at v6.7 1534 lines 48 kB view raw
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 19struct folio_batch; 20 21unsigned long invalidate_mapping_pages(struct address_space *mapping, 22 pgoff_t start, pgoff_t end); 23 24static 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} 30int invalidate_inode_pages2(struct address_space *mapping); 31int invalidate_inode_pages2_range(struct address_space *mapping, 32 pgoff_t start, pgoff_t end); 33int kiocb_invalidate_pages(struct kiocb *iocb, size_t count); 34void kiocb_invalidate_post_direct_write(struct kiocb *iocb, size_t count); 35 36int write_inode_now(struct inode *, int sync); 37int filemap_fdatawrite(struct address_space *); 38int filemap_flush(struct address_space *); 39int filemap_fdatawait_keep_errors(struct address_space *mapping); 40int filemap_fdatawait_range(struct address_space *, loff_t lstart, loff_t lend); 41int filemap_fdatawait_range_keep_errors(struct address_space *mapping, 42 loff_t start_byte, loff_t end_byte); 43 44static inline int filemap_fdatawait(struct address_space *mapping) 45{ 46 return filemap_fdatawait_range(mapping, 0, LLONG_MAX); 47} 48 49bool filemap_range_has_page(struct address_space *, loff_t lstart, loff_t lend); 50int filemap_write_and_wait_range(struct address_space *mapping, 51 loff_t lstart, loff_t lend); 52int __filemap_fdatawrite_range(struct address_space *mapping, 53 loff_t start, loff_t end, int sync_mode); 54int filemap_fdatawrite_range(struct address_space *mapping, 55 loff_t start, loff_t end); 56int filemap_check_errors(struct address_space *mapping); 57void __filemap_set_wb_err(struct address_space *mapping, int err); 58int filemap_fdatawrite_wbc(struct address_space *mapping, 59 struct writeback_control *wbc); 60int kiocb_write_and_wait(struct kiocb *iocb, size_t count); 61 62static inline int filemap_write_and_wait(struct address_space *mapping) 63{ 64 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX); 65} 66 67/** 68 * filemap_set_wb_err - set a writeback error on an address_space 69 * @mapping: mapping in which to set writeback error 70 * @err: error to be set in mapping 71 * 72 * When writeback fails in some way, we must record that error so that 73 * userspace can be informed when fsync and the like are called. We endeavor 74 * to report errors on any file that was open at the time of the error. Some 75 * internal callers also need to know when writeback errors have occurred. 76 * 77 * When a writeback error occurs, most filesystems will want to call 78 * filemap_set_wb_err to record the error in the mapping so that it will be 79 * automatically reported whenever fsync is called on the file. 80 */ 81static inline void filemap_set_wb_err(struct address_space *mapping, int err) 82{ 83 /* Fastpath for common case of no error */ 84 if (unlikely(err)) 85 __filemap_set_wb_err(mapping, err); 86} 87 88/** 89 * filemap_check_wb_err - has an error occurred since the mark was sampled? 90 * @mapping: mapping to check for writeback errors 91 * @since: previously-sampled errseq_t 92 * 93 * Grab the errseq_t value from the mapping, and see if it has changed "since" 94 * the given value was sampled. 95 * 96 * If it has then report the latest error set, otherwise return 0. 97 */ 98static inline int filemap_check_wb_err(struct address_space *mapping, 99 errseq_t since) 100{ 101 return errseq_check(&mapping->wb_err, since); 102} 103 104/** 105 * filemap_sample_wb_err - sample the current errseq_t to test for later errors 106 * @mapping: mapping to be sampled 107 * 108 * Writeback errors are always reported relative to a particular sample point 109 * in the past. This function provides those sample points. 110 */ 111static inline errseq_t filemap_sample_wb_err(struct address_space *mapping) 112{ 113 return errseq_sample(&mapping->wb_err); 114} 115 116/** 117 * file_sample_sb_err - sample the current errseq_t to test for later errors 118 * @file: file pointer to be sampled 119 * 120 * Grab the most current superblock-level errseq_t value for the given 121 * struct file. 122 */ 123static inline errseq_t file_sample_sb_err(struct file *file) 124{ 125 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err); 126} 127 128/* 129 * Flush file data before changing attributes. Caller must hold any locks 130 * required to prevent further writes to this file until we're done setting 131 * flags. 132 */ 133static inline int inode_drain_writes(struct inode *inode) 134{ 135 inode_dio_wait(inode); 136 return filemap_write_and_wait(inode->i_mapping); 137} 138 139static inline bool mapping_empty(struct address_space *mapping) 140{ 141 return xa_empty(&mapping->i_pages); 142} 143 144/* 145 * mapping_shrinkable - test if page cache state allows inode reclaim 146 * @mapping: the page cache mapping 147 * 148 * This checks the mapping's cache state for the pupose of inode 149 * reclaim and LRU management. 150 * 151 * The caller is expected to hold the i_lock, but is not required to 152 * hold the i_pages lock, which usually protects cache state. That's 153 * because the i_lock and the list_lru lock that protect the inode and 154 * its LRU state don't nest inside the irq-safe i_pages lock. 155 * 156 * Cache deletions are performed under the i_lock, which ensures that 157 * when an inode goes empty, it will reliably get queued on the LRU. 158 * 159 * Cache additions do not acquire the i_lock and may race with this 160 * check, in which case we'll report the inode as shrinkable when it 161 * has cache pages. This is okay: the shrinker also checks the 162 * refcount and the referenced bit, which will be elevated or set in 163 * the process of adding new cache pages to an inode. 164 */ 165static inline bool mapping_shrinkable(struct address_space *mapping) 166{ 167 void *head; 168 169 /* 170 * On highmem systems, there could be lowmem pressure from the 171 * inodes before there is highmem pressure from the page 172 * cache. Make inodes shrinkable regardless of cache state. 173 */ 174 if (IS_ENABLED(CONFIG_HIGHMEM)) 175 return true; 176 177 /* Cache completely empty? Shrink away. */ 178 head = rcu_access_pointer(mapping->i_pages.xa_head); 179 if (!head) 180 return true; 181 182 /* 183 * The xarray stores single offset-0 entries directly in the 184 * head pointer, which allows non-resident page cache entries 185 * to escape the shadow shrinker's list of xarray nodes. The 186 * inode shrinker needs to pick them up under memory pressure. 187 */ 188 if (!xa_is_node(head) && xa_is_value(head)) 189 return true; 190 191 return false; 192} 193 194/* 195 * Bits in mapping->flags. 196 */ 197enum mapping_flags { 198 AS_EIO = 0, /* IO error on async write */ 199 AS_ENOSPC = 1, /* ENOSPC on async write */ 200 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */ 201 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */ 202 AS_EXITING = 4, /* final truncate in progress */ 203 /* writeback related tags are not used */ 204 AS_NO_WRITEBACK_TAGS = 5, 205 AS_LARGE_FOLIO_SUPPORT = 6, 206 AS_RELEASE_ALWAYS, /* Call ->release_folio(), even if no private data */ 207 AS_STABLE_WRITES, /* must wait for writeback before modifying 208 folio contents */ 209}; 210 211/** 212 * mapping_set_error - record a writeback error in the address_space 213 * @mapping: the mapping in which an error should be set 214 * @error: the error to set in the mapping 215 * 216 * When writeback fails in some way, we must record that error so that 217 * userspace can be informed when fsync and the like are called. We endeavor 218 * to report errors on any file that was open at the time of the error. Some 219 * internal callers also need to know when writeback errors have occurred. 220 * 221 * When a writeback error occurs, most filesystems will want to call 222 * mapping_set_error to record the error in the mapping so that it can be 223 * reported when the application calls fsync(2). 224 */ 225static inline void mapping_set_error(struct address_space *mapping, int error) 226{ 227 if (likely(!error)) 228 return; 229 230 /* Record in wb_err for checkers using errseq_t based tracking */ 231 __filemap_set_wb_err(mapping, error); 232 233 /* Record it in superblock */ 234 if (mapping->host) 235 errseq_set(&mapping->host->i_sb->s_wb_err, error); 236 237 /* Record it in flags for now, for legacy callers */ 238 if (error == -ENOSPC) 239 set_bit(AS_ENOSPC, &mapping->flags); 240 else 241 set_bit(AS_EIO, &mapping->flags); 242} 243 244static inline void mapping_set_unevictable(struct address_space *mapping) 245{ 246 set_bit(AS_UNEVICTABLE, &mapping->flags); 247} 248 249static inline void mapping_clear_unevictable(struct address_space *mapping) 250{ 251 clear_bit(AS_UNEVICTABLE, &mapping->flags); 252} 253 254static inline bool mapping_unevictable(struct address_space *mapping) 255{ 256 return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags); 257} 258 259static inline void mapping_set_exiting(struct address_space *mapping) 260{ 261 set_bit(AS_EXITING, &mapping->flags); 262} 263 264static inline int mapping_exiting(struct address_space *mapping) 265{ 266 return test_bit(AS_EXITING, &mapping->flags); 267} 268 269static inline void mapping_set_no_writeback_tags(struct address_space *mapping) 270{ 271 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); 272} 273 274static inline int mapping_use_writeback_tags(struct address_space *mapping) 275{ 276 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags); 277} 278 279static inline bool mapping_release_always(const struct address_space *mapping) 280{ 281 return test_bit(AS_RELEASE_ALWAYS, &mapping->flags); 282} 283 284static inline void mapping_set_release_always(struct address_space *mapping) 285{ 286 set_bit(AS_RELEASE_ALWAYS, &mapping->flags); 287} 288 289static inline void mapping_clear_release_always(struct address_space *mapping) 290{ 291 clear_bit(AS_RELEASE_ALWAYS, &mapping->flags); 292} 293 294static inline bool mapping_stable_writes(const struct address_space *mapping) 295{ 296 return test_bit(AS_STABLE_WRITES, &mapping->flags); 297} 298 299static inline void mapping_set_stable_writes(struct address_space *mapping) 300{ 301 set_bit(AS_STABLE_WRITES, &mapping->flags); 302} 303 304static inline void mapping_clear_stable_writes(struct address_space *mapping) 305{ 306 clear_bit(AS_STABLE_WRITES, &mapping->flags); 307} 308 309static inline gfp_t mapping_gfp_mask(struct address_space * mapping) 310{ 311 return mapping->gfp_mask; 312} 313 314/* Restricts the given gfp_mask to what the mapping allows. */ 315static inline gfp_t mapping_gfp_constraint(struct address_space *mapping, 316 gfp_t gfp_mask) 317{ 318 return mapping_gfp_mask(mapping) & gfp_mask; 319} 320 321/* 322 * This is non-atomic. Only to be used before the mapping is activated. 323 * Probably needs a barrier... 324 */ 325static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask) 326{ 327 m->gfp_mask = mask; 328} 329 330/** 331 * mapping_set_large_folios() - Indicate the file supports large folios. 332 * @mapping: The file. 333 * 334 * The filesystem should call this function in its inode constructor to 335 * indicate that the VFS can use large folios to cache the contents of 336 * the file. 337 * 338 * Context: This should not be called while the inode is active as it 339 * is non-atomic. 340 */ 341static inline void mapping_set_large_folios(struct address_space *mapping) 342{ 343 __set_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); 344} 345 346/* 347 * Large folio support currently depends on THP. These dependencies are 348 * being worked on but are not yet fixed. 349 */ 350static inline bool mapping_large_folio_support(struct address_space *mapping) 351{ 352 return IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) && 353 test_bit(AS_LARGE_FOLIO_SUPPORT, &mapping->flags); 354} 355 356static inline int filemap_nr_thps(struct address_space *mapping) 357{ 358#ifdef CONFIG_READ_ONLY_THP_FOR_FS 359 return atomic_read(&mapping->nr_thps); 360#else 361 return 0; 362#endif 363} 364 365static inline void filemap_nr_thps_inc(struct address_space *mapping) 366{ 367#ifdef CONFIG_READ_ONLY_THP_FOR_FS 368 if (!mapping_large_folio_support(mapping)) 369 atomic_inc(&mapping->nr_thps); 370#else 371 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); 372#endif 373} 374 375static inline void filemap_nr_thps_dec(struct address_space *mapping) 376{ 377#ifdef CONFIG_READ_ONLY_THP_FOR_FS 378 if (!mapping_large_folio_support(mapping)) 379 atomic_dec(&mapping->nr_thps); 380#else 381 WARN_ON_ONCE(mapping_large_folio_support(mapping) == 0); 382#endif 383} 384 385struct address_space *page_mapping(struct page *); 386struct address_space *folio_mapping(struct folio *); 387struct address_space *swapcache_mapping(struct folio *); 388 389/** 390 * folio_file_mapping - Find the mapping this folio belongs to. 391 * @folio: The folio. 392 * 393 * For folios which are in the page cache, return the mapping that this 394 * page belongs to. Folios in the swap cache return the mapping of the 395 * swap file or swap device where the data is stored. This is different 396 * from the mapping returned by folio_mapping(). The only reason to 397 * use it is if, like NFS, you return 0 from ->activate_swapfile. 398 * 399 * Do not call this for folios which aren't in the page cache or swap cache. 400 */ 401static inline struct address_space *folio_file_mapping(struct folio *folio) 402{ 403 if (unlikely(folio_test_swapcache(folio))) 404 return swapcache_mapping(folio); 405 406 return folio->mapping; 407} 408 409/** 410 * folio_flush_mapping - Find the file mapping this folio belongs to. 411 * @folio: The folio. 412 * 413 * For folios which are in the page cache, return the mapping that this 414 * page belongs to. Anonymous folios return NULL, even if they're in 415 * the swap cache. Other kinds of folio also return NULL. 416 * 417 * This is ONLY used by architecture cache flushing code. If you aren't 418 * writing cache flushing code, you want either folio_mapping() or 419 * folio_file_mapping(). 420 */ 421static inline struct address_space *folio_flush_mapping(struct folio *folio) 422{ 423 if (unlikely(folio_test_swapcache(folio))) 424 return NULL; 425 426 return folio_mapping(folio); 427} 428 429static inline struct address_space *page_file_mapping(struct page *page) 430{ 431 return folio_file_mapping(page_folio(page)); 432} 433 434/** 435 * folio_inode - Get the host inode for this folio. 436 * @folio: The folio. 437 * 438 * For folios which are in the page cache, return the inode that this folio 439 * belongs to. 440 * 441 * Do not call this for folios which aren't in the page cache. 442 */ 443static inline struct inode *folio_inode(struct folio *folio) 444{ 445 return folio->mapping->host; 446} 447 448/** 449 * folio_attach_private - Attach private data to a folio. 450 * @folio: Folio to attach data to. 451 * @data: Data to attach to folio. 452 * 453 * Attaching private data to a folio increments the page's reference count. 454 * The data must be detached before the folio will be freed. 455 */ 456static inline void folio_attach_private(struct folio *folio, void *data) 457{ 458 folio_get(folio); 459 folio->private = data; 460 folio_set_private(folio); 461} 462 463/** 464 * folio_change_private - Change private data on a folio. 465 * @folio: Folio to change the data on. 466 * @data: Data to set on the folio. 467 * 468 * Change the private data attached to a folio and return the old 469 * data. The page must previously have had data attached and the data 470 * must be detached before the folio will be freed. 471 * 472 * Return: Data that was previously attached to the folio. 473 */ 474static inline void *folio_change_private(struct folio *folio, void *data) 475{ 476 void *old = folio_get_private(folio); 477 478 folio->private = data; 479 return old; 480} 481 482/** 483 * folio_detach_private - Detach private data from a folio. 484 * @folio: Folio to detach data from. 485 * 486 * Removes the data that was previously attached to the folio and decrements 487 * the refcount on the page. 488 * 489 * Return: Data that was attached to the folio. 490 */ 491static inline void *folio_detach_private(struct folio *folio) 492{ 493 void *data = folio_get_private(folio); 494 495 if (!folio_test_private(folio)) 496 return NULL; 497 folio_clear_private(folio); 498 folio->private = NULL; 499 folio_put(folio); 500 501 return data; 502} 503 504static inline void attach_page_private(struct page *page, void *data) 505{ 506 folio_attach_private(page_folio(page), data); 507} 508 509static inline void *detach_page_private(struct page *page) 510{ 511 return folio_detach_private(page_folio(page)); 512} 513 514/* 515 * There are some parts of the kernel which assume that PMD entries 516 * are exactly HPAGE_PMD_ORDER. Those should be fixed, but until then, 517 * limit the maximum allocation order to PMD size. I'm not aware of any 518 * assumptions about maximum order if THP are disabled, but 8 seems like 519 * a good order (that's 1MB if you're using 4kB pages) 520 */ 521#ifdef CONFIG_TRANSPARENT_HUGEPAGE 522#define MAX_PAGECACHE_ORDER HPAGE_PMD_ORDER 523#else 524#define MAX_PAGECACHE_ORDER 8 525#endif 526 527#ifdef CONFIG_NUMA 528struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order); 529#else 530static inline struct folio *filemap_alloc_folio(gfp_t gfp, unsigned int order) 531{ 532 return folio_alloc(gfp, order); 533} 534#endif 535 536static inline struct page *__page_cache_alloc(gfp_t gfp) 537{ 538 return &filemap_alloc_folio(gfp, 0)->page; 539} 540 541static inline struct page *page_cache_alloc(struct address_space *x) 542{ 543 return __page_cache_alloc(mapping_gfp_mask(x)); 544} 545 546static inline gfp_t readahead_gfp_mask(struct address_space *x) 547{ 548 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN; 549} 550 551typedef int filler_t(struct file *, struct folio *); 552 553pgoff_t page_cache_next_miss(struct address_space *mapping, 554 pgoff_t index, unsigned long max_scan); 555pgoff_t page_cache_prev_miss(struct address_space *mapping, 556 pgoff_t index, unsigned long max_scan); 557 558/** 559 * typedef fgf_t - Flags for getting folios from the page cache. 560 * 561 * Most users of the page cache will not need to use these flags; 562 * there are convenience functions such as filemap_get_folio() and 563 * filemap_lock_folio(). For users which need more control over exactly 564 * what is done with the folios, these flags to __filemap_get_folio() 565 * are available. 566 * 567 * * %FGP_ACCESSED - The folio will be marked accessed. 568 * * %FGP_LOCK - The folio is returned locked. 569 * * %FGP_CREAT - If no folio is present then a new folio is allocated, 570 * added to the page cache and the VM's LRU list. The folio is 571 * returned locked. 572 * * %FGP_FOR_MMAP - The caller wants to do its own locking dance if the 573 * folio is already in cache. If the folio was allocated, unlock it 574 * before returning so the caller can do the same dance. 575 * * %FGP_WRITE - The folio will be written to by the caller. 576 * * %FGP_NOFS - __GFP_FS will get cleared in gfp. 577 * * %FGP_NOWAIT - Don't block on the folio lock. 578 * * %FGP_STABLE - Wait for the folio to be stable (finished writeback) 579 * * %FGP_WRITEBEGIN - The flags to use in a filesystem write_begin() 580 * implementation. 581 */ 582typedef unsigned int __bitwise fgf_t; 583 584#define FGP_ACCESSED ((__force fgf_t)0x00000001) 585#define FGP_LOCK ((__force fgf_t)0x00000002) 586#define FGP_CREAT ((__force fgf_t)0x00000004) 587#define FGP_WRITE ((__force fgf_t)0x00000008) 588#define FGP_NOFS ((__force fgf_t)0x00000010) 589#define FGP_NOWAIT ((__force fgf_t)0x00000020) 590#define FGP_FOR_MMAP ((__force fgf_t)0x00000040) 591#define FGP_STABLE ((__force fgf_t)0x00000080) 592#define FGF_GET_ORDER(fgf) (((__force unsigned)fgf) >> 26) /* top 6 bits */ 593 594#define FGP_WRITEBEGIN (FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE) 595 596/** 597 * fgf_set_order - Encode a length in the fgf_t flags. 598 * @size: The suggested size of the folio to create. 599 * 600 * The caller of __filemap_get_folio() can use this to suggest a preferred 601 * size for the folio that is created. If there is already a folio at 602 * the index, it will be returned, no matter what its size. If a folio 603 * is freshly created, it may be of a different size than requested 604 * due to alignment constraints, memory pressure, or the presence of 605 * other folios at nearby indices. 606 */ 607static inline fgf_t fgf_set_order(size_t size) 608{ 609 unsigned int shift = ilog2(size); 610 611 if (shift <= PAGE_SHIFT) 612 return 0; 613 return (__force fgf_t)((shift - PAGE_SHIFT) << 26); 614} 615 616void *filemap_get_entry(struct address_space *mapping, pgoff_t index); 617struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, 618 fgf_t fgp_flags, gfp_t gfp); 619struct page *pagecache_get_page(struct address_space *mapping, pgoff_t index, 620 fgf_t fgp_flags, gfp_t gfp); 621 622/** 623 * filemap_get_folio - Find and get a folio. 624 * @mapping: The address_space to search. 625 * @index: The page index. 626 * 627 * Looks up the page cache entry at @mapping & @index. If a folio is 628 * present, it is returned with an increased refcount. 629 * 630 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for 631 * this index. Will not return a shadow, swap or DAX entry. 632 */ 633static inline struct folio *filemap_get_folio(struct address_space *mapping, 634 pgoff_t index) 635{ 636 return __filemap_get_folio(mapping, index, 0, 0); 637} 638 639/** 640 * filemap_lock_folio - Find and lock 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 locked with an increased refcount. 646 * 647 * Context: May sleep. 648 * Return: A folio or ERR_PTR(-ENOENT) if there is no folio in the cache for 649 * this index. Will not return a shadow, swap or DAX entry. 650 */ 651static inline struct folio *filemap_lock_folio(struct address_space *mapping, 652 pgoff_t index) 653{ 654 return __filemap_get_folio(mapping, index, FGP_LOCK, 0); 655} 656 657/** 658 * filemap_grab_folio - grab a folio from the page cache 659 * @mapping: The address space to search 660 * @index: The page index 661 * 662 * Looks up the page cache entry at @mapping & @index. If no folio is found, 663 * a new folio is created. The folio is locked, marked as accessed, and 664 * returned. 665 * 666 * Return: A found or created folio. ERR_PTR(-ENOMEM) if no folio is found 667 * and failed to create a folio. 668 */ 669static inline struct folio *filemap_grab_folio(struct address_space *mapping, 670 pgoff_t index) 671{ 672 return __filemap_get_folio(mapping, index, 673 FGP_LOCK | FGP_ACCESSED | FGP_CREAT, 674 mapping_gfp_mask(mapping)); 675} 676 677/** 678 * find_get_page - find and get a page reference 679 * @mapping: the address_space to search 680 * @offset: the page index 681 * 682 * Looks up the page cache slot at @mapping & @offset. If there is a 683 * page cache page, it is returned with an increased refcount. 684 * 685 * Otherwise, %NULL is returned. 686 */ 687static inline struct page *find_get_page(struct address_space *mapping, 688 pgoff_t offset) 689{ 690 return pagecache_get_page(mapping, offset, 0, 0); 691} 692 693static inline struct page *find_get_page_flags(struct address_space *mapping, 694 pgoff_t offset, fgf_t fgp_flags) 695{ 696 return pagecache_get_page(mapping, offset, fgp_flags, 0); 697} 698 699/** 700 * find_lock_page - locate, pin and lock a pagecache page 701 * @mapping: the address_space to search 702 * @index: the page index 703 * 704 * Looks up the page cache entry at @mapping & @index. If there is a 705 * page cache page, it is returned locked and with an increased 706 * refcount. 707 * 708 * Context: May sleep. 709 * Return: A struct page or %NULL if there is no page in the cache for this 710 * index. 711 */ 712static inline struct page *find_lock_page(struct address_space *mapping, 713 pgoff_t index) 714{ 715 return pagecache_get_page(mapping, index, FGP_LOCK, 0); 716} 717 718/** 719 * find_or_create_page - locate or add a pagecache page 720 * @mapping: the page's address_space 721 * @index: the page's index into the mapping 722 * @gfp_mask: page allocation mode 723 * 724 * Looks up the page cache slot at @mapping & @offset. If there is a 725 * page cache page, it is returned locked and with an increased 726 * refcount. 727 * 728 * If the page is not present, a new page is allocated using @gfp_mask 729 * and added to the page cache and the VM's LRU list. The page is 730 * returned locked and with an increased refcount. 731 * 732 * On memory exhaustion, %NULL is returned. 733 * 734 * find_or_create_page() may sleep, even if @gfp_flags specifies an 735 * atomic allocation! 736 */ 737static inline struct page *find_or_create_page(struct address_space *mapping, 738 pgoff_t index, gfp_t gfp_mask) 739{ 740 return pagecache_get_page(mapping, index, 741 FGP_LOCK|FGP_ACCESSED|FGP_CREAT, 742 gfp_mask); 743} 744 745/** 746 * grab_cache_page_nowait - returns locked page at given index in given cache 747 * @mapping: target address_space 748 * @index: the page index 749 * 750 * Same as grab_cache_page(), but do not wait if the page is unavailable. 751 * This is intended for speculative data generators, where the data can 752 * be regenerated if the page couldn't be grabbed. This routine should 753 * be safe to call while holding the lock for another page. 754 * 755 * Clear __GFP_FS when allocating the page to avoid recursion into the fs 756 * and deadlock against the caller's locked page. 757 */ 758static inline struct page *grab_cache_page_nowait(struct address_space *mapping, 759 pgoff_t index) 760{ 761 return pagecache_get_page(mapping, index, 762 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT, 763 mapping_gfp_mask(mapping)); 764} 765 766#define swapcache_index(folio) __page_file_index(&(folio)->page) 767 768/** 769 * folio_index - File index of a folio. 770 * @folio: The folio. 771 * 772 * For a folio which is either in the page cache or the swap cache, 773 * return its index within the address_space it belongs to. If you know 774 * the page is definitely in the page cache, you can look at the folio's 775 * index directly. 776 * 777 * Return: The index (offset in units of pages) of a folio in its file. 778 */ 779static inline pgoff_t folio_index(struct folio *folio) 780{ 781 if (unlikely(folio_test_swapcache(folio))) 782 return swapcache_index(folio); 783 return folio->index; 784} 785 786/** 787 * folio_next_index - Get the index of the next folio. 788 * @folio: The current folio. 789 * 790 * Return: The index of the folio which follows this folio in the file. 791 */ 792static inline pgoff_t folio_next_index(struct folio *folio) 793{ 794 return folio->index + folio_nr_pages(folio); 795} 796 797/** 798 * folio_file_page - The page for a particular index. 799 * @folio: The folio which contains this index. 800 * @index: The index we want to look up. 801 * 802 * Sometimes after looking up a folio in the page cache, we need to 803 * obtain the specific page for an index (eg a page fault). 804 * 805 * Return: The page containing the file data for this index. 806 */ 807static inline struct page *folio_file_page(struct folio *folio, pgoff_t index) 808{ 809 return folio_page(folio, index & (folio_nr_pages(folio) - 1)); 810} 811 812/** 813 * folio_contains - Does this folio contain this index? 814 * @folio: The folio. 815 * @index: The page index within the file. 816 * 817 * Context: The caller should have the page locked in order to prevent 818 * (eg) shmem from moving the page between the page cache and swap cache 819 * and changing its index in the middle of the operation. 820 * Return: true or false. 821 */ 822static inline bool folio_contains(struct folio *folio, pgoff_t index) 823{ 824 return index - folio_index(folio) < folio_nr_pages(folio); 825} 826 827/* 828 * Given the page we found in the page cache, return the page corresponding 829 * to this index in the file 830 */ 831static inline struct page *find_subpage(struct page *head, pgoff_t index) 832{ 833 /* HugeTLBfs wants the head page regardless */ 834 if (PageHuge(head)) 835 return head; 836 837 return head + (index & (thp_nr_pages(head) - 1)); 838} 839 840unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start, 841 pgoff_t end, struct folio_batch *fbatch); 842unsigned filemap_get_folios_contig(struct address_space *mapping, 843 pgoff_t *start, pgoff_t end, struct folio_batch *fbatch); 844unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start, 845 pgoff_t end, xa_mark_t tag, struct folio_batch *fbatch); 846 847struct page *grab_cache_page_write_begin(struct address_space *mapping, 848 pgoff_t index); 849 850/* 851 * Returns locked page at given index in given cache, creating it if needed. 852 */ 853static inline struct page *grab_cache_page(struct address_space *mapping, 854 pgoff_t index) 855{ 856 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping)); 857} 858 859struct folio *read_cache_folio(struct address_space *, pgoff_t index, 860 filler_t *filler, struct file *file); 861struct folio *mapping_read_folio_gfp(struct address_space *, pgoff_t index, 862 gfp_t flags); 863struct page *read_cache_page(struct address_space *, pgoff_t index, 864 filler_t *filler, struct file *file); 865extern struct page * read_cache_page_gfp(struct address_space *mapping, 866 pgoff_t index, gfp_t gfp_mask); 867 868static inline struct page *read_mapping_page(struct address_space *mapping, 869 pgoff_t index, struct file *file) 870{ 871 return read_cache_page(mapping, index, NULL, file); 872} 873 874static inline struct folio *read_mapping_folio(struct address_space *mapping, 875 pgoff_t index, struct file *file) 876{ 877 return read_cache_folio(mapping, index, NULL, file); 878} 879 880/* 881 * Get the offset in PAGE_SIZE (even for hugetlb pages). 882 */ 883static inline pgoff_t page_to_pgoff(struct page *page) 884{ 885 struct page *head; 886 887 if (likely(!PageTransTail(page))) 888 return page->index; 889 890 head = compound_head(page); 891 /* 892 * We don't initialize ->index for tail pages: calculate based on 893 * head page 894 */ 895 return head->index + page - head; 896} 897 898/* 899 * Return byte-offset into filesystem object for page. 900 */ 901static inline loff_t page_offset(struct page *page) 902{ 903 return ((loff_t)page->index) << PAGE_SHIFT; 904} 905 906static inline loff_t page_file_offset(struct page *page) 907{ 908 return ((loff_t)page_index(page)) << PAGE_SHIFT; 909} 910 911/** 912 * folio_pos - Returns the byte position of this folio in its file. 913 * @folio: The folio. 914 */ 915static inline loff_t folio_pos(struct folio *folio) 916{ 917 return page_offset(&folio->page); 918} 919 920/** 921 * folio_file_pos - Returns the byte position of this folio in its file. 922 * @folio: The folio. 923 * 924 * This differs from folio_pos() for folios which belong to a swap file. 925 * NFS is the only filesystem today which needs to use folio_file_pos(). 926 */ 927static inline loff_t folio_file_pos(struct folio *folio) 928{ 929 return page_file_offset(&folio->page); 930} 931 932/* 933 * Get the offset in PAGE_SIZE (even for hugetlb folios). 934 */ 935static inline pgoff_t folio_pgoff(struct folio *folio) 936{ 937 return folio->index; 938} 939 940static inline pgoff_t linear_page_index(struct vm_area_struct *vma, 941 unsigned long address) 942{ 943 pgoff_t pgoff; 944 pgoff = (address - vma->vm_start) >> PAGE_SHIFT; 945 pgoff += vma->vm_pgoff; 946 return pgoff; 947} 948 949struct wait_page_key { 950 struct folio *folio; 951 int bit_nr; 952 int page_match; 953}; 954 955struct wait_page_queue { 956 struct folio *folio; 957 int bit_nr; 958 wait_queue_entry_t wait; 959}; 960 961static inline bool wake_page_match(struct wait_page_queue *wait_page, 962 struct wait_page_key *key) 963{ 964 if (wait_page->folio != key->folio) 965 return false; 966 key->page_match = 1; 967 968 if (wait_page->bit_nr != key->bit_nr) 969 return false; 970 971 return true; 972} 973 974void __folio_lock(struct folio *folio); 975int __folio_lock_killable(struct folio *folio); 976vm_fault_t __folio_lock_or_retry(struct folio *folio, struct vm_fault *vmf); 977void unlock_page(struct page *page); 978void folio_unlock(struct folio *folio); 979 980/** 981 * folio_trylock() - Attempt to lock a folio. 982 * @folio: The folio to attempt to lock. 983 * 984 * Sometimes it is undesirable to wait for a folio to be unlocked (eg 985 * when the locks are being taken in the wrong order, or if making 986 * progress through a batch of folios is more important than processing 987 * them in order). Usually folio_lock() is the correct function to call. 988 * 989 * Context: Any context. 990 * Return: Whether the lock was successfully acquired. 991 */ 992static inline bool folio_trylock(struct folio *folio) 993{ 994 return likely(!test_and_set_bit_lock(PG_locked, folio_flags(folio, 0))); 995} 996 997/* 998 * Return true if the page was successfully locked 999 */ 1000static inline int trylock_page(struct page *page) 1001{ 1002 return folio_trylock(page_folio(page)); 1003} 1004 1005/** 1006 * folio_lock() - Lock this folio. 1007 * @folio: The folio to lock. 1008 * 1009 * The folio lock protects against many things, probably more than it 1010 * should. It is primarily held while a folio is being brought uptodate, 1011 * either from its backing file or from swap. It is also held while a 1012 * folio is being truncated from its address_space, so holding the lock 1013 * is sufficient to keep folio->mapping stable. 1014 * 1015 * The folio lock is also held while write() is modifying the page to 1016 * provide POSIX atomicity guarantees (as long as the write does not 1017 * cross a page boundary). Other modifications to the data in the folio 1018 * do not hold the folio lock and can race with writes, eg DMA and stores 1019 * to mapped pages. 1020 * 1021 * Context: May sleep. If you need to acquire the locks of two or 1022 * more folios, they must be in order of ascending index, if they are 1023 * in the same address_space. If they are in different address_spaces, 1024 * acquire the lock of the folio which belongs to the address_space which 1025 * has the lowest address in memory first. 1026 */ 1027static inline void folio_lock(struct folio *folio) 1028{ 1029 might_sleep(); 1030 if (!folio_trylock(folio)) 1031 __folio_lock(folio); 1032} 1033 1034/** 1035 * lock_page() - Lock the folio containing this page. 1036 * @page: The page to lock. 1037 * 1038 * See folio_lock() for a description of what the lock protects. 1039 * This is a legacy function and new code should probably use folio_lock() 1040 * instead. 1041 * 1042 * Context: May sleep. Pages in the same folio share a lock, so do not 1043 * attempt to lock two pages which share a folio. 1044 */ 1045static inline void lock_page(struct page *page) 1046{ 1047 struct folio *folio; 1048 might_sleep(); 1049 1050 folio = page_folio(page); 1051 if (!folio_trylock(folio)) 1052 __folio_lock(folio); 1053} 1054 1055/** 1056 * folio_lock_killable() - Lock this folio, interruptible by a fatal signal. 1057 * @folio: The folio to lock. 1058 * 1059 * Attempts to lock the folio, like folio_lock(), except that the sleep 1060 * to acquire the lock is interruptible by a fatal signal. 1061 * 1062 * Context: May sleep; see folio_lock(). 1063 * Return: 0 if the lock was acquired; -EINTR if a fatal signal was received. 1064 */ 1065static inline int folio_lock_killable(struct folio *folio) 1066{ 1067 might_sleep(); 1068 if (!folio_trylock(folio)) 1069 return __folio_lock_killable(folio); 1070 return 0; 1071} 1072 1073/* 1074 * folio_lock_or_retry - Lock the folio, unless this would block and the 1075 * caller indicated that it can handle a retry. 1076 * 1077 * Return value and mmap_lock implications depend on flags; see 1078 * __folio_lock_or_retry(). 1079 */ 1080static inline vm_fault_t folio_lock_or_retry(struct folio *folio, 1081 struct vm_fault *vmf) 1082{ 1083 might_sleep(); 1084 if (!folio_trylock(folio)) 1085 return __folio_lock_or_retry(folio, vmf); 1086 return 0; 1087} 1088 1089/* 1090 * This is exported only for folio_wait_locked/folio_wait_writeback, etc., 1091 * and should not be used directly. 1092 */ 1093void folio_wait_bit(struct folio *folio, int bit_nr); 1094int folio_wait_bit_killable(struct folio *folio, int bit_nr); 1095 1096/* 1097 * Wait for a folio to be unlocked. 1098 * 1099 * This must be called with the caller "holding" the folio, 1100 * ie with increased folio reference count so that the folio won't 1101 * go away during the wait. 1102 */ 1103static inline void folio_wait_locked(struct folio *folio) 1104{ 1105 if (folio_test_locked(folio)) 1106 folio_wait_bit(folio, PG_locked); 1107} 1108 1109static inline int folio_wait_locked_killable(struct folio *folio) 1110{ 1111 if (!folio_test_locked(folio)) 1112 return 0; 1113 return folio_wait_bit_killable(folio, PG_locked); 1114} 1115 1116static inline void wait_on_page_locked(struct page *page) 1117{ 1118 folio_wait_locked(page_folio(page)); 1119} 1120 1121void folio_end_read(struct folio *folio, bool success); 1122void wait_on_page_writeback(struct page *page); 1123void folio_wait_writeback(struct folio *folio); 1124int folio_wait_writeback_killable(struct folio *folio); 1125void end_page_writeback(struct page *page); 1126void folio_end_writeback(struct folio *folio); 1127void wait_for_stable_page(struct page *page); 1128void folio_wait_stable(struct folio *folio); 1129void __folio_mark_dirty(struct folio *folio, struct address_space *, int warn); 1130static inline void __set_page_dirty(struct page *page, 1131 struct address_space *mapping, int warn) 1132{ 1133 __folio_mark_dirty(page_folio(page), mapping, warn); 1134} 1135void folio_account_cleaned(struct folio *folio, struct bdi_writeback *wb); 1136void __folio_cancel_dirty(struct folio *folio); 1137static inline void folio_cancel_dirty(struct folio *folio) 1138{ 1139 /* Avoid atomic ops, locking, etc. when not actually needed. */ 1140 if (folio_test_dirty(folio)) 1141 __folio_cancel_dirty(folio); 1142} 1143bool folio_clear_dirty_for_io(struct folio *folio); 1144bool clear_page_dirty_for_io(struct page *page); 1145void folio_invalidate(struct folio *folio, size_t offset, size_t length); 1146int __set_page_dirty_nobuffers(struct page *page); 1147bool noop_dirty_folio(struct address_space *mapping, struct folio *folio); 1148 1149#ifdef CONFIG_MIGRATION 1150int filemap_migrate_folio(struct address_space *mapping, struct folio *dst, 1151 struct folio *src, enum migrate_mode mode); 1152#else 1153#define filemap_migrate_folio NULL 1154#endif 1155void folio_end_private_2(struct folio *folio); 1156void folio_wait_private_2(struct folio *folio); 1157int folio_wait_private_2_killable(struct folio *folio); 1158 1159/* 1160 * Add an arbitrary waiter to a page's wait queue 1161 */ 1162void folio_add_wait_queue(struct folio *folio, wait_queue_entry_t *waiter); 1163 1164/* 1165 * Fault in userspace address range. 1166 */ 1167size_t fault_in_writeable(char __user *uaddr, size_t size); 1168size_t fault_in_subpage_writeable(char __user *uaddr, size_t size); 1169size_t fault_in_safe_writeable(const char __user *uaddr, size_t size); 1170size_t fault_in_readable(const char __user *uaddr, size_t size); 1171 1172int add_to_page_cache_lru(struct page *page, struct address_space *mapping, 1173 pgoff_t index, gfp_t gfp); 1174int filemap_add_folio(struct address_space *mapping, struct folio *folio, 1175 pgoff_t index, gfp_t gfp); 1176void filemap_remove_folio(struct folio *folio); 1177void __filemap_remove_folio(struct folio *folio, void *shadow); 1178void replace_page_cache_folio(struct folio *old, struct folio *new); 1179void delete_from_page_cache_batch(struct address_space *mapping, 1180 struct folio_batch *fbatch); 1181bool filemap_release_folio(struct folio *folio, gfp_t gfp); 1182loff_t mapping_seek_hole_data(struct address_space *, loff_t start, loff_t end, 1183 int whence); 1184 1185/* Must be non-static for BPF error injection */ 1186int __filemap_add_folio(struct address_space *mapping, struct folio *folio, 1187 pgoff_t index, gfp_t gfp, void **shadowp); 1188 1189bool filemap_range_has_writeback(struct address_space *mapping, 1190 loff_t start_byte, loff_t end_byte); 1191 1192/** 1193 * filemap_range_needs_writeback - check if range potentially needs writeback 1194 * @mapping: address space within which to check 1195 * @start_byte: offset in bytes where the range starts 1196 * @end_byte: offset in bytes where the range ends (inclusive) 1197 * 1198 * Find at least one page in the range supplied, usually used to check if 1199 * direct writing in this range will trigger a writeback. Used by O_DIRECT 1200 * read/write with IOCB_NOWAIT, to see if the caller needs to do 1201 * filemap_write_and_wait_range() before proceeding. 1202 * 1203 * Return: %true if the caller should do filemap_write_and_wait_range() before 1204 * doing O_DIRECT to a page in this range, %false otherwise. 1205 */ 1206static inline bool filemap_range_needs_writeback(struct address_space *mapping, 1207 loff_t start_byte, 1208 loff_t end_byte) 1209{ 1210 if (!mapping->nrpages) 1211 return false; 1212 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY) && 1213 !mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) 1214 return false; 1215 return filemap_range_has_writeback(mapping, start_byte, end_byte); 1216} 1217 1218/** 1219 * struct readahead_control - Describes a readahead request. 1220 * 1221 * A readahead request is for consecutive pages. Filesystems which 1222 * implement the ->readahead method should call readahead_page() or 1223 * readahead_page_batch() in a loop and attempt to start I/O against 1224 * each page in the request. 1225 * 1226 * Most of the fields in this struct are private and should be accessed 1227 * by the functions below. 1228 * 1229 * @file: The file, used primarily by network filesystems for authentication. 1230 * May be NULL if invoked internally by the filesystem. 1231 * @mapping: Readahead this filesystem object. 1232 * @ra: File readahead state. May be NULL. 1233 */ 1234struct readahead_control { 1235 struct file *file; 1236 struct address_space *mapping; 1237 struct file_ra_state *ra; 1238/* private: use the readahead_* accessors instead */ 1239 pgoff_t _index; 1240 unsigned int _nr_pages; 1241 unsigned int _batch_count; 1242 bool _workingset; 1243 unsigned long _pflags; 1244}; 1245 1246#define DEFINE_READAHEAD(ractl, f, r, m, i) \ 1247 struct readahead_control ractl = { \ 1248 .file = f, \ 1249 .mapping = m, \ 1250 .ra = r, \ 1251 ._index = i, \ 1252 } 1253 1254#define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE) 1255 1256void page_cache_ra_unbounded(struct readahead_control *, 1257 unsigned long nr_to_read, unsigned long lookahead_count); 1258void page_cache_sync_ra(struct readahead_control *, unsigned long req_count); 1259void page_cache_async_ra(struct readahead_control *, struct folio *, 1260 unsigned long req_count); 1261void readahead_expand(struct readahead_control *ractl, 1262 loff_t new_start, size_t new_len); 1263 1264/** 1265 * page_cache_sync_readahead - generic file readahead 1266 * @mapping: address_space which holds the pagecache and I/O vectors 1267 * @ra: file_ra_state which holds the readahead state 1268 * @file: Used by the filesystem for authentication. 1269 * @index: Index of first page to be read. 1270 * @req_count: Total number of pages being read by the caller. 1271 * 1272 * page_cache_sync_readahead() should be called when a cache miss happened: 1273 * it will submit the read. The readahead logic may decide to piggyback more 1274 * pages onto the read request if access patterns suggest it will improve 1275 * performance. 1276 */ 1277static inline 1278void page_cache_sync_readahead(struct address_space *mapping, 1279 struct file_ra_state *ra, struct file *file, pgoff_t index, 1280 unsigned long req_count) 1281{ 1282 DEFINE_READAHEAD(ractl, file, ra, mapping, index); 1283 page_cache_sync_ra(&ractl, req_count); 1284} 1285 1286/** 1287 * page_cache_async_readahead - file readahead for marked pages 1288 * @mapping: address_space which holds the pagecache and I/O vectors 1289 * @ra: file_ra_state which holds the readahead state 1290 * @file: Used by the filesystem for authentication. 1291 * @folio: The folio at @index which triggered the readahead call. 1292 * @index: Index of first page to be read. 1293 * @req_count: Total number of pages being read by the caller. 1294 * 1295 * page_cache_async_readahead() should be called when a page is used which 1296 * is marked as PageReadahead; this is a marker to suggest that the application 1297 * has used up enough of the readahead window that we should start pulling in 1298 * more pages. 1299 */ 1300static inline 1301void page_cache_async_readahead(struct address_space *mapping, 1302 struct file_ra_state *ra, struct file *file, 1303 struct folio *folio, pgoff_t index, unsigned long req_count) 1304{ 1305 DEFINE_READAHEAD(ractl, file, ra, mapping, index); 1306 page_cache_async_ra(&ractl, folio, req_count); 1307} 1308 1309static inline struct folio *__readahead_folio(struct readahead_control *ractl) 1310{ 1311 struct folio *folio; 1312 1313 BUG_ON(ractl->_batch_count > ractl->_nr_pages); 1314 ractl->_nr_pages -= ractl->_batch_count; 1315 ractl->_index += ractl->_batch_count; 1316 1317 if (!ractl->_nr_pages) { 1318 ractl->_batch_count = 0; 1319 return NULL; 1320 } 1321 1322 folio = xa_load(&ractl->mapping->i_pages, ractl->_index); 1323 VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); 1324 ractl->_batch_count = folio_nr_pages(folio); 1325 1326 return folio; 1327} 1328 1329/** 1330 * readahead_page - Get the next page to read. 1331 * @ractl: The current readahead request. 1332 * 1333 * Context: The page is locked and has an elevated refcount. The caller 1334 * should decreases the refcount once the page has been submitted for I/O 1335 * and unlock the page once all I/O to that page has completed. 1336 * Return: A pointer to the next page, or %NULL if we are done. 1337 */ 1338static inline struct page *readahead_page(struct readahead_control *ractl) 1339{ 1340 struct folio *folio = __readahead_folio(ractl); 1341 1342 return &folio->page; 1343} 1344 1345/** 1346 * readahead_folio - Get the next folio to read. 1347 * @ractl: The current readahead request. 1348 * 1349 * Context: The folio is locked. The caller should unlock the folio once 1350 * all I/O to that folio has completed. 1351 * Return: A pointer to the next folio, or %NULL if we are done. 1352 */ 1353static inline struct folio *readahead_folio(struct readahead_control *ractl) 1354{ 1355 struct folio *folio = __readahead_folio(ractl); 1356 1357 if (folio) 1358 folio_put(folio); 1359 return folio; 1360} 1361 1362static inline unsigned int __readahead_batch(struct readahead_control *rac, 1363 struct page **array, unsigned int array_sz) 1364{ 1365 unsigned int i = 0; 1366 XA_STATE(xas, &rac->mapping->i_pages, 0); 1367 struct page *page; 1368 1369 BUG_ON(rac->_batch_count > rac->_nr_pages); 1370 rac->_nr_pages -= rac->_batch_count; 1371 rac->_index += rac->_batch_count; 1372 rac->_batch_count = 0; 1373 1374 xas_set(&xas, rac->_index); 1375 rcu_read_lock(); 1376 xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) { 1377 if (xas_retry(&xas, page)) 1378 continue; 1379 VM_BUG_ON_PAGE(!PageLocked(page), page); 1380 VM_BUG_ON_PAGE(PageTail(page), page); 1381 array[i++] = page; 1382 rac->_batch_count += thp_nr_pages(page); 1383 if (i == array_sz) 1384 break; 1385 } 1386 rcu_read_unlock(); 1387 1388 return i; 1389} 1390 1391/** 1392 * readahead_page_batch - Get a batch of pages to read. 1393 * @rac: The current readahead request. 1394 * @array: An array of pointers to struct page. 1395 * 1396 * Context: The pages are locked and have an elevated refcount. The caller 1397 * should decreases the refcount once the page has been submitted for I/O 1398 * and unlock the page once all I/O to that page has completed. 1399 * Return: The number of pages placed in the array. 0 indicates the request 1400 * is complete. 1401 */ 1402#define readahead_page_batch(rac, array) \ 1403 __readahead_batch(rac, array, ARRAY_SIZE(array)) 1404 1405/** 1406 * readahead_pos - The byte offset into the file of this readahead request. 1407 * @rac: The readahead request. 1408 */ 1409static inline loff_t readahead_pos(struct readahead_control *rac) 1410{ 1411 return (loff_t)rac->_index * PAGE_SIZE; 1412} 1413 1414/** 1415 * readahead_length - The number of bytes in this readahead request. 1416 * @rac: The readahead request. 1417 */ 1418static inline size_t readahead_length(struct readahead_control *rac) 1419{ 1420 return rac->_nr_pages * PAGE_SIZE; 1421} 1422 1423/** 1424 * readahead_index - The index of the first page in this readahead request. 1425 * @rac: The readahead request. 1426 */ 1427static inline pgoff_t readahead_index(struct readahead_control *rac) 1428{ 1429 return rac->_index; 1430} 1431 1432/** 1433 * readahead_count - The number of pages in this readahead request. 1434 * @rac: The readahead request. 1435 */ 1436static inline unsigned int readahead_count(struct readahead_control *rac) 1437{ 1438 return rac->_nr_pages; 1439} 1440 1441/** 1442 * readahead_batch_length - The number of bytes in the current batch. 1443 * @rac: The readahead request. 1444 */ 1445static inline size_t readahead_batch_length(struct readahead_control *rac) 1446{ 1447 return rac->_batch_count * PAGE_SIZE; 1448} 1449 1450static inline unsigned long dir_pages(struct inode *inode) 1451{ 1452 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >> 1453 PAGE_SHIFT; 1454} 1455 1456/** 1457 * folio_mkwrite_check_truncate - check if folio was truncated 1458 * @folio: the folio to check 1459 * @inode: the inode to check the folio against 1460 * 1461 * Return: the number of bytes in the folio up to EOF, 1462 * or -EFAULT if the folio was truncated. 1463 */ 1464static inline ssize_t folio_mkwrite_check_truncate(struct folio *folio, 1465 struct inode *inode) 1466{ 1467 loff_t size = i_size_read(inode); 1468 pgoff_t index = size >> PAGE_SHIFT; 1469 size_t offset = offset_in_folio(folio, size); 1470 1471 if (!folio->mapping) 1472 return -EFAULT; 1473 1474 /* folio is wholly inside EOF */ 1475 if (folio_next_index(folio) - 1 < index) 1476 return folio_size(folio); 1477 /* folio is wholly past EOF */ 1478 if (folio->index > index || !offset) 1479 return -EFAULT; 1480 /* folio is partially inside EOF */ 1481 return offset; 1482} 1483 1484/** 1485 * page_mkwrite_check_truncate - check if page was truncated 1486 * @page: the page to check 1487 * @inode: the inode to check the page against 1488 * 1489 * Returns the number of bytes in the page up to EOF, 1490 * or -EFAULT if the page was truncated. 1491 */ 1492static inline int page_mkwrite_check_truncate(struct page *page, 1493 struct inode *inode) 1494{ 1495 loff_t size = i_size_read(inode); 1496 pgoff_t index = size >> PAGE_SHIFT; 1497 int offset = offset_in_page(size); 1498 1499 if (page->mapping != inode->i_mapping) 1500 return -EFAULT; 1501 1502 /* page is wholly inside EOF */ 1503 if (page->index < index) 1504 return PAGE_SIZE; 1505 /* page is wholly past EOF */ 1506 if (page->index > index || !offset) 1507 return -EFAULT; 1508 /* page is partially inside EOF */ 1509 return offset; 1510} 1511 1512/** 1513 * i_blocks_per_folio - How many blocks fit in this folio. 1514 * @inode: The inode which contains the blocks. 1515 * @folio: The folio. 1516 * 1517 * If the block size is larger than the size of this folio, return zero. 1518 * 1519 * Context: The caller should hold a refcount on the folio to prevent it 1520 * from being split. 1521 * Return: The number of filesystem blocks covered by this folio. 1522 */ 1523static inline 1524unsigned int i_blocks_per_folio(struct inode *inode, struct folio *folio) 1525{ 1526 return folio_size(folio) >> inode->i_blkbits; 1527} 1528 1529static inline 1530unsigned int i_blocks_per_page(struct inode *inode, struct page *page) 1531{ 1532 return i_blocks_per_folio(inode, page_folio(page)); 1533} 1534#endif /* _LINUX_PAGEMAP_H */