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/atomic.h> 14 #include <linux/debug_locks.h> 15 #include <linux/mm_types.h> 16 #include <linux/range.h> 17 #include <linux/pfn.h> 18 #include <linux/bit_spinlock.h> 19 #include <linux/shrinker.h> 20 21 struct mempolicy; 22 struct anon_vma; 23 struct file_ra_state; 24 struct user_struct; 25 struct writeback_control; 26 27 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 28 extern unsigned long max_mapnr; 29 #endif 30 31 extern unsigned long num_physpages; 32 extern unsigned long totalram_pages; 33 extern void * high_memory; 34 extern int page_cluster; 35 36 #ifdef CONFIG_SYSCTL 37 extern int sysctl_legacy_va_layout; 38 #else 39 #define sysctl_legacy_va_layout 0 40 #endif 41 42 #include <asm/page.h> 43 #include <asm/pgtable.h> 44 #include <asm/processor.h> 45 46 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 47 48 /* to align the pointer to the (next) page boundary */ 49 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 50 51 /* 52 * Linux kernel virtual memory manager primitives. 53 * The idea being to have a "virtual" mm in the same way 54 * we have a virtual fs - giving a cleaner interface to the 55 * mm details, and allowing different kinds of memory mappings 56 * (from shared memory to executable loading to arbitrary 57 * mmap() functions). 58 */ 59 60 extern struct kmem_cache *vm_area_cachep; 61 62 #ifndef CONFIG_MMU 63 extern struct rb_root nommu_region_tree; 64 extern struct rw_semaphore nommu_region_sem; 65 66 extern unsigned int kobjsize(const void *objp); 67 #endif 68 69 /* 70 * vm_flags in vm_area_struct, see mm_types.h. 71 */ 72 #define VM_READ 0x00000001 /* currently active flags */ 73 #define VM_WRITE 0x00000002 74 #define VM_EXEC 0x00000004 75 #define VM_SHARED 0x00000008 76 77 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 78 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 79 #define VM_MAYWRITE 0x00000020 80 #define VM_MAYEXEC 0x00000040 81 #define VM_MAYSHARE 0x00000080 82 83 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 84 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64) 85 #define VM_GROWSUP 0x00000200 86 #else 87 #define VM_GROWSUP 0x00000000 88 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */ 89 #endif 90 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 91 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 92 93 #define VM_EXECUTABLE 0x00001000 94 #define VM_LOCKED 0x00002000 95 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 96 97 /* Used by sys_madvise() */ 98 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 99 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 100 101 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 102 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 103 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ 104 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 105 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 106 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 107 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 108 #ifndef CONFIG_TRANSPARENT_HUGEPAGE 109 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ 110 #else 111 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */ 112 #endif 113 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ 114 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ 115 116 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ 117 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 118 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ 119 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ 120 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 121 122 /* Bits set in the VMA until the stack is in its final location */ 123 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) 124 125 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 126 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 127 #endif 128 129 #ifdef CONFIG_STACK_GROWSUP 130 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 131 #else 132 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 133 #endif 134 135 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 136 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 137 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 138 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 139 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 140 141 /* 142 * Special vmas that are non-mergable, non-mlock()able. 143 * Note: mm/huge_memory.c VM_NO_THP depends on this definition. 144 */ 145 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 146 147 /* 148 * mapping from the currently active vm_flags protection bits (the 149 * low four bits) to a page protection mask.. 150 */ 151 extern pgprot_t protection_map[16]; 152 153 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 154 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 155 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 156 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ 157 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ 158 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ 159 160 /* 161 * This interface is used by x86 PAT code to identify a pfn mapping that is 162 * linear over entire vma. This is to optimize PAT code that deals with 163 * marking the physical region with a particular prot. This is not for generic 164 * mm use. Note also that this check will not work if the pfn mapping is 165 * linear for a vma starting at physical address 0. In which case PAT code 166 * falls back to slow path of reserving physical range page by page. 167 */ 168 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 169 { 170 return !!(vma->vm_flags & VM_PFN_AT_MMAP); 171 } 172 173 static inline int is_pfn_mapping(struct vm_area_struct *vma) 174 { 175 return !!(vma->vm_flags & VM_PFNMAP); 176 } 177 178 /* 179 * vm_fault is filled by the the pagefault handler and passed to the vma's 180 * ->fault function. The vma's ->fault is responsible for returning a bitmask 181 * of VM_FAULT_xxx flags that give details about how the fault was handled. 182 * 183 * pgoff should be used in favour of virtual_address, if possible. If pgoff 184 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 185 * mapping support. 186 */ 187 struct vm_fault { 188 unsigned int flags; /* FAULT_FLAG_xxx flags */ 189 pgoff_t pgoff; /* Logical page offset based on vma */ 190 void __user *virtual_address; /* Faulting virtual address */ 191 192 struct page *page; /* ->fault handlers should return a 193 * page here, unless VM_FAULT_NOPAGE 194 * is set (which is also implied by 195 * VM_FAULT_ERROR). 196 */ 197 }; 198 199 /* 200 * These are the virtual MM functions - opening of an area, closing and 201 * unmapping it (needed to keep files on disk up-to-date etc), pointer 202 * to the functions called when a no-page or a wp-page exception occurs. 203 */ 204 struct vm_operations_struct { 205 void (*open)(struct vm_area_struct * area); 206 void (*close)(struct vm_area_struct * area); 207 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 208 209 /* notification that a previously read-only page is about to become 210 * writable, if an error is returned it will cause a SIGBUS */ 211 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 212 213 /* called by access_process_vm when get_user_pages() fails, typically 214 * for use by special VMAs that can switch between memory and hardware 215 */ 216 int (*access)(struct vm_area_struct *vma, unsigned long addr, 217 void *buf, int len, int write); 218 #ifdef CONFIG_NUMA 219 /* 220 * set_policy() op must add a reference to any non-NULL @new mempolicy 221 * to hold the policy upon return. Caller should pass NULL @new to 222 * remove a policy and fall back to surrounding context--i.e. do not 223 * install a MPOL_DEFAULT policy, nor the task or system default 224 * mempolicy. 225 */ 226 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 227 228 /* 229 * get_policy() op must add reference [mpol_get()] to any policy at 230 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 231 * in mm/mempolicy.c will do this automatically. 232 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 233 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 234 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 235 * must return NULL--i.e., do not "fallback" to task or system default 236 * policy. 237 */ 238 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 239 unsigned long addr); 240 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 241 const nodemask_t *to, unsigned long flags); 242 #endif 243 }; 244 245 struct mmu_gather; 246 struct inode; 247 248 #define page_private(page) ((page)->private) 249 #define set_page_private(page, v) ((page)->private = (v)) 250 251 /* 252 * FIXME: take this include out, include page-flags.h in 253 * files which need it (119 of them) 254 */ 255 #include <linux/page-flags.h> 256 #include <linux/huge_mm.h> 257 258 /* 259 * Methods to modify the page usage count. 260 * 261 * What counts for a page usage: 262 * - cache mapping (page->mapping) 263 * - private data (page->private) 264 * - page mapped in a task's page tables, each mapping 265 * is counted separately 266 * 267 * Also, many kernel routines increase the page count before a critical 268 * routine so they can be sure the page doesn't go away from under them. 269 */ 270 271 /* 272 * Drop a ref, return true if the refcount fell to zero (the page has no users) 273 */ 274 static inline int put_page_testzero(struct page *page) 275 { 276 VM_BUG_ON(atomic_read(&page->_count) == 0); 277 return atomic_dec_and_test(&page->_count); 278 } 279 280 /* 281 * Try to grab a ref unless the page has a refcount of zero, return false if 282 * that is the case. 283 */ 284 static inline int get_page_unless_zero(struct page *page) 285 { 286 return atomic_inc_not_zero(&page->_count); 287 } 288 289 extern int page_is_ram(unsigned long pfn); 290 291 /* Support for virtually mapped pages */ 292 struct page *vmalloc_to_page(const void *addr); 293 unsigned long vmalloc_to_pfn(const void *addr); 294 295 /* 296 * Determine if an address is within the vmalloc range 297 * 298 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 299 * is no special casing required. 300 */ 301 static inline int is_vmalloc_addr(const void *x) 302 { 303 #ifdef CONFIG_MMU 304 unsigned long addr = (unsigned long)x; 305 306 return addr >= VMALLOC_START && addr < VMALLOC_END; 307 #else 308 return 0; 309 #endif 310 } 311 #ifdef CONFIG_MMU 312 extern int is_vmalloc_or_module_addr(const void *x); 313 #else 314 static inline int is_vmalloc_or_module_addr(const void *x) 315 { 316 return 0; 317 } 318 #endif 319 320 static inline void compound_lock(struct page *page) 321 { 322 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 323 bit_spin_lock(PG_compound_lock, &page->flags); 324 #endif 325 } 326 327 static inline void compound_unlock(struct page *page) 328 { 329 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 330 bit_spin_unlock(PG_compound_lock, &page->flags); 331 #endif 332 } 333 334 static inline unsigned long compound_lock_irqsave(struct page *page) 335 { 336 unsigned long uninitialized_var(flags); 337 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 338 local_irq_save(flags); 339 compound_lock(page); 340 #endif 341 return flags; 342 } 343 344 static inline void compound_unlock_irqrestore(struct page *page, 345 unsigned long flags) 346 { 347 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 348 compound_unlock(page); 349 local_irq_restore(flags); 350 #endif 351 } 352 353 static inline struct page *compound_head(struct page *page) 354 { 355 if (unlikely(PageTail(page))) 356 return page->first_page; 357 return page; 358 } 359 360 /* 361 * The atomic page->_mapcount, starts from -1: so that transitions 362 * both from it and to it can be tracked, using atomic_inc_and_test 363 * and atomic_add_negative(-1). 364 */ 365 static inline void reset_page_mapcount(struct page *page) 366 { 367 atomic_set(&(page)->_mapcount, -1); 368 } 369 370 static inline int page_mapcount(struct page *page) 371 { 372 return atomic_read(&(page)->_mapcount) + 1; 373 } 374 375 static inline int page_count(struct page *page) 376 { 377 return atomic_read(&compound_head(page)->_count); 378 } 379 380 static inline void get_huge_page_tail(struct page *page) 381 { 382 /* 383 * __split_huge_page_refcount() cannot run 384 * from under us. 385 */ 386 VM_BUG_ON(page_mapcount(page) < 0); 387 VM_BUG_ON(atomic_read(&page->_count) != 0); 388 atomic_inc(&page->_mapcount); 389 } 390 391 extern bool __get_page_tail(struct page *page); 392 393 static inline void get_page(struct page *page) 394 { 395 if (unlikely(PageTail(page))) 396 if (likely(__get_page_tail(page))) 397 return; 398 /* 399 * Getting a normal page or the head of a compound page 400 * requires to already have an elevated page->_count. 401 */ 402 VM_BUG_ON(atomic_read(&page->_count) <= 0); 403 atomic_inc(&page->_count); 404 } 405 406 static inline struct page *virt_to_head_page(const void *x) 407 { 408 struct page *page = virt_to_page(x); 409 return compound_head(page); 410 } 411 412 /* 413 * Setup the page count before being freed into the page allocator for 414 * the first time (boot or memory hotplug) 415 */ 416 static inline void init_page_count(struct page *page) 417 { 418 atomic_set(&page->_count, 1); 419 } 420 421 /* 422 * PageBuddy() indicate that the page is free and in the buddy system 423 * (see mm/page_alloc.c). 424 * 425 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to 426 * -2 so that an underflow of the page_mapcount() won't be mistaken 427 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very 428 * efficiently by most CPU architectures. 429 */ 430 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) 431 432 static inline int PageBuddy(struct page *page) 433 { 434 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; 435 } 436 437 static inline void __SetPageBuddy(struct page *page) 438 { 439 VM_BUG_ON(atomic_read(&page->_mapcount) != -1); 440 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); 441 } 442 443 static inline void __ClearPageBuddy(struct page *page) 444 { 445 VM_BUG_ON(!PageBuddy(page)); 446 atomic_set(&page->_mapcount, -1); 447 } 448 449 void put_page(struct page *page); 450 void put_pages_list(struct list_head *pages); 451 452 void split_page(struct page *page, unsigned int order); 453 int split_free_page(struct page *page); 454 455 /* 456 * Compound pages have a destructor function. Provide a 457 * prototype for that function and accessor functions. 458 * These are _only_ valid on the head of a PG_compound page. 459 */ 460 typedef void compound_page_dtor(struct page *); 461 462 static inline void set_compound_page_dtor(struct page *page, 463 compound_page_dtor *dtor) 464 { 465 page[1].lru.next = (void *)dtor; 466 } 467 468 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 469 { 470 return (compound_page_dtor *)page[1].lru.next; 471 } 472 473 static inline int compound_order(struct page *page) 474 { 475 if (!PageHead(page)) 476 return 0; 477 return (unsigned long)page[1].lru.prev; 478 } 479 480 static inline int compound_trans_order(struct page *page) 481 { 482 int order; 483 unsigned long flags; 484 485 if (!PageHead(page)) 486 return 0; 487 488 flags = compound_lock_irqsave(page); 489 order = compound_order(page); 490 compound_unlock_irqrestore(page, flags); 491 return order; 492 } 493 494 static inline void set_compound_order(struct page *page, unsigned long order) 495 { 496 page[1].lru.prev = (void *)order; 497 } 498 499 #ifdef CONFIG_MMU 500 /* 501 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when 502 * servicing faults for write access. In the normal case, do always want 503 * pte_mkwrite. But get_user_pages can cause write faults for mappings 504 * that do not have writing enabled, when used by access_process_vm. 505 */ 506 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) 507 { 508 if (likely(vma->vm_flags & VM_WRITE)) 509 pte = pte_mkwrite(pte); 510 return pte; 511 } 512 #endif 513 514 /* 515 * Multiple processes may "see" the same page. E.g. for untouched 516 * mappings of /dev/null, all processes see the same page full of 517 * zeroes, and text pages of executables and shared libraries have 518 * only one copy in memory, at most, normally. 519 * 520 * For the non-reserved pages, page_count(page) denotes a reference count. 521 * page_count() == 0 means the page is free. page->lru is then used for 522 * freelist management in the buddy allocator. 523 * page_count() > 0 means the page has been allocated. 524 * 525 * Pages are allocated by the slab allocator in order to provide memory 526 * to kmalloc and kmem_cache_alloc. In this case, the management of the 527 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 528 * unless a particular usage is carefully commented. (the responsibility of 529 * freeing the kmalloc memory is the caller's, of course). 530 * 531 * A page may be used by anyone else who does a __get_free_page(). 532 * In this case, page_count still tracks the references, and should only 533 * be used through the normal accessor functions. The top bits of page->flags 534 * and page->virtual store page management information, but all other fields 535 * are unused and could be used privately, carefully. The management of this 536 * page is the responsibility of the one who allocated it, and those who have 537 * subsequently been given references to it. 538 * 539 * The other pages (we may call them "pagecache pages") are completely 540 * managed by the Linux memory manager: I/O, buffers, swapping etc. 541 * The following discussion applies only to them. 542 * 543 * A pagecache page contains an opaque `private' member, which belongs to the 544 * page's address_space. Usually, this is the address of a circular list of 545 * the page's disk buffers. PG_private must be set to tell the VM to call 546 * into the filesystem to release these pages. 547 * 548 * A page may belong to an inode's memory mapping. In this case, page->mapping 549 * is the pointer to the inode, and page->index is the file offset of the page, 550 * in units of PAGE_CACHE_SIZE. 551 * 552 * If pagecache pages are not associated with an inode, they are said to be 553 * anonymous pages. These may become associated with the swapcache, and in that 554 * case PG_swapcache is set, and page->private is an offset into the swapcache. 555 * 556 * In either case (swapcache or inode backed), the pagecache itself holds one 557 * reference to the page. Setting PG_private should also increment the 558 * refcount. The each user mapping also has a reference to the page. 559 * 560 * The pagecache pages are stored in a per-mapping radix tree, which is 561 * rooted at mapping->page_tree, and indexed by offset. 562 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 563 * lists, we instead now tag pages as dirty/writeback in the radix tree. 564 * 565 * All pagecache pages may be subject to I/O: 566 * - inode pages may need to be read from disk, 567 * - inode pages which have been modified and are MAP_SHARED may need 568 * to be written back to the inode on disk, 569 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 570 * modified may need to be swapped out to swap space and (later) to be read 571 * back into memory. 572 */ 573 574 /* 575 * The zone field is never updated after free_area_init_core() 576 * sets it, so none of the operations on it need to be atomic. 577 */ 578 579 580 /* 581 * page->flags layout: 582 * 583 * There are three possibilities for how page->flags get 584 * laid out. The first is for the normal case, without 585 * sparsemem. The second is for sparsemem when there is 586 * plenty of space for node and section. The last is when 587 * we have run out of space and have to fall back to an 588 * alternate (slower) way of determining the node. 589 * 590 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 591 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 592 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 593 */ 594 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 595 #define SECTIONS_WIDTH SECTIONS_SHIFT 596 #else 597 #define SECTIONS_WIDTH 0 598 #endif 599 600 #define ZONES_WIDTH ZONES_SHIFT 601 602 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 603 #define NODES_WIDTH NODES_SHIFT 604 #else 605 #ifdef CONFIG_SPARSEMEM_VMEMMAP 606 #error "Vmemmap: No space for nodes field in page flags" 607 #endif 608 #define NODES_WIDTH 0 609 #endif 610 611 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 612 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 613 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 614 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 615 616 /* 617 * We are going to use the flags for the page to node mapping if its in 618 * there. This includes the case where there is no node, so it is implicit. 619 */ 620 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 621 #define NODE_NOT_IN_PAGE_FLAGS 622 #endif 623 624 /* 625 * Define the bit shifts to access each section. For non-existent 626 * sections we define the shift as 0; that plus a 0 mask ensures 627 * the compiler will optimise away reference to them. 628 */ 629 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 630 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 631 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 632 633 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ 634 #ifdef NODE_NOT_IN_PAGE_FLAGS 635 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 636 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 637 SECTIONS_PGOFF : ZONES_PGOFF) 638 #else 639 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 640 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 641 NODES_PGOFF : ZONES_PGOFF) 642 #endif 643 644 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 645 646 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 647 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 648 #endif 649 650 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 651 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 652 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 653 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 654 655 static inline enum zone_type page_zonenum(const struct page *page) 656 { 657 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 658 } 659 660 /* 661 * The identification function is only used by the buddy allocator for 662 * determining if two pages could be buddies. We are not really 663 * identifying a zone since we could be using a the section number 664 * id if we have not node id available in page flags. 665 * We guarantee only that it will return the same value for two 666 * combinable pages in a zone. 667 */ 668 static inline int page_zone_id(struct page *page) 669 { 670 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 671 } 672 673 static inline int zone_to_nid(struct zone *zone) 674 { 675 #ifdef CONFIG_NUMA 676 return zone->node; 677 #else 678 return 0; 679 #endif 680 } 681 682 #ifdef NODE_NOT_IN_PAGE_FLAGS 683 extern int page_to_nid(const struct page *page); 684 #else 685 static inline int page_to_nid(const struct page *page) 686 { 687 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 688 } 689 #endif 690 691 static inline struct zone *page_zone(const struct page *page) 692 { 693 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 694 } 695 696 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 697 static inline void set_page_section(struct page *page, unsigned long section) 698 { 699 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 700 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 701 } 702 703 static inline unsigned long page_to_section(const struct page *page) 704 { 705 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 706 } 707 #endif 708 709 static inline void set_page_zone(struct page *page, enum zone_type zone) 710 { 711 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 712 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 713 } 714 715 static inline void set_page_node(struct page *page, unsigned long node) 716 { 717 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 718 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 719 } 720 721 static inline void set_page_links(struct page *page, enum zone_type zone, 722 unsigned long node, unsigned long pfn) 723 { 724 set_page_zone(page, zone); 725 set_page_node(page, node); 726 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 727 set_page_section(page, pfn_to_section_nr(pfn)); 728 #endif 729 } 730 731 /* 732 * Some inline functions in vmstat.h depend on page_zone() 733 */ 734 #include <linux/vmstat.h> 735 736 static __always_inline void *lowmem_page_address(const struct page *page) 737 { 738 return __va(PFN_PHYS(page_to_pfn(page))); 739 } 740 741 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 742 #define HASHED_PAGE_VIRTUAL 743 #endif 744 745 #if defined(WANT_PAGE_VIRTUAL) 746 #define page_address(page) ((page)->virtual) 747 #define set_page_address(page, address) \ 748 do { \ 749 (page)->virtual = (address); \ 750 } while(0) 751 #define page_address_init() do { } while(0) 752 #endif 753 754 #if defined(HASHED_PAGE_VIRTUAL) 755 void *page_address(const struct page *page); 756 void set_page_address(struct page *page, void *virtual); 757 void page_address_init(void); 758 #endif 759 760 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 761 #define page_address(page) lowmem_page_address(page) 762 #define set_page_address(page, address) do { } while(0) 763 #define page_address_init() do { } while(0) 764 #endif 765 766 /* 767 * On an anonymous page mapped into a user virtual memory area, 768 * page->mapping points to its anon_vma, not to a struct address_space; 769 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 770 * 771 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 772 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 773 * and then page->mapping points, not to an anon_vma, but to a private 774 * structure which KSM associates with that merged page. See ksm.h. 775 * 776 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 777 * 778 * Please note that, confusingly, "page_mapping" refers to the inode 779 * address_space which maps the page from disk; whereas "page_mapped" 780 * refers to user virtual address space into which the page is mapped. 781 */ 782 #define PAGE_MAPPING_ANON 1 783 #define PAGE_MAPPING_KSM 2 784 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 785 786 extern struct address_space swapper_space; 787 static inline struct address_space *page_mapping(struct page *page) 788 { 789 struct address_space *mapping = page->mapping; 790 791 VM_BUG_ON(PageSlab(page)); 792 if (unlikely(PageSwapCache(page))) 793 mapping = &swapper_space; 794 else if ((unsigned long)mapping & PAGE_MAPPING_ANON) 795 mapping = NULL; 796 return mapping; 797 } 798 799 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 800 static inline void *page_rmapping(struct page *page) 801 { 802 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 803 } 804 805 static inline int PageAnon(struct page *page) 806 { 807 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 808 } 809 810 /* 811 * Return the pagecache index of the passed page. Regular pagecache pages 812 * use ->index whereas swapcache pages use ->private 813 */ 814 static inline pgoff_t page_index(struct page *page) 815 { 816 if (unlikely(PageSwapCache(page))) 817 return page_private(page); 818 return page->index; 819 } 820 821 /* 822 * Return true if this page is mapped into pagetables. 823 */ 824 static inline int page_mapped(struct page *page) 825 { 826 return atomic_read(&(page)->_mapcount) >= 0; 827 } 828 829 /* 830 * Different kinds of faults, as returned by handle_mm_fault(). 831 * Used to decide whether a process gets delivered SIGBUS or 832 * just gets major/minor fault counters bumped up. 833 */ 834 835 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 836 837 #define VM_FAULT_OOM 0x0001 838 #define VM_FAULT_SIGBUS 0x0002 839 #define VM_FAULT_MAJOR 0x0004 840 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 841 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ 842 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ 843 844 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 845 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 846 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ 847 848 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ 849 850 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \ 851 VM_FAULT_HWPOISON_LARGE) 852 853 /* Encode hstate index for a hwpoisoned large page */ 854 #define VM_FAULT_SET_HINDEX(x) ((x) << 12) 855 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) 856 857 /* 858 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 859 */ 860 extern void pagefault_out_of_memory(void); 861 862 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 863 864 /* 865 * Flags passed to show_mem() and show_free_areas() to suppress output in 866 * various contexts. 867 */ 868 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */ 869 870 extern void show_free_areas(unsigned int flags); 871 extern bool skip_free_areas_node(unsigned int flags, int nid); 872 873 int shmem_lock(struct file *file, int lock, struct user_struct *user); 874 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 875 int shmem_zero_setup(struct vm_area_struct *); 876 877 extern int can_do_mlock(void); 878 extern int user_shm_lock(size_t, struct user_struct *); 879 extern void user_shm_unlock(size_t, struct user_struct *); 880 881 /* 882 * Parameter block passed down to zap_pte_range in exceptional cases. 883 */ 884 struct zap_details { 885 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 886 struct address_space *check_mapping; /* Check page->mapping if set */ 887 pgoff_t first_index; /* Lowest page->index to unmap */ 888 pgoff_t last_index; /* Highest page->index to unmap */ 889 }; 890 891 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 892 pte_t pte); 893 894 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 895 unsigned long size); 896 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 897 unsigned long size, struct zap_details *); 898 unsigned long unmap_vmas(struct mmu_gather *tlb, 899 struct vm_area_struct *start_vma, unsigned long start_addr, 900 unsigned long end_addr, unsigned long *nr_accounted, 901 struct zap_details *); 902 903 /** 904 * mm_walk - callbacks for walk_page_range 905 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 906 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 907 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 908 * this handler is required to be able to handle 909 * pmd_trans_huge() pmds. They may simply choose to 910 * split_huge_page() instead of handling it explicitly. 911 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 912 * @pte_hole: if set, called for each hole at all levels 913 * @hugetlb_entry: if set, called for each hugetlb entry 914 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry 915 * is used. 916 * 917 * (see walk_page_range for more details) 918 */ 919 struct mm_walk { 920 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 921 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 922 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 923 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 924 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 925 int (*hugetlb_entry)(pte_t *, unsigned long, 926 unsigned long, unsigned long, struct mm_walk *); 927 struct mm_struct *mm; 928 void *private; 929 }; 930 931 int walk_page_range(unsigned long addr, unsigned long end, 932 struct mm_walk *walk); 933 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 934 unsigned long end, unsigned long floor, unsigned long ceiling); 935 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 936 struct vm_area_struct *vma); 937 void unmap_mapping_range(struct address_space *mapping, 938 loff_t const holebegin, loff_t const holelen, int even_cows); 939 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 940 unsigned long *pfn); 941 int follow_phys(struct vm_area_struct *vma, unsigned long address, 942 unsigned int flags, unsigned long *prot, resource_size_t *phys); 943 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 944 void *buf, int len, int write); 945 946 static inline void unmap_shared_mapping_range(struct address_space *mapping, 947 loff_t const holebegin, loff_t const holelen) 948 { 949 unmap_mapping_range(mapping, holebegin, holelen, 0); 950 } 951 952 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 953 extern void truncate_setsize(struct inode *inode, loff_t newsize); 954 extern int vmtruncate(struct inode *inode, loff_t offset); 955 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 956 957 int truncate_inode_page(struct address_space *mapping, struct page *page); 958 int generic_error_remove_page(struct address_space *mapping, struct page *page); 959 960 int invalidate_inode_page(struct page *page); 961 962 #ifdef CONFIG_MMU 963 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 964 unsigned long address, unsigned int flags); 965 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, 966 unsigned long address, unsigned int fault_flags); 967 #else 968 static inline int handle_mm_fault(struct mm_struct *mm, 969 struct vm_area_struct *vma, unsigned long address, 970 unsigned int flags) 971 { 972 /* should never happen if there's no MMU */ 973 BUG(); 974 return VM_FAULT_SIGBUS; 975 } 976 static inline int fixup_user_fault(struct task_struct *tsk, 977 struct mm_struct *mm, unsigned long address, 978 unsigned int fault_flags) 979 { 980 /* should never happen if there's no MMU */ 981 BUG(); 982 return -EFAULT; 983 } 984 #endif 985 986 extern int make_pages_present(unsigned long addr, unsigned long end); 987 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 988 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, 989 void *buf, int len, int write); 990 991 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 992 unsigned long start, int len, unsigned int foll_flags, 993 struct page **pages, struct vm_area_struct **vmas, 994 int *nonblocking); 995 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 996 unsigned long start, int nr_pages, int write, int force, 997 struct page **pages, struct vm_area_struct **vmas); 998 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 999 struct page **pages); 1000 struct page *get_dump_page(unsigned long addr); 1001 1002 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 1003 extern void do_invalidatepage(struct page *page, unsigned long offset); 1004 1005 int __set_page_dirty_nobuffers(struct page *page); 1006 int __set_page_dirty_no_writeback(struct page *page); 1007 int redirty_page_for_writepage(struct writeback_control *wbc, 1008 struct page *page); 1009 void account_page_dirtied(struct page *page, struct address_space *mapping); 1010 void account_page_writeback(struct page *page); 1011 int set_page_dirty(struct page *page); 1012 int set_page_dirty_lock(struct page *page); 1013 int clear_page_dirty_for_io(struct page *page); 1014 1015 /* Is the vma a continuation of the stack vma above it? */ 1016 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) 1017 { 1018 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); 1019 } 1020 1021 static inline int stack_guard_page_start(struct vm_area_struct *vma, 1022 unsigned long addr) 1023 { 1024 return (vma->vm_flags & VM_GROWSDOWN) && 1025 (vma->vm_start == addr) && 1026 !vma_growsdown(vma->vm_prev, addr); 1027 } 1028 1029 /* Is the vma a continuation of the stack vma below it? */ 1030 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) 1031 { 1032 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); 1033 } 1034 1035 static inline int stack_guard_page_end(struct vm_area_struct *vma, 1036 unsigned long addr) 1037 { 1038 return (vma->vm_flags & VM_GROWSUP) && 1039 (vma->vm_end == addr) && 1040 !vma_growsup(vma->vm_next, addr); 1041 } 1042 1043 extern unsigned long move_page_tables(struct vm_area_struct *vma, 1044 unsigned long old_addr, struct vm_area_struct *new_vma, 1045 unsigned long new_addr, unsigned long len); 1046 extern unsigned long do_mremap(unsigned long addr, 1047 unsigned long old_len, unsigned long new_len, 1048 unsigned long flags, unsigned long new_addr); 1049 extern int mprotect_fixup(struct vm_area_struct *vma, 1050 struct vm_area_struct **pprev, unsigned long start, 1051 unsigned long end, unsigned long newflags); 1052 1053 /* 1054 * doesn't attempt to fault and will return short. 1055 */ 1056 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 1057 struct page **pages); 1058 /* 1059 * per-process(per-mm_struct) statistics. 1060 */ 1061 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 1062 { 1063 atomic_long_set(&mm->rss_stat.count[member], value); 1064 } 1065 1066 #if defined(SPLIT_RSS_COUNTING) 1067 unsigned long get_mm_counter(struct mm_struct *mm, int member); 1068 #else 1069 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 1070 { 1071 return atomic_long_read(&mm->rss_stat.count[member]); 1072 } 1073 #endif 1074 1075 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 1076 { 1077 atomic_long_add(value, &mm->rss_stat.count[member]); 1078 } 1079 1080 static inline void inc_mm_counter(struct mm_struct *mm, int member) 1081 { 1082 atomic_long_inc(&mm->rss_stat.count[member]); 1083 } 1084 1085 static inline void dec_mm_counter(struct mm_struct *mm, int member) 1086 { 1087 atomic_long_dec(&mm->rss_stat.count[member]); 1088 } 1089 1090 static inline unsigned long get_mm_rss(struct mm_struct *mm) 1091 { 1092 return get_mm_counter(mm, MM_FILEPAGES) + 1093 get_mm_counter(mm, MM_ANONPAGES); 1094 } 1095 1096 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 1097 { 1098 return max(mm->hiwater_rss, get_mm_rss(mm)); 1099 } 1100 1101 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 1102 { 1103 return max(mm->hiwater_vm, mm->total_vm); 1104 } 1105 1106 static inline void update_hiwater_rss(struct mm_struct *mm) 1107 { 1108 unsigned long _rss = get_mm_rss(mm); 1109 1110 if ((mm)->hiwater_rss < _rss) 1111 (mm)->hiwater_rss = _rss; 1112 } 1113 1114 static inline void update_hiwater_vm(struct mm_struct *mm) 1115 { 1116 if (mm->hiwater_vm < mm->total_vm) 1117 mm->hiwater_vm = mm->total_vm; 1118 } 1119 1120 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 1121 struct mm_struct *mm) 1122 { 1123 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 1124 1125 if (*maxrss < hiwater_rss) 1126 *maxrss = hiwater_rss; 1127 } 1128 1129 #if defined(SPLIT_RSS_COUNTING) 1130 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); 1131 #else 1132 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) 1133 { 1134 } 1135 #endif 1136 1137 int vma_wants_writenotify(struct vm_area_struct *vma); 1138 1139 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, 1140 spinlock_t **ptl); 1141 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, 1142 spinlock_t **ptl) 1143 { 1144 pte_t *ptep; 1145 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); 1146 return ptep; 1147 } 1148 1149 #ifdef __PAGETABLE_PUD_FOLDED 1150 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1151 unsigned long address) 1152 { 1153 return 0; 1154 } 1155 #else 1156 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1157 #endif 1158 1159 #ifdef __PAGETABLE_PMD_FOLDED 1160 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1161 unsigned long address) 1162 { 1163 return 0; 1164 } 1165 #else 1166 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1167 #endif 1168 1169 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, 1170 pmd_t *pmd, unsigned long address); 1171 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1172 1173 /* 1174 * The following ifdef needed to get the 4level-fixup.h header to work. 1175 * Remove it when 4level-fixup.h has been removed. 1176 */ 1177 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1178 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1179 { 1180 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1181 NULL: pud_offset(pgd, address); 1182 } 1183 1184 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1185 { 1186 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1187 NULL: pmd_offset(pud, address); 1188 } 1189 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1190 1191 #if USE_SPLIT_PTLOCKS 1192 /* 1193 * We tuck a spinlock to guard each pagetable page into its struct page, 1194 * at page->private, with BUILD_BUG_ON to make sure that this will not 1195 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 1196 * When freeing, reset page->mapping so free_pages_check won't complain. 1197 */ 1198 #define __pte_lockptr(page) &((page)->ptl) 1199 #define pte_lock_init(_page) do { \ 1200 spin_lock_init(__pte_lockptr(_page)); \ 1201 } while (0) 1202 #define pte_lock_deinit(page) ((page)->mapping = NULL) 1203 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 1204 #else /* !USE_SPLIT_PTLOCKS */ 1205 /* 1206 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1207 */ 1208 #define pte_lock_init(page) do {} while (0) 1209 #define pte_lock_deinit(page) do {} while (0) 1210 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 1211 #endif /* USE_SPLIT_PTLOCKS */ 1212 1213 static inline void pgtable_page_ctor(struct page *page) 1214 { 1215 pte_lock_init(page); 1216 inc_zone_page_state(page, NR_PAGETABLE); 1217 } 1218 1219 static inline void pgtable_page_dtor(struct page *page) 1220 { 1221 pte_lock_deinit(page); 1222 dec_zone_page_state(page, NR_PAGETABLE); 1223 } 1224 1225 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1226 ({ \ 1227 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1228 pte_t *__pte = pte_offset_map(pmd, address); \ 1229 *(ptlp) = __ptl; \ 1230 spin_lock(__ptl); \ 1231 __pte; \ 1232 }) 1233 1234 #define pte_unmap_unlock(pte, ptl) do { \ 1235 spin_unlock(ptl); \ 1236 pte_unmap(pte); \ 1237 } while (0) 1238 1239 #define pte_alloc_map(mm, vma, pmd, address) \ 1240 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ 1241 pmd, address))? \ 1242 NULL: pte_offset_map(pmd, address)) 1243 1244 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1245 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ 1246 pmd, address))? \ 1247 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1248 1249 #define pte_alloc_kernel(pmd, address) \ 1250 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1251 NULL: pte_offset_kernel(pmd, address)) 1252 1253 extern void free_area_init(unsigned long * zones_size); 1254 extern void free_area_init_node(int nid, unsigned long * zones_size, 1255 unsigned long zone_start_pfn, unsigned long *zholes_size); 1256 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 1257 /* 1258 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its 1259 * zones, allocate the backing mem_map and account for memory holes in a more 1260 * architecture independent manner. This is a substitute for creating the 1261 * zone_sizes[] and zholes_size[] arrays and passing them to 1262 * free_area_init_node() 1263 * 1264 * An architecture is expected to register range of page frames backed by 1265 * physical memory with memblock_add[_node]() before calling 1266 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1267 * usage, an architecture is expected to do something like 1268 * 1269 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1270 * max_highmem_pfn}; 1271 * for_each_valid_physical_page_range() 1272 * memblock_add_node(base, size, nid) 1273 * free_area_init_nodes(max_zone_pfns); 1274 * 1275 * free_bootmem_with_active_regions() calls free_bootmem_node() for each 1276 * registered physical page range. Similarly 1277 * sparse_memory_present_with_active_regions() calls memory_present() for 1278 * each range when SPARSEMEM is enabled. 1279 * 1280 * See mm/page_alloc.c for more information on each function exposed by 1281 * CONFIG_HAVE_MEMBLOCK_NODE_MAP. 1282 */ 1283 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1284 unsigned long node_map_pfn_alignment(void); 1285 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1286 unsigned long end_pfn); 1287 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1288 unsigned long end_pfn); 1289 extern void get_pfn_range_for_nid(unsigned int nid, 1290 unsigned long *start_pfn, unsigned long *end_pfn); 1291 extern unsigned long find_min_pfn_with_active_regions(void); 1292 extern void free_bootmem_with_active_regions(int nid, 1293 unsigned long max_low_pfn); 1294 int add_from_early_node_map(struct range *range, int az, 1295 int nr_range, int nid); 1296 extern void sparse_memory_present_with_active_regions(int nid); 1297 1298 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ 1299 1300 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ 1301 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1302 static inline int __early_pfn_to_nid(unsigned long pfn) 1303 { 1304 return 0; 1305 } 1306 #else 1307 /* please see mm/page_alloc.c */ 1308 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1309 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1310 /* there is a per-arch backend function. */ 1311 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1312 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1313 #endif 1314 1315 extern void set_dma_reserve(unsigned long new_dma_reserve); 1316 extern void memmap_init_zone(unsigned long, int, unsigned long, 1317 unsigned long, enum memmap_context); 1318 extern void setup_per_zone_wmarks(void); 1319 extern int __meminit init_per_zone_wmark_min(void); 1320 extern void mem_init(void); 1321 extern void __init mmap_init(void); 1322 extern void show_mem(unsigned int flags); 1323 extern void si_meminfo(struct sysinfo * val); 1324 extern void si_meminfo_node(struct sysinfo *val, int nid); 1325 extern int after_bootmem; 1326 1327 extern __printf(3, 4) 1328 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); 1329 1330 extern void setup_per_cpu_pageset(void); 1331 1332 extern void zone_pcp_update(struct zone *zone); 1333 1334 /* nommu.c */ 1335 extern atomic_long_t mmap_pages_allocated; 1336 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1337 1338 /* prio_tree.c */ 1339 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1340 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1341 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1342 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1343 struct prio_tree_iter *iter); 1344 1345 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1346 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1347 (vma = vma_prio_tree_next(vma, iter)); ) 1348 1349 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1350 struct list_head *list) 1351 { 1352 vma->shared.vm_set.parent = NULL; 1353 list_add_tail(&vma->shared.vm_set.list, list); 1354 } 1355 1356 /* mmap.c */ 1357 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1358 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1359 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1360 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1361 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1362 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1363 struct mempolicy *); 1364 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1365 extern int split_vma(struct mm_struct *, 1366 struct vm_area_struct *, unsigned long addr, int new_below); 1367 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1368 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1369 struct rb_node **, struct rb_node *); 1370 extern void unlink_file_vma(struct vm_area_struct *); 1371 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1372 unsigned long addr, unsigned long len, pgoff_t pgoff); 1373 extern void exit_mmap(struct mm_struct *); 1374 1375 extern int mm_take_all_locks(struct mm_struct *mm); 1376 extern void mm_drop_all_locks(struct mm_struct *mm); 1377 1378 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1379 extern void added_exe_file_vma(struct mm_struct *mm); 1380 extern void removed_exe_file_vma(struct mm_struct *mm); 1381 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); 1382 extern struct file *get_mm_exe_file(struct mm_struct *mm); 1383 1384 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1385 extern int install_special_mapping(struct mm_struct *mm, 1386 unsigned long addr, unsigned long len, 1387 unsigned long flags, struct page **pages); 1388 1389 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1390 1391 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1392 unsigned long len, unsigned long prot, 1393 unsigned long flag, unsigned long pgoff); 1394 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1395 unsigned long len, unsigned long flags, 1396 vm_flags_t vm_flags, unsigned long pgoff); 1397 1398 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1399 unsigned long len, unsigned long prot, 1400 unsigned long flag, unsigned long offset) 1401 { 1402 unsigned long ret = -EINVAL; 1403 if ((offset + PAGE_ALIGN(len)) < offset) 1404 goto out; 1405 if (!(offset & ~PAGE_MASK)) 1406 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1407 out: 1408 return ret; 1409 } 1410 1411 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1412 1413 extern unsigned long do_brk(unsigned long, unsigned long); 1414 1415 /* truncate.c */ 1416 extern void truncate_inode_pages(struct address_space *, loff_t); 1417 extern void truncate_inode_pages_range(struct address_space *, 1418 loff_t lstart, loff_t lend); 1419 1420 /* generic vm_area_ops exported for stackable file systems */ 1421 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1422 1423 /* mm/page-writeback.c */ 1424 int write_one_page(struct page *page, int wait); 1425 void task_dirty_inc(struct task_struct *tsk); 1426 1427 /* readahead.c */ 1428 #define VM_MAX_READAHEAD 128 /* kbytes */ 1429 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1430 1431 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1432 pgoff_t offset, unsigned long nr_to_read); 1433 1434 void page_cache_sync_readahead(struct address_space *mapping, 1435 struct file_ra_state *ra, 1436 struct file *filp, 1437 pgoff_t offset, 1438 unsigned long size); 1439 1440 void page_cache_async_readahead(struct address_space *mapping, 1441 struct file_ra_state *ra, 1442 struct file *filp, 1443 struct page *pg, 1444 pgoff_t offset, 1445 unsigned long size); 1446 1447 unsigned long max_sane_readahead(unsigned long nr); 1448 unsigned long ra_submit(struct file_ra_state *ra, 1449 struct address_space *mapping, 1450 struct file *filp); 1451 1452 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ 1453 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1454 1455 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ 1456 extern int expand_downwards(struct vm_area_struct *vma, 1457 unsigned long address); 1458 #if VM_GROWSUP 1459 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1460 #else 1461 #define expand_upwards(vma, address) do { } while (0) 1462 #endif 1463 1464 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1465 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1466 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1467 struct vm_area_struct **pprev); 1468 1469 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1470 NULL if none. Assume start_addr < end_addr. */ 1471 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1472 { 1473 struct vm_area_struct * vma = find_vma(mm,start_addr); 1474 1475 if (vma && end_addr <= vma->vm_start) 1476 vma = NULL; 1477 return vma; 1478 } 1479 1480 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1481 { 1482 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1483 } 1484 1485 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ 1486 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, 1487 unsigned long vm_start, unsigned long vm_end) 1488 { 1489 struct vm_area_struct *vma = find_vma(mm, vm_start); 1490 1491 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) 1492 vma = NULL; 1493 1494 return vma; 1495 } 1496 1497 #ifdef CONFIG_MMU 1498 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1499 #else 1500 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) 1501 { 1502 return __pgprot(0); 1503 } 1504 #endif 1505 1506 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1507 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1508 unsigned long pfn, unsigned long size, pgprot_t); 1509 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1510 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1511 unsigned long pfn); 1512 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1513 unsigned long pfn); 1514 1515 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1516 unsigned int foll_flags); 1517 #define FOLL_WRITE 0x01 /* check pte is writable */ 1518 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1519 #define FOLL_GET 0x04 /* do get_page on page */ 1520 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1521 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1522 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO 1523 * and return without waiting upon it */ 1524 #define FOLL_MLOCK 0x40 /* mark page as mlocked */ 1525 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ 1526 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ 1527 1528 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1529 void *data); 1530 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1531 unsigned long size, pte_fn_t fn, void *data); 1532 1533 #ifdef CONFIG_PROC_FS 1534 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1535 #else 1536 static inline void vm_stat_account(struct mm_struct *mm, 1537 unsigned long flags, struct file *file, long pages) 1538 { 1539 } 1540 #endif /* CONFIG_PROC_FS */ 1541 1542 #ifdef CONFIG_DEBUG_PAGEALLOC 1543 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1544 #ifdef CONFIG_HIBERNATION 1545 extern bool kernel_page_present(struct page *page); 1546 #endif /* CONFIG_HIBERNATION */ 1547 #else 1548 static inline void 1549 kernel_map_pages(struct page *page, int numpages, int enable) {} 1550 #ifdef CONFIG_HIBERNATION 1551 static inline bool kernel_page_present(struct page *page) { return true; } 1552 #endif /* CONFIG_HIBERNATION */ 1553 #endif 1554 1555 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); 1556 #ifdef __HAVE_ARCH_GATE_AREA 1557 int in_gate_area_no_mm(unsigned long addr); 1558 int in_gate_area(struct mm_struct *mm, unsigned long addr); 1559 #else 1560 int in_gate_area_no_mm(unsigned long addr); 1561 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) 1562 #endif /* __HAVE_ARCH_GATE_AREA */ 1563 1564 int drop_caches_sysctl_handler(struct ctl_table *, int, 1565 void __user *, size_t *, loff_t *); 1566 unsigned long shrink_slab(struct shrink_control *shrink, 1567 unsigned long nr_pages_scanned, 1568 unsigned long lru_pages); 1569 1570 #ifndef CONFIG_MMU 1571 #define randomize_va_space 0 1572 #else 1573 extern int randomize_va_space; 1574 #endif 1575 1576 const char * arch_vma_name(struct vm_area_struct *vma); 1577 void print_vma_addr(char *prefix, unsigned long rip); 1578 1579 void sparse_mem_maps_populate_node(struct page **map_map, 1580 unsigned long pnum_begin, 1581 unsigned long pnum_end, 1582 unsigned long map_count, 1583 int nodeid); 1584 1585 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1586 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1587 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1588 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1589 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1590 void *vmemmap_alloc_block(unsigned long size, int node); 1591 void *vmemmap_alloc_block_buf(unsigned long size, int node); 1592 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1593 int vmemmap_populate_basepages(struct page *start_page, 1594 unsigned long pages, int node); 1595 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1596 void vmemmap_populate_print_last(void); 1597 1598 1599 enum mf_flags { 1600 MF_COUNT_INCREASED = 1 << 0, 1601 }; 1602 extern void memory_failure(unsigned long pfn, int trapno); 1603 extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1604 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); 1605 extern int unpoison_memory(unsigned long pfn); 1606 extern int sysctl_memory_failure_early_kill; 1607 extern int sysctl_memory_failure_recovery; 1608 extern void shake_page(struct page *p, int access); 1609 extern atomic_long_t mce_bad_pages; 1610 extern int soft_offline_page(struct page *page, int flags); 1611 1612 extern void dump_page(struct page *page); 1613 1614 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) 1615 extern void clear_huge_page(struct page *page, 1616 unsigned long addr, 1617 unsigned int pages_per_huge_page); 1618 extern void copy_user_huge_page(struct page *dst, struct page *src, 1619 unsigned long addr, struct vm_area_struct *vma, 1620 unsigned int pages_per_huge_page); 1621 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ 1622 1623 #ifdef CONFIG_DEBUG_PAGEALLOC 1624 extern unsigned int _debug_guardpage_minorder; 1625 1626 static inline unsigned int debug_guardpage_minorder(void) 1627 { 1628 return _debug_guardpage_minorder; 1629 } 1630 1631 static inline bool page_is_guard(struct page *page) 1632 { 1633 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags); 1634 } 1635 #else 1636 static inline unsigned int debug_guardpage_minorder(void) { return 0; } 1637 static inline bool page_is_guard(struct page *page) { return false; } 1638 #endif /* CONFIG_DEBUG_PAGEALLOC */ 1639 1640 #endif /* __KERNEL__ */ 1641 #endif /* _LINUX_MM_H */ 1642