xref: /openbmc/linux/fs/btrfs/extent_io.h (revision aac5987a)
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_CLEAR_DATA_RESV	(1U << 17)
24 #define EXTENT_IOBITS		(EXTENT_LOCKED | EXTENT_WRITEBACK)
25 #define EXTENT_CTLBITS		(EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
26 
27 /*
28  * flags for bio submission. The high bits indicate the compression
29  * type for this bio
30  */
31 #define EXTENT_BIO_COMPRESSED 1
32 #define EXTENT_BIO_TREE_LOG 2
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 __process_pages_contig */
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 #define PAGE_LOCK		(1 << 6)
56 
57 /*
58  * page->private values.  Every page that is controlled by the extent
59  * map has page->private set to one.
60  */
61 #define EXTENT_PAGE_PRIVATE 1
62 
63 /*
64  * The extent buffer bitmap operations are done with byte granularity instead of
65  * word granularity for two reasons:
66  * 1. The bitmaps must be little-endian on disk.
67  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
68  *    single word in a bitmap may straddle two pages in the extent buffer.
69  */
70 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
71 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
72 #define BITMAP_FIRST_BYTE_MASK(start) \
73 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
74 #define BITMAP_LAST_BYTE_MASK(nbits) \
75 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
76 
77 static inline int le_test_bit(int nr, const u8 *addr)
78 {
79 	return 1U & (addr[BIT_BYTE(nr)] >> (nr & (BITS_PER_BYTE-1)));
80 }
81 
82 extern void le_bitmap_set(u8 *map, unsigned int start, int len);
83 extern void le_bitmap_clear(u8 *map, unsigned int start, int len);
84 
85 struct extent_state;
86 struct btrfs_root;
87 struct btrfs_inode;
88 struct btrfs_io_bio;
89 struct io_failure_record;
90 
91 typedef	int (extent_submit_bio_hook_t)(struct inode *inode, struct bio *bio,
92 				       int mirror_num, unsigned long bio_flags,
93 				       u64 bio_offset);
94 struct extent_io_ops {
95 	/*
96 	 * The following callbacks must be allways defined, the function
97 	 * pointer will be called unconditionally.
98 	 */
99 	extent_submit_bio_hook_t *submit_bio_hook;
100 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
101 				    struct page *page, u64 start, u64 end,
102 				    int mirror);
103 	int (*merge_bio_hook)(struct page *page, unsigned long offset,
104 			      size_t size, struct bio *bio,
105 			      unsigned long bio_flags);
106 	int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
107 
108 	/*
109 	 * Optional hooks, called if the pointer is not NULL
110 	 */
111 	int (*fill_delalloc)(struct inode *inode, struct page *locked_page,
112 			     u64 start, u64 end, int *page_started,
113 			     unsigned long *nr_written);
114 
115 	int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
116 	void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
117 				      struct extent_state *state, int uptodate);
118 	void (*set_bit_hook)(struct inode *inode, struct extent_state *state,
119 			     unsigned *bits);
120 	void (*clear_bit_hook)(struct btrfs_inode *inode,
121 			struct extent_state *state,
122 			unsigned *bits);
123 	void (*merge_extent_hook)(struct inode *inode,
124 				  struct extent_state *new,
125 				  struct extent_state *other);
126 	void (*split_extent_hook)(struct inode *inode,
127 				  struct extent_state *orig, u64 split);
128 };
129 
130 struct extent_io_tree {
131 	struct rb_root state;
132 	struct address_space *mapping;
133 	u64 dirty_bytes;
134 	int track_uptodate;
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 	atomic_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 	/* count of read lock holders on the extent buffer */
171 	atomic_t write_locks;
172 	atomic_t read_locks;
173 	atomic_t blocking_writers;
174 	atomic_t blocking_readers;
175 	atomic_t spinning_readers;
176 	atomic_t spinning_writers;
177 	short lock_nested;
178 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
179 	short log_index;
180 
181 	/* protects write locks */
182 	rwlock_t lock;
183 
184 	/* readers use lock_wq while they wait for the write
185 	 * lock holders to unlock
186 	 */
187 	wait_queue_head_t write_lock_wq;
188 
189 	/* writers use read_lock_wq while they wait for readers
190 	 * to unlock
191 	 */
192 	wait_queue_head_t read_lock_wq;
193 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
194 #ifdef CONFIG_BTRFS_DEBUG
195 	struct list_head leak_list;
196 #endif
197 };
198 
199 /*
200  * Structure to record how many bytes and which ranges are set/cleared
201  */
202 struct extent_changeset {
203 	/* How many bytes are set/cleared in this operation */
204 	u64 bytes_changed;
205 
206 	/* Changed ranges */
207 	struct ulist range_changed;
208 };
209 
210 static inline void extent_set_compress_type(unsigned long *bio_flags,
211 					    int compress_type)
212 {
213 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
214 }
215 
216 static inline int extent_compress_type(unsigned long bio_flags)
217 {
218 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
219 }
220 
221 struct extent_map_tree;
222 
223 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
224 					  struct page *page,
225 					  size_t pg_offset,
226 					  u64 start, u64 len,
227 					  int create);
228 
229 void extent_io_tree_init(struct extent_io_tree *tree,
230 			 struct address_space *mapping);
231 int try_release_extent_mapping(struct extent_map_tree *map,
232 			       struct extent_io_tree *tree, struct page *page,
233 			       gfp_t mask);
234 int try_release_extent_buffer(struct page *page);
235 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
236 		     struct extent_state **cached);
237 
238 static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
239 {
240 	return lock_extent_bits(tree, start, end, NULL);
241 }
242 
243 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
244 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
245 			  get_extent_t *get_extent, int mirror_num);
246 int __init extent_io_init(void);
247 void extent_io_exit(void);
248 
249 u64 count_range_bits(struct extent_io_tree *tree,
250 		     u64 *start, u64 search_end,
251 		     u64 max_bytes, unsigned bits, int contig);
252 
253 void free_extent_state(struct extent_state *state);
254 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
255 		   unsigned bits, int filled,
256 		   struct extent_state *cached_state);
257 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
258 		unsigned bits, struct extent_changeset *changeset);
259 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
260 		     unsigned bits, int wake, int delete,
261 		     struct extent_state **cached, gfp_t mask);
262 
263 static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
264 {
265 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
266 				GFP_NOFS);
267 }
268 
269 static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
270 		u64 end, struct extent_state **cached, gfp_t mask)
271 {
272 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
273 				mask);
274 }
275 
276 static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
277 		u64 end, unsigned bits)
278 {
279 	int wake = 0;
280 
281 	if (bits & EXTENT_LOCKED)
282 		wake = 1;
283 
284 	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL,
285 			GFP_NOFS);
286 }
287 
288 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
289 			   unsigned bits, struct extent_changeset *changeset);
290 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
291 		   unsigned bits, u64 *failed_start,
292 		   struct extent_state **cached_state, gfp_t mask);
293 
294 static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
295 		u64 end, unsigned bits)
296 {
297 	return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
298 }
299 
300 static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
301 		u64 end, struct extent_state **cached_state, gfp_t mask)
302 {
303 	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
304 				cached_state, mask);
305 }
306 
307 static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
308 		u64 end, gfp_t mask)
309 {
310 	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
311 			      NULL, mask);
312 }
313 
314 static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
315 		u64 end)
316 {
317 	return clear_extent_bit(tree, start, end,
318 				EXTENT_DIRTY | EXTENT_DELALLOC |
319 				EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
320 }
321 
322 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
323 		       unsigned bits, unsigned clear_bits,
324 		       struct extent_state **cached_state);
325 
326 static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
327 		u64 end, struct extent_state **cached_state)
328 {
329 	return set_extent_bit(tree, start, end,
330 			      EXTENT_DELALLOC | EXTENT_UPTODATE,
331 			      NULL, cached_state, GFP_NOFS);
332 }
333 
334 static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
335 		u64 end, struct extent_state **cached_state)
336 {
337 	return set_extent_bit(tree, start, end,
338 			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
339 			      NULL, cached_state, GFP_NOFS);
340 }
341 
342 static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
343 		u64 end)
344 {
345 	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
346 			GFP_NOFS);
347 }
348 
349 static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
350 		u64 end, struct extent_state **cached_state, gfp_t mask)
351 {
352 	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
353 			      cached_state, mask);
354 }
355 
356 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
357 			  u64 *start_ret, u64 *end_ret, unsigned bits,
358 			  struct extent_state **cached_state);
359 int extent_invalidatepage(struct extent_io_tree *tree,
360 			  struct page *page, unsigned long offset);
361 int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
362 			  get_extent_t *get_extent,
363 			  struct writeback_control *wbc);
364 int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
365 			      u64 start, u64 end, get_extent_t *get_extent,
366 			      int mode);
367 int extent_writepages(struct extent_io_tree *tree,
368 		      struct address_space *mapping,
369 		      get_extent_t *get_extent,
370 		      struct writeback_control *wbc);
371 int btree_write_cache_pages(struct address_space *mapping,
372 			    struct writeback_control *wbc);
373 int extent_readpages(struct extent_io_tree *tree,
374 		     struct address_space *mapping,
375 		     struct list_head *pages, unsigned nr_pages,
376 		     get_extent_t get_extent);
377 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
378 		__u64 start, __u64 len, get_extent_t *get_extent);
379 void set_page_extent_mapped(struct page *page);
380 
381 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
382 					  u64 start);
383 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
384 						  u64 start, unsigned long len);
385 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
386 						u64 start);
387 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
388 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
389 					 u64 start);
390 void free_extent_buffer(struct extent_buffer *eb);
391 void free_extent_buffer_stale(struct extent_buffer *eb);
392 #define WAIT_NONE	0
393 #define WAIT_COMPLETE	1
394 #define WAIT_PAGE_LOCK	2
395 int read_extent_buffer_pages(struct extent_io_tree *tree,
396 			     struct extent_buffer *eb, int wait,
397 			     get_extent_t *get_extent, int mirror_num);
398 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
399 
400 static inline unsigned long num_extent_pages(u64 start, u64 len)
401 {
402 	return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) -
403 		(start >> PAGE_SHIFT);
404 }
405 
406 static inline void extent_buffer_get(struct extent_buffer *eb)
407 {
408 	atomic_inc(&eb->refs);
409 }
410 
411 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
412 			  unsigned long start,
413 			  unsigned long len);
414 void read_extent_buffer(struct extent_buffer *eb, void *dst,
415 			unsigned long start,
416 			unsigned long len);
417 int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dst,
418 			       unsigned long start,
419 			       unsigned long len);
420 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
421 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
422 		const void *src);
423 void write_extent_buffer(struct extent_buffer *eb, const void *src,
424 			 unsigned long start, unsigned long len);
425 void copy_extent_buffer_full(struct extent_buffer *dst,
426 			     struct extent_buffer *src);
427 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
428 			unsigned long dst_offset, unsigned long src_offset,
429 			unsigned long len);
430 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
431 			   unsigned long src_offset, unsigned long len);
432 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
433 			   unsigned long src_offset, unsigned long len);
434 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
435 			   unsigned long len);
436 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
437 			   unsigned long pos);
438 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
439 			      unsigned long pos, unsigned long len);
440 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
441 				unsigned long pos, unsigned long len);
442 void clear_extent_buffer_dirty(struct extent_buffer *eb);
443 int set_extent_buffer_dirty(struct extent_buffer *eb);
444 void set_extent_buffer_uptodate(struct extent_buffer *eb);
445 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
446 int extent_buffer_uptodate(struct extent_buffer *eb);
447 int extent_buffer_under_io(struct extent_buffer *eb);
448 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
449 		      unsigned long min_len, char **map,
450 		      unsigned long *map_start,
451 		      unsigned long *map_len);
452 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
453 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
454 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
455 				 u64 delalloc_end, struct page *locked_page,
456 				 unsigned bits_to_clear,
457 				 unsigned long page_ops);
458 struct bio *
459 btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
460 		gfp_t gfp_flags);
461 struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs);
462 struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask);
463 
464 struct btrfs_fs_info;
465 struct btrfs_inode;
466 
467 int repair_io_failure(struct btrfs_inode *inode, u64 start, u64 length,
468 		u64 logical, struct page *page,
469 		unsigned int pg_offset, int mirror_num);
470 int clean_io_failure(struct btrfs_inode *inode, u64 start,
471 		struct page *page, unsigned int pg_offset);
472 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
473 int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
474 			 struct extent_buffer *eb, int mirror_num);
475 
476 /*
477  * When IO fails, either with EIO or csum verification fails, we
478  * try other mirrors that might have a good copy of the data.  This
479  * io_failure_record is used to record state as we go through all the
480  * mirrors.  If another mirror has good data, the page is set up to date
481  * and things continue.  If a good mirror can't be found, the original
482  * bio end_io callback is called to indicate things have failed.
483  */
484 struct io_failure_record {
485 	struct page *page;
486 	u64 start;
487 	u64 len;
488 	u64 logical;
489 	unsigned long bio_flags;
490 	int this_mirror;
491 	int failed_mirror;
492 	int in_validation;
493 };
494 
495 
496 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
497 		u64 end);
498 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
499 				struct io_failure_record **failrec_ret);
500 int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
501 			   struct io_failure_record *failrec, int fail_mirror);
502 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
503 				    struct io_failure_record *failrec,
504 				    struct page *page, int pg_offset, int icsum,
505 				    bio_end_io_t *endio_func, void *data);
506 int free_io_failure(struct btrfs_inode *inode, struct io_failure_record *rec);
507 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
508 noinline u64 find_lock_delalloc_range(struct inode *inode,
509 				      struct extent_io_tree *tree,
510 				      struct page *locked_page, u64 *start,
511 				      u64 *end, u64 max_bytes);
512 #endif
513 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
514 					       u64 start);
515 #endif
516