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 18struct mempolicy; 19struct anon_vma; 20struct file_ra_state; 21struct user_struct; 22struct writeback_control; 23 24#ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 25extern unsigned long max_mapnr; 26#endif 27 28extern unsigned long num_physpages; 29extern unsigned long totalram_pages; 30extern void * high_memory; 31extern int page_cluster; 32 33#ifdef CONFIG_SYSCTL 34extern 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 57extern struct kmem_cache *vm_area_cachep; 58 59#ifndef CONFIG_MMU 60extern struct rb_root nommu_region_tree; 61extern struct rw_semaphore nommu_region_sem; 62 63extern 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 */ 142extern 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 */ 156static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 157{ 158 return (vma->vm_flags & VM_PFN_AT_MMAP); 159} 160 161static 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 */ 175struct 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 */ 192struct 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 233struct mmu_gather; 234struct inode; 235 236#define page_private(page) ((page)->private) 237#define set_page_private(page, v) ((page)->private = (v)) 238 239#include <linux/page-flags.h> 240 241/* 242 * Methods to modify the page usage count. 243 * 244 * What counts for a page usage: 245 * - cache mapping (page->mapping) 246 * - private data (page->private) 247 * - page mapped in a task's page tables, each mapping 248 * is counted separately 249 * 250 * Also, many kernel routines increase the page count before a critical 251 * routine so they can be sure the page doesn't go away from under them. 252 */ 253 254/* 255 * Drop a ref, return true if the refcount fell to zero (the page has no users) 256 */ 257static inline int put_page_testzero(struct page *page) 258{ 259 VM_BUG_ON(atomic_read(&page->_count) == 0); 260 return atomic_dec_and_test(&page->_count); 261} 262 263/* 264 * Try to grab a ref unless the page has a refcount of zero, return false if 265 * that is the case. 266 */ 267static inline int get_page_unless_zero(struct page *page) 268{ 269 return atomic_inc_not_zero(&page->_count); 270} 271 272extern int page_is_ram(unsigned long pfn); 273 274/* Support for virtually mapped pages */ 275struct page *vmalloc_to_page(const void *addr); 276unsigned long vmalloc_to_pfn(const void *addr); 277 278/* 279 * Determine if an address is within the vmalloc range 280 * 281 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 282 * is no special casing required. 283 */ 284static inline int is_vmalloc_addr(const void *x) 285{ 286#ifdef CONFIG_MMU 287 unsigned long addr = (unsigned long)x; 288 289 return addr >= VMALLOC_START && addr < VMALLOC_END; 290#else 291 return 0; 292#endif 293} 294#ifdef CONFIG_MMU 295extern int is_vmalloc_or_module_addr(const void *x); 296#else 297static inline int is_vmalloc_or_module_addr(const void *x) 298{ 299 return 0; 300} 301#endif 302 303static inline struct page *compound_head(struct page *page) 304{ 305 if (unlikely(PageTail(page))) 306 return page->first_page; 307 return page; 308} 309 310static inline int page_count(struct page *page) 311{ 312 return atomic_read(&compound_head(page)->_count); 313} 314 315static inline void get_page(struct page *page) 316{ 317 page = compound_head(page); 318 VM_BUG_ON(atomic_read(&page->_count) == 0); 319 atomic_inc(&page->_count); 320} 321 322static inline struct page *virt_to_head_page(const void *x) 323{ 324 struct page *page = virt_to_page(x); 325 return compound_head(page); 326} 327 328/* 329 * Setup the page count before being freed into the page allocator for 330 * the first time (boot or memory hotplug) 331 */ 332static inline void init_page_count(struct page *page) 333{ 334 atomic_set(&page->_count, 1); 335} 336 337void put_page(struct page *page); 338void put_pages_list(struct list_head *pages); 339 340void split_page(struct page *page, unsigned int order); 341int split_free_page(struct page *page); 342 343/* 344 * Compound pages have a destructor function. Provide a 345 * prototype for that function and accessor functions. 346 * These are _only_ valid on the head of a PG_compound page. 347 */ 348typedef void compound_page_dtor(struct page *); 349 350static inline void set_compound_page_dtor(struct page *page, 351 compound_page_dtor *dtor) 352{ 353 page[1].lru.next = (void *)dtor; 354} 355 356static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 357{ 358 return (compound_page_dtor *)page[1].lru.next; 359} 360 361static inline int compound_order(struct page *page) 362{ 363 if (!PageHead(page)) 364 return 0; 365 return (unsigned long)page[1].lru.prev; 366} 367 368static inline void set_compound_order(struct page *page, unsigned long order) 369{ 370 page[1].lru.prev = (void *)order; 371} 372 373/* 374 * Multiple processes may "see" the same page. E.g. for untouched 375 * mappings of /dev/null, all processes see the same page full of 376 * zeroes, and text pages of executables and shared libraries have 377 * only one copy in memory, at most, normally. 378 * 379 * For the non-reserved pages, page_count(page) denotes a reference count. 380 * page_count() == 0 means the page is free. page->lru is then used for 381 * freelist management in the buddy allocator. 382 * page_count() > 0 means the page has been allocated. 383 * 384 * Pages are allocated by the slab allocator in order to provide memory 385 * to kmalloc and kmem_cache_alloc. In this case, the management of the 386 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 387 * unless a particular usage is carefully commented. (the responsibility of 388 * freeing the kmalloc memory is the caller's, of course). 389 * 390 * A page may be used by anyone else who does a __get_free_page(). 391 * In this case, page_count still tracks the references, and should only 392 * be used through the normal accessor functions. The top bits of page->flags 393 * and page->virtual store page management information, but all other fields 394 * are unused and could be used privately, carefully. The management of this 395 * page is the responsibility of the one who allocated it, and those who have 396 * subsequently been given references to it. 397 * 398 * The other pages (we may call them "pagecache pages") are completely 399 * managed by the Linux memory manager: I/O, buffers, swapping etc. 400 * The following discussion applies only to them. 401 * 402 * A pagecache page contains an opaque `private' member, which belongs to the 403 * page's address_space. Usually, this is the address of a circular list of 404 * the page's disk buffers. PG_private must be set to tell the VM to call 405 * into the filesystem to release these pages. 406 * 407 * A page may belong to an inode's memory mapping. In this case, page->mapping 408 * is the pointer to the inode, and page->index is the file offset of the page, 409 * in units of PAGE_CACHE_SIZE. 410 * 411 * If pagecache pages are not associated with an inode, they are said to be 412 * anonymous pages. These may become associated with the swapcache, and in that 413 * case PG_swapcache is set, and page->private is an offset into the swapcache. 414 * 415 * In either case (swapcache or inode backed), the pagecache itself holds one 416 * reference to the page. Setting PG_private should also increment the 417 * refcount. The each user mapping also has a reference to the page. 418 * 419 * The pagecache pages are stored in a per-mapping radix tree, which is 420 * rooted at mapping->page_tree, and indexed by offset. 421 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 422 * lists, we instead now tag pages as dirty/writeback in the radix tree. 423 * 424 * All pagecache pages may be subject to I/O: 425 * - inode pages may need to be read from disk, 426 * - inode pages which have been modified and are MAP_SHARED may need 427 * to be written back to the inode on disk, 428 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 429 * modified may need to be swapped out to swap space and (later) to be read 430 * back into memory. 431 */ 432 433/* 434 * The zone field is never updated after free_area_init_core() 435 * sets it, so none of the operations on it need to be atomic. 436 */ 437 438 439/* 440 * page->flags layout: 441 * 442 * There are three possibilities for how page->flags get 443 * laid out. The first is for the normal case, without 444 * sparsemem. The second is for sparsemem when there is 445 * plenty of space for node and section. The last is when 446 * we have run out of space and have to fall back to an 447 * alternate (slower) way of determining the node. 448 * 449 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 450 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 451 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 452 */ 453#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 454#define SECTIONS_WIDTH SECTIONS_SHIFT 455#else 456#define SECTIONS_WIDTH 0 457#endif 458 459#define ZONES_WIDTH ZONES_SHIFT 460 461#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 462#define NODES_WIDTH NODES_SHIFT 463#else 464#ifdef CONFIG_SPARSEMEM_VMEMMAP 465#error "Vmemmap: No space for nodes field in page flags" 466#endif 467#define NODES_WIDTH 0 468#endif 469 470/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 471#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 472#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 473#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 474 475/* 476 * We are going to use the flags for the page to node mapping if its in 477 * there. This includes the case where there is no node, so it is implicit. 478 */ 479#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 480#define NODE_NOT_IN_PAGE_FLAGS 481#endif 482 483#ifndef PFN_SECTION_SHIFT 484#define PFN_SECTION_SHIFT 0 485#endif 486 487/* 488 * Define the bit shifts to access each section. For non-existant 489 * sections we define the shift as 0; that plus a 0 mask ensures 490 * the compiler will optimise away reference to them. 491 */ 492#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 493#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 494#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 495 496/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 497#ifdef NODE_NOT_IN_PAGEFLAGS 498#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 499#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 500 SECTIONS_PGOFF : ZONES_PGOFF) 501#else 502#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 503#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 504 NODES_PGOFF : ZONES_PGOFF) 505#endif 506 507#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 508 509#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 510#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 511#endif 512 513#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 514#define NODES_MASK ((1UL << NODES_WIDTH) - 1) 515#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 516#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 517 518static inline enum zone_type page_zonenum(struct page *page) 519{ 520 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 521} 522 523/* 524 * The identification function is only used by the buddy allocator for 525 * determining if two pages could be buddies. We are not really 526 * identifying a zone since we could be using a the section number 527 * id if we have not node id available in page flags. 528 * We guarantee only that it will return the same value for two 529 * combinable pages in a zone. 530 */ 531static inline int page_zone_id(struct page *page) 532{ 533 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 534} 535 536static inline int zone_to_nid(struct zone *zone) 537{ 538#ifdef CONFIG_NUMA 539 return zone->node; 540#else 541 return 0; 542#endif 543} 544 545#ifdef NODE_NOT_IN_PAGE_FLAGS 546extern int page_to_nid(struct page *page); 547#else 548static inline int page_to_nid(struct page *page) 549{ 550 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 551} 552#endif 553 554static inline struct zone *page_zone(struct page *page) 555{ 556 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 557} 558 559#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 560static inline unsigned long page_to_section(struct page *page) 561{ 562 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 563} 564#endif 565 566static inline void set_page_zone(struct page *page, enum zone_type zone) 567{ 568 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 569 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 570} 571 572static inline void set_page_node(struct page *page, unsigned long node) 573{ 574 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 575 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 576} 577 578static inline void set_page_section(struct page *page, unsigned long section) 579{ 580 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 581 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 582} 583 584static inline void set_page_links(struct page *page, enum zone_type zone, 585 unsigned long node, unsigned long pfn) 586{ 587 set_page_zone(page, zone); 588 set_page_node(page, node); 589 set_page_section(page, pfn_to_section_nr(pfn)); 590} 591 592/* 593 * Some inline functions in vmstat.h depend on page_zone() 594 */ 595#include <linux/vmstat.h> 596 597static __always_inline void *lowmem_page_address(struct page *page) 598{ 599 return __va(PFN_PHYS(page_to_pfn(page))); 600} 601 602#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 603#define HASHED_PAGE_VIRTUAL 604#endif 605 606#if defined(WANT_PAGE_VIRTUAL) 607#define page_address(page) ((page)->virtual) 608#define set_page_address(page, address) \ 609 do { \ 610 (page)->virtual = (address); \ 611 } while(0) 612#define page_address_init() do { } while(0) 613#endif 614 615#if defined(HASHED_PAGE_VIRTUAL) 616void *page_address(struct page *page); 617void set_page_address(struct page *page, void *virtual); 618void page_address_init(void); 619#endif 620 621#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 622#define page_address(page) lowmem_page_address(page) 623#define set_page_address(page, address) do { } while(0) 624#define page_address_init() do { } while(0) 625#endif 626 627/* 628 * On an anonymous page mapped into a user virtual memory area, 629 * page->mapping points to its anon_vma, not to a struct address_space; 630 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 631 * 632 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 633 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 634 * and then page->mapping points, not to an anon_vma, but to a private 635 * structure which KSM associates with that merged page. See ksm.h. 636 * 637 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 638 * 639 * Please note that, confusingly, "page_mapping" refers to the inode 640 * address_space which maps the page from disk; whereas "page_mapped" 641 * refers to user virtual address space into which the page is mapped. 642 */ 643#define PAGE_MAPPING_ANON 1 644#define PAGE_MAPPING_KSM 2 645#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 646 647extern struct address_space swapper_space; 648static inline struct address_space *page_mapping(struct page *page) 649{ 650 struct address_space *mapping = page->mapping; 651 652 VM_BUG_ON(PageSlab(page)); 653 if (unlikely(PageSwapCache(page))) 654 mapping = &swapper_space; 655 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 656 mapping = NULL; 657 return mapping; 658} 659 660/* Neutral page->mapping pointer to address_space or anon_vma or other */ 661static inline void *page_rmapping(struct page *page) 662{ 663 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 664} 665 666static inline int PageAnon(struct page *page) 667{ 668 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 669} 670 671/* 672 * Return the pagecache index of the passed page. Regular pagecache pages 673 * use ->index whereas swapcache pages use ->private 674 */ 675static inline pgoff_t page_index(struct page *page) 676{ 677 if (unlikely(PageSwapCache(page))) 678 return page_private(page); 679 return page->index; 680} 681 682/* 683 * The atomic page->_mapcount, like _count, starts from -1: 684 * so that transitions both from it and to it can be tracked, 685 * using atomic_inc_and_test and atomic_add_negative(-1). 686 */ 687static inline void reset_page_mapcount(struct page *page) 688{ 689 atomic_set(&(page)->_mapcount, -1); 690} 691 692static inline int page_mapcount(struct page *page) 693{ 694 return atomic_read(&(page)->_mapcount) + 1; 695} 696 697/* 698 * Return true if this page is mapped into pagetables. 699 */ 700static inline int page_mapped(struct page *page) 701{ 702 return atomic_read(&(page)->_mapcount) >= 0; 703} 704 705/* 706 * Different kinds of faults, as returned by handle_mm_fault(). 707 * Used to decide whether a process gets delivered SIGBUS or 708 * just gets major/minor fault counters bumped up. 709 */ 710 711#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 712 713#define VM_FAULT_OOM 0x0001 714#define VM_FAULT_SIGBUS 0x0002 715#define VM_FAULT_MAJOR 0x0004 716#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 717#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */ 718 719#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 720#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 721 722#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON) 723 724/* 725 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 726 */ 727extern void pagefault_out_of_memory(void); 728 729#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 730 731extern void show_free_areas(void); 732 733int shmem_lock(struct file *file, int lock, struct user_struct *user); 734struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 735int shmem_zero_setup(struct vm_area_struct *); 736 737#ifndef CONFIG_MMU 738extern unsigned long shmem_get_unmapped_area(struct file *file, 739 unsigned long addr, 740 unsigned long len, 741 unsigned long pgoff, 742 unsigned long flags); 743#endif 744 745extern int can_do_mlock(void); 746extern int user_shm_lock(size_t, struct user_struct *); 747extern void user_shm_unlock(size_t, struct user_struct *); 748 749/* 750 * Parameter block passed down to zap_pte_range in exceptional cases. 751 */ 752struct zap_details { 753 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 754 struct address_space *check_mapping; /* Check page->mapping if set */ 755 pgoff_t first_index; /* Lowest page->index to unmap */ 756 pgoff_t last_index; /* Highest page->index to unmap */ 757 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 758 unsigned long truncate_count; /* Compare vm_truncate_count */ 759}; 760 761struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 762 pte_t pte); 763 764int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 765 unsigned long size); 766unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 767 unsigned long size, struct zap_details *); 768unsigned long unmap_vmas(struct mmu_gather **tlb, 769 struct vm_area_struct *start_vma, unsigned long start_addr, 770 unsigned long end_addr, unsigned long *nr_accounted, 771 struct zap_details *); 772 773/** 774 * mm_walk - callbacks for walk_page_range 775 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 776 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 777 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 778 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 779 * @pte_hole: if set, called for each hole at all levels 780 * @hugetlb_entry: if set, called for each hugetlb entry 781 * 782 * (see walk_page_range for more details) 783 */ 784struct mm_walk { 785 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 786 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 787 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 788 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 789 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 790 int (*hugetlb_entry)(pte_t *, unsigned long, 791 unsigned long, unsigned long, struct mm_walk *); 792 struct mm_struct *mm; 793 void *private; 794}; 795 796int walk_page_range(unsigned long addr, unsigned long end, 797 struct mm_walk *walk); 798void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 799 unsigned long end, unsigned long floor, unsigned long ceiling); 800int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 801 struct vm_area_struct *vma); 802void unmap_mapping_range(struct address_space *mapping, 803 loff_t const holebegin, loff_t const holelen, int even_cows); 804int follow_pfn(struct vm_area_struct *vma, unsigned long address, 805 unsigned long *pfn); 806int follow_phys(struct vm_area_struct *vma, unsigned long address, 807 unsigned int flags, unsigned long *prot, resource_size_t *phys); 808int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 809 void *buf, int len, int write); 810 811static inline void unmap_shared_mapping_range(struct address_space *mapping, 812 loff_t const holebegin, loff_t const holelen) 813{ 814 unmap_mapping_range(mapping, holebegin, holelen, 0); 815} 816 817extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 818extern void truncate_setsize(struct inode *inode, loff_t newsize); 819extern int vmtruncate(struct inode *inode, loff_t offset); 820extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 821 822int truncate_inode_page(struct address_space *mapping, struct page *page); 823int generic_error_remove_page(struct address_space *mapping, struct page *page); 824 825int invalidate_inode_page(struct page *page); 826 827#ifdef CONFIG_MMU 828extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 829 unsigned long address, unsigned int flags); 830#else 831static inline int handle_mm_fault(struct mm_struct *mm, 832 struct vm_area_struct *vma, unsigned long address, 833 unsigned int flags) 834{ 835 /* should never happen if there's no MMU */ 836 BUG(); 837 return VM_FAULT_SIGBUS; 838} 839#endif 840 841extern int make_pages_present(unsigned long addr, unsigned long end); 842extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 843 844int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 845 unsigned long start, int nr_pages, int write, int force, 846 struct page **pages, struct vm_area_struct **vmas); 847int get_user_pages_fast(unsigned long start, int nr_pages, int write, 848 struct page **pages); 849struct page *get_dump_page(unsigned long addr); 850 851extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 852extern void do_invalidatepage(struct page *page, unsigned long offset); 853 854int __set_page_dirty_nobuffers(struct page *page); 855int __set_page_dirty_no_writeback(struct page *page); 856int redirty_page_for_writepage(struct writeback_control *wbc, 857 struct page *page); 858void account_page_dirtied(struct page *page, struct address_space *mapping); 859int set_page_dirty(struct page *page); 860int set_page_dirty_lock(struct page *page); 861int clear_page_dirty_for_io(struct page *page); 862 863/* Is the vma a continuation of the stack vma above it? */ 864static inline int vma_stack_continue(struct vm_area_struct *vma, unsigned long addr) 865{ 866 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); 867} 868 869extern unsigned long move_page_tables(struct vm_area_struct *vma, 870 unsigned long old_addr, struct vm_area_struct *new_vma, 871 unsigned long new_addr, unsigned long len); 872extern unsigned long do_mremap(unsigned long addr, 873 unsigned long old_len, unsigned long new_len, 874 unsigned long flags, unsigned long new_addr); 875extern int mprotect_fixup(struct vm_area_struct *vma, 876 struct vm_area_struct **pprev, unsigned long start, 877 unsigned long end, unsigned long newflags); 878 879/* 880 * doesn't attempt to fault and will return short. 881 */ 882int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 883 struct page **pages); 884/* 885 * per-process(per-mm_struct) statistics. 886 */ 887#if defined(SPLIT_RSS_COUNTING) 888/* 889 * The mm counters are not protected by its page_table_lock, 890 * so must be incremented atomically. 891 */ 892static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 893{ 894 atomic_long_set(&mm->rss_stat.count[member], value); 895} 896 897unsigned long get_mm_counter(struct mm_struct *mm, int member); 898 899static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 900{ 901 atomic_long_add(value, &mm->rss_stat.count[member]); 902} 903 904static inline void inc_mm_counter(struct mm_struct *mm, int member) 905{ 906 atomic_long_inc(&mm->rss_stat.count[member]); 907} 908 909static inline void dec_mm_counter(struct mm_struct *mm, int member) 910{ 911 atomic_long_dec(&mm->rss_stat.count[member]); 912} 913 914#else /* !USE_SPLIT_PTLOCKS */ 915/* 916 * The mm counters are protected by its page_table_lock, 917 * so can be incremented directly. 918 */ 919static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 920{ 921 mm->rss_stat.count[member] = value; 922} 923 924static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 925{ 926 return mm->rss_stat.count[member]; 927} 928 929static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 930{ 931 mm->rss_stat.count[member] += value; 932} 933 934static inline void inc_mm_counter(struct mm_struct *mm, int member) 935{ 936 mm->rss_stat.count[member]++; 937} 938 939static inline void dec_mm_counter(struct mm_struct *mm, int member) 940{ 941 mm->rss_stat.count[member]--; 942} 943 944#endif /* !USE_SPLIT_PTLOCKS */ 945 946static inline unsigned long get_mm_rss(struct mm_struct *mm) 947{ 948 return get_mm_counter(mm, MM_FILEPAGES) + 949 get_mm_counter(mm, MM_ANONPAGES); 950} 951 952static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 953{ 954 return max(mm->hiwater_rss, get_mm_rss(mm)); 955} 956 957static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 958{ 959 return max(mm->hiwater_vm, mm->total_vm); 960} 961 962static inline void update_hiwater_rss(struct mm_struct *mm) 963{ 964 unsigned long _rss = get_mm_rss(mm); 965 966 if ((mm)->hiwater_rss < _rss) 967 (mm)->hiwater_rss = _rss; 968} 969 970static inline void update_hiwater_vm(struct mm_struct *mm) 971{ 972 if (mm->hiwater_vm < mm->total_vm) 973 mm->hiwater_vm = mm->total_vm; 974} 975 976static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 977 struct mm_struct *mm) 978{ 979 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 980 981 if (*maxrss < hiwater_rss) 982 *maxrss = hiwater_rss; 983} 984 985#if defined(SPLIT_RSS_COUNTING) 986void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); 987#else 988static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) 989{ 990} 991#endif 992 993/* 994 * A callback you can register to apply pressure to ageable caches. 995 * 996 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 997 * look through the least-recently-used 'nr_to_scan' entries and 998 * attempt to free them up. It should return the number of objects 999 * which remain in the cache. If it returns -1, it means it cannot do 1000 * any scanning at this time (eg. there is a risk of deadlock). 1001 * 1002 * The 'gfpmask' refers to the allocation we are currently trying to 1003 * fulfil. 1004 * 1005 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 1006 * querying the cache size, so a fastpath for that case is appropriate. 1007 */ 1008struct shrinker { 1009 int (*shrink)(struct shrinker *, int nr_to_scan, gfp_t gfp_mask); 1010 int seeks; /* seeks to recreate an obj */ 1011 1012 /* These are for internal use */ 1013 struct list_head list; 1014 long nr; /* objs pending delete */ 1015}; 1016#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 1017extern void register_shrinker(struct shrinker *); 1018extern void unregister_shrinker(struct shrinker *); 1019 1020int vma_wants_writenotify(struct vm_area_struct *vma); 1021 1022extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 1023 1024#ifdef __PAGETABLE_PUD_FOLDED 1025static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1026 unsigned long address) 1027{ 1028 return 0; 1029} 1030#else 1031int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1032#endif 1033 1034#ifdef __PAGETABLE_PMD_FOLDED 1035static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1036 unsigned long address) 1037{ 1038 return 0; 1039} 1040#else 1041int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1042#endif 1043 1044int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 1045int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1046 1047/* 1048 * The following ifdef needed to get the 4level-fixup.h header to work. 1049 * Remove it when 4level-fixup.h has been removed. 1050 */ 1051#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1052static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1053{ 1054 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1055 NULL: pud_offset(pgd, address); 1056} 1057 1058static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1059{ 1060 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1061 NULL: pmd_offset(pud, address); 1062} 1063#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1064 1065#if USE_SPLIT_PTLOCKS 1066/* 1067 * We tuck a spinlock to guard each pagetable page into its struct page, 1068 * at page->private, with BUILD_BUG_ON to make sure that this will not 1069 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 1070 * When freeing, reset page->mapping so free_pages_check won't complain. 1071 */ 1072#define __pte_lockptr(page) &((page)->ptl) 1073#define pte_lock_init(_page) do { \ 1074 spin_lock_init(__pte_lockptr(_page)); \ 1075} while (0) 1076#define pte_lock_deinit(page) ((page)->mapping = NULL) 1077#define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 1078#else /* !USE_SPLIT_PTLOCKS */ 1079/* 1080 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1081 */ 1082#define pte_lock_init(page) do {} while (0) 1083#define pte_lock_deinit(page) do {} while (0) 1084#define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 1085#endif /* USE_SPLIT_PTLOCKS */ 1086 1087static inline void pgtable_page_ctor(struct page *page) 1088{ 1089 pte_lock_init(page); 1090 inc_zone_page_state(page, NR_PAGETABLE); 1091} 1092 1093static inline void pgtable_page_dtor(struct page *page) 1094{ 1095 pte_lock_deinit(page); 1096 dec_zone_page_state(page, NR_PAGETABLE); 1097} 1098 1099#define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1100({ \ 1101 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1102 pte_t *__pte = pte_offset_map(pmd, address); \ 1103 *(ptlp) = __ptl; \ 1104 spin_lock(__ptl); \ 1105 __pte; \ 1106}) 1107 1108#define pte_unmap_unlock(pte, ptl) do { \ 1109 spin_unlock(ptl); \ 1110 pte_unmap(pte); \ 1111} while (0) 1112 1113#define pte_alloc_map(mm, pmd, address) \ 1114 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1115 NULL: pte_offset_map(pmd, address)) 1116 1117#define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1118 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1119 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1120 1121#define pte_alloc_kernel(pmd, address) \ 1122 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1123 NULL: pte_offset_kernel(pmd, address)) 1124 1125extern void free_area_init(unsigned long * zones_size); 1126extern void free_area_init_node(int nid, unsigned long * zones_size, 1127 unsigned long zone_start_pfn, unsigned long *zholes_size); 1128#ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1129/* 1130 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1131 * zones, allocate the backing mem_map and account for memory holes in a more 1132 * architecture independent manner. This is a substitute for creating the 1133 * zone_sizes[] and zholes_size[] arrays and passing them to 1134 * free_area_init_node() 1135 * 1136 * An architecture is expected to register range of page frames backed by 1137 * physical memory with add_active_range() before calling 1138 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1139 * usage, an architecture is expected to do something like 1140 * 1141 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1142 * max_highmem_pfn}; 1143 * for_each_valid_physical_page_range() 1144 * add_active_range(node_id, start_pfn, end_pfn) 1145 * free_area_init_nodes(max_zone_pfns); 1146 * 1147 * If the architecture guarantees that there are no holes in the ranges 1148 * registered with add_active_range(), free_bootmem_active_regions() 1149 * will call free_bootmem_node() for each registered physical page range. 1150 * Similarly sparse_memory_present_with_active_regions() calls 1151 * memory_present() for each range when SPARSEMEM is enabled. 1152 * 1153 * See mm/page_alloc.c for more information on each function exposed by 1154 * CONFIG_ARCH_POPULATES_NODE_MAP 1155 */ 1156extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1157extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1158 unsigned long end_pfn); 1159extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1160 unsigned long end_pfn); 1161extern void remove_all_active_ranges(void); 1162void sort_node_map(void); 1163unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1164 unsigned long end_pfn); 1165extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1166 unsigned long end_pfn); 1167extern void get_pfn_range_for_nid(unsigned int nid, 1168 unsigned long *start_pfn, unsigned long *end_pfn); 1169extern unsigned long find_min_pfn_with_active_regions(void); 1170extern void free_bootmem_with_active_regions(int nid, 1171 unsigned long max_low_pfn); 1172int add_from_early_node_map(struct range *range, int az, 1173 int nr_range, int nid); 1174void *__alloc_memory_core_early(int nodeid, u64 size, u64 align, 1175 u64 goal, u64 limit); 1176typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1177extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1178extern void sparse_memory_present_with_active_regions(int nid); 1179#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1180 1181#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ 1182 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1183static inline int __early_pfn_to_nid(unsigned long pfn) 1184{ 1185 return 0; 1186} 1187#else 1188/* please see mm/page_alloc.c */ 1189extern int __meminit early_pfn_to_nid(unsigned long pfn); 1190#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1191/* there is a per-arch backend function. */ 1192extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1193#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1194#endif 1195 1196extern void set_dma_reserve(unsigned long new_dma_reserve); 1197extern void memmap_init_zone(unsigned long, int, unsigned long, 1198 unsigned long, enum memmap_context); 1199extern void setup_per_zone_wmarks(void); 1200extern void calculate_zone_inactive_ratio(struct zone *zone); 1201extern void mem_init(void); 1202extern void __init mmap_init(void); 1203extern void show_mem(void); 1204extern void si_meminfo(struct sysinfo * val); 1205extern void si_meminfo_node(struct sysinfo *val, int nid); 1206extern int after_bootmem; 1207 1208extern void setup_per_cpu_pageset(void); 1209 1210extern void zone_pcp_update(struct zone *zone); 1211 1212/* nommu.c */ 1213extern atomic_long_t mmap_pages_allocated; 1214extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1215 1216/* prio_tree.c */ 1217void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1218void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1219void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1220struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1221 struct prio_tree_iter *iter); 1222 1223#define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1224 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1225 (vma = vma_prio_tree_next(vma, iter)); ) 1226 1227static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1228 struct list_head *list) 1229{ 1230 vma->shared.vm_set.parent = NULL; 1231 list_add_tail(&vma->shared.vm_set.list, list); 1232} 1233 1234/* mmap.c */ 1235extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1236extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1237 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1238extern struct vm_area_struct *vma_merge(struct mm_struct *, 1239 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1240 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1241 struct mempolicy *); 1242extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1243extern int split_vma(struct mm_struct *, 1244 struct vm_area_struct *, unsigned long addr, int new_below); 1245extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1246extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1247 struct rb_node **, struct rb_node *); 1248extern void unlink_file_vma(struct vm_area_struct *); 1249extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1250 unsigned long addr, unsigned long len, pgoff_t pgoff); 1251extern void exit_mmap(struct mm_struct *); 1252 1253extern int mm_take_all_locks(struct mm_struct *mm); 1254extern void mm_drop_all_locks(struct mm_struct *mm); 1255 1256#ifdef CONFIG_PROC_FS 1257/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1258extern void added_exe_file_vma(struct mm_struct *mm); 1259extern void removed_exe_file_vma(struct mm_struct *mm); 1260#else 1261static inline void added_exe_file_vma(struct mm_struct *mm) 1262{} 1263 1264static inline void removed_exe_file_vma(struct mm_struct *mm) 1265{} 1266#endif /* CONFIG_PROC_FS */ 1267 1268extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1269extern int install_special_mapping(struct mm_struct *mm, 1270 unsigned long addr, unsigned long len, 1271 unsigned long flags, struct page **pages); 1272 1273extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1274 1275extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1276 unsigned long len, unsigned long prot, 1277 unsigned long flag, unsigned long pgoff); 1278extern unsigned long mmap_region(struct file *file, unsigned long addr, 1279 unsigned long len, unsigned long flags, 1280 unsigned int vm_flags, unsigned long pgoff); 1281 1282static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1283 unsigned long len, unsigned long prot, 1284 unsigned long flag, unsigned long offset) 1285{ 1286 unsigned long ret = -EINVAL; 1287 if ((offset + PAGE_ALIGN(len)) < offset) 1288 goto out; 1289 if (!(offset & ~PAGE_MASK)) 1290 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1291out: 1292 return ret; 1293} 1294 1295extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1296 1297extern unsigned long do_brk(unsigned long, unsigned long); 1298 1299/* filemap.c */ 1300extern unsigned long page_unuse(struct page *); 1301extern void truncate_inode_pages(struct address_space *, loff_t); 1302extern void truncate_inode_pages_range(struct address_space *, 1303 loff_t lstart, loff_t lend); 1304 1305/* generic vm_area_ops exported for stackable file systems */ 1306extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1307 1308/* mm/page-writeback.c */ 1309int write_one_page(struct page *page, int wait); 1310void task_dirty_inc(struct task_struct *tsk); 1311 1312/* readahead.c */ 1313#define VM_MAX_READAHEAD 128 /* kbytes */ 1314#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1315 1316int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1317 pgoff_t offset, unsigned long nr_to_read); 1318 1319void page_cache_sync_readahead(struct address_space *mapping, 1320 struct file_ra_state *ra, 1321 struct file *filp, 1322 pgoff_t offset, 1323 unsigned long size); 1324 1325void page_cache_async_readahead(struct address_space *mapping, 1326 struct file_ra_state *ra, 1327 struct file *filp, 1328 struct page *pg, 1329 pgoff_t offset, 1330 unsigned long size); 1331 1332unsigned long max_sane_readahead(unsigned long nr); 1333unsigned long ra_submit(struct file_ra_state *ra, 1334 struct address_space *mapping, 1335 struct file *filp); 1336 1337/* Do stack extension */ 1338extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1339#if VM_GROWSUP 1340extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1341#else 1342 #define expand_upwards(vma, address) do { } while (0) 1343#endif 1344extern int expand_stack_downwards(struct vm_area_struct *vma, 1345 unsigned long address); 1346 1347/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1348extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1349extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1350 struct vm_area_struct **pprev); 1351 1352/* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1353 NULL if none. Assume start_addr < end_addr. */ 1354static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1355{ 1356 struct vm_area_struct * vma = find_vma(mm,start_addr); 1357 1358 if (vma && end_addr <= vma->vm_start) 1359 vma = NULL; 1360 return vma; 1361} 1362 1363static inline unsigned long vma_pages(struct vm_area_struct *vma) 1364{ 1365 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1366} 1367 1368#ifdef CONFIG_MMU 1369pgprot_t vm_get_page_prot(unsigned long vm_flags); 1370#else 1371static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) 1372{ 1373 return __pgprot(0); 1374} 1375#endif 1376 1377struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1378int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1379 unsigned long pfn, unsigned long size, pgprot_t); 1380int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1381int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1382 unsigned long pfn); 1383int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1384 unsigned long pfn); 1385 1386struct page *follow_page(struct vm_area_struct *, unsigned long address, 1387 unsigned int foll_flags); 1388#define FOLL_WRITE 0x01 /* check pte is writable */ 1389#define FOLL_TOUCH 0x02 /* mark page accessed */ 1390#define FOLL_GET 0x04 /* do get_page on page */ 1391#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1392#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1393 1394typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1395 void *data); 1396extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1397 unsigned long size, pte_fn_t fn, void *data); 1398 1399#ifdef CONFIG_PROC_FS 1400void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1401#else 1402static inline void vm_stat_account(struct mm_struct *mm, 1403 unsigned long flags, struct file *file, long pages) 1404{ 1405} 1406#endif /* CONFIG_PROC_FS */ 1407 1408#ifdef CONFIG_DEBUG_PAGEALLOC 1409extern int debug_pagealloc_enabled; 1410 1411extern void kernel_map_pages(struct page *page, int numpages, int enable); 1412 1413static inline void enable_debug_pagealloc(void) 1414{ 1415 debug_pagealloc_enabled = 1; 1416} 1417#ifdef CONFIG_HIBERNATION 1418extern bool kernel_page_present(struct page *page); 1419#endif /* CONFIG_HIBERNATION */ 1420#else 1421static inline void 1422kernel_map_pages(struct page *page, int numpages, int enable) {} 1423static inline void enable_debug_pagealloc(void) 1424{ 1425} 1426#ifdef CONFIG_HIBERNATION 1427static inline bool kernel_page_present(struct page *page) { return true; } 1428#endif /* CONFIG_HIBERNATION */ 1429#endif 1430 1431extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1432#ifdef __HAVE_ARCH_GATE_AREA 1433int in_gate_area_no_task(unsigned long addr); 1434int in_gate_area(struct task_struct *task, unsigned long addr); 1435#else 1436int in_gate_area_no_task(unsigned long addr); 1437#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1438#endif /* __HAVE_ARCH_GATE_AREA */ 1439 1440int drop_caches_sysctl_handler(struct ctl_table *, int, 1441 void __user *, size_t *, loff_t *); 1442unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1443 unsigned long lru_pages); 1444 1445#ifndef CONFIG_MMU 1446#define randomize_va_space 0 1447#else 1448extern int randomize_va_space; 1449#endif 1450 1451const char * arch_vma_name(struct vm_area_struct *vma); 1452void print_vma_addr(char *prefix, unsigned long rip); 1453 1454void sparse_mem_maps_populate_node(struct page **map_map, 1455 unsigned long pnum_begin, 1456 unsigned long pnum_end, 1457 unsigned long map_count, 1458 int nodeid); 1459 1460struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1461pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1462pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1463pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1464pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1465void *vmemmap_alloc_block(unsigned long size, int node); 1466void *vmemmap_alloc_block_buf(unsigned long size, int node); 1467void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1468int vmemmap_populate_basepages(struct page *start_page, 1469 unsigned long pages, int node); 1470int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1471void vmemmap_populate_print_last(void); 1472 1473 1474enum mf_flags { 1475 MF_COUNT_INCREASED = 1 << 0, 1476}; 1477extern void memory_failure(unsigned long pfn, int trapno); 1478extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1479extern int unpoison_memory(unsigned long pfn); 1480extern int sysctl_memory_failure_early_kill; 1481extern int sysctl_memory_failure_recovery; 1482extern void shake_page(struct page *p, int access); 1483extern atomic_long_t mce_bad_pages; 1484extern int soft_offline_page(struct page *page, int flags); 1485#ifdef CONFIG_MEMORY_FAILURE 1486int is_hwpoison_address(unsigned long addr); 1487#else 1488static inline int is_hwpoison_address(unsigned long addr) 1489{ 1490 return 0; 1491} 1492#endif 1493 1494extern void dump_page(struct page *page); 1495 1496#endif /* __KERNEL__ */ 1497#endif /* _LINUX_MM_H */ 1498