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>
23 struct anon_vma_chain
;
26 struct writeback_control
;
28 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr
;
32 extern unsigned long num_physpages
;
33 extern unsigned long totalram_pages
;
34 extern void * high_memory
;
35 extern int page_cluster
;
38 extern int sysctl_legacy_va_layout
;
40 #define sysctl_legacy_va_layout 0
44 #include <asm/pgtable.h>
45 #include <asm/processor.h>
47 extern unsigned long sysctl_user_reserve_kbytes
;
48 extern unsigned long sysctl_admin_reserve_kbytes
;
50 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
52 /* to align the pointer to the (next) page boundary */
53 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
56 * Linux kernel virtual memory manager primitives.
57 * The idea being to have a "virtual" mm in the same way
58 * we have a virtual fs - giving a cleaner interface to the
59 * mm details, and allowing different kinds of memory mappings
60 * (from shared memory to executable loading to arbitrary
64 extern struct kmem_cache
*vm_area_cachep
;
67 extern struct rb_root nommu_region_tree
;
68 extern struct rw_semaphore nommu_region_sem
;
70 extern unsigned int kobjsize(const void *objp
);
74 * vm_flags in vm_area_struct, see mm_types.h.
76 #define VM_NONE 0x00000000
78 #define VM_READ 0x00000001 /* currently active flags */
79 #define VM_WRITE 0x00000002
80 #define VM_EXEC 0x00000004
81 #define VM_SHARED 0x00000008
83 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
84 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
85 #define VM_MAYWRITE 0x00000020
86 #define VM_MAYEXEC 0x00000040
87 #define VM_MAYSHARE 0x00000080
89 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
90 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
91 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
93 #define VM_LOCKED 0x00002000
94 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
96 /* Used by sys_madvise() */
97 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
98 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
100 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
101 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
102 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
103 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
104 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
105 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
106 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
107 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
108 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
110 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
111 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
112 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
113 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
115 #if defined(CONFIG_X86)
116 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
117 #elif defined(CONFIG_PPC)
118 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
119 #elif defined(CONFIG_PARISC)
120 # define VM_GROWSUP VM_ARCH_1
121 #elif defined(CONFIG_METAG)
122 # define VM_GROWSUP VM_ARCH_1
123 #elif defined(CONFIG_IA64)
124 # define VM_GROWSUP VM_ARCH_1
125 #elif !defined(CONFIG_MMU)
126 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
130 # define VM_GROWSUP VM_NONE
133 /* Bits set in the VMA until the stack is in its final location */
134 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
136 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
137 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
140 #ifdef CONFIG_STACK_GROWSUP
141 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
143 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
146 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
147 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
148 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
149 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
150 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
153 * Special vmas that are non-mergable, non-mlock()able.
154 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
156 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
159 * mapping from the currently active vm_flags protection bits (the
160 * low four bits) to a page protection mask..
162 extern pgprot_t protection_map
[16];
164 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
165 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
166 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
167 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
168 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
169 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
170 #define FAULT_FLAG_TRIED 0x40 /* second try */
173 * vm_fault is filled by the the pagefault handler and passed to the vma's
174 * ->fault function. The vma's ->fault is responsible for returning a bitmask
175 * of VM_FAULT_xxx flags that give details about how the fault was handled.
177 * pgoff should be used in favour of virtual_address, if possible. If pgoff
178 * is used, one may implement ->remap_pages to get nonlinear mapping support.
181 unsigned int flags
; /* FAULT_FLAG_xxx flags */
182 pgoff_t pgoff
; /* Logical page offset based on vma */
183 void __user
*virtual_address
; /* Faulting virtual address */
185 struct page
*page
; /* ->fault handlers should return a
186 * page here, unless VM_FAULT_NOPAGE
187 * is set (which is also implied by
193 * These are the virtual MM functions - opening of an area, closing and
194 * unmapping it (needed to keep files on disk up-to-date etc), pointer
195 * to the functions called when a no-page or a wp-page exception occurs.
197 struct vm_operations_struct
{
198 void (*open
)(struct vm_area_struct
* area
);
199 void (*close
)(struct vm_area_struct
* area
);
200 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
202 /* notification that a previously read-only page is about to become
203 * writable, if an error is returned it will cause a SIGBUS */
204 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
206 /* called by access_process_vm when get_user_pages() fails, typically
207 * for use by special VMAs that can switch between memory and hardware
209 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
210 void *buf
, int len
, int write
);
213 * set_policy() op must add a reference to any non-NULL @new mempolicy
214 * to hold the policy upon return. Caller should pass NULL @new to
215 * remove a policy and fall back to surrounding context--i.e. do not
216 * install a MPOL_DEFAULT policy, nor the task or system default
219 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
222 * get_policy() op must add reference [mpol_get()] to any policy at
223 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
224 * in mm/mempolicy.c will do this automatically.
225 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
226 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
227 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
228 * must return NULL--i.e., do not "fallback" to task or system default
231 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
233 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
234 const nodemask_t
*to
, unsigned long flags
);
236 /* called by sys_remap_file_pages() to populate non-linear mapping */
237 int (*remap_pages
)(struct vm_area_struct
*vma
, unsigned long addr
,
238 unsigned long size
, pgoff_t pgoff
);
244 #define page_private(page) ((page)->private)
245 #define set_page_private(page, v) ((page)->private = (v))
247 /* It's valid only if the page is free path or free_list */
248 static inline void set_freepage_migratetype(struct page
*page
, int migratetype
)
250 page
->index
= migratetype
;
253 /* It's valid only if the page is free path or free_list */
254 static inline int get_freepage_migratetype(struct page
*page
)
260 * FIXME: take this include out, include page-flags.h in
261 * files which need it (119 of them)
263 #include <linux/page-flags.h>
264 #include <linux/huge_mm.h>
267 * Methods to modify the page usage count.
269 * What counts for a page usage:
270 * - cache mapping (page->mapping)
271 * - private data (page->private)
272 * - page mapped in a task's page tables, each mapping
273 * is counted separately
275 * Also, many kernel routines increase the page count before a critical
276 * routine so they can be sure the page doesn't go away from under them.
280 * Drop a ref, return true if the refcount fell to zero (the page has no users)
282 static inline int put_page_testzero(struct page
*page
)
284 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
285 return atomic_dec_and_test(&page
->_count
);
289 * Try to grab a ref unless the page has a refcount of zero, return false if
292 static inline int get_page_unless_zero(struct page
*page
)
294 return atomic_inc_not_zero(&page
->_count
);
297 extern int page_is_ram(unsigned long pfn
);
299 /* Support for virtually mapped pages */
300 struct page
*vmalloc_to_page(const void *addr
);
301 unsigned long vmalloc_to_pfn(const void *addr
);
304 * Determine if an address is within the vmalloc range
306 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
307 * is no special casing required.
309 static inline int is_vmalloc_addr(const void *x
)
312 unsigned long addr
= (unsigned long)x
;
314 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
320 extern int is_vmalloc_or_module_addr(const void *x
);
322 static inline int is_vmalloc_or_module_addr(const void *x
)
328 extern void kvfree(const void *addr
);
330 static inline void compound_lock(struct page
*page
)
332 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
333 VM_BUG_ON(PageSlab(page
));
334 bit_spin_lock(PG_compound_lock
, &page
->flags
);
338 static inline void compound_unlock(struct page
*page
)
340 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
341 VM_BUG_ON(PageSlab(page
));
342 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
346 static inline unsigned long compound_lock_irqsave(struct page
*page
)
348 unsigned long uninitialized_var(flags
);
349 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
350 local_irq_save(flags
);
356 static inline void compound_unlock_irqrestore(struct page
*page
,
359 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
360 compound_unlock(page
);
361 local_irq_restore(flags
);
365 static inline struct page
*compound_head(struct page
*page
)
367 if (unlikely(PageTail(page
))) {
368 struct page
*head
= page
->first_page
;
371 * page->first_page may be a dangling pointer to an old
372 * compound page, so recheck that it is still a tail
373 * page before returning.
376 if (likely(PageTail(page
)))
383 * The atomic page->_mapcount, starts from -1: so that transitions
384 * both from it and to it can be tracked, using atomic_inc_and_test
385 * and atomic_add_negative(-1).
387 static inline void page_mapcount_reset(struct page
*page
)
389 atomic_set(&(page
)->_mapcount
, -1);
392 static inline int page_mapcount(struct page
*page
)
394 return atomic_read(&(page
)->_mapcount
) + 1;
397 static inline int page_count(struct page
*page
)
399 return atomic_read(&compound_head(page
)->_count
);
402 static inline void get_huge_page_tail(struct page
*page
)
405 * __split_huge_page_refcount() cannot run
408 VM_BUG_ON(page_mapcount(page
) < 0);
409 VM_BUG_ON(atomic_read(&page
->_count
) != 0);
410 atomic_inc(&page
->_mapcount
);
413 extern bool __get_page_tail(struct page
*page
);
415 static inline void get_page(struct page
*page
)
417 if (unlikely(PageTail(page
)))
418 if (likely(__get_page_tail(page
)))
421 * Getting a normal page or the head of a compound page
422 * requires to already have an elevated page->_count.
424 VM_BUG_ON(atomic_read(&page
->_count
) <= 0);
425 atomic_inc(&page
->_count
);
428 static inline struct page
*virt_to_head_page(const void *x
)
430 struct page
*page
= virt_to_page(x
);
431 return compound_head(page
);
435 * Setup the page count before being freed into the page allocator for
436 * the first time (boot or memory hotplug)
438 static inline void init_page_count(struct page
*page
)
440 atomic_set(&page
->_count
, 1);
444 * PageBuddy() indicate that the page is free and in the buddy system
445 * (see mm/page_alloc.c).
447 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
448 * -2 so that an underflow of the page_mapcount() won't be mistaken
449 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
450 * efficiently by most CPU architectures.
452 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
454 static inline int PageBuddy(struct page
*page
)
456 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
459 static inline void __SetPageBuddy(struct page
*page
)
461 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
462 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
465 static inline void __ClearPageBuddy(struct page
*page
)
467 VM_BUG_ON(!PageBuddy(page
));
468 atomic_set(&page
->_mapcount
, -1);
471 void put_page(struct page
*page
);
472 void put_pages_list(struct list_head
*pages
);
474 void split_page(struct page
*page
, unsigned int order
);
475 int split_free_page(struct page
*page
);
478 * Compound pages have a destructor function. Provide a
479 * prototype for that function and accessor functions.
480 * These are _only_ valid on the head of a PG_compound page.
482 typedef void compound_page_dtor(struct page
*);
484 static inline void set_compound_page_dtor(struct page
*page
,
485 compound_page_dtor
*dtor
)
487 page
[1].lru
.next
= (void *)dtor
;
490 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
492 return (compound_page_dtor
*)page
[1].lru
.next
;
495 static inline int compound_order(struct page
*page
)
499 return (unsigned long)page
[1].lru
.prev
;
502 static inline int compound_trans_order(struct page
*page
)
510 flags
= compound_lock_irqsave(page
);
511 order
= compound_order(page
);
512 compound_unlock_irqrestore(page
, flags
);
516 static inline void set_compound_order(struct page
*page
, unsigned long order
)
518 page
[1].lru
.prev
= (void *)order
;
523 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
524 * servicing faults for write access. In the normal case, do always want
525 * pte_mkwrite. But get_user_pages can cause write faults for mappings
526 * that do not have writing enabled, when used by access_process_vm.
528 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
530 if (likely(vma
->vm_flags
& VM_WRITE
))
531 pte
= pte_mkwrite(pte
);
537 * Multiple processes may "see" the same page. E.g. for untouched
538 * mappings of /dev/null, all processes see the same page full of
539 * zeroes, and text pages of executables and shared libraries have
540 * only one copy in memory, at most, normally.
542 * For the non-reserved pages, page_count(page) denotes a reference count.
543 * page_count() == 0 means the page is free. page->lru is then used for
544 * freelist management in the buddy allocator.
545 * page_count() > 0 means the page has been allocated.
547 * Pages are allocated by the slab allocator in order to provide memory
548 * to kmalloc and kmem_cache_alloc. In this case, the management of the
549 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
550 * unless a particular usage is carefully commented. (the responsibility of
551 * freeing the kmalloc memory is the caller's, of course).
553 * A page may be used by anyone else who does a __get_free_page().
554 * In this case, page_count still tracks the references, and should only
555 * be used through the normal accessor functions. The top bits of page->flags
556 * and page->virtual store page management information, but all other fields
557 * are unused and could be used privately, carefully. The management of this
558 * page is the responsibility of the one who allocated it, and those who have
559 * subsequently been given references to it.
561 * The other pages (we may call them "pagecache pages") are completely
562 * managed by the Linux memory manager: I/O, buffers, swapping etc.
563 * The following discussion applies only to them.
565 * A pagecache page contains an opaque `private' member, which belongs to the
566 * page's address_space. Usually, this is the address of a circular list of
567 * the page's disk buffers. PG_private must be set to tell the VM to call
568 * into the filesystem to release these pages.
570 * A page may belong to an inode's memory mapping. In this case, page->mapping
571 * is the pointer to the inode, and page->index is the file offset of the page,
572 * in units of PAGE_CACHE_SIZE.
574 * If pagecache pages are not associated with an inode, they are said to be
575 * anonymous pages. These may become associated with the swapcache, and in that
576 * case PG_swapcache is set, and page->private is an offset into the swapcache.
578 * In either case (swapcache or inode backed), the pagecache itself holds one
579 * reference to the page. Setting PG_private should also increment the
580 * refcount. The each user mapping also has a reference to the page.
582 * The pagecache pages are stored in a per-mapping radix tree, which is
583 * rooted at mapping->page_tree, and indexed by offset.
584 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
585 * lists, we instead now tag pages as dirty/writeback in the radix tree.
587 * All pagecache pages may be subject to I/O:
588 * - inode pages may need to be read from disk,
589 * - inode pages which have been modified and are MAP_SHARED may need
590 * to be written back to the inode on disk,
591 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
592 * modified may need to be swapped out to swap space and (later) to be read
597 * The zone field is never updated after free_area_init_core()
598 * sets it, so none of the operations on it need to be atomic.
601 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
602 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
603 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
604 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
605 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
608 * Define the bit shifts to access each section. For non-existent
609 * sections we define the shift as 0; that plus a 0 mask ensures
610 * the compiler will optimise away reference to them.
612 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
613 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
614 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
615 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
617 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
618 #ifdef NODE_NOT_IN_PAGE_FLAGS
619 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
620 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
621 SECTIONS_PGOFF : ZONES_PGOFF)
623 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
624 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
625 NODES_PGOFF : ZONES_PGOFF)
628 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
630 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
631 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
634 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
635 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
636 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
637 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
638 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
640 static inline enum zone_type
page_zonenum(const struct page
*page
)
642 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
645 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
646 #define SECTION_IN_PAGE_FLAGS
650 * The identification function is only used by the buddy allocator for
651 * determining if two pages could be buddies. We are not really
652 * identifying a zone since we could be using a the section number
653 * id if we have not node id available in page flags.
654 * We guarantee only that it will return the same value for two
655 * combinable pages in a zone.
657 static inline int page_zone_id(struct page
*page
)
659 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
662 static inline int zone_to_nid(struct zone
*zone
)
671 #ifdef NODE_NOT_IN_PAGE_FLAGS
672 extern int page_to_nid(const struct page
*page
);
674 static inline int page_to_nid(const struct page
*page
)
676 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
680 #ifdef CONFIG_NUMA_BALANCING
681 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
682 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
684 return xchg(&page
->_last_nid
, nid
);
687 static inline int page_nid_last(struct page
*page
)
689 return page
->_last_nid
;
691 static inline void page_nid_reset_last(struct page
*page
)
693 page
->_last_nid
= -1;
696 static inline int page_nid_last(struct page
*page
)
698 return (page
->flags
>> LAST_NID_PGSHIFT
) & LAST_NID_MASK
;
701 extern int page_nid_xchg_last(struct page
*page
, int nid
);
703 static inline void page_nid_reset_last(struct page
*page
)
705 int nid
= (1 << LAST_NID_SHIFT
) - 1;
707 page
->flags
&= ~(LAST_NID_MASK
<< LAST_NID_PGSHIFT
);
708 page
->flags
|= (nid
& LAST_NID_MASK
) << LAST_NID_PGSHIFT
;
710 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
712 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
714 return page_to_nid(page
);
717 static inline int page_nid_last(struct page
*page
)
719 return page_to_nid(page
);
722 static inline void page_nid_reset_last(struct page
*page
)
727 static inline struct zone
*page_zone(const struct page
*page
)
729 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
732 #ifdef SECTION_IN_PAGE_FLAGS
733 static inline void set_page_section(struct page
*page
, unsigned long section
)
735 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
736 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
739 static inline unsigned long page_to_section(const struct page
*page
)
741 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
745 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
747 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
748 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
751 static inline void set_page_node(struct page
*page
, unsigned long node
)
753 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
754 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
757 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
758 unsigned long node
, unsigned long pfn
)
760 set_page_zone(page
, zone
);
761 set_page_node(page
, node
);
762 #ifdef SECTION_IN_PAGE_FLAGS
763 set_page_section(page
, pfn_to_section_nr(pfn
));
768 * Some inline functions in vmstat.h depend on page_zone()
770 #include <linux/vmstat.h>
772 static __always_inline
void *lowmem_page_address(const struct page
*page
)
774 return __va(PFN_PHYS(page_to_pfn(page
)));
777 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
778 #define HASHED_PAGE_VIRTUAL
781 #if defined(WANT_PAGE_VIRTUAL)
782 static inline void *page_address(const struct page
*page
)
784 return page
->virtual;
786 static inline void set_page_address(struct page
*page
, void *address
)
788 page
->virtual = address
;
790 #define page_address_init() do { } while(0)
793 #if defined(HASHED_PAGE_VIRTUAL)
794 void *page_address(const struct page
*page
);
795 void set_page_address(struct page
*page
, void *virtual);
796 void page_address_init(void);
799 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
800 #define page_address(page) lowmem_page_address(page)
801 #define set_page_address(page, address) do { } while(0)
802 #define page_address_init() do { } while(0)
806 * On an anonymous page mapped into a user virtual memory area,
807 * page->mapping points to its anon_vma, not to a struct address_space;
808 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
810 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
811 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
812 * and then page->mapping points, not to an anon_vma, but to a private
813 * structure which KSM associates with that merged page. See ksm.h.
815 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
817 * Please note that, confusingly, "page_mapping" refers to the inode
818 * address_space which maps the page from disk; whereas "page_mapped"
819 * refers to user virtual address space into which the page is mapped.
821 #define PAGE_MAPPING_ANON 1
822 #define PAGE_MAPPING_KSM 2
823 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
825 extern struct address_space
*page_mapping(struct page
*page
);
827 /* Neutral page->mapping pointer to address_space or anon_vma or other */
828 static inline void *page_rmapping(struct page
*page
)
830 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
833 extern struct address_space
*__page_file_mapping(struct page
*);
836 struct address_space
*page_file_mapping(struct page
*page
)
838 if (unlikely(PageSwapCache(page
)))
839 return __page_file_mapping(page
);
841 return page
->mapping
;
844 static inline int PageAnon(struct page
*page
)
846 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
850 * Return the pagecache index of the passed page. Regular pagecache pages
851 * use ->index whereas swapcache pages use ->private
853 static inline pgoff_t
page_index(struct page
*page
)
855 if (unlikely(PageSwapCache(page
)))
856 return page_private(page
);
860 extern pgoff_t
__page_file_index(struct page
*page
);
863 * Return the file index of the page. Regular pagecache pages use ->index
864 * whereas swapcache pages use swp_offset(->private)
866 static inline pgoff_t
page_file_index(struct page
*page
)
868 if (unlikely(PageSwapCache(page
)))
869 return __page_file_index(page
);
875 * Return true if this page is mapped into pagetables.
877 static inline int page_mapped(struct page
*page
)
879 return atomic_read(&(page
)->_mapcount
) >= 0;
883 * Different kinds of faults, as returned by handle_mm_fault().
884 * Used to decide whether a process gets delivered SIGBUS or
885 * just gets major/minor fault counters bumped up.
888 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
890 #define VM_FAULT_OOM 0x0001
891 #define VM_FAULT_SIGBUS 0x0002
892 #define VM_FAULT_MAJOR 0x0004
893 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
894 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
895 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
897 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
898 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
899 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
901 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
903 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
904 VM_FAULT_HWPOISON_LARGE)
906 /* Encode hstate index for a hwpoisoned large page */
907 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
908 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
911 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
913 extern void pagefault_out_of_memory(void);
915 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
918 * Flags passed to show_mem() and show_free_areas() to suppress output in
921 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
922 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
924 extern void show_free_areas(unsigned int flags
);
925 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
927 void shmem_set_file(struct vm_area_struct
*vma
, struct file
*file
);
928 int shmem_zero_setup(struct vm_area_struct
*);
930 extern int can_do_mlock(void);
931 extern int user_shm_lock(size_t, struct user_struct
*);
932 extern void user_shm_unlock(size_t, struct user_struct
*);
935 * Parameter block passed down to zap_pte_range in exceptional cases.
938 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
939 struct address_space
*check_mapping
; /* Check page->mapping if set */
940 pgoff_t first_index
; /* Lowest page->index to unmap */
941 pgoff_t last_index
; /* Highest page->index to unmap */
944 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
947 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
949 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
950 unsigned long size
, struct zap_details
*);
951 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
952 unsigned long start
, unsigned long end
);
955 * mm_walk - callbacks for walk_page_range
956 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
957 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
958 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
959 * this handler is required to be able to handle
960 * pmd_trans_huge() pmds. They may simply choose to
961 * split_huge_page() instead of handling it explicitly.
962 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
963 * @pte_hole: if set, called for each hole at all levels
964 * @hugetlb_entry: if set, called for each hugetlb entry
965 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
968 * (see walk_page_range for more details)
971 int (*pgd_entry
)(pgd_t
*pgd
, unsigned long addr
,
972 unsigned long next
, struct mm_walk
*walk
);
973 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
974 unsigned long next
, struct mm_walk
*walk
);
975 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
976 unsigned long next
, struct mm_walk
*walk
);
977 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
978 unsigned long next
, struct mm_walk
*walk
);
979 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
980 struct mm_walk
*walk
);
981 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
982 unsigned long addr
, unsigned long next
,
983 struct mm_walk
*walk
);
984 struct mm_struct
*mm
;
988 int walk_page_range(unsigned long addr
, unsigned long end
,
989 struct mm_walk
*walk
);
990 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
991 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
992 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
993 struct vm_area_struct
*vma
);
994 void unmap_mapping_range(struct address_space
*mapping
,
995 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
996 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
998 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
999 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1000 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1001 void *buf
, int len
, int write
);
1003 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1004 loff_t
const holebegin
, loff_t
const holelen
)
1006 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1009 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
1010 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1011 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1012 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1013 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1014 int invalidate_inode_page(struct page
*page
);
1017 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1018 unsigned long address
, unsigned int flags
);
1019 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1020 unsigned long address
, unsigned int fault_flags
);
1022 static inline int handle_mm_fault(struct mm_struct
*mm
,
1023 struct vm_area_struct
*vma
, unsigned long address
,
1026 /* should never happen if there's no MMU */
1028 return VM_FAULT_SIGBUS
;
1030 static inline int fixup_user_fault(struct task_struct
*tsk
,
1031 struct mm_struct
*mm
, unsigned long address
,
1032 unsigned int fault_flags
)
1034 /* should never happen if there's no MMU */
1040 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1041 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1042 void *buf
, int len
, int write
);
1044 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1045 unsigned long start
, unsigned long nr_pages
,
1046 unsigned int foll_flags
, struct page
**pages
,
1047 struct vm_area_struct
**vmas
, int *nonblocking
);
1048 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1049 unsigned long start
, unsigned long nr_pages
,
1050 int write
, int force
, struct page
**pages
,
1051 struct vm_area_struct
**vmas
);
1052 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1053 struct page
**pages
);
1055 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1056 struct page
**pages
);
1057 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1058 struct page
*get_dump_page(unsigned long addr
);
1060 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1061 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
1063 int __set_page_dirty_nobuffers(struct page
*page
);
1064 int __set_page_dirty_no_writeback(struct page
*page
);
1065 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1067 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1068 void account_page_writeback(struct page
*page
);
1069 int set_page_dirty(struct page
*page
);
1070 int set_page_dirty_lock(struct page
*page
);
1071 int clear_page_dirty_for_io(struct page
*page
);
1073 /* Is the vma a continuation of the stack vma above it? */
1074 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1076 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1079 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1082 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1083 (vma
->vm_start
== addr
) &&
1084 !vma_growsdown(vma
->vm_prev
, addr
);
1087 /* Is the vma a continuation of the stack vma below it? */
1088 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1090 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1093 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1096 return (vma
->vm_flags
& VM_GROWSUP
) &&
1097 (vma
->vm_end
== addr
) &&
1098 !vma_growsup(vma
->vm_next
, addr
);
1102 vm_is_stack(struct task_struct
*task
, struct vm_area_struct
*vma
, int in_group
);
1104 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1105 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1106 unsigned long new_addr
, unsigned long len
,
1107 bool need_rmap_locks
);
1108 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1109 unsigned long end
, pgprot_t newprot
,
1110 int dirty_accountable
, int prot_numa
);
1111 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1112 struct vm_area_struct
**pprev
, unsigned long start
,
1113 unsigned long end
, unsigned long newflags
);
1116 * doesn't attempt to fault and will return short.
1118 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1119 struct page
**pages
);
1121 * per-process(per-mm_struct) statistics.
1123 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1125 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1127 #ifdef SPLIT_RSS_COUNTING
1129 * counter is updated in asynchronous manner and may go to minus.
1130 * But it's never be expected number for users.
1135 return (unsigned long)val
;
1138 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1140 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1143 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1145 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1148 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1150 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1153 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1155 return get_mm_counter(mm
, MM_FILEPAGES
) +
1156 get_mm_counter(mm
, MM_ANONPAGES
);
1159 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1161 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1164 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1166 return max(mm
->hiwater_vm
, mm
->total_vm
);
1169 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1171 unsigned long _rss
= get_mm_rss(mm
);
1173 if ((mm
)->hiwater_rss
< _rss
)
1174 (mm
)->hiwater_rss
= _rss
;
1177 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1179 if (mm
->hiwater_vm
< mm
->total_vm
)
1180 mm
->hiwater_vm
= mm
->total_vm
;
1183 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1184 struct mm_struct
*mm
)
1186 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1188 if (*maxrss
< hiwater_rss
)
1189 *maxrss
= hiwater_rss
;
1192 #if defined(SPLIT_RSS_COUNTING)
1193 void sync_mm_rss(struct mm_struct
*mm
);
1195 static inline void sync_mm_rss(struct mm_struct
*mm
)
1200 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1202 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1204 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1208 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1212 #ifdef __PAGETABLE_PUD_FOLDED
1213 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1214 unsigned long address
)
1219 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1222 #ifdef __PAGETABLE_PMD_FOLDED
1223 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1224 unsigned long address
)
1229 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1232 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1233 pmd_t
*pmd
, unsigned long address
);
1234 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1237 * The following ifdef needed to get the 4level-fixup.h header to work.
1238 * Remove it when 4level-fixup.h has been removed.
1240 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1241 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1243 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1244 NULL
: pud_offset(pgd
, address
);
1247 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1249 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1250 NULL
: pmd_offset(pud
, address
);
1252 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1254 #if USE_SPLIT_PTLOCKS
1256 * We tuck a spinlock to guard each pagetable page into its struct page,
1257 * at page->private, with BUILD_BUG_ON to make sure that this will not
1258 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1259 * When freeing, reset page->mapping so free_pages_check won't complain.
1261 #define __pte_lockptr(page) &((page)->ptl)
1262 #define pte_lock_init(_page) do { \
1263 spin_lock_init(__pte_lockptr(_page)); \
1265 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1266 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1267 #else /* !USE_SPLIT_PTLOCKS */
1269 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1271 #define pte_lock_init(page) do {} while (0)
1272 #define pte_lock_deinit(page) do {} while (0)
1273 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1274 #endif /* USE_SPLIT_PTLOCKS */
1276 static inline void pgtable_page_ctor(struct page
*page
)
1278 pte_lock_init(page
);
1279 inc_zone_page_state(page
, NR_PAGETABLE
);
1282 static inline void pgtable_page_dtor(struct page
*page
)
1284 pte_lock_deinit(page
);
1285 dec_zone_page_state(page
, NR_PAGETABLE
);
1288 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1290 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1291 pte_t *__pte = pte_offset_map(pmd, address); \
1297 #define pte_unmap_unlock(pte, ptl) do { \
1302 #define pte_alloc_map(mm, vma, pmd, address) \
1303 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1305 NULL: pte_offset_map(pmd, address))
1307 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1308 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1310 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1312 #define pte_alloc_kernel(pmd, address) \
1313 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1314 NULL: pte_offset_kernel(pmd, address))
1316 extern void free_area_init(unsigned long * zones_size
);
1317 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1318 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1319 extern void free_initmem(void);
1322 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1323 * into the buddy system. The freed pages will be poisoned with pattern
1324 * "poison" if it's non-zero.
1325 * Return pages freed into the buddy system.
1327 extern unsigned long free_reserved_area(unsigned long start
, unsigned long end
,
1328 int poison
, char *s
);
1329 #ifdef CONFIG_HIGHMEM
1331 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1332 * and totalram_pages.
1334 extern void free_highmem_page(struct page
*page
);
1337 static inline void adjust_managed_page_count(struct page
*page
, long count
)
1339 totalram_pages
+= count
;
1342 /* Free the reserved page into the buddy system, so it gets managed. */
1343 static inline void __free_reserved_page(struct page
*page
)
1345 ClearPageReserved(page
);
1346 init_page_count(page
);
1350 static inline void free_reserved_page(struct page
*page
)
1352 __free_reserved_page(page
);
1353 adjust_managed_page_count(page
, 1);
1356 static inline void mark_page_reserved(struct page
*page
)
1358 SetPageReserved(page
);
1359 adjust_managed_page_count(page
, -1);
1363 * Default method to free all the __init memory into the buddy system.
1364 * The freed pages will be poisoned with pattern "poison" if it is
1365 * non-zero. Return pages freed into the buddy system.
1367 static inline unsigned long free_initmem_default(int poison
)
1369 extern char __init_begin
[], __init_end
[];
1371 return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin
) ,
1372 ((unsigned long)&__init_end
) & PAGE_MASK
,
1373 poison
, "unused kernel");
1376 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1378 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1379 * zones, allocate the backing mem_map and account for memory holes in a more
1380 * architecture independent manner. This is a substitute for creating the
1381 * zone_sizes[] and zholes_size[] arrays and passing them to
1382 * free_area_init_node()
1384 * An architecture is expected to register range of page frames backed by
1385 * physical memory with memblock_add[_node]() before calling
1386 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1387 * usage, an architecture is expected to do something like
1389 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1391 * for_each_valid_physical_page_range()
1392 * memblock_add_node(base, size, nid)
1393 * free_area_init_nodes(max_zone_pfns);
1395 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1396 * registered physical page range. Similarly
1397 * sparse_memory_present_with_active_regions() calls memory_present() for
1398 * each range when SPARSEMEM is enabled.
1400 * See mm/page_alloc.c for more information on each function exposed by
1401 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1403 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1404 unsigned long node_map_pfn_alignment(void);
1405 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1406 unsigned long end_pfn
);
1407 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1408 unsigned long end_pfn
);
1409 extern void get_pfn_range_for_nid(unsigned int nid
,
1410 unsigned long *start_pfn
, unsigned long *end_pfn
);
1411 extern unsigned long find_min_pfn_with_active_regions(void);
1412 extern void free_bootmem_with_active_regions(int nid
,
1413 unsigned long max_low_pfn
);
1414 extern void sparse_memory_present_with_active_regions(int nid
);
1416 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1418 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1419 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1420 static inline int __early_pfn_to_nid(unsigned long pfn
)
1425 /* please see mm/page_alloc.c */
1426 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1427 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1428 /* there is a per-arch backend function. */
1429 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1430 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1433 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1434 extern void memmap_init_zone(unsigned long, int, unsigned long,
1435 unsigned long, enum memmap_context
);
1436 extern void setup_per_zone_wmarks(void);
1437 extern int __meminit
init_per_zone_wmark_min(void);
1438 extern void mem_init(void);
1439 extern void __init
mmap_init(void);
1440 extern void show_mem(unsigned int flags
);
1441 extern void si_meminfo(struct sysinfo
* val
);
1442 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1444 extern __printf(3, 4)
1445 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1447 extern void setup_per_cpu_pageset(void);
1449 extern void zone_pcp_update(struct zone
*zone
);
1450 extern void zone_pcp_reset(struct zone
*zone
);
1453 extern int min_free_kbytes
;
1456 extern atomic_long_t mmap_pages_allocated
;
1457 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1459 /* interval_tree.c */
1460 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1461 struct rb_root
*root
);
1462 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1463 struct vm_area_struct
*prev
,
1464 struct rb_root
*root
);
1465 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1466 struct rb_root
*root
);
1467 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1468 unsigned long start
, unsigned long last
);
1469 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1470 unsigned long start
, unsigned long last
);
1472 #define vma_interval_tree_foreach(vma, root, start, last) \
1473 for (vma = vma_interval_tree_iter_first(root, start, last); \
1474 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1476 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1477 struct list_head
*list
)
1479 list_add_tail(&vma
->shared
.nonlinear
, list
);
1482 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1483 struct rb_root
*root
);
1484 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1485 struct rb_root
*root
);
1486 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1487 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1488 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1489 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1490 #ifdef CONFIG_DEBUG_VM_RB
1491 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1494 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1495 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1496 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1499 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1500 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1501 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1502 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1503 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1504 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1505 struct mempolicy
*, const char __user
*);
1506 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1507 extern int split_vma(struct mm_struct
*,
1508 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1509 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1510 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1511 struct rb_node
**, struct rb_node
*);
1512 extern void unlink_file_vma(struct vm_area_struct
*);
1513 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1514 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1515 bool *need_rmap_locks
);
1516 extern void exit_mmap(struct mm_struct
*);
1518 extern int mm_take_all_locks(struct mm_struct
*mm
);
1519 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1521 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1522 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1524 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1525 extern int install_special_mapping(struct mm_struct
*mm
,
1526 unsigned long addr
, unsigned long len
,
1527 unsigned long flags
, struct page
**pages
);
1529 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1531 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1532 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1533 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1534 unsigned long len
, unsigned long prot
, unsigned long flags
,
1535 unsigned long pgoff
, unsigned long *populate
);
1536 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1539 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1541 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1544 (void) __mm_populate(addr
, len
, 1);
1547 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1550 /* These take the mm semaphore themselves */
1551 extern unsigned long vm_brk(unsigned long, unsigned long);
1552 extern int vm_munmap(unsigned long, size_t);
1553 extern unsigned long vm_mmap(struct file
*, unsigned long,
1554 unsigned long, unsigned long,
1555 unsigned long, unsigned long);
1557 struct vm_unmapped_area_info
{
1558 #define VM_UNMAPPED_AREA_TOPDOWN 1
1559 unsigned long flags
;
1560 unsigned long length
;
1561 unsigned long low_limit
;
1562 unsigned long high_limit
;
1563 unsigned long align_mask
;
1564 unsigned long align_offset
;
1567 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1568 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1571 * Search for an unmapped address range.
1573 * We are looking for a range that:
1574 * - does not intersect with any VMA;
1575 * - is contained within the [low_limit, high_limit) interval;
1576 * - is at least the desired size.
1577 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1579 static inline unsigned long
1580 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1582 if (!(info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
))
1583 return unmapped_area(info
);
1585 return unmapped_area_topdown(info
);
1589 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1590 extern void truncate_inode_pages_range(struct address_space
*,
1591 loff_t lstart
, loff_t lend
);
1593 /* generic vm_area_ops exported for stackable file systems */
1594 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1595 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1597 /* mm/page-writeback.c */
1598 int write_one_page(struct page
*page
, int wait
);
1599 void task_dirty_inc(struct task_struct
*tsk
);
1602 #define VM_MAX_READAHEAD 128 /* kbytes */
1603 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1605 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1606 pgoff_t offset
, unsigned long nr_to_read
);
1608 void page_cache_sync_readahead(struct address_space
*mapping
,
1609 struct file_ra_state
*ra
,
1612 unsigned long size
);
1614 void page_cache_async_readahead(struct address_space
*mapping
,
1615 struct file_ra_state
*ra
,
1619 unsigned long size
);
1621 unsigned long max_sane_readahead(unsigned long nr
);
1622 unsigned long ra_submit(struct file_ra_state
*ra
,
1623 struct address_space
*mapping
,
1626 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1627 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1629 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1630 extern int expand_downwards(struct vm_area_struct
*vma
,
1631 unsigned long address
);
1633 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1635 #define expand_upwards(vma, address) do { } while (0)
1638 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1639 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1640 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1641 struct vm_area_struct
**pprev
);
1643 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1644 NULL if none. Assume start_addr < end_addr. */
1645 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1647 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1649 if (vma
&& end_addr
<= vma
->vm_start
)
1654 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1656 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1659 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1660 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
1661 unsigned long vm_start
, unsigned long vm_end
)
1663 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
1665 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
1672 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1674 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1680 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1681 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
1682 unsigned long start
, unsigned long end
);
1685 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1686 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1687 unsigned long pfn
, unsigned long size
, pgprot_t
);
1688 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1689 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1691 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1693 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
1696 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
1697 unsigned long address
, unsigned int foll_flags
,
1698 unsigned int *page_mask
);
1700 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
1701 unsigned long address
, unsigned int foll_flags
)
1703 unsigned int unused_page_mask
;
1704 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
1707 #define FOLL_WRITE 0x01 /* check pte is writable */
1708 #define FOLL_TOUCH 0x02 /* mark page accessed */
1709 #define FOLL_GET 0x04 /* do get_page on page */
1710 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1711 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1712 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1713 * and return without waiting upon it */
1714 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1715 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1716 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1717 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1718 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1719 #define FOLL_COW 0x4000 /* internal GUP flag */
1721 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1723 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1724 unsigned long size
, pte_fn_t fn
, void *data
);
1726 #ifdef CONFIG_PROC_FS
1727 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1729 static inline void vm_stat_account(struct mm_struct
*mm
,
1730 unsigned long flags
, struct file
*file
, long pages
)
1732 mm
->total_vm
+= pages
;
1734 #endif /* CONFIG_PROC_FS */
1736 #ifdef CONFIG_DEBUG_PAGEALLOC
1737 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1738 #ifdef CONFIG_HIBERNATION
1739 extern bool kernel_page_present(struct page
*page
);
1740 #endif /* CONFIG_HIBERNATION */
1743 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1744 #ifdef CONFIG_HIBERNATION
1745 static inline bool kernel_page_present(struct page
*page
) { return true; }
1746 #endif /* CONFIG_HIBERNATION */
1749 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
1750 #ifdef __HAVE_ARCH_GATE_AREA
1751 int in_gate_area_no_mm(unsigned long addr
);
1752 int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
1754 int in_gate_area_no_mm(unsigned long addr
);
1755 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1756 #endif /* __HAVE_ARCH_GATE_AREA */
1758 #ifdef CONFIG_SYSCTL
1759 extern int sysctl_drop_caches
;
1760 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1761 void __user
*, size_t *, loff_t
*);
1764 unsigned long shrink_slab(struct shrink_control
*shrink
,
1765 unsigned long nr_pages_scanned
,
1766 unsigned long lru_pages
);
1767 void drop_pagecache(void);
1770 #define randomize_va_space 0
1772 extern int randomize_va_space
;
1775 const char * arch_vma_name(struct vm_area_struct
*vma
);
1776 void print_vma_addr(char *prefix
, unsigned long rip
);
1778 void sparse_mem_maps_populate_node(struct page
**map_map
,
1779 unsigned long pnum_begin
,
1780 unsigned long pnum_end
,
1781 unsigned long map_count
,
1784 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1785 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1786 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1787 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1788 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1789 void *vmemmap_alloc_block(unsigned long size
, int node
);
1790 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1791 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1792 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
1794 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
1795 void vmemmap_populate_print_last(void);
1796 #ifdef CONFIG_MEMORY_HOTPLUG
1797 void vmemmap_free(unsigned long start
, unsigned long end
);
1799 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
1800 unsigned long size
);
1803 MF_COUNT_INCREASED
= 1 << 0,
1804 MF_ACTION_REQUIRED
= 1 << 1,
1805 MF_MUST_KILL
= 1 << 2,
1807 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
1808 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
1809 extern int unpoison_memory(unsigned long pfn
);
1810 extern int sysctl_memory_failure_early_kill
;
1811 extern int sysctl_memory_failure_recovery
;
1812 extern void shake_page(struct page
*p
, int access
);
1813 extern atomic_long_t num_poisoned_pages
;
1814 extern int soft_offline_page(struct page
*page
, int flags
);
1816 extern void dump_page(struct page
*page
);
1818 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1819 extern void clear_huge_page(struct page
*page
,
1821 unsigned int pages_per_huge_page
);
1822 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1823 unsigned long addr
, struct vm_area_struct
*vma
,
1824 unsigned int pages_per_huge_page
);
1825 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1827 #ifdef CONFIG_DEBUG_PAGEALLOC
1828 extern unsigned int _debug_guardpage_minorder
;
1830 static inline unsigned int debug_guardpage_minorder(void)
1832 return _debug_guardpage_minorder
;
1835 static inline bool page_is_guard(struct page
*page
)
1837 return test_bit(PAGE_DEBUG_FLAG_GUARD
, &page
->debug_flags
);
1840 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1841 static inline bool page_is_guard(struct page
*page
) { return false; }
1842 #endif /* CONFIG_DEBUG_PAGEALLOC */
1844 #if MAX_NUMNODES > 1
1845 void __init
setup_nr_node_ids(void);
1847 static inline void setup_nr_node_ids(void) {}
1850 #endif /* __KERNEL__ */
1851 #endif /* _LINUX_MM_H */