writeback: split writeback_inodes_wb
[linux-2.6/next.git] / include / linux / mm.h
blobb969efb03787ee69546995ad378ac54b6bc37e3b
1 #ifndef _LINUX_MM_H
2 #define _LINUX_MM_H
4 #include <linux/errno.h>
6 #ifdef __KERNEL__
8 #include <linux/gfp.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>
18 struct mempolicy;
19 struct anon_vma;
20 struct file_ra_state;
21 struct user_struct;
22 struct writeback_control;
24 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
25 extern unsigned long max_mapnr;
26 #endif
28 extern unsigned long num_physpages;
29 extern unsigned long totalram_pages;
30 extern void * high_memory;
31 extern int page_cluster;
33 #ifdef CONFIG_SYSCTL
34 extern int sysctl_legacy_va_layout;
35 #else
36 #define sysctl_legacy_va_layout 0
37 #endif
39 #include <asm/page.h>
40 #include <asm/pgtable.h>
41 #include <asm/processor.h>
43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
45 /* to align the pointer to the (next) page boundary */
46 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
49 * Linux kernel virtual memory manager primitives.
50 * The idea being to have a "virtual" mm in the same way
51 * we have a virtual fs - giving a cleaner interface to the
52 * mm details, and allowing different kinds of memory mappings
53 * (from shared memory to executable loading to arbitrary
54 * mmap() functions).
57 extern struct kmem_cache *vm_area_cachep;
59 #ifndef CONFIG_MMU
60 extern struct rb_root nommu_region_tree;
61 extern struct rw_semaphore nommu_region_sem;
63 extern unsigned int kobjsize(const void *objp);
64 #endif
67 * vm_flags in vm_area_struct, see mm_types.h.
69 #define VM_READ 0x00000001 /* currently active flags */
70 #define VM_WRITE 0x00000002
71 #define VM_EXEC 0x00000004
72 #define VM_SHARED 0x00000008
74 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
75 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
76 #define VM_MAYWRITE 0x00000020
77 #define VM_MAYEXEC 0x00000040
78 #define VM_MAYSHARE 0x00000080
80 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
81 #define VM_GROWSUP 0x00000200
82 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
83 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
85 #define VM_EXECUTABLE 0x00001000
86 #define VM_LOCKED 0x00002000
87 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
89 /* Used by sys_madvise() */
90 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
91 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
93 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
94 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
95 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
96 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
97 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
98 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
99 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
100 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
101 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
102 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
104 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
105 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
106 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
107 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
108 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
110 /* Bits set in the VMA until the stack is in its final location */
111 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
113 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
114 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
115 #endif
117 #ifdef CONFIG_STACK_GROWSUP
118 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
119 #else
120 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
121 #endif
123 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
124 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
125 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
126 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
127 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
130 * special vmas that are non-mergable, non-mlock()able
132 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
135 * mapping from the currently active vm_flags protection bits (the
136 * low four bits) to a page protection mask..
138 extern pgprot_t protection_map[16];
140 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
141 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
142 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
145 * This interface is used by x86 PAT code to identify a pfn mapping that is
146 * linear over entire vma. This is to optimize PAT code that deals with
147 * marking the physical region with a particular prot. This is not for generic
148 * mm use. Note also that this check will not work if the pfn mapping is
149 * linear for a vma starting at physical address 0. In which case PAT code
150 * falls back to slow path of reserving physical range page by page.
152 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
154 return (vma->vm_flags & VM_PFN_AT_MMAP);
157 static inline int is_pfn_mapping(struct vm_area_struct *vma)
159 return (vma->vm_flags & VM_PFNMAP);
163 * vm_fault is filled by the the pagefault handler and passed to the vma's
164 * ->fault function. The vma's ->fault is responsible for returning a bitmask
165 * of VM_FAULT_xxx flags that give details about how the fault was handled.
167 * pgoff should be used in favour of virtual_address, if possible. If pgoff
168 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
169 * mapping support.
171 struct vm_fault {
172 unsigned int flags; /* FAULT_FLAG_xxx flags */
173 pgoff_t pgoff; /* Logical page offset based on vma */
174 void __user *virtual_address; /* Faulting virtual address */
176 struct page *page; /* ->fault handlers should return a
177 * page here, unless VM_FAULT_NOPAGE
178 * is set (which is also implied by
179 * VM_FAULT_ERROR).
184 * These are the virtual MM functions - opening of an area, closing and
185 * unmapping it (needed to keep files on disk up-to-date etc), pointer
186 * to the functions called when a no-page or a wp-page exception occurs.
188 struct vm_operations_struct {
189 void (*open)(struct vm_area_struct * area);
190 void (*close)(struct vm_area_struct * area);
191 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
193 /* notification that a previously read-only page is about to become
194 * writable, if an error is returned it will cause a SIGBUS */
195 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
197 /* called by access_process_vm when get_user_pages() fails, typically
198 * for use by special VMAs that can switch between memory and hardware
200 int (*access)(struct vm_area_struct *vma, unsigned long addr,
201 void *buf, int len, int write);
202 #ifdef CONFIG_NUMA
204 * set_policy() op must add a reference to any non-NULL @new mempolicy
205 * to hold the policy upon return. Caller should pass NULL @new to
206 * remove a policy and fall back to surrounding context--i.e. do not
207 * install a MPOL_DEFAULT policy, nor the task or system default
208 * mempolicy.
210 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
213 * get_policy() op must add reference [mpol_get()] to any policy at
214 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
215 * in mm/mempolicy.c will do this automatically.
216 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
217 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
218 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
219 * must return NULL--i.e., do not "fallback" to task or system default
220 * policy.
222 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
223 unsigned long addr);
224 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
225 const nodemask_t *to, unsigned long flags);
226 #endif
229 struct mmu_gather;
230 struct inode;
232 #define page_private(page) ((page)->private)
233 #define set_page_private(page, v) ((page)->private = (v))
236 * FIXME: take this include out, include page-flags.h in
237 * files which need it (119 of them)
239 #include <linux/page-flags.h>
242 * Methods to modify the page usage count.
244 * What counts for a page usage:
245 * - cache mapping (page->mapping)
246 * - private data (page->private)
247 * - page mapped in a task's page tables, each mapping
248 * is counted separately
250 * Also, many kernel routines increase the page count before a critical
251 * routine so they can be sure the page doesn't go away from under them.
255 * Drop a ref, return true if the refcount fell to zero (the page has no users)
257 static inline int put_page_testzero(struct page *page)
259 VM_BUG_ON(atomic_read(&page->_count) == 0);
260 return atomic_dec_and_test(&page->_count);
264 * Try to grab a ref unless the page has a refcount of zero, return false if
265 * that is the case.
267 static inline int get_page_unless_zero(struct page *page)
269 return atomic_inc_not_zero(&page->_count);
272 extern int page_is_ram(unsigned long pfn);
274 /* Support for virtually mapped pages */
275 struct page *vmalloc_to_page(const void *addr);
276 unsigned long vmalloc_to_pfn(const void *addr);
279 * Determine if an address is within the vmalloc range
281 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
282 * is no special casing required.
284 static inline int is_vmalloc_addr(const void *x)
286 #ifdef CONFIG_MMU
287 unsigned long addr = (unsigned long)x;
289 return addr >= VMALLOC_START && addr < VMALLOC_END;
290 #else
291 return 0;
292 #endif
294 #ifdef CONFIG_MMU
295 extern int is_vmalloc_or_module_addr(const void *x);
296 #else
297 static inline int is_vmalloc_or_module_addr(const void *x)
299 return 0;
301 #endif
303 static inline struct page *compound_head(struct page *page)
305 if (unlikely(PageTail(page)))
306 return page->first_page;
307 return page;
310 static inline int page_count(struct page *page)
312 return atomic_read(&compound_head(page)->_count);
315 static inline void get_page(struct page *page)
317 page = compound_head(page);
318 VM_BUG_ON(atomic_read(&page->_count) == 0);
319 atomic_inc(&page->_count);
322 static inline struct page *virt_to_head_page(const void *x)
324 struct page *page = virt_to_page(x);
325 return compound_head(page);
329 * Setup the page count before being freed into the page allocator for
330 * the first time (boot or memory hotplug)
332 static inline void init_page_count(struct page *page)
334 atomic_set(&page->_count, 1);
337 void put_page(struct page *page);
338 void put_pages_list(struct list_head *pages);
340 void split_page(struct page *page, unsigned int order);
341 int split_free_page(struct page *page);
344 * Compound pages have a destructor function. Provide a
345 * prototype for that function and accessor functions.
346 * These are _only_ valid on the head of a PG_compound page.
348 typedef void compound_page_dtor(struct page *);
350 static inline void set_compound_page_dtor(struct page *page,
351 compound_page_dtor *dtor)
353 page[1].lru.next = (void *)dtor;
356 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
358 return (compound_page_dtor *)page[1].lru.next;
361 static inline int compound_order(struct page *page)
363 if (!PageHead(page))
364 return 0;
365 return (unsigned long)page[1].lru.prev;
368 static inline void set_compound_order(struct page *page, unsigned long order)
370 page[1].lru.prev = (void *)order;
374 * Multiple processes may "see" the same page. E.g. for untouched
375 * mappings of /dev/null, all processes see the same page full of
376 * zeroes, and text pages of executables and shared libraries have
377 * only one copy in memory, at most, normally.
379 * For the non-reserved pages, page_count(page) denotes a reference count.
380 * page_count() == 0 means the page is free. page->lru is then used for
381 * freelist management in the buddy allocator.
382 * page_count() > 0 means the page has been allocated.
384 * Pages are allocated by the slab allocator in order to provide memory
385 * to kmalloc and kmem_cache_alloc. In this case, the management of the
386 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
387 * unless a particular usage is carefully commented. (the responsibility of
388 * freeing the kmalloc memory is the caller's, of course).
390 * A page may be used by anyone else who does a __get_free_page().
391 * In this case, page_count still tracks the references, and should only
392 * be used through the normal accessor functions. The top bits of page->flags
393 * and page->virtual store page management information, but all other fields
394 * are unused and could be used privately, carefully. The management of this
395 * page is the responsibility of the one who allocated it, and those who have
396 * subsequently been given references to it.
398 * The other pages (we may call them "pagecache pages") are completely
399 * managed by the Linux memory manager: I/O, buffers, swapping etc.
400 * The following discussion applies only to them.
402 * A pagecache page contains an opaque `private' member, which belongs to the
403 * page's address_space. Usually, this is the address of a circular list of
404 * the page's disk buffers. PG_private must be set to tell the VM to call
405 * into the filesystem to release these pages.
407 * A page may belong to an inode's memory mapping. In this case, page->mapping
408 * is the pointer to the inode, and page->index is the file offset of the page,
409 * in units of PAGE_CACHE_SIZE.
411 * If pagecache pages are not associated with an inode, they are said to be
412 * anonymous pages. These may become associated with the swapcache, and in that
413 * case PG_swapcache is set, and page->private is an offset into the swapcache.
415 * In either case (swapcache or inode backed), the pagecache itself holds one
416 * reference to the page. Setting PG_private should also increment the
417 * refcount. The each user mapping also has a reference to the page.
419 * The pagecache pages are stored in a per-mapping radix tree, which is
420 * rooted at mapping->page_tree, and indexed by offset.
421 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
422 * lists, we instead now tag pages as dirty/writeback in the radix tree.
424 * All pagecache pages may be subject to I/O:
425 * - inode pages may need to be read from disk,
426 * - inode pages which have been modified and are MAP_SHARED may need
427 * to be written back to the inode on disk,
428 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
429 * modified may need to be swapped out to swap space and (later) to be read
430 * back into memory.
434 * The zone field is never updated after free_area_init_core()
435 * sets it, so none of the operations on it need to be atomic.
440 * page->flags layout:
442 * There are three possibilities for how page->flags get
443 * laid out. The first is for the normal case, without
444 * sparsemem. The second is for sparsemem when there is
445 * plenty of space for node and section. The last is when
446 * we have run out of space and have to fall back to an
447 * alternate (slower) way of determining the node.
449 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
450 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
451 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
453 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
454 #define SECTIONS_WIDTH SECTIONS_SHIFT
455 #else
456 #define SECTIONS_WIDTH 0
457 #endif
459 #define ZONES_WIDTH ZONES_SHIFT
461 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
462 #define NODES_WIDTH NODES_SHIFT
463 #else
464 #ifdef CONFIG_SPARSEMEM_VMEMMAP
465 #error "Vmemmap: No space for nodes field in page flags"
466 #endif
467 #define NODES_WIDTH 0
468 #endif
470 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
471 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
472 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
473 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
476 * We are going to use the flags for the page to node mapping if its in
477 * there. This includes the case where there is no node, so it is implicit.
479 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
480 #define NODE_NOT_IN_PAGE_FLAGS
481 #endif
483 #ifndef PFN_SECTION_SHIFT
484 #define PFN_SECTION_SHIFT 0
485 #endif
488 * Define the bit shifts to access each section. For non-existant
489 * sections we define the shift as 0; that plus a 0 mask ensures
490 * the compiler will optimise away reference to them.
492 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
493 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
494 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
496 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
497 #ifdef NODE_NOT_IN_PAGEFLAGS
498 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
499 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
500 SECTIONS_PGOFF : ZONES_PGOFF)
501 #else
502 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
503 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
504 NODES_PGOFF : ZONES_PGOFF)
505 #endif
507 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
509 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
510 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
511 #endif
513 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
514 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
515 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
516 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
518 static inline enum zone_type page_zonenum(struct page *page)
520 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
524 * The identification function is only used by the buddy allocator for
525 * determining if two pages could be buddies. We are not really
526 * identifying a zone since we could be using a the section number
527 * id if we have not node id available in page flags.
528 * We guarantee only that it will return the same value for two
529 * combinable pages in a zone.
531 static inline int page_zone_id(struct page *page)
533 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
536 static inline int zone_to_nid(struct zone *zone)
538 #ifdef CONFIG_NUMA
539 return zone->node;
540 #else
541 return 0;
542 #endif
545 #ifdef NODE_NOT_IN_PAGE_FLAGS
546 extern int page_to_nid(struct page *page);
547 #else
548 static inline int page_to_nid(struct page *page)
550 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
552 #endif
554 static inline struct zone *page_zone(struct page *page)
556 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
559 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
560 static inline unsigned long page_to_section(struct page *page)
562 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
564 #endif
566 static inline void set_page_zone(struct page *page, enum zone_type zone)
568 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
569 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
572 static inline void set_page_node(struct page *page, unsigned long node)
574 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
575 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
578 static inline void set_page_section(struct page *page, unsigned long section)
580 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
581 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
584 static inline void set_page_links(struct page *page, enum zone_type zone,
585 unsigned long node, unsigned long pfn)
587 set_page_zone(page, zone);
588 set_page_node(page, node);
589 set_page_section(page, pfn_to_section_nr(pfn));
593 * Some inline functions in vmstat.h depend on page_zone()
595 #include <linux/vmstat.h>
597 static __always_inline void *lowmem_page_address(struct page *page)
599 return __va(PFN_PHYS(page_to_pfn(page)));
602 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
603 #define HASHED_PAGE_VIRTUAL
604 #endif
606 #if defined(WANT_PAGE_VIRTUAL)
607 #define page_address(page) ((page)->virtual)
608 #define set_page_address(page, address) \
609 do { \
610 (page)->virtual = (address); \
611 } while(0)
612 #define page_address_init() do { } while(0)
613 #endif
615 #if defined(HASHED_PAGE_VIRTUAL)
616 void *page_address(struct page *page);
617 void set_page_address(struct page *page, void *virtual);
618 void page_address_init(void);
619 #endif
621 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
622 #define page_address(page) lowmem_page_address(page)
623 #define set_page_address(page, address) do { } while(0)
624 #define page_address_init() do { } while(0)
625 #endif
628 * On an anonymous page mapped into a user virtual memory area,
629 * page->mapping points to its anon_vma, not to a struct address_space;
630 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
632 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
633 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
634 * and then page->mapping points, not to an anon_vma, but to a private
635 * structure which KSM associates with that merged page. See ksm.h.
637 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
639 * Please note that, confusingly, "page_mapping" refers to the inode
640 * address_space which maps the page from disk; whereas "page_mapped"
641 * refers to user virtual address space into which the page is mapped.
643 #define PAGE_MAPPING_ANON 1
644 #define PAGE_MAPPING_KSM 2
645 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
647 extern struct address_space swapper_space;
648 static inline struct address_space *page_mapping(struct page *page)
650 struct address_space *mapping = page->mapping;
652 VM_BUG_ON(PageSlab(page));
653 if (unlikely(PageSwapCache(page)))
654 mapping = &swapper_space;
655 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
656 mapping = NULL;
657 return mapping;
660 /* Neutral page->mapping pointer to address_space or anon_vma or other */
661 static inline void *page_rmapping(struct page *page)
663 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
666 static inline int PageAnon(struct page *page)
668 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
672 * Return the pagecache index of the passed page. Regular pagecache pages
673 * use ->index whereas swapcache pages use ->private
675 static inline pgoff_t page_index(struct page *page)
677 if (unlikely(PageSwapCache(page)))
678 return page_private(page);
679 return page->index;
683 * The atomic page->_mapcount, like _count, starts from -1:
684 * so that transitions both from it and to it can be tracked,
685 * using atomic_inc_and_test and atomic_add_negative(-1).
687 static inline void reset_page_mapcount(struct page *page)
689 atomic_set(&(page)->_mapcount, -1);
692 static inline int page_mapcount(struct page *page)
694 return atomic_read(&(page)->_mapcount) + 1;
698 * Return true if this page is mapped into pagetables.
700 static inline int page_mapped(struct page *page)
702 return atomic_read(&(page)->_mapcount) >= 0;
706 * Different kinds of faults, as returned by handle_mm_fault().
707 * Used to decide whether a process gets delivered SIGBUS or
708 * just gets major/minor fault counters bumped up.
711 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
713 #define VM_FAULT_OOM 0x0001
714 #define VM_FAULT_SIGBUS 0x0002
715 #define VM_FAULT_MAJOR 0x0004
716 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
717 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */
719 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
720 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
722 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON)
725 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
727 extern void pagefault_out_of_memory(void);
729 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
731 extern void show_free_areas(void);
733 int shmem_lock(struct file *file, int lock, struct user_struct *user);
734 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
735 int shmem_zero_setup(struct vm_area_struct *);
737 #ifndef CONFIG_MMU
738 extern unsigned long shmem_get_unmapped_area(struct file *file,
739 unsigned long addr,
740 unsigned long len,
741 unsigned long pgoff,
742 unsigned long flags);
743 #endif
745 extern int can_do_mlock(void);
746 extern int user_shm_lock(size_t, struct user_struct *);
747 extern void user_shm_unlock(size_t, struct user_struct *);
750 * Parameter block passed down to zap_pte_range in exceptional cases.
752 struct zap_details {
753 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
754 struct address_space *check_mapping; /* Check page->mapping if set */
755 pgoff_t first_index; /* Lowest page->index to unmap */
756 pgoff_t last_index; /* Highest page->index to unmap */
757 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
758 unsigned long truncate_count; /* Compare vm_truncate_count */
761 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
762 pte_t pte);
764 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
765 unsigned long size);
766 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
767 unsigned long size, struct zap_details *);
768 unsigned long unmap_vmas(struct mmu_gather **tlb,
769 struct vm_area_struct *start_vma, unsigned long start_addr,
770 unsigned long end_addr, unsigned long *nr_accounted,
771 struct zap_details *);
774 * mm_walk - callbacks for walk_page_range
775 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
776 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
777 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
778 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
779 * @pte_hole: if set, called for each hole at all levels
780 * @hugetlb_entry: if set, called for each hugetlb entry
782 * (see walk_page_range for more details)
784 struct mm_walk {
785 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
786 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
787 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
788 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
789 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
790 int (*hugetlb_entry)(pte_t *, unsigned long,
791 unsigned long, unsigned long, struct mm_walk *);
792 struct mm_struct *mm;
793 void *private;
796 int walk_page_range(unsigned long addr, unsigned long end,
797 struct mm_walk *walk);
798 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
799 unsigned long end, unsigned long floor, unsigned long ceiling);
800 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
801 struct vm_area_struct *vma);
802 void unmap_mapping_range(struct address_space *mapping,
803 loff_t const holebegin, loff_t const holelen, int even_cows);
804 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
805 unsigned long *pfn);
806 int follow_phys(struct vm_area_struct *vma, unsigned long address,
807 unsigned int flags, unsigned long *prot, resource_size_t *phys);
808 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
809 void *buf, int len, int write);
811 static inline void unmap_shared_mapping_range(struct address_space *mapping,
812 loff_t const holebegin, loff_t const holelen)
814 unmap_mapping_range(mapping, holebegin, holelen, 0);
817 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
818 extern int vmtruncate(struct inode *inode, loff_t offset);
819 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
821 int truncate_inode_page(struct address_space *mapping, struct page *page);
822 int generic_error_remove_page(struct address_space *mapping, struct page *page);
824 int invalidate_inode_page(struct page *page);
826 #ifdef CONFIG_MMU
827 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
828 unsigned long address, unsigned int flags);
829 #else
830 static inline int handle_mm_fault(struct mm_struct *mm,
831 struct vm_area_struct *vma, unsigned long address,
832 unsigned int flags)
834 /* should never happen if there's no MMU */
835 BUG();
836 return VM_FAULT_SIGBUS;
838 #endif
840 extern int make_pages_present(unsigned long addr, unsigned long end);
841 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
843 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
844 unsigned long start, int nr_pages, int write, int force,
845 struct page **pages, struct vm_area_struct **vmas);
846 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
847 struct page **pages);
848 struct page *get_dump_page(unsigned long addr);
850 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
851 extern void do_invalidatepage(struct page *page, unsigned long offset);
853 int __set_page_dirty_nobuffers(struct page *page);
854 int __set_page_dirty_no_writeback(struct page *page);
855 int redirty_page_for_writepage(struct writeback_control *wbc,
856 struct page *page);
857 void account_page_dirtied(struct page *page, struct address_space *mapping);
858 int set_page_dirty(struct page *page);
859 int set_page_dirty_lock(struct page *page);
860 int clear_page_dirty_for_io(struct page *page);
862 extern unsigned long move_page_tables(struct vm_area_struct *vma,
863 unsigned long old_addr, struct vm_area_struct *new_vma,
864 unsigned long new_addr, unsigned long len);
865 extern unsigned long do_mremap(unsigned long addr,
866 unsigned long old_len, unsigned long new_len,
867 unsigned long flags, unsigned long new_addr);
868 extern int mprotect_fixup(struct vm_area_struct *vma,
869 struct vm_area_struct **pprev, unsigned long start,
870 unsigned long end, unsigned long newflags);
873 * doesn't attempt to fault and will return short.
875 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
876 struct page **pages);
878 * per-process(per-mm_struct) statistics.
880 #if defined(SPLIT_RSS_COUNTING)
882 * The mm counters are not protected by its page_table_lock,
883 * so must be incremented atomically.
885 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
887 atomic_long_set(&mm->rss_stat.count[member], value);
890 unsigned long get_mm_counter(struct mm_struct *mm, int member);
892 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
894 atomic_long_add(value, &mm->rss_stat.count[member]);
897 static inline void inc_mm_counter(struct mm_struct *mm, int member)
899 atomic_long_inc(&mm->rss_stat.count[member]);
902 static inline void dec_mm_counter(struct mm_struct *mm, int member)
904 atomic_long_dec(&mm->rss_stat.count[member]);
907 #else /* !USE_SPLIT_PTLOCKS */
909 * The mm counters are protected by its page_table_lock,
910 * so can be incremented directly.
912 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
914 mm->rss_stat.count[member] = value;
917 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
919 return mm->rss_stat.count[member];
922 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
924 mm->rss_stat.count[member] += value;
927 static inline void inc_mm_counter(struct mm_struct *mm, int member)
929 mm->rss_stat.count[member]++;
932 static inline void dec_mm_counter(struct mm_struct *mm, int member)
934 mm->rss_stat.count[member]--;
937 #endif /* !USE_SPLIT_PTLOCKS */
939 static inline unsigned long get_mm_rss(struct mm_struct *mm)
941 return get_mm_counter(mm, MM_FILEPAGES) +
942 get_mm_counter(mm, MM_ANONPAGES);
945 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
947 return max(mm->hiwater_rss, get_mm_rss(mm));
950 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
952 return max(mm->hiwater_vm, mm->total_vm);
955 static inline void update_hiwater_rss(struct mm_struct *mm)
957 unsigned long _rss = get_mm_rss(mm);
959 if ((mm)->hiwater_rss < _rss)
960 (mm)->hiwater_rss = _rss;
963 static inline void update_hiwater_vm(struct mm_struct *mm)
965 if (mm->hiwater_vm < mm->total_vm)
966 mm->hiwater_vm = mm->total_vm;
969 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
970 struct mm_struct *mm)
972 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
974 if (*maxrss < hiwater_rss)
975 *maxrss = hiwater_rss;
978 #if defined(SPLIT_RSS_COUNTING)
979 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
980 #else
981 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
984 #endif
987 * A callback you can register to apply pressure to ageable caches.
989 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should
990 * look through the least-recently-used 'nr_to_scan' entries and
991 * attempt to free them up. It should return the number of objects
992 * which remain in the cache. If it returns -1, it means it cannot do
993 * any scanning at this time (eg. there is a risk of deadlock).
995 * The 'gfpmask' refers to the allocation we are currently trying to
996 * fulfil.
998 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
999 * querying the cache size, so a fastpath for that case is appropriate.
1001 struct shrinker {
1002 int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
1003 int seeks; /* seeks to recreate an obj */
1005 /* These are for internal use */
1006 struct list_head list;
1007 long nr; /* objs pending delete */
1009 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
1010 extern void register_shrinker(struct shrinker *);
1011 extern void unregister_shrinker(struct shrinker *);
1013 int vma_wants_writenotify(struct vm_area_struct *vma);
1015 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
1017 #ifdef __PAGETABLE_PUD_FOLDED
1018 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1019 unsigned long address)
1021 return 0;
1023 #else
1024 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1025 #endif
1027 #ifdef __PAGETABLE_PMD_FOLDED
1028 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1029 unsigned long address)
1031 return 0;
1033 #else
1034 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1035 #endif
1037 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1038 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1041 * The following ifdef needed to get the 4level-fixup.h header to work.
1042 * Remove it when 4level-fixup.h has been removed.
1044 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1045 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1047 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1048 NULL: pud_offset(pgd, address);
1051 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1053 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1054 NULL: pmd_offset(pud, address);
1056 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1058 #if USE_SPLIT_PTLOCKS
1060 * We tuck a spinlock to guard each pagetable page into its struct page,
1061 * at page->private, with BUILD_BUG_ON to make sure that this will not
1062 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1063 * When freeing, reset page->mapping so free_pages_check won't complain.
1065 #define __pte_lockptr(page) &((page)->ptl)
1066 #define pte_lock_init(_page) do { \
1067 spin_lock_init(__pte_lockptr(_page)); \
1068 } while (0)
1069 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1070 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1071 #else /* !USE_SPLIT_PTLOCKS */
1073 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1075 #define pte_lock_init(page) do {} while (0)
1076 #define pte_lock_deinit(page) do {} while (0)
1077 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1078 #endif /* USE_SPLIT_PTLOCKS */
1080 static inline void pgtable_page_ctor(struct page *page)
1082 pte_lock_init(page);
1083 inc_zone_page_state(page, NR_PAGETABLE);
1086 static inline void pgtable_page_dtor(struct page *page)
1088 pte_lock_deinit(page);
1089 dec_zone_page_state(page, NR_PAGETABLE);
1092 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1093 ({ \
1094 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1095 pte_t *__pte = pte_offset_map(pmd, address); \
1096 *(ptlp) = __ptl; \
1097 spin_lock(__ptl); \
1098 __pte; \
1101 #define pte_unmap_unlock(pte, ptl) do { \
1102 spin_unlock(ptl); \
1103 pte_unmap(pte); \
1104 } while (0)
1106 #define pte_alloc_map(mm, pmd, address) \
1107 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1108 NULL: pte_offset_map(pmd, address))
1110 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1111 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
1112 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1114 #define pte_alloc_kernel(pmd, address) \
1115 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1116 NULL: pte_offset_kernel(pmd, address))
1118 extern void free_area_init(unsigned long * zones_size);
1119 extern void free_area_init_node(int nid, unsigned long * zones_size,
1120 unsigned long zone_start_pfn, unsigned long *zholes_size);
1121 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1123 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1124 * zones, allocate the backing mem_map and account for memory holes in a more
1125 * architecture independent manner. This is a substitute for creating the
1126 * zone_sizes[] and zholes_size[] arrays and passing them to
1127 * free_area_init_node()
1129 * An architecture is expected to register range of page frames backed by
1130 * physical memory with add_active_range() before calling
1131 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1132 * usage, an architecture is expected to do something like
1134 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1135 * max_highmem_pfn};
1136 * for_each_valid_physical_page_range()
1137 * add_active_range(node_id, start_pfn, end_pfn)
1138 * free_area_init_nodes(max_zone_pfns);
1140 * If the architecture guarantees that there are no holes in the ranges
1141 * registered with add_active_range(), free_bootmem_active_regions()
1142 * will call free_bootmem_node() for each registered physical page range.
1143 * Similarly sparse_memory_present_with_active_regions() calls
1144 * memory_present() for each range when SPARSEMEM is enabled.
1146 * See mm/page_alloc.c for more information on each function exposed by
1147 * CONFIG_ARCH_POPULATES_NODE_MAP
1149 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1150 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1151 unsigned long end_pfn);
1152 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1153 unsigned long end_pfn);
1154 extern void remove_all_active_ranges(void);
1155 void sort_node_map(void);
1156 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1157 unsigned long end_pfn);
1158 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1159 unsigned long end_pfn);
1160 extern void get_pfn_range_for_nid(unsigned int nid,
1161 unsigned long *start_pfn, unsigned long *end_pfn);
1162 extern unsigned long find_min_pfn_with_active_regions(void);
1163 extern void free_bootmem_with_active_regions(int nid,
1164 unsigned long max_low_pfn);
1165 int add_from_early_node_map(struct range *range, int az,
1166 int nr_range, int nid);
1167 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align,
1168 u64 goal, u64 limit);
1169 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1170 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1171 extern void sparse_memory_present_with_active_regions(int nid);
1172 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1174 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1175 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1176 static inline int __early_pfn_to_nid(unsigned long pfn)
1178 return 0;
1180 #else
1181 /* please see mm/page_alloc.c */
1182 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1183 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1184 /* there is a per-arch backend function. */
1185 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1186 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1187 #endif
1189 extern void set_dma_reserve(unsigned long new_dma_reserve);
1190 extern void memmap_init_zone(unsigned long, int, unsigned long,
1191 unsigned long, enum memmap_context);
1192 extern void setup_per_zone_wmarks(void);
1193 extern void calculate_zone_inactive_ratio(struct zone *zone);
1194 extern void mem_init(void);
1195 extern void __init mmap_init(void);
1196 extern void show_mem(void);
1197 extern void si_meminfo(struct sysinfo * val);
1198 extern void si_meminfo_node(struct sysinfo *val, int nid);
1199 extern int after_bootmem;
1201 extern void setup_per_cpu_pageset(void);
1203 extern void zone_pcp_update(struct zone *zone);
1205 /* nommu.c */
1206 extern atomic_long_t mmap_pages_allocated;
1207 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1209 /* prio_tree.c */
1210 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1211 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1212 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1213 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1214 struct prio_tree_iter *iter);
1216 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1217 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1218 (vma = vma_prio_tree_next(vma, iter)); )
1220 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1221 struct list_head *list)
1223 vma->shared.vm_set.parent = NULL;
1224 list_add_tail(&vma->shared.vm_set.list, list);
1227 /* mmap.c */
1228 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1229 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1230 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1231 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1232 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1233 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1234 struct mempolicy *);
1235 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1236 extern int split_vma(struct mm_struct *,
1237 struct vm_area_struct *, unsigned long addr, int new_below);
1238 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1239 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1240 struct rb_node **, struct rb_node *);
1241 extern void unlink_file_vma(struct vm_area_struct *);
1242 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1243 unsigned long addr, unsigned long len, pgoff_t pgoff);
1244 extern void exit_mmap(struct mm_struct *);
1246 extern int mm_take_all_locks(struct mm_struct *mm);
1247 extern void mm_drop_all_locks(struct mm_struct *mm);
1249 #ifdef CONFIG_PROC_FS
1250 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1251 extern void added_exe_file_vma(struct mm_struct *mm);
1252 extern void removed_exe_file_vma(struct mm_struct *mm);
1253 #else
1254 static inline void added_exe_file_vma(struct mm_struct *mm)
1257 static inline void removed_exe_file_vma(struct mm_struct *mm)
1259 #endif /* CONFIG_PROC_FS */
1261 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1262 extern int install_special_mapping(struct mm_struct *mm,
1263 unsigned long addr, unsigned long len,
1264 unsigned long flags, struct page **pages);
1266 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1268 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1269 unsigned long len, unsigned long prot,
1270 unsigned long flag, unsigned long pgoff);
1271 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1272 unsigned long len, unsigned long flags,
1273 unsigned int vm_flags, unsigned long pgoff);
1275 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1276 unsigned long len, unsigned long prot,
1277 unsigned long flag, unsigned long offset)
1279 unsigned long ret = -EINVAL;
1280 if ((offset + PAGE_ALIGN(len)) < offset)
1281 goto out;
1282 if (!(offset & ~PAGE_MASK))
1283 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1284 out:
1285 return ret;
1288 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1290 extern unsigned long do_brk(unsigned long, unsigned long);
1292 /* filemap.c */
1293 extern unsigned long page_unuse(struct page *);
1294 extern void truncate_inode_pages(struct address_space *, loff_t);
1295 extern void truncate_inode_pages_range(struct address_space *,
1296 loff_t lstart, loff_t lend);
1298 /* generic vm_area_ops exported for stackable file systems */
1299 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1301 /* mm/page-writeback.c */
1302 int write_one_page(struct page *page, int wait);
1303 void task_dirty_inc(struct task_struct *tsk);
1305 /* readahead.c */
1306 #define VM_MAX_READAHEAD 128 /* kbytes */
1307 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1309 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1310 pgoff_t offset, unsigned long nr_to_read);
1312 void page_cache_sync_readahead(struct address_space *mapping,
1313 struct file_ra_state *ra,
1314 struct file *filp,
1315 pgoff_t offset,
1316 unsigned long size);
1318 void page_cache_async_readahead(struct address_space *mapping,
1319 struct file_ra_state *ra,
1320 struct file *filp,
1321 struct page *pg,
1322 pgoff_t offset,
1323 unsigned long size);
1325 unsigned long max_sane_readahead(unsigned long nr);
1326 unsigned long ra_submit(struct file_ra_state *ra,
1327 struct address_space *mapping,
1328 struct file *filp);
1330 /* Do stack extension */
1331 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1332 #ifdef CONFIG_IA64
1333 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1334 #endif
1335 extern int expand_stack_downwards(struct vm_area_struct *vma,
1336 unsigned long address);
1338 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1339 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1340 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1341 struct vm_area_struct **pprev);
1343 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1344 NULL if none. Assume start_addr < end_addr. */
1345 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1347 struct vm_area_struct * vma = find_vma(mm,start_addr);
1349 if (vma && end_addr <= vma->vm_start)
1350 vma = NULL;
1351 return vma;
1354 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1356 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1359 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1360 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1361 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1362 unsigned long pfn, unsigned long size, pgprot_t);
1363 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1364 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1365 unsigned long pfn);
1366 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1367 unsigned long pfn);
1369 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1370 unsigned int foll_flags);
1371 #define FOLL_WRITE 0x01 /* check pte is writable */
1372 #define FOLL_TOUCH 0x02 /* mark page accessed */
1373 #define FOLL_GET 0x04 /* do get_page on page */
1374 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1375 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1377 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1378 void *data);
1379 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1380 unsigned long size, pte_fn_t fn, void *data);
1382 #ifdef CONFIG_PROC_FS
1383 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1384 #else
1385 static inline void vm_stat_account(struct mm_struct *mm,
1386 unsigned long flags, struct file *file, long pages)
1389 #endif /* CONFIG_PROC_FS */
1391 #ifdef CONFIG_DEBUG_PAGEALLOC
1392 extern int debug_pagealloc_enabled;
1394 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1396 static inline void enable_debug_pagealloc(void)
1398 debug_pagealloc_enabled = 1;
1400 #ifdef CONFIG_HIBERNATION
1401 extern bool kernel_page_present(struct page *page);
1402 #endif /* CONFIG_HIBERNATION */
1403 #else
1404 static inline void
1405 kernel_map_pages(struct page *page, int numpages, int enable) {}
1406 static inline void enable_debug_pagealloc(void)
1409 #ifdef CONFIG_HIBERNATION
1410 static inline bool kernel_page_present(struct page *page) { return true; }
1411 #endif /* CONFIG_HIBERNATION */
1412 #endif
1414 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1415 #ifdef __HAVE_ARCH_GATE_AREA
1416 int in_gate_area_no_task(unsigned long addr);
1417 int in_gate_area(struct task_struct *task, unsigned long addr);
1418 #else
1419 int in_gate_area_no_task(unsigned long addr);
1420 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1421 #endif /* __HAVE_ARCH_GATE_AREA */
1423 int drop_caches_sysctl_handler(struct ctl_table *, int,
1424 void __user *, size_t *, loff_t *);
1425 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1426 unsigned long lru_pages);
1428 #ifndef CONFIG_MMU
1429 #define randomize_va_space 0
1430 #else
1431 extern int randomize_va_space;
1432 #endif
1434 const char * arch_vma_name(struct vm_area_struct *vma);
1435 void print_vma_addr(char *prefix, unsigned long rip);
1437 void sparse_mem_maps_populate_node(struct page **map_map,
1438 unsigned long pnum_begin,
1439 unsigned long pnum_end,
1440 unsigned long map_count,
1441 int nodeid);
1443 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1444 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1445 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1446 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1447 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1448 void *vmemmap_alloc_block(unsigned long size, int node);
1449 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1450 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1451 int vmemmap_populate_basepages(struct page *start_page,
1452 unsigned long pages, int node);
1453 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1454 void vmemmap_populate_print_last(void);
1457 enum mf_flags {
1458 MF_COUNT_INCREASED = 1 << 0,
1460 extern void memory_failure(unsigned long pfn, int trapno);
1461 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1462 extern int unpoison_memory(unsigned long pfn);
1463 extern int sysctl_memory_failure_early_kill;
1464 extern int sysctl_memory_failure_recovery;
1465 extern void shake_page(struct page *p, int access);
1466 extern atomic_long_t mce_bad_pages;
1467 extern int soft_offline_page(struct page *page, int flags);
1469 extern void dump_page(struct page *page);
1471 #endif /* __KERNEL__ */
1472 #endif /* _LINUX_MM_H */