1 #ifndef __EXTENTIO__ 2 #define __EXTENTIO__ 3 4 #include <linux/rbtree.h> 5 #include "ulist.h" 6 7 /* bits for the extent state */ 8 #define EXTENT_DIRTY (1U << 0) 9 #define EXTENT_WRITEBACK (1U << 1) 10 #define EXTENT_UPTODATE (1U << 2) 11 #define EXTENT_LOCKED (1U << 3) 12 #define EXTENT_NEW (1U << 4) 13 #define EXTENT_DELALLOC (1U << 5) 14 #define EXTENT_DEFRAG (1U << 6) 15 #define EXTENT_BOUNDARY (1U << 9) 16 #define EXTENT_NODATASUM (1U << 10) 17 #define EXTENT_DO_ACCOUNTING (1U << 11) 18 #define EXTENT_FIRST_DELALLOC (1U << 12) 19 #define EXTENT_NEED_WAIT (1U << 13) 20 #define EXTENT_DAMAGED (1U << 14) 21 #define EXTENT_NORESERVE (1U << 15) 22 #define EXTENT_QGROUP_RESERVED (1U << 16) 23 #define EXTENT_CLEAR_DATA_RESV (1U << 17) 24 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) 25 #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC) 26 27 /* 28 * flags for bio submission. The high bits indicate the compression 29 * type for this bio 30 */ 31 #define EXTENT_BIO_COMPRESSED 1 32 #define EXTENT_BIO_TREE_LOG 2 33 #define EXTENT_BIO_FLAG_SHIFT 16 34 35 /* these are bit numbers for test/set bit */ 36 #define EXTENT_BUFFER_UPTODATE 0 37 #define EXTENT_BUFFER_DIRTY 2 38 #define EXTENT_BUFFER_CORRUPT 3 39 #define EXTENT_BUFFER_READAHEAD 4 /* this got triggered by readahead */ 40 #define EXTENT_BUFFER_TREE_REF 5 41 #define EXTENT_BUFFER_STALE 6 42 #define EXTENT_BUFFER_WRITEBACK 7 43 #define EXTENT_BUFFER_READ_ERR 8 /* read IO error */ 44 #define EXTENT_BUFFER_DUMMY 9 45 #define EXTENT_BUFFER_IN_TREE 10 46 #define EXTENT_BUFFER_WRITE_ERR 11 /* write IO error */ 47 48 /* these are flags for extent_clear_unlock_delalloc */ 49 #define PAGE_UNLOCK (1 << 0) 50 #define PAGE_CLEAR_DIRTY (1 << 1) 51 #define PAGE_SET_WRITEBACK (1 << 2) 52 #define PAGE_END_WRITEBACK (1 << 3) 53 #define PAGE_SET_PRIVATE2 (1 << 4) 54 #define PAGE_SET_ERROR (1 << 5) 55 56 /* 57 * page->private values. Every page that is controlled by the extent 58 * map has page->private set to one. 59 */ 60 #define EXTENT_PAGE_PRIVATE 1 61 62 /* 63 * The extent buffer bitmap operations are done with byte granularity instead of 64 * word granularity for two reasons: 65 * 1. The bitmaps must be little-endian on disk. 66 * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a 67 * single word in a bitmap may straddle two pages in the extent buffer. 68 */ 69 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE) 70 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1) 71 #define BITMAP_FIRST_BYTE_MASK(start) \ 72 ((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK) 73 #define BITMAP_LAST_BYTE_MASK(nbits) \ 74 (BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1))) 75 76 static inline int le_test_bit(int nr, const u8 *addr) 77 { 78 return 1U & (addr[BIT_BYTE(nr)] >> (nr & (BITS_PER_BYTE-1))); 79 } 80 81 extern void le_bitmap_set(u8 *map, unsigned int start, int len); 82 extern void le_bitmap_clear(u8 *map, unsigned int start, int len); 83 84 struct extent_state; 85 struct btrfs_root; 86 struct btrfs_io_bio; 87 struct io_failure_record; 88 89 typedef int (extent_submit_bio_hook_t)(struct inode *inode, struct bio *bio, 90 int mirror_num, unsigned long bio_flags, 91 u64 bio_offset); 92 struct extent_io_ops { 93 int (*fill_delalloc)(struct inode *inode, struct page *locked_page, 94 u64 start, u64 end, int *page_started, 95 unsigned long *nr_written); 96 int (*writepage_start_hook)(struct page *page, u64 start, u64 end); 97 extent_submit_bio_hook_t *submit_bio_hook; 98 int (*merge_bio_hook)(struct page *page, unsigned long offset, 99 size_t size, struct bio *bio, 100 unsigned long bio_flags); 101 int (*readpage_io_failed_hook)(struct page *page, int failed_mirror); 102 int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset, 103 struct page *page, u64 start, u64 end, 104 int mirror); 105 int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, 106 struct extent_state *state, int uptodate); 107 void (*set_bit_hook)(struct inode *inode, struct extent_state *state, 108 unsigned *bits); 109 void (*clear_bit_hook)(struct inode *inode, struct extent_state *state, 110 unsigned *bits); 111 void (*merge_extent_hook)(struct inode *inode, 112 struct extent_state *new, 113 struct extent_state *other); 114 void (*split_extent_hook)(struct inode *inode, 115 struct extent_state *orig, u64 split); 116 }; 117 118 struct extent_io_tree { 119 struct rb_root state; 120 struct address_space *mapping; 121 u64 dirty_bytes; 122 int track_uptodate; 123 spinlock_t lock; 124 const struct extent_io_ops *ops; 125 }; 126 127 struct extent_state { 128 u64 start; 129 u64 end; /* inclusive */ 130 struct rb_node rb_node; 131 132 /* ADD NEW ELEMENTS AFTER THIS */ 133 wait_queue_head_t wq; 134 atomic_t refs; 135 unsigned state; 136 137 struct io_failure_record *failrec; 138 139 #ifdef CONFIG_BTRFS_DEBUG 140 struct list_head leak_list; 141 #endif 142 }; 143 144 #define INLINE_EXTENT_BUFFER_PAGES 16 145 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE) 146 struct extent_buffer { 147 u64 start; 148 unsigned long len; 149 unsigned long bflags; 150 struct btrfs_fs_info *fs_info; 151 spinlock_t refs_lock; 152 atomic_t refs; 153 atomic_t io_pages; 154 int read_mirror; 155 struct rcu_head rcu_head; 156 pid_t lock_owner; 157 158 /* count of read lock holders on the extent buffer */ 159 atomic_t write_locks; 160 atomic_t read_locks; 161 atomic_t blocking_writers; 162 atomic_t blocking_readers; 163 atomic_t spinning_readers; 164 atomic_t spinning_writers; 165 short lock_nested; 166 /* >= 0 if eb belongs to a log tree, -1 otherwise */ 167 short log_index; 168 169 /* protects write locks */ 170 rwlock_t lock; 171 172 /* readers use lock_wq while they wait for the write 173 * lock holders to unlock 174 */ 175 wait_queue_head_t write_lock_wq; 176 177 /* writers use read_lock_wq while they wait for readers 178 * to unlock 179 */ 180 wait_queue_head_t read_lock_wq; 181 struct page *pages[INLINE_EXTENT_BUFFER_PAGES]; 182 #ifdef CONFIG_BTRFS_DEBUG 183 struct list_head leak_list; 184 #endif 185 }; 186 187 /* 188 * Structure to record how many bytes and which ranges are set/cleared 189 */ 190 struct extent_changeset { 191 /* How many bytes are set/cleared in this operation */ 192 u64 bytes_changed; 193 194 /* Changed ranges */ 195 struct ulist *range_changed; 196 }; 197 198 static inline void extent_set_compress_type(unsigned long *bio_flags, 199 int compress_type) 200 { 201 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT; 202 } 203 204 static inline int extent_compress_type(unsigned long bio_flags) 205 { 206 return bio_flags >> EXTENT_BIO_FLAG_SHIFT; 207 } 208 209 struct extent_map_tree; 210 211 typedef struct extent_map *(get_extent_t)(struct inode *inode, 212 struct page *page, 213 size_t pg_offset, 214 u64 start, u64 len, 215 int create); 216 217 void extent_io_tree_init(struct extent_io_tree *tree, 218 struct address_space *mapping); 219 int try_release_extent_mapping(struct extent_map_tree *map, 220 struct extent_io_tree *tree, struct page *page, 221 gfp_t mask); 222 int try_release_extent_buffer(struct page *page); 223 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 224 struct extent_state **cached); 225 226 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end) 227 { 228 return lock_extent_bits(tree, start, end, NULL); 229 } 230 231 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end); 232 int extent_read_full_page(struct extent_io_tree *tree, struct page *page, 233 get_extent_t *get_extent, int mirror_num); 234 int __init extent_io_init(void); 235 void extent_io_exit(void); 236 237 u64 count_range_bits(struct extent_io_tree *tree, 238 u64 *start, u64 search_end, 239 u64 max_bytes, unsigned bits, int contig); 240 241 void free_extent_state(struct extent_state *state); 242 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, 243 unsigned bits, int filled, 244 struct extent_state *cached_state); 245 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 246 unsigned bits, struct extent_changeset *changeset); 247 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 248 unsigned bits, int wake, int delete, 249 struct extent_state **cached, gfp_t mask); 250 251 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end) 252 { 253 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL, 254 GFP_NOFS); 255 } 256 257 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start, 258 u64 end, struct extent_state **cached, gfp_t mask) 259 { 260 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached, 261 mask); 262 } 263 264 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start, 265 u64 end, unsigned bits) 266 { 267 int wake = 0; 268 269 if (bits & EXTENT_LOCKED) 270 wake = 1; 271 272 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL, 273 GFP_NOFS); 274 } 275 276 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, 277 unsigned bits, struct extent_changeset *changeset); 278 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 279 unsigned bits, u64 *failed_start, 280 struct extent_state **cached_state, gfp_t mask); 281 282 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start, 283 u64 end, unsigned bits) 284 { 285 return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS); 286 } 287 288 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, 289 u64 end, struct extent_state **cached_state, gfp_t mask) 290 { 291 return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, 292 cached_state, mask); 293 } 294 295 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start, 296 u64 end, gfp_t mask) 297 { 298 return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL, 299 NULL, mask); 300 } 301 302 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start, 303 u64 end) 304 { 305 return clear_extent_bit(tree, start, end, 306 EXTENT_DIRTY | EXTENT_DELALLOC | 307 EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS); 308 } 309 310 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, 311 unsigned bits, unsigned clear_bits, 312 struct extent_state **cached_state); 313 314 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start, 315 u64 end, struct extent_state **cached_state) 316 { 317 return set_extent_bit(tree, start, end, 318 EXTENT_DELALLOC | EXTENT_UPTODATE, 319 NULL, cached_state, GFP_NOFS); 320 } 321 322 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start, 323 u64 end, struct extent_state **cached_state) 324 { 325 return set_extent_bit(tree, start, end, 326 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG, 327 NULL, cached_state, GFP_NOFS); 328 } 329 330 static inline int set_extent_new(struct extent_io_tree *tree, u64 start, 331 u64 end) 332 { 333 return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL, 334 GFP_NOFS); 335 } 336 337 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start, 338 u64 end, struct extent_state **cached_state, gfp_t mask) 339 { 340 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL, 341 cached_state, mask); 342 } 343 344 int find_first_extent_bit(struct extent_io_tree *tree, u64 start, 345 u64 *start_ret, u64 *end_ret, unsigned bits, 346 struct extent_state **cached_state); 347 int extent_invalidatepage(struct extent_io_tree *tree, 348 struct page *page, unsigned long offset); 349 int extent_write_full_page(struct extent_io_tree *tree, struct page *page, 350 get_extent_t *get_extent, 351 struct writeback_control *wbc); 352 int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, 353 u64 start, u64 end, get_extent_t *get_extent, 354 int mode); 355 int extent_writepages(struct extent_io_tree *tree, 356 struct address_space *mapping, 357 get_extent_t *get_extent, 358 struct writeback_control *wbc); 359 int btree_write_cache_pages(struct address_space *mapping, 360 struct writeback_control *wbc); 361 int extent_readpages(struct extent_io_tree *tree, 362 struct address_space *mapping, 363 struct list_head *pages, unsigned nr_pages, 364 get_extent_t get_extent); 365 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 366 __u64 start, __u64 len, get_extent_t *get_extent); 367 void set_page_extent_mapped(struct page *page); 368 369 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info, 370 u64 start); 371 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 372 u64 start, unsigned long len); 373 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, 374 u64 start, u32 nodesize); 375 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src); 376 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info, 377 u64 start); 378 void free_extent_buffer(struct extent_buffer *eb); 379 void free_extent_buffer_stale(struct extent_buffer *eb); 380 #define WAIT_NONE 0 381 #define WAIT_COMPLETE 1 382 #define WAIT_PAGE_LOCK 2 383 int read_extent_buffer_pages(struct extent_io_tree *tree, 384 struct extent_buffer *eb, int wait, 385 get_extent_t *get_extent, int mirror_num); 386 void wait_on_extent_buffer_writeback(struct extent_buffer *eb); 387 388 static inline unsigned long num_extent_pages(u64 start, u64 len) 389 { 390 return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) - 391 (start >> PAGE_SHIFT); 392 } 393 394 static inline void extent_buffer_get(struct extent_buffer *eb) 395 { 396 atomic_inc(&eb->refs); 397 } 398 399 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, 400 unsigned long start, 401 unsigned long len); 402 void read_extent_buffer(struct extent_buffer *eb, void *dst, 403 unsigned long start, 404 unsigned long len); 405 int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dst, 406 unsigned long start, 407 unsigned long len); 408 void write_extent_buffer(struct extent_buffer *eb, const void *src, 409 unsigned long start, unsigned long len); 410 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, 411 unsigned long dst_offset, unsigned long src_offset, 412 unsigned long len); 413 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, 414 unsigned long src_offset, unsigned long len); 415 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, 416 unsigned long src_offset, unsigned long len); 417 void memset_extent_buffer(struct extent_buffer *eb, char c, 418 unsigned long start, unsigned long len); 419 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start, 420 unsigned long pos); 421 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start, 422 unsigned long pos, unsigned long len); 423 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start, 424 unsigned long pos, unsigned long len); 425 void clear_extent_buffer_dirty(struct extent_buffer *eb); 426 int set_extent_buffer_dirty(struct extent_buffer *eb); 427 void set_extent_buffer_uptodate(struct extent_buffer *eb); 428 void clear_extent_buffer_uptodate(struct extent_buffer *eb); 429 int extent_buffer_uptodate(struct extent_buffer *eb); 430 int extent_buffer_under_io(struct extent_buffer *eb); 431 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset, 432 unsigned long min_len, char **map, 433 unsigned long *map_start, 434 unsigned long *map_len); 435 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end); 436 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end); 437 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end, 438 u64 delalloc_end, struct page *locked_page, 439 unsigned bits_to_clear, 440 unsigned long page_ops); 441 struct bio * 442 btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, 443 gfp_t gfp_flags); 444 struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs); 445 struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask); 446 447 struct btrfs_fs_info; 448 449 int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical, 450 struct page *page, unsigned int pg_offset, 451 int mirror_num); 452 int clean_io_failure(struct inode *inode, u64 start, struct page *page, 453 unsigned int pg_offset); 454 void end_extent_writepage(struct page *page, int err, u64 start, u64 end); 455 int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb, 456 int mirror_num); 457 458 /* 459 * When IO fails, either with EIO or csum verification fails, we 460 * try other mirrors that might have a good copy of the data. This 461 * io_failure_record is used to record state as we go through all the 462 * mirrors. If another mirror has good data, the page is set up to date 463 * and things continue. If a good mirror can't be found, the original 464 * bio end_io callback is called to indicate things have failed. 465 */ 466 struct io_failure_record { 467 struct page *page; 468 u64 start; 469 u64 len; 470 u64 logical; 471 unsigned long bio_flags; 472 int this_mirror; 473 int failed_mirror; 474 int in_validation; 475 }; 476 477 void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end); 478 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end, 479 struct io_failure_record **failrec_ret); 480 int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio, 481 struct io_failure_record *failrec, int fail_mirror); 482 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio, 483 struct io_failure_record *failrec, 484 struct page *page, int pg_offset, int icsum, 485 bio_end_io_t *endio_func, void *data); 486 int free_io_failure(struct inode *inode, struct io_failure_record *rec); 487 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 488 noinline u64 find_lock_delalloc_range(struct inode *inode, 489 struct extent_io_tree *tree, 490 struct page *locked_page, u64 *start, 491 u64 *end, u64 max_bytes); 492 #endif 493 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info, 494 u64 start, u32 nodesize); 495 #endif 496