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