4 #include <linux/errno.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/range.h>
16 #include <linux/pfn.h>
17 #include <linux/bit_spinlock.h>
23 struct writeback_control
;
25 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
26 extern unsigned long max_mapnr
;
29 extern unsigned long num_physpages
;
30 extern unsigned long totalram_pages
;
31 extern void * high_memory
;
32 extern int page_cluster
;
35 extern int sysctl_legacy_va_layout
;
37 #define sysctl_legacy_va_layout 0
41 #include <asm/pgtable.h>
42 #include <asm/processor.h>
44 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
46 /* to align the pointer to the (next) page boundary */
47 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
50 * Linux kernel virtual memory manager primitives.
51 * The idea being to have a "virtual" mm in the same way
52 * we have a virtual fs - giving a cleaner interface to the
53 * mm details, and allowing different kinds of memory mappings
54 * (from shared memory to executable loading to arbitrary
58 extern struct kmem_cache
*vm_area_cachep
;
61 extern struct rb_root nommu_region_tree
;
62 extern struct rw_semaphore nommu_region_sem
;
64 extern unsigned int kobjsize(const void *objp
);
68 * vm_flags in vm_area_struct, see mm_types.h.
70 #define VM_READ 0x00000001 /* currently active flags */
71 #define VM_WRITE 0x00000002
72 #define VM_EXEC 0x00000004
73 #define VM_SHARED 0x00000008
75 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
76 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
77 #define VM_MAYWRITE 0x00000020
78 #define VM_MAYEXEC 0x00000040
79 #define VM_MAYSHARE 0x00000080
81 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
82 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
83 #define VM_GROWSUP 0x00000200
85 #define VM_GROWSUP 0x00000000
86 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
88 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
89 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
91 #define VM_EXECUTABLE 0x00001000
92 #define VM_LOCKED 0x00002000
93 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
95 /* Used by sys_madvise() */
96 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
97 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
99 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
100 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
101 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
102 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
103 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
104 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
105 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
106 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
107 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
109 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
111 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
112 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
114 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
115 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
116 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
117 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
118 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
120 /* Bits set in the VMA until the stack is in its final location */
121 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
123 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
124 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
127 #ifdef CONFIG_STACK_GROWSUP
128 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
130 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
133 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
134 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
135 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
136 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
137 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
140 * special vmas that are non-mergable, non-mlock()able
142 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
145 * mapping from the currently active vm_flags protection bits (the
146 * low four bits) to a page protection mask..
148 extern pgprot_t protection_map
[16];
150 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
151 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
152 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
153 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
156 * This interface is used by x86 PAT code to identify a pfn mapping that is
157 * linear over entire vma. This is to optimize PAT code that deals with
158 * marking the physical region with a particular prot. This is not for generic
159 * mm use. Note also that this check will not work if the pfn mapping is
160 * linear for a vma starting at physical address 0. In which case PAT code
161 * falls back to slow path of reserving physical range page by page.
163 static inline int is_linear_pfn_mapping(struct vm_area_struct
*vma
)
165 return (vma
->vm_flags
& VM_PFN_AT_MMAP
);
168 static inline int is_pfn_mapping(struct vm_area_struct
*vma
)
170 return (vma
->vm_flags
& VM_PFNMAP
);
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 set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
183 unsigned int flags
; /* FAULT_FLAG_xxx flags */
184 pgoff_t pgoff
; /* Logical page offset based on vma */
185 void __user
*virtual_address
; /* Faulting virtual address */
187 struct page
*page
; /* ->fault handlers should return a
188 * page here, unless VM_FAULT_NOPAGE
189 * is set (which is also implied by
195 * These are the virtual MM functions - opening of an area, closing and
196 * unmapping it (needed to keep files on disk up-to-date etc), pointer
197 * to the functions called when a no-page or a wp-page exception occurs.
199 struct vm_operations_struct
{
200 void (*open
)(struct vm_area_struct
* area
);
201 void (*close
)(struct vm_area_struct
* area
);
202 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
204 /* notification that a previously read-only page is about to become
205 * writable, if an error is returned it will cause a SIGBUS */
206 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
208 /* called by access_process_vm when get_user_pages() fails, typically
209 * for use by special VMAs that can switch between memory and hardware
211 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
212 void *buf
, int len
, int write
);
215 * set_policy() op must add a reference to any non-NULL @new mempolicy
216 * to hold the policy upon return. Caller should pass NULL @new to
217 * remove a policy and fall back to surrounding context--i.e. do not
218 * install a MPOL_DEFAULT policy, nor the task or system default
221 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
224 * get_policy() op must add reference [mpol_get()] to any policy at
225 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
226 * in mm/mempolicy.c will do this automatically.
227 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
228 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
229 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
230 * must return NULL--i.e., do not "fallback" to task or system default
233 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
235 int (*migrate
)(struct vm_area_struct
*vma
, const nodemask_t
*from
,
236 const nodemask_t
*to
, unsigned long flags
);
243 #define page_private(page) ((page)->private)
244 #define set_page_private(page, v) ((page)->private = (v))
247 * FIXME: take this include out, include page-flags.h in
248 * files which need it (119 of them)
250 #include <linux/page-flags.h>
251 #include <linux/huge_mm.h>
254 * Methods to modify the page usage count.
256 * What counts for a page usage:
257 * - cache mapping (page->mapping)
258 * - private data (page->private)
259 * - page mapped in a task's page tables, each mapping
260 * is counted separately
262 * Also, many kernel routines increase the page count before a critical
263 * routine so they can be sure the page doesn't go away from under them.
267 * Drop a ref, return true if the refcount fell to zero (the page has no users)
269 static inline int put_page_testzero(struct page
*page
)
271 VM_BUG_ON(atomic_read(&page
->_count
) == 0);
272 return atomic_dec_and_test(&page
->_count
);
276 * Try to grab a ref unless the page has a refcount of zero, return false if
279 static inline int get_page_unless_zero(struct page
*page
)
281 return atomic_inc_not_zero(&page
->_count
);
284 extern int page_is_ram(unsigned long pfn
);
286 /* Support for virtually mapped pages */
287 struct page
*vmalloc_to_page(const void *addr
);
288 unsigned long vmalloc_to_pfn(const void *addr
);
291 * Determine if an address is within the vmalloc range
293 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
294 * is no special casing required.
296 static inline int is_vmalloc_addr(const void *x
)
299 unsigned long addr
= (unsigned long)x
;
301 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
307 extern int is_vmalloc_or_module_addr(const void *x
);
309 static inline int is_vmalloc_or_module_addr(const void *x
)
315 static inline void compound_lock(struct page
*page
)
317 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
318 bit_spin_lock(PG_compound_lock
, &page
->flags
);
322 static inline void compound_unlock(struct page
*page
)
324 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
325 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
329 static inline unsigned long compound_lock_irqsave(struct page
*page
)
331 unsigned long uninitialized_var(flags
);
332 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
333 local_irq_save(flags
);
339 static inline void compound_unlock_irqrestore(struct page
*page
,
342 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
343 compound_unlock(page
);
344 local_irq_restore(flags
);
348 static inline struct page
*compound_head(struct page
*page
)
350 if (unlikely(PageTail(page
)))
351 return page
->first_page
;
355 static inline int page_count(struct page
*page
)
357 return atomic_read(&compound_head(page
)->_count
);
360 static inline void get_page(struct page
*page
)
363 * Getting a normal page or the head of a compound page
364 * requires to already have an elevated page->_count. Only if
365 * we're getting a tail page, the elevated page->_count is
366 * required only in the head page, so for tail pages the
367 * bugcheck only verifies that the page->_count isn't
370 VM_BUG_ON(atomic_read(&page
->_count
) < !PageTail(page
));
371 atomic_inc(&page
->_count
);
373 * Getting a tail page will elevate both the head and tail
376 if (unlikely(PageTail(page
))) {
378 * This is safe only because
379 * __split_huge_page_refcount can't run under
382 VM_BUG_ON(atomic_read(&page
->first_page
->_count
) <= 0);
383 atomic_inc(&page
->first_page
->_count
);
387 static inline struct page
*virt_to_head_page(const void *x
)
389 struct page
*page
= virt_to_page(x
);
390 return compound_head(page
);
394 * Setup the page count before being freed into the page allocator for
395 * the first time (boot or memory hotplug)
397 static inline void init_page_count(struct page
*page
)
399 atomic_set(&page
->_count
, 1);
403 * PageBuddy() indicate that the page is free and in the buddy system
404 * (see mm/page_alloc.c).
406 static inline int PageBuddy(struct page
*page
)
408 return atomic_read(&page
->_mapcount
) == -2;
411 static inline void __SetPageBuddy(struct page
*page
)
413 VM_BUG_ON(atomic_read(&page
->_mapcount
) != -1);
414 atomic_set(&page
->_mapcount
, -2);
417 static inline void __ClearPageBuddy(struct page
*page
)
419 VM_BUG_ON(!PageBuddy(page
));
420 atomic_set(&page
->_mapcount
, -1);
423 void put_page(struct page
*page
);
424 void put_pages_list(struct list_head
*pages
);
426 void split_page(struct page
*page
, unsigned int order
);
427 int split_free_page(struct page
*page
);
430 * Compound pages have a destructor function. Provide a
431 * prototype for that function and accessor functions.
432 * These are _only_ valid on the head of a PG_compound page.
434 typedef void compound_page_dtor(struct page
*);
436 static inline void set_compound_page_dtor(struct page
*page
,
437 compound_page_dtor
*dtor
)
439 page
[1].lru
.next
= (void *)dtor
;
442 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
444 return (compound_page_dtor
*)page
[1].lru
.next
;
447 static inline int compound_order(struct page
*page
)
451 return (unsigned long)page
[1].lru
.prev
;
454 static inline int compound_trans_order(struct page
*page
)
462 flags
= compound_lock_irqsave(page
);
463 order
= compound_order(page
);
464 compound_unlock_irqrestore(page
, flags
);
468 static inline void set_compound_order(struct page
*page
, unsigned long order
)
470 page
[1].lru
.prev
= (void *)order
;
474 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
475 * servicing faults for write access. In the normal case, do always want
476 * pte_mkwrite. But get_user_pages can cause write faults for mappings
477 * that do not have writing enabled, when used by access_process_vm.
479 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
481 if (likely(vma
->vm_flags
& VM_WRITE
))
482 pte
= pte_mkwrite(pte
);
487 * Multiple processes may "see" the same page. E.g. for untouched
488 * mappings of /dev/null, all processes see the same page full of
489 * zeroes, and text pages of executables and shared libraries have
490 * only one copy in memory, at most, normally.
492 * For the non-reserved pages, page_count(page) denotes a reference count.
493 * page_count() == 0 means the page is free. page->lru is then used for
494 * freelist management in the buddy allocator.
495 * page_count() > 0 means the page has been allocated.
497 * Pages are allocated by the slab allocator in order to provide memory
498 * to kmalloc and kmem_cache_alloc. In this case, the management of the
499 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
500 * unless a particular usage is carefully commented. (the responsibility of
501 * freeing the kmalloc memory is the caller's, of course).
503 * A page may be used by anyone else who does a __get_free_page().
504 * In this case, page_count still tracks the references, and should only
505 * be used through the normal accessor functions. The top bits of page->flags
506 * and page->virtual store page management information, but all other fields
507 * are unused and could be used privately, carefully. The management of this
508 * page is the responsibility of the one who allocated it, and those who have
509 * subsequently been given references to it.
511 * The other pages (we may call them "pagecache pages") are completely
512 * managed by the Linux memory manager: I/O, buffers, swapping etc.
513 * The following discussion applies only to them.
515 * A pagecache page contains an opaque `private' member, which belongs to the
516 * page's address_space. Usually, this is the address of a circular list of
517 * the page's disk buffers. PG_private must be set to tell the VM to call
518 * into the filesystem to release these pages.
520 * A page may belong to an inode's memory mapping. In this case, page->mapping
521 * is the pointer to the inode, and page->index is the file offset of the page,
522 * in units of PAGE_CACHE_SIZE.
524 * If pagecache pages are not associated with an inode, they are said to be
525 * anonymous pages. These may become associated with the swapcache, and in that
526 * case PG_swapcache is set, and page->private is an offset into the swapcache.
528 * In either case (swapcache or inode backed), the pagecache itself holds one
529 * reference to the page. Setting PG_private should also increment the
530 * refcount. The each user mapping also has a reference to the page.
532 * The pagecache pages are stored in a per-mapping radix tree, which is
533 * rooted at mapping->page_tree, and indexed by offset.
534 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
535 * lists, we instead now tag pages as dirty/writeback in the radix tree.
537 * All pagecache pages may be subject to I/O:
538 * - inode pages may need to be read from disk,
539 * - inode pages which have been modified and are MAP_SHARED may need
540 * to be written back to the inode on disk,
541 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
542 * modified may need to be swapped out to swap space and (later) to be read
547 * The zone field is never updated after free_area_init_core()
548 * sets it, so none of the operations on it need to be atomic.
553 * page->flags layout:
555 * There are three possibilities for how page->flags get
556 * laid out. The first is for the normal case, without
557 * sparsemem. The second is for sparsemem when there is
558 * plenty of space for node and section. The last is when
559 * we have run out of space and have to fall back to an
560 * alternate (slower) way of determining the node.
562 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
563 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
564 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
566 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
567 #define SECTIONS_WIDTH SECTIONS_SHIFT
569 #define SECTIONS_WIDTH 0
572 #define ZONES_WIDTH ZONES_SHIFT
574 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
575 #define NODES_WIDTH NODES_SHIFT
577 #ifdef CONFIG_SPARSEMEM_VMEMMAP
578 #error "Vmemmap: No space for nodes field in page flags"
580 #define NODES_WIDTH 0
583 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
584 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
585 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
586 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
589 * We are going to use the flags for the page to node mapping if its in
590 * there. This includes the case where there is no node, so it is implicit.
592 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
593 #define NODE_NOT_IN_PAGE_FLAGS
596 #ifndef PFN_SECTION_SHIFT
597 #define PFN_SECTION_SHIFT 0
601 * Define the bit shifts to access each section. For non-existant
602 * sections we define the shift as 0; that plus a 0 mask ensures
603 * the compiler will optimise away reference to them.
605 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
606 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
607 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
609 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
610 #ifdef NODE_NOT_IN_PAGE_FLAGS
611 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
612 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
613 SECTIONS_PGOFF : ZONES_PGOFF)
615 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
616 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
617 NODES_PGOFF : ZONES_PGOFF)
620 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
622 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
623 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
626 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
627 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
628 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
629 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
631 static inline enum zone_type
page_zonenum(struct page
*page
)
633 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
637 * The identification function is only used by the buddy allocator for
638 * determining if two pages could be buddies. We are not really
639 * identifying a zone since we could be using a the section number
640 * id if we have not node id available in page flags.
641 * We guarantee only that it will return the same value for two
642 * combinable pages in a zone.
644 static inline int page_zone_id(struct page
*page
)
646 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
649 static inline int zone_to_nid(struct zone
*zone
)
658 #ifdef NODE_NOT_IN_PAGE_FLAGS
659 extern int page_to_nid(struct page
*page
);
661 static inline int page_to_nid(struct page
*page
)
663 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
667 static inline struct zone
*page_zone(struct page
*page
)
669 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
672 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
673 static inline unsigned long page_to_section(struct page
*page
)
675 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
679 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
681 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
682 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
685 static inline void set_page_node(struct page
*page
, unsigned long node
)
687 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
688 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
691 static inline void set_page_section(struct page
*page
, unsigned long section
)
693 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
694 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
697 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
698 unsigned long node
, unsigned long pfn
)
700 set_page_zone(page
, zone
);
701 set_page_node(page
, node
);
702 set_page_section(page
, pfn_to_section_nr(pfn
));
706 * Some inline functions in vmstat.h depend on page_zone()
708 #include <linux/vmstat.h>
710 static __always_inline
void *lowmem_page_address(struct page
*page
)
712 return __va(PFN_PHYS(page_to_pfn(page
)));
715 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
716 #define HASHED_PAGE_VIRTUAL
719 #if defined(WANT_PAGE_VIRTUAL)
720 #define page_address(page) ((page)->virtual)
721 #define set_page_address(page, address) \
723 (page)->virtual = (address); \
725 #define page_address_init() do { } while(0)
728 #if defined(HASHED_PAGE_VIRTUAL)
729 void *page_address(struct page
*page
);
730 void set_page_address(struct page
*page
, void *virtual);
731 void page_address_init(void);
734 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
735 #define page_address(page) lowmem_page_address(page)
736 #define set_page_address(page, address) do { } while(0)
737 #define page_address_init() do { } while(0)
741 * On an anonymous page mapped into a user virtual memory area,
742 * page->mapping points to its anon_vma, not to a struct address_space;
743 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
745 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
746 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
747 * and then page->mapping points, not to an anon_vma, but to a private
748 * structure which KSM associates with that merged page. See ksm.h.
750 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
752 * Please note that, confusingly, "page_mapping" refers to the inode
753 * address_space which maps the page from disk; whereas "page_mapped"
754 * refers to user virtual address space into which the page is mapped.
756 #define PAGE_MAPPING_ANON 1
757 #define PAGE_MAPPING_KSM 2
758 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
760 extern struct address_space swapper_space
;
761 static inline struct address_space
*page_mapping(struct page
*page
)
763 struct address_space
*mapping
= page
->mapping
;
765 VM_BUG_ON(PageSlab(page
));
766 if (unlikely(PageSwapCache(page
)))
767 mapping
= &swapper_space
;
768 else if ((unsigned long)mapping
& PAGE_MAPPING_ANON
)
773 /* Neutral page->mapping pointer to address_space or anon_vma or other */
774 static inline void *page_rmapping(struct page
*page
)
776 return (void *)((unsigned long)page
->mapping
& ~PAGE_MAPPING_FLAGS
);
779 static inline int PageAnon(struct page
*page
)
781 return ((unsigned long)page
->mapping
& PAGE_MAPPING_ANON
) != 0;
785 * Return the pagecache index of the passed page. Regular pagecache pages
786 * use ->index whereas swapcache pages use ->private
788 static inline pgoff_t
page_index(struct page
*page
)
790 if (unlikely(PageSwapCache(page
)))
791 return page_private(page
);
796 * The atomic page->_mapcount, like _count, starts from -1:
797 * so that transitions both from it and to it can be tracked,
798 * using atomic_inc_and_test and atomic_add_negative(-1).
800 static inline void reset_page_mapcount(struct page
*page
)
802 atomic_set(&(page
)->_mapcount
, -1);
805 static inline int page_mapcount(struct page
*page
)
807 return atomic_read(&(page
)->_mapcount
) + 1;
811 * Return true if this page is mapped into pagetables.
813 static inline int page_mapped(struct page
*page
)
815 return atomic_read(&(page
)->_mapcount
) >= 0;
819 * Different kinds of faults, as returned by handle_mm_fault().
820 * Used to decide whether a process gets delivered SIGBUS or
821 * just gets major/minor fault counters bumped up.
824 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
826 #define VM_FAULT_OOM 0x0001
827 #define VM_FAULT_SIGBUS 0x0002
828 #define VM_FAULT_MAJOR 0x0004
829 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
830 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
831 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
833 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
834 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
835 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
837 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
839 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
840 VM_FAULT_HWPOISON_LARGE)
842 /* Encode hstate index for a hwpoisoned large page */
843 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
844 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
847 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
849 extern void pagefault_out_of_memory(void);
851 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
853 extern void show_free_areas(void);
855 int shmem_lock(struct file
*file
, int lock
, struct user_struct
*user
);
856 struct file
*shmem_file_setup(const char *name
, loff_t size
, unsigned long flags
);
857 int shmem_zero_setup(struct vm_area_struct
*);
860 extern unsigned long shmem_get_unmapped_area(struct file
*file
,
864 unsigned long flags
);
867 extern int can_do_mlock(void);
868 extern int user_shm_lock(size_t, struct user_struct
*);
869 extern void user_shm_unlock(size_t, struct user_struct
*);
872 * Parameter block passed down to zap_pte_range in exceptional cases.
875 struct vm_area_struct
*nonlinear_vma
; /* Check page->index if set */
876 struct address_space
*check_mapping
; /* Check page->mapping if set */
877 pgoff_t first_index
; /* Lowest page->index to unmap */
878 pgoff_t last_index
; /* Highest page->index to unmap */
879 spinlock_t
*i_mmap_lock
; /* For unmap_mapping_range: */
880 unsigned long truncate_count
; /* Compare vm_truncate_count */
883 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
886 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
888 unsigned long zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
889 unsigned long size
, struct zap_details
*);
890 unsigned long unmap_vmas(struct mmu_gather
**tlb
,
891 struct vm_area_struct
*start_vma
, unsigned long start_addr
,
892 unsigned long end_addr
, unsigned long *nr_accounted
,
893 struct zap_details
*);
896 * mm_walk - callbacks for walk_page_range
897 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
898 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
899 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
900 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
901 * @pte_hole: if set, called for each hole at all levels
902 * @hugetlb_entry: if set, called for each hugetlb entry
904 * (see walk_page_range for more details)
907 int (*pgd_entry
)(pgd_t
*, unsigned long, unsigned long, struct mm_walk
*);
908 int (*pud_entry
)(pud_t
*, unsigned long, unsigned long, struct mm_walk
*);
909 int (*pmd_entry
)(pmd_t
*, unsigned long, unsigned long, struct mm_walk
*);
910 int (*pte_entry
)(pte_t
*, unsigned long, unsigned long, struct mm_walk
*);
911 int (*pte_hole
)(unsigned long, unsigned long, struct mm_walk
*);
912 int (*hugetlb_entry
)(pte_t
*, unsigned long,
913 unsigned long, unsigned long, struct mm_walk
*);
914 struct mm_struct
*mm
;
918 int walk_page_range(unsigned long addr
, unsigned long end
,
919 struct mm_walk
*walk
);
920 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
921 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
922 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
923 struct vm_area_struct
*vma
);
924 void unmap_mapping_range(struct address_space
*mapping
,
925 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
926 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
928 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
929 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
930 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
931 void *buf
, int len
, int write
);
933 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
934 loff_t
const holebegin
, loff_t
const holelen
)
936 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
939 extern void truncate_pagecache(struct inode
*inode
, loff_t old
, loff_t
new);
940 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
941 extern int vmtruncate(struct inode
*inode
, loff_t offset
);
942 extern int vmtruncate_range(struct inode
*inode
, loff_t offset
, loff_t end
);
944 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
945 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
947 int invalidate_inode_page(struct page
*page
);
950 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
951 unsigned long address
, unsigned int flags
);
953 static inline int handle_mm_fault(struct mm_struct
*mm
,
954 struct vm_area_struct
*vma
, unsigned long address
,
957 /* should never happen if there's no MMU */
959 return VM_FAULT_SIGBUS
;
963 extern int make_pages_present(unsigned long addr
, unsigned long end
);
964 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
966 int get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
967 unsigned long start
, int nr_pages
, int write
, int force
,
968 struct page
**pages
, struct vm_area_struct
**vmas
);
969 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
970 struct page
**pages
);
971 struct page
*get_dump_page(unsigned long addr
);
973 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
974 extern void do_invalidatepage(struct page
*page
, unsigned long offset
);
976 int __set_page_dirty_nobuffers(struct page
*page
);
977 int __set_page_dirty_no_writeback(struct page
*page
);
978 int redirty_page_for_writepage(struct writeback_control
*wbc
,
980 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
981 void account_page_writeback(struct page
*page
);
982 int set_page_dirty(struct page
*page
);
983 int set_page_dirty_lock(struct page
*page
);
984 int clear_page_dirty_for_io(struct page
*page
);
986 /* Is the vma a continuation of the stack vma above it? */
987 static inline int vma_stack_continue(struct vm_area_struct
*vma
, unsigned long addr
)
989 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
992 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
993 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
994 unsigned long new_addr
, unsigned long len
);
995 extern unsigned long do_mremap(unsigned long addr
,
996 unsigned long old_len
, unsigned long new_len
,
997 unsigned long flags
, unsigned long new_addr
);
998 extern int mprotect_fixup(struct vm_area_struct
*vma
,
999 struct vm_area_struct
**pprev
, unsigned long start
,
1000 unsigned long end
, unsigned long newflags
);
1003 * doesn't attempt to fault and will return short.
1005 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1006 struct page
**pages
);
1008 * per-process(per-mm_struct) statistics.
1010 #if defined(SPLIT_RSS_COUNTING)
1012 * The mm counters are not protected by its page_table_lock,
1013 * so must be incremented atomically.
1015 static inline void set_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1017 atomic_long_set(&mm
->rss_stat
.count
[member
], value
);
1020 unsigned long get_mm_counter(struct mm_struct
*mm
, int member
);
1022 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1024 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1027 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1029 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1032 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1034 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1037 #else /* !USE_SPLIT_PTLOCKS */
1039 * The mm counters are protected by its page_table_lock,
1040 * so can be incremented directly.
1042 static inline void set_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1044 mm
->rss_stat
.count
[member
] = value
;
1047 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1049 return mm
->rss_stat
.count
[member
];
1052 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1054 mm
->rss_stat
.count
[member
] += value
;
1057 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1059 mm
->rss_stat
.count
[member
]++;
1062 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1064 mm
->rss_stat
.count
[member
]--;
1067 #endif /* !USE_SPLIT_PTLOCKS */
1069 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1071 return get_mm_counter(mm
, MM_FILEPAGES
) +
1072 get_mm_counter(mm
, MM_ANONPAGES
);
1075 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1077 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1080 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1082 return max(mm
->hiwater_vm
, mm
->total_vm
);
1085 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1087 unsigned long _rss
= get_mm_rss(mm
);
1089 if ((mm
)->hiwater_rss
< _rss
)
1090 (mm
)->hiwater_rss
= _rss
;
1093 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1095 if (mm
->hiwater_vm
< mm
->total_vm
)
1096 mm
->hiwater_vm
= mm
->total_vm
;
1099 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1100 struct mm_struct
*mm
)
1102 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1104 if (*maxrss
< hiwater_rss
)
1105 *maxrss
= hiwater_rss
;
1108 #if defined(SPLIT_RSS_COUNTING)
1109 void sync_mm_rss(struct task_struct
*task
, struct mm_struct
*mm
);
1111 static inline void sync_mm_rss(struct task_struct
*task
, struct mm_struct
*mm
)
1117 * A callback you can register to apply pressure to ageable caches.
1119 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
1120 * look through the least-recently-used 'nr_to_scan' entries and
1121 * attempt to free them up. It should return the number of objects
1122 * which remain in the cache. If it returns -1, it means it cannot do
1123 * any scanning at this time (eg. there is a risk of deadlock).
1125 * The 'gfpmask' refers to the allocation we are currently trying to
1128 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
1129 * querying the cache size, so a fastpath for that case is appropriate.
1132 int (*shrink
)(struct shrinker
*, int nr_to_scan
, gfp_t gfp_mask
);
1133 int seeks
; /* seeks to recreate an obj */
1135 /* These are for internal use */
1136 struct list_head list
;
1137 long nr
; /* objs pending delete */
1139 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1140 extern void register_shrinker(struct shrinker
*);
1141 extern void unregister_shrinker(struct shrinker
*);
1143 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1145 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1147 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1151 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1155 #ifdef __PAGETABLE_PUD_FOLDED
1156 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1157 unsigned long address
)
1162 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1165 #ifdef __PAGETABLE_PMD_FOLDED
1166 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1167 unsigned long address
)
1172 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1175 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1176 pmd_t
*pmd
, unsigned long address
);
1177 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1180 * The following ifdef needed to get the 4level-fixup.h header to work.
1181 * Remove it when 4level-fixup.h has been removed.
1183 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1184 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1186 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1187 NULL
: pud_offset(pgd
, address
);
1190 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1192 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1193 NULL
: pmd_offset(pud
, address
);
1195 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1197 #if USE_SPLIT_PTLOCKS
1199 * We tuck a spinlock to guard each pagetable page into its struct page,
1200 * at page->private, with BUILD_BUG_ON to make sure that this will not
1201 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1202 * When freeing, reset page->mapping so free_pages_check won't complain.
1204 #define __pte_lockptr(page) &((page)->ptl)
1205 #define pte_lock_init(_page) do { \
1206 spin_lock_init(__pte_lockptr(_page)); \
1208 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1209 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1210 #else /* !USE_SPLIT_PTLOCKS */
1212 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1214 #define pte_lock_init(page) do {} while (0)
1215 #define pte_lock_deinit(page) do {} while (0)
1216 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1217 #endif /* USE_SPLIT_PTLOCKS */
1219 static inline void pgtable_page_ctor(struct page
*page
)
1221 pte_lock_init(page
);
1222 inc_zone_page_state(page
, NR_PAGETABLE
);
1225 static inline void pgtable_page_dtor(struct page
*page
)
1227 pte_lock_deinit(page
);
1228 dec_zone_page_state(page
, NR_PAGETABLE
);
1231 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1233 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1234 pte_t *__pte = pte_offset_map(pmd, address); \
1240 #define pte_unmap_unlock(pte, ptl) do { \
1245 #define pte_alloc_map(mm, vma, pmd, address) \
1246 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1248 NULL: pte_offset_map(pmd, address))
1250 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1251 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1253 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1255 #define pte_alloc_kernel(pmd, address) \
1256 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1257 NULL: pte_offset_kernel(pmd, address))
1259 extern void free_area_init(unsigned long * zones_size
);
1260 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1261 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1262 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1264 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1265 * zones, allocate the backing mem_map and account for memory holes in a more
1266 * architecture independent manner. This is a substitute for creating the
1267 * zone_sizes[] and zholes_size[] arrays and passing them to
1268 * free_area_init_node()
1270 * An architecture is expected to register range of page frames backed by
1271 * physical memory with add_active_range() before calling
1272 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1273 * usage, an architecture is expected to do something like
1275 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1277 * for_each_valid_physical_page_range()
1278 * add_active_range(node_id, start_pfn, end_pfn)
1279 * free_area_init_nodes(max_zone_pfns);
1281 * If the architecture guarantees that there are no holes in the ranges
1282 * registered with add_active_range(), free_bootmem_active_regions()
1283 * will call free_bootmem_node() for each registered physical page range.
1284 * Similarly sparse_memory_present_with_active_regions() calls
1285 * memory_present() for each range when SPARSEMEM is enabled.
1287 * See mm/page_alloc.c for more information on each function exposed by
1288 * CONFIG_ARCH_POPULATES_NODE_MAP
1290 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1291 extern void add_active_range(unsigned int nid
, unsigned long start_pfn
,
1292 unsigned long end_pfn
);
1293 extern void remove_active_range(unsigned int nid
, unsigned long start_pfn
,
1294 unsigned long end_pfn
);
1295 extern void remove_all_active_ranges(void);
1296 void sort_node_map(void);
1297 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1298 unsigned long end_pfn
);
1299 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1300 unsigned long end_pfn
);
1301 extern void get_pfn_range_for_nid(unsigned int nid
,
1302 unsigned long *start_pfn
, unsigned long *end_pfn
);
1303 extern unsigned long find_min_pfn_with_active_regions(void);
1304 extern void free_bootmem_with_active_regions(int nid
,
1305 unsigned long max_low_pfn
);
1306 int add_from_early_node_map(struct range
*range
, int az
,
1307 int nr_range
, int nid
);
1308 u64 __init
find_memory_core_early(int nid
, u64 size
, u64 align
,
1309 u64 goal
, u64 limit
);
1310 void *__alloc_memory_core_early(int nodeid
, u64 size
, u64 align
,
1311 u64 goal
, u64 limit
);
1312 typedef int (*work_fn_t
)(unsigned long, unsigned long, void *);
1313 extern void work_with_active_regions(int nid
, work_fn_t work_fn
, void *data
);
1314 extern void sparse_memory_present_with_active_regions(int nid
);
1315 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1317 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1318 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1319 static inline int __early_pfn_to_nid(unsigned long pfn
)
1324 /* please see mm/page_alloc.c */
1325 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1326 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1327 /* there is a per-arch backend function. */
1328 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
);
1329 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1332 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1333 extern void memmap_init_zone(unsigned long, int, unsigned long,
1334 unsigned long, enum memmap_context
);
1335 extern void setup_per_zone_wmarks(void);
1336 extern void calculate_zone_inactive_ratio(struct zone
*zone
);
1337 extern void mem_init(void);
1338 extern void __init
mmap_init(void);
1339 extern void show_mem(void);
1340 extern void si_meminfo(struct sysinfo
* val
);
1341 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1342 extern int after_bootmem
;
1344 extern void setup_per_cpu_pageset(void);
1346 extern void zone_pcp_update(struct zone
*zone
);
1349 extern atomic_long_t mmap_pages_allocated
;
1350 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1353 void vma_prio_tree_add(struct vm_area_struct
*, struct vm_area_struct
*old
);
1354 void vma_prio_tree_insert(struct vm_area_struct
*, struct prio_tree_root
*);
1355 void vma_prio_tree_remove(struct vm_area_struct
*, struct prio_tree_root
*);
1356 struct vm_area_struct
*vma_prio_tree_next(struct vm_area_struct
*vma
,
1357 struct prio_tree_iter
*iter
);
1359 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1360 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1361 (vma = vma_prio_tree_next(vma, iter)); )
1363 static inline void vma_nonlinear_insert(struct vm_area_struct
*vma
,
1364 struct list_head
*list
)
1366 vma
->shared
.vm_set
.parent
= NULL
;
1367 list_add_tail(&vma
->shared
.vm_set
.list
, list
);
1371 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1372 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1373 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1374 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1375 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1376 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1377 struct mempolicy
*);
1378 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1379 extern int split_vma(struct mm_struct
*,
1380 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1381 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1382 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1383 struct rb_node
**, struct rb_node
*);
1384 extern void unlink_file_vma(struct vm_area_struct
*);
1385 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1386 unsigned long addr
, unsigned long len
, pgoff_t pgoff
);
1387 extern void exit_mmap(struct mm_struct
*);
1389 extern int mm_take_all_locks(struct mm_struct
*mm
);
1390 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1392 #ifdef CONFIG_PROC_FS
1393 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1394 extern void added_exe_file_vma(struct mm_struct
*mm
);
1395 extern void removed_exe_file_vma(struct mm_struct
*mm
);
1397 static inline void added_exe_file_vma(struct mm_struct
*mm
)
1400 static inline void removed_exe_file_vma(struct mm_struct
*mm
)
1402 #endif /* CONFIG_PROC_FS */
1404 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1405 extern int install_special_mapping(struct mm_struct
*mm
,
1406 unsigned long addr
, unsigned long len
,
1407 unsigned long flags
, struct page
**pages
);
1409 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1411 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1412 unsigned long len
, unsigned long prot
,
1413 unsigned long flag
, unsigned long pgoff
);
1414 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1415 unsigned long len
, unsigned long flags
,
1416 unsigned int vm_flags
, unsigned long pgoff
);
1418 static inline unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1419 unsigned long len
, unsigned long prot
,
1420 unsigned long flag
, unsigned long offset
)
1422 unsigned long ret
= -EINVAL
;
1423 if ((offset
+ PAGE_ALIGN(len
)) < offset
)
1425 if (!(offset
& ~PAGE_MASK
))
1426 ret
= do_mmap_pgoff(file
, addr
, len
, prot
, flag
, offset
>> PAGE_SHIFT
);
1431 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1433 extern unsigned long do_brk(unsigned long, unsigned long);
1436 extern unsigned long page_unuse(struct page
*);
1437 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1438 extern void truncate_inode_pages_range(struct address_space
*,
1439 loff_t lstart
, loff_t lend
);
1441 /* generic vm_area_ops exported for stackable file systems */
1442 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1444 /* mm/page-writeback.c */
1445 int write_one_page(struct page
*page
, int wait
);
1446 void task_dirty_inc(struct task_struct
*tsk
);
1449 #define VM_MAX_READAHEAD 128 /* kbytes */
1450 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1452 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1453 pgoff_t offset
, unsigned long nr_to_read
);
1455 void page_cache_sync_readahead(struct address_space
*mapping
,
1456 struct file_ra_state
*ra
,
1459 unsigned long size
);
1461 void page_cache_async_readahead(struct address_space
*mapping
,
1462 struct file_ra_state
*ra
,
1466 unsigned long size
);
1468 unsigned long max_sane_readahead(unsigned long nr
);
1469 unsigned long ra_submit(struct file_ra_state
*ra
,
1470 struct address_space
*mapping
,
1473 /* Do stack extension */
1474 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1476 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1478 #define expand_upwards(vma, address) do { } while (0)
1480 extern int expand_stack_downwards(struct vm_area_struct
*vma
,
1481 unsigned long address
);
1483 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1484 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1485 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1486 struct vm_area_struct
**pprev
);
1488 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1489 NULL if none. Assume start_addr < end_addr. */
1490 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1492 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1494 if (vma
&& end_addr
<= vma
->vm_start
)
1499 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1501 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1505 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
1507 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
1513 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
1514 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
1515 unsigned long pfn
, unsigned long size
, pgprot_t
);
1516 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
1517 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
1519 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
1522 struct page
*follow_page(struct vm_area_struct
*, unsigned long address
,
1523 unsigned int foll_flags
);
1524 #define FOLL_WRITE 0x01 /* check pte is writable */
1525 #define FOLL_TOUCH 0x02 /* mark page accessed */
1526 #define FOLL_GET 0x04 /* do get_page on page */
1527 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1528 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1529 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1530 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1532 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
1534 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
1535 unsigned long size
, pte_fn_t fn
, void *data
);
1537 #ifdef CONFIG_PROC_FS
1538 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
1540 static inline void vm_stat_account(struct mm_struct
*mm
,
1541 unsigned long flags
, struct file
*file
, long pages
)
1544 #endif /* CONFIG_PROC_FS */
1546 #ifdef CONFIG_DEBUG_PAGEALLOC
1547 extern int debug_pagealloc_enabled
;
1549 extern void kernel_map_pages(struct page
*page
, int numpages
, int enable
);
1551 static inline void enable_debug_pagealloc(void)
1553 debug_pagealloc_enabled
= 1;
1555 #ifdef CONFIG_HIBERNATION
1556 extern bool kernel_page_present(struct page
*page
);
1557 #endif /* CONFIG_HIBERNATION */
1560 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
1561 static inline void enable_debug_pagealloc(void)
1564 #ifdef CONFIG_HIBERNATION
1565 static inline bool kernel_page_present(struct page
*page
) { return true; }
1566 #endif /* CONFIG_HIBERNATION */
1569 extern struct vm_area_struct
*get_gate_vma(struct task_struct
*tsk
);
1570 #ifdef __HAVE_ARCH_GATE_AREA
1571 int in_gate_area_no_task(unsigned long addr
);
1572 int in_gate_area(struct task_struct
*task
, unsigned long addr
);
1574 int in_gate_area_no_task(unsigned long addr
);
1575 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1576 #endif /* __HAVE_ARCH_GATE_AREA */
1578 int drop_caches_sysctl_handler(struct ctl_table
*, int,
1579 void __user
*, size_t *, loff_t
*);
1580 unsigned long shrink_slab(unsigned long scanned
, gfp_t gfp_mask
,
1581 unsigned long lru_pages
);
1584 #define randomize_va_space 0
1586 extern int randomize_va_space
;
1589 const char * arch_vma_name(struct vm_area_struct
*vma
);
1590 void print_vma_addr(char *prefix
, unsigned long rip
);
1592 void sparse_mem_maps_populate_node(struct page
**map_map
,
1593 unsigned long pnum_begin
,
1594 unsigned long pnum_end
,
1595 unsigned long map_count
,
1598 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
1599 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
1600 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
1601 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
1602 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
1603 void *vmemmap_alloc_block(unsigned long size
, int node
);
1604 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
1605 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
1606 int vmemmap_populate_basepages(struct page
*start_page
,
1607 unsigned long pages
, int node
);
1608 int vmemmap_populate(struct page
*start_page
, unsigned long pages
, int node
);
1609 void vmemmap_populate_print_last(void);
1613 MF_COUNT_INCREASED
= 1 << 0,
1615 extern void memory_failure(unsigned long pfn
, int trapno
);
1616 extern int __memory_failure(unsigned long pfn
, int trapno
, int flags
);
1617 extern int unpoison_memory(unsigned long pfn
);
1618 extern int sysctl_memory_failure_early_kill
;
1619 extern int sysctl_memory_failure_recovery
;
1620 extern void shake_page(struct page
*p
, int access
);
1621 extern atomic_long_t mce_bad_pages
;
1622 extern int soft_offline_page(struct page
*page
, int flags
);
1623 #ifdef CONFIG_MEMORY_FAILURE
1624 int is_hwpoison_address(unsigned long addr
);
1626 static inline int is_hwpoison_address(unsigned long addr
)
1632 extern void dump_page(struct page
*page
);
1634 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1635 extern void clear_huge_page(struct page
*page
,
1637 unsigned int pages_per_huge_page
);
1638 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
1639 unsigned long addr
, struct vm_area_struct
*vma
,
1640 unsigned int pages_per_huge_page
);
1641 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1643 #endif /* __KERNEL__ */
1644 #endif /* _LINUX_MM_H */