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_ACCOUNT 0x00100000 /* Is a VM accounted object */
103 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
104 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
105 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
106 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
107 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
109 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
110 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
111 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
112 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
114 #if defined(CONFIG_X86)
115 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
116 #elif defined(CONFIG_PPC)
117 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
118 #elif defined(CONFIG_PARISC)
119 # define VM_GROWSUP VM_ARCH_1
120 #elif defined(CONFIG_METAG)
121 # define VM_GROWSUP VM_ARCH_1
122 #elif defined(CONFIG_IA64)
123 # define VM_GROWSUP VM_ARCH_1
124 #elif !defined(CONFIG_MMU)
125 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
129 # define VM_GROWSUP VM_NONE
132 /* Bits set in the VMA until the stack is in its final location */
133 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
135 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
136 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
139 #ifdef CONFIG_STACK_GROWSUP
140 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
142 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
145 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
146 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
147 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
148 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
149 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
152 * Special vmas that are non-mergable, non-mlock()able.
153 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
155 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
158 * mapping from the currently active vm_flags protection bits (the
159 * low four bits) to a page protection mask..
161 extern pgprot_t protection_map
[16];
163 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
164 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
165 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
166 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
167 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
168 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
169 #define FAULT_FLAG_TRIED 0x40 /* second try */
170 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
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 static inline void compound_lock(struct page
*page
)
330 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
331 VM_BUG_ON(PageSlab(page
));
332 bit_spin_lock(PG_compound_lock
, &page
->flags
);
336 static inline void compound_unlock(struct page
*page
)
338 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
339 VM_BUG_ON(PageSlab(page
));
340 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
344 static inline unsigned long compound_lock_irqsave(struct page
*page
)
346 unsigned long uninitialized_var(flags
);
347 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
348 local_irq_save(flags
);
354 static inline void compound_unlock_irqrestore(struct page
*page
,
357 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
358 compound_unlock(page
);
359 local_irq_restore(flags
);
363 static inline struct page
*compound_head(struct page
*page
)
365 if (unlikely(PageTail(page
))) {
366 struct page
*head
= page
->first_page
;
369 * page->first_page may be a dangling pointer to an old
370 * compound page, so recheck that it is still a tail
371 * page before returning.
374 if (likely(PageTail(page
)))
381 * The atomic page->_mapcount, starts from -1: so that transitions
382 * both from it and to it can be tracked, using atomic_inc_and_test
383 * and atomic_add_negative(-1).
385 static inline void page_mapcount_reset(struct page
*page
)
387 atomic_set(&(page
)->_mapcount
, -1);
390 static inline int page_mapcount(struct page
*page
)
392 return atomic_read(&(page
)->_mapcount
) + 1;
395 static inline int page_count(struct page
*page
)
397 return atomic_read(&compound_head(page
)->_count
);
400 static inline void get_huge_page_tail(struct page
*page
)
403 * __split_huge_page_refcount() cannot run
406 VM_BUG_ON(page_mapcount(page
) < 0);
407 VM_BUG_ON(atomic_read(&page
->_count
) != 0);
408 atomic_inc(&page
->_mapcount
);
411 extern bool __get_page_tail(struct page
*page
);
413 static inline void get_page(struct page
*page
)
415 if (unlikely(PageTail(page
)))
416 if (likely(__get_page_tail(page
)))
419 * Getting a normal page or the head of a compound page
420 * requires to already have an elevated page->_count.
422 VM_BUG_ON(atomic_read(&page
->_count
) <= 0);
423 atomic_inc(&page
->_count
);
426 static inline struct page
*virt_to_head_page(const void *x
)
428 struct page
*page
= virt_to_page(x
);
429 return compound_head(page
);
433 * Setup the page count before being freed into the page allocator for
434 * the first time (boot or memory hotplug)
436 static inline void init_page_count(struct page
*page
)
438 atomic_set(&page
->_count
, 1);
442 * PageBuddy() indicate that the page is free and in the buddy system
443 * (see mm/page_alloc.c).
445 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
446 * -2 so that an underflow of the page_mapcount() won't be mistaken
447 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
448 * efficiently by most CPU architectures.
450 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
452 static inline int PageBuddy(struct page
*page
)
454 return atomic_read(&page
->_mapcount
) == PAGE_BUDDY_MAPCOUNT_VALUE
;
457 static inline void __SetPageBuddy(struct page
*page
)
459 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
460 atomic_set(&page
->_mapcount
, PAGE_BUDDY_MAPCOUNT_VALUE
);
463 static inline void __ClearPageBuddy(struct page
*page
)
465 VM_BUG_ON(!PageBuddy(page
));
466 atomic_set(&page
->_mapcount
, -1);
469 void put_page(struct page
*page
);
470 void put_pages_list(struct list_head
*pages
);
472 void split_page(struct page
*page
, unsigned int order
);
473 int split_free_page(struct page
*page
);
476 * Compound pages have a destructor function. Provide a
477 * prototype for that function and accessor functions.
478 * These are _only_ valid on the head of a PG_compound page.
480 typedef void compound_page_dtor(struct page
*);
482 static inline void set_compound_page_dtor(struct page
*page
,
483 compound_page_dtor
*dtor
)
485 page
[1].lru
.next
= (void *)dtor
;
488 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
490 return (compound_page_dtor
*)page
[1].lru
.next
;
493 static inline int compound_order(struct page
*page
)
497 return (unsigned long)page
[1].lru
.prev
;
500 static inline int compound_trans_order(struct page
*page
)
508 flags
= compound_lock_irqsave(page
);
509 order
= compound_order(page
);
510 compound_unlock_irqrestore(page
, flags
);
514 static inline void set_compound_order(struct page
*page
, unsigned long order
)
516 page
[1].lru
.prev
= (void *)order
;
521 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
522 * servicing faults for write access. In the normal case, do always want
523 * pte_mkwrite. But get_user_pages can cause write faults for mappings
524 * that do not have writing enabled, when used by access_process_vm.
526 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
528 if (likely(vma
->vm_flags
& VM_WRITE
))
529 pte
= pte_mkwrite(pte
);
535 * Multiple processes may "see" the same page. E.g. for untouched
536 * mappings of /dev/null, all processes see the same page full of
537 * zeroes, and text pages of executables and shared libraries have
538 * only one copy in memory, at most, normally.
540 * For the non-reserved pages, page_count(page) denotes a reference count.
541 * page_count() == 0 means the page is free. page->lru is then used for
542 * freelist management in the buddy allocator.
543 * page_count() > 0 means the page has been allocated.
545 * Pages are allocated by the slab allocator in order to provide memory
546 * to kmalloc and kmem_cache_alloc. In this case, the management of the
547 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
548 * unless a particular usage is carefully commented. (the responsibility of
549 * freeing the kmalloc memory is the caller's, of course).
551 * A page may be used by anyone else who does a __get_free_page().
552 * In this case, page_count still tracks the references, and should only
553 * be used through the normal accessor functions. The top bits of page->flags
554 * and page->virtual store page management information, but all other fields
555 * are unused and could be used privately, carefully. The management of this
556 * page is the responsibility of the one who allocated it, and those who have
557 * subsequently been given references to it.
559 * The other pages (we may call them "pagecache pages") are completely
560 * managed by the Linux memory manager: I/O, buffers, swapping etc.
561 * The following discussion applies only to them.
563 * A pagecache page contains an opaque `private' member, which belongs to the
564 * page's address_space. Usually, this is the address of a circular list of
565 * the page's disk buffers. PG_private must be set to tell the VM to call
566 * into the filesystem to release these pages.
568 * A page may belong to an inode's memory mapping. In this case, page->mapping
569 * is the pointer to the inode, and page->index is the file offset of the page,
570 * in units of PAGE_CACHE_SIZE.
572 * If pagecache pages are not associated with an inode, they are said to be
573 * anonymous pages. These may become associated with the swapcache, and in that
574 * case PG_swapcache is set, and page->private is an offset into the swapcache.
576 * In either case (swapcache or inode backed), the pagecache itself holds one
577 * reference to the page. Setting PG_private should also increment the
578 * refcount. The each user mapping also has a reference to the page.
580 * The pagecache pages are stored in a per-mapping radix tree, which is
581 * rooted at mapping->page_tree, and indexed by offset.
582 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
583 * lists, we instead now tag pages as dirty/writeback in the radix tree.
585 * All pagecache pages may be subject to I/O:
586 * - inode pages may need to be read from disk,
587 * - inode pages which have been modified and are MAP_SHARED may need
588 * to be written back to the inode on disk,
589 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
590 * modified may need to be swapped out to swap space and (later) to be read
595 * The zone field is never updated after free_area_init_core()
596 * sets it, so none of the operations on it need to be atomic.
599 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
600 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
601 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
602 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
603 #define LAST_NID_PGOFF (ZONES_PGOFF - LAST_NID_WIDTH)
606 * Define the bit shifts to access each section. For non-existent
607 * sections we define the shift as 0; that plus a 0 mask ensures
608 * the compiler will optimise away reference to them.
610 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
611 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
612 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
613 #define LAST_NID_PGSHIFT (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
615 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
616 #ifdef NODE_NOT_IN_PAGE_FLAGS
617 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
618 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
619 SECTIONS_PGOFF : ZONES_PGOFF)
621 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
622 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
623 NODES_PGOFF : ZONES_PGOFF)
626 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
628 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
629 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
632 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
633 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
634 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
635 #define LAST_NID_MASK ((1UL << LAST_NID_WIDTH) - 1)
636 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
638 static inline enum zone_type
page_zonenum(const struct page
*page
)
640 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
643 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
644 #define SECTION_IN_PAGE_FLAGS
648 * The identification function is only used by the buddy allocator for
649 * determining if two pages could be buddies. We are not really
650 * identifying a zone since we could be using a the section number
651 * id if we have not node id available in page flags.
652 * We guarantee only that it will return the same value for two
653 * combinable pages in a zone.
655 static inline int page_zone_id(struct page
*page
)
657 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
660 static inline int zone_to_nid(struct zone
*zone
)
669 #ifdef NODE_NOT_IN_PAGE_FLAGS
670 extern int page_to_nid(const struct page
*page
);
672 static inline int page_to_nid(const struct page
*page
)
674 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
678 #ifdef CONFIG_NUMA_BALANCING
679 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
680 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
682 return xchg(&page
->_last_nid
, nid
);
685 static inline int page_nid_last(struct page
*page
)
687 return page
->_last_nid
;
689 static inline void page_nid_reset_last(struct page
*page
)
691 page
->_last_nid
= -1;
694 static inline int page_nid_last(struct page
*page
)
696 return (page
->flags
>> LAST_NID_PGSHIFT
) & LAST_NID_MASK
;
699 extern int page_nid_xchg_last(struct page
*page
, int nid
);
701 static inline void page_nid_reset_last(struct page
*page
)
703 int nid
= (1 << LAST_NID_SHIFT
) - 1;
705 page
->flags
&= ~(LAST_NID_MASK
<< LAST_NID_PGSHIFT
);
706 page
->flags
|= (nid
& LAST_NID_MASK
) << LAST_NID_PGSHIFT
;
708 #endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
710 static inline int page_nid_xchg_last(struct page
*page
, int nid
)
712 return page_to_nid(page
);
715 static inline int page_nid_last(struct page
*page
)
717 return page_to_nid(page
);
720 static inline void page_nid_reset_last(struct page
*page
)
725 static inline struct zone
*page_zone(const struct page
*page
)
727 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
730 #ifdef SECTION_IN_PAGE_FLAGS
731 static inline void set_page_section(struct page
*page
, unsigned long section
)
733 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
734 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
737 static inline unsigned long page_to_section(const struct page
*page
)
739 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
743 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
745 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
746 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
749 static inline void set_page_node(struct page
*page
, unsigned long node
)
751 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
752 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
755 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
756 unsigned long node
, unsigned long pfn
)
758 set_page_zone(page
, zone
);
759 set_page_node(page
, node
);
760 #ifdef SECTION_IN_PAGE_FLAGS
761 set_page_section(page
, pfn_to_section_nr(pfn
));
766 * Some inline functions in vmstat.h depend on page_zone()
768 #include <linux/vmstat.h>
770 static __always_inline
void *lowmem_page_address(const struct page
*page
)
772 return __va(PFN_PHYS(page_to_pfn(page
)));
775 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
776 #define HASHED_PAGE_VIRTUAL
779 #if defined(WANT_PAGE_VIRTUAL)
780 static inline void *page_address(const struct page
*page
)
782 return page
->virtual;
784 static inline void set_page_address(struct page
*page
, void *address
)
786 page
->virtual = address
;
788 #define page_address_init() do { } while(0)
791 #if defined(HASHED_PAGE_VIRTUAL)
792 void *page_address(const struct page
*page
);
793 void set_page_address(struct page
*page
, void *virtual);
794 void page_address_init(void);
797 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
798 #define page_address(page) lowmem_page_address(page)
799 #define set_page_address(page, address) do { } while(0)
800 #define page_address_init() do { } while(0)
804 * On an anonymous page mapped into a user virtual memory area,
805 * page->mapping points to its anon_vma, not to a struct address_space;
806 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
808 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
809 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
810 * and then page->mapping points, not to an anon_vma, but to a private
811 * structure which KSM associates with that merged page. See ksm.h.
813 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
815 * Please note that, confusingly, "page_mapping" refers to the inode
816 * address_space which maps the page from disk; whereas "page_mapped"
817 * refers to user virtual address space into which the page is mapped.
819 #define PAGE_MAPPING_ANON 1
820 #define PAGE_MAPPING_KSM 2
821 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
823 extern struct address_space
*page_mapping(struct page
*page
);
825 /* Neutral page->mapping pointer to address_space or anon_vma or other */
826 static inline void *page_rmapping(struct page
*page
)
828 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
831 extern struct address_space
*__page_file_mapping(struct page
*);
834 struct address_space
*page_file_mapping(struct page
*page
)
836 if (unlikely(PageSwapCache(page
)))
837 return __page_file_mapping(page
);
839 return page
->mapping
;
842 static inline int PageAnon(struct page
*page
)
844 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
848 * Return the pagecache index of the passed page. Regular pagecache pages
849 * use ->index whereas swapcache pages use ->private
851 static inline pgoff_t
page_index(struct page
*page
)
853 if (unlikely(PageSwapCache(page
)))
854 return page_private(page
);
858 extern pgoff_t
__page_file_index(struct page
*page
);
861 * Return the file index of the page. Regular pagecache pages use ->index
862 * whereas swapcache pages use swp_offset(->private)
864 static inline pgoff_t
page_file_index(struct page
*page
)
866 if (unlikely(PageSwapCache(page
)))
867 return __page_file_index(page
);
873 * Return true if this page is mapped into pagetables.
875 static inline int page_mapped(struct page
*page
)
877 return atomic_read(&(page
)->_mapcount
) >= 0;
881 * Different kinds of faults, as returned by handle_mm_fault().
882 * Used to decide whether a process gets delivered SIGBUS or
883 * just gets major/minor fault counters bumped up.
886 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
888 #define VM_FAULT_OOM 0x0001
889 #define VM_FAULT_SIGBUS 0x0002
890 #define VM_FAULT_MAJOR 0x0004
891 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
892 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
893 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
894 #define VM_FAULT_SIGSEGV 0x0040
896 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
897 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
898 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
900 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
902 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
903 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)
905 /* Encode hstate index for a hwpoisoned large page */
906 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
907 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
910 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
912 extern void pagefault_out_of_memory(void);
914 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
917 * Flags passed to show_mem() and show_free_areas() to suppress output in
920 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
921 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
923 extern void show_free_areas(unsigned int flags
);
924 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
926 int shmem_zero_setup(struct vm_area_struct
*);
928 extern int can_do_mlock(void);
929 extern int user_shm_lock(size_t, struct user_struct
*);
930 extern void user_shm_unlock(size_t, struct user_struct
*);
933 * Parameter block passed down to zap_pte_range in exceptional cases.
936 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
937 struct address_space
*check_mapping
; /* Check page->mapping if set */
938 pgoff_t first_index
; /* Lowest page->index to unmap */
939 pgoff_t last_index
; /* Highest page->index to unmap */
942 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
945 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
947 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
948 unsigned long size
, struct zap_details
*);
949 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
950 unsigned long start
, unsigned long end
);
953 * mm_walk - callbacks for walk_page_range
954 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
955 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
956 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
957 * this handler is required to be able to handle
958 * pmd_trans_huge() pmds. They may simply choose to
959 * split_huge_page() instead of handling it explicitly.
960 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
961 * @pte_hole: if set, called for each hole at all levels
962 * @hugetlb_entry: if set, called for each hugetlb entry
963 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
966 * (see walk_page_range for more details)
969 int (*pgd_entry
)(pgd_t
*pgd
, unsigned long addr
,
970 unsigned long next
, struct mm_walk
*walk
);
971 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
972 unsigned long next
, struct mm_walk
*walk
);
973 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
974 unsigned long next
, struct mm_walk
*walk
);
975 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
976 unsigned long next
, struct mm_walk
*walk
);
977 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
978 struct mm_walk
*walk
);
979 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
980 unsigned long addr
, unsigned long next
,
981 struct mm_walk
*walk
);
982 struct mm_struct
*mm
;
986 int walk_page_range(unsigned long addr
, unsigned long end
,
987 struct mm_walk
*walk
);
988 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
989 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
990 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
991 struct vm_area_struct
*vma
);
992 void unmap_mapping_range(struct address_space
*mapping
,
993 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
994 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
996 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
997 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
998 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
999 void *buf
, int len
, int write
);
1001 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1002 loff_t
const holebegin
, loff_t
const holelen
)
1004 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1007 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
1008 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1009 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1010 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1011 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1012 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1013 int invalidate_inode_page(struct page
*page
);
1016 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1017 unsigned long address
, unsigned int flags
);
1018 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1019 unsigned long address
, unsigned int fault_flags
);
1021 static inline int handle_mm_fault(struct mm_struct
*mm
,
1022 struct vm_area_struct
*vma
, unsigned long address
,
1025 /* should never happen if there's no MMU */
1027 return VM_FAULT_SIGBUS
;
1029 static inline int fixup_user_fault(struct task_struct
*tsk
,
1030 struct mm_struct
*mm
, unsigned long address
,
1031 unsigned int fault_flags
)
1033 /* should never happen if there's no MMU */
1039 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1040 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1041 void *buf
, int len
, int write
);
1043 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1044 unsigned long start
, unsigned long nr_pages
,
1045 unsigned int foll_flags
, struct page
**pages
,
1046 struct vm_area_struct
**vmas
, int *nonblocking
);
1047 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1048 unsigned long start
, unsigned long nr_pages
,
1049 int write
, int force
, struct page
**pages
,
1050 struct vm_area_struct
**vmas
);
1051 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1052 struct page
**pages
);
1054 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1055 struct page
**pages
);
1056 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1057 struct page
*get_dump_page(unsigned long addr
);
1059 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1060 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
1062 int __set_page_dirty_nobuffers(struct page
*page
);
1063 int __set_page_dirty_no_writeback(struct page
*page
);
1064 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1066 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1067 void account_page_writeback(struct page
*page
);
1068 int set_page_dirty(struct page
*page
);
1069 int set_page_dirty_lock(struct page
*page
);
1070 int clear_page_dirty_for_io(struct page
*page
);
1072 /* Is the vma a continuation of the stack vma above it? */
1073 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1075 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1078 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1081 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1082 (vma
->vm_start
== addr
) &&
1083 !vma_growsdown(vma
->vm_prev
, addr
);
1086 /* Is the vma a continuation of the stack vma below it? */
1087 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1089 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1092 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1095 return (vma
->vm_flags
& VM_GROWSUP
) &&
1096 (vma
->vm_end
== addr
) &&
1097 !vma_growsup(vma
->vm_next
, addr
);
1101 vm_is_stack(struct task_struct
*task
, struct vm_area_struct
*vma
, int in_group
);
1103 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1104 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1105 unsigned long new_addr
, unsigned long len
,
1106 bool need_rmap_locks
);
1107 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1108 unsigned long end
, pgprot_t newprot
,
1109 int dirty_accountable
, int prot_numa
);
1110 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1111 struct vm_area_struct
**pprev
, unsigned long start
,
1112 unsigned long end
, unsigned long newflags
);
1115 * doesn't attempt to fault and will return short.
1117 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1118 struct page
**pages
);
1120 * per-process(per-mm_struct) statistics.
1122 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1124 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1126 #ifdef SPLIT_RSS_COUNTING
1128 * counter is updated in asynchronous manner and may go to minus.
1129 * But it's never be expected number for users.
1134 return (unsigned long)val
;
1137 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1139 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1142 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1144 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1147 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1149 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1152 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1154 return get_mm_counter(mm
, MM_FILEPAGES
) +
1155 get_mm_counter(mm
, MM_ANONPAGES
);
1158 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1160 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1163 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1165 return max(mm
->hiwater_vm
, mm
->total_vm
);
1168 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1170 unsigned long _rss
= get_mm_rss(mm
);
1172 if ((mm
)->hiwater_rss
< _rss
)
1173 (mm
)->hiwater_rss
= _rss
;
1176 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1178 if (mm
->hiwater_vm
< mm
->total_vm
)
1179 mm
->hiwater_vm
= mm
->total_vm
;
1182 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1183 struct mm_struct
*mm
)
1185 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1187 if (*maxrss
< hiwater_rss
)
1188 *maxrss
= hiwater_rss
;
1191 #if defined(SPLIT_RSS_COUNTING)
1192 void sync_mm_rss(struct mm_struct
*mm
);
1194 static inline void sync_mm_rss(struct mm_struct
*mm
)
1199 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1201 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1203 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1207 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1211 #ifdef __PAGETABLE_PUD_FOLDED
1212 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1213 unsigned long address
)
1218 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1221 #ifdef __PAGETABLE_PMD_FOLDED
1222 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1223 unsigned long address
)
1228 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1231 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1232 pmd_t
*pmd
, unsigned long address
);
1233 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1236 * The following ifdef needed to get the 4level-fixup.h header to work.
1237 * Remove it when 4level-fixup.h has been removed.
1239 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1240 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1242 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1243 NULL
: pud_offset(pgd
, address
);
1246 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1248 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1249 NULL
: pmd_offset(pud
, address
);
1251 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1253 #if USE_SPLIT_PTLOCKS
1255 * We tuck a spinlock to guard each pagetable page into its struct page,
1256 * at page->private, with BUILD_BUG_ON to make sure that this will not
1257 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1258 * When freeing, reset page->mapping so free_pages_check won't complain.
1260 #define __pte_lockptr(page) &((page)->ptl)
1261 #define pte_lock_init(_page) do { \
1262 spin_lock_init(__pte_lockptr(_page)); \
1264 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1265 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1266 #else /* !USE_SPLIT_PTLOCKS */
1268 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1270 #define pte_lock_init(page) do {} while (0)
1271 #define pte_lock_deinit(page) do {} while (0)
1272 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1273 #endif /* USE_SPLIT_PTLOCKS */
1275 static inline void pgtable_page_ctor(struct page
*page
)
1277 pte_lock_init(page
);
1278 inc_zone_page_state(page
, NR_PAGETABLE
);
1281 static inline void pgtable_page_dtor(struct page
*page
)
1283 pte_lock_deinit(page
);
1284 dec_zone_page_state(page
, NR_PAGETABLE
);
1287 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1289 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1290 pte_t *__pte = pte_offset_map(pmd, address); \
1296 #define pte_unmap_unlock(pte, ptl) do { \
1301 #define pte_alloc_map(mm, vma, pmd, address) \
1302 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1304 NULL: pte_offset_map(pmd, address))
1306 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1307 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1309 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1311 #define pte_alloc_kernel(pmd, address) \
1312 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1313 NULL: pte_offset_kernel(pmd, address))
1315 extern void free_area_init(unsigned long * zones_size
);
1316 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1317 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1318 extern void free_initmem(void);
1321 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1322 * into the buddy system. The freed pages will be poisoned with pattern
1323 * "poison" if it's non-zero.
1324 * Return pages freed into the buddy system.
1326 extern unsigned long free_reserved_area(unsigned long start
, unsigned long end
,
1327 int poison
, char *s
);
1328 #ifdef CONFIG_HIGHMEM
1330 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1331 * and totalram_pages.
1333 extern void free_highmem_page(struct page
*page
);
1336 static inline void adjust_managed_page_count(struct page
*page
, long count
)
1338 totalram_pages
+= count
;
1341 /* Free the reserved page into the buddy system, so it gets managed. */
1342 static inline void __free_reserved_page(struct page
*page
)
1344 ClearPageReserved(page
);
1345 init_page_count(page
);
1349 static inline void free_reserved_page(struct page
*page
)
1351 __free_reserved_page(page
);
1352 adjust_managed_page_count(page
, 1);
1355 static inline void mark_page_reserved(struct page
*page
)
1357 SetPageReserved(page
);
1358 adjust_managed_page_count(page
, -1);
1362 * Default method to free all the __init memory into the buddy system.
1363 * The freed pages will be poisoned with pattern "poison" if it is
1364 * non-zero. Return pages freed into the buddy system.
1366 static inline unsigned long free_initmem_default(int poison
)
1368 extern char __init_begin
[], __init_end
[];
1370 return free_reserved_area(PAGE_ALIGN((unsigned long)&__init_begin
) ,
1371 ((unsigned long)&__init_end
) & PAGE_MASK
,
1372 poison
, "unused kernel");
1375 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1377 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1378 * zones, allocate the backing mem_map and account for memory holes in a more
1379 * architecture independent manner. This is a substitute for creating the
1380 * zone_sizes[] and zholes_size[] arrays and passing them to
1381 * free_area_init_node()
1383 * An architecture is expected to register range of page frames backed by
1384 * physical memory with memblock_add[_node]() before calling
1385 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1386 * usage, an architecture is expected to do something like
1388 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1390 * for_each_valid_physical_page_range()
1391 * memblock_add_node(base, size, nid)
1392 * free_area_init_nodes(max_zone_pfns);
1394 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1395 * registered physical page range. Similarly
1396 * sparse_memory_present_with_active_regions() calls memory_present() for
1397 * each range when SPARSEMEM is enabled.
1399 * See mm/page_alloc.c for more information on each function exposed by
1400 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1402 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1403 unsigned long node_map_pfn_alignment(void);
1404 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1405 unsigned long end_pfn
);
1406 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1407 unsigned long end_pfn
);
1408 extern void get_pfn_range_for_nid(unsigned int nid
,
1409 unsigned long *start_pfn
, unsigned long *end_pfn
);
1410 extern unsigned long find_min_pfn_with_active_regions(void);
1411 extern void free_bootmem_with_active_regions(int nid
,
1412 unsigned long max_low_pfn
);
1413 extern void sparse_memory_present_with_active_regions(int nid
);
1415 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1417 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1418 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1419 static inline int __early_pfn_to_nid(unsigned long pfn
)
1424 /* please see mm/page_alloc.c */
1425 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1426 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1427 /* there is a per-arch backend function. */
1428 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1429 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1432 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1433 extern void memmap_init_zone(unsigned long, int, unsigned long,
1434 unsigned long, enum memmap_context
);
1435 extern void setup_per_zone_wmarks(void);
1436 extern int __meminit
init_per_zone_wmark_min(void);
1437 extern void mem_init(void);
1438 extern void __init
mmap_init(void);
1439 extern void show_mem(unsigned int flags
);
1440 extern void si_meminfo(struct sysinfo
* val
);
1441 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1443 extern __printf(3, 4)
1444 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1446 extern void setup_per_cpu_pageset(void);
1448 extern void zone_pcp_update(struct zone
*zone
);
1449 extern void zone_pcp_reset(struct zone
*zone
);
1452 extern int min_free_kbytes
;
1455 extern atomic_long_t mmap_pages_allocated
;
1456 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1458 /* interval_tree.c */
1459 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1460 struct rb_root
*root
);
1461 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1462 struct vm_area_struct
*prev
,
1463 struct rb_root
*root
);
1464 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1465 struct rb_root
*root
);
1466 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1467 unsigned long start
, unsigned long last
);
1468 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1469 unsigned long start
, unsigned long last
);
1471 #define vma_interval_tree_foreach(vma, root, start, last) \
1472 for (vma = vma_interval_tree_iter_first(root, start, last); \
1473 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1475 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1476 struct list_head
*list
)
1478 list_add_tail(&vma
->shared
.nonlinear
, list
);
1481 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1482 struct rb_root
*root
);
1483 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1484 struct rb_root
*root
);
1485 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1486 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1487 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1488 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1489 #ifdef CONFIG_DEBUG_VM_RB
1490 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1493 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1494 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1495 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1498 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1499 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1500 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1501 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1502 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1503 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1504 struct mempolicy
*);
1505 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1506 extern int split_vma(struct mm_struct
*,
1507 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1508 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1509 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1510 struct rb_node
**, struct rb_node
*);
1511 extern void unlink_file_vma(struct vm_area_struct
*);
1512 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1513 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1514 bool *need_rmap_locks
);
1515 extern void exit_mmap(struct mm_struct
*);
1517 extern int mm_take_all_locks(struct mm_struct
*mm
);
1518 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1520 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1521 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1523 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1524 extern int install_special_mapping(struct mm_struct
*mm
,
1525 unsigned long addr
, unsigned long len
,
1526 unsigned long flags
, struct page
**pages
);
1528 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1530 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1531 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1532 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1533 unsigned long len
, unsigned long prot
, unsigned long flags
,
1534 unsigned long pgoff
, unsigned long *populate
);
1535 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1538 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1540 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1543 (void) __mm_populate(addr
, len
, 1);
1546 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1549 /* These take the mm semaphore themselves */
1550 extern unsigned long vm_brk(unsigned long, unsigned long);
1551 extern int vm_munmap(unsigned long, size_t);
1552 extern unsigned long vm_mmap(struct file
*, unsigned long,
1553 unsigned long, unsigned long,
1554 unsigned long, unsigned long);
1556 struct vm_unmapped_area_info
{
1557 #define VM_UNMAPPED_AREA_TOPDOWN 1
1558 unsigned long flags
;
1559 unsigned long length
;
1560 unsigned long low_limit
;
1561 unsigned long high_limit
;
1562 unsigned long align_mask
;
1563 unsigned long align_offset
;
1566 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1567 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1570 * Search for an unmapped address range.
1572 * We are looking for a range that:
1573 * - does not intersect with any VMA;
1574 * - is contained within the [low_limit, high_limit) interval;
1575 * - is at least the desired size.
1576 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1578 static inline unsigned long
1579 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1581 if (!(info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
))
1582 return unmapped_area(info
);
1584 return unmapped_area_topdown(info
);
1588 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1589 extern void truncate_inode_pages_range(struct address_space
*,
1590 loff_t lstart
, loff_t lend
);
1592 /* generic vm_area_ops exported for stackable file systems */
1593 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1594 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1596 /* mm/page-writeback.c */
1597 int write_one_page(struct page
*page
, int wait
);
1598 void task_dirty_inc(struct task_struct
*tsk
);
1601 #define VM_MAX_READAHEAD 128 /* kbytes */
1602 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1604 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1605 pgoff_t offset
, unsigned long nr_to_read
);
1607 void page_cache_sync_readahead(struct address_space
*mapping
,
1608 struct file_ra_state
*ra
,
1611 unsigned long size
);
1613 void page_cache_async_readahead(struct address_space
*mapping
,
1614 struct file_ra_state
*ra
,
1618 unsigned long size
);
1620 unsigned long max_sane_readahead(unsigned long nr
);
1621 unsigned long ra_submit(struct file_ra_state
*ra
,
1622 struct address_space
*mapping
,
1625 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1626 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1628 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1629 extern int expand_downwards(struct vm_area_struct
*vma
,
1630 unsigned long address
);
1632 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1634 #define expand_upwards(vma, address) (0)
1637 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1638 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1639 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1640 struct vm_area_struct
**pprev
);
1642 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1643 NULL if none. Assume start_addr < end_addr. */
1644 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1646 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1648 if (vma
&& end_addr
<= vma
->vm_start
)
1653 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1655 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1658 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1659 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
1660 unsigned long vm_start
, unsigned long vm_end
)
1662 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
1664 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
1671 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1673 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1679 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1680 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
1681 unsigned long start
, unsigned long end
);
1684 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1685 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1686 unsigned long pfn
, unsigned long size
, pgprot_t
);
1687 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1688 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1690 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1692 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
1695 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
1696 unsigned long address
, unsigned int foll_flags
,
1697 unsigned int *page_mask
);
1699 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
1700 unsigned long address
, unsigned int foll_flags
)
1702 unsigned int unused_page_mask
;
1703 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
1706 #define FOLL_WRITE 0x01 /* check pte is writable */
1707 #define FOLL_TOUCH 0x02 /* mark page accessed */
1708 #define FOLL_GET 0x04 /* do get_page on page */
1709 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1710 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1711 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1712 * and return without waiting upon it */
1713 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1714 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1715 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1716 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1717 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1719 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1721 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1722 unsigned long size
, pte_fn_t fn
, void *data
);
1724 #ifdef CONFIG_PROC_FS
1725 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1727 static inline void vm_stat_account(struct mm_struct
*mm
,
1728 unsigned long flags
, struct file
*file
, long pages
)
1730 mm
->total_vm
+= pages
;
1732 #endif /* CONFIG_PROC_FS */
1734 #ifdef CONFIG_DEBUG_PAGEALLOC
1735 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1736 #ifdef CONFIG_HIBERNATION
1737 extern bool kernel_page_present(struct page
*page
);
1738 #endif /* CONFIG_HIBERNATION */
1741 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1742 #ifdef CONFIG_HIBERNATION
1743 static inline bool kernel_page_present(struct page
*page
) { return true; }
1744 #endif /* CONFIG_HIBERNATION */
1747 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
1748 #ifdef __HAVE_ARCH_GATE_AREA
1749 int in_gate_area_no_mm(unsigned long addr
);
1750 int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
1752 int in_gate_area_no_mm(unsigned long addr
);
1753 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1754 #endif /* __HAVE_ARCH_GATE_AREA */
1756 #ifdef CONFIG_SYSCTL
1757 extern int sysctl_drop_caches
;
1758 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1759 void __user
*, size_t *, loff_t
*);
1762 unsigned long shrink_slab(struct shrink_control
*shrink
,
1763 unsigned long nr_pages_scanned
,
1764 unsigned long lru_pages
);
1767 #define randomize_va_space 0
1769 extern int randomize_va_space
;
1772 const char * arch_vma_name(struct vm_area_struct
*vma
);
1773 void print_vma_addr(char *prefix
, unsigned long rip
);
1775 void sparse_mem_maps_populate_node(struct page
**map_map
,
1776 unsigned long pnum_begin
,
1777 unsigned long pnum_end
,
1778 unsigned long map_count
,
1781 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1782 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1783 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1784 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1785 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1786 void *vmemmap_alloc_block(unsigned long size
, int node
);
1787 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1788 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1789 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
1791 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
1792 void vmemmap_populate_print_last(void);
1793 #ifdef CONFIG_MEMORY_HOTPLUG
1794 void vmemmap_free(unsigned long start
, unsigned long end
);
1796 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
1797 unsigned long size
);
1800 MF_COUNT_INCREASED
= 1 << 0,
1801 MF_ACTION_REQUIRED
= 1 << 1,
1802 MF_MUST_KILL
= 1 << 2,
1804 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
1805 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
1806 extern int unpoison_memory(unsigned long pfn
);
1807 extern int sysctl_memory_failure_early_kill
;
1808 extern int sysctl_memory_failure_recovery
;
1809 extern void shake_page(struct page
*p
, int access
);
1810 extern atomic_long_t num_poisoned_pages
;
1811 extern int soft_offline_page(struct page
*page
, int flags
);
1813 extern void dump_page(struct page
*page
);
1815 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1816 extern void clear_huge_page(struct page
*page
,
1818 unsigned int pages_per_huge_page
);
1819 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1820 unsigned long addr
, struct vm_area_struct
*vma
,
1821 unsigned int pages_per_huge_page
);
1822 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1824 #ifdef CONFIG_DEBUG_PAGEALLOC
1825 extern unsigned int _debug_guardpage_minorder
;
1827 static inline unsigned int debug_guardpage_minorder(void)
1829 return _debug_guardpage_minorder
;
1832 static inline bool page_is_guard(struct page
*page
)
1834 return test_bit(PAGE_DEBUG_FLAG_GUARD
, &page
->debug_flags
);
1837 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1838 static inline bool page_is_guard(struct page
*page
) { return false; }
1839 #endif /* CONFIG_DEBUG_PAGEALLOC */
1841 #if MAX_NUMNODES > 1
1842 void __init
setup_nr_node_ids(void);
1844 static inline void setup_nr_node_ids(void) {}
1847 #endif /* __KERNEL__ */
1848 #endif /* _LINUX_MM_H */