xref: /openbmc/linux/fs/btrfs/extent_io.h (revision a36954f5)
1 #ifndef __EXTENTIO__
2 #define __EXTENTIO__
3 
4 #include <linux/rbtree.h>
5 #include <linux/refcount.h>
6 #include "ulist.h"
7 
8 /* bits for the extent state */
9 #define EXTENT_DIRTY		(1U << 0)
10 #define EXTENT_WRITEBACK	(1U << 1)
11 #define EXTENT_UPTODATE		(1U << 2)
12 #define EXTENT_LOCKED		(1U << 3)
13 #define EXTENT_NEW		(1U << 4)
14 #define EXTENT_DELALLOC		(1U << 5)
15 #define EXTENT_DEFRAG		(1U << 6)
16 #define EXTENT_BOUNDARY		(1U << 9)
17 #define EXTENT_NODATASUM	(1U << 10)
18 #define EXTENT_CLEAR_META_RESV	(1U << 11)
19 #define EXTENT_FIRST_DELALLOC	(1U << 12)
20 #define EXTENT_NEED_WAIT	(1U << 13)
21 #define EXTENT_DAMAGED		(1U << 14)
22 #define EXTENT_NORESERVE	(1U << 15)
23 #define EXTENT_QGROUP_RESERVED	(1U << 16)
24 #define EXTENT_CLEAR_DATA_RESV	(1U << 17)
25 #define EXTENT_DELALLOC_NEW	(1U << 18)
26 #define EXTENT_IOBITS		(EXTENT_LOCKED | EXTENT_WRITEBACK)
27 #define EXTENT_DO_ACCOUNTING    (EXTENT_CLEAR_META_RESV | \
28 				 EXTENT_CLEAR_DATA_RESV)
29 #define EXTENT_CTLBITS		(EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
30 
31 /*
32  * flags for bio submission. The high bits indicate the compression
33  * type for this bio
34  */
35 #define EXTENT_BIO_COMPRESSED 1
36 #define EXTENT_BIO_TREE_LOG 2
37 #define EXTENT_BIO_FLAG_SHIFT 16
38 
39 /* these are bit numbers for test/set bit */
40 #define EXTENT_BUFFER_UPTODATE 0
41 #define EXTENT_BUFFER_DIRTY 2
42 #define EXTENT_BUFFER_CORRUPT 3
43 #define EXTENT_BUFFER_READAHEAD 4	/* this got triggered by readahead */
44 #define EXTENT_BUFFER_TREE_REF 5
45 #define EXTENT_BUFFER_STALE 6
46 #define EXTENT_BUFFER_WRITEBACK 7
47 #define EXTENT_BUFFER_READ_ERR 8        /* read IO error */
48 #define EXTENT_BUFFER_DUMMY 9
49 #define EXTENT_BUFFER_IN_TREE 10
50 #define EXTENT_BUFFER_WRITE_ERR 11    /* write IO error */
51 
52 /* these are flags for __process_pages_contig */
53 #define PAGE_UNLOCK		(1 << 0)
54 #define PAGE_CLEAR_DIRTY	(1 << 1)
55 #define PAGE_SET_WRITEBACK	(1 << 2)
56 #define PAGE_END_WRITEBACK	(1 << 3)
57 #define PAGE_SET_PRIVATE2	(1 << 4)
58 #define PAGE_SET_ERROR		(1 << 5)
59 #define PAGE_LOCK		(1 << 6)
60 
61 /*
62  * page->private values.  Every page that is controlled by the extent
63  * map has page->private set to one.
64  */
65 #define EXTENT_PAGE_PRIVATE 1
66 
67 /*
68  * The extent buffer bitmap operations are done with byte granularity instead of
69  * word granularity for two reasons:
70  * 1. The bitmaps must be little-endian on disk.
71  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
72  *    single word in a bitmap may straddle two pages in the extent buffer.
73  */
74 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
75 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
76 #define BITMAP_FIRST_BYTE_MASK(start) \
77 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
78 #define BITMAP_LAST_BYTE_MASK(nbits) \
79 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
80 
81 static inline int le_test_bit(int nr, const u8 *addr)
82 {
83 	return 1U & (addr[BIT_BYTE(nr)] >> (nr & (BITS_PER_BYTE-1)));
84 }
85 
86 extern void le_bitmap_set(u8 *map, unsigned int start, int len);
87 extern void le_bitmap_clear(u8 *map, unsigned int start, int len);
88 
89 struct extent_state;
90 struct btrfs_root;
91 struct btrfs_inode;
92 struct btrfs_io_bio;
93 struct io_failure_record;
94 
95 typedef	int (extent_submit_bio_hook_t)(struct inode *inode, struct bio *bio,
96 				       int mirror_num, unsigned long bio_flags,
97 				       u64 bio_offset);
98 struct extent_io_ops {
99 	/*
100 	 * The following callbacks must be allways defined, the function
101 	 * pointer will be called unconditionally.
102 	 */
103 	extent_submit_bio_hook_t *submit_bio_hook;
104 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
105 				    struct page *page, u64 start, u64 end,
106 				    int mirror);
107 	int (*merge_bio_hook)(struct page *page, unsigned long offset,
108 			      size_t size, struct bio *bio,
109 			      unsigned long bio_flags);
110 	int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
111 
112 	/*
113 	 * Optional hooks, called if the pointer is not NULL
114 	 */
115 	int (*fill_delalloc)(struct inode *inode, struct page *locked_page,
116 			     u64 start, u64 end, int *page_started,
117 			     unsigned long *nr_written);
118 
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)(struct inode *inode, struct extent_state *state,
123 			     unsigned *bits);
124 	void (*clear_bit_hook)(struct btrfs_inode *inode,
125 			struct extent_state *state,
126 			unsigned *bits);
127 	void (*merge_extent_hook)(struct inode *inode,
128 				  struct extent_state *new,
129 				  struct extent_state *other);
130 	void (*split_extent_hook)(struct inode *inode,
131 				  struct extent_state *orig, u64 split);
132 };
133 
134 struct extent_io_tree {
135 	struct rb_root state;
136 	struct address_space *mapping;
137 	u64 dirty_bytes;
138 	int track_uptodate;
139 	spinlock_t lock;
140 	const struct extent_io_ops *ops;
141 };
142 
143 struct extent_state {
144 	u64 start;
145 	u64 end; /* inclusive */
146 	struct rb_node rb_node;
147 
148 	/* ADD NEW ELEMENTS AFTER THIS */
149 	wait_queue_head_t wq;
150 	refcount_t refs;
151 	unsigned state;
152 
153 	struct io_failure_record *failrec;
154 
155 #ifdef CONFIG_BTRFS_DEBUG
156 	struct list_head leak_list;
157 #endif
158 };
159 
160 #define INLINE_EXTENT_BUFFER_PAGES 16
161 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
162 struct extent_buffer {
163 	u64 start;
164 	unsigned long len;
165 	unsigned long bflags;
166 	struct btrfs_fs_info *fs_info;
167 	spinlock_t refs_lock;
168 	atomic_t refs;
169 	atomic_t io_pages;
170 	int read_mirror;
171 	struct rcu_head rcu_head;
172 	pid_t lock_owner;
173 
174 	/* count of read lock holders on the extent buffer */
175 	atomic_t write_locks;
176 	atomic_t read_locks;
177 	atomic_t blocking_writers;
178 	atomic_t blocking_readers;
179 	atomic_t spinning_readers;
180 	atomic_t spinning_writers;
181 	short lock_nested;
182 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
183 	short log_index;
184 
185 	/* protects write locks */
186 	rwlock_t lock;
187 
188 	/* readers use lock_wq while they wait for the write
189 	 * lock holders to unlock
190 	 */
191 	wait_queue_head_t write_lock_wq;
192 
193 	/* writers use read_lock_wq while they wait for readers
194 	 * to unlock
195 	 */
196 	wait_queue_head_t read_lock_wq;
197 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
198 #ifdef CONFIG_BTRFS_DEBUG
199 	struct list_head leak_list;
200 #endif
201 };
202 
203 /*
204  * Structure to record how many bytes and which ranges are set/cleared
205  */
206 struct extent_changeset {
207 	/* How many bytes are set/cleared in this operation */
208 	u64 bytes_changed;
209 
210 	/* Changed ranges */
211 	struct ulist range_changed;
212 };
213 
214 static inline void extent_set_compress_type(unsigned long *bio_flags,
215 					    int compress_type)
216 {
217 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
218 }
219 
220 static inline int extent_compress_type(unsigned long bio_flags)
221 {
222 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
223 }
224 
225 struct extent_map_tree;
226 
227 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
228 					  struct page *page,
229 					  size_t pg_offset,
230 					  u64 start, u64 len,
231 					  int create);
232 
233 void extent_io_tree_init(struct extent_io_tree *tree,
234 			 struct address_space *mapping);
235 int try_release_extent_mapping(struct extent_map_tree *map,
236 			       struct extent_io_tree *tree, struct page *page,
237 			       gfp_t mask);
238 int try_release_extent_buffer(struct page *page);
239 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
240 		     struct extent_state **cached);
241 
242 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
243 {
244 	return lock_extent_bits(tree, start, end, NULL);
245 }
246 
247 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
248 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
249 			  get_extent_t *get_extent, int mirror_num);
250 int __init extent_io_init(void);
251 void extent_io_exit(void);
252 
253 u64 count_range_bits(struct extent_io_tree *tree,
254 		     u64 *start, u64 search_end,
255 		     u64 max_bytes, unsigned bits, int contig);
256 
257 void free_extent_state(struct extent_state *state);
258 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
259 		   unsigned bits, int filled,
260 		   struct extent_state *cached_state);
261 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
262 		unsigned bits, struct extent_changeset *changeset);
263 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
264 		     unsigned bits, int wake, int delete,
265 		     struct extent_state **cached, gfp_t mask);
266 
267 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
268 {
269 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
270 				GFP_NOFS);
271 }
272 
273 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
274 		u64 end, struct extent_state **cached, gfp_t mask)
275 {
276 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
277 				mask);
278 }
279 
280 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
281 		u64 end, unsigned bits)
282 {
283 	int wake = 0;
284 
285 	if (bits & EXTENT_LOCKED)
286 		wake = 1;
287 
288 	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL,
289 			GFP_NOFS);
290 }
291 
292 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
293 			   unsigned bits, struct extent_changeset *changeset);
294 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
295 		   unsigned bits, u64 *failed_start,
296 		   struct extent_state **cached_state, gfp_t mask);
297 
298 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
299 		u64 end, unsigned bits)
300 {
301 	return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
302 }
303 
304 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
305 		u64 end, struct extent_state **cached_state, gfp_t mask)
306 {
307 	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
308 				cached_state, mask);
309 }
310 
311 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
312 		u64 end, gfp_t mask)
313 {
314 	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
315 			      NULL, mask);
316 }
317 
318 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
319 		u64 end)
320 {
321 	return clear_extent_bit(tree, start, end,
322 				EXTENT_DIRTY | EXTENT_DELALLOC |
323 				EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
324 }
325 
326 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
327 		       unsigned bits, unsigned clear_bits,
328 		       struct extent_state **cached_state);
329 
330 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
331 		u64 end, struct extent_state **cached_state)
332 {
333 	return set_extent_bit(tree, start, end,
334 			      EXTENT_DELALLOC | EXTENT_UPTODATE,
335 			      NULL, cached_state, GFP_NOFS);
336 }
337 
338 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
339 		u64 end, struct extent_state **cached_state)
340 {
341 	return set_extent_bit(tree, start, end,
342 			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
343 			      NULL, cached_state, GFP_NOFS);
344 }
345 
346 static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
347 		u64 end)
348 {
349 	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
350 			GFP_NOFS);
351 }
352 
353 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
354 		u64 end, struct extent_state **cached_state, gfp_t mask)
355 {
356 	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
357 			      cached_state, mask);
358 }
359 
360 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
361 			  u64 *start_ret, u64 *end_ret, unsigned bits,
362 			  struct extent_state **cached_state);
363 int extent_invalidatepage(struct extent_io_tree *tree,
364 			  struct page *page, unsigned long offset);
365 int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
366 			  get_extent_t *get_extent,
367 			  struct writeback_control *wbc);
368 int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
369 			      u64 start, u64 end, get_extent_t *get_extent,
370 			      int mode);
371 int extent_writepages(struct extent_io_tree *tree,
372 		      struct address_space *mapping,
373 		      get_extent_t *get_extent,
374 		      struct writeback_control *wbc);
375 int btree_write_cache_pages(struct address_space *mapping,
376 			    struct writeback_control *wbc);
377 int extent_readpages(struct extent_io_tree *tree,
378 		     struct address_space *mapping,
379 		     struct list_head *pages, unsigned nr_pages,
380 		     get_extent_t get_extent);
381 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
382 		__u64 start, __u64 len, get_extent_t *get_extent);
383 void set_page_extent_mapped(struct page *page);
384 
385 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
386 					  u64 start);
387 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
388 						  u64 start, unsigned long len);
389 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
390 						u64 start);
391 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
392 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
393 					 u64 start);
394 void free_extent_buffer(struct extent_buffer *eb);
395 void free_extent_buffer_stale(struct extent_buffer *eb);
396 #define WAIT_NONE	0
397 #define WAIT_COMPLETE	1
398 #define WAIT_PAGE_LOCK	2
399 int read_extent_buffer_pages(struct extent_io_tree *tree,
400 			     struct extent_buffer *eb, int wait,
401 			     get_extent_t *get_extent, int mirror_num);
402 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
403 
404 static inline unsigned long num_extent_pages(u64 start, u64 len)
405 {
406 	return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) -
407 		(start >> PAGE_SHIFT);
408 }
409 
410 static inline void extent_buffer_get(struct extent_buffer *eb)
411 {
412 	atomic_inc(&eb->refs);
413 }
414 
415 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
416 			  unsigned long start,
417 			  unsigned long len);
418 void read_extent_buffer(struct extent_buffer *eb, void *dst,
419 			unsigned long start,
420 			unsigned long len);
421 int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dst,
422 			       unsigned long start,
423 			       unsigned long len);
424 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
425 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
426 		const void *src);
427 void write_extent_buffer(struct extent_buffer *eb, const void *src,
428 			 unsigned long start, unsigned long len);
429 void copy_extent_buffer_full(struct extent_buffer *dst,
430 			     struct extent_buffer *src);
431 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
432 			unsigned long dst_offset, unsigned long src_offset,
433 			unsigned long len);
434 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
435 			   unsigned long src_offset, unsigned long len);
436 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
437 			   unsigned long src_offset, unsigned long len);
438 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
439 			   unsigned long len);
440 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
441 			   unsigned long pos);
442 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
443 			      unsigned long pos, unsigned long len);
444 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
445 				unsigned long pos, unsigned long len);
446 void clear_extent_buffer_dirty(struct extent_buffer *eb);
447 int set_extent_buffer_dirty(struct extent_buffer *eb);
448 void set_extent_buffer_uptodate(struct extent_buffer *eb);
449 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
450 int extent_buffer_uptodate(struct extent_buffer *eb);
451 int extent_buffer_under_io(struct extent_buffer *eb);
452 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
453 		      unsigned long min_len, char **map,
454 		      unsigned long *map_start,
455 		      unsigned long *map_len);
456 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
457 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
458 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
459 				 u64 delalloc_end, struct page *locked_page,
460 				 unsigned bits_to_clear,
461 				 unsigned long page_ops);
462 struct bio *
463 btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
464 		gfp_t gfp_flags);
465 struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs);
466 struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask);
467 
468 struct btrfs_fs_info;
469 struct btrfs_inode;
470 
471 int repair_io_failure(struct btrfs_inode *inode, u64 start, u64 length,
472 		u64 logical, struct page *page,
473 		unsigned int pg_offset, int mirror_num);
474 int clean_io_failure(struct btrfs_inode *inode, u64 start,
475 		struct page *page, unsigned int pg_offset);
476 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
477 int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
478 			 struct extent_buffer *eb, int mirror_num);
479 
480 /*
481  * When IO fails, either with EIO or csum verification fails, we
482  * try other mirrors that might have a good copy of the data.  This
483  * io_failure_record is used to record state as we go through all the
484  * mirrors.  If another mirror has good data, the page is set up to date
485  * and things continue.  If a good mirror can't be found, the original
486  * bio end_io callback is called to indicate things have failed.
487  */
488 struct io_failure_record {
489 	struct page *page;
490 	u64 start;
491 	u64 len;
492 	u64 logical;
493 	unsigned long bio_flags;
494 	int this_mirror;
495 	int failed_mirror;
496 	int in_validation;
497 };
498 
499 
500 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
501 		u64 end);
502 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
503 				struct io_failure_record **failrec_ret);
504 int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
505 			   struct io_failure_record *failrec, int fail_mirror);
506 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
507 				    struct io_failure_record *failrec,
508 				    struct page *page, int pg_offset, int icsum,
509 				    bio_end_io_t *endio_func, void *data);
510 int free_io_failure(struct btrfs_inode *inode, struct io_failure_record *rec);
511 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
512 noinline u64 find_lock_delalloc_range(struct inode *inode,
513 				      struct extent_io_tree *tree,
514 				      struct page *locked_page, u64 *start,
515 				      u64 *end, u64 max_bytes);
516 #endif
517 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
518 					       u64 start);
519 #endif
520