Loading...
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 extent_clear_unlock_delalloc */
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
56/*
57 * page->private values. Every page that is controlled by the extent
58 * map has page->private set to one.
59 */
60#define EXTENT_PAGE_PRIVATE 1
61
62/*
63 * The extent buffer bitmap operations are done with byte granularity instead of
64 * word granularity for two reasons:
65 * 1. The bitmaps must be little-endian on disk.
66 * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
67 * single word in a bitmap may straddle two pages in the extent buffer.
68 */
69#define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
70#define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
71#define BITMAP_FIRST_BYTE_MASK(start) \
72 ((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
73#define BITMAP_LAST_BYTE_MASK(nbits) \
74 (BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
75
76static inline int le_test_bit(int nr, const u8 *addr)
77{
78 return 1U & (addr[BIT_BYTE(nr)] >> (nr & (BITS_PER_BYTE-1)));
79}
80
81extern void le_bitmap_set(u8 *map, unsigned int start, int len);
82extern void le_bitmap_clear(u8 *map, unsigned int start, int len);
83
84struct extent_state;
85struct btrfs_root;
86struct btrfs_io_bio;
87struct io_failure_record;
88
89typedef int (extent_submit_bio_hook_t)(struct inode *inode, struct bio *bio,
90 int mirror_num, unsigned long bio_flags,
91 u64 bio_offset);
92struct extent_io_ops {
93 int (*fill_delalloc)(struct inode *inode, struct page *locked_page,
94 u64 start, u64 end, int *page_started,
95 unsigned long *nr_written);
96 int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
97 extent_submit_bio_hook_t *submit_bio_hook;
98 int (*merge_bio_hook)(struct page *page, unsigned long offset,
99 size_t size, struct bio *bio,
100 unsigned long bio_flags);
101 int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
102 int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
103 struct page *page, u64 start, u64 end,
104 int mirror);
105 int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
106 struct extent_state *state, int uptodate);
107 void (*set_bit_hook)(struct inode *inode, struct extent_state *state,
108 unsigned *bits);
109 void (*clear_bit_hook)(struct inode *inode, struct extent_state *state,
110 unsigned *bits);
111 void (*merge_extent_hook)(struct inode *inode,
112 struct extent_state *new,
113 struct extent_state *other);
114 void (*split_extent_hook)(struct inode *inode,
115 struct extent_state *orig, u64 split);
116};
117
118struct extent_io_tree {
119 struct rb_root state;
120 struct address_space *mapping;
121 u64 dirty_bytes;
122 int track_uptodate;
123 spinlock_t lock;
124 const struct extent_io_ops *ops;
125};
126
127struct extent_state {
128 u64 start;
129 u64 end; /* inclusive */
130 struct rb_node rb_node;
131
132 /* ADD NEW ELEMENTS AFTER THIS */
133 wait_queue_head_t wq;
134 atomic_t refs;
135 unsigned state;
136
137 struct io_failure_record *failrec;
138
139#ifdef CONFIG_BTRFS_DEBUG
140 struct list_head leak_list;
141#endif
142};
143
144#define INLINE_EXTENT_BUFFER_PAGES 16
145#define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
146struct extent_buffer {
147 u64 start;
148 unsigned long len;
149 unsigned long bflags;
150 struct btrfs_fs_info *fs_info;
151 spinlock_t refs_lock;
152 atomic_t refs;
153 atomic_t io_pages;
154 int read_mirror;
155 struct rcu_head rcu_head;
156 pid_t lock_owner;
157
158 /* count of read lock holders on the extent buffer */
159 atomic_t write_locks;
160 atomic_t read_locks;
161 atomic_t blocking_writers;
162 atomic_t blocking_readers;
163 atomic_t spinning_readers;
164 atomic_t spinning_writers;
165 short lock_nested;
166 /* >= 0 if eb belongs to a log tree, -1 otherwise */
167 short log_index;
168
169 /* protects write locks */
170 rwlock_t lock;
171
172 /* readers use lock_wq while they wait for the write
173 * lock holders to unlock
174 */
175 wait_queue_head_t write_lock_wq;
176
177 /* writers use read_lock_wq while they wait for readers
178 * to unlock
179 */
180 wait_queue_head_t read_lock_wq;
181 struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
182#ifdef CONFIG_BTRFS_DEBUG
183 struct list_head leak_list;
184#endif
185};
186
187/*
188 * Structure to record how many bytes and which ranges are set/cleared
189 */
190struct extent_changeset {
191 /* How many bytes are set/cleared in this operation */
192 u64 bytes_changed;
193
194 /* Changed ranges */
195 struct ulist *range_changed;
196};
197
198static inline void extent_set_compress_type(unsigned long *bio_flags,
199 int compress_type)
200{
201 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
202}
203
204static inline int extent_compress_type(unsigned long bio_flags)
205{
206 return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
207}
208
209struct extent_map_tree;
210
211typedef struct extent_map *(get_extent_t)(struct inode *inode,
212 struct page *page,
213 size_t pg_offset,
214 u64 start, u64 len,
215 int create);
216
217void extent_io_tree_init(struct extent_io_tree *tree,
218 struct address_space *mapping);
219int try_release_extent_mapping(struct extent_map_tree *map,
220 struct extent_io_tree *tree, struct page *page,
221 gfp_t mask);
222int try_release_extent_buffer(struct page *page);
223int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
224 struct extent_state **cached);
225
226static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
227{
228 return lock_extent_bits(tree, start, end, NULL);
229}
230
231int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
232int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
233 get_extent_t *get_extent, int mirror_num);
234int __init extent_io_init(void);
235void extent_io_exit(void);
236
237u64 count_range_bits(struct extent_io_tree *tree,
238 u64 *start, u64 search_end,
239 u64 max_bytes, unsigned bits, int contig);
240
241void free_extent_state(struct extent_state *state);
242int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
243 unsigned bits, int filled,
244 struct extent_state *cached_state);
245int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
246 unsigned bits, struct extent_changeset *changeset);
247int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
248 unsigned bits, int wake, int delete,
249 struct extent_state **cached, gfp_t mask);
250
251static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
252{
253 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
254 GFP_NOFS);
255}
256
257static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
258 u64 end, struct extent_state **cached, gfp_t mask)
259{
260 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
261 mask);
262}
263
264static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
265 u64 end, unsigned bits)
266{
267 int wake = 0;
268
269 if (bits & EXTENT_LOCKED)
270 wake = 1;
271
272 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL,
273 GFP_NOFS);
274}
275
276int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
277 unsigned bits, struct extent_changeset *changeset);
278int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
279 unsigned bits, u64 *failed_start,
280 struct extent_state **cached_state, gfp_t mask);
281
282static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
283 u64 end, unsigned bits)
284{
285 return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
286}
287
288static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
289 u64 end, struct extent_state **cached_state, gfp_t mask)
290{
291 return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
292 cached_state, mask);
293}
294
295static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
296 u64 end, gfp_t mask)
297{
298 return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
299 NULL, mask);
300}
301
302static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
303 u64 end)
304{
305 return clear_extent_bit(tree, start, end,
306 EXTENT_DIRTY | EXTENT_DELALLOC |
307 EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
308}
309
310int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
311 unsigned bits, unsigned clear_bits,
312 struct extent_state **cached_state);
313
314static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
315 u64 end, struct extent_state **cached_state)
316{
317 return set_extent_bit(tree, start, end,
318 EXTENT_DELALLOC | EXTENT_UPTODATE,
319 NULL, cached_state, GFP_NOFS);
320}
321
322static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
323 u64 end, struct extent_state **cached_state)
324{
325 return set_extent_bit(tree, start, end,
326 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
327 NULL, cached_state, GFP_NOFS);
328}
329
330static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
331 u64 end)
332{
333 return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
334 GFP_NOFS);
335}
336
337static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
338 u64 end, struct extent_state **cached_state, gfp_t mask)
339{
340 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
341 cached_state, mask);
342}
343
344int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
345 u64 *start_ret, u64 *end_ret, unsigned bits,
346 struct extent_state **cached_state);
347int extent_invalidatepage(struct extent_io_tree *tree,
348 struct page *page, unsigned long offset);
349int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
350 get_extent_t *get_extent,
351 struct writeback_control *wbc);
352int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
353 u64 start, u64 end, get_extent_t *get_extent,
354 int mode);
355int extent_writepages(struct extent_io_tree *tree,
356 struct address_space *mapping,
357 get_extent_t *get_extent,
358 struct writeback_control *wbc);
359int btree_write_cache_pages(struct address_space *mapping,
360 struct writeback_control *wbc);
361int extent_readpages(struct extent_io_tree *tree,
362 struct address_space *mapping,
363 struct list_head *pages, unsigned nr_pages,
364 get_extent_t get_extent);
365int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
366 __u64 start, __u64 len, get_extent_t *get_extent);
367void set_page_extent_mapped(struct page *page);
368
369struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
370 u64 start);
371struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
372 u64 start, unsigned long len);
373struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
374 u64 start);
375struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
376struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
377 u64 start);
378void free_extent_buffer(struct extent_buffer *eb);
379void free_extent_buffer_stale(struct extent_buffer *eb);
380#define WAIT_NONE 0
381#define WAIT_COMPLETE 1
382#define WAIT_PAGE_LOCK 2
383int read_extent_buffer_pages(struct extent_io_tree *tree,
384 struct extent_buffer *eb, int wait,
385 get_extent_t *get_extent, int mirror_num);
386void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
387
388static inline unsigned long num_extent_pages(u64 start, u64 len)
389{
390 return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) -
391 (start >> PAGE_SHIFT);
392}
393
394static inline void extent_buffer_get(struct extent_buffer *eb)
395{
396 atomic_inc(&eb->refs);
397}
398
399int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
400 unsigned long start,
401 unsigned long len);
402void read_extent_buffer(struct extent_buffer *eb, void *dst,
403 unsigned long start,
404 unsigned long len);
405int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dst,
406 unsigned long start,
407 unsigned long len);
408void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
409void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
410 const void *src);
411void write_extent_buffer(struct extent_buffer *eb, const void *src,
412 unsigned long start, unsigned long len);
413void copy_extent_buffer_full(struct extent_buffer *dst,
414 struct extent_buffer *src);
415void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
416 unsigned long dst_offset, unsigned long src_offset,
417 unsigned long len);
418void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
419 unsigned long src_offset, unsigned long len);
420void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
421 unsigned long src_offset, unsigned long len);
422void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
423 unsigned long len);
424int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
425 unsigned long pos);
426void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
427 unsigned long pos, unsigned long len);
428void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
429 unsigned long pos, unsigned long len);
430void clear_extent_buffer_dirty(struct extent_buffer *eb);
431int set_extent_buffer_dirty(struct extent_buffer *eb);
432void set_extent_buffer_uptodate(struct extent_buffer *eb);
433void clear_extent_buffer_uptodate(struct extent_buffer *eb);
434int extent_buffer_uptodate(struct extent_buffer *eb);
435int extent_buffer_under_io(struct extent_buffer *eb);
436int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
437 unsigned long min_len, char **map,
438 unsigned long *map_start,
439 unsigned long *map_len);
440void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
441void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
442void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
443 u64 delalloc_end, struct page *locked_page,
444 unsigned bits_to_clear,
445 unsigned long page_ops);
446struct bio *
447btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
448 gfp_t gfp_flags);
449struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs);
450struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask);
451
452struct btrfs_fs_info;
453
454int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
455 struct page *page, unsigned int pg_offset,
456 int mirror_num);
457int clean_io_failure(struct inode *inode, u64 start, struct page *page,
458 unsigned int pg_offset);
459void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
460int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
461 struct extent_buffer *eb, int mirror_num);
462
463/*
464 * When IO fails, either with EIO or csum verification fails, we
465 * try other mirrors that might have a good copy of the data. This
466 * io_failure_record is used to record state as we go through all the
467 * mirrors. If another mirror has good data, the page is set up to date
468 * and things continue. If a good mirror can't be found, the original
469 * bio end_io callback is called to indicate things have failed.
470 */
471struct io_failure_record {
472 struct page *page;
473 u64 start;
474 u64 len;
475 u64 logical;
476 unsigned long bio_flags;
477 int this_mirror;
478 int failed_mirror;
479 int in_validation;
480};
481
482void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end);
483int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
484 struct io_failure_record **failrec_ret);
485int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
486 struct io_failure_record *failrec, int fail_mirror);
487struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
488 struct io_failure_record *failrec,
489 struct page *page, int pg_offset, int icsum,
490 bio_end_io_t *endio_func, void *data);
491int free_io_failure(struct inode *inode, struct io_failure_record *rec);
492#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
493noinline u64 find_lock_delalloc_range(struct inode *inode,
494 struct extent_io_tree *tree,
495 struct page *locked_page, u64 *start,
496 u64 *end, u64 max_bytes);
497#endif
498struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
499 u64 start);
500#endif
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/* 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_DUMMY 9
50#define EXTENT_BUFFER_IN_TREE 10
51#define EXTENT_BUFFER_WRITE_ERR 11 /* write IO error */
52
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)
61
62/*
63 * page->private values. Every page that is controlled by the extent
64 * map has page->private set to one.
65 */
66#define EXTENT_PAGE_PRIVATE 1
67
68/*
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.
74 */
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)))
81
82static inline int le_test_bit(int nr, const u8 *addr)
83{
84 return 1U & (addr[BIT_BYTE(nr)] >> (nr & (BITS_PER_BYTE-1)));
85}
86
87void le_bitmap_set(u8 *map, unsigned int start, int len);
88void le_bitmap_clear(u8 *map, unsigned int start, int len);
89
90struct extent_state;
91struct btrfs_root;
92struct btrfs_inode;
93struct btrfs_io_bio;
94struct io_failure_record;
95
96typedef blk_status_t (extent_submit_bio_hook_t)(void *private_data, struct bio *bio,
97 int mirror_num, unsigned long bio_flags,
98 u64 bio_offset);
99
100typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
101 struct bio *bio, u64 bio_offset);
102
103typedef blk_status_t (extent_submit_bio_done_t)(void *private_data,
104 struct bio *bio, int mirror_num);
105
106struct extent_io_ops {
107 /*
108 * The following callbacks must be allways defined, the function
109 * pointer will be called unconditionally.
110 */
111 extent_submit_bio_hook_t *submit_bio_hook;
112 int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
113 struct page *page, u64 start, u64 end,
114 int mirror);
115 int (*merge_bio_hook)(struct page *page, unsigned long offset,
116 size_t size, struct bio *bio,
117 unsigned long bio_flags);
118 int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
119 struct btrfs_fs_info *(*tree_fs_info)(void *private_data);
120 void (*set_range_writeback)(void *private_data, u64 start, u64 end);
121
122 /*
123 * Optional hooks, called if the pointer is not NULL
124 */
125 int (*fill_delalloc)(void *private_data, struct page *locked_page,
126 u64 start, u64 end, int *page_started,
127 unsigned long *nr_written,
128 struct writeback_control *wbc);
129
130 int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
131 void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
132 struct extent_state *state, int uptodate);
133 void (*set_bit_hook)(void *private_data, struct extent_state *state,
134 unsigned *bits);
135 void (*clear_bit_hook)(void *private_data,
136 struct extent_state *state,
137 unsigned *bits);
138 void (*merge_extent_hook)(void *private_data,
139 struct extent_state *new,
140 struct extent_state *other);
141 void (*split_extent_hook)(void *private_data,
142 struct extent_state *orig, u64 split);
143 void (*check_extent_io_range)(void *private_data, const char *caller,
144 u64 start, u64 end);
145};
146
147struct extent_io_tree {
148 struct rb_root state;
149 void *private_data;
150 u64 dirty_bytes;
151 int track_uptodate;
152 spinlock_t lock;
153 const struct extent_io_ops *ops;
154};
155
156struct extent_state {
157 u64 start;
158 u64 end; /* inclusive */
159 struct rb_node rb_node;
160
161 /* ADD NEW ELEMENTS AFTER THIS */
162 wait_queue_head_t wq;
163 refcount_t refs;
164 unsigned state;
165
166 struct io_failure_record *failrec;
167
168#ifdef CONFIG_BTRFS_DEBUG
169 struct list_head leak_list;
170#endif
171};
172
173#define INLINE_EXTENT_BUFFER_PAGES 16
174#define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
175struct extent_buffer {
176 u64 start;
177 unsigned long len;
178 unsigned long bflags;
179 struct btrfs_fs_info *fs_info;
180 spinlock_t refs_lock;
181 atomic_t refs;
182 atomic_t io_pages;
183 int read_mirror;
184 struct rcu_head rcu_head;
185 pid_t lock_owner;
186
187 /* count of read lock holders on the extent buffer */
188 atomic_t write_locks;
189 atomic_t read_locks;
190 atomic_t blocking_writers;
191 atomic_t blocking_readers;
192 atomic_t spinning_readers;
193 atomic_t spinning_writers;
194 short lock_nested;
195 /* >= 0 if eb belongs to a log tree, -1 otherwise */
196 short log_index;
197
198 /* protects write locks */
199 rwlock_t lock;
200
201 /* readers use lock_wq while they wait for the write
202 * lock holders to unlock
203 */
204 wait_queue_head_t write_lock_wq;
205
206 /* writers use read_lock_wq while they wait for readers
207 * to unlock
208 */
209 wait_queue_head_t read_lock_wq;
210 struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
211#ifdef CONFIG_BTRFS_DEBUG
212 struct list_head leak_list;
213#endif
214};
215
216/*
217 * Structure to record how many bytes and which ranges are set/cleared
218 */
219struct extent_changeset {
220 /* How many bytes are set/cleared in this operation */
221 unsigned int bytes_changed;
222
223 /* Changed ranges */
224 struct ulist range_changed;
225};
226
227static inline void extent_changeset_init(struct extent_changeset *changeset)
228{
229 changeset->bytes_changed = 0;
230 ulist_init(&changeset->range_changed);
231}
232
233static inline struct extent_changeset *extent_changeset_alloc(void)
234{
235 struct extent_changeset *ret;
236
237 ret = kmalloc(sizeof(*ret), GFP_KERNEL);
238 if (!ret)
239 return NULL;
240
241 extent_changeset_init(ret);
242 return ret;
243}
244
245static inline void extent_changeset_release(struct extent_changeset *changeset)
246{
247 if (!changeset)
248 return;
249 changeset->bytes_changed = 0;
250 ulist_release(&changeset->range_changed);
251}
252
253static inline void extent_changeset_free(struct extent_changeset *changeset)
254{
255 if (!changeset)
256 return;
257 extent_changeset_release(changeset);
258 kfree(changeset);
259}
260
261static inline void extent_set_compress_type(unsigned long *bio_flags,
262 int compress_type)
263{
264 *bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
265}
266
267static inline int extent_compress_type(unsigned long bio_flags)
268{
269 return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
270}
271
272struct extent_map_tree;
273
274typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
275 struct page *page,
276 size_t pg_offset,
277 u64 start, u64 len,
278 int create);
279
280void extent_io_tree_init(struct extent_io_tree *tree, void *private_data);
281int try_release_extent_mapping(struct extent_map_tree *map,
282 struct extent_io_tree *tree, struct page *page,
283 gfp_t mask);
284int try_release_extent_buffer(struct page *page);
285int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
286 struct extent_state **cached);
287
288static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
289{
290 return lock_extent_bits(tree, start, end, NULL);
291}
292
293int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
294int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
295 get_extent_t *get_extent, int mirror_num);
296int __init extent_io_init(void);
297void __cold extent_io_exit(void);
298
299u64 count_range_bits(struct extent_io_tree *tree,
300 u64 *start, u64 search_end,
301 u64 max_bytes, unsigned bits, int contig);
302
303void free_extent_state(struct extent_state *state);
304int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
305 unsigned bits, int filled,
306 struct extent_state *cached_state);
307int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
308 unsigned bits, struct extent_changeset *changeset);
309int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
310 unsigned bits, int wake, int delete,
311 struct extent_state **cached);
312int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
313 unsigned bits, int wake, int delete,
314 struct extent_state **cached, gfp_t mask,
315 struct extent_changeset *changeset);
316
317static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
318{
319 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL);
320}
321
322static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
323 u64 end, struct extent_state **cached)
324{
325 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
326 GFP_NOFS, NULL);
327}
328
329static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree,
330 u64 start, u64 end, struct extent_state **cached)
331{
332 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
333 GFP_ATOMIC, NULL);
334}
335
336static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
337 u64 end, unsigned bits)
338{
339 int wake = 0;
340
341 if (bits & EXTENT_LOCKED)
342 wake = 1;
343
344 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL);
345}
346
347int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
348 unsigned bits, struct extent_changeset *changeset);
349int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
350 unsigned bits, u64 *failed_start,
351 struct extent_state **cached_state, gfp_t mask);
352
353static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
354 u64 end, unsigned bits)
355{
356 return set_extent_bit(tree, start, end, bits, NULL, NULL, GFP_NOFS);
357}
358
359static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
360 u64 end, struct extent_state **cached_state)
361{
362 return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
363 cached_state, GFP_NOFS, NULL);
364}
365
366static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
367 u64 end, gfp_t mask)
368{
369 return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
370 NULL, mask);
371}
372
373static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
374 u64 end)
375{
376 return clear_extent_bit(tree, start, end,
377 EXTENT_DIRTY | EXTENT_DELALLOC |
378 EXTENT_DO_ACCOUNTING, 0, 0, NULL);
379}
380
381int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
382 unsigned bits, unsigned clear_bits,
383 struct extent_state **cached_state);
384
385static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
386 u64 end, unsigned int extra_bits,
387 struct extent_state **cached_state)
388{
389 return set_extent_bit(tree, start, end,
390 EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits,
391 NULL, cached_state, GFP_NOFS);
392}
393
394static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
395 u64 end, struct extent_state **cached_state)
396{
397 return set_extent_bit(tree, start, end,
398 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
399 NULL, cached_state, GFP_NOFS);
400}
401
402static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
403 u64 end)
404{
405 return set_extent_bit(tree, start, end, EXTENT_NEW, NULL, NULL,
406 GFP_NOFS);
407}
408
409static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
410 u64 end, struct extent_state **cached_state, gfp_t mask)
411{
412 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
413 cached_state, mask);
414}
415
416int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
417 u64 *start_ret, u64 *end_ret, unsigned bits,
418 struct extent_state **cached_state);
419int extent_invalidatepage(struct extent_io_tree *tree,
420 struct page *page, unsigned long offset);
421int extent_write_full_page(struct page *page, struct writeback_control *wbc);
422int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
423 int mode);
424int extent_writepages(struct extent_io_tree *tree,
425 struct address_space *mapping,
426 struct writeback_control *wbc);
427int btree_write_cache_pages(struct address_space *mapping,
428 struct writeback_control *wbc);
429int extent_readpages(struct extent_io_tree *tree,
430 struct address_space *mapping,
431 struct list_head *pages, unsigned nr_pages);
432int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
433 __u64 start, __u64 len);
434void set_page_extent_mapped(struct page *page);
435
436struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
437 u64 start);
438struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
439 u64 start, unsigned long len);
440struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
441 u64 start);
442struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
443struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
444 u64 start);
445void free_extent_buffer(struct extent_buffer *eb);
446void free_extent_buffer_stale(struct extent_buffer *eb);
447#define WAIT_NONE 0
448#define WAIT_COMPLETE 1
449#define WAIT_PAGE_LOCK 2
450int read_extent_buffer_pages(struct extent_io_tree *tree,
451 struct extent_buffer *eb, int wait,
452 int mirror_num);
453void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
454
455static inline unsigned long num_extent_pages(u64 start, u64 len)
456{
457 return ((start + len + PAGE_SIZE - 1) >> PAGE_SHIFT) -
458 (start >> PAGE_SHIFT);
459}
460
461static inline void extent_buffer_get(struct extent_buffer *eb)
462{
463 atomic_inc(&eb->refs);
464}
465
466static inline int extent_buffer_uptodate(struct extent_buffer *eb)
467{
468 return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
469}
470
471int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
472 unsigned long start, unsigned long len);
473void read_extent_buffer(const struct extent_buffer *eb, void *dst,
474 unsigned long start,
475 unsigned long len);
476int read_extent_buffer_to_user(const struct extent_buffer *eb,
477 void __user *dst, unsigned long start,
478 unsigned long len);
479void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
480void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
481 const void *src);
482void write_extent_buffer(struct extent_buffer *eb, const void *src,
483 unsigned long start, unsigned long len);
484void copy_extent_buffer_full(struct extent_buffer *dst,
485 struct extent_buffer *src);
486void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
487 unsigned long dst_offset, unsigned long src_offset,
488 unsigned long len);
489void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
490 unsigned long src_offset, unsigned long len);
491void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
492 unsigned long src_offset, unsigned long len);
493void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
494 unsigned long len);
495int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
496 unsigned long pos);
497void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
498 unsigned long pos, unsigned long len);
499void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
500 unsigned long pos, unsigned long len);
501void clear_extent_buffer_dirty(struct extent_buffer *eb);
502int set_extent_buffer_dirty(struct extent_buffer *eb);
503void set_extent_buffer_uptodate(struct extent_buffer *eb);
504void clear_extent_buffer_uptodate(struct extent_buffer *eb);
505int extent_buffer_under_io(struct extent_buffer *eb);
506int map_private_extent_buffer(const struct extent_buffer *eb,
507 unsigned long offset, unsigned long min_len,
508 char **map, unsigned long *map_start,
509 unsigned long *map_len);
510void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
511void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
512void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
513 u64 delalloc_end, struct page *locked_page,
514 unsigned bits_to_clear,
515 unsigned long page_ops);
516struct bio *btrfs_bio_alloc(struct block_device *bdev, u64 first_byte);
517struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
518struct bio *btrfs_bio_clone(struct bio *bio);
519struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
520
521struct btrfs_fs_info;
522struct btrfs_inode;
523
524int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
525 u64 length, u64 logical, struct page *page,
526 unsigned int pg_offset, int mirror_num);
527int clean_io_failure(struct btrfs_fs_info *fs_info,
528 struct extent_io_tree *failure_tree,
529 struct extent_io_tree *io_tree, u64 start,
530 struct page *page, u64 ino, unsigned int pg_offset);
531void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
532int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
533 struct extent_buffer *eb, int mirror_num);
534
535/*
536 * When IO fails, either with EIO or csum verification fails, we
537 * try other mirrors that might have a good copy of the data. This
538 * io_failure_record is used to record state as we go through all the
539 * mirrors. If another mirror has good data, the page is set up to date
540 * and things continue. If a good mirror can't be found, the original
541 * bio end_io callback is called to indicate things have failed.
542 */
543struct io_failure_record {
544 struct page *page;
545 u64 start;
546 u64 len;
547 u64 logical;
548 unsigned long bio_flags;
549 int this_mirror;
550 int failed_mirror;
551 int in_validation;
552};
553
554
555void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
556 u64 end);
557int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
558 struct io_failure_record **failrec_ret);
559bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
560 struct io_failure_record *failrec, int fail_mirror);
561struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
562 struct io_failure_record *failrec,
563 struct page *page, int pg_offset, int icsum,
564 bio_end_io_t *endio_func, void *data);
565int free_io_failure(struct extent_io_tree *failure_tree,
566 struct extent_io_tree *io_tree,
567 struct io_failure_record *rec);
568#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
569noinline u64 find_lock_delalloc_range(struct inode *inode,
570 struct extent_io_tree *tree,
571 struct page *locked_page, u64 *start,
572 u64 *end, u64 max_bytes);
573#endif
574struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
575 u64 start);
576
577#endif