1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_PAGEMAP_H 3 #define _LINUX_PAGEMAP_H 4 5 /* 6 * Copyright 1995 Linus Torvalds 7 */ 8 #include <linux/mm.h> 9 #include <linux/fs.h> 10 #include <linux/list.h> 11 #include <linux/highmem.h> 12 #include <linux/compiler.h> 13 #include <linux/uaccess.h> 14 #include <linux/gfp.h> 15 #include <linux/bitops.h> 16 #include <linux/hardirq.h> /* for in_interrupt() */ 17 #include <linux/hugetlb_inline.h> 18 19 struct folio_batch; 20 21 unsigned long invalidate_mapping_pages(struct address_space *mapping, 22 pgoff_t start, pgoff_t end); 23 24 static inline void invalidate_remote_inode(struct inode *inode) 25 { 26 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 27 S_ISLNK(inode->i_mode)) 28 invalidate_mapping_pages(inode->i_mapping, 0, -1); 29 } 30 int invalidate_inode_pages2(struct address_space *mapping); 31 int invalidate_inode_pages2_range(struct address_space *mapping, 32 pgoff_t start, pgoff_t end); 33 int kiocb_invalidate_pages(struct kiocb *iocb, size_t count); 34 void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count); 35 36 int write_inode_now(struct inode *, int sync); 37 int filemap_fdatawrite(struct address_space *); 38 int filemap_flush(struct address_space *); 39 int filemap_fdatawait_keep_errors(struct address_space *mapping); 40 int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend); 41 int filemap_fdatawait_range_keep_errors(struct address_space *mapping, 42 loff_t start_byte, loff_t end_byte); 43 int filemap_invalidate_inode(struct inode *inode, bool flush, 44 loff_t start, loff_t end); 45 46 static inline int filemap_fdatawait(struct address_space *mapping) 47 { 48 return filemap_fdatawait_range(mapping, 0, LLONG_MAX); 49 } 50 51 bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend); 52 int filemap_write_and_wait_range(struct address_space *mapping, 53 loff_t lstart, loff_t lend); 54 int __filemap_fdatawrite_range(struct address_space *mapping, 55 loff_t start, loff_t end, int sync_mode); 56 int filemap_fdatawrite_range(struct address_space *mapping, 57 loff_t start, loff_t end); 58 int filemap_check_errors(struct address_space *mapping); 59 void __filemap_set_wb_err(struct address_space *mapping, int err); 60 int filemap_fdatawrite_wbc(struct address_space *mapping, 61 struct writeback_control *wbc); 62 int kiocb_write_and_wait(struct kiocb *iocb, size_t count); 63 64 static inline int filemap_write_and_wait(struct address_space *mapping) 65 { 66 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX); 67 } 68 69 /** 70 * filemap_set_wb_err - set a writeback error on an address_space 71 * @mapping: mapping in which to set writeback error 72 * @err: error to be set in mapping 73 * 74 * When writeback fails in some way, we must record that error so that 75 * userspace can be informed when fsync and the like are called. We endeavor 76 * to report errors on any file that was open at the time of the error. Some 77 * internal callers also need to know when writeback errors have occurred. 78 * 79 * When a writeback error occurs, most filesystems will want to call 80 * filemap_set_wb_err to record the error in the mapping so that it will be 81 * automatically reported whenever fsync is called on the file. 82 */ 83 static inline void filemap_set_wb_err(struct address_space *mapping, int err) 84 { 85 /* Fastpath for common case of no error */ 86 if (unlikely(err)) 87 __filemap_set_wb_err(mapping, err); 88 } 89 90 /** 91 * filemap_check_wb_err - has an error occurred since the mark was sampled? 92 * @mapping: mapping to check for writeback errors 93 * @since: previously-sampled errseq_t 94 * 95 * Grab the errseq_t value from the mapping, and see if it has changed "since" 96 * the given value was sampled. 97 * 98 * If it has then report the latest error set, otherwise return 0. 99 */ 100 static inline int filemap_check_wb_err(struct address_space *mapping, 101 errseq_t since) 102 { 103 return errseq_check(&mapping->wb_err, since); 104 } 105 106 /** 107 * filemap_sample_wb_err - sample the current errseq_t to test for later errors 108 * @mapping: mapping to be sampled 109 * 110 * Writeback errors are always reported relative to a particular sample point 111 * in the past. This function provides those sample points. 112 */ 113 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping) 114 { 115 return errseq_sample(&mapping->wb_err); 116 } 117 118 /** 119 * file_sample_sb_err - sample the current errseq_t to test for later errors 120 * @file: file pointer to be sampled 121 * 122 * Grab the most current superblock-level errseq_t value for the given 123 * struct file. 124 */ 125 static inline errseq_t file_sample_sb_err(struct file *file) 126 { 127 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err); 128 } 129 130 /* 131 * Flush file data before changing attributes. Caller must hold any locks 132 * required to prevent further writes to this file until we're done setting 133 * flags. 134 */ 135 static inline int inode_drain_writes(struct inode *inode) 136 { 137 inode_dio_wait(inode); 138 return filemap_write_and_wait(inode->i_mapping); 139 } 140 141 static inline bool mapping_empty(struct address_space *mapping) 142 { 143 return xa_empty(&mapping->i_pages); 144 } 145 146 /* 147 * mapping_shrinkable - test if page cache state allows inode reclaim 148 * @mapping: the page cache mapping 149 * 150 * This checks the mapping's cache state for the pupose of inode 151 * reclaim and LRU management. 152 * 153 * The caller is expected to hold the i_lock, but is not required to 154 * hold the i_pages lock, which usually protects cache state. That's 155 * because the i_lock and the list_lru lock that protect the inode and 156 * its LRU state don't nest inside the irq-safe i_pages lock. 157 * 158 * Cache deletions are performed under the i_lock, which ensures that 159 * when an inode goes empty, it will reliably get queued on the LRU. 160 * 161 * Cache additions do not acquire the i_lock and may race with this 162 * check, in which case we'll report the inode as shrinkable when it 163 * has cache pages. This is okay: the shrinker also checks the 164 * refcount and the referenced bit, which will be elevated or set in 165 * the process of adding new cache pages to an inode. 166 */ 167 static inline bool mapping_shrinkable(struct address_space *mapping) 168 { 169 void *head; 170 171 /* 172 * On highmem systems, there could be lowmem pressure from the 173 * inodes before there is highmem pressure from the page 174 * cache. Make inodes shrinkable regardless of cache state. 175 */ 176 if (IS_ENABLED(CONFIG_HIGHMEM)) 177 return true; 178 179 /* Cache completely empty? Shrink away. */ 180 head = rcu_access_pointer(mapping->i_pages.xa_head); 181 if (!head) 182 return true; 183 184 /* 185 * The xarray stores single offset-0 entries directly in the 186 * head pointer, which allows non-resident page cache entries 187 * to escape the shadow shrinker's list of xarray nodes. The 188 * inode shrinker needs to pick them up under memory pressure. 189 */ 190 if (!xa_is_node(head) && xa_is_value(head)) 191 return true; 192 193 return false; 194 } 195 196 /* 197 * Bits in mapping->flags. 198 */ 199 enum mapping_flags { 200 AS_EIO = 0, /* IO error on async write */ 201 AS_ENOSPC = 1, /* ENOSPC on async write */ 202 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ 203 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ 204 AS_EXITING = 4, /* final truncate in progress */ 205 /* writeback related tags are not used */ 206 AS_NO_WRITEBACK_TAGS = 5, 207 AS_LARGE_FOLIO_SUPPORT = 6, 208 AS_RELEASE_ALWAYS, /* Call ->release_folio(), even if no private data */ 209 AS_STABLE_WRITES, /* must wait for writeback before modifying 210 folio contents */ 211 AS_UNMOVABLE, /* The mapping cannot be moved, ever */ 212 }; 213 214 /** 215 * mapping_set_error - record a writeback error in the address_space 216 * @mapping: the mapping in which an error should be set 217 * @error: the error to set in the mapping 218 * 219 * When writeback fails in some way, we must record that error so that 220 * userspace can be informed when fsync and the like are called. We endeavor 221 * to report errors on any file that was open at the time of the error. Some 222 * internal callers also need to know when writeback errors have occurred. 223 * 224 * When a writeback error occurs, most filesystems will want to call 225 * mapping_set_error to record the error in the mapping so that it can be 226 * reported when the application calls fsync(2). 227 */ 228 static inline void mapping_set_error(struct address_space *mapping, int error) 229 { 230 if (likely(!error)) 231 return; 232 233 /* Record in wb_err for checkers using errseq_t based tracking */ 234 __filemap_set_wb_err(mapping, error); 235 236 /* Record it in superblock */ 237 if (mapping->host) 238 errseq_set(&mapping->host->i_sb->s_wb_err, error); 239 240 /* Record it in flags for now, for legacy callers */ 241 if (error == -ENOSPC) 242 set_bit(AS_ENOSPC, &mapping->flags); 243 else 244 set_bit(AS_EIO, &mapping->flags); 245 } 246 247 static inline void mapping_set_unevictable(struct address_space *mapping) 248 { 249 set_bit(AS_UNEVICTABLE, &mapping->flags); 250 } 251 252 static inline void mapping_clear_unevictable(struct address_space *mapping) 253 { 254 clear_bit(AS_UNEVICTABLE, &mapping->flags); 255 } 256 257 static inline bool mapping_unevictable(struct address_space *mapping) 258 { 259 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags); 260 } 261 262 static inline void mapping_set_exiting(struct address_space *mapping) 263 { 264 set_bit(AS_EXITING, &mapping->flags); 265 } 266 267 static inline int mapping_exiting(struct address_space *mapping) 268 { 269 return test_bit(AS_EXITING, &mapping->flags); 270 } 271 272 static inline void mapping_set_no_writeback_tags(struct address_space *mapping) 273 { 274 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); 275 } 276 277 static inline int mapping_use_writeback_tags(struct address_space *mapping) 278 { 279 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); 280 } 281 282 static inline bool mapping_release_always(const struct address_space *mapping) 283 { 284 return test_bit(AS_RELEASE_ALWAYS, &mapping->flags); 285 } 286 287 static inline void mapping_set_release_always(struct address_space *mapping) 288 { 289 set_bit(AS_RELEASE_ALWAYS, &mapping->flags); 290 } 291 292 static inline void mapping_clear_release_always(struct address_space *mapping) 293 { 294 clear_bit(AS_RELEASE_ALWAYS, &mapping->flags); 295 } 296 297 static inline bool mapping_stable_writes(const struct address_space *mapping) 298 { 299 return test_bit(AS_STABLE_WRITES, &mapping->flags); 300 } 301 302 static inline void mapping_set_stable_writes(struct address_space *mapping) 303 { 304 set_bit(AS_STABLE_WRITES, &mapping->flags); 305 } 306 307 static inline void mapping_clear_stable_writes(struct address_space *mapping) 308 { 309 clear_bit(AS_STABLE_WRITES, &mapping->flags); 310 } 311 312 static inline void mapping_set_unmovable(struct address_space *mapping) 313 { 314 /* 315 * It's expected unmovable mappings are also unevictable. Compaction 316 * migrate scanner (isolate_migratepages_block()) relies on this to 317 * reduce page locking. 318 */ 319 set_bit(AS_UNEVICTABLE, &mapping->flags); 320 set_bit(AS_UNMOVABLE, &mapping->flags); 321 } 322 323 static inline bool mapping_unmovable(struct address_space *mapping) 324 { 325 return test_bit(AS_UNMOVABLE, &mapping->flags); 326 } 327 328 static inline gfp_t mapping_gfp_mask(struct address_space * mapping) 329 { 330 return mapping->gfp_mask; 331 } 332 333 /* Restricts the given gfp_mask to what the mapping allows. */ 334 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, 335 gfp_t gfp_mask) 336 { 337 return mapping_gfp_mask(mapping) & gfp_mask; 338 } 339 340 /* 341 * This is non-atomic. Only to be used before the mapping is activated. 342 * Probably needs a barrier... 343 */ 344 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) 345 { 346 m->gfp_mask = mask; 347 } 348 349 /** 350 * mapping_set_large_folios() - Indicate the file supports large folios. 351 * @mapping: The file. 352 * 353 * The filesystem should call this function in its inode constructor to 354 * indicate that the VFS can use large folios to cache the contents of 355 * the file. 356 * 357 * Context: This should not be called while the inode is active as it 358 * is non-atomic. 359 */ 360 static inline void mapping_set_large_folios(struct address_space *mapping) 361 { 362 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); 363 } 364 365 /* 366 * Large folio support currently depends on THP. These dependencies are 367 * being worked on but are not yet fixed. 368 */ 369 static inline bool mapping_large_folio_support(struct address_space *mapping) 370 { 371 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) && 372 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); 373 } 374 375 static inline int filemap_nr_thps(struct address_space *mapping) 376 { 377 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 378 return atomic_read(&mapping->nr_thps); 379 #else 380 return 0; 381 #endif 382 } 383 384 static inline void filemap_nr_thps_inc(struct address_space *mapping) 385 { 386 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 387 if (!mapping_large_folio_support(mapping)) 388 atomic_inc(&mapping->nr_thps); 389 #else 390 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); 391 #endif 392 } 393 394 static inline void filemap_nr_thps_dec(struct address_space *mapping) 395 { 396 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 397 if (!mapping_large_folio_support(mapping)) 398 atomic_dec(&mapping->nr_thps); 399 #else 400 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); 401 #endif 402 } 403 404 struct address_space *page_mapping(struct page *); 405 struct address_space *folio_mapping(struct folio *); 406 struct address_space *swapcache_mapping(struct folio *); 407 408 /** 409 * folio_file_mapping - Find the mapping this folio belongs to. 410 * @folio: The folio. 411 * 412 * For folios which are in the page cache, return the mapping that this 413 * page belongs to. Folios in the swap cache return the mapping of the 414 * swap file or swap device where the data is stored. This is different 415 * from the mapping returned by folio_mapping(). The only reason to 416 * use it is if, like NFS, you return 0 from ->activate_swapfile. 417 * 418 * Do not call this for folios which aren't in the page cache or swap cache. 419 */ 420 static inline struct address_space *folio_file_mapping(struct folio *folio) 421 { 422 if (unlikely(folio_test_swapcache(folio))) 423 return swapcache_mapping(folio); 424 425 return folio->mapping; 426 } 427 428 /** 429 * folio_flush_mapping - Find the file mapping this folio belongs to. 430 * @folio: The folio. 431 * 432 * For folios which are in the page cache, return the mapping that this 433 * page belongs to. Anonymous folios return NULL, even if they're in 434 * the swap cache. Other kinds of folio also return NULL. 435 * 436 * This is ONLY used by architecture cache flushing code. If you aren't 437 * writing cache flushing code, you want either folio_mapping() or 438 * folio_file_mapping(). 439 */ 440 static inline struct address_space *folio_flush_mapping(struct folio *folio) 441 { 442 if (unlikely(folio_test_swapcache(folio))) 443 return NULL; 444 445 return folio_mapping(folio); 446 } 447 448 static inline struct address_space *page_file_mapping(struct page *page) 449 { 450 return folio_file_mapping(page_folio(page)); 451 } 452 453 /** 454 * folio_inode - Get the host inode for this folio. 455 * @folio: The folio. 456 * 457 * For folios which are in the page cache, return the inode that this folio 458 * belongs to. 459 * 460 * Do not call this for folios which aren't in the page cache. 461 */ 462 static inline struct inode *folio_inode(struct folio *folio) 463 { 464 return folio->mapping->host; 465 } 466 467 /** 468 * folio_attach_private - Attach private data to a folio. 469 * @folio: Folio to attach data to. 470 * @data: Data to attach to folio. 471 * 472 * Attaching private data to a folio increments the page's reference count. 473 * The data must be detached before the folio will be freed. 474 */ 475 static inline void folio_attach_private(struct folio *folio, void *data) 476 { 477 folio_get(folio); 478 folio->private = data; 479 folio_set_private(folio); 480 } 481 482 /** 483 * folio_change_private - Change private data on a folio. 484 * @folio: Folio to change the data on. 485 * @data: Data to set on the folio. 486 * 487 * Change the private data attached to a folio and return the old 488 * data. The page must previously have had data attached and the data 489 * must be detached before the folio will be freed. 490 * 491 * Return: Data that was previously attached to the folio. 492 */ 493 static inline void *folio_change_private(struct folio *folio, void *data) 494 { 495 void *old = folio_get_private(folio); 496 497 folio->private = data; 498 return old; 499 } 500 501 /** 502 * folio_detach_private - Detach private data from a folio. 503 * @folio: Folio to detach data from. 504 * 505 * Removes the data that was previously attached to the folio and decrements 506 * the refcount on the page. 507 * 508 * Return: Data that was attached to the folio. 509 */ 510 static inline void *folio_detach_private(struct folio *folio) 511 { 512 void *data = folio_get_private(folio); 513 514 if (!folio_test_private(folio)) 515 return NULL; 516 folio_clear_private(folio); 517 folio->private = NULL; 518 folio_put(folio); 519 520 return data; 521 } 522 523 static inline void attach_page_private(struct page *page, void *data) 524 { 525 folio_attach_private(page_folio(page), data); 526 } 527 528 static inline void *detach_page_private(struct page *page) 529 { 530 return folio_detach_private(page_folio(page)); 531 } 532 533 /* 534 * There are some parts of the kernel which assume that PMD entries 535 * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then, 536 * limit the maximum allocation order to PMD size. I'm not aware of any 537 * assumptions about maximum order if THP are disabled, but 8 seems like 538 * a good order (that's 1MB if you're using 4kB pages) 539 */ 540 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 541 #define MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER 542 #else 543 #define MAX_PAGECACHE_ORDER 8 544 #endif 545 546 #ifdef CONFIG_NUMA 547 struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order); 548 #else 549 static inline struct folio *filemap_alloc_folio_noprof(gfp_t gfp, unsigned int order) 550 { 551 return folio_alloc_noprof(gfp, order); 552 } 553 #endif 554 555 #define filemap_alloc_folio(...) \ 556 alloc_hooks(filemap_alloc_folio_noprof(__VA_ARGS__)) 557 558 static inline struct page *__page_cache_alloc(gfp_t gfp) 559 { 560 return &filemap_alloc_folio(gfp, 0)->page; 561 } 562 563 static inline gfp_t readahead_gfp_mask(struct address_space *x) 564 { 565 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN; 566 } 567 568 typedef int filler_t(struct file *, struct folio *); 569 570 pgoff_t page_cache_next_miss(struct address_space *mapping, 571 pgoff_t index, unsigned long max_scan); 572 pgoff_t page_cache_prev_miss(struct address_space *mapping, 573 pgoff_t index, unsigned long max_scan); 574 575 /** 576 * typedef fgf_t - Flags for getting folios from the page cache. 577 * 578 * Most users of the page cache will not need to use these flags; 579 * there are convenience functions such as filemap_get_folio() and 580 * filemap_lock_folio(). For users which need more control over exactly 581 * what is done with the folios, these flags to __filemap_get_folio() 582 * are available. 583 * 584 * * %FGP_ACCESSED - The folio will be marked accessed. 585 * * %FGP_LOCK - The folio is returned locked. 586 * * %FGP_CREAT - If no folio is present then a new folio is allocated, 587 * added to the page cache and the VM's LRU list. The folio is 588 * returned locked. 589 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the 590 * folio is already in cache. If the folio was allocated, unlock it 591 * before returning so the caller can do the same dance. 592 * * %FGP_WRITE - The folio will be written to by the caller. 593 * * %FGP_NOFS - __GFP_FS will get cleared in gfp. 594 * * %FGP_NOWAIT - Don't block on the folio lock. 595 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback) 596 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin() 597 * implementation. 598 */ 599 typedef unsigned int __bitwise fgf_t; 600 601 #define FGP_ACCESSED ((__force fgf_t)0x00000001) 602 #define FGP_LOCK ((__force fgf_t)0x00000002) 603 #define FGP_CREAT ((__force fgf_t)0x00000004) 604 #define FGP_WRITE ((__force fgf_t)0x00000008) 605 #define FGP_NOFS ((__force fgf_t)0x00000010) 606 #define FGP_NOWAIT ((__force fgf_t)0x00000020) 607 #define FGP_FOR_MMAP ((__force fgf_t)0x00000040) 608 #define FGP_STABLE ((__force fgf_t)0x00000080) 609 #define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */ 610 611 #define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE) 612 613 /** 614 * fgf_set_order - Encode a length in the fgf_t flags. 615 * @size: The suggested size of the folio to create. 616 * 617 * The caller of __filemap_get_folio() can use this to suggest a preferred 618 * size for the folio that is created. If there is already a folio at 619 * the index, it will be returned, no matter what its size. If a folio 620 * is freshly created, it may be of a different size than requested 621 * due to alignment constraints, memory pressure, or the presence of 622 * other folios at nearby indices. 623 */ 624 static inline fgf_t fgf_set_order(size_t size) 625 { 626 unsigned int shift = ilog2(size); 627 628 if (shift <= PAGE_SHIFT) 629 return 0; 630 return (__force fgf_t)((shift - PAGE_SHIFT) << 26); 631 } 632 633 void *filemap_get_entry(struct address_space *mapping, pgoff_t index); 634 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, 635 fgf_t fgp_flags, gfp_t gfp); 636 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index, 637 fgf_t fgp_flags, gfp_t gfp); 638 639 /** 640 * filemap_get_folio - Find and get a folio. 641 * @mapping: The address_space to search. 642 * @index: The page index. 643 * 644 * Looks up the page cache entry at @mapping & @index. If a folio is 645 * present, it is returned with an increased refcount. 646 * 647 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for 648 * this index. Will not return a shadow, swap or DAX entry. 649 */ 650 static inline struct folio *filemap_get_folio(struct address_space *mapping, 651 pgoff_t index) 652 { 653 return __filemap_get_folio(mapping, index, 0, 0); 654 } 655 656 /** 657 * filemap_lock_folio - Find and lock a folio. 658 * @mapping: The address_space to search. 659 * @index: The page index. 660 * 661 * Looks up the page cache entry at @mapping & @index. If a folio is 662 * present, it is returned locked with an increased refcount. 663 * 664 * Context: May sleep. 665 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for 666 * this index. Will not return a shadow, swap or DAX entry. 667 */ 668 static inline struct folio *filemap_lock_folio(struct address_space *mapping, 669 pgoff_t index) 670 { 671 return __filemap_get_folio(mapping, index, FGP_LOCK, 0); 672 } 673 674 /** 675 * filemap_grab_folio - grab a folio from the page cache 676 * @mapping: The address space to search 677 * @index: The page index 678 * 679 * Looks up the page cache entry at @mapping & @index. If no folio is found, 680 * a new folio is created. The folio is locked, marked as accessed, and 681 * returned. 682 * 683 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found 684 * and failed to create a folio. 685 */ 686 static inline struct folio *filemap_grab_folio(struct address_space *mapping, 687 pgoff_t index) 688 { 689 return __filemap_get_folio(mapping, index, 690 FGP_LOCK | FGP_ACCESSED | FGP_CREAT, 691 mapping_gfp_mask(mapping)); 692 } 693 694 /** 695 * find_get_page - find and get a page reference 696 * @mapping: the address_space to search 697 * @offset: the page index 698 * 699 * Looks up the page cache slot at @mapping & @offset. If there is a 700 * page cache page, it is returned with an increased refcount. 701 * 702 * Otherwise, %NULL is returned. 703 */ 704 static inline struct page *find_get_page(struct address_space *mapping, 705 pgoff_t offset) 706 { 707 return pagecache_get_page(mapping, offset, 0, 0); 708 } 709 710 static inline struct page *find_get_page_flags(struct address_space *mapping, 711 pgoff_t offset, fgf_t fgp_flags) 712 { 713 return pagecache_get_page(mapping, offset, fgp_flags, 0); 714 } 715 716 /** 717 * find_lock_page - locate, pin and lock a pagecache page 718 * @mapping: the address_space to search 719 * @index: the page index 720 * 721 * Looks up the page cache entry at @mapping & @index. If there is a 722 * page cache page, it is returned locked and with an increased 723 * refcount. 724 * 725 * Context: May sleep. 726 * Return: A struct page or %NULL if there is no page in the cache for this 727 * index. 728 */ 729 static inline struct page *find_lock_page(struct address_space *mapping, 730 pgoff_t index) 731 { 732 return pagecache_get_page(mapping, index, FGP_LOCK, 0); 733 } 734 735 /** 736 * find_or_create_page - locate or add a pagecache page 737 * @mapping: the page's address_space 738 * @index: the page's index into the mapping 739 * @gfp_mask: page allocation mode 740 * 741 * Looks up the page cache slot at @mapping & @offset. If there is a 742 * page cache page, it is returned locked and with an increased 743 * refcount. 744 * 745 * If the page is not present, a new page is allocated using @gfp_mask 746 * and added to the page cache and the VM's LRU list. The page is 747 * returned locked and with an increased refcount. 748 * 749 * On memory exhaustion, %NULL is returned. 750 * 751 * find_or_create_page() may sleep, even if @gfp_flags specifies an 752 * atomic allocation! 753 */ 754 static inline struct page *find_or_create_page(struct address_space *mapping, 755 pgoff_t index, gfp_t gfp_mask) 756 { 757 return pagecache_get_page(mapping, index, 758 FGP_LOCK|FGP_ACCESSED|FGP_CREAT, 759 gfp_mask); 760 } 761 762 /** 763 * grab_cache_page_nowait - returns locked page at given index in given cache 764 * @mapping: target address_space 765 * @index: the page index 766 * 767 * Same as grab_cache_page(), but do not wait if the page is unavailable. 768 * This is intended for speculative data generators, where the data can 769 * be regenerated if the page couldn't be grabbed. This routine should 770 * be safe to call while holding the lock for another page. 771 * 772 * Clear __GFP_FS when allocating the page to avoid recursion into the fs 773 * and deadlock against the caller's locked page. 774 */ 775 static inline struct page *grab_cache_page_nowait(struct address_space *mapping, 776 pgoff_t index) 777 { 778 return pagecache_get_page(mapping, index, 779 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, 780 mapping_gfp_mask(mapping)); 781 } 782 783 #define swapcache_index(folio) __page_file_index(&(folio)->page) 784 785 /** 786 * folio_index - File index of a folio. 787 * @folio: The folio. 788 * 789 * For a folio which is either in the page cache or the swap cache, 790 * return its index within the address_space it belongs to. If you know 791 * the page is definitely in the page cache, you can look at the folio's 792 * index directly. 793 * 794 * Return: The index (offset in units of pages) of a folio in its file. 795 */ 796 static inline pgoff_t folio_index(struct folio *folio) 797 { 798 if (unlikely(folio_test_swapcache(folio))) 799 return swapcache_index(folio); 800 return folio->index; 801 } 802 803 /** 804 * folio_next_index - Get the index of the next folio. 805 * @folio: The current folio. 806 * 807 * Return: The index of the folio which follows this folio in the file. 808 */ 809 static inline pgoff_t folio_next_index(struct folio *folio) 810 { 811 return folio->index + folio_nr_pages(folio); 812 } 813 814 /** 815 * folio_file_page - The page for a particular index. 816 * @folio: The folio which contains this index. 817 * @index: The index we want to look up. 818 * 819 * Sometimes after looking up a folio in the page cache, we need to 820 * obtain the specific page for an index (eg a page fault). 821 * 822 * Return: The page containing the file data for this index. 823 */ 824 static inline struct page *folio_file_page(struct folio *folio, pgoff_t index) 825 { 826 return folio_page(folio, index & (folio_nr_pages(folio) - 1)); 827 } 828 829 /** 830 * folio_contains - Does this folio contain this index? 831 * @folio: The folio. 832 * @index: The page index within the file. 833 * 834 * Context: The caller should have the page locked in order to prevent 835 * (eg) shmem from moving the page between the page cache and swap cache 836 * and changing its index in the middle of the operation. 837 * Return: true or false. 838 */ 839 static inline bool folio_contains(struct folio *folio, pgoff_t index) 840 { 841 return index - folio_index(folio) < folio_nr_pages(folio); 842 } 843 844 /* 845 * Given the page we found in the page cache, return the page corresponding 846 * to this index in the file 847 */ 848 static inline struct page *find_subpage(struct page *head, pgoff_t index) 849 { 850 /* HugeTLBfs wants the head page regardless */ 851 if (PageHuge(head)) 852 return head; 853 854 return head + (index & (thp_nr_pages(head) - 1)); 855 } 856 857 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start, 858 pgoff_t end, struct folio_batch *fbatch); 859 unsigned filemap_get_folios_contig(struct address_space *mapping, 860 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch); 861 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start, 862 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch); 863 864 struct page *grab_cache_page_write_begin(struct address_space *mapping, 865 pgoff_t index); 866 867 /* 868 * Returns locked page at given index in given cache, creating it if needed. 869 */ 870 static inline struct page *grab_cache_page(struct address_space *mapping, 871 pgoff_t index) 872 { 873 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); 874 } 875 876 struct folio *read_cache_folio(struct address_space *, pgoff_t index, 877 filler_t *filler, struct file *file); 878 struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index, 879 gfp_t flags); 880 struct page *read_cache_page(struct address_space *, pgoff_t index, 881 filler_t *filler, struct file *file); 882 extern struct page * read_cache_page_gfp(struct address_space *mapping, 883 pgoff_t index, gfp_t gfp_mask); 884 885 static inline struct page *read_mapping_page(struct address_space *mapping, 886 pgoff_t index, struct file *file) 887 { 888 return read_cache_page(mapping, index, NULL, file); 889 } 890 891 static inline struct folio *read_mapping_folio(struct address_space *mapping, 892 pgoff_t index, struct file *file) 893 { 894 return read_cache_folio(mapping, index, NULL, file); 895 } 896 897 /* 898 * Get the offset in PAGE_SIZE (even for hugetlb pages). 899 */ 900 static inline pgoff_t page_to_pgoff(struct page *page) 901 { 902 struct page *head; 903 904 if (likely(!PageTransTail(page))) 905 return page->index; 906 907 head = compound_head(page); 908 /* 909 * We don't initialize ->index for tail pages: calculate based on 910 * head page 911 */ 912 return head->index + page - head; 913 } 914 915 /* 916 * Return byte-offset into filesystem object for page. 917 */ 918 static inline loff_t page_offset(struct page *page) 919 { 920 return ((loff_t)page->index) << PAGE_SHIFT; 921 } 922 923 static inline loff_t page_file_offset(struct page *page) 924 { 925 return ((loff_t)page_index(page)) << PAGE_SHIFT; 926 } 927 928 /** 929 * folio_pos - Returns the byte position of this folio in its file. 930 * @folio: The folio. 931 */ 932 static inline loff_t folio_pos(struct folio *folio) 933 { 934 return page_offset(&folio->page); 935 } 936 937 /** 938 * folio_file_pos - Returns the byte position of this folio in its file. 939 * @folio: The folio. 940 * 941 * This differs from folio_pos() for folios which belong to a swap file. 942 * NFS is the only filesystem today which needs to use folio_file_pos(). 943 */ 944 static inline loff_t folio_file_pos(struct folio *folio) 945 { 946 return page_file_offset(&folio->page); 947 } 948 949 /* 950 * Get the offset in PAGE_SIZE (even for hugetlb folios). 951 */ 952 static inline pgoff_t folio_pgoff(struct folio *folio) 953 { 954 return folio->index; 955 } 956 957 static inline pgoff_t linear_page_index(struct vm_area_struct *vma, 958 unsigned long address) 959 { 960 pgoff_t pgoff; 961 pgoff = (address - vma->vm_start) >> PAGE_SHIFT; 962 pgoff += vma->vm_pgoff; 963 return pgoff; 964 } 965 966 struct wait_page_key { 967 struct folio *folio; 968 int bit_nr; 969 int page_match; 970 }; 971 972 struct wait_page_queue { 973 struct folio *folio; 974 int bit_nr; 975 wait_queue_entry_t wait; 976 }; 977 978 static inline bool wake_page_match(struct wait_page_queue *wait_page, 979 struct wait_page_key *key) 980 { 981 if (wait_page->folio != key->folio) 982 return false; 983 key->page_match = 1; 984 985 if (wait_page->bit_nr != key->bit_nr) 986 return false; 987 988 return true; 989 } 990 991 void __folio_lock(struct folio *folio); 992 int __folio_lock_killable(struct folio *folio); 993 vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf); 994 void unlock_page(struct page *page); 995 void folio_unlock(struct folio *folio); 996 997 /** 998 * folio_trylock() - Attempt to lock a folio. 999 * @folio: The folio to attempt to lock. 1000 * 1001 * Sometimes it is undesirable to wait for a folio to be unlocked (eg 1002 * when the locks are being taken in the wrong order, or if making 1003 * progress through a batch of folios is more important than processing 1004 * them in order). Usually folio_lock() is the correct function to call. 1005 * 1006 * Context: Any context. 1007 * Return: Whether the lock was successfully acquired. 1008 */ 1009 static inline bool folio_trylock(struct folio *folio) 1010 { 1011 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0))); 1012 } 1013 1014 /* 1015 * Return true if the page was successfully locked 1016 */ 1017 static inline bool trylock_page(struct page *page) 1018 { 1019 return folio_trylock(page_folio(page)); 1020 } 1021 1022 /** 1023 * folio_lock() - Lock this folio. 1024 * @folio: The folio to lock. 1025 * 1026 * The folio lock protects against many things, probably more than it 1027 * should. It is primarily held while a folio is being brought uptodate, 1028 * either from its backing file or from swap. It is also held while a 1029 * folio is being truncated from its address_space, so holding the lock 1030 * is sufficient to keep folio->mapping stable. 1031 * 1032 * The folio lock is also held while write() is modifying the page to 1033 * provide POSIX atomicity guarantees (as long as the write does not 1034 * cross a page boundary). Other modifications to the data in the folio 1035 * do not hold the folio lock and can race with writes, eg DMA and stores 1036 * to mapped pages. 1037 * 1038 * Context: May sleep. If you need to acquire the locks of two or 1039 * more folios, they must be in order of ascending index, if they are 1040 * in the same address_space. If they are in different address_spaces, 1041 * acquire the lock of the folio which belongs to the address_space which 1042 * has the lowest address in memory first. 1043 */ 1044 static inline void folio_lock(struct folio *folio) 1045 { 1046 might_sleep(); 1047 if (!folio_trylock(folio)) 1048 __folio_lock(folio); 1049 } 1050 1051 /** 1052 * lock_page() - Lock the folio containing this page. 1053 * @page: The page to lock. 1054 * 1055 * See folio_lock() for a description of what the lock protects. 1056 * This is a legacy function and new code should probably use folio_lock() 1057 * instead. 1058 * 1059 * Context: May sleep. Pages in the same folio share a lock, so do not 1060 * attempt to lock two pages which share a folio. 1061 */ 1062 static inline void lock_page(struct page *page) 1063 { 1064 struct folio *folio; 1065 might_sleep(); 1066 1067 folio = page_folio(page); 1068 if (!folio_trylock(folio)) 1069 __folio_lock(folio); 1070 } 1071 1072 /** 1073 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal. 1074 * @folio: The folio to lock. 1075 * 1076 * Attempts to lock the folio, like folio_lock(), except that the sleep 1077 * to acquire the lock is interruptible by a fatal signal. 1078 * 1079 * Context: May sleep; see folio_lock(). 1080 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received. 1081 */ 1082 static inline int folio_lock_killable(struct folio *folio) 1083 { 1084 might_sleep(); 1085 if (!folio_trylock(folio)) 1086 return __folio_lock_killable(folio); 1087 return 0; 1088 } 1089 1090 /* 1091 * folio_lock_or_retry - Lock the folio, unless this would block and the 1092 * caller indicated that it can handle a retry. 1093 * 1094 * Return value and mmap_lock implications depend on flags; see 1095 * __folio_lock_or_retry(). 1096 */ 1097 static inline vm_fault_t folio_lock_or_retry(struct folio *folio, 1098 struct vm_fault *vmf) 1099 { 1100 might_sleep(); 1101 if (!folio_trylock(folio)) 1102 return __folio_lock_or_retry(folio, vmf); 1103 return 0; 1104 } 1105 1106 /* 1107 * This is exported only for folio_wait_locked/folio_wait_writeback, etc., 1108 * and should not be used directly. 1109 */ 1110 void folio_wait_bit(struct folio *folio, int bit_nr); 1111 int folio_wait_bit_killable(struct folio *folio, int bit_nr); 1112 1113 /* 1114 * Wait for a folio to be unlocked. 1115 * 1116 * This must be called with the caller "holding" the folio, 1117 * ie with increased folio reference count so that the folio won't 1118 * go away during the wait. 1119 */ 1120 static inline void folio_wait_locked(struct folio *folio) 1121 { 1122 if (folio_test_locked(folio)) 1123 folio_wait_bit(folio, PG_locked); 1124 } 1125 1126 static inline int folio_wait_locked_killable(struct folio *folio) 1127 { 1128 if (!folio_test_locked(folio)) 1129 return 0; 1130 return folio_wait_bit_killable(folio, PG_locked); 1131 } 1132 1133 static inline void wait_on_page_locked(struct page *page) 1134 { 1135 folio_wait_locked(page_folio(page)); 1136 } 1137 1138 void folio_end_read(struct folio *folio, bool success); 1139 void wait_on_page_writeback(struct page *page); 1140 void folio_wait_writeback(struct folio *folio); 1141 int folio_wait_writeback_killable(struct folio *folio); 1142 void end_page_writeback(struct page *page); 1143 void folio_end_writeback(struct folio *folio); 1144 void wait_for_stable_page(struct page *page); 1145 void folio_wait_stable(struct folio *folio); 1146 void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn); 1147 void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb); 1148 void __folio_cancel_dirty(struct folio *folio); 1149 static inline void folio_cancel_dirty(struct folio *folio) 1150 { 1151 /* Avoid atomic ops, locking, etc. when not actually needed. */ 1152 if (folio_test_dirty(folio)) 1153 __folio_cancel_dirty(folio); 1154 } 1155 bool folio_clear_dirty_for_io(struct folio *folio); 1156 bool clear_page_dirty_for_io(struct page *page); 1157 void folio_invalidate(struct folio *folio, size_t offset, size_t length); 1158 bool noop_dirty_folio(struct address_space *mapping, struct folio *folio); 1159 1160 #ifdef CONFIG_MIGRATION 1161 int filemap_migrate_folio(struct address_space *mapping, struct folio *dst, 1162 struct folio *src, enum migrate_mode mode); 1163 #else 1164 #define filemap_migrate_folio NULL 1165 #endif 1166 void folio_end_private_2(struct folio *folio); 1167 void folio_wait_private_2(struct folio *folio); 1168 int folio_wait_private_2_killable(struct folio *folio); 1169 1170 /* 1171 * Add an arbitrary waiter to a page's wait queue 1172 */ 1173 void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter); 1174 1175 /* 1176 * Fault in userspace address range. 1177 */ 1178 size_t fault_in_writeable(char __user *uaddr, size_t size); 1179 size_t fault_in_subpage_writeable(char __user *uaddr, size_t size); 1180 size_t fault_in_safe_writeable(const char __user *uaddr, size_t size); 1181 size_t fault_in_readable(const char __user *uaddr, size_t size); 1182 1183 int add_to_page_cache_lru(struct page *page, struct address_space *mapping, 1184 pgoff_t index, gfp_t gfp); 1185 int filemap_add_folio(struct address_space *mapping, struct folio *folio, 1186 pgoff_t index, gfp_t gfp); 1187 void filemap_remove_folio(struct folio *folio); 1188 void __filemap_remove_folio(struct folio *folio, void *shadow); 1189 void replace_page_cache_folio(struct folio *old, struct folio *new); 1190 void delete_from_page_cache_batch(struct address_space *mapping, 1191 struct folio_batch *fbatch); 1192 bool filemap_release_folio(struct folio *folio, gfp_t gfp); 1193 loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end, 1194 int whence); 1195 1196 /* Must be non-static for BPF error injection */ 1197 int __filemap_add_folio(struct address_space *mapping, struct folio *folio, 1198 pgoff_t index, gfp_t gfp, void **shadowp); 1199 1200 bool filemap_range_has_writeback(struct address_space *mapping, 1201 loff_t start_byte, loff_t end_byte); 1202 1203 /** 1204 * filemap_range_needs_writeback - check if range potentially needs writeback 1205 * @mapping: address space within which to check 1206 * @start_byte: offset in bytes where the range starts 1207 * @end_byte: offset in bytes where the range ends (inclusive) 1208 * 1209 * Find at least one page in the range supplied, usually used to check if 1210 * direct writing in this range will trigger a writeback. Used by O_DIRECT 1211 * read/write with IOCB_NOWAIT, to see if the caller needs to do 1212 * filemap_write_and_wait_range() before proceeding. 1213 * 1214 * Return: %true if the caller should do filemap_write_and_wait_range() before 1215 * doing O_DIRECT to a page in this range, %false otherwise. 1216 */ 1217 static inline bool filemap_range_needs_writeback(struct address_space *mapping, 1218 loff_t start_byte, 1219 loff_t end_byte) 1220 { 1221 if (!mapping->nrpages) 1222 return false; 1223 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) && 1224 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) 1225 return false; 1226 return filemap_range_has_writeback(mapping, start_byte, end_byte); 1227 } 1228 1229 /** 1230 * struct readahead_control - Describes a readahead request. 1231 * 1232 * A readahead request is for consecutive pages. Filesystems which 1233 * implement the ->readahead method should call readahead_page() or 1234 * readahead_page_batch() in a loop and attempt to start I/O against 1235 * each page in the request. 1236 * 1237 * Most of the fields in this struct are private and should be accessed 1238 * by the functions below. 1239 * 1240 * @file: The file, used primarily by network filesystems for authentication. 1241 * May be NULL if invoked internally by the filesystem. 1242 * @mapping: Readahead this filesystem object. 1243 * @ra: File readahead state. May be NULL. 1244 */ 1245 struct readahead_control { 1246 struct file *file; 1247 struct address_space *mapping; 1248 struct file_ra_state *ra; 1249 /* private: use the readahead_* accessors instead */ 1250 pgoff_t _index; 1251 unsigned int _nr_pages; 1252 unsigned int _batch_count; 1253 bool _workingset; 1254 unsigned long _pflags; 1255 }; 1256 1257 #define DEFINE_READAHEAD(ractl, f, r, m, i) \ 1258 struct readahead_control ractl = { \ 1259 .file = f, \ 1260 .mapping = m, \ 1261 .ra = r, \ 1262 ._index = i, \ 1263 } 1264 1265 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) 1266 1267 void page_cache_ra_unbounded(struct readahead_control *, 1268 unsigned long nr_to_read, unsigned long lookahead_count); 1269 void page_cache_sync_ra(struct readahead_control *, unsigned long req_count); 1270 void page_cache_async_ra(struct readahead_control *, struct folio *, 1271 unsigned long req_count); 1272 void readahead_expand(struct readahead_control *ractl, 1273 loff_t new_start, size_t new_len); 1274 1275 /** 1276 * page_cache_sync_readahead - generic file readahead 1277 * @mapping: address_space which holds the pagecache and I/O vectors 1278 * @ra: file_ra_state which holds the readahead state 1279 * @file: Used by the filesystem for authentication. 1280 * @index: Index of first page to be read. 1281 * @req_count: Total number of pages being read by the caller. 1282 * 1283 * page_cache_sync_readahead() should be called when a cache miss happened: 1284 * it will submit the read. The readahead logic may decide to piggyback more 1285 * pages onto the read request if access patterns suggest it will improve 1286 * performance. 1287 */ 1288 static inline 1289 void page_cache_sync_readahead(struct address_space *mapping, 1290 struct file_ra_state *ra, struct file *file, pgoff_t index, 1291 unsigned long req_count) 1292 { 1293 DEFINE_READAHEAD(ractl, file, ra, mapping, index); 1294 page_cache_sync_ra(&ractl, req_count); 1295 } 1296 1297 /** 1298 * page_cache_async_readahead - file readahead for marked pages 1299 * @mapping: address_space which holds the pagecache and I/O vectors 1300 * @ra: file_ra_state which holds the readahead state 1301 * @file: Used by the filesystem for authentication. 1302 * @folio: The folio at @index which triggered the readahead call. 1303 * @index: Index of first page to be read. 1304 * @req_count: Total number of pages being read by the caller. 1305 * 1306 * page_cache_async_readahead() should be called when a page is used which 1307 * is marked as PageReadahead; this is a marker to suggest that the application 1308 * has used up enough of the readahead window that we should start pulling in 1309 * more pages. 1310 */ 1311 static inline 1312 void page_cache_async_readahead(struct address_space *mapping, 1313 struct file_ra_state *ra, struct file *file, 1314 struct folio *folio, pgoff_t index, unsigned long req_count) 1315 { 1316 DEFINE_READAHEAD(ractl, file, ra, mapping, index); 1317 page_cache_async_ra(&ractl, folio, req_count); 1318 } 1319 1320 static inline struct folio *__readahead_folio(struct readahead_control *ractl) 1321 { 1322 struct folio *folio; 1323 1324 BUG_ON(ractl->_batch_count > ractl->_nr_pages); 1325 ractl->_nr_pages -= ractl->_batch_count; 1326 ractl->_index += ractl->_batch_count; 1327 1328 if (!ractl->_nr_pages) { 1329 ractl->_batch_count = 0; 1330 return NULL; 1331 } 1332 1333 folio = xa_load(&ractl->mapping->i_pages, ractl->_index); 1334 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); 1335 ractl->_batch_count = folio_nr_pages(folio); 1336 1337 return folio; 1338 } 1339 1340 /** 1341 * readahead_page - Get the next page to read. 1342 * @ractl: The current readahead request. 1343 * 1344 * Context: The page is locked and has an elevated refcount. The caller 1345 * should decreases the refcount once the page has been submitted for I/O 1346 * and unlock the page once all I/O to that page has completed. 1347 * Return: A pointer to the next page, or %NULL if we are done. 1348 */ 1349 static inline struct page *readahead_page(struct readahead_control *ractl) 1350 { 1351 struct folio *folio = __readahead_folio(ractl); 1352 1353 return &folio->page; 1354 } 1355 1356 /** 1357 * readahead_folio - Get the next folio to read. 1358 * @ractl: The current readahead request. 1359 * 1360 * Context: The folio is locked. The caller should unlock the folio once 1361 * all I/O to that folio has completed. 1362 * Return: A pointer to the next folio, or %NULL if we are done. 1363 */ 1364 static inline struct folio *readahead_folio(struct readahead_control *ractl) 1365 { 1366 struct folio *folio = __readahead_folio(ractl); 1367 1368 if (folio) 1369 folio_put(folio); 1370 return folio; 1371 } 1372 1373 static inline unsigned int __readahead_batch(struct readahead_control *rac, 1374 struct page **array, unsigned int array_sz) 1375 { 1376 unsigned int i = 0; 1377 XA_STATE(xas, &rac->mapping->i_pages, 0); 1378 struct page *page; 1379 1380 BUG_ON(rac->_batch_count > rac->_nr_pages); 1381 rac->_nr_pages -= rac->_batch_count; 1382 rac->_index += rac->_batch_count; 1383 rac->_batch_count = 0; 1384 1385 xas_set(&xas, rac->_index); 1386 rcu_read_lock(); 1387 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) { 1388 if (xas_retry(&xas, page)) 1389 continue; 1390 VM_BUG_ON_PAGE(!PageLocked(page), page); 1391 VM_BUG_ON_PAGE(PageTail(page), page); 1392 array[i++] = page; 1393 rac->_batch_count += thp_nr_pages(page); 1394 if (i == array_sz) 1395 break; 1396 } 1397 rcu_read_unlock(); 1398 1399 return i; 1400 } 1401 1402 /** 1403 * readahead_page_batch - Get a batch of pages to read. 1404 * @rac: The current readahead request. 1405 * @array: An array of pointers to struct page. 1406 * 1407 * Context: The pages are locked and have an elevated refcount. The caller 1408 * should decreases the refcount once the page has been submitted for I/O 1409 * and unlock the page once all I/O to that page has completed. 1410 * Return: The number of pages placed in the array. 0 indicates the request 1411 * is complete. 1412 */ 1413 #define readahead_page_batch(rac, array) \ 1414 __readahead_batch(rac, array, ARRAY_SIZE(array)) 1415 1416 /** 1417 * readahead_pos - The byte offset into the file of this readahead request. 1418 * @rac: The readahead request. 1419 */ 1420 static inline loff_t readahead_pos(struct readahead_control *rac) 1421 { 1422 return (loff_t)rac->_index * PAGE_SIZE; 1423 } 1424 1425 /** 1426 * readahead_length - The number of bytes in this readahead request. 1427 * @rac: The readahead request. 1428 */ 1429 static inline size_t readahead_length(struct readahead_control *rac) 1430 { 1431 return rac->_nr_pages * PAGE_SIZE; 1432 } 1433 1434 /** 1435 * readahead_index - The index of the first page in this readahead request. 1436 * @rac: The readahead request. 1437 */ 1438 static inline pgoff_t readahead_index(struct readahead_control *rac) 1439 { 1440 return rac->_index; 1441 } 1442 1443 /** 1444 * readahead_count - The number of pages in this readahead request. 1445 * @rac: The readahead request. 1446 */ 1447 static inline unsigned int readahead_count(struct readahead_control *rac) 1448 { 1449 return rac->_nr_pages; 1450 } 1451 1452 /** 1453 * readahead_batch_length - The number of bytes in the current batch. 1454 * @rac: The readahead request. 1455 */ 1456 static inline size_t readahead_batch_length(struct readahead_control *rac) 1457 { 1458 return rac->_batch_count * PAGE_SIZE; 1459 } 1460 1461 static inline unsigned long dir_pages(struct inode *inode) 1462 { 1463 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> 1464 PAGE_SHIFT; 1465 } 1466 1467 /** 1468 * folio_mkwrite_check_truncate - check if folio was truncated 1469 * @folio: the folio to check 1470 * @inode: the inode to check the folio against 1471 * 1472 * Return: the number of bytes in the folio up to EOF, 1473 * or -EFAULT if the folio was truncated. 1474 */ 1475 static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio, 1476 struct inode *inode) 1477 { 1478 loff_t size = i_size_read(inode); 1479 pgoff_t index = size >> PAGE_SHIFT; 1480 size_t offset = offset_in_folio(folio, size); 1481 1482 if (!folio->mapping) 1483 return -EFAULT; 1484 1485 /* folio is wholly inside EOF */ 1486 if (folio_next_index(folio) - 1 < index) 1487 return folio_size(folio); 1488 /* folio is wholly past EOF */ 1489 if (folio->index > index || !offset) 1490 return -EFAULT; 1491 /* folio is partially inside EOF */ 1492 return offset; 1493 } 1494 1495 /** 1496 * page_mkwrite_check_truncate - check if page was truncated 1497 * @page: the page to check 1498 * @inode: the inode to check the page against 1499 * 1500 * Returns the number of bytes in the page up to EOF, 1501 * or -EFAULT if the page was truncated. 1502 */ 1503 static inline int page_mkwrite_check_truncate(struct page *page, 1504 struct inode *inode) 1505 { 1506 loff_t size = i_size_read(inode); 1507 pgoff_t index = size >> PAGE_SHIFT; 1508 int offset = offset_in_page(size); 1509 1510 if (page->mapping != inode->i_mapping) 1511 return -EFAULT; 1512 1513 /* page is wholly inside EOF */ 1514 if (page->index < index) 1515 return PAGE_SIZE; 1516 /* page is wholly past EOF */ 1517 if (page->index > index || !offset) 1518 return -EFAULT; 1519 /* page is partially inside EOF */ 1520 return offset; 1521 } 1522 1523 /** 1524 * i_blocks_per_folio - How many blocks fit in this folio. 1525 * @inode: The inode which contains the blocks. 1526 * @folio: The folio. 1527 * 1528 * If the block size is larger than the size of this folio, return zero. 1529 * 1530 * Context: The caller should hold a refcount on the folio to prevent it 1531 * from being split. 1532 * Return: The number of filesystem blocks covered by this folio. 1533 */ 1534 static inline 1535 unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio) 1536 { 1537 return folio_size(folio) >> inode->i_blkbits; 1538 } 1539 1540 static inline 1541 unsigned int i_blocks_per_page(struct inode *inode, struct page *page) 1542 { 1543 return i_blocks_per_folio(inode, page_folio(page)); 1544 } 1545 #endif /* _LINUX_PAGEMAP_H */ 1546