4 #include <linux/sched.h>
5 #include <linux/errno.h>
9 #include <linux/config.h>
10 #include <linux/gfp.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/vinline.h>
17 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
18 extern unsigned long max_mapnr;
21 extern unsigned long num_physpages;
22 extern void * high_memory;
23 extern int page_cluster;
26 #include <asm/pgtable.h>
27 #include <asm/processor.h>
28 #include <asm/atomic.h>
31 #define MM_VM_SIZE(mm) TASK_SIZE
35 * Linux kernel virtual memory manager primitives.
36 * The idea being to have a "virtual" mm in the same way
37 * we have a virtual fs - giving a cleaner interface to the
38 * mm details, and allowing different kinds of memory mappings
39 * (from shared memory to executable loading to arbitrary
44 * This struct defines a memory VMM memory area. There is one of these
45 * per VM-area/task. A VM area is any part of the process virtual memory
46 * space that has a special rule for the page-fault handlers (ie a shared
47 * library, the executable area etc).
49 * This structure is exactly 64 bytes on ia32. Please think very, very hard
50 * before adding anything to it.
52 struct vm_area_struct {
53 struct mm_struct * vm_mm; /* The address space we belong to. */
54 unsigned long vm_start; /* Our start address within vm_mm. */
55 unsigned long vm_end; /* The first byte after our end address
58 /* linked list of VM areas per task, sorted by address */
59 struct vm_area_struct *vm_next;
61 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
62 unsigned long vm_flags; /* Flags, listed below. */
67 * For areas with an address space and backing store,
68 * one of the address_space->i_mmap{,shared} lists,
69 * for shm areas, the list of attaches, otherwise unused.
71 struct list_head shared;
73 /* Function pointers to deal with this struct. */
74 struct vm_operations_struct * vm_ops;
76 /* Information about our backing store: */
77 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
78 units, *not* PAGE_CACHE_SIZE */
79 struct file * vm_file; /* File we map to (can be NULL). */
80 void * vm_private_data; /* was vm_pte (shared mem) */
86 #define VM_READ 0x00000001 /* currently active flags */
87 #define VM_WRITE 0x00000002
88 #define VM_EXEC 0x00000004
89 #define VM_SHARED 0x00000008
91 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
92 #define VM_MAYWRITE 0x00000020
93 #define VM_MAYEXEC 0x00000040
94 #define VM_MAYSHARE 0x00000080
96 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
97 #define VM_GROWSUP 0x00000200
98 #define VM_SHM 0x00000400 /* shared memory area, don't swap out */
99 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
101 #define VM_EXECUTABLE 0x00001000
102 #define VM_LOCKED 0x00002000
103 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
105 /* Used by sys_madvise() */
106 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
107 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
109 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
110 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
111 #define VM_RESERVED 0x00080000 /* Don't unmap it from swap_out */
112 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
113 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
114 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
116 /* It makes sense to apply VM_ACCOUNT to this vma. */
117 #define VM_MAYACCT(vma) (!!((vma)->vm_flags & VM_HUGETLB))
119 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
120 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
123 #ifdef CONFIG_STACK_GROWSUP
124 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
126 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
129 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
130 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
131 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
132 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
133 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
136 * mapping from the currently active vm_flags protection bits (the
137 * low four bits) to a page protection mask..
139 extern pgprot_t protection_map[16];
143 * These are the virtual MM functions - opening of an area, closing and
144 * unmapping it (needed to keep files on disk up-to-date etc), pointer
145 * to the functions called when a no-page or a wp-page exception occurs.
147 struct vm_operations_struct {
148 void (*open)(struct vm_area_struct * area);
149 void (*close)(struct vm_area_struct * area);
150 struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
151 int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
154 /* forward declaration; pte_chain is meant to be internal to rmap.c */
159 #ifdef ARCH_HAS_ATOMIC_UNSIGNED
160 typedef unsigned page_flags_t;
162 typedef unsigned long page_flags_t;
166 * Each physical page in the system has a struct page associated with
167 * it to keep track of whatever it is we are using the page for at the
168 * moment. Note that we have no way to track which tasks are using
171 * Try to keep the most commonly accessed fields in single cache lines
172 * here (16 bytes or greater). This ordering should be particularly
173 * beneficial on 32-bit processors.
175 * The first line is data used in page cache lookup, the second line
176 * is used for linear searches (eg. clock algorithm scans).
178 * TODO: make this structure smaller, it could be as small as 32 bytes.
181 page_flags_t flags; /* atomic flags, some possibly
182 updated asynchronously */
183 atomic_t count; /* Usage count, see below. */
184 struct address_space *mapping; /* The inode (or ...) we belong to. */
185 pgoff_t index; /* Our offset within mapping. */
186 struct list_head lru; /* Pageout list, eg. active_list;
187 protected by zone->lru_lock !! */
189 struct pte_chain *chain;/* Reverse pte mapping pointer.
190 * protected by PG_chainlock */
193 unsigned long private; /* Mapping-private opaque data:
194 * usually used for buffer_heads
195 * if PagePrivate set; used for
196 * swp_entry_t if PageSwapCache
199 * On machines where all RAM is mapped into kernel address space,
200 * we can simply calculate the virtual address. On machines with
201 * highmem some memory is mapped into kernel virtual memory
202 * dynamically, so we need a place to store that address.
203 * Note that this field could be 16 bits on x86 ... ;)
205 * Architectures with slow multiplication can define
206 * WANT_PAGE_VIRTUAL in asm/page.h
208 #if defined(WANT_PAGE_VIRTUAL)
209 void *virtual; /* Kernel virtual address (NULL if
210 not kmapped, ie. highmem) */
211 #endif /* WANT_PAGE_VIRTUAL */
215 * FIXME: take this include out, include page-flags.h in
216 * files which need it (119 of them)
218 #include <linux/page-flags.h>
221 * Methods to modify the page usage count.
223 * What counts for a page usage:
224 * - cache mapping (page->mapping)
225 * - private data (page->private)
226 * - page mapped in a task's page tables, each mapping
227 * is counted separately
229 * Also, many kernel routines increase the page count before a critical
230 * routine so they can be sure the page doesn't go away from under them.
232 #define put_page_testzero(p) \
234 BUG_ON(page_count(p) == 0); \
235 atomic_dec_and_test(&(p)->count); \
238 #define set_page_count(p,v) atomic_set(&(p)->count, v)
239 #define __put_page(p) atomic_dec(&(p)->count)
241 extern void FASTCALL(__page_cache_release(struct page *));
243 #ifdef CONFIG_HUGETLB_PAGE
245 static inline int page_count(struct page *p)
248 p = (struct page *)p->private;
249 return atomic_read(&(p)->count);
252 static inline void get_page(struct page *page)
254 if (unlikely(PageCompound(page)))
255 page = (struct page *)page->private;
256 atomic_inc(&page->count);
259 void put_page(struct page *page);
261 #else /* CONFIG_HUGETLB_PAGE */
263 #define page_count(p) atomic_read(&(p)->count)
265 static inline void get_page(struct page *page)
267 atomic_inc(&page->count);
270 static inline void put_page(struct page *page)
272 if (!PageReserved(page) && put_page_testzero(page))
273 __page_cache_release(page);
276 #endif /* CONFIG_HUGETLB_PAGE */
279 * Multiple processes may "see" the same page. E.g. for untouched
280 * mappings of /dev/null, all processes see the same page full of
281 * zeroes, and text pages of executables and shared libraries have
282 * only one copy in memory, at most, normally.
284 * For the non-reserved pages, page->count denotes a reference count.
285 * page->count == 0 means the page is free.
286 * page->count == 1 means the page is used for exactly one purpose
287 * (e.g. a private data page of one process).
289 * A page may be used for kmalloc() or anyone else who does a
290 * __get_free_page(). In this case the page->count is at least 1, and
291 * all other fields are unused but should be 0 or NULL. The
292 * management of this page is the responsibility of the one who uses
295 * The other pages (we may call them "process pages") are completely
296 * managed by the Linux memory manager: I/O, buffers, swapping etc.
297 * The following discussion applies only to them.
299 * A page may belong to an inode's memory mapping. In this case,
300 * page->mapping is the pointer to the inode, and page->index is the
301 * file offset of the page, in units of PAGE_CACHE_SIZE.
303 * A page contains an opaque `private' member, which belongs to the
304 * page's address_space. Usually, this is the address of a circular
305 * list of the page's disk buffers.
307 * For pages belonging to inodes, the page->count is the number of
308 * attaches, plus 1 if `private' contains something, plus one for
309 * the page cache itself.
311 * All pages belonging to an inode are in these doubly linked lists:
312 * mapping->clean_pages, mapping->dirty_pages and mapping->locked_pages;
313 * using the page->list list_head. These fields are also used for
314 * freelist managemet (when page->count==0).
316 * There is also a per-mapping radix tree mapping index to the page
317 * in memory if present. The tree is rooted at mapping->root.
319 * All process pages can do I/O:
320 * - inode pages may need to be read from disk,
321 * - inode pages which have been modified and are MAP_SHARED may need
322 * to be written to disk,
323 * - private pages which have been modified may need to be swapped out
324 * to swap space and (later) to be read back into memory.
328 * The zone field is never updated after free_area_init_core()
329 * sets it, so none of the operations on it need to be atomic.
330 * We'll have up to (MAX_NUMNODES * MAX_NR_ZONES) zones total,
331 * so we use (MAX_NODES_SHIFT + MAX_ZONES_SHIFT) here to get enough bits.
333 #define NODEZONE_SHIFT (sizeof(page_flags_t)*8 - MAX_NODES_SHIFT - MAX_ZONES_SHIFT)
334 #define NODEZONE(node, zone) ((node << ZONES_SHIFT) | zone)
336 static inline unsigned long page_zonenum(struct page *page)
338 return (page->flags >> NODEZONE_SHIFT) & (~(~0UL << ZONES_SHIFT));
340 static inline unsigned long page_to_nid(struct page *page)
342 return (page->flags >> (NODEZONE_SHIFT + ZONES_SHIFT));
346 extern struct zone *zone_table[];
348 static inline struct zone *page_zone(struct page *page)
350 return zone_table[page->flags >> NODEZONE_SHIFT];
353 static inline void set_page_zone(struct page *page, unsigned long nodezone_num)
355 page->flags &= ~(~0UL << NODEZONE_SHIFT);
356 page->flags |= nodezone_num << NODEZONE_SHIFT;
359 #ifndef CONFIG_DISCONTIGMEM
360 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
361 extern struct page *mem_map;
364 static inline void *lowmem_page_address(struct page *page)
366 return __va(page_to_pfn(page) << PAGE_SHIFT);
369 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
370 #define HASHED_PAGE_VIRTUAL
373 #if defined(WANT_PAGE_VIRTUAL)
374 #define page_address(page) ((page)->virtual)
375 #define set_page_address(page, address) \
377 (page)->virtual = (address); \
379 #define page_address_init() do { } while(0)
382 #if defined(HASHED_PAGE_VIRTUAL)
383 void *page_address(struct page *page);
384 void set_page_address(struct page *page, void *virtual);
385 void page_address_init(void);
388 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
389 #define page_address(page) lowmem_page_address(page)
390 #define set_page_address(page, address) do { } while(0)
391 #define page_address_init() do { } while(0)
395 * On an anonymous page mapped into a user virtual memory area,
396 * page->mapping points to its anon_vma, not to a struct address_space.
398 * Please note that, confusingly, "page_mapping" refers to the inode
399 * address_space which maps the page from disk; whereas "page_mapped"
400 * refers to user virtual address space into which the page is mapped.
402 static inline struct address_space *page_mapping(struct page *page)
404 return PageAnon(page)? NULL: page->mapping;
408 * Return true if this page is mapped into pagetables. Subtle: test pte.direct
409 * rather than pte.chain. Because sometimes pte.direct is 64-bit, and .chain
412 static inline int page_mapped(struct page *page)
414 return page->pte.direct != 0;
418 * Error return values for the *_nopage functions
420 #define NOPAGE_SIGBUS (NULL)
421 #define NOPAGE_OOM ((struct page *) (-1))
424 * Different kinds of faults, as returned by handle_mm_fault().
425 * Used to decide whether a process gets delivered SIGBUS or
426 * just gets major/minor fault counters bumped up.
428 #define VM_FAULT_OOM (-1)
429 #define VM_FAULT_SIGBUS 0
430 #define VM_FAULT_MINOR 1
431 #define VM_FAULT_MAJOR 2
433 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
435 extern void show_free_areas(void);
437 struct page *shmem_nopage(struct vm_area_struct * vma,
438 unsigned long address, int *type);
439 struct file *shmem_file_setup(char * name, loff_t size, unsigned long flags);
440 void shmem_lock(struct file * file, int lock);
441 int shmem_zero_setup(struct vm_area_struct *);
444 * Parameter block passed down to zap_pte_range in exceptional cases.
447 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
448 struct address_space *check_mapping; /* Check page->mapping if set */
449 pgoff_t first_index; /* Lowest page->index to unmap */
450 pgoff_t last_index; /* Highest page->index to unmap */
453 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
454 unsigned long size, struct zap_details *);
455 int unmap_vmas(struct mmu_gather **tlbp, struct mm_struct *mm,
456 struct vm_area_struct *start_vma, unsigned long start_addr,
457 unsigned long end_addr, unsigned long *nr_accounted,
458 struct zap_details *);
459 void clear_page_tables(struct mmu_gather *tlb, unsigned long first, int nr);
460 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
461 struct vm_area_struct *vma);
462 int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
463 unsigned long size, pgprot_t prot);
464 void unmap_mapping_range(struct address_space *mapping,
465 loff_t const holebegin, loff_t const holelen, int even_cows);
467 static inline void unmap_shared_mapping_range(struct address_space *mapping,
468 loff_t const holebegin, loff_t const holelen)
470 unmap_mapping_range(mapping, holebegin, holelen, 0);
473 extern int vmtruncate(struct inode * inode, loff_t offset);
474 extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address));
475 extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address));
476 extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address));
477 extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
478 extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
479 extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access);
480 extern int make_pages_present(unsigned long addr, unsigned long end);
481 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
482 void put_dirty_page(struct task_struct *tsk, struct page *page,
483 unsigned long address, pgprot_t prot);
485 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
486 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
488 int __set_page_dirty_buffers(struct page *page);
489 int __set_page_dirty_nobuffers(struct page *page);
490 int redirty_page_for_writepage(struct writeback_control *wbc,
492 int FASTCALL(set_page_dirty(struct page *page));
493 int set_page_dirty_lock(struct page *page);
494 int clear_page_dirty_for_io(struct page *page);
497 * Prototype to add a shrinker callback for ageable caches.
499 * These functions are passed a count `nr_to_scan' and a gfpmask. They should
500 * scan `nr_to_scan' objects, attempting to free them.
502 * The callback must the number of objects which remain in the cache.
504 * The callback will be passes nr_to_scan == 0 when the VM is querying the
505 * cache size, so a fastpath for that case is appropriate.
507 typedef int (*shrinker_t)(int nr_to_scan, unsigned int gfp_mask);
510 * Add an aging callback. The int is the number of 'seeks' it takes
511 * to recreate one of the objects that these functions age.
514 #define DEFAULT_SEEKS 2
516 extern struct shrinker *set_shrinker(int, shrinker_t);
517 extern void remove_shrinker(struct shrinker *shrinker);
520 * On a two-level page table, this ends up being trivial. Thus the
521 * inlining and the symmetry break with pte_alloc_map() that does all
522 * of this out-of-line.
524 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
527 return __pmd_alloc(mm, pgd, address);
528 return pmd_offset(pgd, address);
531 extern void free_area_init(unsigned long * zones_size);
532 extern void free_area_init_node(int nid, pg_data_t *pgdat, struct page *pmap,
533 unsigned long * zones_size, unsigned long zone_start_pfn,
534 unsigned long *zholes_size);
535 extern void memmap_init_zone(struct page *, unsigned long, int,
536 unsigned long, unsigned long);
537 extern void mem_init(void);
538 extern void show_mem(void);
539 extern void si_meminfo(struct sysinfo * val);
540 extern void si_meminfo_node(struct sysinfo *val, int nid);
543 extern void insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
544 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
545 struct rb_node **, struct rb_node *);
546 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
547 unsigned long addr, unsigned long len, unsigned long pgoff);
548 extern void exit_mmap(struct mm_struct *);
550 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
552 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
553 unsigned long len, unsigned long prot,
554 unsigned long flag, unsigned long pgoff);
556 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
557 unsigned long len, unsigned long prot,
558 unsigned long flag, unsigned long offset)
560 unsigned long ret = -EINVAL;
561 if ((offset + PAGE_ALIGN(len)) < offset)
563 if (!(offset & ~PAGE_MASK))
564 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
569 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
571 extern unsigned long do_brk(unsigned long, unsigned long);
574 __vma_unlink(struct mm_struct *mm, struct vm_area_struct *vma,
575 struct vm_area_struct *prev)
577 prev->vm_next = vma->vm_next;
578 rb_erase(&vma->vm_rb, &mm->mm_rb);
579 if (mm->mmap_cache == vma)
580 mm->mmap_cache = prev;
584 can_vma_merge(struct vm_area_struct *vma, unsigned long vm_flags)
587 if (!vma->vm_file && vma->vm_flags == vm_flags)
594 extern unsigned long page_unuse(struct page *);
595 extern void truncate_inode_pages(struct address_space *, loff_t);
597 /* generic vm_area_ops exported for stackable file systems */
598 struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
600 /* mm/page-writeback.c */
601 int write_one_page(struct page *page, int wait);
604 #define VM_MAX_READAHEAD 128 /* kbytes */
605 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
607 int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
608 unsigned long offset, unsigned long nr_to_read);
609 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
610 unsigned long offset, unsigned long nr_to_read);
611 void page_cache_readahead(struct address_space *mapping,
612 struct file_ra_state *ra,
614 unsigned long offset);
615 void handle_ra_miss(struct address_space *mapping,
616 struct file_ra_state *ra, pgoff_t offset);
617 unsigned long max_sane_readahead(unsigned long nr);
619 /* Do stack extension */
620 extern int expand_stack(struct vm_area_struct * vma, unsigned long address);
622 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
623 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
624 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
625 struct vm_area_struct **pprev);
626 extern int split_vma(struct mm_struct * mm, struct vm_area_struct * vma,
627 unsigned long addr, int new_below);
629 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
630 NULL if none. Assume start_addr < end_addr. */
631 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
633 struct vm_area_struct * vma = find_vma(mm,start_addr);
635 if (vma && end_addr <= vma->vm_start)
640 extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr);
642 extern unsigned int nr_used_zone_pages(void);
644 extern struct page * vmalloc_to_page(void *addr);
645 extern struct page * follow_page(struct mm_struct *mm, unsigned long address,
647 extern int remap_page_range(struct vm_area_struct *vma, unsigned long from,
648 unsigned long to, unsigned long size, pgprot_t prot);
650 #ifndef CONFIG_DEBUG_PAGEALLOC
652 kernel_map_pages(struct page *page, int numpages, int enable)
657 #ifndef CONFIG_ARCH_GATE_AREA
658 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
659 int in_gate_area(struct task_struct *task, unsigned long addr);
662 #endif /* __KERNEL__ */
663 #endif /* _LINUX_MM_H */