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 */
171 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
174 * vm_fault is filled by the the pagefault handler and passed to the vma's
175 * ->fault function. The vma's ->fault is responsible for returning a bitmask
176 * of VM_FAULT_xxx flags that give details about how the fault was handled.
178 * pgoff should be used in favour of virtual_address, if possible. If pgoff
179 * is used, one may implement ->remap_pages to get nonlinear mapping support.
182 unsigned int flags
; /* FAULT_FLAG_xxx flags */
183 pgoff_t pgoff
; /* Logical page offset based on vma */
184 void __user
*virtual_address
; /* Faulting virtual address */
186 struct page
*page
; /* ->fault handlers should return a
187 * page here, unless VM_FAULT_NOPAGE
188 * is set (which is also implied by
194 * These are the virtual MM functions - opening of an area, closing and
195 * unmapping it (needed to keep files on disk up-to-date etc), pointer
196 * to the functions called when a no-page or a wp-page exception occurs.
198 struct vm_operations_struct
{
199 void (*open
)(struct vm_area_struct
* area
);
200 void (*close
)(struct vm_area_struct
* area
);
201 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
203 /* notification that a previously read-only page is about to become
204 * writable, if an error is returned it will cause a SIGBUS */
205 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
207 /* called by access_process_vm when get_user_pages() fails, typically
208 * for use by special VMAs that can switch between memory and hardware
210 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
211 void *buf
, int len
, int write
);
214 * set_policy() op must add a reference to any non-NULL @new mempolicy
215 * to hold the policy upon return. Caller should pass NULL @new to
216 * remove a policy and fall back to surrounding context--i.e. do not
217 * install a MPOL_DEFAULT policy, nor the task or system default
220 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
223 * get_policy() op must add reference [mpol_get()] to any policy at
224 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
225 * in mm/mempolicy.c will do this automatically.
226 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
227 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
228 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
229 * must return NULL--i.e., do not "fallback" to task or system default
232 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
234 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
235 const nodemask_t
*to
, unsigned long flags
);
237 /* called by sys_remap_file_pages() to populate non-linear mapping */
238 int (*remap_pages
)(struct vm_area_struct
*vma
, unsigned long addr
,
239 unsigned long size
, pgoff_t pgoff
);
245 #define page_private(page) ((page)->private)
246 #define set_page_private(page, v) ((page)->private = (v))
248 /* It's valid only if the page is free path or free_list */
249 static inline void set_freepage_migratetype(struct page
*page
, int migratetype
)
251 page
->index
= migratetype
;
254 /* It's valid only if the page is free path or free_list */
255 static inline int get_freepage_migratetype(struct page
*page
)
261 * FIXME: take this include out, include page-flags.h in
262 * files which need it (119 of them)
264 #include <linux/page-flags.h>
265 #include <linux/huge_mm.h>
268 * Methods to modify the page usage count.
270 * What counts for a page usage:
271 * - cache mapping (page->mapping)
272 * - private data (page->private)
273 * - page mapped in a task's page tables, each mapping
274 * is counted separately
276 * Also, many kernel routines increase the page count before a critical
277 * routine so they can be sure the page doesn't go away from under them.
281 * Drop a ref, return true if the refcount fell to zero (the page has no users)
283 static inline int put_page_testzero(struct page
*page
)
285 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
286 return atomic_dec_and_test(&page
->_count
);
290 * Try to grab a ref unless the page has a refcount of zero, return false if
293 static inline int get_page_unless_zero(struct page
*page
)
295 return atomic_inc_not_zero(&page
->_count
);
298 extern int page_is_ram(unsigned long pfn
);
300 /* Support for virtually mapped pages */
301 struct page
*vmalloc_to_page(const void *addr
);
302 unsigned long vmalloc_to_pfn(const void *addr
);
305 * Determine if an address is within the vmalloc range
307 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
308 * is no special casing required.
310 static inline int is_vmalloc_addr(const void *x
)
313 unsigned long addr
= (unsigned long)x
;
315 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
321 extern int is_vmalloc_or_module_addr(const void *x
);
323 static inline int is_vmalloc_or_module_addr(const void *x
)
329 extern void kvfree(const void *addr
);
331 static inline void compound_lock(struct page
*page
)
333 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
334 VM_BUG_ON(PageSlab(page
));
335 bit_spin_lock(PG_compound_lock
, &page
->flags
);
339 static inline void compound_unlock(struct page
*page
)
341 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
342 VM_BUG_ON(PageSlab(page
));
343 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
347 static inline unsigned long compound_lock_irqsave(struct page
*page
)
349 unsigned long uninitialized_var(flags
);
350 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
351 local_irq_save(flags
);
357 static inline void compound_unlock_irqrestore(struct page
*page
,
360 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
361 compound_unlock(page
);
362 local_irq_restore(flags
);
366 static inline struct page
*compound_head(struct page
*page
)
368 if (unlikely(PageTail(page
))) {
369 struct page
*head
= page
->first_page
;
372 * page->first_page may be a dangling pointer to an old
373 * compound page, so recheck that it is still a tail
374 * page before returning.
377 if (likely(PageTail(page
)))
384 * The atomic page->_mapcount, starts from -1: so that transitions
385 * both from it and to it can be tracked, using atomic_inc_and_test
386 * and atomic_add_negative(-1).
388 static inline void page_mapcount_reset(struct page
*page
)
390 atomic_set(&(page
)->_mapcount
, -1);
393 static inline int page_mapcount(struct page
*page
)
395 return atomic_read(&(page
)->_mapcount
) + 1;
398 static inline int page_count(struct page
*page
)
400 return atomic_read(&compound_head(page
)->_count
);
403 static inline void get_huge_page_tail(struct page
*page
)
406 * __split_huge_page_refcount() cannot run
409 VM_BUG_ON(page_mapcount(page
) < 0);
410 VM_BUG_ON(atomic_read(&page
->_count
) != 0);
411 atomic_inc(&page
->_mapcount
);
414 extern bool __get_page_tail(struct page
*page
);
416 static inline void get_page(struct page
*page
)
418 if (unlikely(PageTail(page
)))
419 if (likely(__get_page_tail(page
)))
422 * Getting a normal page or the head of a compound page
423 * requires to already have an elevated page->_count.
425 VM_BUG_ON(atomic_read(&page
->_count
) <= 0);
426 atomic_inc(&page
->_count
);
429 static inline struct page
*virt_to_head_page(const void *x
)
431 struct page
*page
= virt_to_page(x
);
432 return compound_head(page
);
436 * Setup the page count before being freed into the page allocator for
437 * the first time (boot or memory hotplug)
439 static inline void init_page_count(struct page
*page
)
441 atomic_set(&page
->_count
, 1);
445 * PageBuddy() indicate that the page is free and in the buddy system
446 * (see mm/page_alloc.c).
448 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
449 * -2 so that an underflow of the page_mapcount() won't be mistaken
450 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
451 * efficiently by most CPU architectures.
453 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
455 static inline int PageBuddy(struct page
*page
)
457 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
460 static inline void __SetPageBuddy(struct page
*page
)
462 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
463 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
466 static inline void __ClearPageBuddy(struct page
*page
)
468 VM_BUG_ON(!PageBuddy(page
));
469 atomic_set(&page
->_mapcount
, -1);
472 void put_page(struct page
*page
);
473 void put_pages_list(struct list_head
*pages
);
475 void split_page(struct page
*page
, unsigned int order
);
476 int split_free_page(struct page
*page
);
479 * Compound pages have a destructor function. Provide a
480 * prototype for that function and accessor functions.
481 * These are _only_ valid on the head of a PG_compound page.
483 typedef void compound_page_dtor(struct page
*);
485 static inline void set_compound_page_dtor(struct page
*page
,
486 compound_page_dtor
*dtor
)
488 page
[1].lru
.next
= (void *)dtor
;
491 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
493 return (compound_page_dtor
*)page
[1].lru
.next
;
496 static inline int compound_order(struct page
*page
)
500 return (unsigned long)page
[1].lru
.prev
;
503 static inline int compound_trans_order(struct page
*page
)
511 flags
= compound_lock_irqsave(page
);
512 order
= compound_order(page
);
513 compound_unlock_irqrestore(page
, flags
);
517 static inline void set_compound_order(struct page
*page
, unsigned long order
)
519 page
[1].lru
.prev
= (void *)order
;
524 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
525 * servicing faults for write access. In the normal case, do always want
526 * pte_mkwrite. But get_user_pages can cause write faults for mappings
527 * that do not have writing enabled, when used by access_process_vm.
529 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
531 if (likely(vma
->vm_flags
& VM_WRITE
))
532 pte
= pte_mkwrite(pte
);
538 * Multiple processes may "see" the same page. E.g. for untouched
539 * mappings of /dev/null, all processes see the same page full of
540 * zeroes, and text pages of executables and shared libraries have
541 * only one copy in memory, at most, normally.
543 * For the non-reserved pages, page_count(page) denotes a reference count.
544 * page_count() == 0 means the page is free. page->lru is then used for
545 * freelist management in the buddy allocator.
546 * page_count() > 0 means the page has been allocated.
548 * Pages are allocated by the slab allocator in order to provide memory
549 * to kmalloc and kmem_cache_alloc. In this case, the management of the
550 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
551 * unless a particular usage is carefully commented. (the responsibility of
552 * freeing the kmalloc memory is the caller's, of course).
554 * A page may be used by anyone else who does a __get_free_page().
555 * In this case, page_count still tracks the references, and should only
556 * be used through the normal accessor functions. The top bits of page->flags
557 * and page->virtual store page management information, but all other fields
558 * are unused and could be used privately, carefully. The management of this
559 * page is the responsibility of the one who allocated it, and those who have
560 * subsequently been given references to it.
562 * The other pages (we may call them "pagecache pages") are completely
563 * managed by the Linux memory manager: I/O, buffers, swapping etc.
564 * The following discussion applies only to them.
566 * A pagecache page contains an opaque `private' member, which belongs to the
567 * page's address_space. Usually, this is the address of a circular list of
568 * the page's disk buffers. PG_private must be set to tell the VM to call
569 * into the filesystem to release these pages.
571 * A page may belong to an inode's memory mapping. In this case, page->mapping
572 * is the pointer to the inode, and page->index is the file offset of the page,
573 * in units of PAGE_CACHE_SIZE.
575 * If pagecache pages are not associated with an inode, they are said to be
576 * anonymous pages. These may become associated with the swapcache, and in that
577 * case PG_swapcache is set, and page->private is an offset into the swapcache.
579 * In either case (swapcache or inode backed), the pagecache itself holds one
580 * reference to the page. Setting PG_private should also increment the
581 * refcount. The each user mapping also has a reference to the page.
583 * The pagecache pages are stored in a per-mapping radix tree, which is
584 * rooted at mapping->page_tree, and indexed by offset.
585 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
586 * lists, we instead now tag pages as dirty/writeback in the radix tree.
588 * All pagecache pages may be subject to I/O:
589 * - inode pages may need to be read from disk,
590 * - inode pages which have been modified and are MAP_SHARED may need
591 * to be written back to the inode on disk,
592 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
593 * modified may need to be swapped out to swap space and (later) to be read
598 * The zone field is never updated after free_area_init_core()
599 * sets it, so none of the operations on it need to be atomic.
602 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | 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)
606 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
609 * Define the bit shifts to access each section. For non-existent
610 * sections we define the shift as 0; that plus a 0 mask ensures
611 * the compiler will optimise away reference to them.
613 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
614 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
615 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
616 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
618 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
619 #ifdef NODE_NOT_IN_PAGE_FLAGS
620 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
621 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
622 SECTIONS_PGOFF : ZONES_PGOFF)
624 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
625 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
626 NODES_PGOFF : ZONES_PGOFF)
629 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
631 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
632 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
635 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
636 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
637 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
638 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
639 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
641 static inline enum zone_type
page_zonenum(const struct page
*page
)
643 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
646 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
647 #define SECTION_IN_PAGE_FLAGS
651 * The identification function is only used by the buddy allocator for
652 * determining if two pages could be buddies. We are not really
653 * identifying a zone since we could be using a the section number
654 * id if we have not node id available in page flags.
655 * We guarantee only that it will return the same value for two
656 * combinable pages in a zone.
658 static inline int page_zone_id(struct page
*page
)
660 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
663 static inline int zone_to_nid(struct zone
*zone
)
672 #ifdef NODE_NOT_IN_PAGE_FLAGS
673 extern int page_to_nid(const struct page
*page
);
675 static inline int page_to_nid(const struct page
*page
)
677 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
681 #ifdef CONFIG_NUMA_BALANCING
682 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
683 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
685 return xchg(&page
->_last_nid
, nid
);
688 static inline int page_nid_last(struct page
*page
)
690 return page
->_last_nid
;
692 static inline void page_nid_reset_last(struct page
*page
)
694 page
->_last_nid
= -1;
697 static inline int page_nid_last(struct page
*page
)
699 return (page
->flags
>> LAST_NID_PGSHIFT
) & LAST_NID_MASK
;
702 extern int page_nid_xchg_last(struct page
*page
, int nid
);
704 static inline void page_nid_reset_last(struct page
*page
)
706 int nid
= (1 << LAST_NID_SHIFT
) - 1;
708 page
->flags
&= ~(LAST_NID_MASK
<< LAST_NID_PGSHIFT
);
709 page
->flags
|= (nid
& LAST_NID_MASK
) << LAST_NID_PGSHIFT
;
711 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
713 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
715 return page_to_nid(page
);
718 static inline int page_nid_last(struct page
*page
)
720 return page_to_nid(page
);
723 static inline void page_nid_reset_last(struct page
*page
)
728 static inline struct zone
*page_zone(const struct page
*page
)
730 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
733 #ifdef SECTION_IN_PAGE_FLAGS
734 static inline void set_page_section(struct page
*page
, unsigned long section
)
736 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
737 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
740 static inline unsigned long page_to_section(const struct page
*page
)
742 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
746 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
748 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
749 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
752 static inline void set_page_node(struct page
*page
, unsigned long node
)
754 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
755 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
758 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
759 unsigned long node
, unsigned long pfn
)
761 set_page_zone(page
, zone
);
762 set_page_node(page
, node
);
763 #ifdef SECTION_IN_PAGE_FLAGS
764 set_page_section(page
, pfn_to_section_nr(pfn
));
769 * Some inline functions in vmstat.h depend on page_zone()
771 #include <linux/vmstat.h>
773 static __always_inline
void *lowmem_page_address(const struct page
*page
)
775 return __va(PFN_PHYS(page_to_pfn(page
)));
778 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
779 #define HASHED_PAGE_VIRTUAL
782 #if defined(WANT_PAGE_VIRTUAL)
783 static inline void *page_address(const struct page
*page
)
785 return page
->virtual;
787 static inline void set_page_address(struct page
*page
, void *address
)
789 page
->virtual = address
;
791 #define page_address_init() do { } while(0)
794 #if defined(HASHED_PAGE_VIRTUAL)
795 void *page_address(const struct page
*page
);
796 void set_page_address(struct page
*page
, void *virtual);
797 void page_address_init(void);
800 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
801 #define page_address(page) lowmem_page_address(page)
802 #define set_page_address(page, address) do { } while(0)
803 #define page_address_init() do { } while(0)
807 * On an anonymous page mapped into a user virtual memory area,
808 * page->mapping points to its anon_vma, not to a struct address_space;
809 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
811 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
812 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
813 * and then page->mapping points, not to an anon_vma, but to a private
814 * structure which KSM associates with that merged page. See ksm.h.
816 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
818 * Please note that, confusingly, "page_mapping" refers to the inode
819 * address_space which maps the page from disk; whereas "page_mapped"
820 * refers to user virtual address space into which the page is mapped.
822 #define PAGE_MAPPING_ANON 1
823 #define PAGE_MAPPING_KSM 2
824 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
826 extern struct address_space
*page_mapping(struct page
*page
);
828 /* Neutral page->mapping pointer to address_space or anon_vma or other */
829 static inline void *page_rmapping(struct page
*page
)
831 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
834 extern struct address_space
*__page_file_mapping(struct page
*);
837 struct address_space
*page_file_mapping(struct page
*page
)
839 if (unlikely(PageSwapCache(page
)))
840 return __page_file_mapping(page
);
842 return page
->mapping
;
845 static inline int PageAnon(struct page
*page
)
847 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
851 * Return the pagecache index of the passed page. Regular pagecache pages
852 * use ->index whereas swapcache pages use ->private
854 static inline pgoff_t
page_index(struct page
*page
)
856 if (unlikely(PageSwapCache(page
)))
857 return page_private(page
);
861 extern pgoff_t
__page_file_index(struct page
*page
);
864 * Return the file index of the page. Regular pagecache pages use ->index
865 * whereas swapcache pages use swp_offset(->private)
867 static inline pgoff_t
page_file_index(struct page
*page
)
869 if (unlikely(PageSwapCache(page
)))
870 return __page_file_index(page
);
876 * Return true if this page is mapped into pagetables.
878 static inline int page_mapped(struct page
*page
)
880 return atomic_read(&(page
)->_mapcount
) >= 0;
884 * Different kinds of faults, as returned by handle_mm_fault().
885 * Used to decide whether a process gets delivered SIGBUS or
886 * just gets major/minor fault counters bumped up.
889 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
891 #define VM_FAULT_OOM 0x0001
892 #define VM_FAULT_SIGBUS 0x0002
893 #define VM_FAULT_MAJOR 0x0004
894 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
895 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
896 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
897 #define VM_FAULT_SIGSEGV 0x0040
899 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
900 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
901 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
903 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
905 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
906 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)
908 /* Encode hstate index for a hwpoisoned large page */
909 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
910 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
913 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
915 extern void pagefault_out_of_memory(void);
917 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
920 * Flags passed to show_mem() and show_free_areas() to suppress output in
923 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
924 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
926 extern void show_free_areas(unsigned int flags
);
927 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
929 void shmem_set_file(struct vm_area_struct
*vma
, struct file
*file
);
930 int shmem_zero_setup(struct vm_area_struct
*);
932 extern int can_do_mlock(void);
933 extern int user_shm_lock(size_t, struct user_struct
*);
934 extern void user_shm_unlock(size_t, struct user_struct
*);
937 * Parameter block passed down to zap_pte_range in exceptional cases.
940 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
941 struct address_space
*check_mapping
; /* Check page->mapping if set */
942 pgoff_t first_index
; /* Lowest page->index to unmap */
943 pgoff_t last_index
; /* Highest page->index to unmap */
946 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
949 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
951 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
952 unsigned long size
, struct zap_details
*);
953 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
954 unsigned long start
, unsigned long end
);
957 * mm_walk - callbacks for walk_page_range
958 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
959 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
960 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
961 * this handler is required to be able to handle
962 * pmd_trans_huge() pmds. They may simply choose to
963 * split_huge_page() instead of handling it explicitly.
964 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
965 * @pte_hole: if set, called for each hole at all levels
966 * @hugetlb_entry: if set, called for each hugetlb entry
967 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
970 * (see walk_page_range for more details)
973 int (*pgd_entry
)(pgd_t
*pgd
, unsigned long addr
,
974 unsigned long next
, struct mm_walk
*walk
);
975 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
976 unsigned long next
, struct mm_walk
*walk
);
977 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
978 unsigned long next
, struct mm_walk
*walk
);
979 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
980 unsigned long next
, struct mm_walk
*walk
);
981 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
982 struct mm_walk
*walk
);
983 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
984 unsigned long addr
, unsigned long next
,
985 struct mm_walk
*walk
);
986 struct mm_struct
*mm
;
990 int walk_page_range(unsigned long addr
, unsigned long end
,
991 struct mm_walk
*walk
);
992 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
993 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
994 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
995 struct vm_area_struct
*vma
);
996 void unmap_mapping_range(struct address_space
*mapping
,
997 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
998 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1000 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1001 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1002 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1003 void *buf
, int len
, int write
);
1005 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1006 loff_t
const holebegin
, loff_t
const holelen
)
1008 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1011 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
1012 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1013 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1014 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1015 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1016 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1017 int invalidate_inode_page(struct page
*page
);
1020 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1021 unsigned long address
, unsigned int flags
);
1022 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1023 unsigned long address
, unsigned int fault_flags
);
1025 static inline int handle_mm_fault(struct mm_struct
*mm
,
1026 struct vm_area_struct
*vma
, unsigned long address
,
1029 /* should never happen if there's no MMU */
1031 return VM_FAULT_SIGBUS
;
1033 static inline int fixup_user_fault(struct task_struct
*tsk
,
1034 struct mm_struct
*mm
, unsigned long address
,
1035 unsigned int fault_flags
)
1037 /* should never happen if there's no MMU */
1043 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1044 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1045 void *buf
, int len
, int write
);
1047 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1048 unsigned long start
, unsigned long nr_pages
,
1049 unsigned int foll_flags
, struct page
**pages
,
1050 struct vm_area_struct
**vmas
, int *nonblocking
);
1051 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1052 unsigned long start
, unsigned long nr_pages
,
1053 int write
, int force
, struct page
**pages
,
1054 struct vm_area_struct
**vmas
);
1055 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1056 struct page
**pages
);
1058 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1059 struct page
**pages
);
1060 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1061 struct page
*get_dump_page(unsigned long addr
);
1063 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1064 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
1066 int __set_page_dirty_nobuffers(struct page
*page
);
1067 int __set_page_dirty_no_writeback(struct page
*page
);
1068 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1070 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1071 void account_page_writeback(struct page
*page
);
1072 int set_page_dirty(struct page
*page
);
1073 int set_page_dirty_lock(struct page
*page
);
1074 int clear_page_dirty_for_io(struct page
*page
);
1076 /* Is the vma a continuation of the stack vma above it? */
1077 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1079 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1082 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1085 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1086 (vma
->vm_start
== addr
) &&
1087 !vma_growsdown(vma
->vm_prev
, addr
);
1090 /* Is the vma a continuation of the stack vma below it? */
1091 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1093 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1096 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1099 return (vma
->vm_flags
& VM_GROWSUP
) &&
1100 (vma
->vm_end
== addr
) &&
1101 !vma_growsup(vma
->vm_next
, addr
);
1105 vm_is_stack(struct task_struct
*task
, struct vm_area_struct
*vma
, int in_group
);
1107 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1108 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1109 unsigned long new_addr
, unsigned long len
,
1110 bool need_rmap_locks
);
1111 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1112 unsigned long end
, pgprot_t newprot
,
1113 int dirty_accountable
, int prot_numa
);
1114 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1115 struct vm_area_struct
**pprev
, unsigned long start
,
1116 unsigned long end
, unsigned long newflags
);
1119 * doesn't attempt to fault and will return short.
1121 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1122 struct page
**pages
);
1124 * per-process(per-mm_struct) statistics.
1126 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1128 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1130 #ifdef SPLIT_RSS_COUNTING
1132 * counter is updated in asynchronous manner and may go to minus.
1133 * But it's never be expected number for users.
1138 return (unsigned long)val
;
1141 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1143 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1146 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1148 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1151 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1153 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1156 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1158 return get_mm_counter(mm
, MM_FILEPAGES
) +
1159 get_mm_counter(mm
, MM_ANONPAGES
);
1162 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1164 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1167 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1169 return max(mm
->hiwater_vm
, mm
->total_vm
);
1172 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1174 unsigned long _rss
= get_mm_rss(mm
);
1176 if ((mm
)->hiwater_rss
< _rss
)
1177 (mm
)->hiwater_rss
= _rss
;
1180 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1182 if (mm
->hiwater_vm
< mm
->total_vm
)
1183 mm
->hiwater_vm
= mm
->total_vm
;
1186 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1187 struct mm_struct
*mm
)
1189 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1191 if (*maxrss
< hiwater_rss
)
1192 *maxrss
= hiwater_rss
;
1195 #if defined(SPLIT_RSS_COUNTING)
1196 void sync_mm_rss(struct mm_struct
*mm
);
1198 static inline void sync_mm_rss(struct mm_struct
*mm
)
1203 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1205 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1207 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1211 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1215 #ifdef __PAGETABLE_PUD_FOLDED
1216 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1217 unsigned long address
)
1222 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1225 #ifdef __PAGETABLE_PMD_FOLDED
1226 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1227 unsigned long address
)
1232 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1235 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1236 pmd_t
*pmd
, unsigned long address
);
1237 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1240 * The following ifdef needed to get the 4level-fixup.h header to work.
1241 * Remove it when 4level-fixup.h has been removed.
1243 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1244 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1246 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1247 NULL
: pud_offset(pgd
, address
);
1250 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1252 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1253 NULL
: pmd_offset(pud
, address
);
1255 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1257 #if USE_SPLIT_PTLOCKS
1259 * We tuck a spinlock to guard each pagetable page into its struct page,
1260 * at page->private, with BUILD_BUG_ON to make sure that this will not
1261 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1262 * When freeing, reset page->mapping so free_pages_check won't complain.
1264 #define __pte_lockptr(page) &((page)->ptl)
1265 #define pte_lock_init(_page) do { \
1266 spin_lock_init(__pte_lockptr(_page)); \
1268 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1269 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1270 #else /* !USE_SPLIT_PTLOCKS */
1272 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1274 #define pte_lock_init(page) do {} while (0)
1275 #define pte_lock_deinit(page) do {} while (0)
1276 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1277 #endif /* USE_SPLIT_PTLOCKS */
1279 static inline void pgtable_page_ctor(struct page
*page
)
1281 pte_lock_init(page
);
1282 inc_zone_page_state(page
, NR_PAGETABLE
);
1285 static inline void pgtable_page_dtor(struct page
*page
)
1287 pte_lock_deinit(page
);
1288 dec_zone_page_state(page
, NR_PAGETABLE
);
1291 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1293 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1294 pte_t *__pte = pte_offset_map(pmd, address); \
1300 #define pte_unmap_unlock(pte, ptl) do { \
1305 #define pte_alloc_map(mm, vma, pmd, address) \
1306 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1308 NULL: pte_offset_map(pmd, address))
1310 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1311 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1313 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1315 #define pte_alloc_kernel(pmd, address) \
1316 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1317 NULL: pte_offset_kernel(pmd, address))
1319 extern void free_area_init(unsigned long * zones_size
);
1320 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1321 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1322 extern void free_initmem(void);
1325 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1326 * into the buddy system. The freed pages will be poisoned with pattern
1327 * "poison" if it's non-zero.
1328 * Return pages freed into the buddy system.
1330 extern unsigned long free_reserved_area(unsigned long start
, unsigned long end
,
1331 int poison
, char *s
);
1332 #ifdef CONFIG_HIGHMEM
1334 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1335 * and totalram_pages.
1337 extern void free_highmem_page(struct page
*page
);
1340 static inline void adjust_managed_page_count(struct page
*page
, long count
)
1342 totalram_pages
+= count
;
1345 /* Free the reserved page into the buddy system, so it gets managed. */
1346 static inline void __free_reserved_page(struct page
*page
)
1348 ClearPageReserved(page
);
1349 init_page_count(page
);
1353 static inline void free_reserved_page(struct page
*page
)
1355 __free_reserved_page(page
);
1356 adjust_managed_page_count(page
, 1);
1359 static inline void mark_page_reserved(struct page
*page
)
1361 SetPageReserved(page
);
1362 adjust_managed_page_count(page
, -1);
1366 * Default method to free all the __init memory into the buddy system.
1367 * The freed pages will be poisoned with pattern "poison" if it is
1368 * non-zero. Return pages freed into the buddy system.
1370 static inline unsigned long free_initmem_default(int poison
)
1372 extern char __init_begin
[], __init_end
[];
1374 return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin
) ,
1375 ((unsigned long)&__init_end
) & PAGE_MASK
,
1376 poison
, "unused kernel");
1379 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1381 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1382 * zones, allocate the backing mem_map and account for memory holes in a more
1383 * architecture independent manner. This is a substitute for creating the
1384 * zone_sizes[] and zholes_size[] arrays and passing them to
1385 * free_area_init_node()
1387 * An architecture is expected to register range of page frames backed by
1388 * physical memory with memblock_add[_node]() before calling
1389 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1390 * usage, an architecture is expected to do something like
1392 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1394 * for_each_valid_physical_page_range()
1395 * memblock_add_node(base, size, nid)
1396 * free_area_init_nodes(max_zone_pfns);
1398 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1399 * registered physical page range. Similarly
1400 * sparse_memory_present_with_active_regions() calls memory_present() for
1401 * each range when SPARSEMEM is enabled.
1403 * See mm/page_alloc.c for more information on each function exposed by
1404 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1406 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1407 unsigned long node_map_pfn_alignment(void);
1408 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1409 unsigned long end_pfn
);
1410 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1411 unsigned long end_pfn
);
1412 extern void get_pfn_range_for_nid(unsigned int nid
,
1413 unsigned long *start_pfn
, unsigned long *end_pfn
);
1414 extern unsigned long find_min_pfn_with_active_regions(void);
1415 extern void free_bootmem_with_active_regions(int nid
,
1416 unsigned long max_low_pfn
);
1417 extern void sparse_memory_present_with_active_regions(int nid
);
1419 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1421 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1422 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1423 static inline int __early_pfn_to_nid(unsigned long pfn
)
1428 /* please see mm/page_alloc.c */
1429 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1430 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1431 /* there is a per-arch backend function. */
1432 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1433 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1436 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1437 extern void memmap_init_zone(unsigned long, int, unsigned long,
1438 unsigned long, enum memmap_context
);
1439 extern void setup_per_zone_wmarks(void);
1440 extern int __meminit
init_per_zone_wmark_min(void);
1441 extern void mem_init(void);
1442 extern void __init
mmap_init(void);
1443 extern void show_mem(unsigned int flags
);
1444 extern void si_meminfo(struct sysinfo
* val
);
1445 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1447 extern __printf(3, 4)
1448 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1450 extern void setup_per_cpu_pageset(void);
1452 extern void zone_pcp_update(struct zone
*zone
);
1453 extern void zone_pcp_reset(struct zone
*zone
);
1456 extern int min_free_kbytes
;
1459 extern atomic_long_t mmap_pages_allocated
;
1460 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1462 /* interval_tree.c */
1463 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1464 struct rb_root
*root
);
1465 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1466 struct vm_area_struct
*prev
,
1467 struct rb_root
*root
);
1468 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1469 struct rb_root
*root
);
1470 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1471 unsigned long start
, unsigned long last
);
1472 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1473 unsigned long start
, unsigned long last
);
1475 #define vma_interval_tree_foreach(vma, root, start, last) \
1476 for (vma = vma_interval_tree_iter_first(root, start, last); \
1477 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1479 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1480 struct list_head
*list
)
1482 list_add_tail(&vma
->shared
.nonlinear
, list
);
1485 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1486 struct rb_root
*root
);
1487 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1488 struct rb_root
*root
);
1489 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1490 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1491 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1492 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1493 #ifdef CONFIG_DEBUG_VM_RB
1494 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1497 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1498 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1499 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1502 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1503 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1504 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1505 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1506 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1507 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1508 struct mempolicy
*, const char __user
*);
1509 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1510 extern int split_vma(struct mm_struct
*,
1511 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1512 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1513 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1514 struct rb_node
**, struct rb_node
*);
1515 extern void unlink_file_vma(struct vm_area_struct
*);
1516 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1517 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1518 bool *need_rmap_locks
);
1519 extern void exit_mmap(struct mm_struct
*);
1521 extern int mm_take_all_locks(struct mm_struct
*mm
);
1522 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1524 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1525 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1527 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1528 extern int install_special_mapping(struct mm_struct
*mm
,
1529 unsigned long addr
, unsigned long len
,
1530 unsigned long flags
, struct page
**pages
);
1532 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1534 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1535 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1536 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1537 unsigned long len
, unsigned long prot
, unsigned long flags
,
1538 unsigned long pgoff
, unsigned long *populate
);
1539 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1542 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1544 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1547 (void) __mm_populate(addr
, len
, 1);
1550 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1553 /* These take the mm semaphore themselves */
1554 extern unsigned long vm_brk(unsigned long, unsigned long);
1555 extern int vm_munmap(unsigned long, size_t);
1556 extern unsigned long vm_mmap(struct file
*, unsigned long,
1557 unsigned long, unsigned long,
1558 unsigned long, unsigned long);
1560 struct vm_unmapped_area_info
{
1561 #define VM_UNMAPPED_AREA_TOPDOWN 1
1562 unsigned long flags
;
1563 unsigned long length
;
1564 unsigned long low_limit
;
1565 unsigned long high_limit
;
1566 unsigned long align_mask
;
1567 unsigned long align_offset
;
1570 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1571 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1574 * Search for an unmapped address range.
1576 * We are looking for a range that:
1577 * - does not intersect with any VMA;
1578 * - is contained within the [low_limit, high_limit) interval;
1579 * - is at least the desired size.
1580 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1582 static inline unsigned long
1583 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1585 if (!(info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
))
1586 return unmapped_area(info
);
1588 return unmapped_area_topdown(info
);
1592 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1593 extern void truncate_inode_pages_range(struct address_space
*,
1594 loff_t lstart
, loff_t lend
);
1596 /* generic vm_area_ops exported for stackable file systems */
1597 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1598 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1600 /* mm/page-writeback.c */
1601 int write_one_page(struct page
*page
, int wait
);
1602 void task_dirty_inc(struct task_struct
*tsk
);
1605 #define VM_MAX_READAHEAD 128 /* kbytes */
1606 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1608 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1609 pgoff_t offset
, unsigned long nr_to_read
);
1611 void page_cache_sync_readahead(struct address_space
*mapping
,
1612 struct file_ra_state
*ra
,
1615 unsigned long size
);
1617 void page_cache_async_readahead(struct address_space
*mapping
,
1618 struct file_ra_state
*ra
,
1622 unsigned long size
);
1624 unsigned long max_sane_readahead(unsigned long nr
);
1625 unsigned long ra_submit(struct file_ra_state
*ra
,
1626 struct address_space
*mapping
,
1629 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1630 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1632 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1633 extern int expand_downwards(struct vm_area_struct
*vma
,
1634 unsigned long address
);
1636 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1638 #define expand_upwards(vma, address) (0)
1641 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1642 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1643 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1644 struct vm_area_struct
**pprev
);
1646 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1647 NULL if none. Assume start_addr < end_addr. */
1648 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1650 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1652 if (vma
&& end_addr
<= vma
->vm_start
)
1657 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1659 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1662 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1663 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
1664 unsigned long vm_start
, unsigned long vm_end
)
1666 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
1668 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
1675 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1677 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1683 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1684 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
1685 unsigned long start
, unsigned long end
);
1688 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1689 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1690 unsigned long pfn
, unsigned long size
, pgprot_t
);
1691 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1692 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1694 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1696 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
1699 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
1700 unsigned long address
, unsigned int foll_flags
,
1701 unsigned int *page_mask
);
1703 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
1704 unsigned long address
, unsigned int foll_flags
)
1706 unsigned int unused_page_mask
;
1707 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
1710 #define FOLL_WRITE 0x01 /* check pte is writable */
1711 #define FOLL_TOUCH 0x02 /* mark page accessed */
1712 #define FOLL_GET 0x04 /* do get_page on page */
1713 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1714 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1715 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1716 * and return without waiting upon it */
1717 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1718 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1719 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1720 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1721 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1722 #define FOLL_COW 0x4000 /* internal GUP flag */
1724 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1726 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1727 unsigned long size
, pte_fn_t fn
, void *data
);
1729 #ifdef CONFIG_PROC_FS
1730 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1732 static inline void vm_stat_account(struct mm_struct
*mm
,
1733 unsigned long flags
, struct file
*file
, long pages
)
1735 mm
->total_vm
+= pages
;
1737 #endif /* CONFIG_PROC_FS */
1739 #ifdef CONFIG_DEBUG_PAGEALLOC
1740 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1741 #ifdef CONFIG_HIBERNATION
1742 extern bool kernel_page_present(struct page
*page
);
1743 #endif /* CONFIG_HIBERNATION */
1746 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1747 #ifdef CONFIG_HIBERNATION
1748 static inline bool kernel_page_present(struct page
*page
) { return true; }
1749 #endif /* CONFIG_HIBERNATION */
1752 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
1753 #ifdef __HAVE_ARCH_GATE_AREA
1754 int in_gate_area_no_mm(unsigned long addr
);
1755 int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
1757 int in_gate_area_no_mm(unsigned long addr
);
1758 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1759 #endif /* __HAVE_ARCH_GATE_AREA */
1761 #ifdef CONFIG_SYSCTL
1762 extern int sysctl_drop_caches
;
1763 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1764 void __user
*, size_t *, loff_t
*);
1767 unsigned long shrink_slab(struct shrink_control
*shrink
,
1768 unsigned long nr_pages_scanned
,
1769 unsigned long lru_pages
);
1770 void drop_pagecache(void);
1773 #define randomize_va_space 0
1775 extern int randomize_va_space
;
1778 const char * arch_vma_name(struct vm_area_struct
*vma
);
1779 void print_vma_addr(char *prefix
, unsigned long rip
);
1781 void sparse_mem_maps_populate_node(struct page
**map_map
,
1782 unsigned long pnum_begin
,
1783 unsigned long pnum_end
,
1784 unsigned long map_count
,
1787 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1788 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1789 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1790 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1791 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1792 void *vmemmap_alloc_block(unsigned long size
, int node
);
1793 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1794 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1795 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
1797 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
1798 void vmemmap_populate_print_last(void);
1799 #ifdef CONFIG_MEMORY_HOTPLUG
1800 void vmemmap_free(unsigned long start
, unsigned long end
);
1802 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
1803 unsigned long size
);
1806 MF_COUNT_INCREASED
= 1 << 0,
1807 MF_ACTION_REQUIRED
= 1 << 1,
1808 MF_MUST_KILL
= 1 << 2,
1810 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
1811 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
1812 extern int unpoison_memory(unsigned long pfn
);
1813 extern int sysctl_memory_failure_early_kill
;
1814 extern int sysctl_memory_failure_recovery
;
1815 extern void shake_page(struct page
*p
, int access
);
1816 extern atomic_long_t num_poisoned_pages
;
1817 extern int soft_offline_page(struct page
*page
, int flags
);
1819 extern void dump_page(struct page
*page
);
1821 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1822 extern void clear_huge_page(struct page
*page
,
1824 unsigned int pages_per_huge_page
);
1825 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1826 unsigned long addr
, struct vm_area_struct
*vma
,
1827 unsigned int pages_per_huge_page
);
1828 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1830 #ifdef CONFIG_DEBUG_PAGEALLOC
1831 extern unsigned int _debug_guardpage_minorder
;
1833 static inline unsigned int debug_guardpage_minorder(void)
1835 return _debug_guardpage_minorder
;
1838 static inline bool page_is_guard(struct page
*page
)
1840 return test_bit(PAGE_DEBUG_FLAG_GUARD
, &page
->debug_flags
);
1843 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1844 static inline bool page_is_guard(struct page
*page
) { return false; }
1845 #endif /* CONFIG_DEBUG_PAGEALLOC */
1847 #if MAX_NUMNODES > 1
1848 void __init
setup_nr_node_ids(void);
1850 static inline void setup_nr_node_ids(void) {}
1853 #endif /* __KERNEL__ */
1854 #endif /* _LINUX_MM_H */