1 #ifndef _LINUX_MM_H 2 #define _LINUX_MM_H 3 4 #include <linux/errno.h> 5 6 #ifdef __KERNEL__ 7 8 #include <linux/mmdebug.h> 9 #include <linux/gfp.h> 10 #include <linux/bug.h> 11 #include <linux/list.h> 12 #include <linux/mmzone.h> 13 #include <linux/rbtree.h> 14 #include <linux/atomic.h> 15 #include <linux/debug_locks.h> 16 #include <linux/mm_types.h> 17 #include <linux/range.h> 18 #include <linux/pfn.h> 19 #include <linux/bit_spinlock.h> 20 #include <linux/shrinker.h> 21 22 struct mempolicy; 23 struct anon_vma; 24 struct anon_vma_chain; 25 struct file_ra_state; 26 struct user_struct; 27 struct writeback_control; 28 29 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ 30 extern unsigned long max_mapnr; 31 32 static inline void set_max_mapnr(unsigned long limit) 33 { 34 max_mapnr = limit; 35 } 36 #else 37 static inline void set_max_mapnr(unsigned long limit) { } 38 #endif 39 40 extern unsigned long totalram_pages; 41 extern void * high_memory; 42 extern int page_cluster; 43 44 #ifdef CONFIG_SYSCTL 45 extern int sysctl_legacy_va_layout; 46 #else 47 #define sysctl_legacy_va_layout 0 48 #endif 49 50 #include <asm/page.h> 51 #include <asm/pgtable.h> 52 #include <asm/processor.h> 53 54 #ifndef __pa_symbol 55 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) 56 #endif 57 58 extern unsigned long sysctl_user_reserve_kbytes; 59 extern unsigned long sysctl_admin_reserve_kbytes; 60 61 extern int sysctl_overcommit_memory; 62 extern int sysctl_overcommit_ratio; 63 extern unsigned long sysctl_overcommit_kbytes; 64 65 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, 66 size_t *, loff_t *); 67 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, 68 size_t *, loff_t *); 69 70 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 71 72 /* to align the pointer to the (next) page boundary */ 73 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 74 75 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ 76 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE) 77 78 /* 79 * Linux kernel virtual memory manager primitives. 80 * The idea being to have a "virtual" mm in the same way 81 * we have a virtual fs - giving a cleaner interface to the 82 * mm details, and allowing different kinds of memory mappings 83 * (from shared memory to executable loading to arbitrary 84 * mmap() functions). 85 */ 86 87 extern struct kmem_cache *vm_area_cachep; 88 89 #ifndef CONFIG_MMU 90 extern struct rb_root nommu_region_tree; 91 extern struct rw_semaphore nommu_region_sem; 92 93 extern unsigned int kobjsize(const void *objp); 94 #endif 95 96 /* 97 * vm_flags in vm_area_struct, see mm_types.h. 98 */ 99 #define VM_NONE 0x00000000 100 101 #define VM_READ 0x00000001 /* currently active flags */ 102 #define VM_WRITE 0x00000002 103 #define VM_EXEC 0x00000004 104 #define VM_SHARED 0x00000008 105 106 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 107 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 108 #define VM_MAYWRITE 0x00000020 109 #define VM_MAYEXEC 0x00000040 110 #define VM_MAYSHARE 0x00000080 111 112 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 113 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 114 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 115 116 #define VM_LOCKED 0x00002000 117 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 118 119 /* Used by sys_madvise() */ 120 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 121 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 122 123 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 124 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 125 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 126 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 127 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 128 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 129 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ 130 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ 131 132 #ifdef CONFIG_MEM_SOFT_DIRTY 133 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ 134 #else 135 # define VM_SOFTDIRTY 0 136 #endif 137 138 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 139 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ 140 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ 141 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 142 143 #if defined(CONFIG_X86) 144 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ 145 #elif defined(CONFIG_PPC) 146 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ 147 #elif defined(CONFIG_PARISC) 148 # define VM_GROWSUP VM_ARCH_1 149 #elif defined(CONFIG_METAG) 150 # define VM_GROWSUP VM_ARCH_1 151 #elif defined(CONFIG_IA64) 152 # define VM_GROWSUP VM_ARCH_1 153 #elif !defined(CONFIG_MMU) 154 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ 155 #endif 156 157 #ifndef VM_GROWSUP 158 # define VM_GROWSUP VM_NONE 159 #endif 160 161 /* Bits set in the VMA until the stack is in its final location */ 162 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) 163 164 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 165 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 166 #endif 167 168 #ifdef CONFIG_STACK_GROWSUP 169 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 170 #else 171 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 172 #endif 173 174 /* 175 * Special vmas that are non-mergable, non-mlock()able. 176 * Note: mm/huge_memory.c VM_NO_THP depends on this definition. 177 */ 178 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) 179 180 /* 181 * mapping from the currently active vm_flags protection bits (the 182 * low four bits) to a page protection mask.. 183 */ 184 extern pgprot_t protection_map[16]; 185 186 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 187 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 188 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 189 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */ 190 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */ 191 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */ 192 #define FAULT_FLAG_TRIED 0x40 /* second try */ 193 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */ 194 195 /* 196 * vm_fault is filled by the the pagefault handler and passed to the vma's 197 * ->fault function. The vma's ->fault is responsible for returning a bitmask 198 * of VM_FAULT_xxx flags that give details about how the fault was handled. 199 * 200 * pgoff should be used in favour of virtual_address, if possible. If pgoff 201 * is used, one may implement ->remap_pages to get nonlinear mapping support. 202 */ 203 struct vm_fault { 204 unsigned int flags; /* FAULT_FLAG_xxx flags */ 205 pgoff_t pgoff; /* Logical page offset based on vma */ 206 void __user *virtual_address; /* Faulting virtual address */ 207 208 struct page *page; /* ->fault handlers should return a 209 * page here, unless VM_FAULT_NOPAGE 210 * is set (which is also implied by 211 * VM_FAULT_ERROR). 212 */ 213 }; 214 215 /* 216 * These are the virtual MM functions - opening of an area, closing and 217 * unmapping it (needed to keep files on disk up-to-date etc), pointer 218 * to the functions called when a no-page or a wp-page exception occurs. 219 */ 220 struct vm_operations_struct { 221 void (*open)(struct vm_area_struct * area); 222 void (*close)(struct vm_area_struct * area); 223 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 224 225 /* notification that a previously read-only page is about to become 226 * writable, if an error is returned it will cause a SIGBUS */ 227 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 228 229 /* called by access_process_vm when get_user_pages() fails, typically 230 * for use by special VMAs that can switch between memory and hardware 231 */ 232 int (*access)(struct vm_area_struct *vma, unsigned long addr, 233 void *buf, int len, int write); 234 #ifdef CONFIG_NUMA 235 /* 236 * set_policy() op must add a reference to any non-NULL @new mempolicy 237 * to hold the policy upon return. Caller should pass NULL @new to 238 * remove a policy and fall back to surrounding context--i.e. do not 239 * install a MPOL_DEFAULT policy, nor the task or system default 240 * mempolicy. 241 */ 242 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 243 244 /* 245 * get_policy() op must add reference [mpol_get()] to any policy at 246 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 247 * in mm/mempolicy.c will do this automatically. 248 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 249 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 250 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 251 * must return NULL--i.e., do not "fallback" to task or system default 252 * policy. 253 */ 254 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 255 unsigned long addr); 256 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 257 const nodemask_t *to, unsigned long flags); 258 #endif 259 /* called by sys_remap_file_pages() to populate non-linear mapping */ 260 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr, 261 unsigned long size, pgoff_t pgoff); 262 }; 263 264 struct mmu_gather; 265 struct inode; 266 267 #define page_private(page) ((page)->private) 268 #define set_page_private(page, v) ((page)->private = (v)) 269 270 /* It's valid only if the page is free path or free_list */ 271 static inline void set_freepage_migratetype(struct page *page, int migratetype) 272 { 273 page->index = migratetype; 274 } 275 276 /* It's valid only if the page is free path or free_list */ 277 static inline int get_freepage_migratetype(struct page *page) 278 { 279 return page->index; 280 } 281 282 /* 283 * FIXME: take this include out, include page-flags.h in 284 * files which need it (119 of them) 285 */ 286 #include <linux/page-flags.h> 287 #include <linux/huge_mm.h> 288 289 /* 290 * Methods to modify the page usage count. 291 * 292 * What counts for a page usage: 293 * - cache mapping (page->mapping) 294 * - private data (page->private) 295 * - page mapped in a task's page tables, each mapping 296 * is counted separately 297 * 298 * Also, many kernel routines increase the page count before a critical 299 * routine so they can be sure the page doesn't go away from under them. 300 */ 301 302 /* 303 * Drop a ref, return true if the refcount fell to zero (the page has no users) 304 */ 305 static inline int put_page_testzero(struct page *page) 306 { 307 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page); 308 return atomic_dec_and_test(&page->_count); 309 } 310 311 /* 312 * Try to grab a ref unless the page has a refcount of zero, return false if 313 * that is the case. 314 * This can be called when MMU is off so it must not access 315 * any of the virtual mappings. 316 */ 317 static inline int get_page_unless_zero(struct page *page) 318 { 319 return atomic_inc_not_zero(&page->_count); 320 } 321 322 /* 323 * Try to drop a ref unless the page has a refcount of one, return false if 324 * that is the case. 325 * This is to make sure that the refcount won't become zero after this drop. 326 * This can be called when MMU is off so it must not access 327 * any of the virtual mappings. 328 */ 329 static inline int put_page_unless_one(struct page *page) 330 { 331 return atomic_add_unless(&page->_count, -1, 1); 332 } 333 334 extern int page_is_ram(unsigned long pfn); 335 336 /* Support for virtually mapped pages */ 337 struct page *vmalloc_to_page(const void *addr); 338 unsigned long vmalloc_to_pfn(const void *addr); 339 340 /* 341 * Determine if an address is within the vmalloc range 342 * 343 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 344 * is no special casing required. 345 */ 346 static inline int is_vmalloc_addr(const void *x) 347 { 348 #ifdef CONFIG_MMU 349 unsigned long addr = (unsigned long)x; 350 351 return addr >= VMALLOC_START && addr < VMALLOC_END; 352 #else 353 return 0; 354 #endif 355 } 356 #ifdef CONFIG_MMU 357 extern int is_vmalloc_or_module_addr(const void *x); 358 #else 359 static inline int is_vmalloc_or_module_addr(const void *x) 360 { 361 return 0; 362 } 363 #endif 364 365 static inline void compound_lock(struct page *page) 366 { 367 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 368 VM_BUG_ON_PAGE(PageSlab(page), page); 369 bit_spin_lock(PG_compound_lock, &page->flags); 370 #endif 371 } 372 373 static inline void compound_unlock(struct page *page) 374 { 375 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 376 VM_BUG_ON_PAGE(PageSlab(page), page); 377 bit_spin_unlock(PG_compound_lock, &page->flags); 378 #endif 379 } 380 381 static inline unsigned long compound_lock_irqsave(struct page *page) 382 { 383 unsigned long uninitialized_var(flags); 384 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 385 local_irq_save(flags); 386 compound_lock(page); 387 #endif 388 return flags; 389 } 390 391 static inline void compound_unlock_irqrestore(struct page *page, 392 unsigned long flags) 393 { 394 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 395 compound_unlock(page); 396 local_irq_restore(flags); 397 #endif 398 } 399 400 static inline struct page *compound_head(struct page *page) 401 { 402 if (unlikely(PageTail(page))) { 403 struct page *head = page->first_page; 404 405 /* 406 * page->first_page may be a dangling pointer to an old 407 * compound page, so recheck that it is still a tail 408 * page before returning. 409 */ 410 smp_rmb(); 411 if (likely(PageTail(page))) 412 return head; 413 } 414 return page; 415 } 416 417 /* 418 * The atomic page->_mapcount, starts from -1: so that transitions 419 * both from it and to it can be tracked, using atomic_inc_and_test 420 * and atomic_add_negative(-1). 421 */ 422 static inline void page_mapcount_reset(struct page *page) 423 { 424 atomic_set(&(page)->_mapcount, -1); 425 } 426 427 static inline int page_mapcount(struct page *page) 428 { 429 return atomic_read(&(page)->_mapcount) + 1; 430 } 431 432 static inline int page_count(struct page *page) 433 { 434 return atomic_read(&compound_head(page)->_count); 435 } 436 437 #ifdef CONFIG_HUGETLB_PAGE 438 extern int PageHeadHuge(struct page *page_head); 439 #else /* CONFIG_HUGETLB_PAGE */ 440 static inline int PageHeadHuge(struct page *page_head) 441 { 442 return 0; 443 } 444 #endif /* CONFIG_HUGETLB_PAGE */ 445 446 static inline bool __compound_tail_refcounted(struct page *page) 447 { 448 return !PageSlab(page) && !PageHeadHuge(page); 449 } 450 451 /* 452 * This takes a head page as parameter and tells if the 453 * tail page reference counting can be skipped. 454 * 455 * For this to be safe, PageSlab and PageHeadHuge must remain true on 456 * any given page where they return true here, until all tail pins 457 * have been released. 458 */ 459 static inline bool compound_tail_refcounted(struct page *page) 460 { 461 VM_BUG_ON_PAGE(!PageHead(page), page); 462 return __compound_tail_refcounted(page); 463 } 464 465 static inline void get_huge_page_tail(struct page *page) 466 { 467 /* 468 * __split_huge_page_refcount() cannot run from under us. 469 */ 470 VM_BUG_ON_PAGE(!PageTail(page), page); 471 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); 472 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page); 473 if (compound_tail_refcounted(page->first_page)) 474 atomic_inc(&page->_mapcount); 475 } 476 477 extern bool __get_page_tail(struct page *page); 478 479 static inline void get_page(struct page *page) 480 { 481 if (unlikely(PageTail(page))) 482 if (likely(__get_page_tail(page))) 483 return; 484 /* 485 * Getting a normal page or the head of a compound page 486 * requires to already have an elevated page->_count. 487 */ 488 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); 489 atomic_inc(&page->_count); 490 } 491 492 static inline struct page *virt_to_head_page(const void *x) 493 { 494 struct page *page = virt_to_page(x); 495 return compound_head(page); 496 } 497 498 /* 499 * Setup the page count before being freed into the page allocator for 500 * the first time (boot or memory hotplug) 501 */ 502 static inline void init_page_count(struct page *page) 503 { 504 atomic_set(&page->_count, 1); 505 } 506 507 /* 508 * PageBuddy() indicate that the page is free and in the buddy system 509 * (see mm/page_alloc.c). 510 * 511 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to 512 * -2 so that an underflow of the page_mapcount() won't be mistaken 513 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very 514 * efficiently by most CPU architectures. 515 */ 516 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128) 517 518 static inline int PageBuddy(struct page *page) 519 { 520 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE; 521 } 522 523 static inline void __SetPageBuddy(struct page *page) 524 { 525 VM_BUG_ON_PAGE(atomic_read(&page->_mapcount) != -1, page); 526 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE); 527 } 528 529 static inline void __ClearPageBuddy(struct page *page) 530 { 531 VM_BUG_ON_PAGE(!PageBuddy(page), page); 532 atomic_set(&page->_mapcount, -1); 533 } 534 535 void put_page(struct page *page); 536 void put_pages_list(struct list_head *pages); 537 538 void split_page(struct page *page, unsigned int order); 539 int split_free_page(struct page *page); 540 541 /* 542 * Compound pages have a destructor function. Provide a 543 * prototype for that function and accessor functions. 544 * These are _only_ valid on the head of a PG_compound page. 545 */ 546 typedef void compound_page_dtor(struct page *); 547 548 static inline void set_compound_page_dtor(struct page *page, 549 compound_page_dtor *dtor) 550 { 551 page[1].lru.next = (void *)dtor; 552 } 553 554 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 555 { 556 return (compound_page_dtor *)page[1].lru.next; 557 } 558 559 static inline int compound_order(struct page *page) 560 { 561 if (!PageHead(page)) 562 return 0; 563 return (unsigned long)page[1].lru.prev; 564 } 565 566 static inline void set_compound_order(struct page *page, unsigned long order) 567 { 568 page[1].lru.prev = (void *)order; 569 } 570 571 #ifdef CONFIG_MMU 572 /* 573 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when 574 * servicing faults for write access. In the normal case, do always want 575 * pte_mkwrite. But get_user_pages can cause write faults for mappings 576 * that do not have writing enabled, when used by access_process_vm. 577 */ 578 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) 579 { 580 if (likely(vma->vm_flags & VM_WRITE)) 581 pte = pte_mkwrite(pte); 582 return pte; 583 } 584 #endif 585 586 /* 587 * Multiple processes may "see" the same page. E.g. for untouched 588 * mappings of /dev/null, all processes see the same page full of 589 * zeroes, and text pages of executables and shared libraries have 590 * only one copy in memory, at most, normally. 591 * 592 * For the non-reserved pages, page_count(page) denotes a reference count. 593 * page_count() == 0 means the page is free. page->lru is then used for 594 * freelist management in the buddy allocator. 595 * page_count() > 0 means the page has been allocated. 596 * 597 * Pages are allocated by the slab allocator in order to provide memory 598 * to kmalloc and kmem_cache_alloc. In this case, the management of the 599 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 600 * unless a particular usage is carefully commented. (the responsibility of 601 * freeing the kmalloc memory is the caller's, of course). 602 * 603 * A page may be used by anyone else who does a __get_free_page(). 604 * In this case, page_count still tracks the references, and should only 605 * be used through the normal accessor functions. The top bits of page->flags 606 * and page->virtual store page management information, but all other fields 607 * are unused and could be used privately, carefully. The management of this 608 * page is the responsibility of the one who allocated it, and those who have 609 * subsequently been given references to it. 610 * 611 * The other pages (we may call them "pagecache pages") are completely 612 * managed by the Linux memory manager: I/O, buffers, swapping etc. 613 * The following discussion applies only to them. 614 * 615 * A pagecache page contains an opaque `private' member, which belongs to the 616 * page's address_space. Usually, this is the address of a circular list of 617 * the page's disk buffers. PG_private must be set to tell the VM to call 618 * into the filesystem to release these pages. 619 * 620 * A page may belong to an inode's memory mapping. In this case, page->mapping 621 * is the pointer to the inode, and page->index is the file offset of the page, 622 * in units of PAGE_CACHE_SIZE. 623 * 624 * If pagecache pages are not associated with an inode, they are said to be 625 * anonymous pages. These may become associated with the swapcache, and in that 626 * case PG_swapcache is set, and page->private is an offset into the swapcache. 627 * 628 * In either case (swapcache or inode backed), the pagecache itself holds one 629 * reference to the page. Setting PG_private should also increment the 630 * refcount. The each user mapping also has a reference to the page. 631 * 632 * The pagecache pages are stored in a per-mapping radix tree, which is 633 * rooted at mapping->page_tree, and indexed by offset. 634 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 635 * lists, we instead now tag pages as dirty/writeback in the radix tree. 636 * 637 * All pagecache pages may be subject to I/O: 638 * - inode pages may need to be read from disk, 639 * - inode pages which have been modified and are MAP_SHARED may need 640 * to be written back to the inode on disk, 641 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 642 * modified may need to be swapped out to swap space and (later) to be read 643 * back into memory. 644 */ 645 646 /* 647 * The zone field is never updated after free_area_init_core() 648 * sets it, so none of the operations on it need to be atomic. 649 */ 650 651 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ 652 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 653 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 654 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 655 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) 656 657 /* 658 * Define the bit shifts to access each section. For non-existent 659 * sections we define the shift as 0; that plus a 0 mask ensures 660 * the compiler will optimise away reference to them. 661 */ 662 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 663 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 664 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 665 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) 666 667 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ 668 #ifdef NODE_NOT_IN_PAGE_FLAGS 669 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 670 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 671 SECTIONS_PGOFF : ZONES_PGOFF) 672 #else 673 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 674 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 675 NODES_PGOFF : ZONES_PGOFF) 676 #endif 677 678 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 679 680 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 681 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 682 #endif 683 684 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 685 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 686 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 687 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1) 688 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 689 690 static inline enum zone_type page_zonenum(const struct page *page) 691 { 692 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 693 } 694 695 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 696 #define SECTION_IN_PAGE_FLAGS 697 #endif 698 699 /* 700 * The identification function is mainly used by the buddy allocator for 701 * determining if two pages could be buddies. We are not really identifying 702 * the zone since we could be using the section number id if we do not have 703 * node id available in page flags. 704 * We only guarantee that it will return the same value for two combinable 705 * pages in a zone. 706 */ 707 static inline int page_zone_id(struct page *page) 708 { 709 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 710 } 711 712 static inline int zone_to_nid(struct zone *zone) 713 { 714 #ifdef CONFIG_NUMA 715 return zone->node; 716 #else 717 return 0; 718 #endif 719 } 720 721 #ifdef NODE_NOT_IN_PAGE_FLAGS 722 extern int page_to_nid(const struct page *page); 723 #else 724 static inline int page_to_nid(const struct page *page) 725 { 726 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 727 } 728 #endif 729 730 #ifdef CONFIG_NUMA_BALANCING 731 static inline int cpu_pid_to_cpupid(int cpu, int pid) 732 { 733 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); 734 } 735 736 static inline int cpupid_to_pid(int cpupid) 737 { 738 return cpupid & LAST__PID_MASK; 739 } 740 741 static inline int cpupid_to_cpu(int cpupid) 742 { 743 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; 744 } 745 746 static inline int cpupid_to_nid(int cpupid) 747 { 748 return cpu_to_node(cpupid_to_cpu(cpupid)); 749 } 750 751 static inline bool cpupid_pid_unset(int cpupid) 752 { 753 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); 754 } 755 756 static inline bool cpupid_cpu_unset(int cpupid) 757 { 758 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); 759 } 760 761 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) 762 { 763 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); 764 } 765 766 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) 767 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS 768 static inline int page_cpupid_xchg_last(struct page *page, int cpupid) 769 { 770 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); 771 } 772 773 static inline int page_cpupid_last(struct page *page) 774 { 775 return page->_last_cpupid; 776 } 777 static inline void page_cpupid_reset_last(struct page *page) 778 { 779 page->_last_cpupid = -1 & LAST_CPUPID_MASK; 780 } 781 #else 782 static inline int page_cpupid_last(struct page *page) 783 { 784 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; 785 } 786 787 extern int page_cpupid_xchg_last(struct page *page, int cpupid); 788 789 static inline void page_cpupid_reset_last(struct page *page) 790 { 791 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1; 792 793 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT); 794 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT; 795 } 796 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ 797 #else /* !CONFIG_NUMA_BALANCING */ 798 static inline int page_cpupid_xchg_last(struct page *page, int cpupid) 799 { 800 return page_to_nid(page); /* XXX */ 801 } 802 803 static inline int page_cpupid_last(struct page *page) 804 { 805 return page_to_nid(page); /* XXX */ 806 } 807 808 static inline int cpupid_to_nid(int cpupid) 809 { 810 return -1; 811 } 812 813 static inline int cpupid_to_pid(int cpupid) 814 { 815 return -1; 816 } 817 818 static inline int cpupid_to_cpu(int cpupid) 819 { 820 return -1; 821 } 822 823 static inline int cpu_pid_to_cpupid(int nid, int pid) 824 { 825 return -1; 826 } 827 828 static inline bool cpupid_pid_unset(int cpupid) 829 { 830 return 1; 831 } 832 833 static inline void page_cpupid_reset_last(struct page *page) 834 { 835 } 836 837 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) 838 { 839 return false; 840 } 841 #endif /* CONFIG_NUMA_BALANCING */ 842 843 static inline struct zone *page_zone(const struct page *page) 844 { 845 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 846 } 847 848 #ifdef SECTION_IN_PAGE_FLAGS 849 static inline void set_page_section(struct page *page, unsigned long section) 850 { 851 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 852 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 853 } 854 855 static inline unsigned long page_to_section(const struct page *page) 856 { 857 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 858 } 859 #endif 860 861 static inline void set_page_zone(struct page *page, enum zone_type zone) 862 { 863 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 864 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 865 } 866 867 static inline void set_page_node(struct page *page, unsigned long node) 868 { 869 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 870 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 871 } 872 873 static inline void set_page_links(struct page *page, enum zone_type zone, 874 unsigned long node, unsigned long pfn) 875 { 876 set_page_zone(page, zone); 877 set_page_node(page, node); 878 #ifdef SECTION_IN_PAGE_FLAGS 879 set_page_section(page, pfn_to_section_nr(pfn)); 880 #endif 881 } 882 883 /* 884 * Some inline functions in vmstat.h depend on page_zone() 885 */ 886 #include <linux/vmstat.h> 887 888 static __always_inline void *lowmem_page_address(const struct page *page) 889 { 890 return __va(PFN_PHYS(page_to_pfn(page))); 891 } 892 893 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 894 #define HASHED_PAGE_VIRTUAL 895 #endif 896 897 #if defined(WANT_PAGE_VIRTUAL) 898 static inline void *page_address(const struct page *page) 899 { 900 return page->virtual; 901 } 902 static inline void set_page_address(struct page *page, void *address) 903 { 904 page->virtual = address; 905 } 906 #define page_address_init() do { } while(0) 907 #endif 908 909 #if defined(HASHED_PAGE_VIRTUAL) 910 void *page_address(const struct page *page); 911 void set_page_address(struct page *page, void *virtual); 912 void page_address_init(void); 913 #endif 914 915 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 916 #define page_address(page) lowmem_page_address(page) 917 #define set_page_address(page, address) do { } while(0) 918 #define page_address_init() do { } while(0) 919 #endif 920 921 /* 922 * On an anonymous page mapped into a user virtual memory area, 923 * page->mapping points to its anon_vma, not to a struct address_space; 924 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 925 * 926 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 927 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 928 * and then page->mapping points, not to an anon_vma, but to a private 929 * structure which KSM associates with that merged page. See ksm.h. 930 * 931 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 932 * 933 * Please note that, confusingly, "page_mapping" refers to the inode 934 * address_space which maps the page from disk; whereas "page_mapped" 935 * refers to user virtual address space into which the page is mapped. 936 */ 937 #define PAGE_MAPPING_ANON 1 938 #define PAGE_MAPPING_KSM 2 939 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 940 941 extern struct address_space *page_mapping(struct page *page); 942 943 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 944 static inline void *page_rmapping(struct page *page) 945 { 946 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 947 } 948 949 extern struct address_space *__page_file_mapping(struct page *); 950 951 static inline 952 struct address_space *page_file_mapping(struct page *page) 953 { 954 if (unlikely(PageSwapCache(page))) 955 return __page_file_mapping(page); 956 957 return page->mapping; 958 } 959 960 static inline int PageAnon(struct page *page) 961 { 962 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 963 } 964 965 /* 966 * Return the pagecache index of the passed page. Regular pagecache pages 967 * use ->index whereas swapcache pages use ->private 968 */ 969 static inline pgoff_t page_index(struct page *page) 970 { 971 if (unlikely(PageSwapCache(page))) 972 return page_private(page); 973 return page->index; 974 } 975 976 extern pgoff_t __page_file_index(struct page *page); 977 978 /* 979 * Return the file index of the page. Regular pagecache pages use ->index 980 * whereas swapcache pages use swp_offset(->private) 981 */ 982 static inline pgoff_t page_file_index(struct page *page) 983 { 984 if (unlikely(PageSwapCache(page))) 985 return __page_file_index(page); 986 987 return page->index; 988 } 989 990 /* 991 * Return true if this page is mapped into pagetables. 992 */ 993 static inline int page_mapped(struct page *page) 994 { 995 return atomic_read(&(page)->_mapcount) >= 0; 996 } 997 998 /* 999 * Different kinds of faults, as returned by handle_mm_fault(). 1000 * Used to decide whether a process gets delivered SIGBUS or 1001 * just gets major/minor fault counters bumped up. 1002 */ 1003 1004 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 1005 1006 #define VM_FAULT_OOM 0x0001 1007 #define VM_FAULT_SIGBUS 0x0002 1008 #define VM_FAULT_MAJOR 0x0004 1009 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 1010 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ 1011 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ 1012 #define VM_FAULT_SIGSEGV 0x0040 1013 1014 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 1015 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 1016 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ 1017 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ 1018 1019 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ 1020 1021 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \ 1022 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \ 1023 VM_FAULT_FALLBACK) 1024 1025 /* Encode hstate index for a hwpoisoned large page */ 1026 #define VM_FAULT_SET_HINDEX(x) ((x) << 12) 1027 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) 1028 1029 /* 1030 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 1031 */ 1032 extern void pagefault_out_of_memory(void); 1033 1034 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 1035 1036 /* 1037 * Flags passed to show_mem() and show_free_areas() to suppress output in 1038 * various contexts. 1039 */ 1040 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ 1041 1042 extern void show_free_areas(unsigned int flags); 1043 extern bool skip_free_areas_node(unsigned int flags, int nid); 1044 1045 int shmem_zero_setup(struct vm_area_struct *); 1046 #ifdef CONFIG_SHMEM 1047 bool shmem_mapping(struct address_space *mapping); 1048 #else 1049 static inline bool shmem_mapping(struct address_space *mapping) 1050 { 1051 return false; 1052 } 1053 #endif 1054 1055 extern int can_do_mlock(void); 1056 extern int user_shm_lock(size_t, struct user_struct *); 1057 extern void user_shm_unlock(size_t, struct user_struct *); 1058 1059 /* 1060 * Parameter block passed down to zap_pte_range in exceptional cases. 1061 */ 1062 struct zap_details { 1063 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 1064 struct address_space *check_mapping; /* Check page->mapping if set */ 1065 pgoff_t first_index; /* Lowest page->index to unmap */ 1066 pgoff_t last_index; /* Highest page->index to unmap */ 1067 }; 1068 1069 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 1070 pte_t pte); 1071 1072 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 1073 unsigned long size); 1074 void zap_page_range(struct vm_area_struct *vma, unsigned long address, 1075 unsigned long size, struct zap_details *); 1076 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, 1077 unsigned long start, unsigned long end); 1078 1079 /** 1080 * mm_walk - callbacks for walk_page_range 1081 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 1082 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 1083 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 1084 * this handler is required to be able to handle 1085 * pmd_trans_huge() pmds. They may simply choose to 1086 * split_huge_page() instead of handling it explicitly. 1087 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 1088 * @pte_hole: if set, called for each hole at all levels 1089 * @hugetlb_entry: if set, called for each hugetlb entry 1090 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry 1091 * is used. 1092 * 1093 * (see walk_page_range for more details) 1094 */ 1095 struct mm_walk { 1096 int (*pgd_entry)(pgd_t *pgd, unsigned long addr, 1097 unsigned long next, struct mm_walk *walk); 1098 int (*pud_entry)(pud_t *pud, unsigned long addr, 1099 unsigned long next, struct mm_walk *walk); 1100 int (*pmd_entry)(pmd_t *pmd, unsigned long addr, 1101 unsigned long next, struct mm_walk *walk); 1102 int (*pte_entry)(pte_t *pte, unsigned long addr, 1103 unsigned long next, struct mm_walk *walk); 1104 int (*pte_hole)(unsigned long addr, unsigned long next, 1105 struct mm_walk *walk); 1106 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, 1107 unsigned long addr, unsigned long next, 1108 struct mm_walk *walk); 1109 struct mm_struct *mm; 1110 void *private; 1111 }; 1112 1113 int walk_page_range(unsigned long addr, unsigned long end, 1114 struct mm_walk *walk); 1115 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 1116 unsigned long end, unsigned long floor, unsigned long ceiling); 1117 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 1118 struct vm_area_struct *vma); 1119 void unmap_mapping_range(struct address_space *mapping, 1120 loff_t const holebegin, loff_t const holelen, int even_cows); 1121 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 1122 unsigned long *pfn); 1123 int follow_phys(struct vm_area_struct *vma, unsigned long address, 1124 unsigned int flags, unsigned long *prot, resource_size_t *phys); 1125 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 1126 void *buf, int len, int write); 1127 1128 static inline void unmap_shared_mapping_range(struct address_space *mapping, 1129 loff_t const holebegin, loff_t const holelen) 1130 { 1131 unmap_mapping_range(mapping, holebegin, holelen, 0); 1132 } 1133 1134 extern void truncate_pagecache(struct inode *inode, loff_t new); 1135 extern void truncate_setsize(struct inode *inode, loff_t newsize); 1136 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); 1137 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); 1138 int truncate_inode_page(struct address_space *mapping, struct page *page); 1139 int generic_error_remove_page(struct address_space *mapping, struct page *page); 1140 int invalidate_inode_page(struct page *page); 1141 1142 #ifdef CONFIG_MMU 1143 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 1144 unsigned long address, unsigned int flags); 1145 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, 1146 unsigned long address, unsigned int fault_flags); 1147 #else 1148 static inline int handle_mm_fault(struct mm_struct *mm, 1149 struct vm_area_struct *vma, unsigned long address, 1150 unsigned int flags) 1151 { 1152 /* should never happen if there's no MMU */ 1153 BUG(); 1154 return VM_FAULT_SIGBUS; 1155 } 1156 static inline int fixup_user_fault(struct task_struct *tsk, 1157 struct mm_struct *mm, unsigned long address, 1158 unsigned int fault_flags) 1159 { 1160 /* should never happen if there's no MMU */ 1161 BUG(); 1162 return -EFAULT; 1163 } 1164 #endif 1165 1166 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 1167 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, 1168 void *buf, int len, int write); 1169 1170 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 1171 unsigned long start, unsigned long nr_pages, 1172 unsigned int foll_flags, struct page **pages, 1173 struct vm_area_struct **vmas, int *nonblocking); 1174 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 1175 unsigned long start, unsigned long nr_pages, 1176 int write, int force, struct page **pages, 1177 struct vm_area_struct **vmas); 1178 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 1179 struct page **pages); 1180 struct kvec; 1181 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, 1182 struct page **pages); 1183 int get_kernel_page(unsigned long start, int write, struct page **pages); 1184 struct page *get_dump_page(unsigned long addr); 1185 1186 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 1187 extern void do_invalidatepage(struct page *page, unsigned int offset, 1188 unsigned int length); 1189 1190 int __set_page_dirty_nobuffers(struct page *page); 1191 int __set_page_dirty_no_writeback(struct page *page); 1192 int redirty_page_for_writepage(struct writeback_control *wbc, 1193 struct page *page); 1194 void account_page_dirtied(struct page *page, struct address_space *mapping); 1195 void account_page_writeback(struct page *page); 1196 int set_page_dirty(struct page *page); 1197 int set_page_dirty_lock(struct page *page); 1198 int clear_page_dirty_for_io(struct page *page); 1199 1200 /* Is the vma a continuation of the stack vma above it? */ 1201 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) 1202 { 1203 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); 1204 } 1205 1206 static inline int stack_guard_page_start(struct vm_area_struct *vma, 1207 unsigned long addr) 1208 { 1209 return (vma->vm_flags & VM_GROWSDOWN) && 1210 (vma->vm_start == addr) && 1211 !vma_growsdown(vma->vm_prev, addr); 1212 } 1213 1214 /* Is the vma a continuation of the stack vma below it? */ 1215 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) 1216 { 1217 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); 1218 } 1219 1220 static inline int stack_guard_page_end(struct vm_area_struct *vma, 1221 unsigned long addr) 1222 { 1223 return (vma->vm_flags & VM_GROWSUP) && 1224 (vma->vm_end == addr) && 1225 !vma_growsup(vma->vm_next, addr); 1226 } 1227 1228 extern pid_t 1229 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group); 1230 1231 extern unsigned long move_page_tables(struct vm_area_struct *vma, 1232 unsigned long old_addr, struct vm_area_struct *new_vma, 1233 unsigned long new_addr, unsigned long len, 1234 bool need_rmap_locks); 1235 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, 1236 unsigned long end, pgprot_t newprot, 1237 int dirty_accountable, int prot_numa); 1238 extern int mprotect_fixup(struct vm_area_struct *vma, 1239 struct vm_area_struct **pprev, unsigned long start, 1240 unsigned long end, unsigned long newflags); 1241 1242 /* 1243 * doesn't attempt to fault and will return short. 1244 */ 1245 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 1246 struct page **pages); 1247 /* 1248 * per-process(per-mm_struct) statistics. 1249 */ 1250 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 1251 { 1252 long val = atomic_long_read(&mm->rss_stat.count[member]); 1253 1254 #ifdef SPLIT_RSS_COUNTING 1255 /* 1256 * counter is updated in asynchronous manner and may go to minus. 1257 * But it's never be expected number for users. 1258 */ 1259 if (val < 0) 1260 val = 0; 1261 #endif 1262 return (unsigned long)val; 1263 } 1264 1265 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 1266 { 1267 atomic_long_add(value, &mm->rss_stat.count[member]); 1268 } 1269 1270 static inline void inc_mm_counter(struct mm_struct *mm, int member) 1271 { 1272 atomic_long_inc(&mm->rss_stat.count[member]); 1273 } 1274 1275 static inline void dec_mm_counter(struct mm_struct *mm, int member) 1276 { 1277 atomic_long_dec(&mm->rss_stat.count[member]); 1278 } 1279 1280 static inline unsigned long get_mm_rss(struct mm_struct *mm) 1281 { 1282 return get_mm_counter(mm, MM_FILEPAGES) + 1283 get_mm_counter(mm, MM_ANONPAGES); 1284 } 1285 1286 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 1287 { 1288 return max(mm->hiwater_rss, get_mm_rss(mm)); 1289 } 1290 1291 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 1292 { 1293 return max(mm->hiwater_vm, mm->total_vm); 1294 } 1295 1296 static inline void update_hiwater_rss(struct mm_struct *mm) 1297 { 1298 unsigned long _rss = get_mm_rss(mm); 1299 1300 if ((mm)->hiwater_rss < _rss) 1301 (mm)->hiwater_rss = _rss; 1302 } 1303 1304 static inline void update_hiwater_vm(struct mm_struct *mm) 1305 { 1306 if (mm->hiwater_vm < mm->total_vm) 1307 mm->hiwater_vm = mm->total_vm; 1308 } 1309 1310 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 1311 struct mm_struct *mm) 1312 { 1313 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 1314 1315 if (*maxrss < hiwater_rss) 1316 *maxrss = hiwater_rss; 1317 } 1318 1319 #if defined(SPLIT_RSS_COUNTING) 1320 void sync_mm_rss(struct mm_struct *mm); 1321 #else 1322 static inline void sync_mm_rss(struct mm_struct *mm) 1323 { 1324 } 1325 #endif 1326 1327 int vma_wants_writenotify(struct vm_area_struct *vma); 1328 1329 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, 1330 spinlock_t **ptl); 1331 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, 1332 spinlock_t **ptl) 1333 { 1334 pte_t *ptep; 1335 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); 1336 return ptep; 1337 } 1338 1339 #ifdef __PAGETABLE_PUD_FOLDED 1340 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1341 unsigned long address) 1342 { 1343 return 0; 1344 } 1345 #else 1346 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1347 #endif 1348 1349 #ifdef __PAGETABLE_PMD_FOLDED 1350 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1351 unsigned long address) 1352 { 1353 return 0; 1354 } 1355 #else 1356 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1357 #endif 1358 1359 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, 1360 pmd_t *pmd, unsigned long address); 1361 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1362 1363 /* 1364 * The following ifdef needed to get the 4level-fixup.h header to work. 1365 * Remove it when 4level-fixup.h has been removed. 1366 */ 1367 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1368 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1369 { 1370 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1371 NULL: pud_offset(pgd, address); 1372 } 1373 1374 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1375 { 1376 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1377 NULL: pmd_offset(pud, address); 1378 } 1379 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1380 1381 #if USE_SPLIT_PTE_PTLOCKS 1382 #if ALLOC_SPLIT_PTLOCKS 1383 void __init ptlock_cache_init(void); 1384 extern bool ptlock_alloc(struct page *page); 1385 extern void ptlock_free(struct page *page); 1386 1387 static inline spinlock_t *ptlock_ptr(struct page *page) 1388 { 1389 return page->ptl; 1390 } 1391 #else /* ALLOC_SPLIT_PTLOCKS */ 1392 static inline void ptlock_cache_init(void) 1393 { 1394 } 1395 1396 static inline bool ptlock_alloc(struct page *page) 1397 { 1398 return true; 1399 } 1400 1401 static inline void ptlock_free(struct page *page) 1402 { 1403 } 1404 1405 static inline spinlock_t *ptlock_ptr(struct page *page) 1406 { 1407 return &page->ptl; 1408 } 1409 #endif /* ALLOC_SPLIT_PTLOCKS */ 1410 1411 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) 1412 { 1413 return ptlock_ptr(pmd_page(*pmd)); 1414 } 1415 1416 static inline bool ptlock_init(struct page *page) 1417 { 1418 /* 1419 * prep_new_page() initialize page->private (and therefore page->ptl) 1420 * with 0. Make sure nobody took it in use in between. 1421 * 1422 * It can happen if arch try to use slab for page table allocation: 1423 * slab code uses page->slab_cache and page->first_page (for tail 1424 * pages), which share storage with page->ptl. 1425 */ 1426 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); 1427 if (!ptlock_alloc(page)) 1428 return false; 1429 spin_lock_init(ptlock_ptr(page)); 1430 return true; 1431 } 1432 1433 /* Reset page->mapping so free_pages_check won't complain. */ 1434 static inline void pte_lock_deinit(struct page *page) 1435 { 1436 page->mapping = NULL; 1437 ptlock_free(page); 1438 } 1439 1440 #else /* !USE_SPLIT_PTE_PTLOCKS */ 1441 /* 1442 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1443 */ 1444 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) 1445 { 1446 return &mm->page_table_lock; 1447 } 1448 static inline void ptlock_cache_init(void) {} 1449 static inline bool ptlock_init(struct page *page) { return true; } 1450 static inline void pte_lock_deinit(struct page *page) {} 1451 #endif /* USE_SPLIT_PTE_PTLOCKS */ 1452 1453 static inline void pgtable_init(void) 1454 { 1455 ptlock_cache_init(); 1456 pgtable_cache_init(); 1457 } 1458 1459 static inline bool pgtable_page_ctor(struct page *page) 1460 { 1461 inc_zone_page_state(page, NR_PAGETABLE); 1462 return ptlock_init(page); 1463 } 1464 1465 static inline void pgtable_page_dtor(struct page *page) 1466 { 1467 pte_lock_deinit(page); 1468 dec_zone_page_state(page, NR_PAGETABLE); 1469 } 1470 1471 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1472 ({ \ 1473 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1474 pte_t *__pte = pte_offset_map(pmd, address); \ 1475 *(ptlp) = __ptl; \ 1476 spin_lock(__ptl); \ 1477 __pte; \ 1478 }) 1479 1480 #define pte_unmap_unlock(pte, ptl) do { \ 1481 spin_unlock(ptl); \ 1482 pte_unmap(pte); \ 1483 } while (0) 1484 1485 #define pte_alloc_map(mm, vma, pmd, address) \ 1486 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \ 1487 pmd, address))? \ 1488 NULL: pte_offset_map(pmd, address)) 1489 1490 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1491 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \ 1492 pmd, address))? \ 1493 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1494 1495 #define pte_alloc_kernel(pmd, address) \ 1496 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1497 NULL: pte_offset_kernel(pmd, address)) 1498 1499 #if USE_SPLIT_PMD_PTLOCKS 1500 1501 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) 1502 { 1503 return ptlock_ptr(virt_to_page(pmd)); 1504 } 1505 1506 static inline bool pgtable_pmd_page_ctor(struct page *page) 1507 { 1508 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 1509 page->pmd_huge_pte = NULL; 1510 #endif 1511 return ptlock_init(page); 1512 } 1513 1514 static inline void pgtable_pmd_page_dtor(struct page *page) 1515 { 1516 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 1517 VM_BUG_ON_PAGE(page->pmd_huge_pte, page); 1518 #endif 1519 ptlock_free(page); 1520 } 1521 1522 #define pmd_huge_pte(mm, pmd) (virt_to_page(pmd)->pmd_huge_pte) 1523 1524 #else 1525 1526 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) 1527 { 1528 return &mm->page_table_lock; 1529 } 1530 1531 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } 1532 static inline void pgtable_pmd_page_dtor(struct page *page) {} 1533 1534 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) 1535 1536 #endif 1537 1538 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) 1539 { 1540 spinlock_t *ptl = pmd_lockptr(mm, pmd); 1541 spin_lock(ptl); 1542 return ptl; 1543 } 1544 1545 extern void free_area_init(unsigned long * zones_size); 1546 extern void free_area_init_node(int nid, unsigned long * zones_size, 1547 unsigned long zone_start_pfn, unsigned long *zholes_size); 1548 extern void free_initmem(void); 1549 1550 /* 1551 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) 1552 * into the buddy system. The freed pages will be poisoned with pattern 1553 * "poison" if it's within range [0, UCHAR_MAX]. 1554 * Return pages freed into the buddy system. 1555 */ 1556 extern unsigned long free_reserved_area(void *start, void *end, 1557 int poison, char *s); 1558 1559 #ifdef CONFIG_HIGHMEM 1560 /* 1561 * Free a highmem page into the buddy system, adjusting totalhigh_pages 1562 * and totalram_pages. 1563 */ 1564 extern void free_highmem_page(struct page *page); 1565 #endif 1566 1567 extern void adjust_managed_page_count(struct page *page, long count); 1568 extern void mem_init_print_info(const char *str); 1569 1570 /* Free the reserved page into the buddy system, so it gets managed. */ 1571 static inline void __free_reserved_page(struct page *page) 1572 { 1573 ClearPageReserved(page); 1574 init_page_count(page); 1575 __free_page(page); 1576 } 1577 1578 static inline void free_reserved_page(struct page *page) 1579 { 1580 __free_reserved_page(page); 1581 adjust_managed_page_count(page, 1); 1582 } 1583 1584 static inline void mark_page_reserved(struct page *page) 1585 { 1586 SetPageReserved(page); 1587 adjust_managed_page_count(page, -1); 1588 } 1589 1590 /* 1591 * Default method to free all the __init memory into the buddy system. 1592 * The freed pages will be poisoned with pattern "poison" if it's within 1593 * range [0, UCHAR_MAX]. 1594 * Return pages freed into the buddy system. 1595 */ 1596 static inline unsigned long free_initmem_default(int poison) 1597 { 1598 extern char __init_begin[], __init_end[]; 1599 1600 return free_reserved_area(&__init_begin, &__init_end, 1601 poison, "unused kernel"); 1602 } 1603 1604 static inline unsigned long get_num_physpages(void) 1605 { 1606 int nid; 1607 unsigned long phys_pages = 0; 1608 1609 for_each_online_node(nid) 1610 phys_pages += node_present_pages(nid); 1611 1612 return phys_pages; 1613 } 1614 1615 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 1616 /* 1617 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its 1618 * zones, allocate the backing mem_map and account for memory holes in a more 1619 * architecture independent manner. This is a substitute for creating the 1620 * zone_sizes[] and zholes_size[] arrays and passing them to 1621 * free_area_init_node() 1622 * 1623 * An architecture is expected to register range of page frames backed by 1624 * physical memory with memblock_add[_node]() before calling 1625 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1626 * usage, an architecture is expected to do something like 1627 * 1628 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1629 * max_highmem_pfn}; 1630 * for_each_valid_physical_page_range() 1631 * memblock_add_node(base, size, nid) 1632 * free_area_init_nodes(max_zone_pfns); 1633 * 1634 * free_bootmem_with_active_regions() calls free_bootmem_node() for each 1635 * registered physical page range. Similarly 1636 * sparse_memory_present_with_active_regions() calls memory_present() for 1637 * each range when SPARSEMEM is enabled. 1638 * 1639 * See mm/page_alloc.c for more information on each function exposed by 1640 * CONFIG_HAVE_MEMBLOCK_NODE_MAP. 1641 */ 1642 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1643 unsigned long node_map_pfn_alignment(void); 1644 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1645 unsigned long end_pfn); 1646 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1647 unsigned long end_pfn); 1648 extern void get_pfn_range_for_nid(unsigned int nid, 1649 unsigned long *start_pfn, unsigned long *end_pfn); 1650 extern unsigned long find_min_pfn_with_active_regions(void); 1651 extern void free_bootmem_with_active_regions(int nid, 1652 unsigned long max_low_pfn); 1653 extern void sparse_memory_present_with_active_regions(int nid); 1654 1655 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ 1656 1657 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ 1658 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1659 static inline int __early_pfn_to_nid(unsigned long pfn) 1660 { 1661 return 0; 1662 } 1663 #else 1664 /* please see mm/page_alloc.c */ 1665 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1666 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1667 /* there is a per-arch backend function. */ 1668 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1669 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1670 #endif 1671 1672 extern void set_dma_reserve(unsigned long new_dma_reserve); 1673 extern void memmap_init_zone(unsigned long, int, unsigned long, 1674 unsigned long, enum memmap_context); 1675 extern void setup_per_zone_wmarks(void); 1676 extern int __meminit init_per_zone_wmark_min(void); 1677 extern void mem_init(void); 1678 extern void __init mmap_init(void); 1679 extern void show_mem(unsigned int flags); 1680 extern void si_meminfo(struct sysinfo * val); 1681 extern void si_meminfo_node(struct sysinfo *val, int nid); 1682 1683 extern __printf(3, 4) 1684 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...); 1685 1686 extern void setup_per_cpu_pageset(void); 1687 1688 extern void zone_pcp_update(struct zone *zone); 1689 extern void zone_pcp_reset(struct zone *zone); 1690 1691 /* page_alloc.c */ 1692 extern int min_free_kbytes; 1693 1694 /* nommu.c */ 1695 extern atomic_long_t mmap_pages_allocated; 1696 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1697 1698 /* interval_tree.c */ 1699 void vma_interval_tree_insert(struct vm_area_struct *node, 1700 struct rb_root *root); 1701 void vma_interval_tree_insert_after(struct vm_area_struct *node, 1702 struct vm_area_struct *prev, 1703 struct rb_root *root); 1704 void vma_interval_tree_remove(struct vm_area_struct *node, 1705 struct rb_root *root); 1706 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, 1707 unsigned long start, unsigned long last); 1708 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, 1709 unsigned long start, unsigned long last); 1710 1711 #define vma_interval_tree_foreach(vma, root, start, last) \ 1712 for (vma = vma_interval_tree_iter_first(root, start, last); \ 1713 vma; vma = vma_interval_tree_iter_next(vma, start, last)) 1714 1715 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1716 struct list_head *list) 1717 { 1718 list_add_tail(&vma->shared.nonlinear, list); 1719 } 1720 1721 void anon_vma_interval_tree_insert(struct anon_vma_chain *node, 1722 struct rb_root *root); 1723 void anon_vma_interval_tree_remove(struct anon_vma_chain *node, 1724 struct rb_root *root); 1725 struct anon_vma_chain *anon_vma_interval_tree_iter_first( 1726 struct rb_root *root, unsigned long start, unsigned long last); 1727 struct anon_vma_chain *anon_vma_interval_tree_iter_next( 1728 struct anon_vma_chain *node, unsigned long start, unsigned long last); 1729 #ifdef CONFIG_DEBUG_VM_RB 1730 void anon_vma_interval_tree_verify(struct anon_vma_chain *node); 1731 #endif 1732 1733 #define anon_vma_interval_tree_foreach(avc, root, start, last) \ 1734 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ 1735 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) 1736 1737 /* mmap.c */ 1738 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1739 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1740 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1741 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1742 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1743 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1744 struct mempolicy *); 1745 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1746 extern int split_vma(struct mm_struct *, 1747 struct vm_area_struct *, unsigned long addr, int new_below); 1748 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1749 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1750 struct rb_node **, struct rb_node *); 1751 extern void unlink_file_vma(struct vm_area_struct *); 1752 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1753 unsigned long addr, unsigned long len, pgoff_t pgoff, 1754 bool *need_rmap_locks); 1755 extern void exit_mmap(struct mm_struct *); 1756 1757 extern int mm_take_all_locks(struct mm_struct *mm); 1758 extern void mm_drop_all_locks(struct mm_struct *mm); 1759 1760 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); 1761 extern struct file *get_mm_exe_file(struct mm_struct *mm); 1762 1763 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1764 extern int install_special_mapping(struct mm_struct *mm, 1765 unsigned long addr, unsigned long len, 1766 unsigned long flags, struct page **pages); 1767 1768 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1769 1770 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1771 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); 1772 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1773 unsigned long len, unsigned long prot, unsigned long flags, 1774 unsigned long pgoff, unsigned long *populate); 1775 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1776 1777 #ifdef CONFIG_MMU 1778 extern int __mm_populate(unsigned long addr, unsigned long len, 1779 int ignore_errors); 1780 static inline void mm_populate(unsigned long addr, unsigned long len) 1781 { 1782 /* Ignore errors */ 1783 (void) __mm_populate(addr, len, 1); 1784 } 1785 #else 1786 static inline void mm_populate(unsigned long addr, unsigned long len) {} 1787 #endif 1788 1789 /* These take the mm semaphore themselves */ 1790 extern unsigned long vm_brk(unsigned long, unsigned long); 1791 extern int vm_munmap(unsigned long, size_t); 1792 extern unsigned long vm_mmap(struct file *, unsigned long, 1793 unsigned long, unsigned long, 1794 unsigned long, unsigned long); 1795 1796 struct vm_unmapped_area_info { 1797 #define VM_UNMAPPED_AREA_TOPDOWN 1 1798 unsigned long flags; 1799 unsigned long length; 1800 unsigned long low_limit; 1801 unsigned long high_limit; 1802 unsigned long align_mask; 1803 unsigned long align_offset; 1804 }; 1805 1806 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); 1807 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); 1808 1809 /* 1810 * Search for an unmapped address range. 1811 * 1812 * We are looking for a range that: 1813 * - does not intersect with any VMA; 1814 * - is contained within the [low_limit, high_limit) interval; 1815 * - is at least the desired size. 1816 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) 1817 */ 1818 static inline unsigned long 1819 vm_unmapped_area(struct vm_unmapped_area_info *info) 1820 { 1821 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN)) 1822 return unmapped_area(info); 1823 else 1824 return unmapped_area_topdown(info); 1825 } 1826 1827 /* truncate.c */ 1828 extern void truncate_inode_pages(struct address_space *, loff_t); 1829 extern void truncate_inode_pages_range(struct address_space *, 1830 loff_t lstart, loff_t lend); 1831 1832 /* generic vm_area_ops exported for stackable file systems */ 1833 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1834 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); 1835 1836 /* mm/page-writeback.c */ 1837 int write_one_page(struct page *page, int wait); 1838 void task_dirty_inc(struct task_struct *tsk); 1839 1840 /* readahead.c */ 1841 #define VM_MAX_READAHEAD 128 /* kbytes */ 1842 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1843 1844 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1845 pgoff_t offset, unsigned long nr_to_read); 1846 1847 void page_cache_sync_readahead(struct address_space *mapping, 1848 struct file_ra_state *ra, 1849 struct file *filp, 1850 pgoff_t offset, 1851 unsigned long size); 1852 1853 void page_cache_async_readahead(struct address_space *mapping, 1854 struct file_ra_state *ra, 1855 struct file *filp, 1856 struct page *pg, 1857 pgoff_t offset, 1858 unsigned long size); 1859 1860 unsigned long max_sane_readahead(unsigned long nr); 1861 1862 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ 1863 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1864 1865 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ 1866 extern int expand_downwards(struct vm_area_struct *vma, 1867 unsigned long address); 1868 #if VM_GROWSUP 1869 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1870 #else 1871 #define expand_upwards(vma, address) (0) 1872 #endif 1873 1874 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1875 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1876 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1877 struct vm_area_struct **pprev); 1878 1879 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1880 NULL if none. Assume start_addr < end_addr. */ 1881 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1882 { 1883 struct vm_area_struct * vma = find_vma(mm,start_addr); 1884 1885 if (vma && end_addr <= vma->vm_start) 1886 vma = NULL; 1887 return vma; 1888 } 1889 1890 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1891 { 1892 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1893 } 1894 1895 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ 1896 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, 1897 unsigned long vm_start, unsigned long vm_end) 1898 { 1899 struct vm_area_struct *vma = find_vma(mm, vm_start); 1900 1901 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) 1902 vma = NULL; 1903 1904 return vma; 1905 } 1906 1907 #ifdef CONFIG_MMU 1908 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1909 #else 1910 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) 1911 { 1912 return __pgprot(0); 1913 } 1914 #endif 1915 1916 #ifdef CONFIG_NUMA_BALANCING 1917 unsigned long change_prot_numa(struct vm_area_struct *vma, 1918 unsigned long start, unsigned long end); 1919 #endif 1920 1921 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1922 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1923 unsigned long pfn, unsigned long size, pgprot_t); 1924 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1925 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1926 unsigned long pfn); 1927 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1928 unsigned long pfn); 1929 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); 1930 1931 1932 struct page *follow_page_mask(struct vm_area_struct *vma, 1933 unsigned long address, unsigned int foll_flags, 1934 unsigned int *page_mask); 1935 1936 static inline struct page *follow_page(struct vm_area_struct *vma, 1937 unsigned long address, unsigned int foll_flags) 1938 { 1939 unsigned int unused_page_mask; 1940 return follow_page_mask(vma, address, foll_flags, &unused_page_mask); 1941 } 1942 1943 #define FOLL_WRITE 0x01 /* check pte is writable */ 1944 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1945 #define FOLL_GET 0x04 /* do get_page on page */ 1946 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1947 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1948 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO 1949 * and return without waiting upon it */ 1950 #define FOLL_MLOCK 0x40 /* mark page as mlocked */ 1951 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ 1952 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ 1953 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ 1954 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ 1955 1956 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1957 void *data); 1958 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1959 unsigned long size, pte_fn_t fn, void *data); 1960 1961 #ifdef CONFIG_PROC_FS 1962 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1963 #else 1964 static inline void vm_stat_account(struct mm_struct *mm, 1965 unsigned long flags, struct file *file, long pages) 1966 { 1967 mm->total_vm += pages; 1968 } 1969 #endif /* CONFIG_PROC_FS */ 1970 1971 #ifdef CONFIG_DEBUG_PAGEALLOC 1972 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1973 #ifdef CONFIG_HIBERNATION 1974 extern bool kernel_page_present(struct page *page); 1975 #endif /* CONFIG_HIBERNATION */ 1976 #else 1977 static inline void 1978 kernel_map_pages(struct page *page, int numpages, int enable) {} 1979 #ifdef CONFIG_HIBERNATION 1980 static inline bool kernel_page_present(struct page *page) { return true; } 1981 #endif /* CONFIG_HIBERNATION */ 1982 #endif 1983 1984 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); 1985 #ifdef __HAVE_ARCH_GATE_AREA 1986 int in_gate_area_no_mm(unsigned long addr); 1987 int in_gate_area(struct mm_struct *mm, unsigned long addr); 1988 #else 1989 int in_gate_area_no_mm(unsigned long addr); 1990 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);}) 1991 #endif /* __HAVE_ARCH_GATE_AREA */ 1992 1993 #ifdef CONFIG_SYSCTL 1994 extern int sysctl_drop_caches; 1995 int drop_caches_sysctl_handler(struct ctl_table *, int, 1996 void __user *, size_t *, loff_t *); 1997 #endif 1998 1999 unsigned long shrink_slab(struct shrink_control *shrink, 2000 unsigned long nr_pages_scanned, 2001 unsigned long lru_pages); 2002 2003 #ifndef CONFIG_MMU 2004 #define randomize_va_space 0 2005 #else 2006 extern int randomize_va_space; 2007 #endif 2008 2009 const char * arch_vma_name(struct vm_area_struct *vma); 2010 void print_vma_addr(char *prefix, unsigned long rip); 2011 2012 void sparse_mem_maps_populate_node(struct page **map_map, 2013 unsigned long pnum_begin, 2014 unsigned long pnum_end, 2015 unsigned long map_count, 2016 int nodeid); 2017 2018 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 2019 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 2020 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 2021 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 2022 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 2023 void *vmemmap_alloc_block(unsigned long size, int node); 2024 void *vmemmap_alloc_block_buf(unsigned long size, int node); 2025 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 2026 int vmemmap_populate_basepages(unsigned long start, unsigned long end, 2027 int node); 2028 int vmemmap_populate(unsigned long start, unsigned long end, int node); 2029 void vmemmap_populate_print_last(void); 2030 #ifdef CONFIG_MEMORY_HOTPLUG 2031 void vmemmap_free(unsigned long start, unsigned long end); 2032 #endif 2033 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, 2034 unsigned long size); 2035 2036 enum mf_flags { 2037 MF_COUNT_INCREASED = 1 << 0, 2038 MF_ACTION_REQUIRED = 1 << 1, 2039 MF_MUST_KILL = 1 << 2, 2040 MF_SOFT_OFFLINE = 1 << 3, 2041 }; 2042 extern int memory_failure(unsigned long pfn, int trapno, int flags); 2043 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); 2044 extern int unpoison_memory(unsigned long pfn); 2045 extern int sysctl_memory_failure_early_kill; 2046 extern int sysctl_memory_failure_recovery; 2047 extern void shake_page(struct page *p, int access); 2048 extern atomic_long_t num_poisoned_pages; 2049 extern int soft_offline_page(struct page *page, int flags); 2050 2051 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) 2052 extern void clear_huge_page(struct page *page, 2053 unsigned long addr, 2054 unsigned int pages_per_huge_page); 2055 extern void copy_user_huge_page(struct page *dst, struct page *src, 2056 unsigned long addr, struct vm_area_struct *vma, 2057 unsigned int pages_per_huge_page); 2058 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ 2059 2060 #ifdef CONFIG_DEBUG_PAGEALLOC 2061 extern unsigned int _debug_guardpage_minorder; 2062 2063 static inline unsigned int debug_guardpage_minorder(void) 2064 { 2065 return _debug_guardpage_minorder; 2066 } 2067 2068 static inline bool page_is_guard(struct page *page) 2069 { 2070 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags); 2071 } 2072 #else 2073 static inline unsigned int debug_guardpage_minorder(void) { return 0; } 2074 static inline bool page_is_guard(struct page *page) { return false; } 2075 #endif /* CONFIG_DEBUG_PAGEALLOC */ 2076 2077 #if MAX_NUMNODES > 1 2078 void __init setup_nr_node_ids(void); 2079 #else 2080 static inline void setup_nr_node_ids(void) {} 2081 #endif 2082 2083 #endif /* __KERNEL__ */ 2084 #endif /* _LINUX_MM_H */ 2085
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.