1 /* SPDX-License-Identifier: GPL-2.0 */ 2 3 #ifndef BTRFS_EXTENT_IO_H 4 #define BTRFS_EXTENT_IO_H 5 6 #include <linux/rbtree.h> 7 #include <linux/refcount.h> 8 #include "ulist.h" 9 10 /* bits for the extent state */ 11 #define EXTENT_DIRTY (1U << 0) 12 #define EXTENT_WRITEBACK (1U << 1) 13 #define EXTENT_UPTODATE (1U << 2) 14 #define EXTENT_LOCKED (1U << 3) 15 #define EXTENT_NEW (1U << 4) 16 #define EXTENT_DELALLOC (1U << 5) 17 #define EXTENT_DEFRAG (1U << 6) 18 #define EXTENT_BOUNDARY (1U << 9) 19 #define EXTENT_NODATASUM (1U << 10) 20 #define EXTENT_CLEAR_META_RESV (1U << 11) 21 #define EXTENT_FIRST_DELALLOC (1U << 12) 22 #define EXTENT_NEED_WAIT (1U << 13) 23 #define EXTENT_DAMAGED (1U << 14) 24 #define EXTENT_NORESERVE (1U << 15) 25 #define EXTENT_QGROUP_RESERVED (1U << 16) 26 #define EXTENT_CLEAR_DATA_RESV (1U << 17) 27 #define EXTENT_DELALLOC_NEW (1U << 18) 28 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) 29 #define EXTENT_DO_ACCOUNTING (EXTENT_CLEAR_META_RESV | \ 30 EXTENT_CLEAR_DATA_RESV) 31 #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC) 32 33 /* 34 * flags for bio submission. The high bits indicate the compression 35 * type for this bio 36 */ 37 #define EXTENT_BIO_COMPRESSED 1 38 #define EXTENT_BIO_FLAG_SHIFT 16 39 40 /* these are bit numbers for test/set bit */ 41 #define EXTENT_BUFFER_UPTODATE 0 42 #define EXTENT_BUFFER_DIRTY 2 43 #define EXTENT_BUFFER_CORRUPT 3 44 #define EXTENT_BUFFER_READAHEAD 4 /* this got triggered by readahead */ 45 #define EXTENT_BUFFER_TREE_REF 5 46 #define EXTENT_BUFFER_STALE 6 47 #define EXTENT_BUFFER_WRITEBACK 7 48 #define EXTENT_BUFFER_READ_ERR 8 /* read IO error */ 49 #define EXTENT_BUFFER_DUMMY 9 50 #define EXTENT_BUFFER_IN_TREE 10 51 #define EXTENT_BUFFER_WRITE_ERR 11 /* write IO error */ 52 53 /* these are flags for __process_pages_contig */ 54 #define PAGE_UNLOCK (1 << 0) 55 #define PAGE_CLEAR_DIRTY (1 << 1) 56 #define PAGE_SET_WRITEBACK (1 << 2) 57 #define PAGE_END_WRITEBACK (1 << 3) 58 #define PAGE_SET_PRIVATE2 (1 << 4) 59 #define PAGE_SET_ERROR (1 << 5) 60 #define PAGE_LOCK (1 << 6) 61 62 /* 63 * page->private values. Every page that is controlled by the extent 64 * map has page->private set to one. 65 */ 66 #define EXTENT_PAGE_PRIVATE 1 67 68 /* 69 * The extent buffer bitmap operations are done with byte granularity instead of 70 * word granularity for two reasons: 71 * 1. The bitmaps must be little-endian on disk. 72 * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a 73 * single word in a bitmap may straddle two pages in the extent buffer. 74 */ 75 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE) 76 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1) 77 #define BITMAP_FIRST_BYTE_MASK(start) \ 78 ((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK) 79 #define BITMAP_LAST_BYTE_MASK(nbits) \ 80 (BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1))) 81 82 struct extent_state; 83 struct btrfs_root; 84 struct btrfs_inode; 85 struct btrfs_io_bio; 86 struct io_failure_record; 87 88 typedef blk_status_t (extent_submit_bio_hook_t)(void *private_data, struct bio *bio, 89 int mirror_num, unsigned long bio_flags, 90 u64 bio_offset); 91 92 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data, 93 struct bio *bio, u64 bio_offset); 94 95 typedef blk_status_t (extent_submit_bio_done_t)(void *private_data, 96 struct bio *bio, int mirror_num); 97 98 struct extent_io_ops { 99 /* 100 * The following callbacks must be allways defined, the function 101 * pointer will be called unconditionally. 102 */ 103 extent_submit_bio_hook_t *submit_bio_hook; 104 int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset, 105 struct page *page, u64 start, u64 end, 106 int mirror); 107 int (*merge_bio_hook)(struct page *page, unsigned long offset, 108 size_t size, struct bio *bio, 109 unsigned long bio_flags); 110 int (*readpage_io_failed_hook)(struct page *page, int failed_mirror); 111 struct btrfs_fs_info *(*tree_fs_info)(void *private_data); 112 void (*set_range_writeback)(void *private_data, u64 start, u64 end); 113 114 /* 115 * Optional hooks, called if the pointer is not NULL 116 */ 117 int (*fill_delalloc)(void *private_data, struct page *locked_page, 118 u64 start, u64 end, int *page_started, 119 unsigned long *nr_written, 120 struct writeback_control *wbc); 121 122 int (*writepage_start_hook)(struct page *page, u64 start, u64 end); 123 void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, 124 struct extent_state *state, int uptodate); 125 void (*set_bit_hook)(void *private_data, struct extent_state *state, 126 unsigned *bits); 127 void (*clear_bit_hook)(void *private_data, 128 struct extent_state *state, 129 unsigned *bits); 130 void (*merge_extent_hook)(void *private_data, 131 struct extent_state *new, 132 struct extent_state *other); 133 void (*split_extent_hook)(void *private_data, 134 struct extent_state *orig, u64 split); 135 void (*check_extent_io_range)(void *private_data, const char *caller, 136 u64 start, u64 end); 137 }; 138 139 struct extent_io_tree { 140 struct rb_root state; 141 void *private_data; 142 u64 dirty_bytes; 143 int track_uptodate; 144 spinlock_t lock; 145 const struct extent_io_ops *ops; 146 }; 147 148 struct extent_state { 149 u64 start; 150 u64 end; /* inclusive */ 151 struct rb_node rb_node; 152 153 /* ADD NEW ELEMENTS AFTER THIS */ 154 wait_queue_head_t wq; 155 refcount_t refs; 156 unsigned state; 157 158 struct io_failure_record *failrec; 159 160 #ifdef CONFIG_BTRFS_DEBUG 161 struct list_head leak_list; 162 #endif 163 }; 164 165 #define INLINE_EXTENT_BUFFER_PAGES 16 166 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE) 167 struct extent_buffer { 168 u64 start; 169 unsigned long len; 170 unsigned long bflags; 171 struct btrfs_fs_info *fs_info; 172 spinlock_t refs_lock; 173 atomic_t refs; 174 atomic_t io_pages; 175 int read_mirror; 176 struct rcu_head rcu_head; 177 pid_t lock_owner; 178 179 /* count of read lock holders on the extent buffer */ 180 atomic_t write_locks; 181 atomic_t read_locks; 182 atomic_t blocking_writers; 183 atomic_t blocking_readers; 184 atomic_t spinning_readers; 185 atomic_t spinning_writers; 186 short lock_nested; 187 /* >= 0 if eb belongs to a log tree, -1 otherwise */ 188 short log_index; 189 190 /* protects write locks */ 191 rwlock_t lock; 192 193 /* readers use lock_wq while they wait for the write 194 * lock holders to unlock 195 */ 196 wait_queue_head_t write_lock_wq; 197 198 /* writers use read_lock_wq while they wait for readers 199 * to unlock 200 */ 201 wait_queue_head_t read_lock_wq; 202 struct page *pages[INLINE_EXTENT_BUFFER_PAGES]; 203 #ifdef CONFIG_BTRFS_DEBUG 204 struct list_head leak_list; 205 #endif 206 }; 207 208 /* 209 * Structure to record how many bytes and which ranges are set/cleared 210 */ 211 struct extent_changeset { 212 /* How many bytes are set/cleared in this operation */ 213 unsigned int bytes_changed; 214 215 /* Changed ranges */ 216 struct ulist range_changed; 217 }; 218 219 static inline void extent_changeset_init(struct extent_changeset *changeset) 220 { 221 changeset->bytes_changed = 0; 222 ulist_init(&changeset->range_changed); 223 } 224 225 static inline struct extent_changeset *extent_changeset_alloc(void) 226 { 227 struct extent_changeset *ret; 228 229 ret = kmalloc(sizeof(*ret), GFP_KERNEL); 230 if (!ret) 231 return NULL; 232 233 extent_changeset_init(ret); 234 return ret; 235 } 236 237 static inline void extent_changeset_release(struct extent_changeset *changeset) 238 { 239 if (!changeset) 240 return; 241 changeset->bytes_changed = 0; 242 ulist_release(&changeset->range_changed); 243 } 244 245 static inline void extent_changeset_free(struct extent_changeset *changeset) 246 { 247 if (!changeset) 248 return; 249 extent_changeset_release(changeset); 250 kfree(changeset); 251 } 252 253 static inline void extent_set_compress_type(unsigned long *bio_flags, 254 int compress_type) 255 { 256 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT; 257 } 258 259 static inline int extent_compress_type(unsigned long bio_flags) 260 { 261 return bio_flags >> EXTENT_BIO_FLAG_SHIFT; 262 } 263 264 struct extent_map_tree; 265 266 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode, 267 struct page *page, 268 size_t pg_offset, 269 u64 start, u64 len, 270 int create); 271 272 void extent_io_tree_init(struct extent_io_tree *tree, void *private_data); 273 int try_release_extent_mapping(struct page *page, gfp_t mask); 274 int try_release_extent_buffer(struct page *page); 275 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 276 struct extent_state **cached); 277 278 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end) 279 { 280 return lock_extent_bits(tree, start, end, NULL); 281 } 282 283 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end); 284 int extent_read_full_page(struct extent_io_tree *tree, struct page *page, 285 get_extent_t *get_extent, int mirror_num); 286 int __init extent_io_init(void); 287 void __cold extent_io_exit(void); 288 289 u64 count_range_bits(struct extent_io_tree *tree, 290 u64 *start, u64 search_end, 291 u64 max_bytes, unsigned bits, int contig); 292 293 void free_extent_state(struct extent_state *state); 294 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, 295 unsigned bits, int filled, 296 struct extent_state *cached_state); 297 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 298 unsigned bits, struct extent_changeset *changeset); 299 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 300 unsigned bits, int wake, int delete, 301 struct extent_state **cached); 302 int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 303 unsigned bits, int wake, int delete, 304 struct extent_state **cached, gfp_t mask, 305 struct extent_changeset *changeset); 306 307 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end) 308 { 309 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL); 310 } 311 312 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start, 313 u64 end, struct extent_state **cached) 314 { 315 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached, 316 GFP_NOFS, NULL); 317 } 318 319 static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree, 320 u64 start, u64 end, struct extent_state **cached) 321 { 322 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached, 323 GFP_ATOMIC, NULL); 324 } 325 326 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start, 327 u64 end, unsigned bits) 328 { 329 int wake = 0; 330 331 if (bits & EXTENT_LOCKED) 332 wake = 1; 333 334 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL); 335 } 336 337 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 338 unsigned bits, struct extent_changeset *changeset); 339 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 340 unsigned bits, u64 *failed_start, 341 struct extent_state **cached_state, gfp_t mask); 342 343 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start, 344 u64 end, unsigned bits) 345 { 346 return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS); 347 } 348 349 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, 350 u64 end, struct extent_state **cached_state) 351 { 352 return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, 353 cached_state, GFP_NOFS, NULL); 354 } 355 356 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start, 357 u64 end, gfp_t mask) 358 { 359 return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL, 360 NULL, mask); 361 } 362 363 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start, 364 u64 end) 365 { 366 return clear_extent_bit(tree, start, end, 367 EXTENT_DIRTY | EXTENT_DELALLOC | 368 EXTENT_DO_ACCOUNTING, 0, 0, NULL); 369 } 370 371 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 372 unsigned bits, unsigned clear_bits, 373 struct extent_state **cached_state); 374 375 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start, 376 u64 end, unsigned int extra_bits, 377 struct extent_state **cached_state) 378 { 379 return set_extent_bit(tree, start, end, 380 EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits, 381 NULL, cached_state, GFP_NOFS); 382 } 383 384 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start, 385 u64 end, struct extent_state **cached_state) 386 { 387 return set_extent_bit(tree, start, end, 388 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG, 389 NULL, cached_state, GFP_NOFS); 390 } 391 392 static inline int set_extent_new(struct extent_io_tree *tree, u64 start, 393 u64 end) 394 { 395 return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL, 396 GFP_NOFS); 397 } 398 399 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start, 400 u64 end, struct extent_state **cached_state, gfp_t mask) 401 { 402 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL, 403 cached_state, mask); 404 } 405 406 int find_first_extent_bit(struct extent_io_tree *tree, u64 start, 407 u64 *start_ret, u64 *end_ret, unsigned bits, 408 struct extent_state **cached_state); 409 int extent_invalidatepage(struct extent_io_tree *tree, 410 struct page *page, unsigned long offset); 411 int extent_write_full_page(struct page *page, struct writeback_control *wbc); 412 int extent_write_locked_range(struct inode *inode, u64 start, u64 end, 413 int mode); 414 int extent_writepages(struct address_space *mapping, 415 struct writeback_control *wbc); 416 int btree_write_cache_pages(struct address_space *mapping, 417 struct writeback_control *wbc); 418 int extent_readpages(struct address_space *mapping, struct list_head *pages, 419 unsigned nr_pages); 420 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 421 __u64 start, __u64 len); 422 void set_page_extent_mapped(struct page *page); 423 424 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info, 425 u64 start); 426 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 427 u64 start, unsigned long len); 428 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 429 u64 start); 430 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src); 431 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info, 432 u64 start); 433 void free_extent_buffer(struct extent_buffer *eb); 434 void free_extent_buffer_stale(struct extent_buffer *eb); 435 #define WAIT_NONE 0 436 #define WAIT_COMPLETE 1 437 #define WAIT_PAGE_LOCK 2 438 int read_extent_buffer_pages(struct extent_io_tree *tree, 439 struct extent_buffer *eb, int wait, 440 int mirror_num); 441 void wait_on_extent_buffer_writeback(struct extent_buffer *eb); 442 443 static inline unsigned long num_extent_pages(u64 start, u64 len) 444 { 445 return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) - 446 (start >> PAGE_SHIFT); 447 } 448 449 static inline void extent_buffer_get(struct extent_buffer *eb) 450 { 451 atomic_inc(&eb->refs); 452 } 453 454 static inline int extent_buffer_uptodate(struct extent_buffer *eb) 455 { 456 return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); 457 } 458 459 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv, 460 unsigned long start, unsigned long len); 461 void read_extent_buffer(const struct extent_buffer *eb, void *dst, 462 unsigned long start, 463 unsigned long len); 464 int read_extent_buffer_to_user(const struct extent_buffer *eb, 465 void __user *dst, unsigned long start, 466 unsigned long len); 467 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src); 468 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb, 469 const void *src); 470 void write_extent_buffer(struct extent_buffer *eb, const void *src, 471 unsigned long start, unsigned long len); 472 void copy_extent_buffer_full(struct extent_buffer *dst, 473 struct extent_buffer *src); 474 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, 475 unsigned long dst_offset, unsigned long src_offset, 476 unsigned long len); 477 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, 478 unsigned long src_offset, unsigned long len); 479 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, 480 unsigned long src_offset, unsigned long len); 481 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start, 482 unsigned long len); 483 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start, 484 unsigned long pos); 485 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start, 486 unsigned long pos, unsigned long len); 487 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start, 488 unsigned long pos, unsigned long len); 489 void clear_extent_buffer_dirty(struct extent_buffer *eb); 490 int set_extent_buffer_dirty(struct extent_buffer *eb); 491 void set_extent_buffer_uptodate(struct extent_buffer *eb); 492 void clear_extent_buffer_uptodate(struct extent_buffer *eb); 493 int extent_buffer_under_io(struct extent_buffer *eb); 494 int map_private_extent_buffer(const struct extent_buffer *eb, 495 unsigned long offset, unsigned long min_len, 496 char **map, unsigned long *map_start, 497 unsigned long *map_len); 498 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end); 499 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end); 500 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end, 501 u64 delalloc_end, struct page *locked_page, 502 unsigned bits_to_clear, 503 unsigned long page_ops); 504 struct bio *btrfs_bio_alloc(struct block_device *bdev, u64 first_byte); 505 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs); 506 struct bio *btrfs_bio_clone(struct bio *bio); 507 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size); 508 509 struct btrfs_fs_info; 510 struct btrfs_inode; 511 512 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start, 513 u64 length, u64 logical, struct page *page, 514 unsigned int pg_offset, int mirror_num); 515 int clean_io_failure(struct btrfs_fs_info *fs_info, 516 struct extent_io_tree *failure_tree, 517 struct extent_io_tree *io_tree, u64 start, 518 struct page *page, u64 ino, unsigned int pg_offset); 519 void end_extent_writepage(struct page *page, int err, u64 start, u64 end); 520 int repair_eb_io_failure(struct btrfs_fs_info *fs_info, 521 struct extent_buffer *eb, int mirror_num); 522 523 /* 524 * When IO fails, either with EIO or csum verification fails, we 525 * try other mirrors that might have a good copy of the data. This 526 * io_failure_record is used to record state as we go through all the 527 * mirrors. If another mirror has good data, the page is set up to date 528 * and things continue. If a good mirror can't be found, the original 529 * bio end_io callback is called to indicate things have failed. 530 */ 531 struct io_failure_record { 532 struct page *page; 533 u64 start; 534 u64 len; 535 u64 logical; 536 unsigned long bio_flags; 537 int this_mirror; 538 int failed_mirror; 539 int in_validation; 540 }; 541 542 543 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start, 544 u64 end); 545 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end, 546 struct io_failure_record **failrec_ret); 547 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages, 548 struct io_failure_record *failrec, int fail_mirror); 549 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio, 550 struct io_failure_record *failrec, 551 struct page *page, int pg_offset, int icsum, 552 bio_end_io_t *endio_func, void *data); 553 int free_io_failure(struct extent_io_tree *failure_tree, 554 struct extent_io_tree *io_tree, 555 struct io_failure_record *rec); 556 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 557 noinline u64 find_lock_delalloc_range(struct inode *inode, 558 struct extent_io_tree *tree, 559 struct page *locked_page, u64 *start, 560 u64 *end, u64 max_bytes); 561 #endif 562 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info, 563 u64 start); 564 565 #endif 566