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 #include <linux/range.h> 16 #include <linux/pfn.h> 17 18 struct mempolicy; 19 struct anon_vma; 20 struct file_ra_state; 21 struct user_struct; 22 struct writeback_control; 23 24 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 25 extern unsigned long max_mapnr; 26 #endif 27 28 extern unsigned long num_physpages; 29 extern unsigned long totalram_pages; 30 extern void * high_memory; 31 extern int page_cluster; 32 33 #ifdef CONFIG_SYSCTL 34 extern int sysctl_legacy_va_layout; 35 #else 36 #define sysctl_legacy_va_layout 0 37 #endif 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 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64) 82 #define VM_GROWSUP 0x00000200 83 #else 84 #define VM_GROWSUP 0x00000000 85 #endif 86 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 87 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 88 89 #define VM_EXECUTABLE 0x00001000 90 #define VM_LOCKED 0x00002000 91 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 92 93 /* Used by sys_madvise() */ 94 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 95 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 96 97 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 98 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 99 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ 100 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 101 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 102 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 103 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 104 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ 105 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ 106 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ 107 108 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ 109 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 110 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ 111 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ 112 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 113 114 /* Bits set in the VMA until the stack is in its final location */ 115 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) 116 117 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 118 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 119 #endif 120 121 #ifdef CONFIG_STACK_GROWSUP 122 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 123 #else 124 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 125 #endif 126 127 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 128 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 129 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 130 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 131 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 132 133 /* 134 * special vmas that are non-mergable, non-mlock()able 135 */ 136 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 137 138 /* 139 * mapping from the currently active vm_flags protection bits (the 140 * low four bits) to a page protection mask.. 141 */ 142 extern pgprot_t protection_map[16]; 143 144 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 145 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 146 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 147 148 /* 149 * This interface is used by x86 PAT code to identify a pfn mapping that is 150 * linear over entire vma. This is to optimize PAT code that deals with 151 * marking the physical region with a particular prot. This is not for generic 152 * mm use. Note also that this check will not work if the pfn mapping is 153 * linear for a vma starting at physical address 0. In which case PAT code 154 * falls back to slow path of reserving physical range page by page. 155 */ 156 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 157 { 158 return (vma->vm_flags & VM_PFN_AT_MMAP); 159 } 160 161 static inline int is_pfn_mapping(struct vm_area_struct *vma) 162 { 163 return (vma->vm_flags & VM_PFNMAP); 164 } 165 166 /* 167 * vm_fault is filled by the the pagefault handler and passed to the vma's 168 * ->fault function. The vma's ->fault is responsible for returning a bitmask 169 * of VM_FAULT_xxx flags that give details about how the fault was handled. 170 * 171 * pgoff should be used in favour of virtual_address, if possible. If pgoff 172 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 173 * mapping support. 174 */ 175 struct vm_fault { 176 unsigned int flags; /* FAULT_FLAG_xxx flags */ 177 pgoff_t pgoff; /* Logical page offset based on vma */ 178 void __user *virtual_address; /* Faulting virtual address */ 179 180 struct page *page; /* ->fault handlers should return a 181 * page here, unless VM_FAULT_NOPAGE 182 * is set (which is also implied by 183 * VM_FAULT_ERROR). 184 */ 185 }; 186 187 /* 188 * These are the virtual MM functions - opening of an area, closing and 189 * unmapping it (needed to keep files on disk up-to-date etc), pointer 190 * to the functions called when a no-page or a wp-page exception occurs. 191 */ 192 struct vm_operations_struct { 193 void (*open)(struct vm_area_struct * area); 194 void (*close)(struct vm_area_struct * area); 195 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 196 197 /* notification that a previously read-only page is about to become 198 * writable, if an error is returned it will cause a SIGBUS */ 199 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 200 201 /* called by access_process_vm when get_user_pages() fails, typically 202 * for use by special VMAs that can switch between memory and hardware 203 */ 204 int (*access)(struct vm_area_struct *vma, unsigned long addr, 205 void *buf, int len, int write); 206 #ifdef CONFIG_NUMA 207 /* 208 * set_policy() op must add a reference to any non-NULL @new mempolicy 209 * to hold the policy upon return. Caller should pass NULL @new to 210 * remove a policy and fall back to surrounding context--i.e. do not 211 * install a MPOL_DEFAULT policy, nor the task or system default 212 * mempolicy. 213 */ 214 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 215 216 /* 217 * get_policy() op must add reference [mpol_get()] to any policy at 218 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 219 * in mm/mempolicy.c will do this automatically. 220 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 221 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 222 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 223 * must return NULL--i.e., do not "fallback" to task or system default 224 * policy. 225 */ 226 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 227 unsigned long addr); 228 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 229 const nodemask_t *to, unsigned long flags); 230 #endif 231 }; 232 233 struct mmu_gather; 234 struct inode; 235 236 #define page_private(page) ((page)->private) 237 #define set_page_private(page, v) ((page)->private = (v)) 238 239 /* 240 * FIXME: take this include out, include page-flags.h in 241 * files which need it (119 of them) 242 */ 243 #include <linux/page-flags.h> 244 245 /* 246 * Methods to modify the page usage count. 247 * 248 * What counts for a page usage: 249 * - cache mapping (page->mapping) 250 * - private data (page->private) 251 * - page mapped in a task's page tables, each mapping 252 * is counted separately 253 * 254 * Also, many kernel routines increase the page count before a critical 255 * routine so they can be sure the page doesn't go away from under them. 256 */ 257 258 /* 259 * Drop a ref, return true if the refcount fell to zero (the page has no users) 260 */ 261 static inline int put_page_testzero(struct page *page) 262 { 263 VM_BUG_ON(atomic_read(&page->_count) == 0); 264 return atomic_dec_and_test(&page->_count); 265 } 266 267 /* 268 * Try to grab a ref unless the page has a refcount of zero, return false if 269 * that is the case. 270 */ 271 static inline int get_page_unless_zero(struct page *page) 272 { 273 return atomic_inc_not_zero(&page->_count); 274 } 275 276 extern int page_is_ram(unsigned long pfn); 277 278 /* Support for virtually mapped pages */ 279 struct page *vmalloc_to_page(const void *addr); 280 unsigned long vmalloc_to_pfn(const void *addr); 281 282 /* 283 * Determine if an address is within the vmalloc range 284 * 285 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 286 * is no special casing required. 287 */ 288 static inline int is_vmalloc_addr(const void *x) 289 { 290 #ifdef CONFIG_MMU 291 unsigned long addr = (unsigned long)x; 292 293 return addr >= VMALLOC_START && addr < VMALLOC_END; 294 #else 295 return 0; 296 #endif 297 } 298 #ifdef CONFIG_MMU 299 extern int is_vmalloc_or_module_addr(const void *x); 300 #else 301 static inline int is_vmalloc_or_module_addr(const void *x) 302 { 303 return 0; 304 } 305 #endif 306 307 static inline struct page *compound_head(struct page *page) 308 { 309 if (unlikely(PageTail(page))) 310 return page->first_page; 311 return page; 312 } 313 314 static inline int page_count(struct page *page) 315 { 316 return atomic_read(&compound_head(page)->_count); 317 } 318 319 static inline void get_page(struct page *page) 320 { 321 page = compound_head(page); 322 VM_BUG_ON(atomic_read(&page->_count) == 0); 323 atomic_inc(&page->_count); 324 } 325 326 static inline struct page *virt_to_head_page(const void *x) 327 { 328 struct page *page = virt_to_page(x); 329 return compound_head(page); 330 } 331 332 /* 333 * Setup the page count before being freed into the page allocator for 334 * the first time (boot or memory hotplug) 335 */ 336 static inline void init_page_count(struct page *page) 337 { 338 atomic_set(&page->_count, 1); 339 } 340 341 void put_page(struct page *page); 342 void put_pages_list(struct list_head *pages); 343 344 void split_page(struct page *page, unsigned int order); 345 int split_free_page(struct page *page); 346 347 /* 348 * Compound pages have a destructor function. Provide a 349 * prototype for that function and accessor functions. 350 * These are _only_ valid on the head of a PG_compound page. 351 */ 352 typedef void compound_page_dtor(struct page *); 353 354 static inline void set_compound_page_dtor(struct page *page, 355 compound_page_dtor *dtor) 356 { 357 page[1].lru.next = (void *)dtor; 358 } 359 360 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 361 { 362 return (compound_page_dtor *)page[1].lru.next; 363 } 364 365 static inline int compound_order(struct page *page) 366 { 367 if (!PageHead(page)) 368 return 0; 369 return (unsigned long)page[1].lru.prev; 370 } 371 372 static inline void set_compound_order(struct page *page, unsigned long order) 373 { 374 page[1].lru.prev = (void *)order; 375 } 376 377 /* 378 * Multiple processes may "see" the same page. E.g. for untouched 379 * mappings of /dev/null, all processes see the same page full of 380 * zeroes, and text pages of executables and shared libraries have 381 * only one copy in memory, at most, normally. 382 * 383 * For the non-reserved pages, page_count(page) denotes a reference count. 384 * page_count() == 0 means the page is free. page->lru is then used for 385 * freelist management in the buddy allocator. 386 * page_count() > 0 means the page has been allocated. 387 * 388 * Pages are allocated by the slab allocator in order to provide memory 389 * to kmalloc and kmem_cache_alloc. In this case, the management of the 390 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 391 * unless a particular usage is carefully commented. (the responsibility of 392 * freeing the kmalloc memory is the caller's, of course). 393 * 394 * A page may be used by anyone else who does a __get_free_page(). 395 * In this case, page_count still tracks the references, and should only 396 * be used through the normal accessor functions. The top bits of page->flags 397 * and page->virtual store page management information, but all other fields 398 * are unused and could be used privately, carefully. The management of this 399 * page is the responsibility of the one who allocated it, and those who have 400 * subsequently been given references to it. 401 * 402 * The other pages (we may call them "pagecache pages") are completely 403 * managed by the Linux memory manager: I/O, buffers, swapping etc. 404 * The following discussion applies only to them. 405 * 406 * A pagecache page contains an opaque `private' member, which belongs to the 407 * page's address_space. Usually, this is the address of a circular list of 408 * the page's disk buffers. PG_private must be set to tell the VM to call 409 * into the filesystem to release these pages. 410 * 411 * A page may belong to an inode's memory mapping. In this case, page->mapping 412 * is the pointer to the inode, and page->index is the file offset of the page, 413 * in units of PAGE_CACHE_SIZE. 414 * 415 * If pagecache pages are not associated with an inode, they are said to be 416 * anonymous pages. These may become associated with the swapcache, and in that 417 * case PG_swapcache is set, and page->private is an offset into the swapcache. 418 * 419 * In either case (swapcache or inode backed), the pagecache itself holds one 420 * reference to the page. Setting PG_private should also increment the 421 * refcount. The each user mapping also has a reference to the page. 422 * 423 * The pagecache pages are stored in a per-mapping radix tree, which is 424 * rooted at mapping->page_tree, and indexed by offset. 425 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 426 * lists, we instead now tag pages as dirty/writeback in the radix tree. 427 * 428 * All pagecache pages may be subject to I/O: 429 * - inode pages may need to be read from disk, 430 * - inode pages which have been modified and are MAP_SHARED may need 431 * to be written back to the inode on disk, 432 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 433 * modified may need to be swapped out to swap space and (later) to be read 434 * back into memory. 435 */ 436 437 /* 438 * The zone field is never updated after free_area_init_core() 439 * sets it, so none of the operations on it need to be atomic. 440 */ 441 442 443 /* 444 * page->flags layout: 445 * 446 * There are three possibilities for how page->flags get 447 * laid out. The first is for the normal case, without 448 * sparsemem. The second is for sparsemem when there is 449 * plenty of space for node and section. The last is when 450 * we have run out of space and have to fall back to an 451 * alternate (slower) way of determining the node. 452 * 453 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 454 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 455 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 456 */ 457 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 458 #define SECTIONS_WIDTH SECTIONS_SHIFT 459 #else 460 #define SECTIONS_WIDTH 0 461 #endif 462 463 #define ZONES_WIDTH ZONES_SHIFT 464 465 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 466 #define NODES_WIDTH NODES_SHIFT 467 #else 468 #ifdef CONFIG_SPARSEMEM_VMEMMAP 469 #error "Vmemmap: No space for nodes field in page flags" 470 #endif 471 #define NODES_WIDTH 0 472 #endif 473 474 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 475 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 476 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 477 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 478 479 /* 480 * We are going to use the flags for the page to node mapping if its in 481 * there. This includes the case where there is no node, so it is implicit. 482 */ 483 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 484 #define NODE_NOT_IN_PAGE_FLAGS 485 #endif 486 487 #ifndef PFN_SECTION_SHIFT 488 #define PFN_SECTION_SHIFT 0 489 #endif 490 491 /* 492 * Define the bit shifts to access each section. For non-existant 493 * sections we define the shift as 0; that plus a 0 mask ensures 494 * the compiler will optimise away reference to them. 495 */ 496 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 497 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 498 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 499 500 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 501 #ifdef NODE_NOT_IN_PAGEFLAGS 502 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 503 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 504 SECTIONS_PGOFF : ZONES_PGOFF) 505 #else 506 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 507 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 508 NODES_PGOFF : ZONES_PGOFF) 509 #endif 510 511 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 512 513 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 514 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 515 #endif 516 517 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 518 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 519 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 520 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 521 522 static inline enum zone_type page_zonenum(struct page *page) 523 { 524 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 525 } 526 527 /* 528 * The identification function is only used by the buddy allocator for 529 * determining if two pages could be buddies. We are not really 530 * identifying a zone since we could be using a the section number 531 * id if we have not node id available in page flags. 532 * We guarantee only that it will return the same value for two 533 * combinable pages in a zone. 534 */ 535 static inline int page_zone_id(struct page *page) 536 { 537 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 538 } 539 540 static inline int zone_to_nid(struct zone *zone) 541 { 542 #ifdef CONFIG_NUMA 543 return zone->node; 544 #else 545 return 0; 546 #endif 547 } 548 549 #ifdef NODE_NOT_IN_PAGE_FLAGS 550 extern int page_to_nid(struct page *page); 551 #else 552 static inline int page_to_nid(struct page *page) 553 { 554 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 555 } 556 #endif 557 558 static inline struct zone *page_zone(struct page *page) 559 { 560 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 561 } 562 563 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 564 static inline unsigned long page_to_section(struct page *page) 565 { 566 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 567 } 568 #endif 569 570 static inline void set_page_zone(struct page *page, enum zone_type zone) 571 { 572 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 573 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 574 } 575 576 static inline void set_page_node(struct page *page, unsigned long node) 577 { 578 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 579 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 580 } 581 582 static inline void set_page_section(struct page *page, unsigned long section) 583 { 584 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 585 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 586 } 587 588 static inline void set_page_links(struct page *page, enum zone_type zone, 589 unsigned long node, unsigned long pfn) 590 { 591 set_page_zone(page, zone); 592 set_page_node(page, node); 593 set_page_section(page, pfn_to_section_nr(pfn)); 594 } 595 596 /* 597 * Some inline functions in vmstat.h depend on page_zone() 598 */ 599 #include <linux/vmstat.h> 600 601 static __always_inline void *lowmem_page_address(struct page *page) 602 { 603 return __va(PFN_PHYS(page_to_pfn(page))); 604 } 605 606 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 607 #define HASHED_PAGE_VIRTUAL 608 #endif 609 610 #if defined(WANT_PAGE_VIRTUAL) 611 #define page_address(page) ((page)->virtual) 612 #define set_page_address(page, address) \ 613 do { \ 614 (page)->virtual = (address); \ 615 } while(0) 616 #define page_address_init() do { } while(0) 617 #endif 618 619 #if defined(HASHED_PAGE_VIRTUAL) 620 void *page_address(struct page *page); 621 void set_page_address(struct page *page, void *virtual); 622 void page_address_init(void); 623 #endif 624 625 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 626 #define page_address(page) lowmem_page_address(page) 627 #define set_page_address(page, address) do { } while(0) 628 #define page_address_init() do { } while(0) 629 #endif 630 631 /* 632 * On an anonymous page mapped into a user virtual memory area, 633 * page->mapping points to its anon_vma, not to a struct address_space; 634 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 635 * 636 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 637 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 638 * and then page->mapping points, not to an anon_vma, but to a private 639 * structure which KSM associates with that merged page. See ksm.h. 640 * 641 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 642 * 643 * Please note that, confusingly, "page_mapping" refers to the inode 644 * address_space which maps the page from disk; whereas "page_mapped" 645 * refers to user virtual address space into which the page is mapped. 646 */ 647 #define PAGE_MAPPING_ANON 1 648 #define PAGE_MAPPING_KSM 2 649 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 650 651 extern struct address_space swapper_space; 652 static inline struct address_space *page_mapping(struct page *page) 653 { 654 struct address_space *mapping = page->mapping; 655 656 VM_BUG_ON(PageSlab(page)); 657 if (unlikely(PageSwapCache(page))) 658 mapping = &swapper_space; 659 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 660 mapping = NULL; 661 return mapping; 662 } 663 664 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 665 static inline void *page_rmapping(struct page *page) 666 { 667 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 668 } 669 670 static inline int PageAnon(struct page *page) 671 { 672 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 673 } 674 675 /* 676 * Return the pagecache index of the passed page. Regular pagecache pages 677 * use ->index whereas swapcache pages use ->private 678 */ 679 static inline pgoff_t page_index(struct page *page) 680 { 681 if (unlikely(PageSwapCache(page))) 682 return page_private(page); 683 return page->index; 684 } 685 686 /* 687 * The atomic page->_mapcount, like _count, starts from -1: 688 * so that transitions both from it and to it can be tracked, 689 * using atomic_inc_and_test and atomic_add_negative(-1). 690 */ 691 static inline void reset_page_mapcount(struct page *page) 692 { 693 atomic_set(&(page)->_mapcount, -1); 694 } 695 696 static inline int page_mapcount(struct page *page) 697 { 698 return atomic_read(&(page)->_mapcount) + 1; 699 } 700 701 /* 702 * Return true if this page is mapped into pagetables. 703 */ 704 static inline int page_mapped(struct page *page) 705 { 706 return atomic_read(&(page)->_mapcount) >= 0; 707 } 708 709 /* 710 * Different kinds of faults, as returned by handle_mm_fault(). 711 * Used to decide whether a process gets delivered SIGBUS or 712 * just gets major/minor fault counters bumped up. 713 */ 714 715 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 716 717 #define VM_FAULT_OOM 0x0001 718 #define VM_FAULT_SIGBUS 0x0002 719 #define VM_FAULT_MAJOR 0x0004 720 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 721 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */ 722 723 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 724 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 725 726 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON) 727 728 /* 729 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 730 */ 731 extern void pagefault_out_of_memory(void); 732 733 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 734 735 extern void show_free_areas(void); 736 737 int shmem_lock(struct file *file, int lock, struct user_struct *user); 738 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 739 int shmem_zero_setup(struct vm_area_struct *); 740 741 #ifndef CONFIG_MMU 742 extern unsigned long shmem_get_unmapped_area(struct file *file, 743 unsigned long addr, 744 unsigned long len, 745 unsigned long pgoff, 746 unsigned long flags); 747 #endif 748 749 extern int can_do_mlock(void); 750 extern int user_shm_lock(size_t, struct user_struct *); 751 extern void user_shm_unlock(size_t, struct user_struct *); 752 753 /* 754 * Parameter block passed down to zap_pte_range in exceptional cases. 755 */ 756 struct zap_details { 757 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 758 struct address_space *check_mapping; /* Check page->mapping if set */ 759 pgoff_t first_index; /* Lowest page->index to unmap */ 760 pgoff_t last_index; /* Highest page->index to unmap */ 761 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 762 unsigned long truncate_count; /* Compare vm_truncate_count */ 763 }; 764 765 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 766 pte_t pte); 767 768 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 769 unsigned long size); 770 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 771 unsigned long size, struct zap_details *); 772 unsigned long unmap_vmas(struct mmu_gather **tlb, 773 struct vm_area_struct *start_vma, unsigned long start_addr, 774 unsigned long end_addr, unsigned long *nr_accounted, 775 struct zap_details *); 776 777 /** 778 * mm_walk - callbacks for walk_page_range 779 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 780 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 781 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 782 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 783 * @pte_hole: if set, called for each hole at all levels 784 * @hugetlb_entry: if set, called for each hugetlb entry 785 * 786 * (see walk_page_range for more details) 787 */ 788 struct mm_walk { 789 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 790 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 791 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 792 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 793 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 794 int (*hugetlb_entry)(pte_t *, unsigned long, 795 unsigned long, unsigned long, struct mm_walk *); 796 struct mm_struct *mm; 797 void *private; 798 }; 799 800 int walk_page_range(unsigned long addr, unsigned long end, 801 struct mm_walk *walk); 802 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 803 unsigned long end, unsigned long floor, unsigned long ceiling); 804 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 805 struct vm_area_struct *vma); 806 void unmap_mapping_range(struct address_space *mapping, 807 loff_t const holebegin, loff_t const holelen, int even_cows); 808 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 809 unsigned long *pfn); 810 int follow_phys(struct vm_area_struct *vma, unsigned long address, 811 unsigned int flags, unsigned long *prot, resource_size_t *phys); 812 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 813 void *buf, int len, int write); 814 815 static inline void unmap_shared_mapping_range(struct address_space *mapping, 816 loff_t const holebegin, loff_t const holelen) 817 { 818 unmap_mapping_range(mapping, holebegin, holelen, 0); 819 } 820 821 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 822 extern void truncate_setsize(struct inode *inode, loff_t newsize); 823 extern int vmtruncate(struct inode *inode, loff_t offset); 824 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 825 826 int truncate_inode_page(struct address_space *mapping, struct page *page); 827 int generic_error_remove_page(struct address_space *mapping, struct page *page); 828 829 int invalidate_inode_page(struct page *page); 830 831 #ifdef CONFIG_MMU 832 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 833 unsigned long address, unsigned int flags); 834 #else 835 static inline int handle_mm_fault(struct mm_struct *mm, 836 struct vm_area_struct *vma, unsigned long address, 837 unsigned int flags) 838 { 839 /* should never happen if there's no MMU */ 840 BUG(); 841 return VM_FAULT_SIGBUS; 842 } 843 #endif 844 845 extern int make_pages_present(unsigned long addr, unsigned long end); 846 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 847 848 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 849 unsigned long start, int nr_pages, int write, int force, 850 struct page **pages, struct vm_area_struct **vmas); 851 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 852 struct page **pages); 853 struct page *get_dump_page(unsigned long addr); 854 855 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 856 extern void do_invalidatepage(struct page *page, unsigned long offset); 857 858 int __set_page_dirty_nobuffers(struct page *page); 859 int __set_page_dirty_no_writeback(struct page *page); 860 int redirty_page_for_writepage(struct writeback_control *wbc, 861 struct page *page); 862 void account_page_dirtied(struct page *page, struct address_space *mapping); 863 int set_page_dirty(struct page *page); 864 int set_page_dirty_lock(struct page *page); 865 int clear_page_dirty_for_io(struct page *page); 866 867 /* Is the vma a continuation of the stack vma above it? */ 868 static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr) 869 { 870 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); 871 } 872 873 extern unsigned long move_page_tables(struct vm_area_struct *vma, 874 unsigned long old_addr, struct vm_area_struct *new_vma, 875 unsigned long new_addr, unsigned long len); 876 extern unsigned long do_mremap(unsigned long addr, 877 unsigned long old_len, unsigned long new_len, 878 unsigned long flags, unsigned long new_addr); 879 extern int mprotect_fixup(struct vm_area_struct *vma, 880 struct vm_area_struct **pprev, unsigned long start, 881 unsigned long end, unsigned long newflags); 882 883 /* 884 * doesn't attempt to fault and will return short. 885 */ 886 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 887 struct page **pages); 888 /* 889 * per-process(per-mm_struct) statistics. 890 */ 891 #if defined(SPLIT_RSS_COUNTING) 892 /* 893 * The mm counters are not protected by its page_table_lock, 894 * so must be incremented atomically. 895 */ 896 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 897 { 898 atomic_long_set(&mm->rss_stat.count[member], value); 899 } 900 901 unsigned long get_mm_counter(struct mm_struct *mm, int member); 902 903 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 904 { 905 atomic_long_add(value, &mm->rss_stat.count[member]); 906 } 907 908 static inline void inc_mm_counter(struct mm_struct *mm, int member) 909 { 910 atomic_long_inc(&mm->rss_stat.count[member]); 911 } 912 913 static inline void dec_mm_counter(struct mm_struct *mm, int member) 914 { 915 atomic_long_dec(&mm->rss_stat.count[member]); 916 } 917 918 #else /* !USE_SPLIT_PTLOCKS */ 919 /* 920 * The mm counters are protected by its page_table_lock, 921 * so can be incremented directly. 922 */ 923 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 924 { 925 mm->rss_stat.count[member] = value; 926 } 927 928 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 929 { 930 return mm->rss_stat.count[member]; 931 } 932 933 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 934 { 935 mm->rss_stat.count[member] += value; 936 } 937 938 static inline void inc_mm_counter(struct mm_struct *mm, int member) 939 { 940 mm->rss_stat.count[member]++; 941 } 942 943 static inline void dec_mm_counter(struct mm_struct *mm, int member) 944 { 945 mm->rss_stat.count[member]--; 946 } 947 948 #endif /* !USE_SPLIT_PTLOCKS */ 949 950 static inline unsigned long get_mm_rss(struct mm_struct *mm) 951 { 952 return get_mm_counter(mm, MM_FILEPAGES) + 953 get_mm_counter(mm, MM_ANONPAGES); 954 } 955 956 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 957 { 958 return max(mm->hiwater_rss, get_mm_rss(mm)); 959 } 960 961 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 962 { 963 return max(mm->hiwater_vm, mm->total_vm); 964 } 965 966 static inline void update_hiwater_rss(struct mm_struct *mm) 967 { 968 unsigned long _rss = get_mm_rss(mm); 969 970 if ((mm)->hiwater_rss < _rss) 971 (mm)->hiwater_rss = _rss; 972 } 973 974 static inline void update_hiwater_vm(struct mm_struct *mm) 975 { 976 if (mm->hiwater_vm < mm->total_vm) 977 mm->hiwater_vm = mm->total_vm; 978 } 979 980 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 981 struct mm_struct *mm) 982 { 983 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 984 985 if (*maxrss < hiwater_rss) 986 *maxrss = hiwater_rss; 987 } 988 989 #if defined(SPLIT_RSS_COUNTING) 990 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); 991 #else 992 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) 993 { 994 } 995 #endif 996 997 /* 998 * A callback you can register to apply pressure to ageable caches. 999 * 1000 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 1001 * look through the least-recently-used 'nr_to_scan' entries and 1002 * attempt to free them up. It should return the number of objects 1003 * which remain in the cache. If it returns -1, it means it cannot do 1004 * any scanning at this time (eg. there is a risk of deadlock). 1005 * 1006 * The 'gfpmask' refers to the allocation we are currently trying to 1007 * fulfil. 1008 * 1009 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 1010 * querying the cache size, so a fastpath for that case is appropriate. 1011 */ 1012 struct shrinker { 1013 int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask); 1014 int seeks; /* seeks to recreate an obj */ 1015 1016 /* These are for internal use */ 1017 struct list_head list; 1018 long nr; /* objs pending delete */ 1019 }; 1020 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 1021 extern void register_shrinker(struct shrinker *); 1022 extern void unregister_shrinker(struct shrinker *); 1023 1024 int vma_wants_writenotify(struct vm_area_struct *vma); 1025 1026 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 1027 1028 #ifdef __PAGETABLE_PUD_FOLDED 1029 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1030 unsigned long address) 1031 { 1032 return 0; 1033 } 1034 #else 1035 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1036 #endif 1037 1038 #ifdef __PAGETABLE_PMD_FOLDED 1039 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1040 unsigned long address) 1041 { 1042 return 0; 1043 } 1044 #else 1045 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1046 #endif 1047 1048 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 1049 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1050 1051 /* 1052 * The following ifdef needed to get the 4level-fixup.h header to work. 1053 * Remove it when 4level-fixup.h has been removed. 1054 */ 1055 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1056 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1057 { 1058 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1059 NULL: pud_offset(pgd, address); 1060 } 1061 1062 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1063 { 1064 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1065 NULL: pmd_offset(pud, address); 1066 } 1067 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1068 1069 #if USE_SPLIT_PTLOCKS 1070 /* 1071 * We tuck a spinlock to guard each pagetable page into its struct page, 1072 * at page->private, with BUILD_BUG_ON to make sure that this will not 1073 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 1074 * When freeing, reset page->mapping so free_pages_check won't complain. 1075 */ 1076 #define __pte_lockptr(page) &((page)->ptl) 1077 #define pte_lock_init(_page) do { \ 1078 spin_lock_init(__pte_lockptr(_page)); \ 1079 } while (0) 1080 #define pte_lock_deinit(page) ((page)->mapping = NULL) 1081 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 1082 #else /* !USE_SPLIT_PTLOCKS */ 1083 /* 1084 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1085 */ 1086 #define pte_lock_init(page) do {} while (0) 1087 #define pte_lock_deinit(page) do {} while (0) 1088 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 1089 #endif /* USE_SPLIT_PTLOCKS */ 1090 1091 static inline void pgtable_page_ctor(struct page *page) 1092 { 1093 pte_lock_init(page); 1094 inc_zone_page_state(page, NR_PAGETABLE); 1095 } 1096 1097 static inline void pgtable_page_dtor(struct page *page) 1098 { 1099 pte_lock_deinit(page); 1100 dec_zone_page_state(page, NR_PAGETABLE); 1101 } 1102 1103 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1104 ({ \ 1105 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1106 pte_t *__pte = pte_offset_map(pmd, address); \ 1107 *(ptlp) = __ptl; \ 1108 spin_lock(__ptl); \ 1109 __pte; \ 1110 }) 1111 1112 #define pte_unmap_unlock(pte, ptl) do { \ 1113 spin_unlock(ptl); \ 1114 pte_unmap(pte); \ 1115 } while (0) 1116 1117 #define pte_alloc_map(mm, pmd, address) \ 1118 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1119 NULL: pte_offset_map(pmd, address)) 1120 1121 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1122 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1123 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1124 1125 #define pte_alloc_kernel(pmd, address) \ 1126 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1127 NULL: pte_offset_kernel(pmd, address)) 1128 1129 extern void free_area_init(unsigned long * zones_size); 1130 extern void free_area_init_node(int nid, unsigned long * zones_size, 1131 unsigned long zone_start_pfn, unsigned long *zholes_size); 1132 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1133 /* 1134 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1135 * zones, allocate the backing mem_map and account for memory holes in a more 1136 * architecture independent manner. This is a substitute for creating the 1137 * zone_sizes[] and zholes_size[] arrays and passing them to 1138 * free_area_init_node() 1139 * 1140 * An architecture is expected to register range of page frames backed by 1141 * physical memory with add_active_range() before calling 1142 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1143 * usage, an architecture is expected to do something like 1144 * 1145 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1146 * max_highmem_pfn}; 1147 * for_each_valid_physical_page_range() 1148 * add_active_range(node_id, start_pfn, end_pfn) 1149 * free_area_init_nodes(max_zone_pfns); 1150 * 1151 * If the architecture guarantees that there are no holes in the ranges 1152 * registered with add_active_range(), free_bootmem_active_regions() 1153 * will call free_bootmem_node() for each registered physical page range. 1154 * Similarly sparse_memory_present_with_active_regions() calls 1155 * memory_present() for each range when SPARSEMEM is enabled. 1156 * 1157 * See mm/page_alloc.c for more information on each function exposed by 1158 * CONFIG_ARCH_POPULATES_NODE_MAP 1159 */ 1160 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1161 extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1162 unsigned long end_pfn); 1163 extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1164 unsigned long end_pfn); 1165 extern void remove_all_active_ranges(void); 1166 void sort_node_map(void); 1167 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1168 unsigned long end_pfn); 1169 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1170 unsigned long end_pfn); 1171 extern void get_pfn_range_for_nid(unsigned int nid, 1172 unsigned long *start_pfn, unsigned long *end_pfn); 1173 extern unsigned long find_min_pfn_with_active_regions(void); 1174 extern void free_bootmem_with_active_regions(int nid, 1175 unsigned long max_low_pfn); 1176 int add_from_early_node_map(struct range *range, int az, 1177 int nr_range, int nid); 1178 u64 __init find_memory_core_early(int nid, u64 size, u64 align, 1179 u64 goal, u64 limit); 1180 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align, 1181 u64 goal, u64 limit); 1182 typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1183 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1184 extern void sparse_memory_present_with_active_regions(int nid); 1185 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1186 1187 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ 1188 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1189 static inline int __early_pfn_to_nid(unsigned long pfn) 1190 { 1191 return 0; 1192 } 1193 #else 1194 /* please see mm/page_alloc.c */ 1195 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1196 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1197 /* there is a per-arch backend function. */ 1198 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1199 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1200 #endif 1201 1202 extern void set_dma_reserve(unsigned long new_dma_reserve); 1203 extern void memmap_init_zone(unsigned long, int, unsigned long, 1204 unsigned long, enum memmap_context); 1205 extern void setup_per_zone_wmarks(void); 1206 extern void calculate_zone_inactive_ratio(struct zone *zone); 1207 extern void mem_init(void); 1208 extern void __init mmap_init(void); 1209 extern void show_mem(void); 1210 extern void si_meminfo(struct sysinfo * val); 1211 extern void si_meminfo_node(struct sysinfo *val, int nid); 1212 extern int after_bootmem; 1213 1214 extern void setup_per_cpu_pageset(void); 1215 1216 extern void zone_pcp_update(struct zone *zone); 1217 1218 /* nommu.c */ 1219 extern atomic_long_t mmap_pages_allocated; 1220 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1221 1222 /* prio_tree.c */ 1223 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1224 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1225 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1226 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1227 struct prio_tree_iter *iter); 1228 1229 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1230 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1231 (vma = vma_prio_tree_next(vma, iter)); ) 1232 1233 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1234 struct list_head *list) 1235 { 1236 vma->shared.vm_set.parent = NULL; 1237 list_add_tail(&vma->shared.vm_set.list, list); 1238 } 1239 1240 /* mmap.c */ 1241 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1242 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1243 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1244 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1245 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1246 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1247 struct mempolicy *); 1248 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1249 extern int split_vma(struct mm_struct *, 1250 struct vm_area_struct *, unsigned long addr, int new_below); 1251 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1252 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1253 struct rb_node **, struct rb_node *); 1254 extern void unlink_file_vma(struct vm_area_struct *); 1255 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1256 unsigned long addr, unsigned long len, pgoff_t pgoff); 1257 extern void exit_mmap(struct mm_struct *); 1258 1259 extern int mm_take_all_locks(struct mm_struct *mm); 1260 extern void mm_drop_all_locks(struct mm_struct *mm); 1261 1262 #ifdef CONFIG_PROC_FS 1263 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1264 extern void added_exe_file_vma(struct mm_struct *mm); 1265 extern void removed_exe_file_vma(struct mm_struct *mm); 1266 #else 1267 static inline void added_exe_file_vma(struct mm_struct *mm) 1268 {} 1269 1270 static inline void removed_exe_file_vma(struct mm_struct *mm) 1271 {} 1272 #endif /* CONFIG_PROC_FS */ 1273 1274 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1275 extern int install_special_mapping(struct mm_struct *mm, 1276 unsigned long addr, unsigned long len, 1277 unsigned long flags, struct page **pages); 1278 1279 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1280 1281 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1282 unsigned long len, unsigned long prot, 1283 unsigned long flag, unsigned long pgoff); 1284 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1285 unsigned long len, unsigned long flags, 1286 unsigned int vm_flags, unsigned long pgoff); 1287 1288 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1289 unsigned long len, unsigned long prot, 1290 unsigned long flag, unsigned long offset) 1291 { 1292 unsigned long ret = -EINVAL; 1293 if ((offset + PAGE_ALIGN(len)) < offset) 1294 goto out; 1295 if (!(offset & ~PAGE_MASK)) 1296 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1297 out: 1298 return ret; 1299 } 1300 1301 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1302 1303 extern unsigned long do_brk(unsigned long, unsigned long); 1304 1305 /* filemap.c */ 1306 extern unsigned long page_unuse(struct page *); 1307 extern void truncate_inode_pages(struct address_space *, loff_t); 1308 extern void truncate_inode_pages_range(struct address_space *, 1309 loff_t lstart, loff_t lend); 1310 1311 /* generic vm_area_ops exported for stackable file systems */ 1312 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1313 1314 /* mm/page-writeback.c */ 1315 int write_one_page(struct page *page, int wait); 1316 void task_dirty_inc(struct task_struct *tsk); 1317 1318 /* readahead.c */ 1319 #define VM_MAX_READAHEAD 128 /* kbytes */ 1320 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1321 1322 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1323 pgoff_t offset, unsigned long nr_to_read); 1324 1325 void page_cache_sync_readahead(struct address_space *mapping, 1326 struct file_ra_state *ra, 1327 struct file *filp, 1328 pgoff_t offset, 1329 unsigned long size); 1330 1331 void page_cache_async_readahead(struct address_space *mapping, 1332 struct file_ra_state *ra, 1333 struct file *filp, 1334 struct page *pg, 1335 pgoff_t offset, 1336 unsigned long size); 1337 1338 unsigned long max_sane_readahead(unsigned long nr); 1339 unsigned long ra_submit(struct file_ra_state *ra, 1340 struct address_space *mapping, 1341 struct file *filp); 1342 1343 /* Do stack extension */ 1344 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1345 #if VM_GROWSUP 1346 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1347 #else 1348 #define expand_upwards(vma, address) do { } while (0) 1349 #endif 1350 extern int expand_stack_downwards(struct vm_area_struct *vma, 1351 unsigned long address); 1352 1353 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1354 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1355 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1356 struct vm_area_struct **pprev); 1357 1358 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1359 NULL if none. Assume start_addr < end_addr. */ 1360 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1361 { 1362 struct vm_area_struct * vma = find_vma(mm,start_addr); 1363 1364 if (vma && end_addr <= vma->vm_start) 1365 vma = NULL; 1366 return vma; 1367 } 1368 1369 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1370 { 1371 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1372 } 1373 1374 #ifdef CONFIG_MMU 1375 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1376 #else 1377 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) 1378 { 1379 return __pgprot(0); 1380 } 1381 #endif 1382 1383 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1384 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1385 unsigned long pfn, unsigned long size, pgprot_t); 1386 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1387 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1388 unsigned long pfn); 1389 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1390 unsigned long pfn); 1391 1392 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1393 unsigned int foll_flags); 1394 #define FOLL_WRITE 0x01 /* check pte is writable */ 1395 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1396 #define FOLL_GET 0x04 /* do get_page on page */ 1397 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1398 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1399 1400 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1401 void *data); 1402 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1403 unsigned long size, pte_fn_t fn, void *data); 1404 1405 #ifdef CONFIG_PROC_FS 1406 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1407 #else 1408 static inline void vm_stat_account(struct mm_struct *mm, 1409 unsigned long flags, struct file *file, long pages) 1410 { 1411 } 1412 #endif /* CONFIG_PROC_FS */ 1413 1414 #ifdef CONFIG_DEBUG_PAGEALLOC 1415 extern int debug_pagealloc_enabled; 1416 1417 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1418 1419 static inline void enable_debug_pagealloc(void) 1420 { 1421 debug_pagealloc_enabled = 1; 1422 } 1423 #ifdef CONFIG_HIBERNATION 1424 extern bool kernel_page_present(struct page *page); 1425 #endif /* CONFIG_HIBERNATION */ 1426 #else 1427 static inline void 1428 kernel_map_pages(struct page *page, int numpages, int enable) {} 1429 static inline void enable_debug_pagealloc(void) 1430 { 1431 } 1432 #ifdef CONFIG_HIBERNATION 1433 static inline bool kernel_page_present(struct page *page) { return true; } 1434 #endif /* CONFIG_HIBERNATION */ 1435 #endif 1436 1437 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1438 #ifdef __HAVE_ARCH_GATE_AREA 1439 int in_gate_area_no_task(unsigned long addr); 1440 int in_gate_area(struct task_struct *task, unsigned long addr); 1441 #else 1442 int in_gate_area_no_task(unsigned long addr); 1443 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1444 #endif /* __HAVE_ARCH_GATE_AREA */ 1445 1446 int drop_caches_sysctl_handler(struct ctl_table *, int, 1447 void __user *, size_t *, loff_t *); 1448 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1449 unsigned long lru_pages); 1450 1451 #ifndef CONFIG_MMU 1452 #define randomize_va_space 0 1453 #else 1454 extern int randomize_va_space; 1455 #endif 1456 1457 const char * arch_vma_name(struct vm_area_struct *vma); 1458 void print_vma_addr(char *prefix, unsigned long rip); 1459 1460 void sparse_mem_maps_populate_node(struct page **map_map, 1461 unsigned long pnum_begin, 1462 unsigned long pnum_end, 1463 unsigned long map_count, 1464 int nodeid); 1465 1466 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1467 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1468 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1469 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1470 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1471 void *vmemmap_alloc_block(unsigned long size, int node); 1472 void *vmemmap_alloc_block_buf(unsigned long size, int node); 1473 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1474 int vmemmap_populate_basepages(struct page *start_page, 1475 unsigned long pages, int node); 1476 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1477 void vmemmap_populate_print_last(void); 1478 1479 1480 enum mf_flags { 1481 MF_COUNT_INCREASED = 1 << 0, 1482 }; 1483 extern void memory_failure(unsigned long pfn, int trapno); 1484 extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1485 extern int unpoison_memory(unsigned long pfn); 1486 extern int sysctl_memory_failure_early_kill; 1487 extern int sysctl_memory_failure_recovery; 1488 extern void shake_page(struct page *p, int access); 1489 extern atomic_long_t mce_bad_pages; 1490 extern int soft_offline_page(struct page *page, int flags); 1491 #ifdef CONFIG_MEMORY_FAILURE 1492 int is_hwpoison_address(unsigned long addr); 1493 #else 1494 static inline int is_hwpoison_address(unsigned long addr) 1495 { 1496 return 0; 1497 } 1498 #endif 1499 1500 extern void dump_page(struct page *page); 1501 1502 #endif /* __KERNEL__ */ 1503 #endif /* _LINUX_MM_H */ 1504