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/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 16 struct mempolicy; 17 struct anon_vma; 18 struct file_ra_state; 19 struct user_struct; 20 struct writeback_control; 21 struct rlimit; 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 unsigned long totalram_pages; 29 extern void * high_memory; 30 extern int page_cluster; 31 32 #ifdef CONFIG_SYSCTL 33 extern int sysctl_legacy_va_layout; 34 #else 35 #define sysctl_legacy_va_layout 0 36 #endif 37 38 #include <asm/page.h> 39 #include <asm/pgtable.h> 40 #include <asm/processor.h> 41 42 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 43 44 /* to align the pointer to the (next) page boundary */ 45 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 46 47 /* 48 * Linux kernel virtual memory manager primitives. 49 * The idea being to have a "virtual" mm in the same way 50 * we have a virtual fs - giving a cleaner interface to the 51 * mm details, and allowing different kinds of memory mappings 52 * (from shared memory to executable loading to arbitrary 53 * mmap() functions). 54 */ 55 56 extern struct kmem_cache *vm_area_cachep; 57 58 #ifndef CONFIG_MMU 59 extern struct rb_root nommu_region_tree; 60 extern struct rw_semaphore nommu_region_sem; 61 62 extern unsigned int kobjsize(const void *objp); 63 #endif 64 65 /* 66 * vm_flags in vm_area_struct, see mm_types.h. 67 */ 68 #define VM_READ 0x00000001 /* currently active flags */ 69 #define VM_WRITE 0x00000002 70 #define VM_EXEC 0x00000004 71 #define VM_SHARED 0x00000008 72 73 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 74 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 75 #define VM_MAYWRITE 0x00000020 76 #define VM_MAYEXEC 0x00000040 77 #define VM_MAYSHARE 0x00000080 78 79 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 80 #define VM_GROWSUP 0x00000200 81 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 82 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 83 84 #define VM_EXECUTABLE 0x00001000 85 #define VM_LOCKED 0x00002000 86 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 87 88 /* Used by sys_madvise() */ 89 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 90 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 91 92 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 93 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 94 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ 95 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 96 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 97 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 98 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 99 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ 100 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ 101 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ 102 103 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ 104 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 105 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ 106 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ 107 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 108 109 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 110 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 111 #endif 112 113 #ifdef CONFIG_STACK_GROWSUP 114 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 115 #else 116 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 117 #endif 118 119 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 120 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 121 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 122 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 123 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 124 125 /* 126 * special vmas that are non-mergable, non-mlock()able 127 */ 128 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 129 130 /* 131 * mapping from the currently active vm_flags protection bits (the 132 * low four bits) to a page protection mask.. 133 */ 134 extern pgprot_t protection_map[16]; 135 136 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 137 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 138 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 139 140 /* 141 * This interface is used by x86 PAT code to identify a pfn mapping that is 142 * linear over entire vma. This is to optimize PAT code that deals with 143 * marking the physical region with a particular prot. This is not for generic 144 * mm use. Note also that this check will not work if the pfn mapping is 145 * linear for a vma starting at physical address 0. In which case PAT code 146 * falls back to slow path of reserving physical range page by page. 147 */ 148 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 149 { 150 return (vma->vm_flags & VM_PFN_AT_MMAP); 151 } 152 153 static inline int is_pfn_mapping(struct vm_area_struct *vma) 154 { 155 return (vma->vm_flags & VM_PFNMAP); 156 } 157 158 /* 159 * vm_fault is filled by the the pagefault handler and passed to the vma's 160 * ->fault function. The vma's ->fault is responsible for returning a bitmask 161 * of VM_FAULT_xxx flags that give details about how the fault was handled. 162 * 163 * pgoff should be used in favour of virtual_address, if possible. If pgoff 164 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 165 * mapping support. 166 */ 167 struct vm_fault { 168 unsigned int flags; /* FAULT_FLAG_xxx flags */ 169 pgoff_t pgoff; /* Logical page offset based on vma */ 170 void __user *virtual_address; /* Faulting virtual address */ 171 172 struct page *page; /* ->fault handlers should return a 173 * page here, unless VM_FAULT_NOPAGE 174 * is set (which is also implied by 175 * VM_FAULT_ERROR). 176 */ 177 }; 178 179 /* 180 * These are the virtual MM functions - opening of an area, closing and 181 * unmapping it (needed to keep files on disk up-to-date etc), pointer 182 * to the functions called when a no-page or a wp-page exception occurs. 183 */ 184 struct vm_operations_struct { 185 void (*open)(struct vm_area_struct * area); 186 void (*close)(struct vm_area_struct * area); 187 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 188 189 /* notification that a previously read-only page is about to become 190 * writable, if an error is returned it will cause a SIGBUS */ 191 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 192 193 /* called by access_process_vm when get_user_pages() fails, typically 194 * for use by special VMAs that can switch between memory and hardware 195 */ 196 int (*access)(struct vm_area_struct *vma, unsigned long addr, 197 void *buf, int len, int write); 198 #ifdef CONFIG_NUMA 199 /* 200 * set_policy() op must add a reference to any non-NULL @new mempolicy 201 * to hold the policy upon return. Caller should pass NULL @new to 202 * remove a policy and fall back to surrounding context--i.e. do not 203 * install a MPOL_DEFAULT policy, nor the task or system default 204 * mempolicy. 205 */ 206 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 207 208 /* 209 * get_policy() op must add reference [mpol_get()] to any policy at 210 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 211 * in mm/mempolicy.c will do this automatically. 212 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 213 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 214 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 215 * must return NULL--i.e., do not "fallback" to task or system default 216 * policy. 217 */ 218 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 219 unsigned long addr); 220 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 221 const nodemask_t *to, unsigned long flags); 222 #endif 223 }; 224 225 struct mmu_gather; 226 struct inode; 227 228 #define page_private(page) ((page)->private) 229 #define set_page_private(page, v) ((page)->private = (v)) 230 231 /* 232 * FIXME: take this include out, include page-flags.h in 233 * files which need it (119 of them) 234 */ 235 #include <linux/page-flags.h> 236 237 /* 238 * Methods to modify the page usage count. 239 * 240 * What counts for a page usage: 241 * - cache mapping (page->mapping) 242 * - private data (page->private) 243 * - page mapped in a task's page tables, each mapping 244 * is counted separately 245 * 246 * Also, many kernel routines increase the page count before a critical 247 * routine so they can be sure the page doesn't go away from under them. 248 */ 249 250 /* 251 * Drop a ref, return true if the refcount fell to zero (the page has no users) 252 */ 253 static inline int put_page_testzero(struct page *page) 254 { 255 VM_BUG_ON(atomic_read(&page->_count) == 0); 256 return atomic_dec_and_test(&page->_count); 257 } 258 259 /* 260 * Try to grab a ref unless the page has a refcount of zero, return false if 261 * that is the case. 262 */ 263 static inline int get_page_unless_zero(struct page *page) 264 { 265 return atomic_inc_not_zero(&page->_count); 266 } 267 268 /* Support for virtually mapped pages */ 269 struct page *vmalloc_to_page(const void *addr); 270 unsigned long vmalloc_to_pfn(const void *addr); 271 272 /* 273 * Determine if an address is within the vmalloc range 274 * 275 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 276 * is no special casing required. 277 */ 278 static inline int is_vmalloc_addr(const void *x) 279 { 280 #ifdef CONFIG_MMU 281 unsigned long addr = (unsigned long)x; 282 283 return addr >= VMALLOC_START && addr < VMALLOC_END; 284 #else 285 return 0; 286 #endif 287 } 288 #ifdef CONFIG_MMU 289 extern int is_vmalloc_or_module_addr(const void *x); 290 #else 291 static inline int is_vmalloc_or_module_addr(const void *x) 292 { 293 return 0; 294 } 295 #endif 296 297 static inline struct page *compound_head(struct page *page) 298 { 299 if (unlikely(PageTail(page))) 300 return page->first_page; 301 return page; 302 } 303 304 static inline int page_count(struct page *page) 305 { 306 return atomic_read(&compound_head(page)->_count); 307 } 308 309 static inline void get_page(struct page *page) 310 { 311 page = compound_head(page); 312 VM_BUG_ON(atomic_read(&page->_count) == 0); 313 atomic_inc(&page->_count); 314 } 315 316 static inline struct page *virt_to_head_page(const void *x) 317 { 318 struct page *page = virt_to_page(x); 319 return compound_head(page); 320 } 321 322 /* 323 * Setup the page count before being freed into the page allocator for 324 * the first time (boot or memory hotplug) 325 */ 326 static inline void init_page_count(struct page *page) 327 { 328 atomic_set(&page->_count, 1); 329 } 330 331 void put_page(struct page *page); 332 void put_pages_list(struct list_head *pages); 333 334 void split_page(struct page *page, unsigned int order); 335 336 /* 337 * Compound pages have a destructor function. Provide a 338 * prototype for that function and accessor functions. 339 * These are _only_ valid on the head of a PG_compound page. 340 */ 341 typedef void compound_page_dtor(struct page *); 342 343 static inline void set_compound_page_dtor(struct page *page, 344 compound_page_dtor *dtor) 345 { 346 page[1].lru.next = (void *)dtor; 347 } 348 349 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 350 { 351 return (compound_page_dtor *)page[1].lru.next; 352 } 353 354 static inline int compound_order(struct page *page) 355 { 356 if (!PageHead(page)) 357 return 0; 358 return (unsigned long)page[1].lru.prev; 359 } 360 361 static inline void set_compound_order(struct page *page, unsigned long order) 362 { 363 page[1].lru.prev = (void *)order; 364 } 365 366 /* 367 * Multiple processes may "see" the same page. E.g. for untouched 368 * mappings of /dev/null, all processes see the same page full of 369 * zeroes, and text pages of executables and shared libraries have 370 * only one copy in memory, at most, normally. 371 * 372 * For the non-reserved pages, page_count(page) denotes a reference count. 373 * page_count() == 0 means the page is free. page->lru is then used for 374 * freelist management in the buddy allocator. 375 * page_count() > 0 means the page has been allocated. 376 * 377 * Pages are allocated by the slab allocator in order to provide memory 378 * to kmalloc and kmem_cache_alloc. In this case, the management of the 379 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 380 * unless a particular usage is carefully commented. (the responsibility of 381 * freeing the kmalloc memory is the caller's, of course). 382 * 383 * A page may be used by anyone else who does a __get_free_page(). 384 * In this case, page_count still tracks the references, and should only 385 * be used through the normal accessor functions. The top bits of page->flags 386 * and page->virtual store page management information, but all other fields 387 * are unused and could be used privately, carefully. The management of this 388 * page is the responsibility of the one who allocated it, and those who have 389 * subsequently been given references to it. 390 * 391 * The other pages (we may call them "pagecache pages") are completely 392 * managed by the Linux memory manager: I/O, buffers, swapping etc. 393 * The following discussion applies only to them. 394 * 395 * A pagecache page contains an opaque `private' member, which belongs to the 396 * page's address_space. Usually, this is the address of a circular list of 397 * the page's disk buffers. PG_private must be set to tell the VM to call 398 * into the filesystem to release these pages. 399 * 400 * A page may belong to an inode's memory mapping. In this case, page->mapping 401 * is the pointer to the inode, and page->index is the file offset of the page, 402 * in units of PAGE_CACHE_SIZE. 403 * 404 * If pagecache pages are not associated with an inode, they are said to be 405 * anonymous pages. These may become associated with the swapcache, and in that 406 * case PG_swapcache is set, and page->private is an offset into the swapcache. 407 * 408 * In either case (swapcache or inode backed), the pagecache itself holds one 409 * reference to the page. Setting PG_private should also increment the 410 * refcount. The each user mapping also has a reference to the page. 411 * 412 * The pagecache pages are stored in a per-mapping radix tree, which is 413 * rooted at mapping->page_tree, and indexed by offset. 414 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 415 * lists, we instead now tag pages as dirty/writeback in the radix tree. 416 * 417 * All pagecache pages may be subject to I/O: 418 * - inode pages may need to be read from disk, 419 * - inode pages which have been modified and are MAP_SHARED may need 420 * to be written back to the inode on disk, 421 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 422 * modified may need to be swapped out to swap space and (later) to be read 423 * back into memory. 424 */ 425 426 /* 427 * The zone field is never updated after free_area_init_core() 428 * sets it, so none of the operations on it need to be atomic. 429 */ 430 431 432 /* 433 * page->flags layout: 434 * 435 * There are three possibilities for how page->flags get 436 * laid out. The first is for the normal case, without 437 * sparsemem. The second is for sparsemem when there is 438 * plenty of space for node and section. The last is when 439 * we have run out of space and have to fall back to an 440 * alternate (slower) way of determining the node. 441 * 442 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 443 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 444 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 445 */ 446 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 447 #define SECTIONS_WIDTH SECTIONS_SHIFT 448 #else 449 #define SECTIONS_WIDTH 0 450 #endif 451 452 #define ZONES_WIDTH ZONES_SHIFT 453 454 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 455 #define NODES_WIDTH NODES_SHIFT 456 #else 457 #ifdef CONFIG_SPARSEMEM_VMEMMAP 458 #error "Vmemmap: No space for nodes field in page flags" 459 #endif 460 #define NODES_WIDTH 0 461 #endif 462 463 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 464 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 465 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 466 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 467 468 /* 469 * We are going to use the flags for the page to node mapping if its in 470 * there. This includes the case where there is no node, so it is implicit. 471 */ 472 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 473 #define NODE_NOT_IN_PAGE_FLAGS 474 #endif 475 476 #ifndef PFN_SECTION_SHIFT 477 #define PFN_SECTION_SHIFT 0 478 #endif 479 480 /* 481 * Define the bit shifts to access each section. For non-existant 482 * sections we define the shift as 0; that plus a 0 mask ensures 483 * the compiler will optimise away reference to them. 484 */ 485 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 486 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 487 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 488 489 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 490 #ifdef NODE_NOT_IN_PAGEFLAGS 491 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 492 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 493 SECTIONS_PGOFF : ZONES_PGOFF) 494 #else 495 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 496 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 497 NODES_PGOFF : ZONES_PGOFF) 498 #endif 499 500 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 501 502 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 503 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 504 #endif 505 506 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 507 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 508 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 509 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 510 511 static inline enum zone_type page_zonenum(struct page *page) 512 { 513 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 514 } 515 516 /* 517 * The identification function is only used by the buddy allocator for 518 * determining if two pages could be buddies. We are not really 519 * identifying a zone since we could be using a the section number 520 * id if we have not node id available in page flags. 521 * We guarantee only that it will return the same value for two 522 * combinable pages in a zone. 523 */ 524 static inline int page_zone_id(struct page *page) 525 { 526 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 527 } 528 529 static inline int zone_to_nid(struct zone *zone) 530 { 531 #ifdef CONFIG_NUMA 532 return zone->node; 533 #else 534 return 0; 535 #endif 536 } 537 538 #ifdef NODE_NOT_IN_PAGE_FLAGS 539 extern int page_to_nid(struct page *page); 540 #else 541 static inline int page_to_nid(struct page *page) 542 { 543 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 544 } 545 #endif 546 547 static inline struct zone *page_zone(struct page *page) 548 { 549 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 550 } 551 552 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 553 static inline unsigned long page_to_section(struct page *page) 554 { 555 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 556 } 557 #endif 558 559 static inline void set_page_zone(struct page *page, enum zone_type zone) 560 { 561 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 562 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 563 } 564 565 static inline void set_page_node(struct page *page, unsigned long node) 566 { 567 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 568 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 569 } 570 571 static inline void set_page_section(struct page *page, unsigned long section) 572 { 573 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 574 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 575 } 576 577 static inline void set_page_links(struct page *page, enum zone_type zone, 578 unsigned long node, unsigned long pfn) 579 { 580 set_page_zone(page, zone); 581 set_page_node(page, node); 582 set_page_section(page, pfn_to_section_nr(pfn)); 583 } 584 585 /* 586 * Some inline functions in vmstat.h depend on page_zone() 587 */ 588 #include <linux/vmstat.h> 589 590 static __always_inline void *lowmem_page_address(struct page *page) 591 { 592 return __va(page_to_pfn(page) << PAGE_SHIFT); 593 } 594 595 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 596 #define HASHED_PAGE_VIRTUAL 597 #endif 598 599 #if defined(WANT_PAGE_VIRTUAL) 600 #define page_address(page) ((page)->virtual) 601 #define set_page_address(page, address) \ 602 do { \ 603 (page)->virtual = (address); \ 604 } while(0) 605 #define page_address_init() do { } while(0) 606 #endif 607 608 #if defined(HASHED_PAGE_VIRTUAL) 609 void *page_address(struct page *page); 610 void set_page_address(struct page *page, void *virtual); 611 void page_address_init(void); 612 #endif 613 614 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 615 #define page_address(page) lowmem_page_address(page) 616 #define set_page_address(page, address) do { } while(0) 617 #define page_address_init() do { } while(0) 618 #endif 619 620 /* 621 * On an anonymous page mapped into a user virtual memory area, 622 * page->mapping points to its anon_vma, not to a struct address_space; 623 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 624 * 625 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 626 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 627 * and then page->mapping points, not to an anon_vma, but to a private 628 * structure which KSM associates with that merged page. See ksm.h. 629 * 630 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 631 * 632 * Please note that, confusingly, "page_mapping" refers to the inode 633 * address_space which maps the page from disk; whereas "page_mapped" 634 * refers to user virtual address space into which the page is mapped. 635 */ 636 #define PAGE_MAPPING_ANON 1 637 #define PAGE_MAPPING_KSM 2 638 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 639 640 extern struct address_space swapper_space; 641 static inline struct address_space *page_mapping(struct page *page) 642 { 643 struct address_space *mapping = page->mapping; 644 645 VM_BUG_ON(PageSlab(page)); 646 if (unlikely(PageSwapCache(page))) 647 mapping = &swapper_space; 648 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 649 mapping = NULL; 650 return mapping; 651 } 652 653 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 654 static inline void *page_rmapping(struct page *page) 655 { 656 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 657 } 658 659 static inline int PageAnon(struct page *page) 660 { 661 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 662 } 663 664 /* 665 * Return the pagecache index of the passed page. Regular pagecache pages 666 * use ->index whereas swapcache pages use ->private 667 */ 668 static inline pgoff_t page_index(struct page *page) 669 { 670 if (unlikely(PageSwapCache(page))) 671 return page_private(page); 672 return page->index; 673 } 674 675 /* 676 * The atomic page->_mapcount, like _count, starts from -1: 677 * so that transitions both from it and to it can be tracked, 678 * using atomic_inc_and_test and atomic_add_negative(-1). 679 */ 680 static inline void reset_page_mapcount(struct page *page) 681 { 682 atomic_set(&(page)->_mapcount, -1); 683 } 684 685 static inline int page_mapcount(struct page *page) 686 { 687 return atomic_read(&(page)->_mapcount) + 1; 688 } 689 690 /* 691 * Return true if this page is mapped into pagetables. 692 */ 693 static inline int page_mapped(struct page *page) 694 { 695 return atomic_read(&(page)->_mapcount) >= 0; 696 } 697 698 /* 699 * Different kinds of faults, as returned by handle_mm_fault(). 700 * Used to decide whether a process gets delivered SIGBUS or 701 * just gets major/minor fault counters bumped up. 702 */ 703 704 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 705 706 #define VM_FAULT_OOM 0x0001 707 #define VM_FAULT_SIGBUS 0x0002 708 #define VM_FAULT_MAJOR 0x0004 709 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 710 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */ 711 712 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 713 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 714 715 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON) 716 717 /* 718 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 719 */ 720 extern void pagefault_out_of_memory(void); 721 722 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 723 724 extern void show_free_areas(void); 725 726 int shmem_lock(struct file *file, int lock, struct user_struct *user); 727 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 728 int shmem_zero_setup(struct vm_area_struct *); 729 730 #ifndef CONFIG_MMU 731 extern unsigned long shmem_get_unmapped_area(struct file *file, 732 unsigned long addr, 733 unsigned long len, 734 unsigned long pgoff, 735 unsigned long flags); 736 #endif 737 738 extern int can_do_mlock(void); 739 extern int user_shm_lock(size_t, struct user_struct *); 740 extern void user_shm_unlock(size_t, struct user_struct *); 741 742 /* 743 * Parameter block passed down to zap_pte_range in exceptional cases. 744 */ 745 struct zap_details { 746 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 747 struct address_space *check_mapping; /* Check page->mapping if set */ 748 pgoff_t first_index; /* Lowest page->index to unmap */ 749 pgoff_t last_index; /* Highest page->index to unmap */ 750 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 751 unsigned long truncate_count; /* Compare vm_truncate_count */ 752 }; 753 754 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 755 pte_t pte); 756 757 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 758 unsigned long size); 759 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 760 unsigned long size, struct zap_details *); 761 unsigned long unmap_vmas(struct mmu_gather **tlb, 762 struct vm_area_struct *start_vma, unsigned long start_addr, 763 unsigned long end_addr, unsigned long *nr_accounted, 764 struct zap_details *); 765 766 /** 767 * mm_walk - callbacks for walk_page_range 768 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 769 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 770 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 771 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 772 * @pte_hole: if set, called for each hole at all levels 773 * @hugetlb_entry: if set, called for each hugetlb entry 774 * 775 * (see walk_page_range for more details) 776 */ 777 struct mm_walk { 778 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 779 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 780 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 781 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 782 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 783 int (*hugetlb_entry)(pte_t *, unsigned long, unsigned long, 784 struct mm_walk *); 785 struct mm_struct *mm; 786 void *private; 787 }; 788 789 int walk_page_range(unsigned long addr, unsigned long end, 790 struct mm_walk *walk); 791 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 792 unsigned long end, unsigned long floor, unsigned long ceiling); 793 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 794 struct vm_area_struct *vma); 795 void unmap_mapping_range(struct address_space *mapping, 796 loff_t const holebegin, loff_t const holelen, int even_cows); 797 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 798 unsigned long *pfn); 799 int follow_phys(struct vm_area_struct *vma, unsigned long address, 800 unsigned int flags, unsigned long *prot, resource_size_t *phys); 801 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 802 void *buf, int len, int write); 803 804 static inline void unmap_shared_mapping_range(struct address_space *mapping, 805 loff_t const holebegin, loff_t const holelen) 806 { 807 unmap_mapping_range(mapping, holebegin, holelen, 0); 808 } 809 810 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 811 extern int vmtruncate(struct inode *inode, loff_t offset); 812 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 813 814 int truncate_inode_page(struct address_space *mapping, struct page *page); 815 int generic_error_remove_page(struct address_space *mapping, struct page *page); 816 817 int invalidate_inode_page(struct page *page); 818 819 #ifdef CONFIG_MMU 820 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 821 unsigned long address, unsigned int flags); 822 #else 823 static inline int handle_mm_fault(struct mm_struct *mm, 824 struct vm_area_struct *vma, unsigned long address, 825 unsigned int flags) 826 { 827 /* should never happen if there's no MMU */ 828 BUG(); 829 return VM_FAULT_SIGBUS; 830 } 831 #endif 832 833 extern int make_pages_present(unsigned long addr, unsigned long end); 834 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 835 836 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 837 unsigned long start, int nr_pages, int write, int force, 838 struct page **pages, struct vm_area_struct **vmas); 839 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 840 struct page **pages); 841 struct page *get_dump_page(unsigned long addr); 842 843 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 844 extern void do_invalidatepage(struct page *page, unsigned long offset); 845 846 int __set_page_dirty_nobuffers(struct page *page); 847 int __set_page_dirty_no_writeback(struct page *page); 848 int redirty_page_for_writepage(struct writeback_control *wbc, 849 struct page *page); 850 void account_page_dirtied(struct page *page, struct address_space *mapping); 851 int set_page_dirty(struct page *page); 852 int set_page_dirty_lock(struct page *page); 853 int clear_page_dirty_for_io(struct page *page); 854 855 extern unsigned long move_page_tables(struct vm_area_struct *vma, 856 unsigned long old_addr, struct vm_area_struct *new_vma, 857 unsigned long new_addr, unsigned long len); 858 extern unsigned long do_mremap(unsigned long addr, 859 unsigned long old_len, unsigned long new_len, 860 unsigned long flags, unsigned long new_addr); 861 extern int mprotect_fixup(struct vm_area_struct *vma, 862 struct vm_area_struct **pprev, unsigned long start, 863 unsigned long end, unsigned long newflags); 864 865 /* 866 * doesn't attempt to fault and will return short. 867 */ 868 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 869 struct page **pages); 870 871 /* 872 * A callback you can register to apply pressure to ageable caches. 873 * 874 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 875 * look through the least-recently-used 'nr_to_scan' entries and 876 * attempt to free them up. It should return the number of objects 877 * which remain in the cache. If it returns -1, it means it cannot do 878 * any scanning at this time (eg. there is a risk of deadlock). 879 * 880 * The 'gfpmask' refers to the allocation we are currently trying to 881 * fulfil. 882 * 883 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 884 * querying the cache size, so a fastpath for that case is appropriate. 885 */ 886 struct shrinker { 887 int (*shrink)(int nr_to_scan, gfp_t gfp_mask); 888 int seeks; /* seeks to recreate an obj */ 889 890 /* These are for internal use */ 891 struct list_head list; 892 long nr; /* objs pending delete */ 893 }; 894 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 895 extern void register_shrinker(struct shrinker *); 896 extern void unregister_shrinker(struct shrinker *); 897 898 int vma_wants_writenotify(struct vm_area_struct *vma); 899 900 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 901 902 #ifdef __PAGETABLE_PUD_FOLDED 903 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 904 unsigned long address) 905 { 906 return 0; 907 } 908 #else 909 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 910 #endif 911 912 #ifdef __PAGETABLE_PMD_FOLDED 913 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 914 unsigned long address) 915 { 916 return 0; 917 } 918 #else 919 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 920 #endif 921 922 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 923 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 924 925 /* 926 * The following ifdef needed to get the 4level-fixup.h header to work. 927 * Remove it when 4level-fixup.h has been removed. 928 */ 929 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 930 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 931 { 932 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 933 NULL: pud_offset(pgd, address); 934 } 935 936 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 937 { 938 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 939 NULL: pmd_offset(pud, address); 940 } 941 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 942 943 #if USE_SPLIT_PTLOCKS 944 /* 945 * We tuck a spinlock to guard each pagetable page into its struct page, 946 * at page->private, with BUILD_BUG_ON to make sure that this will not 947 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 948 * When freeing, reset page->mapping so free_pages_check won't complain. 949 */ 950 #define __pte_lockptr(page) &((page)->ptl) 951 #define pte_lock_init(_page) do { \ 952 spin_lock_init(__pte_lockptr(_page)); \ 953 } while (0) 954 #define pte_lock_deinit(page) ((page)->mapping = NULL) 955 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 956 #else /* !USE_SPLIT_PTLOCKS */ 957 /* 958 * We use mm->page_table_lock to guard all pagetable pages of the mm. 959 */ 960 #define pte_lock_init(page) do {} while (0) 961 #define pte_lock_deinit(page) do {} while (0) 962 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 963 #endif /* USE_SPLIT_PTLOCKS */ 964 965 static inline void pgtable_page_ctor(struct page *page) 966 { 967 pte_lock_init(page); 968 inc_zone_page_state(page, NR_PAGETABLE); 969 } 970 971 static inline void pgtable_page_dtor(struct page *page) 972 { 973 pte_lock_deinit(page); 974 dec_zone_page_state(page, NR_PAGETABLE); 975 } 976 977 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 978 ({ \ 979 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 980 pte_t *__pte = pte_offset_map(pmd, address); \ 981 *(ptlp) = __ptl; \ 982 spin_lock(__ptl); \ 983 __pte; \ 984 }) 985 986 #define pte_unmap_unlock(pte, ptl) do { \ 987 spin_unlock(ptl); \ 988 pte_unmap(pte); \ 989 } while (0) 990 991 #define pte_alloc_map(mm, pmd, address) \ 992 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 993 NULL: pte_offset_map(pmd, address)) 994 995 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 996 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 997 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 998 999 #define pte_alloc_kernel(pmd, address) \ 1000 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1001 NULL: pte_offset_kernel(pmd, address)) 1002 1003 extern void free_area_init(unsigned long * zones_size); 1004 extern void free_area_init_node(int nid, unsigned long * zones_size, 1005 unsigned long zone_start_pfn, unsigned long *zholes_size); 1006 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1007 /* 1008 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1009 * zones, allocate the backing mem_map and account for memory holes in a more 1010 * architecture independent manner. This is a substitute for creating the 1011 * zone_sizes[] and zholes_size[] arrays and passing them to 1012 * free_area_init_node() 1013 * 1014 * An architecture is expected to register range of page frames backed by 1015 * physical memory with add_active_range() before calling 1016 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1017 * usage, an architecture is expected to do something like 1018 * 1019 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1020 * max_highmem_pfn}; 1021 * for_each_valid_physical_page_range() 1022 * add_active_range(node_id, start_pfn, end_pfn) 1023 * free_area_init_nodes(max_zone_pfns); 1024 * 1025 * If the architecture guarantees that there are no holes in the ranges 1026 * registered with add_active_range(), free_bootmem_active_regions() 1027 * will call free_bootmem_node() for each registered physical page range. 1028 * Similarly sparse_memory_present_with_active_regions() calls 1029 * memory_present() for each range when SPARSEMEM is enabled. 1030 * 1031 * See mm/page_alloc.c for more information on each function exposed by 1032 * CONFIG_ARCH_POPULATES_NODE_MAP 1033 */ 1034 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1035 extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1036 unsigned long end_pfn); 1037 extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1038 unsigned long end_pfn); 1039 extern void remove_all_active_ranges(void); 1040 void sort_node_map(void); 1041 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1042 unsigned long end_pfn); 1043 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1044 unsigned long end_pfn); 1045 extern void get_pfn_range_for_nid(unsigned int nid, 1046 unsigned long *start_pfn, unsigned long *end_pfn); 1047 extern unsigned long find_min_pfn_with_active_regions(void); 1048 extern void free_bootmem_with_active_regions(int nid, 1049 unsigned long max_low_pfn); 1050 typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1051 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1052 extern void sparse_memory_present_with_active_regions(int nid); 1053 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1054 1055 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ 1056 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1057 static inline int __early_pfn_to_nid(unsigned long pfn) 1058 { 1059 return 0; 1060 } 1061 #else 1062 /* please see mm/page_alloc.c */ 1063 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1064 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1065 /* there is a per-arch backend function. */ 1066 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1067 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1068 #endif 1069 1070 extern void set_dma_reserve(unsigned long new_dma_reserve); 1071 extern void memmap_init_zone(unsigned long, int, unsigned long, 1072 unsigned long, enum memmap_context); 1073 extern void setup_per_zone_wmarks(void); 1074 extern void calculate_zone_inactive_ratio(struct zone *zone); 1075 extern void mem_init(void); 1076 extern void __init mmap_init(void); 1077 extern void show_mem(void); 1078 extern void si_meminfo(struct sysinfo * val); 1079 extern void si_meminfo_node(struct sysinfo *val, int nid); 1080 extern int after_bootmem; 1081 1082 #ifdef CONFIG_NUMA 1083 extern void setup_per_cpu_pageset(void); 1084 #else 1085 static inline void setup_per_cpu_pageset(void) {} 1086 #endif 1087 1088 extern void zone_pcp_update(struct zone *zone); 1089 1090 /* nommu.c */ 1091 extern atomic_long_t mmap_pages_allocated; 1092 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1093 1094 /* prio_tree.c */ 1095 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1096 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1097 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1098 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1099 struct prio_tree_iter *iter); 1100 1101 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1102 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1103 (vma = vma_prio_tree_next(vma, iter)); ) 1104 1105 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1106 struct list_head *list) 1107 { 1108 vma->shared.vm_set.parent = NULL; 1109 list_add_tail(&vma->shared.vm_set.list, list); 1110 } 1111 1112 /* mmap.c */ 1113 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1114 extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, 1115 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1116 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1117 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1118 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1119 struct mempolicy *); 1120 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1121 extern int split_vma(struct mm_struct *, 1122 struct vm_area_struct *, unsigned long addr, int new_below); 1123 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1124 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1125 struct rb_node **, struct rb_node *); 1126 extern void unlink_file_vma(struct vm_area_struct *); 1127 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1128 unsigned long addr, unsigned long len, pgoff_t pgoff); 1129 extern void exit_mmap(struct mm_struct *); 1130 1131 extern int mm_take_all_locks(struct mm_struct *mm); 1132 extern void mm_drop_all_locks(struct mm_struct *mm); 1133 1134 #ifdef CONFIG_PROC_FS 1135 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1136 extern void added_exe_file_vma(struct mm_struct *mm); 1137 extern void removed_exe_file_vma(struct mm_struct *mm); 1138 #else 1139 static inline void added_exe_file_vma(struct mm_struct *mm) 1140 {} 1141 1142 static inline void removed_exe_file_vma(struct mm_struct *mm) 1143 {} 1144 #endif /* CONFIG_PROC_FS */ 1145 1146 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1147 extern int install_special_mapping(struct mm_struct *mm, 1148 unsigned long addr, unsigned long len, 1149 unsigned long flags, struct page **pages); 1150 1151 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1152 1153 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1154 unsigned long len, unsigned long prot, 1155 unsigned long flag, unsigned long pgoff); 1156 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1157 unsigned long len, unsigned long flags, 1158 unsigned int vm_flags, unsigned long pgoff); 1159 1160 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1161 unsigned long len, unsigned long prot, 1162 unsigned long flag, unsigned long offset) 1163 { 1164 unsigned long ret = -EINVAL; 1165 if ((offset + PAGE_ALIGN(len)) < offset) 1166 goto out; 1167 if (!(offset & ~PAGE_MASK)) 1168 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1169 out: 1170 return ret; 1171 } 1172 1173 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1174 1175 extern unsigned long do_brk(unsigned long, unsigned long); 1176 1177 /* filemap.c */ 1178 extern unsigned long page_unuse(struct page *); 1179 extern void truncate_inode_pages(struct address_space *, loff_t); 1180 extern void truncate_inode_pages_range(struct address_space *, 1181 loff_t lstart, loff_t lend); 1182 1183 /* generic vm_area_ops exported for stackable file systems */ 1184 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1185 1186 /* mm/page-writeback.c */ 1187 int write_one_page(struct page *page, int wait); 1188 void task_dirty_inc(struct task_struct *tsk); 1189 1190 /* readahead.c */ 1191 #define VM_MAX_READAHEAD 128 /* kbytes */ 1192 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1193 1194 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1195 pgoff_t offset, unsigned long nr_to_read); 1196 1197 void page_cache_sync_readahead(struct address_space *mapping, 1198 struct file_ra_state *ra, 1199 struct file *filp, 1200 pgoff_t offset, 1201 unsigned long size); 1202 1203 void page_cache_async_readahead(struct address_space *mapping, 1204 struct file_ra_state *ra, 1205 struct file *filp, 1206 struct page *pg, 1207 pgoff_t offset, 1208 unsigned long size); 1209 1210 unsigned long max_sane_readahead(unsigned long nr); 1211 unsigned long ra_submit(struct file_ra_state *ra, 1212 struct address_space *mapping, 1213 struct file *filp); 1214 1215 /* Do stack extension */ 1216 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1217 #ifdef CONFIG_IA64 1218 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1219 #endif 1220 extern int expand_stack_downwards(struct vm_area_struct *vma, 1221 unsigned long address); 1222 1223 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1224 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1225 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1226 struct vm_area_struct **pprev); 1227 1228 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1229 NULL if none. Assume start_addr < end_addr. */ 1230 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1231 { 1232 struct vm_area_struct * vma = find_vma(mm,start_addr); 1233 1234 if (vma && end_addr <= vma->vm_start) 1235 vma = NULL; 1236 return vma; 1237 } 1238 1239 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1240 { 1241 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1242 } 1243 1244 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1245 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1246 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1247 unsigned long pfn, unsigned long size, pgprot_t); 1248 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1249 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1250 unsigned long pfn); 1251 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1252 unsigned long pfn); 1253 1254 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1255 unsigned int foll_flags); 1256 #define FOLL_WRITE 0x01 /* check pte is writable */ 1257 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1258 #define FOLL_GET 0x04 /* do get_page on page */ 1259 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1260 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1261 1262 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1263 void *data); 1264 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1265 unsigned long size, pte_fn_t fn, void *data); 1266 1267 #ifdef CONFIG_PROC_FS 1268 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1269 #else 1270 static inline void vm_stat_account(struct mm_struct *mm, 1271 unsigned long flags, struct file *file, long pages) 1272 { 1273 } 1274 #endif /* CONFIG_PROC_FS */ 1275 1276 #ifdef CONFIG_DEBUG_PAGEALLOC 1277 extern int debug_pagealloc_enabled; 1278 1279 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1280 1281 static inline void enable_debug_pagealloc(void) 1282 { 1283 debug_pagealloc_enabled = 1; 1284 } 1285 #ifdef CONFIG_HIBERNATION 1286 extern bool kernel_page_present(struct page *page); 1287 #endif /* CONFIG_HIBERNATION */ 1288 #else 1289 static inline void 1290 kernel_map_pages(struct page *page, int numpages, int enable) {} 1291 static inline void enable_debug_pagealloc(void) 1292 { 1293 } 1294 #ifdef CONFIG_HIBERNATION 1295 static inline bool kernel_page_present(struct page *page) { return true; } 1296 #endif /* CONFIG_HIBERNATION */ 1297 #endif 1298 1299 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1300 #ifdef __HAVE_ARCH_GATE_AREA 1301 int in_gate_area_no_task(unsigned long addr); 1302 int in_gate_area(struct task_struct *task, unsigned long addr); 1303 #else 1304 int in_gate_area_no_task(unsigned long addr); 1305 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1306 #endif /* __HAVE_ARCH_GATE_AREA */ 1307 1308 int drop_caches_sysctl_handler(struct ctl_table *, int, 1309 void __user *, size_t *, loff_t *); 1310 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1311 unsigned long lru_pages); 1312 1313 #ifndef CONFIG_MMU 1314 #define randomize_va_space 0 1315 #else 1316 extern int randomize_va_space; 1317 #endif 1318 1319 const char * arch_vma_name(struct vm_area_struct *vma); 1320 void print_vma_addr(char *prefix, unsigned long rip); 1321 1322 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1323 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1324 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1325 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1326 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1327 void *vmemmap_alloc_block(unsigned long size, int node); 1328 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1329 int vmemmap_populate_basepages(struct page *start_page, 1330 unsigned long pages, int node); 1331 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1332 void vmemmap_populate_print_last(void); 1333 1334 extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim, 1335 size_t size); 1336 extern void refund_locked_memory(struct mm_struct *mm, size_t size); 1337 1338 enum mf_flags { 1339 MF_COUNT_INCREASED = 1 << 0, 1340 }; 1341 extern void memory_failure(unsigned long pfn, int trapno); 1342 extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1343 extern int unpoison_memory(unsigned long pfn); 1344 extern int sysctl_memory_failure_early_kill; 1345 extern int sysctl_memory_failure_recovery; 1346 extern void shake_page(struct page *p, int access); 1347 extern atomic_long_t mce_bad_pages; 1348 extern int soft_offline_page(struct page *page, int flags); 1349 1350 #endif /* __KERNEL__ */ 1351 #endif /* _LINUX_MM_H */ 1352