1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_MM_H 3 #define _LINUX_MM_H 4 5 #include <linux/errno.h> 6 7 #ifdef __KERNEL__ 8 9 #include <linux/mmdebug.h> 10 #include <linux/gfp.h> 11 #include <linux/bug.h> 12 #include <linux/list.h> 13 #include <linux/mmzone.h> 14 #include <linux/rbtree.h> 15 #include <linux/atomic.h> 16 #include <linux/debug_locks.h> 17 #include <linux/mm_types.h> 18 #include <linux/range.h> 19 #include <linux/pfn.h> 20 #include <linux/percpu-refcount.h> 21 #include <linux/bit_spinlock.h> 22 #include <linux/shrinker.h> 23 #include <linux/resource.h> 24 #include <linux/page_ext.h> 25 #include <linux/err.h> 26 #include <linux/page_ref.h> 27 #include <linux/memremap.h> 28 29 struct mempolicy; 30 struct anon_vma; 31 struct anon_vma_chain; 32 struct file_ra_state; 33 struct user_struct; 34 struct writeback_control; 35 struct bdi_writeback; 36 37 void init_mm_internals(void); 38 39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ 40 extern unsigned long max_mapnr; 41 42 static inline void set_max_mapnr(unsigned long limit) 43 { 44 max_mapnr = limit; 45 } 46 #else 47 static inline void set_max_mapnr(unsigned long limit) { } 48 #endif 49 50 extern unsigned long totalram_pages; 51 extern void * high_memory; 52 extern int page_cluster; 53 54 #ifdef CONFIG_SYSCTL 55 extern int sysctl_legacy_va_layout; 56 #else 57 #define sysctl_legacy_va_layout 0 58 #endif 59 60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS 61 extern const int mmap_rnd_bits_min; 62 extern const int mmap_rnd_bits_max; 63 extern int mmap_rnd_bits __read_mostly; 64 #endif 65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS 66 extern const int mmap_rnd_compat_bits_min; 67 extern const int mmap_rnd_compat_bits_max; 68 extern int mmap_rnd_compat_bits __read_mostly; 69 #endif 70 71 #include <asm/page.h> 72 #include <asm/pgtable.h> 73 #include <asm/processor.h> 74 75 #ifndef __pa_symbol 76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) 77 #endif 78 79 #ifndef page_to_virt 80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) 81 #endif 82 83 #ifndef lm_alias 84 #define lm_alias(x) __va(__pa_symbol(x)) 85 #endif 86 87 /* 88 * To prevent common memory management code establishing 89 * a zero page mapping on a read fault. 90 * This macro should be defined within <asm/pgtable.h>. 91 * s390 does this to prevent multiplexing of hardware bits 92 * related to the physical page in case of virtualization. 93 */ 94 #ifndef mm_forbids_zeropage 95 #define mm_forbids_zeropage(X) (0) 96 #endif 97 98 /* 99 * On some architectures it is expensive to call memset() for small sizes. 100 * Those architectures should provide their own implementation of "struct page" 101 * zeroing by defining this macro in <asm/pgtable.h>. 102 */ 103 #ifndef mm_zero_struct_page 104 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page))) 105 #endif 106 107 /* 108 * Default maximum number of active map areas, this limits the number of vmas 109 * per mm struct. Users can overwrite this number by sysctl but there is a 110 * problem. 111 * 112 * When a program's coredump is generated as ELF format, a section is created 113 * per a vma. In ELF, the number of sections is represented in unsigned short. 114 * This means the number of sections should be smaller than 65535 at coredump. 115 * Because the kernel adds some informative sections to a image of program at 116 * generating coredump, we need some margin. The number of extra sections is 117 * 1-3 now and depends on arch. We use "5" as safe margin, here. 118 * 119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is 120 * not a hard limit any more. Although some userspace tools can be surprised by 121 * that. 122 */ 123 #define MAPCOUNT_ELF_CORE_MARGIN (5) 124 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) 125 126 extern int sysctl_max_map_count; 127 128 extern unsigned long sysctl_user_reserve_kbytes; 129 extern unsigned long sysctl_admin_reserve_kbytes; 130 131 extern int sysctl_overcommit_memory; 132 extern int sysctl_overcommit_ratio; 133 extern unsigned long sysctl_overcommit_kbytes; 134 135 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, 136 size_t *, loff_t *); 137 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, 138 size_t *, loff_t *); 139 140 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 141 142 /* to align the pointer to the (next) page boundary */ 143 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 144 145 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ 146 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) 147 148 /* 149 * Linux kernel virtual memory manager primitives. 150 * The idea being to have a "virtual" mm in the same way 151 * we have a virtual fs - giving a cleaner interface to the 152 * mm details, and allowing different kinds of memory mappings 153 * (from shared memory to executable loading to arbitrary 154 * mmap() functions). 155 */ 156 157 extern struct kmem_cache *vm_area_cachep; 158 159 #ifndef CONFIG_MMU 160 extern struct rb_root nommu_region_tree; 161 extern struct rw_semaphore nommu_region_sem; 162 163 extern unsigned int kobjsize(const void *objp); 164 #endif 165 166 /* 167 * vm_flags in vm_area_struct, see mm_types.h. 168 * When changing, update also include/trace/events/mmflags.h 169 */ 170 #define VM_NONE 0x00000000 171 172 #define VM_READ 0x00000001 /* currently active flags */ 173 #define VM_WRITE 0x00000002 174 #define VM_EXEC 0x00000004 175 #define VM_SHARED 0x00000008 176 177 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 178 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 179 #define VM_MAYWRITE 0x00000020 180 #define VM_MAYEXEC 0x00000040 181 #define VM_MAYSHARE 0x00000080 182 183 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 184 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ 185 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 186 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 187 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ 188 189 #define VM_LOCKED 0x00002000 190 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 191 192 /* Used by sys_madvise() */ 193 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 194 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 195 196 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 197 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 198 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ 199 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 200 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 201 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 202 #define VM_SYNC 0x00800000 /* Synchronous page faults */ 203 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ 204 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */ 205 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ 206 207 #ifdef CONFIG_MEM_SOFT_DIRTY 208 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ 209 #else 210 # define VM_SOFTDIRTY 0 211 #endif 212 213 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 214 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ 215 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ 216 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 217 218 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS 219 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ 220 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ 221 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ 222 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ 223 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */ 224 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) 225 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) 226 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) 227 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) 228 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4) 229 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ 230 231 #if defined(CONFIG_X86) 232 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ 233 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS) 234 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 235 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ 236 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 237 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 238 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 239 #endif 240 #elif defined(CONFIG_PPC) 241 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ 242 #elif defined(CONFIG_PARISC) 243 # define VM_GROWSUP VM_ARCH_1 244 #elif defined(CONFIG_METAG) 245 # define VM_GROWSUP VM_ARCH_1 246 #elif defined(CONFIG_IA64) 247 # define VM_GROWSUP VM_ARCH_1 248 #elif !defined(CONFIG_MMU) 249 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ 250 #endif 251 252 #if defined(CONFIG_X86_INTEL_MPX) 253 /* MPX specific bounds table or bounds directory */ 254 # define VM_MPX VM_HIGH_ARCH_4 255 #else 256 # define VM_MPX VM_NONE 257 #endif 258 259 #ifndef VM_GROWSUP 260 # define VM_GROWSUP VM_NONE 261 #endif 262 263 /* Bits set in the VMA until the stack is in its final location */ 264 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) 265 266 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 267 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 268 #endif 269 270 #ifdef CONFIG_STACK_GROWSUP 271 #define VM_STACK VM_GROWSUP 272 #else 273 #define VM_STACK VM_GROWSDOWN 274 #endif 275 276 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 277 278 /* 279 * Special vmas that are non-mergable, non-mlock()able. 280 * Note: mm/huge_memory.c VM_NO_THP depends on this definition. 281 */ 282 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) 283 284 /* This mask defines which mm->def_flags a process can inherit its parent */ 285 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE 286 287 /* This mask is used to clear all the VMA flags used by mlock */ 288 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) 289 290 /* 291 * mapping from the currently active vm_flags protection bits (the 292 * low four bits) to a page protection mask.. 293 */ 294 extern pgprot_t protection_map[16]; 295 296 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 297 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */ 298 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */ 299 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */ 300 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */ 301 #define FAULT_FLAG_TRIED 0x20 /* Second try */ 302 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */ 303 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */ 304 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */ 305 306 #define FAULT_FLAG_TRACE \ 307 { FAULT_FLAG_WRITE, "WRITE" }, \ 308 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \ 309 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \ 310 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \ 311 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \ 312 { FAULT_FLAG_TRIED, "TRIED" }, \ 313 { FAULT_FLAG_USER, "USER" }, \ 314 { FAULT_FLAG_REMOTE, "REMOTE" }, \ 315 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" } 316 317 /* 318 * vm_fault is filled by the the pagefault handler and passed to the vma's 319 * ->fault function. The vma's ->fault is responsible for returning a bitmask 320 * of VM_FAULT_xxx flags that give details about how the fault was handled. 321 * 322 * MM layer fills up gfp_mask for page allocations but fault handler might 323 * alter it if its implementation requires a different allocation context. 324 * 325 * pgoff should be used in favour of virtual_address, if possible. 326 */ 327 struct vm_fault { 328 struct vm_area_struct *vma; /* Target VMA */ 329 unsigned int flags; /* FAULT_FLAG_xxx flags */ 330 gfp_t gfp_mask; /* gfp mask to be used for allocations */ 331 pgoff_t pgoff; /* Logical page offset based on vma */ 332 unsigned long address; /* Faulting virtual address */ 333 pmd_t *pmd; /* Pointer to pmd entry matching 334 * the 'address' */ 335 pud_t *pud; /* Pointer to pud entry matching 336 * the 'address' 337 */ 338 pte_t orig_pte; /* Value of PTE at the time of fault */ 339 340 struct page *cow_page; /* Page handler may use for COW fault */ 341 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */ 342 struct page *page; /* ->fault handlers should return a 343 * page here, unless VM_FAULT_NOPAGE 344 * is set (which is also implied by 345 * VM_FAULT_ERROR). 346 */ 347 /* These three entries are valid only while holding ptl lock */ 348 pte_t *pte; /* Pointer to pte entry matching 349 * the 'address'. NULL if the page 350 * table hasn't been allocated. 351 */ 352 spinlock_t *ptl; /* Page table lock. 353 * Protects pte page table if 'pte' 354 * is not NULL, otherwise pmd. 355 */ 356 pgtable_t prealloc_pte; /* Pre-allocated pte page table. 357 * vm_ops->map_pages() calls 358 * alloc_set_pte() from atomic context. 359 * do_fault_around() pre-allocates 360 * page table to avoid allocation from 361 * atomic context. 362 */ 363 }; 364 365 /* page entry size for vm->huge_fault() */ 366 enum page_entry_size { 367 PE_SIZE_PTE = 0, 368 PE_SIZE_PMD, 369 PE_SIZE_PUD, 370 }; 371 372 /* 373 * These are the virtual MM functions - opening of an area, closing and 374 * unmapping it (needed to keep files on disk up-to-date etc), pointer 375 * to the functions called when a no-page or a wp-page exception occurs. 376 */ 377 struct vm_operations_struct { 378 void (*open)(struct vm_area_struct * area); 379 void (*close)(struct vm_area_struct * area); 380 int (*split)(struct vm_area_struct * area, unsigned long addr); 381 int (*mremap)(struct vm_area_struct * area); 382 int (*fault)(struct vm_fault *vmf); 383 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size); 384 void (*map_pages)(struct vm_fault *vmf, 385 pgoff_t start_pgoff, pgoff_t end_pgoff); 386 387 /* notification that a previously read-only page is about to become 388 * writable, if an error is returned it will cause a SIGBUS */ 389 int (*page_mkwrite)(struct vm_fault *vmf); 390 391 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ 392 int (*pfn_mkwrite)(struct vm_fault *vmf); 393 394 /* called by access_process_vm when get_user_pages() fails, typically 395 * for use by special VMAs that can switch between memory and hardware 396 */ 397 int (*access)(struct vm_area_struct *vma, unsigned long addr, 398 void *buf, int len, int write); 399 400 /* Called by the /proc/PID/maps code to ask the vma whether it 401 * has a special name. Returning non-NULL will also cause this 402 * vma to be dumped unconditionally. */ 403 const char *(*name)(struct vm_area_struct *vma); 404 405 #ifdef CONFIG_NUMA 406 /* 407 * set_policy() op must add a reference to any non-NULL @new mempolicy 408 * to hold the policy upon return. Caller should pass NULL @new to 409 * remove a policy and fall back to surrounding context--i.e. do not 410 * install a MPOL_DEFAULT policy, nor the task or system default 411 * mempolicy. 412 */ 413 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 414 415 /* 416 * get_policy() op must add reference [mpol_get()] to any policy at 417 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 418 * in mm/mempolicy.c will do this automatically. 419 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 420 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 421 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 422 * must return NULL--i.e., do not "fallback" to task or system default 423 * policy. 424 */ 425 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 426 unsigned long addr); 427 #endif 428 /* 429 * Called by vm_normal_page() for special PTEs to find the 430 * page for @addr. This is useful if the default behavior 431 * (using pte_page()) would not find the correct page. 432 */ 433 struct page *(*find_special_page)(struct vm_area_struct *vma, 434 unsigned long addr); 435 }; 436 437 struct mmu_gather; 438 struct inode; 439 440 #define page_private(page) ((page)->private) 441 #define set_page_private(page, v) ((page)->private = (v)) 442 443 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE) 444 static inline int pmd_devmap(pmd_t pmd) 445 { 446 return 0; 447 } 448 static inline int pud_devmap(pud_t pud) 449 { 450 return 0; 451 } 452 static inline int pgd_devmap(pgd_t pgd) 453 { 454 return 0; 455 } 456 #endif 457 458 /* 459 * FIXME: take this include out, include page-flags.h in 460 * files which need it (119 of them) 461 */ 462 #include <linux/page-flags.h> 463 #include <linux/huge_mm.h> 464 465 /* 466 * Methods to modify the page usage count. 467 * 468 * What counts for a page usage: 469 * - cache mapping (page->mapping) 470 * - private data (page->private) 471 * - page mapped in a task's page tables, each mapping 472 * is counted separately 473 * 474 * Also, many kernel routines increase the page count before a critical 475 * routine so they can be sure the page doesn't go away from under them. 476 */ 477 478 /* 479 * Drop a ref, return true if the refcount fell to zero (the page has no users) 480 */ 481 static inline int put_page_testzero(struct page *page) 482 { 483 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); 484 return page_ref_dec_and_test(page); 485 } 486 487 /* 488 * Try to grab a ref unless the page has a refcount of zero, return false if 489 * that is the case. 490 * This can be called when MMU is off so it must not access 491 * any of the virtual mappings. 492 */ 493 static inline int get_page_unless_zero(struct page *page) 494 { 495 return page_ref_add_unless(page, 1, 0); 496 } 497 498 extern int page_is_ram(unsigned long pfn); 499 500 enum { 501 REGION_INTERSECTS, 502 REGION_DISJOINT, 503 REGION_MIXED, 504 }; 505 506 int region_intersects(resource_size_t offset, size_t size, unsigned long flags, 507 unsigned long desc); 508 509 /* Support for virtually mapped pages */ 510 struct page *vmalloc_to_page(const void *addr); 511 unsigned long vmalloc_to_pfn(const void *addr); 512 513 /* 514 * Determine if an address is within the vmalloc range 515 * 516 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 517 * is no special casing required. 518 */ 519 static inline bool is_vmalloc_addr(const void *x) 520 { 521 #ifdef CONFIG_MMU 522 unsigned long addr = (unsigned long)x; 523 524 return addr >= VMALLOC_START && addr < VMALLOC_END; 525 #else 526 return false; 527 #endif 528 } 529 #ifdef CONFIG_MMU 530 extern int is_vmalloc_or_module_addr(const void *x); 531 #else 532 static inline int is_vmalloc_or_module_addr(const void *x) 533 { 534 return 0; 535 } 536 #endif 537 538 extern void *kvmalloc_node(size_t size, gfp_t flags, int node); 539 static inline void *kvmalloc(size_t size, gfp_t flags) 540 { 541 return kvmalloc_node(size, flags, NUMA_NO_NODE); 542 } 543 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node) 544 { 545 return kvmalloc_node(size, flags | __GFP_ZERO, node); 546 } 547 static inline void *kvzalloc(size_t size, gfp_t flags) 548 { 549 return kvmalloc(size, flags | __GFP_ZERO); 550 } 551 552 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags) 553 { 554 if (size != 0 && n > SIZE_MAX / size) 555 return NULL; 556 557 return kvmalloc(n * size, flags); 558 } 559 560 extern void kvfree(const void *addr); 561 562 static inline atomic_t *compound_mapcount_ptr(struct page *page) 563 { 564 return &page[1].compound_mapcount; 565 } 566 567 static inline int compound_mapcount(struct page *page) 568 { 569 VM_BUG_ON_PAGE(!PageCompound(page), page); 570 page = compound_head(page); 571 return atomic_read(compound_mapcount_ptr(page)) + 1; 572 } 573 574 /* 575 * The atomic page->_mapcount, starts from -1: so that transitions 576 * both from it and to it can be tracked, using atomic_inc_and_test 577 * and atomic_add_negative(-1). 578 */ 579 static inline void page_mapcount_reset(struct page *page) 580 { 581 atomic_set(&(page)->_mapcount, -1); 582 } 583 584 int __page_mapcount(struct page *page); 585 586 static inline int page_mapcount(struct page *page) 587 { 588 VM_BUG_ON_PAGE(PageSlab(page), page); 589 590 if (unlikely(PageCompound(page))) 591 return __page_mapcount(page); 592 return atomic_read(&page->_mapcount) + 1; 593 } 594 595 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 596 int total_mapcount(struct page *page); 597 int page_trans_huge_mapcount(struct page *page, int *total_mapcount); 598 #else 599 static inline int total_mapcount(struct page *page) 600 { 601 return page_mapcount(page); 602 } 603 static inline int page_trans_huge_mapcount(struct page *page, 604 int *total_mapcount) 605 { 606 int mapcount = page_mapcount(page); 607 if (total_mapcount) 608 *total_mapcount = mapcount; 609 return mapcount; 610 } 611 #endif 612 613 static inline struct page *virt_to_head_page(const void *x) 614 { 615 struct page *page = virt_to_page(x); 616 617 return compound_head(page); 618 } 619 620 void __put_page(struct page *page); 621 622 void put_pages_list(struct list_head *pages); 623 624 void split_page(struct page *page, unsigned int order); 625 626 /* 627 * Compound pages have a destructor function. Provide a 628 * prototype for that function and accessor functions. 629 * These are _only_ valid on the head of a compound page. 630 */ 631 typedef void compound_page_dtor(struct page *); 632 633 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ 634 enum compound_dtor_id { 635 NULL_COMPOUND_DTOR, 636 COMPOUND_PAGE_DTOR, 637 #ifdef CONFIG_HUGETLB_PAGE 638 HUGETLB_PAGE_DTOR, 639 #endif 640 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 641 TRANSHUGE_PAGE_DTOR, 642 #endif 643 NR_COMPOUND_DTORS, 644 }; 645 extern compound_page_dtor * const compound_page_dtors[]; 646 647 static inline void set_compound_page_dtor(struct page *page, 648 enum compound_dtor_id compound_dtor) 649 { 650 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); 651 page[1].compound_dtor = compound_dtor; 652 } 653 654 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 655 { 656 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); 657 return compound_page_dtors[page[1].compound_dtor]; 658 } 659 660 static inline unsigned int compound_order(struct page *page) 661 { 662 if (!PageHead(page)) 663 return 0; 664 return page[1].compound_order; 665 } 666 667 static inline void set_compound_order(struct page *page, unsigned int order) 668 { 669 page[1].compound_order = order; 670 } 671 672 void free_compound_page(struct page *page); 673 674 #ifdef CONFIG_MMU 675 /* 676 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when 677 * servicing faults for write access. In the normal case, do always want 678 * pte_mkwrite. But get_user_pages can cause write faults for mappings 679 * that do not have writing enabled, when used by access_process_vm. 680 */ 681 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) 682 { 683 if (likely(vma->vm_flags & VM_WRITE)) 684 pte = pte_mkwrite(pte); 685 return pte; 686 } 687 688 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg, 689 struct page *page); 690 int finish_fault(struct vm_fault *vmf); 691 int finish_mkwrite_fault(struct vm_fault *vmf); 692 #endif 693 694 /* 695 * Multiple processes may "see" the same page. E.g. for untouched 696 * mappings of /dev/null, all processes see the same page full of 697 * zeroes, and text pages of executables and shared libraries have 698 * only one copy in memory, at most, normally. 699 * 700 * For the non-reserved pages, page_count(page) denotes a reference count. 701 * page_count() == 0 means the page is free. page->lru is then used for 702 * freelist management in the buddy allocator. 703 * page_count() > 0 means the page has been allocated. 704 * 705 * Pages are allocated by the slab allocator in order to provide memory 706 * to kmalloc and kmem_cache_alloc. In this case, the management of the 707 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 708 * unless a particular usage is carefully commented. (the responsibility of 709 * freeing the kmalloc memory is the caller's, of course). 710 * 711 * A page may be used by anyone else who does a __get_free_page(). 712 * In this case, page_count still tracks the references, and should only 713 * be used through the normal accessor functions. The top bits of page->flags 714 * and page->virtual store page management information, but all other fields 715 * are unused and could be used privately, carefully. The management of this 716 * page is the responsibility of the one who allocated it, and those who have 717 * subsequently been given references to it. 718 * 719 * The other pages (we may call them "pagecache pages") are completely 720 * managed by the Linux memory manager: I/O, buffers, swapping etc. 721 * The following discussion applies only to them. 722 * 723 * A pagecache page contains an opaque `private' member, which belongs to the 724 * page's address_space. Usually, this is the address of a circular list of 725 * the page's disk buffers. PG_private must be set to tell the VM to call 726 * into the filesystem to release these pages. 727 * 728 * A page may belong to an inode's memory mapping. In this case, page->mapping 729 * is the pointer to the inode, and page->index is the file offset of the page, 730 * in units of PAGE_SIZE. 731 * 732 * If pagecache pages are not associated with an inode, they are said to be 733 * anonymous pages. These may become associated with the swapcache, and in that 734 * case PG_swapcache is set, and page->private is an offset into the swapcache. 735 * 736 * In either case (swapcache or inode backed), the pagecache itself holds one 737 * reference to the page. Setting PG_private should also increment the 738 * refcount. The each user mapping also has a reference to the page. 739 * 740 * The pagecache pages are stored in a per-mapping radix tree, which is 741 * rooted at mapping->page_tree, and indexed by offset. 742 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 743 * lists, we instead now tag pages as dirty/writeback in the radix tree. 744 * 745 * All pagecache pages may be subject to I/O: 746 * - inode pages may need to be read from disk, 747 * - inode pages which have been modified and are MAP_SHARED may need 748 * to be written back to the inode on disk, 749 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 750 * modified may need to be swapped out to swap space and (later) to be read 751 * back into memory. 752 */ 753 754 /* 755 * The zone field is never updated after free_area_init_core() 756 * sets it, so none of the operations on it need to be atomic. 757 */ 758 759 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ 760 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 761 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 762 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 763 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) 764 765 /* 766 * Define the bit shifts to access each section. For non-existent 767 * sections we define the shift as 0; that plus a 0 mask ensures 768 * the compiler will optimise away reference to them. 769 */ 770 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 771 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 772 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 773 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) 774 775 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ 776 #ifdef NODE_NOT_IN_PAGE_FLAGS 777 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 778 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 779 SECTIONS_PGOFF : ZONES_PGOFF) 780 #else 781 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 782 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 783 NODES_PGOFF : ZONES_PGOFF) 784 #endif 785 786 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 787 788 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 789 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 790 #endif 791 792 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 793 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 794 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 795 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) 796 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 797 798 static inline enum zone_type page_zonenum(const struct page *page) 799 { 800 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 801 } 802 803 #ifdef CONFIG_ZONE_DEVICE 804 static inline bool is_zone_device_page(const struct page *page) 805 { 806 return page_zonenum(page) == ZONE_DEVICE; 807 } 808 #else 809 static inline bool is_zone_device_page(const struct page *page) 810 { 811 return false; 812 } 813 #endif 814 815 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC) 816 void put_zone_device_private_or_public_page(struct page *page); 817 DECLARE_STATIC_KEY_FALSE(device_private_key); 818 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key) 819 static inline bool is_device_private_page(const struct page *page); 820 static inline bool is_device_public_page(const struct page *page); 821 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */ 822 static inline void put_zone_device_private_or_public_page(struct page *page) 823 { 824 } 825 #define IS_HMM_ENABLED 0 826 static inline bool is_device_private_page(const struct page *page) 827 { 828 return false; 829 } 830 static inline bool is_device_public_page(const struct page *page) 831 { 832 return false; 833 } 834 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */ 835 836 837 static inline void get_page(struct page *page) 838 { 839 page = compound_head(page); 840 /* 841 * Getting a normal page or the head of a compound page 842 * requires to already have an elevated page->_refcount. 843 */ 844 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page); 845 page_ref_inc(page); 846 } 847 848 static inline void put_page(struct page *page) 849 { 850 page = compound_head(page); 851 852 /* 853 * For private device pages we need to catch refcount transition from 854 * 2 to 1, when refcount reach one it means the private device page is 855 * free and we need to inform the device driver through callback. See 856 * include/linux/memremap.h and HMM for details. 857 */ 858 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) || 859 unlikely(is_device_public_page(page)))) { 860 put_zone_device_private_or_public_page(page); 861 return; 862 } 863 864 if (put_page_testzero(page)) 865 __put_page(page); 866 } 867 868 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 869 #define SECTION_IN_PAGE_FLAGS 870 #endif 871 872 /* 873 * The identification function is mainly used by the buddy allocator for 874 * determining if two pages could be buddies. We are not really identifying 875 * the zone since we could be using the section number id if we do not have 876 * node id available in page flags. 877 * We only guarantee that it will return the same value for two combinable 878 * pages in a zone. 879 */ 880 static inline int page_zone_id(struct page *page) 881 { 882 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 883 } 884 885 static inline int zone_to_nid(struct zone *zone) 886 { 887 #ifdef CONFIG_NUMA 888 return zone->node; 889 #else 890 return 0; 891 #endif 892 } 893 894 #ifdef NODE_NOT_IN_PAGE_FLAGS 895 extern int page_to_nid(const struct page *page); 896 #else 897 static inline int page_to_nid(const struct page *page) 898 { 899 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 900 } 901 #endif 902 903 #ifdef CONFIG_NUMA_BALANCING 904 static inline int cpu_pid_to_cpupid(int cpu, int pid) 905 { 906 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); 907 } 908 909 static inline int cpupid_to_pid(int cpupid) 910 { 911 return cpupid & LAST__PID_MASK; 912 } 913 914 static inline int cpupid_to_cpu(int cpupid) 915 { 916 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; 917 } 918 919 static inline int cpupid_to_nid(int cpupid) 920 { 921 return cpu_to_node(cpupid_to_cpu(cpupid)); 922 } 923 924 static inline bool cpupid_pid_unset(int cpupid) 925 { 926 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); 927 } 928 929 static inline bool cpupid_cpu_unset(int cpupid) 930 { 931 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); 932 } 933 934 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) 935 { 936 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); 937 } 938 939 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) 940 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS 941 static inline int page_cpupid_xchg_last(struct page *page, int cpupid) 942 { 943 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); 944 } 945 946 static inline int page_cpupid_last(struct page *page) 947 { 948 return page->_last_cpupid; 949 } 950 static inline void page_cpupid_reset_last(struct page *page) 951 { 952 page->_last_cpupid = -1 & LAST_CPUPID_MASK; 953 } 954 #else 955 static inline int page_cpupid_last(struct page *page) 956 { 957 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; 958 } 959 960 extern int page_cpupid_xchg_last(struct page *page, int cpupid); 961 962 static inline void page_cpupid_reset_last(struct page *page) 963 { 964 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; 965 } 966 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ 967 #else /* !CONFIG_NUMA_BALANCING */ 968 static inline int page_cpupid_xchg_last(struct page *page, int cpupid) 969 { 970 return page_to_nid(page); /* XXX */ 971 } 972 973 static inline int page_cpupid_last(struct page *page) 974 { 975 return page_to_nid(page); /* XXX */ 976 } 977 978 static inline int cpupid_to_nid(int cpupid) 979 { 980 return -1; 981 } 982 983 static inline int cpupid_to_pid(int cpupid) 984 { 985 return -1; 986 } 987 988 static inline int cpupid_to_cpu(int cpupid) 989 { 990 return -1; 991 } 992 993 static inline int cpu_pid_to_cpupid(int nid, int pid) 994 { 995 return -1; 996 } 997 998 static inline bool cpupid_pid_unset(int cpupid) 999 { 1000 return 1; 1001 } 1002 1003 static inline void page_cpupid_reset_last(struct page *page) 1004 { 1005 } 1006 1007 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) 1008 { 1009 return false; 1010 } 1011 #endif /* CONFIG_NUMA_BALANCING */ 1012 1013 static inline struct zone *page_zone(const struct page *page) 1014 { 1015 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 1016 } 1017 1018 static inline pg_data_t *page_pgdat(const struct page *page) 1019 { 1020 return NODE_DATA(page_to_nid(page)); 1021 } 1022 1023 #ifdef SECTION_IN_PAGE_FLAGS 1024 static inline void set_page_section(struct page *page, unsigned long section) 1025 { 1026 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 1027 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 1028 } 1029 1030 static inline unsigned long page_to_section(const struct page *page) 1031 { 1032 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 1033 } 1034 #endif 1035 1036 static inline void set_page_zone(struct page *page, enum zone_type zone) 1037 { 1038 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 1039 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 1040 } 1041 1042 static inline void set_page_node(struct page *page, unsigned long node) 1043 { 1044 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 1045 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 1046 } 1047 1048 static inline void set_page_links(struct page *page, enum zone_type zone, 1049 unsigned long node, unsigned long pfn) 1050 { 1051 set_page_zone(page, zone); 1052 set_page_node(page, node); 1053 #ifdef SECTION_IN_PAGE_FLAGS 1054 set_page_section(page, pfn_to_section_nr(pfn)); 1055 #endif 1056 } 1057 1058 #ifdef CONFIG_MEMCG 1059 static inline struct mem_cgroup *page_memcg(struct page *page) 1060 { 1061 return page->mem_cgroup; 1062 } 1063 static inline struct mem_cgroup *page_memcg_rcu(struct page *page) 1064 { 1065 WARN_ON_ONCE(!rcu_read_lock_held()); 1066 return READ_ONCE(page->mem_cgroup); 1067 } 1068 #else 1069 static inline struct mem_cgroup *page_memcg(struct page *page) 1070 { 1071 return NULL; 1072 } 1073 static inline struct mem_cgroup *page_memcg_rcu(struct page *page) 1074 { 1075 WARN_ON_ONCE(!rcu_read_lock_held()); 1076 return NULL; 1077 } 1078 #endif 1079 1080 /* 1081 * Some inline functions in vmstat.h depend on page_zone() 1082 */ 1083 #include <linux/vmstat.h> 1084 1085 static __always_inline void *lowmem_page_address(const struct page *page) 1086 { 1087 return page_to_virt(page); 1088 } 1089 1090 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 1091 #define HASHED_PAGE_VIRTUAL 1092 #endif 1093 1094 #if defined(WANT_PAGE_VIRTUAL) 1095 static inline void *page_address(const struct page *page) 1096 { 1097 return page->virtual; 1098 } 1099 static inline void set_page_address(struct page *page, void *address) 1100 { 1101 page->virtual = address; 1102 } 1103 #define page_address_init() do { } while(0) 1104 #endif 1105 1106 #if defined(HASHED_PAGE_VIRTUAL) 1107 void *page_address(const struct page *page); 1108 void set_page_address(struct page *page, void *virtual); 1109 void page_address_init(void); 1110 #endif 1111 1112 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 1113 #define page_address(page) lowmem_page_address(page) 1114 #define set_page_address(page, address) do { } while(0) 1115 #define page_address_init() do { } while(0) 1116 #endif 1117 1118 extern void *page_rmapping(struct page *page); 1119 extern struct anon_vma *page_anon_vma(struct page *page); 1120 extern struct address_space *page_mapping(struct page *page); 1121 1122 extern struct address_space *__page_file_mapping(struct page *); 1123 1124 static inline 1125 struct address_space *page_file_mapping(struct page *page) 1126 { 1127 if (unlikely(PageSwapCache(page))) 1128 return __page_file_mapping(page); 1129 1130 return page->mapping; 1131 } 1132 1133 extern pgoff_t __page_file_index(struct page *page); 1134 1135 /* 1136 * Return the pagecache index of the passed page. Regular pagecache pages 1137 * use ->index whereas swapcache pages use swp_offset(->private) 1138 */ 1139 static inline pgoff_t page_index(struct page *page) 1140 { 1141 if (unlikely(PageSwapCache(page))) 1142 return __page_file_index(page); 1143 return page->index; 1144 } 1145 1146 bool page_mapped(struct page *page); 1147 struct address_space *page_mapping(struct page *page); 1148 1149 /* 1150 * Return true only if the page has been allocated with 1151 * ALLOC_NO_WATERMARKS and the low watermark was not 1152 * met implying that the system is under some pressure. 1153 */ 1154 static inline bool page_is_pfmemalloc(struct page *page) 1155 { 1156 /* 1157 * Page index cannot be this large so this must be 1158 * a pfmemalloc page. 1159 */ 1160 return page->index == -1UL; 1161 } 1162 1163 /* 1164 * Only to be called by the page allocator on a freshly allocated 1165 * page. 1166 */ 1167 static inline void set_page_pfmemalloc(struct page *page) 1168 { 1169 page->index = -1UL; 1170 } 1171 1172 static inline void clear_page_pfmemalloc(struct page *page) 1173 { 1174 page->index = 0; 1175 } 1176 1177 /* 1178 * Different kinds of faults, as returned by handle_mm_fault(). 1179 * Used to decide whether a process gets delivered SIGBUS or 1180 * just gets major/minor fault counters bumped up. 1181 */ 1182 1183 #define VM_FAULT_OOM 0x0001 1184 #define VM_FAULT_SIGBUS 0x0002 1185 #define VM_FAULT_MAJOR 0x0004 1186 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 1187 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ 1188 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ 1189 #define VM_FAULT_SIGSEGV 0x0040 1190 1191 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 1192 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 1193 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ 1194 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ 1195 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */ 1196 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables 1197 * and needs fsync() to complete (for 1198 * synchronous page faults in DAX) */ 1199 1200 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \ 1201 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \ 1202 VM_FAULT_FALLBACK) 1203 1204 #define VM_FAULT_RESULT_TRACE \ 1205 { VM_FAULT_OOM, "OOM" }, \ 1206 { VM_FAULT_SIGBUS, "SIGBUS" }, \ 1207 { VM_FAULT_MAJOR, "MAJOR" }, \ 1208 { VM_FAULT_WRITE, "WRITE" }, \ 1209 { VM_FAULT_HWPOISON, "HWPOISON" }, \ 1210 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \ 1211 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \ 1212 { VM_FAULT_NOPAGE, "NOPAGE" }, \ 1213 { VM_FAULT_LOCKED, "LOCKED" }, \ 1214 { VM_FAULT_RETRY, "RETRY" }, \ 1215 { VM_FAULT_FALLBACK, "FALLBACK" }, \ 1216 { VM_FAULT_DONE_COW, "DONE_COW" }, \ 1217 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" } 1218 1219 /* Encode hstate index for a hwpoisoned large page */ 1220 #define VM_FAULT_SET_HINDEX(x) ((x) << 12) 1221 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) 1222 1223 /* 1224 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 1225 */ 1226 extern void pagefault_out_of_memory(void); 1227 1228 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 1229 1230 /* 1231 * Flags passed to show_mem() and show_free_areas() to suppress output in 1232 * various contexts. 1233 */ 1234 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ 1235 1236 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask); 1237 1238 extern bool can_do_mlock(void); 1239 extern int user_shm_lock(size_t, struct user_struct *); 1240 extern void user_shm_unlock(size_t, struct user_struct *); 1241 1242 /* 1243 * Parameter block passed down to zap_pte_range in exceptional cases. 1244 */ 1245 struct zap_details { 1246 struct address_space *check_mapping; /* Check page->mapping if set */ 1247 pgoff_t first_index; /* Lowest page->index to unmap */ 1248 pgoff_t last_index; /* Highest page->index to unmap */ 1249 }; 1250 1251 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 1252 pte_t pte, bool with_public_device); 1253 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false) 1254 1255 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, 1256 pmd_t pmd); 1257 1258 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 1259 unsigned long size); 1260 void zap_page_range(struct vm_area_struct *vma, unsigned long address, 1261 unsigned long size); 1262 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, 1263 unsigned long start, unsigned long end); 1264 1265 /** 1266 * mm_walk - callbacks for walk_page_range 1267 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 1268 * this handler should only handle pud_trans_huge() puds. 1269 * the pmd_entry or pte_entry callbacks will be used for 1270 * regular PUDs. 1271 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 1272 * this handler is required to be able to handle 1273 * pmd_trans_huge() pmds. They may simply choose to 1274 * split_huge_page() instead of handling it explicitly. 1275 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 1276 * @pte_hole: if set, called for each hole at all levels 1277 * @hugetlb_entry: if set, called for each hugetlb entry 1278 * @test_walk: caller specific callback function to determine whether 1279 * we walk over the current vma or not. Returning 0 1280 * value means "do page table walk over the current vma," 1281 * and a negative one means "abort current page table walk 1282 * right now." 1 means "skip the current vma." 1283 * @mm: mm_struct representing the target process of page table walk 1284 * @vma: vma currently walked (NULL if walking outside vmas) 1285 * @private: private data for callbacks' usage 1286 * 1287 * (see the comment on walk_page_range() for more details) 1288 */ 1289 struct mm_walk { 1290 int (*pud_entry)(pud_t *pud, unsigned long addr, 1291 unsigned long next, struct mm_walk *walk); 1292 int (*pmd_entry)(pmd_t *pmd, unsigned long addr, 1293 unsigned long next, struct mm_walk *walk); 1294 int (*pte_entry)(pte_t *pte, unsigned long addr, 1295 unsigned long next, struct mm_walk *walk); 1296 int (*pte_hole)(unsigned long addr, unsigned long next, 1297 struct mm_walk *walk); 1298 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, 1299 unsigned long addr, unsigned long next, 1300 struct mm_walk *walk); 1301 int (*test_walk)(unsigned long addr, unsigned long next, 1302 struct mm_walk *walk); 1303 struct mm_struct *mm; 1304 struct vm_area_struct *vma; 1305 void *private; 1306 }; 1307 1308 int walk_page_range(unsigned long addr, unsigned long end, 1309 struct mm_walk *walk); 1310 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk); 1311 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 1312 unsigned long end, unsigned long floor, unsigned long ceiling); 1313 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 1314 struct vm_area_struct *vma); 1315 int follow_pte_pmd(struct mm_struct *mm, unsigned long address, 1316 unsigned long *start, unsigned long *end, 1317 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp); 1318 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 1319 unsigned long *pfn); 1320 int follow_phys(struct vm_area_struct *vma, unsigned long address, 1321 unsigned int flags, unsigned long *prot, resource_size_t *phys); 1322 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 1323 void *buf, int len, int write); 1324 1325 extern void truncate_pagecache(struct inode *inode, loff_t new); 1326 extern void truncate_setsize(struct inode *inode, loff_t newsize); 1327 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); 1328 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); 1329 int truncate_inode_page(struct address_space *mapping, struct page *page); 1330 int generic_error_remove_page(struct address_space *mapping, struct page *page); 1331 int invalidate_inode_page(struct page *page); 1332 1333 #ifdef CONFIG_MMU 1334 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address, 1335 unsigned int flags); 1336 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, 1337 unsigned long address, unsigned int fault_flags, 1338 bool *unlocked); 1339 void unmap_mapping_pages(struct address_space *mapping, 1340 pgoff_t start, pgoff_t nr, bool even_cows); 1341 void unmap_mapping_range(struct address_space *mapping, 1342 loff_t const holebegin, loff_t const holelen, int even_cows); 1343 #else 1344 static inline int handle_mm_fault(struct vm_area_struct *vma, 1345 unsigned long address, unsigned int flags) 1346 { 1347 /* should never happen if there's no MMU */ 1348 BUG(); 1349 return VM_FAULT_SIGBUS; 1350 } 1351 static inline int fixup_user_fault(struct task_struct *tsk, 1352 struct mm_struct *mm, unsigned long address, 1353 unsigned int fault_flags, bool *unlocked) 1354 { 1355 /* should never happen if there's no MMU */ 1356 BUG(); 1357 return -EFAULT; 1358 } 1359 static inline void unmap_mapping_pages(struct address_space *mapping, 1360 pgoff_t start, pgoff_t nr, bool even_cows) { } 1361 static inline void unmap_mapping_range(struct address_space *mapping, 1362 loff_t const holebegin, loff_t const holelen, int even_cows) { } 1363 #endif 1364 1365 static inline void unmap_shared_mapping_range(struct address_space *mapping, 1366 loff_t const holebegin, loff_t const holelen) 1367 { 1368 unmap_mapping_range(mapping, holebegin, holelen, 0); 1369 } 1370 1371 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, 1372 void *buf, int len, unsigned int gup_flags); 1373 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, 1374 void *buf, int len, unsigned int gup_flags); 1375 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm, 1376 unsigned long addr, void *buf, int len, unsigned int gup_flags); 1377 1378 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm, 1379 unsigned long start, unsigned long nr_pages, 1380 unsigned int gup_flags, struct page **pages, 1381 struct vm_area_struct **vmas, int *locked); 1382 long get_user_pages(unsigned long start, unsigned long nr_pages, 1383 unsigned int gup_flags, struct page **pages, 1384 struct vm_area_struct **vmas); 1385 long get_user_pages_locked(unsigned long start, unsigned long nr_pages, 1386 unsigned int gup_flags, struct page **pages, int *locked); 1387 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, 1388 struct page **pages, unsigned int gup_flags); 1389 #ifdef CONFIG_FS_DAX 1390 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages, 1391 unsigned int gup_flags, struct page **pages, 1392 struct vm_area_struct **vmas); 1393 #else 1394 static inline long get_user_pages_longterm(unsigned long start, 1395 unsigned long nr_pages, unsigned int gup_flags, 1396 struct page **pages, struct vm_area_struct **vmas) 1397 { 1398 return get_user_pages(start, nr_pages, gup_flags, pages, vmas); 1399 } 1400 #endif /* CONFIG_FS_DAX */ 1401 1402 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 1403 struct page **pages); 1404 1405 /* Container for pinned pfns / pages */ 1406 struct frame_vector { 1407 unsigned int nr_allocated; /* Number of frames we have space for */ 1408 unsigned int nr_frames; /* Number of frames stored in ptrs array */ 1409 bool got_ref; /* Did we pin pages by getting page ref? */ 1410 bool is_pfns; /* Does array contain pages or pfns? */ 1411 void *ptrs[0]; /* Array of pinned pfns / pages. Use 1412 * pfns_vector_pages() or pfns_vector_pfns() 1413 * for access */ 1414 }; 1415 1416 struct frame_vector *frame_vector_create(unsigned int nr_frames); 1417 void frame_vector_destroy(struct frame_vector *vec); 1418 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, 1419 unsigned int gup_flags, struct frame_vector *vec); 1420 void put_vaddr_frames(struct frame_vector *vec); 1421 int frame_vector_to_pages(struct frame_vector *vec); 1422 void frame_vector_to_pfns(struct frame_vector *vec); 1423 1424 static inline unsigned int frame_vector_count(struct frame_vector *vec) 1425 { 1426 return vec->nr_frames; 1427 } 1428 1429 static inline struct page **frame_vector_pages(struct frame_vector *vec) 1430 { 1431 if (vec->is_pfns) { 1432 int err = frame_vector_to_pages(vec); 1433 1434 if (err) 1435 return ERR_PTR(err); 1436 } 1437 return (struct page **)(vec->ptrs); 1438 } 1439 1440 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) 1441 { 1442 if (!vec->is_pfns) 1443 frame_vector_to_pfns(vec); 1444 return (unsigned long *)(vec->ptrs); 1445 } 1446 1447 struct kvec; 1448 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, 1449 struct page **pages); 1450 int get_kernel_page(unsigned long start, int write, struct page **pages); 1451 struct page *get_dump_page(unsigned long addr); 1452 1453 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 1454 extern void do_invalidatepage(struct page *page, unsigned int offset, 1455 unsigned int length); 1456 1457 int __set_page_dirty_nobuffers(struct page *page); 1458 int __set_page_dirty_no_writeback(struct page *page); 1459 int redirty_page_for_writepage(struct writeback_control *wbc, 1460 struct page *page); 1461 void account_page_dirtied(struct page *page, struct address_space *mapping); 1462 void account_page_cleaned(struct page *page, struct address_space *mapping, 1463 struct bdi_writeback *wb); 1464 int set_page_dirty(struct page *page); 1465 int set_page_dirty_lock(struct page *page); 1466 void __cancel_dirty_page(struct page *page); 1467 static inline void cancel_dirty_page(struct page *page) 1468 { 1469 /* Avoid atomic ops, locking, etc. when not actually needed. */ 1470 if (PageDirty(page)) 1471 __cancel_dirty_page(page); 1472 } 1473 int clear_page_dirty_for_io(struct page *page); 1474 1475 int get_cmdline(struct task_struct *task, char *buffer, int buflen); 1476 1477 static inline bool vma_is_anonymous(struct vm_area_struct *vma) 1478 { 1479 return !vma->vm_ops; 1480 } 1481 1482 #ifdef CONFIG_SHMEM 1483 /* 1484 * The vma_is_shmem is not inline because it is used only by slow 1485 * paths in userfault. 1486 */ 1487 bool vma_is_shmem(struct vm_area_struct *vma); 1488 #else 1489 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; } 1490 #endif 1491 1492 int vma_is_stack_for_current(struct vm_area_struct *vma); 1493 1494 extern unsigned long move_page_tables(struct vm_area_struct *vma, 1495 unsigned long old_addr, struct vm_area_struct *new_vma, 1496 unsigned long new_addr, unsigned long len, 1497 bool need_rmap_locks); 1498 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, 1499 unsigned long end, pgprot_t newprot, 1500 int dirty_accountable, int prot_numa); 1501 extern int mprotect_fixup(struct vm_area_struct *vma, 1502 struct vm_area_struct **pprev, unsigned long start, 1503 unsigned long end, unsigned long newflags); 1504 1505 /* 1506 * doesn't attempt to fault and will return short. 1507 */ 1508 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 1509 struct page **pages); 1510 /* 1511 * per-process(per-mm_struct) statistics. 1512 */ 1513 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 1514 { 1515 long val = atomic_long_read(&mm->rss_stat.count[member]); 1516 1517 #ifdef SPLIT_RSS_COUNTING 1518 /* 1519 * counter is updated in asynchronous manner and may go to minus. 1520 * But it's never be expected number for users. 1521 */ 1522 if (val < 0) 1523 val = 0; 1524 #endif 1525 return (unsigned long)val; 1526 } 1527 1528 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 1529 { 1530 atomic_long_add(value, &mm->rss_stat.count[member]); 1531 } 1532 1533 static inline void inc_mm_counter(struct mm_struct *mm, int member) 1534 { 1535 atomic_long_inc(&mm->rss_stat.count[member]); 1536 } 1537 1538 static inline void dec_mm_counter(struct mm_struct *mm, int member) 1539 { 1540 atomic_long_dec(&mm->rss_stat.count[member]); 1541 } 1542 1543 /* Optimized variant when page is already known not to be PageAnon */ 1544 static inline int mm_counter_file(struct page *page) 1545 { 1546 if (PageSwapBacked(page)) 1547 return MM_SHMEMPAGES; 1548 return MM_FILEPAGES; 1549 } 1550 1551 static inline int mm_counter(struct page *page) 1552 { 1553 if (PageAnon(page)) 1554 return MM_ANONPAGES; 1555 return mm_counter_file(page); 1556 } 1557 1558 static inline unsigned long get_mm_rss(struct mm_struct *mm) 1559 { 1560 return get_mm_counter(mm, MM_FILEPAGES) + 1561 get_mm_counter(mm, MM_ANONPAGES) + 1562 get_mm_counter(mm, MM_SHMEMPAGES); 1563 } 1564 1565 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 1566 { 1567 return max(mm->hiwater_rss, get_mm_rss(mm)); 1568 } 1569 1570 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 1571 { 1572 return max(mm->hiwater_vm, mm->total_vm); 1573 } 1574 1575 static inline void update_hiwater_rss(struct mm_struct *mm) 1576 { 1577 unsigned long _rss = get_mm_rss(mm); 1578 1579 if ((mm)->hiwater_rss < _rss) 1580 (mm)->hiwater_rss = _rss; 1581 } 1582 1583 static inline void update_hiwater_vm(struct mm_struct *mm) 1584 { 1585 if (mm->hiwater_vm < mm->total_vm) 1586 mm->hiwater_vm = mm->total_vm; 1587 } 1588 1589 static inline void reset_mm_hiwater_rss(struct mm_struct *mm) 1590 { 1591 mm->hiwater_rss = get_mm_rss(mm); 1592 } 1593 1594 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 1595 struct mm_struct *mm) 1596 { 1597 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 1598 1599 if (*maxrss < hiwater_rss) 1600 *maxrss = hiwater_rss; 1601 } 1602 1603 #if defined(SPLIT_RSS_COUNTING) 1604 void sync_mm_rss(struct mm_struct *mm); 1605 #else 1606 static inline void sync_mm_rss(struct mm_struct *mm) 1607 { 1608 } 1609 #endif 1610 1611 #ifndef __HAVE_ARCH_PTE_DEVMAP 1612 static inline int pte_devmap(pte_t pte) 1613 { 1614 return 0; 1615 } 1616 #endif 1617 1618 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); 1619 1620 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, 1621 spinlock_t **ptl); 1622 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, 1623 spinlock_t **ptl) 1624 { 1625 pte_t *ptep; 1626 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); 1627 return ptep; 1628 } 1629 1630 #ifdef __PAGETABLE_P4D_FOLDED 1631 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, 1632 unsigned long address) 1633 { 1634 return 0; 1635 } 1636 #else 1637 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1638 #endif 1639 1640 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU) 1641 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, 1642 unsigned long address) 1643 { 1644 return 0; 1645 } 1646 static inline void mm_inc_nr_puds(struct mm_struct *mm) {} 1647 static inline void mm_dec_nr_puds(struct mm_struct *mm) {} 1648 1649 #else 1650 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address); 1651 1652 static inline void mm_inc_nr_puds(struct mm_struct *mm) 1653 { 1654 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); 1655 } 1656 1657 static inline void mm_dec_nr_puds(struct mm_struct *mm) 1658 { 1659 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes); 1660 } 1661 #endif 1662 1663 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) 1664 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1665 unsigned long address) 1666 { 1667 return 0; 1668 } 1669 1670 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} 1671 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} 1672 1673 #else 1674 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1675 1676 static inline void mm_inc_nr_pmds(struct mm_struct *mm) 1677 { 1678 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); 1679 } 1680 1681 static inline void mm_dec_nr_pmds(struct mm_struct *mm) 1682 { 1683 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes); 1684 } 1685 #endif 1686 1687 #ifdef CONFIG_MMU 1688 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) 1689 { 1690 atomic_long_set(&mm->pgtables_bytes, 0); 1691 } 1692 1693 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) 1694 { 1695 return atomic_long_read(&mm->pgtables_bytes); 1696 } 1697 1698 static inline void mm_inc_nr_ptes(struct mm_struct *mm) 1699 { 1700 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); 1701 } 1702 1703 static inline void mm_dec_nr_ptes(struct mm_struct *mm) 1704 { 1705 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes); 1706 } 1707 #else 1708 1709 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {} 1710 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm) 1711 { 1712 return 0; 1713 } 1714 1715 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {} 1716 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {} 1717 #endif 1718 1719 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 1720 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1721 1722 /* 1723 * The following ifdef needed to get the 4level-fixup.h header to work. 1724 * Remove it when 4level-fixup.h has been removed. 1725 */ 1726 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1727 1728 #ifndef __ARCH_HAS_5LEVEL_HACK 1729 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd, 1730 unsigned long address) 1731 { 1732 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ? 1733 NULL : p4d_offset(pgd, address); 1734 } 1735 1736 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d, 1737 unsigned long address) 1738 { 1739 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ? 1740 NULL : pud_offset(p4d, address); 1741 } 1742 #endif /* !__ARCH_HAS_5LEVEL_HACK */ 1743 1744 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1745 { 1746 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1747 NULL: pmd_offset(pud, address); 1748 } 1749 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1750 1751 #if USE_SPLIT_PTE_PTLOCKS 1752 #if ALLOC_SPLIT_PTLOCKS 1753 void __init ptlock_cache_init(void); 1754 extern bool ptlock_alloc(struct page *page); 1755 extern void ptlock_free(struct page *page); 1756 1757 static inline spinlock_t *ptlock_ptr(struct page *page) 1758 { 1759 return page->ptl; 1760 } 1761 #else /* ALLOC_SPLIT_PTLOCKS */ 1762 static inline void ptlock_cache_init(void) 1763 { 1764 } 1765 1766 static inline bool ptlock_alloc(struct page *page) 1767 { 1768 return true; 1769 } 1770 1771 static inline void ptlock_free(struct page *page) 1772 { 1773 } 1774 1775 static inline spinlock_t *ptlock_ptr(struct page *page) 1776 { 1777 return &page->ptl; 1778 } 1779 #endif /* ALLOC_SPLIT_PTLOCKS */ 1780 1781 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) 1782 { 1783 return ptlock_ptr(pmd_page(*pmd)); 1784 } 1785 1786 static inline bool ptlock_init(struct page *page) 1787 { 1788 /* 1789 * prep_new_page() initialize page->private (and therefore page->ptl) 1790 * with 0. Make sure nobody took it in use in between. 1791 * 1792 * It can happen if arch try to use slab for page table allocation: 1793 * slab code uses page->slab_cache, which share storage with page->ptl. 1794 */ 1795 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); 1796 if (!ptlock_alloc(page)) 1797 return false; 1798 spin_lock_init(ptlock_ptr(page)); 1799 return true; 1800 } 1801 1802 /* Reset page->mapping so free_pages_check won't complain. */ 1803 static inline void pte_lock_deinit(struct page *page) 1804 { 1805 page->mapping = NULL; 1806 ptlock_free(page); 1807 } 1808 1809 #else /* !USE_SPLIT_PTE_PTLOCKS */ 1810 /* 1811 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1812 */ 1813 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) 1814 { 1815 return &mm->page_table_lock; 1816 } 1817 static inline void ptlock_cache_init(void) {} 1818 static inline bool ptlock_init(struct page *page) { return true; } 1819 static inline void pte_lock_deinit(struct page *page) {} 1820 #endif /* USE_SPLIT_PTE_PTLOCKS */ 1821 1822 static inline void pgtable_init(void) 1823 { 1824 ptlock_cache_init(); 1825 pgtable_cache_init(); 1826 } 1827 1828 static inline bool pgtable_page_ctor(struct page *page) 1829 { 1830 if (!ptlock_init(page)) 1831 return false; 1832 inc_zone_page_state(page, NR_PAGETABLE); 1833 return true; 1834 } 1835 1836 static inline void pgtable_page_dtor(struct page *page) 1837 { 1838 pte_lock_deinit(page); 1839 dec_zone_page_state(page, NR_PAGETABLE); 1840 } 1841 1842 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1843 ({ \ 1844 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1845 pte_t *__pte = pte_offset_map(pmd, address); \ 1846 *(ptlp) = __ptl; \ 1847 spin_lock(__ptl); \ 1848 __pte; \ 1849 }) 1850 1851 #define pte_unmap_unlock(pte, ptl) do { \ 1852 spin_unlock(ptl); \ 1853 pte_unmap(pte); \ 1854 } while (0) 1855 1856 #define pte_alloc(mm, pmd, address) \ 1857 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address)) 1858 1859 #define pte_alloc_map(mm, pmd, address) \ 1860 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address)) 1861 1862 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1863 (pte_alloc(mm, pmd, address) ? \ 1864 NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) 1865 1866 #define pte_alloc_kernel(pmd, address) \ 1867 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1868 NULL: pte_offset_kernel(pmd, address)) 1869 1870 #if USE_SPLIT_PMD_PTLOCKS 1871 1872 static struct page *pmd_to_page(pmd_t *pmd) 1873 { 1874 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); 1875 return virt_to_page((void *)((unsigned long) pmd & mask)); 1876 } 1877 1878 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) 1879 { 1880 return ptlock_ptr(pmd_to_page(pmd)); 1881 } 1882 1883 static inline bool pgtable_pmd_page_ctor(struct page *page) 1884 { 1885 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 1886 page->pmd_huge_pte = NULL; 1887 #endif 1888 return ptlock_init(page); 1889 } 1890 1891 static inline void pgtable_pmd_page_dtor(struct page *page) 1892 { 1893 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 1894 VM_BUG_ON_PAGE(page->pmd_huge_pte, page); 1895 #endif 1896 ptlock_free(page); 1897 } 1898 1899 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) 1900 1901 #else 1902 1903 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) 1904 { 1905 return &mm->page_table_lock; 1906 } 1907 1908 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } 1909 static inline void pgtable_pmd_page_dtor(struct page *page) {} 1910 1911 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) 1912 1913 #endif 1914 1915 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) 1916 { 1917 spinlock_t *ptl = pmd_lockptr(mm, pmd); 1918 spin_lock(ptl); 1919 return ptl; 1920 } 1921 1922 /* 1923 * No scalability reason to split PUD locks yet, but follow the same pattern 1924 * as the PMD locks to make it easier if we decide to. The VM should not be 1925 * considered ready to switch to split PUD locks yet; there may be places 1926 * which need to be converted from page_table_lock. 1927 */ 1928 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud) 1929 { 1930 return &mm->page_table_lock; 1931 } 1932 1933 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud) 1934 { 1935 spinlock_t *ptl = pud_lockptr(mm, pud); 1936 1937 spin_lock(ptl); 1938 return ptl; 1939 } 1940 1941 extern void __init pagecache_init(void); 1942 extern void free_area_init(unsigned long * zones_size); 1943 extern void free_area_init_node(int nid, unsigned long * zones_size, 1944 unsigned long zone_start_pfn, unsigned long *zholes_size); 1945 extern void free_initmem(void); 1946 1947 /* 1948 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) 1949 * into the buddy system. The freed pages will be poisoned with pattern 1950 * "poison" if it's within range [0, UCHAR_MAX]. 1951 * Return pages freed into the buddy system. 1952 */ 1953 extern unsigned long free_reserved_area(void *start, void *end, 1954 int poison, char *s); 1955 1956 #ifdef CONFIG_HIGHMEM 1957 /* 1958 * Free a highmem page into the buddy system, adjusting totalhigh_pages 1959 * and totalram_pages. 1960 */ 1961 extern void free_highmem_page(struct page *page); 1962 #endif 1963 1964 extern void adjust_managed_page_count(struct page *page, long count); 1965 extern void mem_init_print_info(const char *str); 1966 1967 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); 1968 1969 /* Free the reserved page into the buddy system, so it gets managed. */ 1970 static inline void __free_reserved_page(struct page *page) 1971 { 1972 ClearPageReserved(page); 1973 init_page_count(page); 1974 __free_page(page); 1975 } 1976 1977 static inline void free_reserved_page(struct page *page) 1978 { 1979 __free_reserved_page(page); 1980 adjust_managed_page_count(page, 1); 1981 } 1982 1983 static inline void mark_page_reserved(struct page *page) 1984 { 1985 SetPageReserved(page); 1986 adjust_managed_page_count(page, -1); 1987 } 1988 1989 /* 1990 * Default method to free all the __init memory into the buddy system. 1991 * The freed pages will be poisoned with pattern "poison" if it's within 1992 * range [0, UCHAR_MAX]. 1993 * Return pages freed into the buddy system. 1994 */ 1995 static inline unsigned long free_initmem_default(int poison) 1996 { 1997 extern char __init_begin[], __init_end[]; 1998 1999 return free_reserved_area(&__init_begin, &__init_end, 2000 poison, "unused kernel"); 2001 } 2002 2003 static inline unsigned long get_num_physpages(void) 2004 { 2005 int nid; 2006 unsigned long phys_pages = 0; 2007 2008 for_each_online_node(nid) 2009 phys_pages += node_present_pages(nid); 2010 2011 return phys_pages; 2012 } 2013 2014 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP 2015 /* 2016 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its 2017 * zones, allocate the backing mem_map and account for memory holes in a more 2018 * architecture independent manner. This is a substitute for creating the 2019 * zone_sizes[] and zholes_size[] arrays and passing them to 2020 * free_area_init_node() 2021 * 2022 * An architecture is expected to register range of page frames backed by 2023 * physical memory with memblock_add[_node]() before calling 2024 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 2025 * usage, an architecture is expected to do something like 2026 * 2027 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 2028 * max_highmem_pfn}; 2029 * for_each_valid_physical_page_range() 2030 * memblock_add_node(base, size, nid) 2031 * free_area_init_nodes(max_zone_pfns); 2032 * 2033 * free_bootmem_with_active_regions() calls free_bootmem_node() for each 2034 * registered physical page range. Similarly 2035 * sparse_memory_present_with_active_regions() calls memory_present() for 2036 * each range when SPARSEMEM is enabled. 2037 * 2038 * See mm/page_alloc.c for more information on each function exposed by 2039 * CONFIG_HAVE_MEMBLOCK_NODE_MAP. 2040 */ 2041 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 2042 unsigned long node_map_pfn_alignment(void); 2043 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 2044 unsigned long end_pfn); 2045 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 2046 unsigned long end_pfn); 2047 extern void get_pfn_range_for_nid(unsigned int nid, 2048 unsigned long *start_pfn, unsigned long *end_pfn); 2049 extern unsigned long find_min_pfn_with_active_regions(void); 2050 extern void free_bootmem_with_active_regions(int nid, 2051 unsigned long max_low_pfn); 2052 extern void sparse_memory_present_with_active_regions(int nid); 2053 2054 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ 2055 2056 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ 2057 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 2058 static inline int __early_pfn_to_nid(unsigned long pfn, 2059 struct mminit_pfnnid_cache *state) 2060 { 2061 return 0; 2062 } 2063 #else 2064 /* please see mm/page_alloc.c */ 2065 extern int __meminit early_pfn_to_nid(unsigned long pfn); 2066 /* there is a per-arch backend function. */ 2067 extern int __meminit __early_pfn_to_nid(unsigned long pfn, 2068 struct mminit_pfnnid_cache *state); 2069 #endif 2070 2071 #ifdef CONFIG_HAVE_MEMBLOCK 2072 void zero_resv_unavail(void); 2073 #else 2074 static inline void zero_resv_unavail(void) {} 2075 #endif 2076 2077 extern void set_dma_reserve(unsigned long new_dma_reserve); 2078 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long, 2079 enum memmap_context, struct vmem_altmap *); 2080 extern void setup_per_zone_wmarks(void); 2081 extern int __meminit init_per_zone_wmark_min(void); 2082 extern void mem_init(void); 2083 extern void __init mmap_init(void); 2084 extern void show_mem(unsigned int flags, nodemask_t *nodemask); 2085 extern long si_mem_available(void); 2086 extern void si_meminfo(struct sysinfo * val); 2087 extern void si_meminfo_node(struct sysinfo *val, int nid); 2088 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES 2089 extern unsigned long arch_reserved_kernel_pages(void); 2090 #endif 2091 2092 extern __printf(3, 4) 2093 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...); 2094 2095 extern void setup_per_cpu_pageset(void); 2096 2097 extern void zone_pcp_update(struct zone *zone); 2098 extern void zone_pcp_reset(struct zone *zone); 2099 2100 /* page_alloc.c */ 2101 extern int min_free_kbytes; 2102 extern int watermark_scale_factor; 2103 2104 /* nommu.c */ 2105 extern atomic_long_t mmap_pages_allocated; 2106 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 2107 2108 /* interval_tree.c */ 2109 void vma_interval_tree_insert(struct vm_area_struct *node, 2110 struct rb_root_cached *root); 2111 void vma_interval_tree_insert_after(struct vm_area_struct *node, 2112 struct vm_area_struct *prev, 2113 struct rb_root_cached *root); 2114 void vma_interval_tree_remove(struct vm_area_struct *node, 2115 struct rb_root_cached *root); 2116 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root, 2117 unsigned long start, unsigned long last); 2118 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, 2119 unsigned long start, unsigned long last); 2120 2121 #define vma_interval_tree_foreach(vma, root, start, last) \ 2122 for (vma = vma_interval_tree_iter_first(root, start, last); \ 2123 vma; vma = vma_interval_tree_iter_next(vma, start, last)) 2124 2125 void anon_vma_interval_tree_insert(struct anon_vma_chain *node, 2126 struct rb_root_cached *root); 2127 void anon_vma_interval_tree_remove(struct anon_vma_chain *node, 2128 struct rb_root_cached *root); 2129 struct anon_vma_chain * 2130 anon_vma_interval_tree_iter_first(struct rb_root_cached *root, 2131 unsigned long start, unsigned long last); 2132 struct anon_vma_chain *anon_vma_interval_tree_iter_next( 2133 struct anon_vma_chain *node, unsigned long start, unsigned long last); 2134 #ifdef CONFIG_DEBUG_VM_RB 2135 void anon_vma_interval_tree_verify(struct anon_vma_chain *node); 2136 #endif 2137 2138 #define anon_vma_interval_tree_foreach(avc, root, start, last) \ 2139 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ 2140 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) 2141 2142 /* mmap.c */ 2143 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 2144 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start, 2145 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, 2146 struct vm_area_struct *expand); 2147 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start, 2148 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert) 2149 { 2150 return __vma_adjust(vma, start, end, pgoff, insert, NULL); 2151 } 2152 extern struct vm_area_struct *vma_merge(struct mm_struct *, 2153 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 2154 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 2155 struct mempolicy *, struct vm_userfaultfd_ctx); 2156 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 2157 extern int __split_vma(struct mm_struct *, struct vm_area_struct *, 2158 unsigned long addr, int new_below); 2159 extern int split_vma(struct mm_struct *, struct vm_area_struct *, 2160 unsigned long addr, int new_below); 2161 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 2162 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 2163 struct rb_node **, struct rb_node *); 2164 extern void unlink_file_vma(struct vm_area_struct *); 2165 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 2166 unsigned long addr, unsigned long len, pgoff_t pgoff, 2167 bool *need_rmap_locks); 2168 extern void exit_mmap(struct mm_struct *); 2169 2170 static inline int check_data_rlimit(unsigned long rlim, 2171 unsigned long new, 2172 unsigned long start, 2173 unsigned long end_data, 2174 unsigned long start_data) 2175 { 2176 if (rlim < RLIM_INFINITY) { 2177 if (((new - start) + (end_data - start_data)) > rlim) 2178 return -ENOSPC; 2179 } 2180 2181 return 0; 2182 } 2183 2184 extern int mm_take_all_locks(struct mm_struct *mm); 2185 extern void mm_drop_all_locks(struct mm_struct *mm); 2186 2187 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); 2188 extern struct file *get_mm_exe_file(struct mm_struct *mm); 2189 extern struct file *get_task_exe_file(struct task_struct *task); 2190 2191 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); 2192 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); 2193 2194 extern bool vma_is_special_mapping(const struct vm_area_struct *vma, 2195 const struct vm_special_mapping *sm); 2196 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, 2197 unsigned long addr, unsigned long len, 2198 unsigned long flags, 2199 const struct vm_special_mapping *spec); 2200 /* This is an obsolete alternative to _install_special_mapping. */ 2201 extern int install_special_mapping(struct mm_struct *mm, 2202 unsigned long addr, unsigned long len, 2203 unsigned long flags, struct page **pages); 2204 2205 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 2206 2207 extern unsigned long mmap_region(struct file *file, unsigned long addr, 2208 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff, 2209 struct list_head *uf); 2210 extern unsigned long do_mmap(struct file *file, unsigned long addr, 2211 unsigned long len, unsigned long prot, unsigned long flags, 2212 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate, 2213 struct list_head *uf); 2214 extern int do_munmap(struct mm_struct *, unsigned long, size_t, 2215 struct list_head *uf); 2216 2217 static inline unsigned long 2218 do_mmap_pgoff(struct file *file, unsigned long addr, 2219 unsigned long len, unsigned long prot, unsigned long flags, 2220 unsigned long pgoff, unsigned long *populate, 2221 struct list_head *uf) 2222 { 2223 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf); 2224 } 2225 2226 #ifdef CONFIG_MMU 2227 extern int __mm_populate(unsigned long addr, unsigned long len, 2228 int ignore_errors); 2229 static inline void mm_populate(unsigned long addr, unsigned long len) 2230 { 2231 /* Ignore errors */ 2232 (void) __mm_populate(addr, len, 1); 2233 } 2234 #else 2235 static inline void mm_populate(unsigned long addr, unsigned long len) {} 2236 #endif 2237 2238 /* These take the mm semaphore themselves */ 2239 extern int __must_check vm_brk(unsigned long, unsigned long); 2240 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long); 2241 extern int vm_munmap(unsigned long, size_t); 2242 extern unsigned long __must_check vm_mmap(struct file *, unsigned long, 2243 unsigned long, unsigned long, 2244 unsigned long, unsigned long); 2245 2246 struct vm_unmapped_area_info { 2247 #define VM_UNMAPPED_AREA_TOPDOWN 1 2248 unsigned long flags; 2249 unsigned long length; 2250 unsigned long low_limit; 2251 unsigned long high_limit; 2252 unsigned long align_mask; 2253 unsigned long align_offset; 2254 }; 2255 2256 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); 2257 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); 2258 2259 /* 2260 * Search for an unmapped address range. 2261 * 2262 * We are looking for a range that: 2263 * - does not intersect with any VMA; 2264 * - is contained within the [low_limit, high_limit) interval; 2265 * - is at least the desired size. 2266 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) 2267 */ 2268 static inline unsigned long 2269 vm_unmapped_area(struct vm_unmapped_area_info *info) 2270 { 2271 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN) 2272 return unmapped_area_topdown(info); 2273 else 2274 return unmapped_area(info); 2275 } 2276 2277 /* truncate.c */ 2278 extern void truncate_inode_pages(struct address_space *, loff_t); 2279 extern void truncate_inode_pages_range(struct address_space *, 2280 loff_t lstart, loff_t lend); 2281 extern void truncate_inode_pages_final(struct address_space *); 2282 2283 /* generic vm_area_ops exported for stackable file systems */ 2284 extern int filemap_fault(struct vm_fault *vmf); 2285 extern void filemap_map_pages(struct vm_fault *vmf, 2286 pgoff_t start_pgoff, pgoff_t end_pgoff); 2287 extern int filemap_page_mkwrite(struct vm_fault *vmf); 2288 2289 /* mm/page-writeback.c */ 2290 int __must_check write_one_page(struct page *page); 2291 void task_dirty_inc(struct task_struct *tsk); 2292 2293 /* readahead.c */ 2294 #define VM_MAX_READAHEAD 128 /* kbytes */ 2295 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 2296 2297 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 2298 pgoff_t offset, unsigned long nr_to_read); 2299 2300 void page_cache_sync_readahead(struct address_space *mapping, 2301 struct file_ra_state *ra, 2302 struct file *filp, 2303 pgoff_t offset, 2304 unsigned long size); 2305 2306 void page_cache_async_readahead(struct address_space *mapping, 2307 struct file_ra_state *ra, 2308 struct file *filp, 2309 struct page *pg, 2310 pgoff_t offset, 2311 unsigned long size); 2312 2313 extern unsigned long stack_guard_gap; 2314 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ 2315 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 2316 2317 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ 2318 extern int expand_downwards(struct vm_area_struct *vma, 2319 unsigned long address); 2320 #if VM_GROWSUP 2321 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 2322 #else 2323 #define expand_upwards(vma, address) (0) 2324 #endif 2325 2326 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 2327 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 2328 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 2329 struct vm_area_struct **pprev); 2330 2331 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 2332 NULL if none. Assume start_addr < end_addr. */ 2333 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 2334 { 2335 struct vm_area_struct * vma = find_vma(mm,start_addr); 2336 2337 if (vma && end_addr <= vma->vm_start) 2338 vma = NULL; 2339 return vma; 2340 } 2341 2342 static inline unsigned long vm_start_gap(struct vm_area_struct *vma) 2343 { 2344 unsigned long vm_start = vma->vm_start; 2345 2346 if (vma->vm_flags & VM_GROWSDOWN) { 2347 vm_start -= stack_guard_gap; 2348 if (vm_start > vma->vm_start) 2349 vm_start = 0; 2350 } 2351 return vm_start; 2352 } 2353 2354 static inline unsigned long vm_end_gap(struct vm_area_struct *vma) 2355 { 2356 unsigned long vm_end = vma->vm_end; 2357 2358 if (vma->vm_flags & VM_GROWSUP) { 2359 vm_end += stack_guard_gap; 2360 if (vm_end < vma->vm_end) 2361 vm_end = -PAGE_SIZE; 2362 } 2363 return vm_end; 2364 } 2365 2366 static inline unsigned long vma_pages(struct vm_area_struct *vma) 2367 { 2368 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 2369 } 2370 2371 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ 2372 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, 2373 unsigned long vm_start, unsigned long vm_end) 2374 { 2375 struct vm_area_struct *vma = find_vma(mm, vm_start); 2376 2377 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) 2378 vma = NULL; 2379 2380 return vma; 2381 } 2382 2383 #ifdef CONFIG_MMU 2384 pgprot_t vm_get_page_prot(unsigned long vm_flags); 2385 void vma_set_page_prot(struct vm_area_struct *vma); 2386 #else 2387 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) 2388 { 2389 return __pgprot(0); 2390 } 2391 static inline void vma_set_page_prot(struct vm_area_struct *vma) 2392 { 2393 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); 2394 } 2395 #endif 2396 2397 #ifdef CONFIG_NUMA_BALANCING 2398 unsigned long change_prot_numa(struct vm_area_struct *vma, 2399 unsigned long start, unsigned long end); 2400 #endif 2401 2402 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 2403 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 2404 unsigned long pfn, unsigned long size, pgprot_t); 2405 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 2406 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 2407 unsigned long pfn); 2408 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, 2409 unsigned long pfn, pgprot_t pgprot); 2410 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 2411 pfn_t pfn); 2412 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr, 2413 pfn_t pfn); 2414 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); 2415 2416 2417 struct page *follow_page_mask(struct vm_area_struct *vma, 2418 unsigned long address, unsigned int foll_flags, 2419 unsigned int *page_mask); 2420 2421 static inline struct page *follow_page(struct vm_area_struct *vma, 2422 unsigned long address, unsigned int foll_flags) 2423 { 2424 unsigned int unused_page_mask; 2425 return follow_page_mask(vma, address, foll_flags, &unused_page_mask); 2426 } 2427 2428 #define FOLL_WRITE 0x01 /* check pte is writable */ 2429 #define FOLL_TOUCH 0x02 /* mark page accessed */ 2430 #define FOLL_GET 0x04 /* do get_page on page */ 2431 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 2432 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 2433 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO 2434 * and return without waiting upon it */ 2435 #define FOLL_POPULATE 0x40 /* fault in page */ 2436 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ 2437 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ 2438 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ 2439 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ 2440 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ 2441 #define FOLL_MLOCK 0x1000 /* lock present pages */ 2442 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ 2443 #define FOLL_COW 0x4000 /* internal GUP flag */ 2444 #define FOLL_ANON 0x8000 /* don't do file mappings */ 2445 2446 static inline int vm_fault_to_errno(int vm_fault, int foll_flags) 2447 { 2448 if (vm_fault & VM_FAULT_OOM) 2449 return -ENOMEM; 2450 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) 2451 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT; 2452 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV)) 2453 return -EFAULT; 2454 return 0; 2455 } 2456 2457 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 2458 void *data); 2459 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 2460 unsigned long size, pte_fn_t fn, void *data); 2461 2462 2463 #ifdef CONFIG_PAGE_POISONING 2464 extern bool page_poisoning_enabled(void); 2465 extern void kernel_poison_pages(struct page *page, int numpages, int enable); 2466 extern bool page_is_poisoned(struct page *page); 2467 #else 2468 static inline bool page_poisoning_enabled(void) { return false; } 2469 static inline void kernel_poison_pages(struct page *page, int numpages, 2470 int enable) { } 2471 static inline bool page_is_poisoned(struct page *page) { return false; } 2472 #endif 2473 2474 #ifdef CONFIG_DEBUG_PAGEALLOC 2475 extern bool _debug_pagealloc_enabled; 2476 extern void __kernel_map_pages(struct page *page, int numpages, int enable); 2477 2478 static inline bool debug_pagealloc_enabled(void) 2479 { 2480 return _debug_pagealloc_enabled; 2481 } 2482 2483 static inline void 2484 kernel_map_pages(struct page *page, int numpages, int enable) 2485 { 2486 if (!debug_pagealloc_enabled()) 2487 return; 2488 2489 __kernel_map_pages(page, numpages, enable); 2490 } 2491 #ifdef CONFIG_HIBERNATION 2492 extern bool kernel_page_present(struct page *page); 2493 #endif /* CONFIG_HIBERNATION */ 2494 #else /* CONFIG_DEBUG_PAGEALLOC */ 2495 static inline void 2496 kernel_map_pages(struct page *page, int numpages, int enable) {} 2497 #ifdef CONFIG_HIBERNATION 2498 static inline bool kernel_page_present(struct page *page) { return true; } 2499 #endif /* CONFIG_HIBERNATION */ 2500 static inline bool debug_pagealloc_enabled(void) 2501 { 2502 return false; 2503 } 2504 #endif /* CONFIG_DEBUG_PAGEALLOC */ 2505 2506 #ifdef __HAVE_ARCH_GATE_AREA 2507 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); 2508 extern int in_gate_area_no_mm(unsigned long addr); 2509 extern int in_gate_area(struct mm_struct *mm, unsigned long addr); 2510 #else 2511 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) 2512 { 2513 return NULL; 2514 } 2515 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } 2516 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) 2517 { 2518 return 0; 2519 } 2520 #endif /* __HAVE_ARCH_GATE_AREA */ 2521 2522 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); 2523 2524 #ifdef CONFIG_SYSCTL 2525 extern int sysctl_drop_caches; 2526 int drop_caches_sysctl_handler(struct ctl_table *, int, 2527 void __user *, size_t *, loff_t *); 2528 #endif 2529 2530 void drop_slab(void); 2531 void drop_slab_node(int nid); 2532 2533 #ifndef CONFIG_MMU 2534 #define randomize_va_space 0 2535 #else 2536 extern int randomize_va_space; 2537 #endif 2538 2539 const char * arch_vma_name(struct vm_area_struct *vma); 2540 void print_vma_addr(char *prefix, unsigned long rip); 2541 2542 void sparse_mem_maps_populate_node(struct page **map_map, 2543 unsigned long pnum_begin, 2544 unsigned long pnum_end, 2545 unsigned long map_count, 2546 int nodeid); 2547 2548 struct page *sparse_mem_map_populate(unsigned long pnum, int nid, 2549 struct vmem_altmap *altmap); 2550 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 2551 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node); 2552 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node); 2553 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 2554 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 2555 void *vmemmap_alloc_block(unsigned long size, int node); 2556 struct vmem_altmap; 2557 void *vmemmap_alloc_block_buf(unsigned long size, int node); 2558 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap); 2559 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 2560 int vmemmap_populate_basepages(unsigned long start, unsigned long end, 2561 int node); 2562 int vmemmap_populate(unsigned long start, unsigned long end, int node, 2563 struct vmem_altmap *altmap); 2564 void vmemmap_populate_print_last(void); 2565 #ifdef CONFIG_MEMORY_HOTPLUG 2566 void vmemmap_free(unsigned long start, unsigned long end, 2567 struct vmem_altmap *altmap); 2568 #endif 2569 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, 2570 unsigned long nr_pages); 2571 2572 enum mf_flags { 2573 MF_COUNT_INCREASED = 1 << 0, 2574 MF_ACTION_REQUIRED = 1 << 1, 2575 MF_MUST_KILL = 1 << 2, 2576 MF_SOFT_OFFLINE = 1 << 3, 2577 }; 2578 extern int memory_failure(unsigned long pfn, int flags); 2579 extern void memory_failure_queue(unsigned long pfn, int flags); 2580 extern int unpoison_memory(unsigned long pfn); 2581 extern int get_hwpoison_page(struct page *page); 2582 #define put_hwpoison_page(page) put_page(page) 2583 extern int sysctl_memory_failure_early_kill; 2584 extern int sysctl_memory_failure_recovery; 2585 extern void shake_page(struct page *p, int access); 2586 extern atomic_long_t num_poisoned_pages; 2587 extern int soft_offline_page(struct page *page, int flags); 2588 2589 2590 /* 2591 * Error handlers for various types of pages. 2592 */ 2593 enum mf_result { 2594 MF_IGNORED, /* Error: cannot be handled */ 2595 MF_FAILED, /* Error: handling failed */ 2596 MF_DELAYED, /* Will be handled later */ 2597 MF_RECOVERED, /* Successfully recovered */ 2598 }; 2599 2600 enum mf_action_page_type { 2601 MF_MSG_KERNEL, 2602 MF_MSG_KERNEL_HIGH_ORDER, 2603 MF_MSG_SLAB, 2604 MF_MSG_DIFFERENT_COMPOUND, 2605 MF_MSG_POISONED_HUGE, 2606 MF_MSG_HUGE, 2607 MF_MSG_FREE_HUGE, 2608 MF_MSG_NON_PMD_HUGE, 2609 MF_MSG_UNMAP_FAILED, 2610 MF_MSG_DIRTY_SWAPCACHE, 2611 MF_MSG_CLEAN_SWAPCACHE, 2612 MF_MSG_DIRTY_MLOCKED_LRU, 2613 MF_MSG_CLEAN_MLOCKED_LRU, 2614 MF_MSG_DIRTY_UNEVICTABLE_LRU, 2615 MF_MSG_CLEAN_UNEVICTABLE_LRU, 2616 MF_MSG_DIRTY_LRU, 2617 MF_MSG_CLEAN_LRU, 2618 MF_MSG_TRUNCATED_LRU, 2619 MF_MSG_BUDDY, 2620 MF_MSG_BUDDY_2ND, 2621 MF_MSG_UNKNOWN, 2622 }; 2623 2624 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) 2625 extern void clear_huge_page(struct page *page, 2626 unsigned long addr_hint, 2627 unsigned int pages_per_huge_page); 2628 extern void copy_user_huge_page(struct page *dst, struct page *src, 2629 unsigned long addr, struct vm_area_struct *vma, 2630 unsigned int pages_per_huge_page); 2631 extern long copy_huge_page_from_user(struct page *dst_page, 2632 const void __user *usr_src, 2633 unsigned int pages_per_huge_page, 2634 bool allow_pagefault); 2635 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ 2636 2637 extern struct page_ext_operations debug_guardpage_ops; 2638 2639 #ifdef CONFIG_DEBUG_PAGEALLOC 2640 extern unsigned int _debug_guardpage_minorder; 2641 extern bool _debug_guardpage_enabled; 2642 2643 static inline unsigned int debug_guardpage_minorder(void) 2644 { 2645 return _debug_guardpage_minorder; 2646 } 2647 2648 static inline bool debug_guardpage_enabled(void) 2649 { 2650 return _debug_guardpage_enabled; 2651 } 2652 2653 static inline bool page_is_guard(struct page *page) 2654 { 2655 struct page_ext *page_ext; 2656 2657 if (!debug_guardpage_enabled()) 2658 return false; 2659 2660 page_ext = lookup_page_ext(page); 2661 if (unlikely(!page_ext)) 2662 return false; 2663 2664 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags); 2665 } 2666 #else 2667 static inline unsigned int debug_guardpage_minorder(void) { return 0; } 2668 static inline bool debug_guardpage_enabled(void) { return false; } 2669 static inline bool page_is_guard(struct page *page) { return false; } 2670 #endif /* CONFIG_DEBUG_PAGEALLOC */ 2671 2672 #if MAX_NUMNODES > 1 2673 void __init setup_nr_node_ids(void); 2674 #else 2675 static inline void setup_nr_node_ids(void) {} 2676 #endif 2677 2678 #endif /* __KERNEL__ */ 2679 #endif /* _LINUX_MM_H */ 2680
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.