xref: /openbmc/linux/fs/btrfs/extent_io.h (revision ecd25094)
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_UPTODATE		(1U << 1)
13 #define EXTENT_LOCKED		(1U << 2)
14 #define EXTENT_NEW		(1U << 3)
15 #define EXTENT_DELALLOC		(1U << 4)
16 #define EXTENT_DEFRAG		(1U << 5)
17 #define EXTENT_BOUNDARY		(1U << 6)
18 #define EXTENT_NODATASUM	(1U << 7)
19 #define EXTENT_CLEAR_META_RESV	(1U << 8)
20 #define EXTENT_NEED_WAIT	(1U << 9)
21 #define EXTENT_DAMAGED		(1U << 10)
22 #define EXTENT_NORESERVE	(1U << 11)
23 #define EXTENT_QGROUP_RESERVED	(1U << 12)
24 #define EXTENT_CLEAR_DATA_RESV	(1U << 13)
25 #define EXTENT_DELALLOC_NEW	(1U << 14)
26 #define EXTENT_DO_ACCOUNTING    (EXTENT_CLEAR_META_RESV | \
27 				 EXTENT_CLEAR_DATA_RESV)
28 #define EXTENT_CTLBITS		(EXTENT_DO_ACCOUNTING)
29 
30 /*
31  * Redefined bits above which are used only in the device allocation tree,
32  * shouldn't be using EXTENT_LOCKED / EXTENT_BOUNDARY / EXTENT_CLEAR_META_RESV
33  * / EXTENT_CLEAR_DATA_RESV because they have special meaning to the bit
34  * manipulation functions
35  */
36 #define CHUNK_ALLOCATED EXTENT_DIRTY
37 #define CHUNK_TRIMMED   EXTENT_DEFRAG
38 
39 /*
40  * flags for bio submission. The high bits indicate the compression
41  * type for this bio
42  */
43 #define EXTENT_BIO_COMPRESSED 1
44 #define EXTENT_BIO_FLAG_SHIFT 16
45 
46 enum {
47 	EXTENT_BUFFER_UPTODATE,
48 	EXTENT_BUFFER_DIRTY,
49 	EXTENT_BUFFER_CORRUPT,
50 	/* this got triggered by readahead */
51 	EXTENT_BUFFER_READAHEAD,
52 	EXTENT_BUFFER_TREE_REF,
53 	EXTENT_BUFFER_STALE,
54 	EXTENT_BUFFER_WRITEBACK,
55 	/* read IO error */
56 	EXTENT_BUFFER_READ_ERR,
57 	EXTENT_BUFFER_UNMAPPED,
58 	EXTENT_BUFFER_IN_TREE,
59 	/* write IO error */
60 	EXTENT_BUFFER_WRITE_ERR,
61 };
62 
63 /* these are flags for __process_pages_contig */
64 #define PAGE_UNLOCK		(1 << 0)
65 #define PAGE_CLEAR_DIRTY	(1 << 1)
66 #define PAGE_SET_WRITEBACK	(1 << 2)
67 #define PAGE_END_WRITEBACK	(1 << 3)
68 #define PAGE_SET_PRIVATE2	(1 << 4)
69 #define PAGE_SET_ERROR		(1 << 5)
70 #define PAGE_LOCK		(1 << 6)
71 
72 /*
73  * page->private values.  Every page that is controlled by the extent
74  * map has page->private set to one.
75  */
76 #define EXTENT_PAGE_PRIVATE 1
77 
78 /*
79  * The extent buffer bitmap operations are done with byte granularity instead of
80  * word granularity for two reasons:
81  * 1. The bitmaps must be little-endian on disk.
82  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
83  *    single word in a bitmap may straddle two pages in the extent buffer.
84  */
85 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
86 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
87 #define BITMAP_FIRST_BYTE_MASK(start) \
88 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
89 #define BITMAP_LAST_BYTE_MASK(nbits) \
90 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
91 
92 struct extent_state;
93 struct btrfs_root;
94 struct btrfs_inode;
95 struct btrfs_io_bio;
96 struct io_failure_record;
97 
98 
99 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
100 		struct bio *bio, u64 bio_offset);
101 
102 struct extent_io_ops {
103 	/*
104 	 * The following callbacks must be always defined, the function
105 	 * pointer will be called unconditionally.
106 	 */
107 	blk_status_t (*submit_bio_hook)(struct inode *inode, struct bio *bio,
108 					int mirror_num, unsigned long bio_flags);
109 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
110 				    struct page *page, u64 start, u64 end,
111 				    int mirror);
112 };
113 
114 enum {
115 	IO_TREE_FS_INFO_FREED_EXTENTS0,
116 	IO_TREE_FS_INFO_FREED_EXTENTS1,
117 	IO_TREE_INODE_IO,
118 	IO_TREE_INODE_IO_FAILURE,
119 	IO_TREE_RELOC_BLOCKS,
120 	IO_TREE_TRANS_DIRTY_PAGES,
121 	IO_TREE_ROOT_DIRTY_LOG_PAGES,
122 	IO_TREE_SELFTEST,
123 };
124 
125 struct extent_io_tree {
126 	struct rb_root state;
127 	struct btrfs_fs_info *fs_info;
128 	void *private_data;
129 	u64 dirty_bytes;
130 	bool track_uptodate;
131 
132 	/* Who owns this io tree, should be one of IO_TREE_* */
133 	u8 owner;
134 
135 	spinlock_t lock;
136 	const struct extent_io_ops *ops;
137 };
138 
139 struct extent_state {
140 	u64 start;
141 	u64 end; /* inclusive */
142 	struct rb_node rb_node;
143 
144 	/* ADD NEW ELEMENTS AFTER THIS */
145 	wait_queue_head_t wq;
146 	refcount_t refs;
147 	unsigned state;
148 
149 	struct io_failure_record *failrec;
150 
151 #ifdef CONFIG_BTRFS_DEBUG
152 	struct list_head leak_list;
153 #endif
154 };
155 
156 #define INLINE_EXTENT_BUFFER_PAGES 16
157 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
158 struct extent_buffer {
159 	u64 start;
160 	unsigned long len;
161 	unsigned long bflags;
162 	struct btrfs_fs_info *fs_info;
163 	spinlock_t refs_lock;
164 	atomic_t refs;
165 	atomic_t io_pages;
166 	int read_mirror;
167 	struct rcu_head rcu_head;
168 	pid_t lock_owner;
169 
170 	int blocking_writers;
171 	atomic_t blocking_readers;
172 	bool lock_nested;
173 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
174 	short log_index;
175 
176 	/* protects write locks */
177 	rwlock_t lock;
178 
179 	/* readers use lock_wq while they wait for the write
180 	 * lock holders to unlock
181 	 */
182 	wait_queue_head_t write_lock_wq;
183 
184 	/* writers use read_lock_wq while they wait for readers
185 	 * to unlock
186 	 */
187 	wait_queue_head_t read_lock_wq;
188 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
189 #ifdef CONFIG_BTRFS_DEBUG
190 	int spinning_writers;
191 	atomic_t spinning_readers;
192 	atomic_t read_locks;
193 	int write_locks;
194 	struct list_head leak_list;
195 #endif
196 };
197 
198 /*
199  * Structure to record how many bytes and which ranges are set/cleared
200  */
201 struct extent_changeset {
202 	/* How many bytes are set/cleared in this operation */
203 	unsigned int bytes_changed;
204 
205 	/* Changed ranges */
206 	struct ulist range_changed;
207 };
208 
209 static inline void extent_changeset_init(struct extent_changeset *changeset)
210 {
211 	changeset->bytes_changed = 0;
212 	ulist_init(&changeset->range_changed);
213 }
214 
215 static inline struct extent_changeset *extent_changeset_alloc(void)
216 {
217 	struct extent_changeset *ret;
218 
219 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
220 	if (!ret)
221 		return NULL;
222 
223 	extent_changeset_init(ret);
224 	return ret;
225 }
226 
227 static inline void extent_changeset_release(struct extent_changeset *changeset)
228 {
229 	if (!changeset)
230 		return;
231 	changeset->bytes_changed = 0;
232 	ulist_release(&changeset->range_changed);
233 }
234 
235 static inline void extent_changeset_free(struct extent_changeset *changeset)
236 {
237 	if (!changeset)
238 		return;
239 	extent_changeset_release(changeset);
240 	kfree(changeset);
241 }
242 
243 static inline void extent_set_compress_type(unsigned long *bio_flags,
244 					    int compress_type)
245 {
246 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
247 }
248 
249 static inline int extent_compress_type(unsigned long bio_flags)
250 {
251 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
252 }
253 
254 struct extent_map_tree;
255 
256 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
257 					  struct page *page,
258 					  size_t pg_offset,
259 					  u64 start, u64 len,
260 					  int create);
261 
262 void extent_io_tree_init(struct btrfs_fs_info *fs_info,
263 			 struct extent_io_tree *tree, unsigned int owner,
264 			 void *private_data);
265 void extent_io_tree_release(struct extent_io_tree *tree);
266 int try_release_extent_mapping(struct page *page, gfp_t mask);
267 int try_release_extent_buffer(struct page *page);
268 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
269 		     struct extent_state **cached);
270 
271 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
272 {
273 	return lock_extent_bits(tree, start, end, NULL);
274 }
275 
276 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
277 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
278 			  get_extent_t *get_extent, int mirror_num);
279 int __init extent_io_init(void);
280 void __cold extent_io_exit(void);
281 
282 u64 count_range_bits(struct extent_io_tree *tree,
283 		     u64 *start, u64 search_end,
284 		     u64 max_bytes, unsigned bits, int contig);
285 
286 void free_extent_state(struct extent_state *state);
287 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
288 		   unsigned bits, int filled,
289 		   struct extent_state *cached_state);
290 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
291 		unsigned bits, struct extent_changeset *changeset);
292 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
293 		     unsigned bits, int wake, int delete,
294 		     struct extent_state **cached);
295 int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
296 		     unsigned bits, int wake, int delete,
297 		     struct extent_state **cached, gfp_t mask,
298 		     struct extent_changeset *changeset);
299 
300 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
301 {
302 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL);
303 }
304 
305 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
306 		u64 end, struct extent_state **cached)
307 {
308 	return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
309 				GFP_NOFS, NULL);
310 }
311 
312 static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree,
313 		u64 start, u64 end, struct extent_state **cached)
314 {
315 	return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
316 				GFP_ATOMIC, NULL);
317 }
318 
319 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
320 		u64 end, unsigned bits)
321 {
322 	int wake = 0;
323 
324 	if (bits & EXTENT_LOCKED)
325 		wake = 1;
326 
327 	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL);
328 }
329 
330 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
331 			   unsigned bits, struct extent_changeset *changeset);
332 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
333 		   unsigned bits, u64 *failed_start,
334 		   struct extent_state **cached_state, gfp_t mask);
335 int set_extent_bits_nowait(struct extent_io_tree *tree, u64 start, u64 end,
336 			   unsigned bits);
337 
338 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
339 		u64 end, unsigned bits)
340 {
341 	return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
342 }
343 
344 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
345 		u64 end, struct extent_state **cached_state)
346 {
347 	return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
348 				cached_state, GFP_NOFS, NULL);
349 }
350 
351 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
352 		u64 end, gfp_t mask)
353 {
354 	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
355 			      NULL, mask);
356 }
357 
358 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
359 				     u64 end, struct extent_state **cached)
360 {
361 	return clear_extent_bit(tree, start, end,
362 				EXTENT_DIRTY | EXTENT_DELALLOC |
363 				EXTENT_DO_ACCOUNTING, 0, 0, cached);
364 }
365 
366 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
367 		       unsigned bits, unsigned clear_bits,
368 		       struct extent_state **cached_state);
369 
370 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
371 				      u64 end, unsigned int extra_bits,
372 				      struct extent_state **cached_state)
373 {
374 	return set_extent_bit(tree, start, end,
375 			      EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits,
376 			      NULL, cached_state, GFP_NOFS);
377 }
378 
379 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
380 		u64 end, struct extent_state **cached_state)
381 {
382 	return set_extent_bit(tree, start, end,
383 			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
384 			      NULL, cached_state, GFP_NOFS);
385 }
386 
387 static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
388 		u64 end)
389 {
390 	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
391 			GFP_NOFS);
392 }
393 
394 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
395 		u64 end, struct extent_state **cached_state, gfp_t mask)
396 {
397 	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
398 			      cached_state, mask);
399 }
400 
401 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
402 			  u64 *start_ret, u64 *end_ret, unsigned bits,
403 			  struct extent_state **cached_state);
404 void find_first_clear_extent_bit(struct extent_io_tree *tree, u64 start,
405 				 u64 *start_ret, u64 *end_ret, unsigned bits);
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,
410 			      int mode);
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,
416 		     unsigned nr_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);
420 
421 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
422 					  u64 start);
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,
426 						u64 start);
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,
429 					 u64 start);
430 void free_extent_buffer(struct extent_buffer *eb);
431 void free_extent_buffer_stale(struct extent_buffer *eb);
432 #define WAIT_NONE	0
433 #define WAIT_COMPLETE	1
434 #define WAIT_PAGE_LOCK	2
435 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
436 			     int mirror_num);
437 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
438 
439 static inline int num_extent_pages(const struct extent_buffer *eb)
440 {
441 	return (round_up(eb->start + eb->len, PAGE_SIZE) >> PAGE_SHIFT) -
442 	       (eb->start >> PAGE_SHIFT);
443 }
444 
445 static inline void extent_buffer_get(struct extent_buffer *eb)
446 {
447 	atomic_inc(&eb->refs);
448 }
449 
450 static inline int extent_buffer_uptodate(struct extent_buffer *eb)
451 {
452 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
453 }
454 
455 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
456 			 unsigned long start, unsigned long len);
457 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
458 			unsigned long start,
459 			unsigned long len);
460 int read_extent_buffer_to_user(const struct extent_buffer *eb,
461 			       void __user *dst, unsigned long start,
462 			       unsigned long len);
463 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
464 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
465 		const void *src);
466 void write_extent_buffer(struct extent_buffer *eb, const void *src,
467 			 unsigned long start, unsigned long len);
468 void copy_extent_buffer_full(struct extent_buffer *dst,
469 			     struct extent_buffer *src);
470 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
471 			unsigned long dst_offset, unsigned long src_offset,
472 			unsigned long len);
473 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
474 			   unsigned long src_offset, unsigned long len);
475 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
476 			   unsigned long src_offset, unsigned long len);
477 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
478 			   unsigned long len);
479 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
480 			   unsigned long pos);
481 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
482 			      unsigned long pos, unsigned long len);
483 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
484 				unsigned long pos, unsigned long len);
485 void clear_extent_buffer_dirty(struct extent_buffer *eb);
486 bool set_extent_buffer_dirty(struct extent_buffer *eb);
487 void set_extent_buffer_uptodate(struct extent_buffer *eb);
488 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
489 int extent_buffer_under_io(struct extent_buffer *eb);
490 int map_private_extent_buffer(const struct extent_buffer *eb,
491 			      unsigned long offset, unsigned long min_len,
492 			      char **map, unsigned long *map_start,
493 			      unsigned long *map_len);
494 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
495 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
496 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
497 				  struct page *locked_page,
498 				  unsigned bits_to_clear,
499 				  unsigned long page_ops);
500 struct bio *btrfs_bio_alloc(u64 first_byte);
501 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
502 struct bio *btrfs_bio_clone(struct bio *bio);
503 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
504 
505 struct btrfs_fs_info;
506 struct btrfs_inode;
507 
508 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
509 		      u64 length, u64 logical, struct page *page,
510 		      unsigned int pg_offset, int mirror_num);
511 int clean_io_failure(struct btrfs_fs_info *fs_info,
512 		     struct extent_io_tree *failure_tree,
513 		     struct extent_io_tree *io_tree, u64 start,
514 		     struct page *page, u64 ino, unsigned int pg_offset);
515 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
516 int btrfs_repair_eb_io_failure(struct extent_buffer *eb, int mirror_num);
517 
518 /*
519  * When IO fails, either with EIO or csum verification fails, we
520  * try other mirrors that might have a good copy of the data.  This
521  * io_failure_record is used to record state as we go through all the
522  * mirrors.  If another mirror has good data, the page is set up to date
523  * and things continue.  If a good mirror can't be found, the original
524  * bio end_io callback is called to indicate things have failed.
525  */
526 struct io_failure_record {
527 	struct page *page;
528 	u64 start;
529 	u64 len;
530 	u64 logical;
531 	unsigned long bio_flags;
532 	int this_mirror;
533 	int failed_mirror;
534 	int in_validation;
535 };
536 
537 
538 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
539 		u64 end);
540 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
541 				struct io_failure_record **failrec_ret);
542 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
543 			    struct io_failure_record *failrec, int fail_mirror);
544 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
545 				    struct io_failure_record *failrec,
546 				    struct page *page, int pg_offset, int icsum,
547 				    bio_end_io_t *endio_func, void *data);
548 int free_io_failure(struct extent_io_tree *failure_tree,
549 		    struct extent_io_tree *io_tree,
550 		    struct io_failure_record *rec);
551 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
552 bool find_lock_delalloc_range(struct inode *inode,
553 			     struct page *locked_page, u64 *start,
554 			     u64 *end);
555 #endif
556 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
557 					       u64 start);
558 
559 #endif
560