1 #ifndef _LINUX_MM_H 2 #define _LINUX_MM_H 3 4 #include <linux/errno.h> 5 6 #ifdef __KERNEL__ 7 8 #include <linux/gfp.h> 9 #include <linux/list.h> 10 #include <linux/mmdebug.h> 11 #include <linux/mmzone.h> 12 #include <linux/rbtree.h> 13 #include <linux/prio_tree.h> 14 #include <linux/debug_locks.h> 15 #include <linux/mm_types.h> 16 17 struct mempolicy; 18 struct anon_vma; 19 struct file_ra_state; 20 struct user_struct; 21 struct writeback_control; 22 23 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 24 extern unsigned long max_mapnr; 25 #endif 26 27 extern unsigned long num_physpages; 28 extern void * high_memory; 29 extern int page_cluster; 30 31 #ifdef CONFIG_SYSCTL 32 extern int sysctl_legacy_va_layout; 33 #else 34 #define sysctl_legacy_va_layout 0 35 #endif 36 37 extern unsigned long mmap_min_addr; 38 39 #include <asm/page.h> 40 #include <asm/pgtable.h> 41 #include <asm/processor.h> 42 43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 44 45 /* to align the pointer to the (next) page boundary */ 46 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 47 48 /* 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). 55 */ 56 57 extern struct kmem_cache *vm_area_cachep; 58 59 #ifndef CONFIG_MMU 60 extern struct rb_root nommu_region_tree; 61 extern struct rw_semaphore nommu_region_sem; 62 63 extern unsigned int kobjsize(const void *objp); 64 #endif 65 66 /* 67 * vm_flags in vm_area_struct, see mm_types.h. 68 */ 69 #define VM_READ 0x00000001 /* currently active flags */ 70 #define VM_WRITE 0x00000002 71 #define VM_EXEC 0x00000004 72 #define VM_SHARED 0x00000008 73 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 79 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.. */ 84 85 #define VM_EXECUTABLE 0x00001000 86 #define VM_LOCKED 0x00002000 87 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 88 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 */ 92 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 */ 103 104 #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) */ 107 108 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 109 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 110 #endif 111 112 #ifdef CONFIG_STACK_GROWSUP 113 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 114 #else 115 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 116 #endif 117 118 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 119 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 120 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 121 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 122 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 123 124 /* 125 * special vmas that are non-mergable, non-mlock()able 126 */ 127 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 128 129 /* 130 * mapping from the currently active vm_flags protection bits (the 131 * low four bits) to a page protection mask.. 132 */ 133 extern pgprot_t protection_map[16]; 134 135 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 136 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 137 138 /* 139 * This interface is used by x86 PAT code to identify a pfn mapping that is 140 * linear over entire vma. This is to optimize PAT code that deals with 141 * marking the physical region with a particular prot. This is not for generic 142 * mm use. Note also that this check will not work if the pfn mapping is 143 * linear for a vma starting at physical address 0. In which case PAT code 144 * falls back to slow path of reserving physical range page by page. 145 */ 146 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 147 { 148 return ((vma->vm_flags & VM_PFNMAP) && vma->vm_pgoff); 149 } 150 151 static inline int is_pfn_mapping(struct vm_area_struct *vma) 152 { 153 return (vma->vm_flags & VM_PFNMAP); 154 } 155 156 /* 157 * vm_fault is filled by the the pagefault handler and passed to the vma's 158 * ->fault function. The vma's ->fault is responsible for returning a bitmask 159 * of VM_FAULT_xxx flags that give details about how the fault was handled. 160 * 161 * pgoff should be used in favour of virtual_address, if possible. If pgoff 162 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 163 * mapping support. 164 */ 165 struct vm_fault { 166 unsigned int flags; /* FAULT_FLAG_xxx flags */ 167 pgoff_t pgoff; /* Logical page offset based on vma */ 168 void __user *virtual_address; /* Faulting virtual address */ 169 170 struct page *page; /* ->fault handlers should return a 171 * page here, unless VM_FAULT_NOPAGE 172 * is set (which is also implied by 173 * VM_FAULT_ERROR). 174 */ 175 }; 176 177 /* 178 * These are the virtual MM functions - opening of an area, closing and 179 * unmapping it (needed to keep files on disk up-to-date etc), pointer 180 * to the functions called when a no-page or a wp-page exception occurs. 181 */ 182 struct vm_operations_struct { 183 void (*open)(struct vm_area_struct * area); 184 void (*close)(struct vm_area_struct * area); 185 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 186 187 /* notification that a previously read-only page is about to become 188 * writable, if an error is returned it will cause a SIGBUS */ 189 int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page); 190 191 /* called by access_process_vm when get_user_pages() fails, typically 192 * for use by special VMAs that can switch between memory and hardware 193 */ 194 int (*access)(struct vm_area_struct *vma, unsigned long addr, 195 void *buf, int len, int write); 196 #ifdef CONFIG_NUMA 197 /* 198 * set_policy() op must add a reference to any non-NULL @new mempolicy 199 * to hold the policy upon return. Caller should pass NULL @new to 200 * remove a policy and fall back to surrounding context--i.e. do not 201 * install a MPOL_DEFAULT policy, nor the task or system default 202 * mempolicy. 203 */ 204 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 205 206 /* 207 * get_policy() op must add reference [mpol_get()] to any policy at 208 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 209 * in mm/mempolicy.c will do this automatically. 210 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 211 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 212 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 213 * must return NULL--i.e., do not "fallback" to task or system default 214 * policy. 215 */ 216 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 217 unsigned long addr); 218 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 219 const nodemask_t *to, unsigned long flags); 220 #endif 221 }; 222 223 struct mmu_gather; 224 struct inode; 225 226 #define page_private(page) ((page)->private) 227 #define set_page_private(page, v) ((page)->private = (v)) 228 229 /* 230 * FIXME: take this include out, include page-flags.h in 231 * files which need it (119 of them) 232 */ 233 #include <linux/page-flags.h> 234 235 /* 236 * Methods to modify the page usage count. 237 * 238 * What counts for a page usage: 239 * - cache mapping (page->mapping) 240 * - private data (page->private) 241 * - page mapped in a task's page tables, each mapping 242 * is counted separately 243 * 244 * Also, many kernel routines increase the page count before a critical 245 * routine so they can be sure the page doesn't go away from under them. 246 */ 247 248 /* 249 * Drop a ref, return true if the refcount fell to zero (the page has no users) 250 */ 251 static inline int put_page_testzero(struct page *page) 252 { 253 VM_BUG_ON(atomic_read(&page->_count) == 0); 254 return atomic_dec_and_test(&page->_count); 255 } 256 257 /* 258 * Try to grab a ref unless the page has a refcount of zero, return false if 259 * that is the case. 260 */ 261 static inline int get_page_unless_zero(struct page *page) 262 { 263 VM_BUG_ON(PageTail(page)); 264 return atomic_inc_not_zero(&page->_count); 265 } 266 267 /* Support for virtually mapped pages */ 268 struct page *vmalloc_to_page(const void *addr); 269 unsigned long vmalloc_to_pfn(const void *addr); 270 271 /* 272 * Determine if an address is within the vmalloc range 273 * 274 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 275 * is no special casing required. 276 */ 277 static inline int is_vmalloc_addr(const void *x) 278 { 279 #ifdef CONFIG_MMU 280 unsigned long addr = (unsigned long)x; 281 282 return addr >= VMALLOC_START && addr < VMALLOC_END; 283 #else 284 return 0; 285 #endif 286 } 287 288 static inline struct page *compound_head(struct page *page) 289 { 290 if (unlikely(PageTail(page))) 291 return page->first_page; 292 return page; 293 } 294 295 static inline int page_count(struct page *page) 296 { 297 return atomic_read(&compound_head(page)->_count); 298 } 299 300 static inline void get_page(struct page *page) 301 { 302 page = compound_head(page); 303 VM_BUG_ON(atomic_read(&page->_count) == 0); 304 atomic_inc(&page->_count); 305 } 306 307 static inline struct page *virt_to_head_page(const void *x) 308 { 309 struct page *page = virt_to_page(x); 310 return compound_head(page); 311 } 312 313 /* 314 * Setup the page count before being freed into the page allocator for 315 * the first time (boot or memory hotplug) 316 */ 317 static inline void init_page_count(struct page *page) 318 { 319 atomic_set(&page->_count, 1); 320 } 321 322 void put_page(struct page *page); 323 void put_pages_list(struct list_head *pages); 324 325 void split_page(struct page *page, unsigned int order); 326 327 /* 328 * Compound pages have a destructor function. Provide a 329 * prototype for that function and accessor functions. 330 * These are _only_ valid on the head of a PG_compound page. 331 */ 332 typedef void compound_page_dtor(struct page *); 333 334 static inline void set_compound_page_dtor(struct page *page, 335 compound_page_dtor *dtor) 336 { 337 page[1].lru.next = (void *)dtor; 338 } 339 340 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 341 { 342 return (compound_page_dtor *)page[1].lru.next; 343 } 344 345 static inline int compound_order(struct page *page) 346 { 347 if (!PageHead(page)) 348 return 0; 349 return (unsigned long)page[1].lru.prev; 350 } 351 352 static inline void set_compound_order(struct page *page, unsigned long order) 353 { 354 page[1].lru.prev = (void *)order; 355 } 356 357 /* 358 * Multiple processes may "see" the same page. E.g. for untouched 359 * mappings of /dev/null, all processes see the same page full of 360 * zeroes, and text pages of executables and shared libraries have 361 * only one copy in memory, at most, normally. 362 * 363 * For the non-reserved pages, page_count(page) denotes a reference count. 364 * page_count() == 0 means the page is free. page->lru is then used for 365 * freelist management in the buddy allocator. 366 * page_count() > 0 means the page has been allocated. 367 * 368 * Pages are allocated by the slab allocator in order to provide memory 369 * to kmalloc and kmem_cache_alloc. In this case, the management of the 370 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 371 * unless a particular usage is carefully commented. (the responsibility of 372 * freeing the kmalloc memory is the caller's, of course). 373 * 374 * A page may be used by anyone else who does a __get_free_page(). 375 * In this case, page_count still tracks the references, and should only 376 * be used through the normal accessor functions. The top bits of page->flags 377 * and page->virtual store page management information, but all other fields 378 * are unused and could be used privately, carefully. The management of this 379 * page is the responsibility of the one who allocated it, and those who have 380 * subsequently been given references to it. 381 * 382 * The other pages (we may call them "pagecache pages") are completely 383 * managed by the Linux memory manager: I/O, buffers, swapping etc. 384 * The following discussion applies only to them. 385 * 386 * A pagecache page contains an opaque `private' member, which belongs to the 387 * page's address_space. Usually, this is the address of a circular list of 388 * the page's disk buffers. PG_private must be set to tell the VM to call 389 * into the filesystem to release these pages. 390 * 391 * A page may belong to an inode's memory mapping. In this case, page->mapping 392 * is the pointer to the inode, and page->index is the file offset of the page, 393 * in units of PAGE_CACHE_SIZE. 394 * 395 * If pagecache pages are not associated with an inode, they are said to be 396 * anonymous pages. These may become associated with the swapcache, and in that 397 * case PG_swapcache is set, and page->private is an offset into the swapcache. 398 * 399 * In either case (swapcache or inode backed), the pagecache itself holds one 400 * reference to the page. Setting PG_private should also increment the 401 * refcount. The each user mapping also has a reference to the page. 402 * 403 * The pagecache pages are stored in a per-mapping radix tree, which is 404 * rooted at mapping->page_tree, and indexed by offset. 405 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 406 * lists, we instead now tag pages as dirty/writeback in the radix tree. 407 * 408 * All pagecache pages may be subject to I/O: 409 * - inode pages may need to be read from disk, 410 * - inode pages which have been modified and are MAP_SHARED may need 411 * to be written back to the inode on disk, 412 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 413 * modified may need to be swapped out to swap space and (later) to be read 414 * back into memory. 415 */ 416 417 /* 418 * The zone field is never updated after free_area_init_core() 419 * sets it, so none of the operations on it need to be atomic. 420 */ 421 422 423 /* 424 * page->flags layout: 425 * 426 * There are three possibilities for how page->flags get 427 * laid out. The first is for the normal case, without 428 * sparsemem. The second is for sparsemem when there is 429 * plenty of space for node and section. The last is when 430 * we have run out of space and have to fall back to an 431 * alternate (slower) way of determining the node. 432 * 433 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 434 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 435 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 436 */ 437 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 438 #define SECTIONS_WIDTH SECTIONS_SHIFT 439 #else 440 #define SECTIONS_WIDTH 0 441 #endif 442 443 #define ZONES_WIDTH ZONES_SHIFT 444 445 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 446 #define NODES_WIDTH NODES_SHIFT 447 #else 448 #ifdef CONFIG_SPARSEMEM_VMEMMAP 449 #error "Vmemmap: No space for nodes field in page flags" 450 #endif 451 #define NODES_WIDTH 0 452 #endif 453 454 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 455 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 456 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 457 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 458 459 /* 460 * We are going to use the flags for the page to node mapping if its in 461 * there. This includes the case where there is no node, so it is implicit. 462 */ 463 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 464 #define NODE_NOT_IN_PAGE_FLAGS 465 #endif 466 467 #ifndef PFN_SECTION_SHIFT 468 #define PFN_SECTION_SHIFT 0 469 #endif 470 471 /* 472 * Define the bit shifts to access each section. For non-existant 473 * sections we define the shift as 0; that plus a 0 mask ensures 474 * the compiler will optimise away reference to them. 475 */ 476 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 477 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 478 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 479 480 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 481 #ifdef NODE_NOT_IN_PAGEFLAGS 482 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 483 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 484 SECTIONS_PGOFF : ZONES_PGOFF) 485 #else 486 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 487 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 488 NODES_PGOFF : ZONES_PGOFF) 489 #endif 490 491 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 492 493 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 494 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 495 #endif 496 497 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 498 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 499 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 500 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 501 502 static inline enum zone_type page_zonenum(struct page *page) 503 { 504 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 505 } 506 507 /* 508 * The identification function is only used by the buddy allocator for 509 * determining if two pages could be buddies. We are not really 510 * identifying a zone since we could be using a the section number 511 * id if we have not node id available in page flags. 512 * We guarantee only that it will return the same value for two 513 * combinable pages in a zone. 514 */ 515 static inline int page_zone_id(struct page *page) 516 { 517 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 518 } 519 520 static inline int zone_to_nid(struct zone *zone) 521 { 522 #ifdef CONFIG_NUMA 523 return zone->node; 524 #else 525 return 0; 526 #endif 527 } 528 529 #ifdef NODE_NOT_IN_PAGE_FLAGS 530 extern int page_to_nid(struct page *page); 531 #else 532 static inline int page_to_nid(struct page *page) 533 { 534 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 535 } 536 #endif 537 538 static inline struct zone *page_zone(struct page *page) 539 { 540 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 541 } 542 543 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 544 static inline unsigned long page_to_section(struct page *page) 545 { 546 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 547 } 548 #endif 549 550 static inline void set_page_zone(struct page *page, enum zone_type zone) 551 { 552 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 553 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 554 } 555 556 static inline void set_page_node(struct page *page, unsigned long node) 557 { 558 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 559 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 560 } 561 562 static inline void set_page_section(struct page *page, unsigned long section) 563 { 564 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 565 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 566 } 567 568 static inline void set_page_links(struct page *page, enum zone_type zone, 569 unsigned long node, unsigned long pfn) 570 { 571 set_page_zone(page, zone); 572 set_page_node(page, node); 573 set_page_section(page, pfn_to_section_nr(pfn)); 574 } 575 576 /* 577 * If a hint addr is less than mmap_min_addr change hint to be as 578 * low as possible but still greater than mmap_min_addr 579 */ 580 static inline unsigned long round_hint_to_min(unsigned long hint) 581 { 582 #ifdef CONFIG_SECURITY 583 hint &= PAGE_MASK; 584 if (((void *)hint != NULL) && 585 (hint < mmap_min_addr)) 586 return PAGE_ALIGN(mmap_min_addr); 587 #endif 588 return hint; 589 } 590 591 /* 592 * Some inline functions in vmstat.h depend on page_zone() 593 */ 594 #include <linux/vmstat.h> 595 596 static __always_inline void *lowmem_page_address(struct page *page) 597 { 598 return __va(page_to_pfn(page) << PAGE_SHIFT); 599 } 600 601 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 602 #define HASHED_PAGE_VIRTUAL 603 #endif 604 605 #if defined(WANT_PAGE_VIRTUAL) 606 #define page_address(page) ((page)->virtual) 607 #define set_page_address(page, address) \ 608 do { \ 609 (page)->virtual = (address); \ 610 } while(0) 611 #define page_address_init() do { } while(0) 612 #endif 613 614 #if defined(HASHED_PAGE_VIRTUAL) 615 void *page_address(struct page *page); 616 void set_page_address(struct page *page, void *virtual); 617 void page_address_init(void); 618 #endif 619 620 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 621 #define page_address(page) lowmem_page_address(page) 622 #define set_page_address(page, address) do { } while(0) 623 #define page_address_init() do { } while(0) 624 #endif 625 626 /* 627 * On an anonymous page mapped into a user virtual memory area, 628 * page->mapping points to its anon_vma, not to a struct address_space; 629 * with the PAGE_MAPPING_ANON bit set to distinguish it. 630 * 631 * Please note that, confusingly, "page_mapping" refers to the inode 632 * address_space which maps the page from disk; whereas "page_mapped" 633 * refers to user virtual address space into which the page is mapped. 634 */ 635 #define PAGE_MAPPING_ANON 1 636 637 extern struct address_space swapper_space; 638 static inline struct address_space *page_mapping(struct page *page) 639 { 640 struct address_space *mapping = page->mapping; 641 642 VM_BUG_ON(PageSlab(page)); 643 #ifdef CONFIG_SWAP 644 if (unlikely(PageSwapCache(page))) 645 mapping = &swapper_space; 646 else 647 #endif 648 if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 649 mapping = NULL; 650 return mapping; 651 } 652 653 static inline int PageAnon(struct page *page) 654 { 655 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 656 } 657 658 /* 659 * Return the pagecache index of the passed page. Regular pagecache pages 660 * use ->index whereas swapcache pages use ->private 661 */ 662 static inline pgoff_t page_index(struct page *page) 663 { 664 if (unlikely(PageSwapCache(page))) 665 return page_private(page); 666 return page->index; 667 } 668 669 /* 670 * The atomic page->_mapcount, like _count, starts from -1: 671 * so that transitions both from it and to it can be tracked, 672 * using atomic_inc_and_test and atomic_add_negative(-1). 673 */ 674 static inline void reset_page_mapcount(struct page *page) 675 { 676 atomic_set(&(page)->_mapcount, -1); 677 } 678 679 static inline int page_mapcount(struct page *page) 680 { 681 return atomic_read(&(page)->_mapcount) + 1; 682 } 683 684 /* 685 * Return true if this page is mapped into pagetables. 686 */ 687 static inline int page_mapped(struct page *page) 688 { 689 return atomic_read(&(page)->_mapcount) >= 0; 690 } 691 692 /* 693 * Different kinds of faults, as returned by handle_mm_fault(). 694 * Used to decide whether a process gets delivered SIGBUS or 695 * just gets major/minor fault counters bumped up. 696 */ 697 698 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 699 700 #define VM_FAULT_OOM 0x0001 701 #define VM_FAULT_SIGBUS 0x0002 702 #define VM_FAULT_MAJOR 0x0004 703 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 704 705 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 706 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 707 708 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS) 709 710 /* 711 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 712 */ 713 extern void pagefault_out_of_memory(void); 714 715 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 716 717 extern void show_free_areas(void); 718 719 #ifdef CONFIG_SHMEM 720 extern int shmem_lock(struct file *file, int lock, struct user_struct *user); 721 #else 722 static inline int shmem_lock(struct file *file, int lock, 723 struct user_struct *user) 724 { 725 return 0; 726 } 727 #endif 728 struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); 729 730 int shmem_zero_setup(struct vm_area_struct *); 731 732 #ifndef CONFIG_MMU 733 extern unsigned long shmem_get_unmapped_area(struct file *file, 734 unsigned long addr, 735 unsigned long len, 736 unsigned long pgoff, 737 unsigned long flags); 738 #endif 739 740 extern int can_do_mlock(void); 741 extern int user_shm_lock(size_t, struct user_struct *); 742 extern void user_shm_unlock(size_t, struct user_struct *); 743 744 /* 745 * Parameter block passed down to zap_pte_range in exceptional cases. 746 */ 747 struct zap_details { 748 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 749 struct address_space *check_mapping; /* Check page->mapping if set */ 750 pgoff_t first_index; /* Lowest page->index to unmap */ 751 pgoff_t last_index; /* Highest page->index to unmap */ 752 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 753 unsigned long truncate_count; /* Compare vm_truncate_count */ 754 }; 755 756 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 757 pte_t pte); 758 759 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 760 unsigned long size); 761 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 762 unsigned long size, struct zap_details *); 763 unsigned long unmap_vmas(struct mmu_gather **tlb, 764 struct vm_area_struct *start_vma, unsigned long start_addr, 765 unsigned long end_addr, unsigned long *nr_accounted, 766 struct zap_details *); 767 768 /** 769 * mm_walk - callbacks for walk_page_range 770 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 771 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 772 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 773 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 774 * @pte_hole: if set, called for each hole at all levels 775 * 776 * (see walk_page_range for more details) 777 */ 778 struct mm_walk { 779 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 780 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 781 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 782 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 783 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 784 struct mm_struct *mm; 785 void *private; 786 }; 787 788 int walk_page_range(unsigned long addr, unsigned long end, 789 struct mm_walk *walk); 790 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 791 unsigned long end, unsigned long floor, unsigned long ceiling); 792 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 793 struct vm_area_struct *vma); 794 void unmap_mapping_range(struct address_space *mapping, 795 loff_t const holebegin, loff_t const holelen, int even_cows); 796 int follow_phys(struct vm_area_struct *vma, unsigned long address, 797 unsigned int flags, unsigned long *prot, resource_size_t *phys); 798 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 799 void *buf, int len, int write); 800 801 static inline void unmap_shared_mapping_range(struct address_space *mapping, 802 loff_t const holebegin, loff_t const holelen) 803 { 804 unmap_mapping_range(mapping, holebegin, holelen, 0); 805 } 806 807 extern int vmtruncate(struct inode * inode, loff_t offset); 808 extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end); 809 810 #ifdef CONFIG_MMU 811 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 812 unsigned long address, int write_access); 813 #else 814 static inline int handle_mm_fault(struct mm_struct *mm, 815 struct vm_area_struct *vma, unsigned long address, 816 int write_access) 817 { 818 /* should never happen if there's no MMU */ 819 BUG(); 820 return VM_FAULT_SIGBUS; 821 } 822 #endif 823 824 extern int make_pages_present(unsigned long addr, unsigned long end); 825 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 826 827 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, 828 int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); 829 830 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 831 extern void do_invalidatepage(struct page *page, unsigned long offset); 832 833 int __set_page_dirty_nobuffers(struct page *page); 834 int __set_page_dirty_no_writeback(struct page *page); 835 int redirty_page_for_writepage(struct writeback_control *wbc, 836 struct page *page); 837 int set_page_dirty(struct page *page); 838 int set_page_dirty_lock(struct page *page); 839 int clear_page_dirty_for_io(struct page *page); 840 841 extern unsigned long move_page_tables(struct vm_area_struct *vma, 842 unsigned long old_addr, struct vm_area_struct *new_vma, 843 unsigned long new_addr, unsigned long len); 844 extern unsigned long do_mremap(unsigned long addr, 845 unsigned long old_len, unsigned long new_len, 846 unsigned long flags, unsigned long new_addr); 847 extern int mprotect_fixup(struct vm_area_struct *vma, 848 struct vm_area_struct **pprev, unsigned long start, 849 unsigned long end, unsigned long newflags); 850 851 /* 852 * get_user_pages_fast provides equivalent functionality to get_user_pages, 853 * operating on current and current->mm (force=0 and doesn't return any vmas). 854 * 855 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions 856 * can be made about locking. get_user_pages_fast is to be implemented in a 857 * way that is advantageous (vs get_user_pages()) when the user memory area is 858 * already faulted in and present in ptes. However if the pages have to be 859 * faulted in, it may turn out to be slightly slower). 860 */ 861 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 862 struct page **pages); 863 864 /* 865 * A callback you can register to apply pressure to ageable caches. 866 * 867 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 868 * look through the least-recently-used 'nr_to_scan' entries and 869 * attempt to free them up. It should return the number of objects 870 * which remain in the cache. If it returns -1, it means it cannot do 871 * any scanning at this time (eg. there is a risk of deadlock). 872 * 873 * The 'gfpmask' refers to the allocation we are currently trying to 874 * fulfil. 875 * 876 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 877 * querying the cache size, so a fastpath for that case is appropriate. 878 */ 879 struct shrinker { 880 int (*shrink)(int nr_to_scan, gfp_t gfp_mask); 881 int seeks; /* seeks to recreate an obj */ 882 883 /* These are for internal use */ 884 struct list_head list; 885 long nr; /* objs pending delete */ 886 }; 887 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 888 extern void register_shrinker(struct shrinker *); 889 extern void unregister_shrinker(struct shrinker *); 890 891 int vma_wants_writenotify(struct vm_area_struct *vma); 892 893 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 894 895 #ifdef __PAGETABLE_PUD_FOLDED 896 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 897 unsigned long address) 898 { 899 return 0; 900 } 901 #else 902 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 903 #endif 904 905 #ifdef __PAGETABLE_PMD_FOLDED 906 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 907 unsigned long address) 908 { 909 return 0; 910 } 911 #else 912 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 913 #endif 914 915 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 916 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 917 918 /* 919 * The following ifdef needed to get the 4level-fixup.h header to work. 920 * Remove it when 4level-fixup.h has been removed. 921 */ 922 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 923 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 924 { 925 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 926 NULL: pud_offset(pgd, address); 927 } 928 929 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 930 { 931 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 932 NULL: pmd_offset(pud, address); 933 } 934 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 935 936 #if USE_SPLIT_PTLOCKS 937 /* 938 * We tuck a spinlock to guard each pagetable page into its struct page, 939 * at page->private, with BUILD_BUG_ON to make sure that this will not 940 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 941 * When freeing, reset page->mapping so free_pages_check won't complain. 942 */ 943 #define __pte_lockptr(page) &((page)->ptl) 944 #define pte_lock_init(_page) do { \ 945 spin_lock_init(__pte_lockptr(_page)); \ 946 } while (0) 947 #define pte_lock_deinit(page) ((page)->mapping = NULL) 948 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 949 #else /* !USE_SPLIT_PTLOCKS */ 950 /* 951 * We use mm->page_table_lock to guard all pagetable pages of the mm. 952 */ 953 #define pte_lock_init(page) do {} while (0) 954 #define pte_lock_deinit(page) do {} while (0) 955 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 956 #endif /* USE_SPLIT_PTLOCKS */ 957 958 static inline void pgtable_page_ctor(struct page *page) 959 { 960 pte_lock_init(page); 961 inc_zone_page_state(page, NR_PAGETABLE); 962 } 963 964 static inline void pgtable_page_dtor(struct page *page) 965 { 966 pte_lock_deinit(page); 967 dec_zone_page_state(page, NR_PAGETABLE); 968 } 969 970 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 971 ({ \ 972 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 973 pte_t *__pte = pte_offset_map(pmd, address); \ 974 *(ptlp) = __ptl; \ 975 spin_lock(__ptl); \ 976 __pte; \ 977 }) 978 979 #define pte_unmap_unlock(pte, ptl) do { \ 980 spin_unlock(ptl); \ 981 pte_unmap(pte); \ 982 } while (0) 983 984 #define pte_alloc_map(mm, pmd, address) \ 985 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 986 NULL: pte_offset_map(pmd, address)) 987 988 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 989 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 990 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 991 992 #define pte_alloc_kernel(pmd, address) \ 993 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 994 NULL: pte_offset_kernel(pmd, address)) 995 996 extern void free_area_init(unsigned long * zones_size); 997 extern void free_area_init_node(int nid, unsigned long * zones_size, 998 unsigned long zone_start_pfn, unsigned long *zholes_size); 999 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1000 /* 1001 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1002 * zones, allocate the backing mem_map and account for memory holes in a more 1003 * architecture independent manner. This is a substitute for creating the 1004 * zone_sizes[] and zholes_size[] arrays and passing them to 1005 * free_area_init_node() 1006 * 1007 * An architecture is expected to register range of page frames backed by 1008 * physical memory with add_active_range() before calling 1009 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1010 * usage, an architecture is expected to do something like 1011 * 1012 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1013 * max_highmem_pfn}; 1014 * for_each_valid_physical_page_range() 1015 * add_active_range(node_id, start_pfn, end_pfn) 1016 * free_area_init_nodes(max_zone_pfns); 1017 * 1018 * If the architecture guarantees that there are no holes in the ranges 1019 * registered with add_active_range(), free_bootmem_active_regions() 1020 * will call free_bootmem_node() for each registered physical page range. 1021 * Similarly sparse_memory_present_with_active_regions() calls 1022 * memory_present() for each range when SPARSEMEM is enabled. 1023 * 1024 * See mm/page_alloc.c for more information on each function exposed by 1025 * CONFIG_ARCH_POPULATES_NODE_MAP 1026 */ 1027 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1028 extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1029 unsigned long end_pfn); 1030 extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1031 unsigned long end_pfn); 1032 extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn, 1033 unsigned long end_pfn); 1034 extern void remove_all_active_ranges(void); 1035 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1036 unsigned long end_pfn); 1037 extern void get_pfn_range_for_nid(unsigned int nid, 1038 unsigned long *start_pfn, unsigned long *end_pfn); 1039 extern unsigned long find_min_pfn_with_active_regions(void); 1040 extern void free_bootmem_with_active_regions(int nid, 1041 unsigned long max_low_pfn); 1042 typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1043 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1044 extern void sparse_memory_present_with_active_regions(int nid); 1045 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1046 extern int early_pfn_to_nid(unsigned long pfn); 1047 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1048 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1049 extern void set_dma_reserve(unsigned long new_dma_reserve); 1050 extern void memmap_init_zone(unsigned long, int, unsigned long, 1051 unsigned long, enum memmap_context); 1052 extern void setup_per_zone_pages_min(void); 1053 extern void mem_init(void); 1054 extern void __init mmap_init(void); 1055 extern void show_mem(void); 1056 extern void si_meminfo(struct sysinfo * val); 1057 extern void si_meminfo_node(struct sysinfo *val, int nid); 1058 extern int after_bootmem; 1059 1060 #ifdef CONFIG_NUMA 1061 extern void setup_per_cpu_pageset(void); 1062 #else 1063 static inline void setup_per_cpu_pageset(void) {} 1064 #endif 1065 1066 /* nommu.c */ 1067 extern atomic_t mmap_pages_allocated; 1068 1069 /* prio_tree.c */ 1070 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1071 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1072 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1073 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1074 struct prio_tree_iter *iter); 1075 1076 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1077 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1078 (vma = vma_prio_tree_next(vma, iter)); ) 1079 1080 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1081 struct list_head *list) 1082 { 1083 vma->shared.vm_set.parent = NULL; 1084 list_add_tail(&vma->shared.vm_set.list, list); 1085 } 1086 1087 /* mmap.c */ 1088 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1089 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, 1090 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1091 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1092 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1093 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1094 struct mempolicy *); 1095 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1096 extern int split_vma(struct mm_struct *, 1097 struct vm_area_struct *, unsigned long addr, int new_below); 1098 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1099 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1100 struct rb_node **, struct rb_node *); 1101 extern void unlink_file_vma(struct vm_area_struct *); 1102 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1103 unsigned long addr, unsigned long len, pgoff_t pgoff); 1104 extern void exit_mmap(struct mm_struct *); 1105 1106 extern int mm_take_all_locks(struct mm_struct *mm); 1107 extern void mm_drop_all_locks(struct mm_struct *mm); 1108 1109 #ifdef CONFIG_PROC_FS 1110 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1111 extern void added_exe_file_vma(struct mm_struct *mm); 1112 extern void removed_exe_file_vma(struct mm_struct *mm); 1113 #else 1114 static inline void added_exe_file_vma(struct mm_struct *mm) 1115 {} 1116 1117 static inline void removed_exe_file_vma(struct mm_struct *mm) 1118 {} 1119 #endif /* CONFIG_PROC_FS */ 1120 1121 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1122 extern int install_special_mapping(struct mm_struct *mm, 1123 unsigned long addr, unsigned long len, 1124 unsigned long flags, struct page **pages); 1125 1126 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1127 1128 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1129 unsigned long len, unsigned long prot, 1130 unsigned long flag, unsigned long pgoff); 1131 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1132 unsigned long len, unsigned long flags, 1133 unsigned int vm_flags, unsigned long pgoff, 1134 int accountable); 1135 1136 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1137 unsigned long len, unsigned long prot, 1138 unsigned long flag, unsigned long offset) 1139 { 1140 unsigned long ret = -EINVAL; 1141 if ((offset + PAGE_ALIGN(len)) < offset) 1142 goto out; 1143 if (!(offset & ~PAGE_MASK)) 1144 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1145 out: 1146 return ret; 1147 } 1148 1149 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1150 1151 extern unsigned long do_brk(unsigned long, unsigned long); 1152 1153 /* filemap.c */ 1154 extern unsigned long page_unuse(struct page *); 1155 extern void truncate_inode_pages(struct address_space *, loff_t); 1156 extern void truncate_inode_pages_range(struct address_space *, 1157 loff_t lstart, loff_t lend); 1158 1159 /* generic vm_area_ops exported for stackable file systems */ 1160 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1161 1162 /* mm/page-writeback.c */ 1163 int write_one_page(struct page *page, int wait); 1164 1165 /* readahead.c */ 1166 #define VM_MAX_READAHEAD 128 /* kbytes */ 1167 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1168 1169 int do_page_cache_readahead(struct address_space *mapping, struct file *filp, 1170 pgoff_t offset, unsigned long nr_to_read); 1171 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1172 pgoff_t offset, unsigned long nr_to_read); 1173 1174 void page_cache_sync_readahead(struct address_space *mapping, 1175 struct file_ra_state *ra, 1176 struct file *filp, 1177 pgoff_t offset, 1178 unsigned long size); 1179 1180 void page_cache_async_readahead(struct address_space *mapping, 1181 struct file_ra_state *ra, 1182 struct file *filp, 1183 struct page *pg, 1184 pgoff_t offset, 1185 unsigned long size); 1186 1187 unsigned long max_sane_readahead(unsigned long nr); 1188 1189 /* Do stack extension */ 1190 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1191 #ifdef CONFIG_IA64 1192 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1193 #endif 1194 extern int expand_stack_downwards(struct vm_area_struct *vma, 1195 unsigned long address); 1196 1197 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1198 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1199 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1200 struct vm_area_struct **pprev); 1201 1202 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1203 NULL if none. Assume start_addr < end_addr. */ 1204 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1205 { 1206 struct vm_area_struct * vma = find_vma(mm,start_addr); 1207 1208 if (vma && end_addr <= vma->vm_start) 1209 vma = NULL; 1210 return vma; 1211 } 1212 1213 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1214 { 1215 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1216 } 1217 1218 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1219 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1220 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1221 unsigned long pfn, unsigned long size, pgprot_t); 1222 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1223 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1224 unsigned long pfn); 1225 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1226 unsigned long pfn); 1227 1228 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1229 unsigned int foll_flags); 1230 #define FOLL_WRITE 0x01 /* check pte is writable */ 1231 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1232 #define FOLL_GET 0x04 /* do get_page on page */ 1233 #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */ 1234 1235 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1236 void *data); 1237 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1238 unsigned long size, pte_fn_t fn, void *data); 1239 1240 #ifdef CONFIG_PROC_FS 1241 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1242 #else 1243 static inline void vm_stat_account(struct mm_struct *mm, 1244 unsigned long flags, struct file *file, long pages) 1245 { 1246 } 1247 #endif /* CONFIG_PROC_FS */ 1248 1249 #ifdef CONFIG_DEBUG_PAGEALLOC 1250 extern int debug_pagealloc_enabled; 1251 1252 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1253 1254 static inline void enable_debug_pagealloc(void) 1255 { 1256 debug_pagealloc_enabled = 1; 1257 } 1258 #ifdef CONFIG_HIBERNATION 1259 extern bool kernel_page_present(struct page *page); 1260 #endif /* CONFIG_HIBERNATION */ 1261 #else 1262 static inline void 1263 kernel_map_pages(struct page *page, int numpages, int enable) {} 1264 static inline void enable_debug_pagealloc(void) 1265 { 1266 } 1267 #ifdef CONFIG_HIBERNATION 1268 static inline bool kernel_page_present(struct page *page) { return true; } 1269 #endif /* CONFIG_HIBERNATION */ 1270 #endif 1271 1272 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1273 #ifdef __HAVE_ARCH_GATE_AREA 1274 int in_gate_area_no_task(unsigned long addr); 1275 int in_gate_area(struct task_struct *task, unsigned long addr); 1276 #else 1277 int in_gate_area_no_task(unsigned long addr); 1278 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1279 #endif /* __HAVE_ARCH_GATE_AREA */ 1280 1281 int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *, 1282 void __user *, size_t *, loff_t *); 1283 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1284 unsigned long lru_pages); 1285 1286 #ifndef CONFIG_MMU 1287 #define randomize_va_space 0 1288 #else 1289 extern int randomize_va_space; 1290 #endif 1291 1292 const char * arch_vma_name(struct vm_area_struct *vma); 1293 void print_vma_addr(char *prefix, unsigned long rip); 1294 1295 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1296 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1297 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1298 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1299 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1300 void *vmemmap_alloc_block(unsigned long size, int node); 1301 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1302 int vmemmap_populate_basepages(struct page *start_page, 1303 unsigned long pages, int node); 1304 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1305 void vmemmap_populate_print_last(void); 1306 1307 extern void *alloc_locked_buffer(size_t size); 1308 extern void free_locked_buffer(void *buffer, size_t size); 1309 #endif /* __KERNEL__ */ 1310 #endif /* _LINUX_MM_H */ 1311