4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/bit_spinlock.h>
19 #include <linux/shrinker.h>
25 struct writeback_control
;
27 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
28 extern unsigned long max_mapnr
;
31 extern unsigned long num_physpages
;
32 extern unsigned long totalram_pages
;
33 extern void * high_memory
;
34 extern int page_cluster
;
37 extern int sysctl_legacy_va_layout
;
39 #define sysctl_legacy_va_layout 0
43 #include <asm/pgtable.h>
44 #include <asm/processor.h>
46 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
48 /* to align the pointer to the (next) page boundary */
49 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
52 * Linux kernel virtual memory manager primitives.
53 * The idea being to have a "virtual" mm in the same way
54 * we have a virtual fs - giving a cleaner interface to the
55 * mm details, and allowing different kinds of memory mappings
56 * (from shared memory to executable loading to arbitrary
60 extern struct kmem_cache
*vm_area_cachep
;
63 extern struct rb_root nommu_region_tree
;
64 extern struct rw_semaphore nommu_region_sem
;
66 extern unsigned int kobjsize(const void *objp
);
70 * vm_flags in vm_area_struct, see mm_types.h.
72 #define VM_NONE 0x00000000
74 #define VM_READ 0x00000001 /* currently active flags */
75 #define VM_WRITE 0x00000002
76 #define VM_EXEC 0x00000004
77 #define VM_SHARED 0x00000008
79 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
80 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
81 #define VM_MAYWRITE 0x00000020
82 #define VM_MAYEXEC 0x00000040
83 #define VM_MAYSHARE 0x00000080
85 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
86 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
87 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
89 #define VM_LOCKED 0x00002000
90 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
92 /* Used by sys_madvise() */
93 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
94 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
96 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
97 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
98 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
99 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
100 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
101 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
102 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
103 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
105 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
106 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
107 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
108 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
110 #if defined(CONFIG_X86)
111 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
112 #elif defined(CONFIG_PPC)
113 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
114 #elif defined(CONFIG_PARISC)
115 # define VM_GROWSUP VM_ARCH_1
116 #elif defined(CONFIG_IA64)
117 # define VM_GROWSUP VM_ARCH_1
118 #elif !defined(CONFIG_MMU)
119 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
123 # define VM_GROWSUP VM_NONE
126 /* Bits set in the VMA until the stack is in its final location */
127 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
129 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
130 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
133 #ifdef CONFIG_STACK_GROWSUP
134 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
136 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
139 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
140 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
141 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
142 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
143 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
146 * Special vmas that are non-mergable, non-mlock()able.
147 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
149 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
152 * mapping from the currently active vm_flags protection bits (the
153 * low four bits) to a page protection mask..
155 extern pgprot_t protection_map
[16];
157 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
158 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
159 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
160 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
161 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
162 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
165 * vm_fault is filled by the the pagefault handler and passed to the vma's
166 * ->fault function. The vma's ->fault is responsible for returning a bitmask
167 * of VM_FAULT_xxx flags that give details about how the fault was handled.
169 * pgoff should be used in favour of virtual_address, if possible. If pgoff
170 * is used, one may implement ->remap_pages to get nonlinear mapping support.
173 unsigned int flags
; /* FAULT_FLAG_xxx flags */
174 pgoff_t pgoff
; /* Logical page offset based on vma */
175 void __user
*virtual_address
; /* Faulting virtual address */
177 struct page
*page
; /* ->fault handlers should return a
178 * page here, unless VM_FAULT_NOPAGE
179 * is set (which is also implied by
185 * These are the virtual MM functions - opening of an area, closing and
186 * unmapping it (needed to keep files on disk up-to-date etc), pointer
187 * to the functions called when a no-page or a wp-page exception occurs.
189 struct vm_operations_struct
{
190 void (*open
)(struct vm_area_struct
* area
);
191 void (*close
)(struct vm_area_struct
* area
);
192 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
194 /* notification that a previously read-only page is about to become
195 * writable, if an error is returned it will cause a SIGBUS */
196 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
198 /* called by access_process_vm when get_user_pages() fails, typically
199 * for use by special VMAs that can switch between memory and hardware
201 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
202 void *buf
, int len
, int write
);
205 * set_policy() op must add a reference to any non-NULL @new mempolicy
206 * to hold the policy upon return. Caller should pass NULL @new to
207 * remove a policy and fall back to surrounding context--i.e. do not
208 * install a MPOL_DEFAULT policy, nor the task or system default
211 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
214 * get_policy() op must add reference [mpol_get()] to any policy at
215 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
216 * in mm/mempolicy.c will do this automatically.
217 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
218 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
219 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
220 * must return NULL--i.e., do not "fallback" to task or system default
223 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
225 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
226 const nodemask_t
*to
, unsigned long flags
);
228 /* called by sys_remap_file_pages() to populate non-linear mapping */
229 int (*remap_pages
)(struct vm_area_struct
*vma
, unsigned long addr
,
230 unsigned long size
, pgoff_t pgoff
);
236 #define page_private(page) ((page)->private)
237 #define set_page_private(page, v) ((page)->private = (v))
240 * FIXME: take this include out, include page-flags.h in
241 * files which need it (119 of them)
243 #include <linux/page-flags.h>
244 #include <linux/huge_mm.h>
247 * Methods to modify the page usage count.
249 * What counts for a page usage:
250 * - cache mapping (page->mapping)
251 * - private data (page->private)
252 * - page mapped in a task's page tables, each mapping
253 * is counted separately
255 * Also, many kernel routines increase the page count before a critical
256 * routine so they can be sure the page doesn't go away from under them.
260 * Drop a ref, return true if the refcount fell to zero (the page has no users)
262 static inline int put_page_testzero(struct page
*page
)
264 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
265 return atomic_dec_and_test(&page
->_count
);
269 * Try to grab a ref unless the page has a refcount of zero, return false if
272 static inline int get_page_unless_zero(struct page
*page
)
274 return atomic_inc_not_zero(&page
->_count
);
277 extern int page_is_ram(unsigned long pfn
);
279 /* Support for virtually mapped pages */
280 struct page
*vmalloc_to_page(const void *addr
);
281 unsigned long vmalloc_to_pfn(const void *addr
);
284 * Determine if an address is within the vmalloc range
286 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
287 * is no special casing required.
289 static inline int is_vmalloc_addr(const void *x
)
292 unsigned long addr
= (unsigned long)x
;
294 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
300 extern int is_vmalloc_or_module_addr(const void *x
);
302 static inline int is_vmalloc_or_module_addr(const void *x
)
308 static inline void compound_lock(struct page
*page
)
310 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
311 VM_BUG_ON(PageSlab(page
));
312 bit_spin_lock(PG_compound_lock
, &page
->flags
);
316 static inline void compound_unlock(struct page
*page
)
318 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
319 VM_BUG_ON(PageSlab(page
));
320 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
324 static inline unsigned long compound_lock_irqsave(struct page
*page
)
326 unsigned long uninitialized_var(flags
);
327 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
328 local_irq_save(flags
);
334 static inline void compound_unlock_irqrestore(struct page
*page
,
337 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
338 compound_unlock(page
);
339 local_irq_restore(flags
);
343 static inline struct page
*compound_head(struct page
*page
)
345 if (unlikely(PageTail(page
)))
346 return page
->first_page
;
351 * The atomic page->_mapcount, starts from -1: so that transitions
352 * both from it and to it can be tracked, using atomic_inc_and_test
353 * and atomic_add_negative(-1).
355 static inline void reset_page_mapcount(struct page
*page
)
357 atomic_set(&(page
)->_mapcount
, -1);
360 static inline int page_mapcount(struct page
*page
)
362 return atomic_read(&(page
)->_mapcount
) + 1;
365 static inline int page_count(struct page
*page
)
367 return atomic_read(&compound_head(page
)->_count
);
370 static inline void get_huge_page_tail(struct page
*page
)
373 * __split_huge_page_refcount() cannot run
376 VM_BUG_ON(page_mapcount(page
) < 0);
377 VM_BUG_ON(atomic_read(&page
->_count
) != 0);
378 atomic_inc(&page
->_mapcount
);
381 extern bool __get_page_tail(struct page
*page
);
383 static inline void get_page(struct page
*page
)
385 if (unlikely(PageTail(page
)))
386 if (likely(__get_page_tail(page
)))
389 * Getting a normal page or the head of a compound page
390 * requires to already have an elevated page->_count.
392 VM_BUG_ON(atomic_read(&page
->_count
) <= 0);
393 atomic_inc(&page
->_count
);
396 static inline struct page
*virt_to_head_page(const void *x
)
398 struct page
*page
= virt_to_page(x
);
399 return compound_head(page
);
403 * Setup the page count before being freed into the page allocator for
404 * the first time (boot or memory hotplug)
406 static inline void init_page_count(struct page
*page
)
408 atomic_set(&page
->_count
, 1);
412 * PageBuddy() indicate that the page is free and in the buddy system
413 * (see mm/page_alloc.c).
415 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
416 * -2 so that an underflow of the page_mapcount() won't be mistaken
417 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
418 * efficiently by most CPU architectures.
420 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
422 static inline int PageBuddy(struct page
*page
)
424 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
427 static inline void __SetPageBuddy(struct page
*page
)
429 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
430 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
433 static inline void __ClearPageBuddy(struct page
*page
)
435 VM_BUG_ON(!PageBuddy(page
));
436 atomic_set(&page
->_mapcount
, -1);
439 void put_page(struct page
*page
);
440 void put_pages_list(struct list_head
*pages
);
442 void split_page(struct page
*page
, unsigned int order
);
443 int split_free_page(struct page
*page
);
444 int capture_free_page(struct page
*page
, int alloc_order
, int migratetype
);
447 * Compound pages have a destructor function. Provide a
448 * prototype for that function and accessor functions.
449 * These are _only_ valid on the head of a PG_compound page.
451 typedef void compound_page_dtor(struct page
*);
453 static inline void set_compound_page_dtor(struct page
*page
,
454 compound_page_dtor
*dtor
)
456 page
[1].lru
.next
= (void *)dtor
;
459 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
461 return (compound_page_dtor
*)page
[1].lru
.next
;
464 static inline int compound_order(struct page
*page
)
468 return (unsigned long)page
[1].lru
.prev
;
471 static inline int compound_trans_order(struct page
*page
)
479 flags
= compound_lock_irqsave(page
);
480 order
= compound_order(page
);
481 compound_unlock_irqrestore(page
, flags
);
485 static inline void set_compound_order(struct page
*page
, unsigned long order
)
487 page
[1].lru
.prev
= (void *)order
;
492 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
493 * servicing faults for write access. In the normal case, do always want
494 * pte_mkwrite. But get_user_pages can cause write faults for mappings
495 * that do not have writing enabled, when used by access_process_vm.
497 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
499 if (likely(vma
->vm_flags
& VM_WRITE
))
500 pte
= pte_mkwrite(pte
);
506 * Multiple processes may "see" the same page. E.g. for untouched
507 * mappings of /dev/null, all processes see the same page full of
508 * zeroes, and text pages of executables and shared libraries have
509 * only one copy in memory, at most, normally.
511 * For the non-reserved pages, page_count(page) denotes a reference count.
512 * page_count() == 0 means the page is free. page->lru is then used for
513 * freelist management in the buddy allocator.
514 * page_count() > 0 means the page has been allocated.
516 * Pages are allocated by the slab allocator in order to provide memory
517 * to kmalloc and kmem_cache_alloc. In this case, the management of the
518 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
519 * unless a particular usage is carefully commented. (the responsibility of
520 * freeing the kmalloc memory is the caller's, of course).
522 * A page may be used by anyone else who does a __get_free_page().
523 * In this case, page_count still tracks the references, and should only
524 * be used through the normal accessor functions. The top bits of page->flags
525 * and page->virtual store page management information, but all other fields
526 * are unused and could be used privately, carefully. The management of this
527 * page is the responsibility of the one who allocated it, and those who have
528 * subsequently been given references to it.
530 * The other pages (we may call them "pagecache pages") are completely
531 * managed by the Linux memory manager: I/O, buffers, swapping etc.
532 * The following discussion applies only to them.
534 * A pagecache page contains an opaque `private' member, which belongs to the
535 * page's address_space. Usually, this is the address of a circular list of
536 * the page's disk buffers. PG_private must be set to tell the VM to call
537 * into the filesystem to release these pages.
539 * A page may belong to an inode's memory mapping. In this case, page->mapping
540 * is the pointer to the inode, and page->index is the file offset of the page,
541 * in units of PAGE_CACHE_SIZE.
543 * If pagecache pages are not associated with an inode, they are said to be
544 * anonymous pages. These may become associated with the swapcache, and in that
545 * case PG_swapcache is set, and page->private is an offset into the swapcache.
547 * In either case (swapcache or inode backed), the pagecache itself holds one
548 * reference to the page. Setting PG_private should also increment the
549 * refcount. The each user mapping also has a reference to the page.
551 * The pagecache pages are stored in a per-mapping radix tree, which is
552 * rooted at mapping->page_tree, and indexed by offset.
553 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
554 * lists, we instead now tag pages as dirty/writeback in the radix tree.
556 * All pagecache pages may be subject to I/O:
557 * - inode pages may need to be read from disk,
558 * - inode pages which have been modified and are MAP_SHARED may need
559 * to be written back to the inode on disk,
560 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
561 * modified may need to be swapped out to swap space and (later) to be read
566 * The zone field is never updated after free_area_init_core()
567 * sets it, so none of the operations on it need to be atomic.
572 * page->flags layout:
574 * There are three possibilities for how page->flags get
575 * laid out. The first is for the normal case, without
576 * sparsemem. The second is for sparsemem when there is
577 * plenty of space for node and section. The last is when
578 * we have run out of space and have to fall back to an
579 * alternate (slower) way of determining the node.
581 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
582 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
583 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
585 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
586 #define SECTIONS_WIDTH SECTIONS_SHIFT
588 #define SECTIONS_WIDTH 0
591 #define ZONES_WIDTH ZONES_SHIFT
593 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
594 #define NODES_WIDTH NODES_SHIFT
596 #ifdef CONFIG_SPARSEMEM_VMEMMAP
597 #error "Vmemmap: No space for nodes field in page flags"
599 #define NODES_WIDTH 0
602 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
603 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
604 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
605 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
608 * We are going to use the flags for the page to node mapping if its in
609 * there. This includes the case where there is no node, so it is implicit.
611 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
612 #define NODE_NOT_IN_PAGE_FLAGS
616 * Define the bit shifts to access each section. For non-existent
617 * sections we define the shift as 0; that plus a 0 mask ensures
618 * the compiler will optimise away reference to them.
620 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
621 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
622 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
624 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
625 #ifdef NODE_NOT_IN_PAGE_FLAGS
626 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
627 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
628 SECTIONS_PGOFF : ZONES_PGOFF)
630 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
631 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
632 NODES_PGOFF : ZONES_PGOFF)
635 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
637 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
638 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
641 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
642 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
643 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
644 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
646 static inline enum zone_type
page_zonenum(const struct page
*page
)
648 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
652 * The identification function is only used by the buddy allocator for
653 * determining if two pages could be buddies. We are not really
654 * identifying a zone since we could be using a the section number
655 * id if we have not node id available in page flags.
656 * We guarantee only that it will return the same value for two
657 * combinable pages in a zone.
659 static inline int page_zone_id(struct page
*page
)
661 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
664 static inline int zone_to_nid(struct zone
*zone
)
673 #ifdef NODE_NOT_IN_PAGE_FLAGS
674 extern int page_to_nid(const struct page
*page
);
676 static inline int page_to_nid(const struct page
*page
)
678 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
682 static inline struct zone
*page_zone(const struct page
*page
)
684 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
687 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
688 static inline void set_page_section(struct page
*page
, unsigned long section
)
690 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
691 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
694 static inline unsigned long page_to_section(const struct page
*page
)
696 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
700 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
702 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
703 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
706 static inline void set_page_node(struct page
*page
, unsigned long node
)
708 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
709 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
712 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
713 unsigned long node
, unsigned long pfn
)
715 set_page_zone(page
, zone
);
716 set_page_node(page
, node
);
717 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
718 set_page_section(page
, pfn_to_section_nr(pfn
));
723 * Some inline functions in vmstat.h depend on page_zone()
725 #include <linux/vmstat.h>
727 static __always_inline
void *lowmem_page_address(const struct page
*page
)
729 return __va(PFN_PHYS(page_to_pfn(page
)));
732 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
733 #define HASHED_PAGE_VIRTUAL
736 #if defined(WANT_PAGE_VIRTUAL)
737 #define page_address(page) ((page)->virtual)
738 #define set_page_address(page, address) \
740 (page)->virtual = (address); \
742 #define page_address_init() do { } while(0)
745 #if defined(HASHED_PAGE_VIRTUAL)
746 void *page_address(const struct page
*page
);
747 void set_page_address(struct page
*page
, void *virtual);
748 void page_address_init(void);
751 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
752 #define page_address(page) lowmem_page_address(page)
753 #define set_page_address(page, address) do { } while(0)
754 #define page_address_init() do { } while(0)
758 * On an anonymous page mapped into a user virtual memory area,
759 * page->mapping points to its anon_vma, not to a struct address_space;
760 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
762 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
763 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
764 * and then page->mapping points, not to an anon_vma, but to a private
765 * structure which KSM associates with that merged page. See ksm.h.
767 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
769 * Please note that, confusingly, "page_mapping" refers to the inode
770 * address_space which maps the page from disk; whereas "page_mapped"
771 * refers to user virtual address space into which the page is mapped.
773 #define PAGE_MAPPING_ANON 1
774 #define PAGE_MAPPING_KSM 2
775 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
777 extern struct address_space swapper_space
;
778 static inline struct address_space
*page_mapping(struct page
*page
)
780 struct address_space
*mapping
= page
->mapping
;
782 VM_BUG_ON(PageSlab(page
));
783 if (unlikely(PageSwapCache(page
)))
784 mapping
= &swapper_space
;
785 else if ((unsigned long)mapping
& PAGE_MAPPING_ANON
)
790 /* Neutral page->mapping pointer to address_space or anon_vma or other */
791 static inline void *page_rmapping(struct page
*page
)
793 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
796 extern struct address_space
*__page_file_mapping(struct page
*);
799 struct address_space
*page_file_mapping(struct page
*page
)
801 if (unlikely(PageSwapCache(page
)))
802 return __page_file_mapping(page
);
804 return page
->mapping
;
807 static inline int PageAnon(struct page
*page
)
809 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
813 * Return the pagecache index of the passed page. Regular pagecache pages
814 * use ->index whereas swapcache pages use ->private
816 static inline pgoff_t
page_index(struct page
*page
)
818 if (unlikely(PageSwapCache(page
)))
819 return page_private(page
);
823 extern pgoff_t
__page_file_index(struct page
*page
);
826 * Return the file index of the page. Regular pagecache pages use ->index
827 * whereas swapcache pages use swp_offset(->private)
829 static inline pgoff_t
page_file_index(struct page
*page
)
831 if (unlikely(PageSwapCache(page
)))
832 return __page_file_index(page
);
838 * Return true if this page is mapped into pagetables.
840 static inline int page_mapped(struct page
*page
)
842 return atomic_read(&(page
)->_mapcount
) >= 0;
846 * Different kinds of faults, as returned by handle_mm_fault().
847 * Used to decide whether a process gets delivered SIGBUS or
848 * just gets major/minor fault counters bumped up.
851 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
853 #define VM_FAULT_OOM 0x0001
854 #define VM_FAULT_SIGBUS 0x0002
855 #define VM_FAULT_MAJOR 0x0004
856 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
857 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
858 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
860 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
861 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
862 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
864 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
866 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
867 VM_FAULT_HWPOISON_LARGE)
869 /* Encode hstate index for a hwpoisoned large page */
870 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
871 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
874 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
876 extern void pagefault_out_of_memory(void);
878 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
881 * Flags passed to show_mem() and show_free_areas() to suppress output in
884 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
886 extern void show_free_areas(unsigned int flags
);
887 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
889 int shmem_zero_setup(struct vm_area_struct
*);
891 extern int can_do_mlock(void);
892 extern int user_shm_lock(size_t, struct user_struct
*);
893 extern void user_shm_unlock(size_t, struct user_struct
*);
896 * Parameter block passed down to zap_pte_range in exceptional cases.
899 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
900 struct address_space
*check_mapping
; /* Check page->mapping if set */
901 pgoff_t first_index
; /* Lowest page->index to unmap */
902 pgoff_t last_index
; /* Highest page->index to unmap */
905 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
908 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
910 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
911 unsigned long size
, struct zap_details
*);
912 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
913 unsigned long start
, unsigned long end
);
916 * mm_walk - callbacks for walk_page_range
917 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
918 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
919 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
920 * this handler is required to be able to handle
921 * pmd_trans_huge() pmds. They may simply choose to
922 * split_huge_page() instead of handling it explicitly.
923 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
924 * @pte_hole: if set, called for each hole at all levels
925 * @hugetlb_entry: if set, called for each hugetlb entry
926 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
929 * (see walk_page_range for more details)
932 int (*pgd_entry
)(pgd_t
*, unsigned long, unsigned long, struct mm_walk
*);
933 int (*pud_entry
)(pud_t
*, unsigned long, unsigned long, struct mm_walk
*);
934 int (*pmd_entry
)(pmd_t
*, unsigned long, unsigned long, struct mm_walk
*);
935 int (*pte_entry
)(pte_t
*, unsigned long, unsigned long, struct mm_walk
*);
936 int (*pte_hole
)(unsigned long, unsigned long, struct mm_walk
*);
937 int (*hugetlb_entry
)(pte_t
*, unsigned long,
938 unsigned long, unsigned long, struct mm_walk
*);
939 struct mm_struct
*mm
;
943 int walk_page_range(unsigned long addr
, unsigned long end
,
944 struct mm_walk
*walk
);
945 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
946 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
947 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
948 struct vm_area_struct
*vma
);
949 void unmap_mapping_range(struct address_space
*mapping
,
950 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
951 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
953 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
954 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
955 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
956 void *buf
, int len
, int write
);
958 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
959 loff_t
const holebegin
, loff_t
const holelen
)
961 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
964 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
965 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
966 extern int vmtruncate(struct inode
*inode
, loff_t offset
);
967 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
968 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
969 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
970 int invalidate_inode_page(struct page
*page
);
973 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
974 unsigned long address
, unsigned int flags
);
975 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
976 unsigned long address
, unsigned int fault_flags
);
978 static inline int handle_mm_fault(struct mm_struct
*mm
,
979 struct vm_area_struct
*vma
, unsigned long address
,
982 /* should never happen if there's no MMU */
984 return VM_FAULT_SIGBUS
;
986 static inline int fixup_user_fault(struct task_struct
*tsk
,
987 struct mm_struct
*mm
, unsigned long address
,
988 unsigned int fault_flags
)
990 /* should never happen if there's no MMU */
996 extern int make_pages_present(unsigned long addr
, unsigned long end
);
997 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
998 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
999 void *buf
, int len
, int write
);
1001 int __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1002 unsigned long start
, int len
, unsigned int foll_flags
,
1003 struct page
**pages
, struct vm_area_struct
**vmas
,
1005 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1006 unsigned long start
, int nr_pages
, int write
, int force
,
1007 struct page
**pages
, struct vm_area_struct
**vmas
);
1008 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1009 struct page
**pages
);
1011 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1012 struct page
**pages
);
1013 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1014 struct page
*get_dump_page(unsigned long addr
);
1016 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1017 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
1019 int __set_page_dirty_nobuffers(struct page
*page
);
1020 int __set_page_dirty_no_writeback(struct page
*page
);
1021 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1023 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1024 void account_page_writeback(struct page
*page
);
1025 int set_page_dirty(struct page
*page
);
1026 int set_page_dirty_lock(struct page
*page
);
1027 int clear_page_dirty_for_io(struct page
*page
);
1029 /* Is the vma a continuation of the stack vma above it? */
1030 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1032 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1035 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1038 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1039 (vma
->vm_start
== addr
) &&
1040 !vma_growsdown(vma
->vm_prev
, addr
);
1043 /* Is the vma a continuation of the stack vma below it? */
1044 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1046 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1049 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1052 return (vma
->vm_flags
& VM_GROWSUP
) &&
1053 (vma
->vm_end
== addr
) &&
1054 !vma_growsup(vma
->vm_next
, addr
);
1058 vm_is_stack(struct task_struct
*task
, struct vm_area_struct
*vma
, int in_group
);
1060 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1061 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1062 unsigned long new_addr
, unsigned long len
);
1063 extern unsigned long do_mremap(unsigned long addr
,
1064 unsigned long old_len
, unsigned long new_len
,
1065 unsigned long flags
, unsigned long new_addr
);
1066 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1067 struct vm_area_struct
**pprev
, unsigned long start
,
1068 unsigned long end
, unsigned long newflags
);
1071 * doesn't attempt to fault and will return short.
1073 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1074 struct page
**pages
);
1076 * per-process(per-mm_struct) statistics.
1078 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1080 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1082 #ifdef SPLIT_RSS_COUNTING
1084 * counter is updated in asynchronous manner and may go to minus.
1085 * But it's never be expected number for users.
1090 return (unsigned long)val
;
1093 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1095 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1098 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1100 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1103 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1105 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1108 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1110 return get_mm_counter(mm
, MM_FILEPAGES
) +
1111 get_mm_counter(mm
, MM_ANONPAGES
);
1114 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1116 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1119 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1121 return max(mm
->hiwater_vm
, mm
->total_vm
);
1124 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1126 unsigned long _rss
= get_mm_rss(mm
);
1128 if ((mm
)->hiwater_rss
< _rss
)
1129 (mm
)->hiwater_rss
= _rss
;
1132 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1134 if (mm
->hiwater_vm
< mm
->total_vm
)
1135 mm
->hiwater_vm
= mm
->total_vm
;
1138 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1139 struct mm_struct
*mm
)
1141 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1143 if (*maxrss
< hiwater_rss
)
1144 *maxrss
= hiwater_rss
;
1147 #if defined(SPLIT_RSS_COUNTING)
1148 void sync_mm_rss(struct mm_struct
*mm
);
1150 static inline void sync_mm_rss(struct mm_struct
*mm
)
1155 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1157 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1159 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1163 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1167 #ifdef __PAGETABLE_PUD_FOLDED
1168 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1169 unsigned long address
)
1174 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1177 #ifdef __PAGETABLE_PMD_FOLDED
1178 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1179 unsigned long address
)
1184 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1187 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1188 pmd_t
*pmd
, unsigned long address
);
1189 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1192 * The following ifdef needed to get the 4level-fixup.h header to work.
1193 * Remove it when 4level-fixup.h has been removed.
1195 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1196 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1198 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1199 NULL
: pud_offset(pgd
, address
);
1202 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1204 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1205 NULL
: pmd_offset(pud
, address
);
1207 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1209 #if USE_SPLIT_PTLOCKS
1211 * We tuck a spinlock to guard each pagetable page into its struct page,
1212 * at page->private, with BUILD_BUG_ON to make sure that this will not
1213 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1214 * When freeing, reset page->mapping so free_pages_check won't complain.
1216 #define __pte_lockptr(page) &((page)->ptl)
1217 #define pte_lock_init(_page) do { \
1218 spin_lock_init(__pte_lockptr(_page)); \
1220 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1221 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1222 #else /* !USE_SPLIT_PTLOCKS */
1224 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1226 #define pte_lock_init(page) do {} while (0)
1227 #define pte_lock_deinit(page) do {} while (0)
1228 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1229 #endif /* USE_SPLIT_PTLOCKS */
1231 static inline void pgtable_page_ctor(struct page
*page
)
1233 pte_lock_init(page
);
1234 inc_zone_page_state(page
, NR_PAGETABLE
);
1237 static inline void pgtable_page_dtor(struct page
*page
)
1239 pte_lock_deinit(page
);
1240 dec_zone_page_state(page
, NR_PAGETABLE
);
1243 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1245 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1246 pte_t *__pte = pte_offset_map(pmd, address); \
1252 #define pte_unmap_unlock(pte, ptl) do { \
1257 #define pte_alloc_map(mm, vma, pmd, address) \
1258 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1260 NULL: pte_offset_map(pmd, address))
1262 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1263 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1265 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1267 #define pte_alloc_kernel(pmd, address) \
1268 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1269 NULL: pte_offset_kernel(pmd, address))
1271 extern void free_area_init(unsigned long * zones_size
);
1272 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1273 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1274 extern void free_initmem(void);
1276 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1278 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1279 * zones, allocate the backing mem_map and account for memory holes in a more
1280 * architecture independent manner. This is a substitute for creating the
1281 * zone_sizes[] and zholes_size[] arrays and passing them to
1282 * free_area_init_node()
1284 * An architecture is expected to register range of page frames backed by
1285 * physical memory with memblock_add[_node]() before calling
1286 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1287 * usage, an architecture is expected to do something like
1289 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1291 * for_each_valid_physical_page_range()
1292 * memblock_add_node(base, size, nid)
1293 * free_area_init_nodes(max_zone_pfns);
1295 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1296 * registered physical page range. Similarly
1297 * sparse_memory_present_with_active_regions() calls memory_present() for
1298 * each range when SPARSEMEM is enabled.
1300 * See mm/page_alloc.c for more information on each function exposed by
1301 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1303 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1304 unsigned long node_map_pfn_alignment(void);
1305 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1306 unsigned long end_pfn
);
1307 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1308 unsigned long end_pfn
);
1309 extern void get_pfn_range_for_nid(unsigned int nid
,
1310 unsigned long *start_pfn
, unsigned long *end_pfn
);
1311 extern unsigned long find_min_pfn_with_active_regions(void);
1312 extern void free_bootmem_with_active_regions(int nid
,
1313 unsigned long max_low_pfn
);
1314 extern void sparse_memory_present_with_active_regions(int nid
);
1316 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1318 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1319 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1320 static inline int __early_pfn_to_nid(unsigned long pfn
)
1325 /* please see mm/page_alloc.c */
1326 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1327 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1328 /* there is a per-arch backend function. */
1329 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1330 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1333 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1334 extern void memmap_init_zone(unsigned long, int, unsigned long,
1335 unsigned long, enum memmap_context
);
1336 extern void setup_per_zone_wmarks(void);
1337 extern int __meminit
init_per_zone_wmark_min(void);
1338 extern void mem_init(void);
1339 extern void __init
mmap_init(void);
1340 extern void show_mem(unsigned int flags
);
1341 extern void si_meminfo(struct sysinfo
* val
);
1342 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1343 extern int after_bootmem
;
1345 extern __printf(3, 4)
1346 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1348 extern void setup_per_cpu_pageset(void);
1350 extern void zone_pcp_update(struct zone
*zone
);
1351 extern void zone_pcp_reset(struct zone
*zone
);
1354 extern atomic_long_t mmap_pages_allocated
;
1355 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1357 /* interval_tree.c */
1358 void vma_interval_tree_add(struct vm_area_struct
*vma
,
1359 struct vm_area_struct
*old
,
1360 struct address_space
*mapping
);
1361 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1362 struct rb_root
*root
);
1363 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1364 struct rb_root
*root
);
1365 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1366 unsigned long start
, unsigned long last
);
1367 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1368 unsigned long start
, unsigned long last
);
1370 #define vma_interval_tree_foreach(vma, root, start, last) \
1371 for (vma = vma_interval_tree_iter_first(root, start, last); \
1372 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1374 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1375 struct list_head
*list
)
1377 list_add_tail(&vma
->shared
.nonlinear
, list
);
1381 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1382 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1383 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1384 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1385 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1386 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1387 struct mempolicy
*);
1388 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1389 extern int split_vma(struct mm_struct
*,
1390 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1391 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1392 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1393 struct rb_node
**, struct rb_node
*);
1394 extern void unlink_file_vma(struct vm_area_struct
*);
1395 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1396 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1397 extern void exit_mmap(struct mm_struct
*);
1399 extern int mm_take_all_locks(struct mm_struct
*mm
);
1400 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1402 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1403 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1405 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1406 extern int install_special_mapping(struct mm_struct
*mm
,
1407 unsigned long addr
, unsigned long len
,
1408 unsigned long flags
, struct page
**pages
);
1410 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1412 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1413 unsigned long len
, unsigned long flags
,
1414 vm_flags_t vm_flags
, unsigned long pgoff
);
1415 extern unsigned long do_mmap_pgoff(struct file
*, unsigned long,
1416 unsigned long, unsigned long,
1417 unsigned long, unsigned long);
1418 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1420 /* These take the mm semaphore themselves */
1421 extern unsigned long vm_brk(unsigned long, unsigned long);
1422 extern int vm_munmap(unsigned long, size_t);
1423 extern unsigned long vm_mmap(struct file
*, unsigned long,
1424 unsigned long, unsigned long,
1425 unsigned long, unsigned long);
1428 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1429 extern void truncate_inode_pages_range(struct address_space
*,
1430 loff_t lstart
, loff_t lend
);
1432 /* generic vm_area_ops exported for stackable file systems */
1433 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1434 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1436 /* mm/page-writeback.c */
1437 int write_one_page(struct page
*page
, int wait
);
1438 void task_dirty_inc(struct task_struct
*tsk
);
1441 #define VM_MAX_READAHEAD 128 /* kbytes */
1442 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1444 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1445 pgoff_t offset
, unsigned long nr_to_read
);
1447 void page_cache_sync_readahead(struct address_space
*mapping
,
1448 struct file_ra_state
*ra
,
1451 unsigned long size
);
1453 void page_cache_async_readahead(struct address_space
*mapping
,
1454 struct file_ra_state
*ra
,
1458 unsigned long size
);
1460 unsigned long max_sane_readahead(unsigned long nr
);
1461 unsigned long ra_submit(struct file_ra_state
*ra
,
1462 struct address_space
*mapping
,
1465 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1466 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1468 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1469 extern int expand_downwards(struct vm_area_struct
*vma
,
1470 unsigned long address
);
1472 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1474 #define expand_upwards(vma, address) do { } while (0)
1477 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1478 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1479 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1480 struct vm_area_struct
**pprev
);
1482 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1483 NULL if none. Assume start_addr < end_addr. */
1484 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1486 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1488 if (vma
&& end_addr
<= vma
->vm_start
)
1493 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1495 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1498 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1499 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
1500 unsigned long vm_start
, unsigned long vm_end
)
1502 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
1504 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
1511 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1513 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1519 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1520 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1521 unsigned long pfn
, unsigned long size
, pgprot_t
);
1522 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1523 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1525 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1528 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1529 unsigned int foll_flags
);
1530 #define FOLL_WRITE 0x01 /* check pte is writable */
1531 #define FOLL_TOUCH 0x02 /* mark page accessed */
1532 #define FOLL_GET 0x04 /* do get_page on page */
1533 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1534 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1535 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1536 * and return without waiting upon it */
1537 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1538 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1539 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1541 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1543 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1544 unsigned long size
, pte_fn_t fn
, void *data
);
1546 #ifdef CONFIG_PROC_FS
1547 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1549 static inline void vm_stat_account(struct mm_struct
*mm
,
1550 unsigned long flags
, struct file
*file
, long pages
)
1552 mm
->total_vm
+= pages
;
1554 #endif /* CONFIG_PROC_FS */
1556 #ifdef CONFIG_DEBUG_PAGEALLOC
1557 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1558 #ifdef CONFIG_HIBERNATION
1559 extern bool kernel_page_present(struct page
*page
);
1560 #endif /* CONFIG_HIBERNATION */
1563 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1564 #ifdef CONFIG_HIBERNATION
1565 static inline bool kernel_page_present(struct page
*page
) { return true; }
1566 #endif /* CONFIG_HIBERNATION */
1569 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
1570 #ifdef __HAVE_ARCH_GATE_AREA
1571 int in_gate_area_no_mm(unsigned long addr
);
1572 int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
1574 int in_gate_area_no_mm(unsigned long addr
);
1575 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1576 #endif /* __HAVE_ARCH_GATE_AREA */
1578 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1579 void __user
*, size_t *, loff_t
*);
1580 unsigned long shrink_slab(struct shrink_control
*shrink
,
1581 unsigned long nr_pages_scanned
,
1582 unsigned long lru_pages
);
1585 #define randomize_va_space 0
1587 extern int randomize_va_space
;
1590 const char * arch_vma_name(struct vm_area_struct
*vma
);
1591 void print_vma_addr(char *prefix
, unsigned long rip
);
1593 void sparse_mem_maps_populate_node(struct page
**map_map
,
1594 unsigned long pnum_begin
,
1595 unsigned long pnum_end
,
1596 unsigned long map_count
,
1599 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1600 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1601 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1602 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1603 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1604 void *vmemmap_alloc_block(unsigned long size
, int node
);
1605 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1606 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1607 int vmemmap_populate_basepages(struct page
*start_page
,
1608 unsigned long pages
, int node
);
1609 int vmemmap_populate(struct page
*start_page
, unsigned long pages
, int node
);
1610 void vmemmap_populate_print_last(void);
1614 MF_COUNT_INCREASED
= 1 << 0,
1615 MF_ACTION_REQUIRED
= 1 << 1,
1616 MF_MUST_KILL
= 1 << 2,
1618 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
1619 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
1620 extern int unpoison_memory(unsigned long pfn
);
1621 extern int sysctl_memory_failure_early_kill
;
1622 extern int sysctl_memory_failure_recovery
;
1623 extern void shake_page(struct page
*p
, int access
);
1624 extern atomic_long_t mce_bad_pages
;
1625 extern int soft_offline_page(struct page
*page
, int flags
);
1627 extern void dump_page(struct page
*page
);
1629 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1630 extern void clear_huge_page(struct page
*page
,
1632 unsigned int pages_per_huge_page
);
1633 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1634 unsigned long addr
, struct vm_area_struct
*vma
,
1635 unsigned int pages_per_huge_page
);
1636 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1638 #ifdef CONFIG_DEBUG_PAGEALLOC
1639 extern unsigned int _debug_guardpage_minorder
;
1641 static inline unsigned int debug_guardpage_minorder(void)
1643 return _debug_guardpage_minorder
;
1646 static inline bool page_is_guard(struct page
*page
)
1648 return test_bit(PAGE_DEBUG_FLAG_GUARD
, &page
->debug_flags
);
1651 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1652 static inline bool page_is_guard(struct page
*page
) { return false; }
1653 #endif /* CONFIG_DEBUG_PAGEALLOC */
1655 #endif /* __KERNEL__ */
1656 #endif /* _LINUX_MM_H */