xref: /linux/fs/btrfs/extent_io.h (revision bd628c1bed7902ec1f24ba0fe70758949146abbe)
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 enum {
41 	EXTENT_BUFFER_UPTODATE,
42 	EXTENT_BUFFER_DIRTY,
43 	EXTENT_BUFFER_CORRUPT,
44 	/* this got triggered by readahead */
45 	EXTENT_BUFFER_READAHEAD,
46 	EXTENT_BUFFER_TREE_REF,
47 	EXTENT_BUFFER_STALE,
48 	EXTENT_BUFFER_WRITEBACK,
49 	/* read IO error */
50 	EXTENT_BUFFER_READ_ERR,
51 	EXTENT_BUFFER_UNMAPPED,
52 	EXTENT_BUFFER_IN_TREE,
53 	/* write IO error */
54 	EXTENT_BUFFER_WRITE_ERR,
55 };
56 
57 /* these are flags for __process_pages_contig */
58 #define PAGE_UNLOCK		(1 << 0)
59 #define PAGE_CLEAR_DIRTY	(1 << 1)
60 #define PAGE_SET_WRITEBACK	(1 << 2)
61 #define PAGE_END_WRITEBACK	(1 << 3)
62 #define PAGE_SET_PRIVATE2	(1 << 4)
63 #define PAGE_SET_ERROR		(1 << 5)
64 #define PAGE_LOCK		(1 << 6)
65 
66 /*
67  * page->private values.  Every page that is controlled by the extent
68  * map has page->private set to one.
69  */
70 #define EXTENT_PAGE_PRIVATE 1
71 
72 /*
73  * The extent buffer bitmap operations are done with byte granularity instead of
74  * word granularity for two reasons:
75  * 1. The bitmaps must be little-endian on disk.
76  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
77  *    single word in a bitmap may straddle two pages in the extent buffer.
78  */
79 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
80 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
81 #define BITMAP_FIRST_BYTE_MASK(start) \
82 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
83 #define BITMAP_LAST_BYTE_MASK(nbits) \
84 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
85 
86 struct extent_state;
87 struct btrfs_root;
88 struct btrfs_inode;
89 struct btrfs_io_bio;
90 struct io_failure_record;
91 
92 typedef	blk_status_t (extent_submit_bio_hook_t)(void *private_data, struct bio *bio,
93 				       int mirror_num, unsigned long bio_flags,
94 				       u64 bio_offset);
95 
96 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
97 		struct bio *bio, u64 bio_offset);
98 
99 struct extent_io_ops {
100 	/*
101 	 * The following callbacks must be always defined, the function
102 	 * pointer will be called unconditionally.
103 	 */
104 	extent_submit_bio_hook_t *submit_bio_hook;
105 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
106 				    struct page *page, u64 start, u64 end,
107 				    int mirror);
108 };
109 
110 struct extent_io_tree {
111 	struct rb_root state;
112 	void *private_data;
113 	u64 dirty_bytes;
114 	int track_uptodate;
115 	spinlock_t lock;
116 	const struct extent_io_ops *ops;
117 };
118 
119 struct extent_state {
120 	u64 start;
121 	u64 end; /* inclusive */
122 	struct rb_node rb_node;
123 
124 	/* ADD NEW ELEMENTS AFTER THIS */
125 	wait_queue_head_t wq;
126 	refcount_t refs;
127 	unsigned state;
128 
129 	struct io_failure_record *failrec;
130 
131 #ifdef CONFIG_BTRFS_DEBUG
132 	struct list_head leak_list;
133 #endif
134 };
135 
136 #define INLINE_EXTENT_BUFFER_PAGES 16
137 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
138 struct extent_buffer {
139 	u64 start;
140 	unsigned long len;
141 	unsigned long bflags;
142 	struct btrfs_fs_info *fs_info;
143 	spinlock_t refs_lock;
144 	atomic_t refs;
145 	atomic_t io_pages;
146 	int read_mirror;
147 	struct rcu_head rcu_head;
148 	pid_t lock_owner;
149 
150 	/* count of read lock holders on the extent buffer */
151 	atomic_t write_locks;
152 	atomic_t read_locks;
153 	atomic_t blocking_writers;
154 	atomic_t blocking_readers;
155 	atomic_t spinning_readers;
156 	atomic_t spinning_writers;
157 	short lock_nested;
158 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
159 	short log_index;
160 
161 	/* protects write locks */
162 	rwlock_t lock;
163 
164 	/* readers use lock_wq while they wait for the write
165 	 * lock holders to unlock
166 	 */
167 	wait_queue_head_t write_lock_wq;
168 
169 	/* writers use read_lock_wq while they wait for readers
170 	 * to unlock
171 	 */
172 	wait_queue_head_t read_lock_wq;
173 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
174 #ifdef CONFIG_BTRFS_DEBUG
175 	struct list_head leak_list;
176 #endif
177 };
178 
179 /*
180  * Structure to record how many bytes and which ranges are set/cleared
181  */
182 struct extent_changeset {
183 	/* How many bytes are set/cleared in this operation */
184 	unsigned int bytes_changed;
185 
186 	/* Changed ranges */
187 	struct ulist range_changed;
188 };
189 
190 static inline void extent_changeset_init(struct extent_changeset *changeset)
191 {
192 	changeset->bytes_changed = 0;
193 	ulist_init(&changeset->range_changed);
194 }
195 
196 static inline struct extent_changeset *extent_changeset_alloc(void)
197 {
198 	struct extent_changeset *ret;
199 
200 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
201 	if (!ret)
202 		return NULL;
203 
204 	extent_changeset_init(ret);
205 	return ret;
206 }
207 
208 static inline void extent_changeset_release(struct extent_changeset *changeset)
209 {
210 	if (!changeset)
211 		return;
212 	changeset->bytes_changed = 0;
213 	ulist_release(&changeset->range_changed);
214 }
215 
216 static inline void extent_changeset_free(struct extent_changeset *changeset)
217 {
218 	if (!changeset)
219 		return;
220 	extent_changeset_release(changeset);
221 	kfree(changeset);
222 }
223 
224 static inline void extent_set_compress_type(unsigned long *bio_flags,
225 					    int compress_type)
226 {
227 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
228 }
229 
230 static inline int extent_compress_type(unsigned long bio_flags)
231 {
232 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
233 }
234 
235 struct extent_map_tree;
236 
237 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
238 					  struct page *page,
239 					  size_t pg_offset,
240 					  u64 start, u64 len,
241 					  int create);
242 
243 void extent_io_tree_init(struct extent_io_tree *tree, void *private_data);
244 int try_release_extent_mapping(struct page *page, gfp_t mask);
245 int try_release_extent_buffer(struct page *page);
246 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
247 		     struct extent_state **cached);
248 
249 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
250 {
251 	return lock_extent_bits(tree, start, end, NULL);
252 }
253 
254 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
255 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
256 			  get_extent_t *get_extent, int mirror_num);
257 int __init extent_io_init(void);
258 void __cold extent_io_exit(void);
259 
260 u64 count_range_bits(struct extent_io_tree *tree,
261 		     u64 *start, u64 search_end,
262 		     u64 max_bytes, unsigned bits, int contig);
263 
264 void free_extent_state(struct extent_state *state);
265 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
266 		   unsigned bits, int filled,
267 		   struct extent_state *cached_state);
268 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
269 		unsigned bits, struct extent_changeset *changeset);
270 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
271 		     unsigned bits, int wake, int delete,
272 		     struct extent_state **cached);
273 int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
274 		     unsigned bits, int wake, int delete,
275 		     struct extent_state **cached, gfp_t mask,
276 		     struct extent_changeset *changeset);
277 
278 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
279 {
280 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL);
281 }
282 
283 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
284 		u64 end, struct extent_state **cached)
285 {
286 	return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
287 				GFP_NOFS, NULL);
288 }
289 
290 static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree,
291 		u64 start, u64 end, struct extent_state **cached)
292 {
293 	return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
294 				GFP_ATOMIC, NULL);
295 }
296 
297 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
298 		u64 end, unsigned bits)
299 {
300 	int wake = 0;
301 
302 	if (bits & EXTENT_LOCKED)
303 		wake = 1;
304 
305 	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL);
306 }
307 
308 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
309 			   unsigned bits, struct extent_changeset *changeset);
310 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
311 		   unsigned bits, u64 *failed_start,
312 		   struct extent_state **cached_state, gfp_t mask);
313 
314 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
315 		u64 end, unsigned bits)
316 {
317 	return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
318 }
319 
320 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
321 		u64 end, struct extent_state **cached_state)
322 {
323 	return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
324 				cached_state, GFP_NOFS, NULL);
325 }
326 
327 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
328 		u64 end, gfp_t mask)
329 {
330 	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
331 			      NULL, mask);
332 }
333 
334 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
335 				     u64 end, struct extent_state **cached)
336 {
337 	return clear_extent_bit(tree, start, end,
338 				EXTENT_DIRTY | EXTENT_DELALLOC |
339 				EXTENT_DO_ACCOUNTING, 0, 0, cached);
340 }
341 
342 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
343 		       unsigned bits, unsigned clear_bits,
344 		       struct extent_state **cached_state);
345 
346 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
347 				      u64 end, unsigned int extra_bits,
348 				      struct extent_state **cached_state)
349 {
350 	return set_extent_bit(tree, start, end,
351 			      EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits,
352 			      NULL, cached_state, GFP_NOFS);
353 }
354 
355 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
356 		u64 end, struct extent_state **cached_state)
357 {
358 	return set_extent_bit(tree, start, end,
359 			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
360 			      NULL, cached_state, GFP_NOFS);
361 }
362 
363 static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
364 		u64 end)
365 {
366 	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
367 			GFP_NOFS);
368 }
369 
370 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
371 		u64 end, struct extent_state **cached_state, gfp_t mask)
372 {
373 	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
374 			      cached_state, mask);
375 }
376 
377 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
378 			  u64 *start_ret, u64 *end_ret, unsigned bits,
379 			  struct extent_state **cached_state);
380 int extent_invalidatepage(struct extent_io_tree *tree,
381 			  struct page *page, unsigned long offset);
382 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
383 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
384 			      int mode);
385 int extent_writepages(struct address_space *mapping,
386 		      struct writeback_control *wbc);
387 int btree_write_cache_pages(struct address_space *mapping,
388 			    struct writeback_control *wbc);
389 int extent_readpages(struct address_space *mapping, struct list_head *pages,
390 		     unsigned nr_pages);
391 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
392 		__u64 start, __u64 len);
393 void set_page_extent_mapped(struct page *page);
394 
395 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
396 					  u64 start);
397 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
398 						  u64 start, unsigned long len);
399 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
400 						u64 start);
401 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
402 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
403 					 u64 start);
404 void free_extent_buffer(struct extent_buffer *eb);
405 void free_extent_buffer_stale(struct extent_buffer *eb);
406 #define WAIT_NONE	0
407 #define WAIT_COMPLETE	1
408 #define WAIT_PAGE_LOCK	2
409 int read_extent_buffer_pages(struct extent_io_tree *tree,
410 			     struct extent_buffer *eb, int wait,
411 			     int mirror_num);
412 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
413 
414 static inline int num_extent_pages(const struct extent_buffer *eb)
415 {
416 	return (round_up(eb->start + eb->len, PAGE_SIZE) >> PAGE_SHIFT) -
417 	       (eb->start >> PAGE_SHIFT);
418 }
419 
420 static inline void extent_buffer_get(struct extent_buffer *eb)
421 {
422 	atomic_inc(&eb->refs);
423 }
424 
425 static inline int extent_buffer_uptodate(struct extent_buffer *eb)
426 {
427 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
428 }
429 
430 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
431 			 unsigned long start, unsigned long len);
432 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
433 			unsigned long start,
434 			unsigned long len);
435 int read_extent_buffer_to_user(const struct extent_buffer *eb,
436 			       void __user *dst, unsigned long start,
437 			       unsigned long len);
438 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
439 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
440 		const void *src);
441 void write_extent_buffer(struct extent_buffer *eb, const void *src,
442 			 unsigned long start, unsigned long len);
443 void copy_extent_buffer_full(struct extent_buffer *dst,
444 			     struct extent_buffer *src);
445 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
446 			unsigned long dst_offset, unsigned long src_offset,
447 			unsigned long len);
448 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
449 			   unsigned long src_offset, unsigned long len);
450 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
451 			   unsigned long src_offset, unsigned long len);
452 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
453 			   unsigned long len);
454 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
455 			   unsigned long pos);
456 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
457 			      unsigned long pos, unsigned long len);
458 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
459 				unsigned long pos, unsigned long len);
460 void clear_extent_buffer_dirty(struct extent_buffer *eb);
461 bool set_extent_buffer_dirty(struct extent_buffer *eb);
462 void set_extent_buffer_uptodate(struct extent_buffer *eb);
463 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
464 int extent_buffer_under_io(struct extent_buffer *eb);
465 int map_private_extent_buffer(const struct extent_buffer *eb,
466 			      unsigned long offset, unsigned long min_len,
467 			      char **map, unsigned long *map_start,
468 			      unsigned long *map_len);
469 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
470 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
471 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
472 				 u64 delalloc_end, struct page *locked_page,
473 				 unsigned bits_to_clear,
474 				 unsigned long page_ops);
475 struct bio *btrfs_bio_alloc(struct block_device *bdev, u64 first_byte);
476 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
477 struct bio *btrfs_bio_clone(struct bio *bio);
478 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
479 
480 struct btrfs_fs_info;
481 struct btrfs_inode;
482 
483 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
484 		      u64 length, u64 logical, struct page *page,
485 		      unsigned int pg_offset, int mirror_num);
486 int clean_io_failure(struct btrfs_fs_info *fs_info,
487 		     struct extent_io_tree *failure_tree,
488 		     struct extent_io_tree *io_tree, u64 start,
489 		     struct page *page, u64 ino, unsigned int pg_offset);
490 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
491 int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
492 			 struct extent_buffer *eb, int mirror_num);
493 
494 /*
495  * When IO fails, either with EIO or csum verification fails, we
496  * try other mirrors that might have a good copy of the data.  This
497  * io_failure_record is used to record state as we go through all the
498  * mirrors.  If another mirror has good data, the page is set up to date
499  * and things continue.  If a good mirror can't be found, the original
500  * bio end_io callback is called to indicate things have failed.
501  */
502 struct io_failure_record {
503 	struct page *page;
504 	u64 start;
505 	u64 len;
506 	u64 logical;
507 	unsigned long bio_flags;
508 	int this_mirror;
509 	int failed_mirror;
510 	int in_validation;
511 };
512 
513 
514 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
515 		u64 end);
516 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
517 				struct io_failure_record **failrec_ret);
518 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
519 			    struct io_failure_record *failrec, int fail_mirror);
520 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
521 				    struct io_failure_record *failrec,
522 				    struct page *page, int pg_offset, int icsum,
523 				    bio_end_io_t *endio_func, void *data);
524 int free_io_failure(struct extent_io_tree *failure_tree,
525 		    struct extent_io_tree *io_tree,
526 		    struct io_failure_record *rec);
527 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
528 bool find_lock_delalloc_range(struct inode *inode, struct extent_io_tree *tree,
529 			     struct page *locked_page, u64 *start,
530 			     u64 *end);
531 #endif
532 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
533 					       u64 start);
534 
535 #endif
536