct file_ra_state
*ra
,
1493 struct address_space
*mapping
,
1496 /* Do stack extension */
1497 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1499 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1501 #define expand_upwards(vma, address) do { } while (0)
1503 extern int expand_stack_downwards(struct vm_area_struct
*vma
,
1504 unsigned long address
);
1506 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1507 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1508 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1509 struct vm_area_struct
**pprev
);
1511 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1512 NULL if none. Assume start_addr < end_addr. */
1513 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1515 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1517 if (vma
&& end_addr
<= vma
->vm_start
)
1522 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1524 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1528 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1530 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1536 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1537 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1538 unsigned long pfn
, unsigned long size
, pgprot_t
);
1539 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1540 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1542 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1545 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1546 unsigned int foll_flags
);
1547 #define FOLL_WRITE 0x01 /* check pte is writable */
1548 #define FOLL_TOUCH 0x02 /* mark page accessed */
1549 #define FOLL_GET 0x04 /* do get_page on page */
1550 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1551 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1552 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1553 * and return without waiting upon it */
1554 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1555 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1556 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1558 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1560 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1561 unsigned long size
, pte_fn_t fn
, void *data
);
1563 #ifdef CONFIG_PROC_FS
1564 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1566 static inline void vm_stat_account(struct mm_struct
*mm
,
1567 unsigned long flags
, struct file
*file
, long pages
)
1570 #endif /* CONFIG_PROC_FS */
1572 #ifdef CONFIG_DEBUG_PAGEALLOC
1573 extern int debug_pagealloc_enabled
;
1575 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1577 static inline void enable_debug_pagealloc(void)
1579 debug_pagealloc_enabled
= 1;
1581 #ifdef CONFIG_HIBERNATION
1582 extern bool kernel_page_present(struct page
*page
);
1583 #endif /* CONFIG_HIBERNATION */
1586 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1587 static inline void enable_debug_pagealloc(void)
1590 #ifdef CONFIG_HIBERNATION
1591 static inline bool kernel_page_present(struct page
*page
) { return true; }
1592 #endif /* CONFIG_HIBERNATION */
1595 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1596 #ifdef __HAVE_ARCH_GATE_AREA
1597 int in_gate_area_no_task(unsigned long addr
);
1598 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1600 int in_gate_area_no_task(unsigned long addr
);
1601 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1602 #endif /* __HAVE_ARCH_GATE_AREA */
1604 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1605 void __user
*, size_t *, loff_t
*);
1606 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1607 unsigned long lru_pages
);
1610 #define randomize_va_space 0
1612 extern int randomize_va_space
;
1615 const char * arch_vma_name(struct vm_area_struct
*vma
);
1616 void print_vma_addr(char *prefix
, unsigned long rip
);
1618 void sparse_mem_maps_populate_node(struct page
**map_map
,
1619 unsigned long pnum_begin
,
1620 unsigned long pnum_end
,
1621 unsigned long map_count
,
1624 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1625 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1626 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1627 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1628 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1629 void *vmemmap_alloc_block(unsigned long size
, int node
);
1630 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1631 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1632 int vmemmap_populate_basepages(struct page
*start_page
,
1633 unsigned long pages
, int node
);
1634 int vmemmap_populate(struct page
*start_page
, unsigned long pages
, int node
);
1635 void vmemmap_populate_print_last(void);
1639 MF_COUNT_INCREASED
= 1 << 0,
1641 extern void memory_failure(unsigned long pfn
, int trapno
);
1642 extern int __memory_failure(unsigned long pfn
, int trapno
, int flags
);
1643 extern int unpoison_memory(unsigned long pfn
);
1644 extern int sysctl_memory_failure_early_kill
;
1645 extern int sysctl_memory_failure_recovery
;
1646 extern void shake_page(struct page
*p
, int access
);
1647 extern atomic_long_t mce_bad_pages
;
1648 extern int soft_offline_page(struct page
*page
, int flags
);
1650 extern void dump_page(struct page
*page
);
1652 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1653 extern void clear_huge_page(struct page
*page
,
1655 unsigned int pages_per_huge_page
);
1656 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1657 unsigned long addr
, struct vm_area_struct
*vma
,
1658 unsigned int pages_per_huge_page
);
1659 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1661 #endif /* __KERNEL__ */
1662 #endif /* _LINUX_MM_H */