1 /* SPDX-License-Identifier: GPL-2.0 */
3 #ifndef BTRFS_EXTENT_IO_H
4 #define BTRFS_EXTENT_IO_H
6 #include <linux/rbtree.h>
7 #include <linux/refcount.h>
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)
34 * flags for bio submission. The high bits indicate the compression
37 #define EXTENT_BIO_COMPRESSED 1
38 #define EXTENT_BIO_FLAG_SHIFT 16
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_UNMAPPED 9
50 #define EXTENT_BUFFER_IN_TREE 10
51 #define EXTENT_BUFFER_WRITE_ERR 11 /* write IO error */
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)
63 * page->private values. Every page that is controlled by the extent
64 * map has page->private set to one.
66 #define EXTENT_PAGE_PRIVATE 1
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.
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)))
86 struct io_failure_record;
88 typedef blk_status_t (extent_submit_bio_hook_t)(void *private_data, struct bio *bio,
89 int mirror_num, unsigned long bio_flags,
92 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
93 struct bio *bio, u64 bio_offset);
95 struct extent_io_ops {
97 * The following callbacks must be allways defined, the function
98 * pointer will be called unconditionally.
100 extent_submit_bio_hook_t *submit_bio_hook;
101 int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
102 struct page *page, u64 start, u64 end,
104 int (*merge_bio_hook)(struct page *page, unsigned long offset,
105 size_t size, struct bio *bio,
106 unsigned long bio_flags);
107 int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
108 struct btrfs_fs_info *(*tree_fs_info)(void *private_data);
109 void (*set_range_writeback)(void *private_data, u64 start, u64 end);
112 * Optional hooks, called if the pointer is not NULL
114 int (*fill_delalloc)(void *private_data, struct page *locked_page,
115 u64 start, u64 end, int *page_started,
116 unsigned long *nr_written,
117 struct writeback_control *wbc);
119 int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
120 void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
121 struct extent_state *state, int uptodate);
122 void (*set_bit_hook)(void *private_data, struct extent_state *state,
124 void (*clear_bit_hook)(void *private_data,
125 struct extent_state *state,
127 void (*merge_extent_hook)(void *private_data,
128 struct extent_state *new,
129 struct extent_state *other);
130 void (*split_extent_hook)(void *private_data,
131 struct extent_state *orig, u64 split);
132 void (*check_extent_io_range)(void *private_data, const char *caller,
136 struct extent_io_tree {
137 struct rb_root state;
142 const struct extent_io_ops *ops;
145 struct extent_state {
147 u64 end; /* inclusive */
148 struct rb_node rb_node;
150 /* ADD NEW ELEMENTS AFTER THIS */
151 wait_queue_head_t wq;
155 struct io_failure_record *failrec;
157 #ifdef CONFIG_BTRFS_DEBUG
158 struct list_head leak_list;
162 #define INLINE_EXTENT_BUFFER_PAGES 16
163 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
164 struct extent_buffer {
167 unsigned long bflags;
168 struct btrfs_fs_info *fs_info;
169 spinlock_t refs_lock;
173 struct rcu_head rcu_head;
176 /* count of read lock holders on the extent buffer */
177 atomic_t write_locks;
179 atomic_t blocking_writers;
180 atomic_t blocking_readers;
181 atomic_t spinning_readers;
182 atomic_t spinning_writers;
184 /* >= 0 if eb belongs to a log tree, -1 otherwise */
187 /* protects write locks */
190 /* readers use lock_wq while they wait for the write
191 * lock holders to unlock
193 wait_queue_head_t write_lock_wq;
195 /* writers use read_lock_wq while they wait for readers
198 wait_queue_head_t read_lock_wq;
199 struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
200 #ifdef CONFIG_BTRFS_DEBUG
201 struct list_head leak_list;
206 * Structure to record how many bytes and which ranges are set/cleared
208 struct extent_changeset {
209 /* How many bytes are set/cleared in this operation */
210 unsigned int bytes_changed;
213 struct ulist range_changed;
216 static inline void extent_changeset_init(struct extent_changeset *changeset)
218 changeset->bytes_changed = 0;
219 ulist_init(&changeset->range_changed);
222 static inline struct extent_changeset *extent_changeset_alloc(void)
224 struct extent_changeset *ret;
226 ret = kmalloc(sizeof(*ret), GFP_KERNEL);
230 extent_changeset_init(ret);
234 static inline void extent_changeset_release(struct extent_changeset *changeset)
238 changeset->bytes_changed = 0;
239 ulist_release(&changeset->range_changed);
242 static inline void extent_changeset_free(struct extent_changeset *changeset)
246 extent_changeset_release(changeset);
250 static inline void extent_set_compress_type(unsigned long *bio_flags,
253 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
256 static inline int extent_compress_type(unsigned long bio_flags)
258 return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
261 struct extent_map_tree;
263 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
269 void extent_io_tree_init(struct extent_io_tree *tree, void *private_data);
270 int try_release_extent_mapping(struct page *page, gfp_t mask);
271 int try_release_extent_buffer(struct page *page);
272 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
273 struct extent_state **cached);
275 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
277 return lock_extent_bits(tree, start, end, NULL);
280 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
281 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
282 get_extent_t *get_extent, int mirror_num);
283 int __init extent_io_init(void);
284 void __cold extent_io_exit(void);
286 u64 count_range_bits(struct extent_io_tree *tree,
287 u64 *start, u64 search_end,
288 u64 max_bytes, unsigned bits, int contig);
290 void free_extent_state(struct extent_state *state);
291 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
292 unsigned bits, int filled,
293 struct extent_state *cached_state);
294 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
295 unsigned bits, struct extent_changeset *changeset);
296 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
297 unsigned bits, int wake, int delete,
298 struct extent_state **cached);
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, gfp_t mask,
302 struct extent_changeset *changeset);
304 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
306 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL);
309 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
310 u64 end, struct extent_state **cached)
312 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
316 static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree,
317 u64 start, u64 end, struct extent_state **cached)
319 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
323 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
324 u64 end, unsigned bits)
328 if (bits & EXTENT_LOCKED)
331 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL);
334 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
335 unsigned bits, struct extent_changeset *changeset);
336 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
337 unsigned bits, u64 *failed_start,
338 struct extent_state **cached_state, gfp_t mask);
340 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
341 u64 end, unsigned bits)
343 return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
346 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
347 u64 end, struct extent_state **cached_state)
349 return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
350 cached_state, GFP_NOFS, NULL);
353 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
356 return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
360 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
363 return clear_extent_bit(tree, start, end,
364 EXTENT_DIRTY | EXTENT_DELALLOC |
365 EXTENT_DO_ACCOUNTING, 0, 0, NULL);
368 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
369 unsigned bits, unsigned clear_bits,
370 struct extent_state **cached_state);
372 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
373 u64 end, unsigned int extra_bits,
374 struct extent_state **cached_state)
376 return set_extent_bit(tree, start, end,
377 EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits,
378 NULL, cached_state, GFP_NOFS);
381 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
382 u64 end, struct extent_state **cached_state)
384 return set_extent_bit(tree, start, end,
385 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
386 NULL, cached_state, GFP_NOFS);
389 static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
392 return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
396 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
397 u64 end, struct extent_state **cached_state, gfp_t mask)
399 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
403 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
404 u64 *start_ret, u64 *end_ret, unsigned bits,
405 struct extent_state **cached_state);
406 int extent_invalidatepage(struct extent_io_tree *tree,
407 struct page *page, unsigned long offset);
408 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
409 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
411 int extent_writepages(struct address_space *mapping,
412 struct writeback_control *wbc);
413 int btree_write_cache_pages(struct address_space *mapping,
414 struct writeback_control *wbc);
415 int extent_readpages(struct address_space *mapping, struct list_head *pages,
417 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
418 __u64 start, __u64 len);
419 void set_page_extent_mapped(struct page *page);
421 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
423 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
424 u64 start, unsigned long len);
425 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
427 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
428 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
430 void free_extent_buffer(struct extent_buffer *eb);
431 void free_extent_buffer_stale(struct extent_buffer *eb);
433 #define WAIT_COMPLETE 1
434 #define WAIT_PAGE_LOCK 2
435 int read_extent_buffer_pages(struct extent_io_tree *tree,
436 struct extent_buffer *eb, int wait,
438 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
440 static inline int num_extent_pages(const struct extent_buffer *eb)
442 return (round_up(eb->start + eb->len, PAGE_SIZE) >> PAGE_SHIFT) -
443 (eb->start >> PAGE_SHIFT);
446 static inline void extent_buffer_get(struct extent_buffer *eb)
448 atomic_inc(&eb->refs);
451 static inline int extent_buffer_uptodate(struct extent_buffer *eb)
453 return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
456 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
457 unsigned long start, unsigned long len);
458 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
461 int read_extent_buffer_to_user(const struct extent_buffer *eb,
462 void __user *dst, unsigned long start,
464 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
465 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
467 void write_extent_buffer(struct extent_buffer *eb, const void *src,
468 unsigned long start, unsigned long len);
469 void copy_extent_buffer_full(struct extent_buffer *dst,
470 struct extent_buffer *src);
471 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
472 unsigned long dst_offset, unsigned long src_offset,
474 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
475 unsigned long src_offset, unsigned long len);
476 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
477 unsigned long src_offset, unsigned long len);
478 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
480 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
482 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
483 unsigned long pos, unsigned long len);
484 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
485 unsigned long pos, unsigned long len);
486 void clear_extent_buffer_dirty(struct extent_buffer *eb);
487 int set_extent_buffer_dirty(struct extent_buffer *eb);
488 void set_extent_buffer_uptodate(struct extent_buffer *eb);
489 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
490 int extent_buffer_under_io(struct extent_buffer *eb);
491 int map_private_extent_buffer(const struct extent_buffer *eb,
492 unsigned long offset, unsigned long min_len,
493 char **map, unsigned long *map_start,
494 unsigned long *map_len);
495 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
496 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
497 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
498 u64 delalloc_end, struct page *locked_page,
499 unsigned bits_to_clear,
500 unsigned long page_ops);
501 struct bio *btrfs_bio_alloc(struct block_device *bdev, u64 first_byte);
502 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
503 struct bio *btrfs_bio_clone(struct bio *bio);
504 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
506 struct btrfs_fs_info;
509 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
510 u64 length, u64 logical, struct page *page,
511 unsigned int pg_offset, int mirror_num);
512 int clean_io_failure(struct btrfs_fs_info *fs_info,
513 struct extent_io_tree *failure_tree,
514 struct extent_io_tree *io_tree, u64 start,
515 struct page *page, u64 ino, unsigned int pg_offset);
516 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
517 int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
518 struct extent_buffer *eb, int mirror_num);
521 * When IO fails, either with EIO or csum verification fails, we
522 * try other mirrors that might have a good copy of the data. This
523 * io_failure_record is used to record state as we go through all the
524 * mirrors. If another mirror has good data, the page is set up to date
525 * and things continue. If a good mirror can't be found, the original
526 * bio end_io callback is called to indicate things have failed.
528 struct io_failure_record {
533 unsigned long bio_flags;
540 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
542 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
543 struct io_failure_record **failrec_ret);
544 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
545 struct io_failure_record *failrec, int fail_mirror);
546 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
547 struct io_failure_record *failrec,
548 struct page *page, int pg_offset, int icsum,
549 bio_end_io_t *endio_func, void *data);
550 int free_io_failure(struct extent_io_tree *failure_tree,
551 struct extent_io_tree *io_tree,
552 struct io_failure_record *rec);
553 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
554 noinline u64 find_lock_delalloc_range(struct inode *inode,
555 struct extent_io_tree *tree,
556 struct page *locked_page, u64 *start,
557 u64 *end, u64 max_bytes);
559 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,