xref: /openbmc/linux/fs/btrfs/extent_io.h (revision 37744fee)
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 /*
11  * flags for bio submission. The high bits indicate the compression
12  * type for this bio
13  */
14 #define EXTENT_BIO_COMPRESSED 1
15 #define EXTENT_BIO_FLAG_SHIFT 16
16 
17 enum {
18 	EXTENT_BUFFER_UPTODATE,
19 	EXTENT_BUFFER_DIRTY,
20 	EXTENT_BUFFER_CORRUPT,
21 	/* this got triggered by readahead */
22 	EXTENT_BUFFER_READAHEAD,
23 	EXTENT_BUFFER_TREE_REF,
24 	EXTENT_BUFFER_STALE,
25 	EXTENT_BUFFER_WRITEBACK,
26 	/* read IO error */
27 	EXTENT_BUFFER_READ_ERR,
28 	EXTENT_BUFFER_UNMAPPED,
29 	EXTENT_BUFFER_IN_TREE,
30 	/* write IO error */
31 	EXTENT_BUFFER_WRITE_ERR,
32 };
33 
34 /* these are flags for __process_pages_contig */
35 #define PAGE_UNLOCK		(1 << 0)
36 #define PAGE_CLEAR_DIRTY	(1 << 1)
37 #define PAGE_SET_WRITEBACK	(1 << 2)
38 #define PAGE_END_WRITEBACK	(1 << 3)
39 #define PAGE_SET_PRIVATE2	(1 << 4)
40 #define PAGE_SET_ERROR		(1 << 5)
41 #define PAGE_LOCK		(1 << 6)
42 
43 /*
44  * page->private values.  Every page that is controlled by the extent
45  * map has page->private set to one.
46  */
47 #define EXTENT_PAGE_PRIVATE 1
48 
49 /*
50  * The extent buffer bitmap operations are done with byte granularity instead of
51  * word granularity for two reasons:
52  * 1. The bitmaps must be little-endian on disk.
53  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
54  *    single word in a bitmap may straddle two pages in the extent buffer.
55  */
56 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
57 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
58 #define BITMAP_FIRST_BYTE_MASK(start) \
59 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
60 #define BITMAP_LAST_BYTE_MASK(nbits) \
61 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
62 
63 struct btrfs_root;
64 struct btrfs_inode;
65 struct btrfs_io_bio;
66 struct io_failure_record;
67 struct extent_io_tree;
68 
69 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
70 		struct bio *bio, u64 bio_offset);
71 
72 struct extent_io_ops {
73 	/*
74 	 * The following callbacks must be always defined, the function
75 	 * pointer will be called unconditionally.
76 	 */
77 	blk_status_t (*submit_bio_hook)(struct inode *inode, struct bio *bio,
78 					int mirror_num, unsigned long bio_flags);
79 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
80 				    struct page *page, u64 start, u64 end,
81 				    int mirror);
82 };
83 
84 
85 #define INLINE_EXTENT_BUFFER_PAGES 16
86 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
87 struct extent_buffer {
88 	u64 start;
89 	unsigned long len;
90 	unsigned long bflags;
91 	struct btrfs_fs_info *fs_info;
92 	spinlock_t refs_lock;
93 	atomic_t refs;
94 	atomic_t io_pages;
95 	int read_mirror;
96 	struct rcu_head rcu_head;
97 	pid_t lock_owner;
98 
99 	int blocking_writers;
100 	atomic_t blocking_readers;
101 	bool lock_nested;
102 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
103 	short log_index;
104 
105 	/* protects write locks */
106 	rwlock_t lock;
107 
108 	/* readers use lock_wq while they wait for the write
109 	 * lock holders to unlock
110 	 */
111 	wait_queue_head_t write_lock_wq;
112 
113 	/* writers use read_lock_wq while they wait for readers
114 	 * to unlock
115 	 */
116 	wait_queue_head_t read_lock_wq;
117 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
118 #ifdef CONFIG_BTRFS_DEBUG
119 	int spinning_writers;
120 	atomic_t spinning_readers;
121 	atomic_t read_locks;
122 	int write_locks;
123 	struct list_head leak_list;
124 #endif
125 };
126 
127 /*
128  * Structure to record how many bytes and which ranges are set/cleared
129  */
130 struct extent_changeset {
131 	/* How many bytes are set/cleared in this operation */
132 	unsigned int bytes_changed;
133 
134 	/* Changed ranges */
135 	struct ulist range_changed;
136 };
137 
138 static inline void extent_changeset_init(struct extent_changeset *changeset)
139 {
140 	changeset->bytes_changed = 0;
141 	ulist_init(&changeset->range_changed);
142 }
143 
144 static inline struct extent_changeset *extent_changeset_alloc(void)
145 {
146 	struct extent_changeset *ret;
147 
148 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
149 	if (!ret)
150 		return NULL;
151 
152 	extent_changeset_init(ret);
153 	return ret;
154 }
155 
156 static inline void extent_changeset_release(struct extent_changeset *changeset)
157 {
158 	if (!changeset)
159 		return;
160 	changeset->bytes_changed = 0;
161 	ulist_release(&changeset->range_changed);
162 }
163 
164 static inline void extent_changeset_free(struct extent_changeset *changeset)
165 {
166 	if (!changeset)
167 		return;
168 	extent_changeset_release(changeset);
169 	kfree(changeset);
170 }
171 
172 static inline void extent_set_compress_type(unsigned long *bio_flags,
173 					    int compress_type)
174 {
175 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
176 }
177 
178 static inline int extent_compress_type(unsigned long bio_flags)
179 {
180 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
181 }
182 
183 struct extent_map_tree;
184 
185 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
186 					  struct page *page, size_t pg_offset,
187 					  u64 start, u64 len);
188 
189 int try_release_extent_mapping(struct page *page, gfp_t mask);
190 int try_release_extent_buffer(struct page *page);
191 
192 int extent_read_full_page(struct page *page, get_extent_t *get_extent,
193 			  int mirror_num);
194 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
195 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
196 			      int mode);
197 int extent_writepages(struct address_space *mapping,
198 		      struct writeback_control *wbc);
199 int btree_write_cache_pages(struct address_space *mapping,
200 			    struct writeback_control *wbc);
201 int extent_readpages(struct address_space *mapping, struct list_head *pages,
202 		     unsigned nr_pages);
203 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
204 		__u64 start, __u64 len);
205 void set_page_extent_mapped(struct page *page);
206 
207 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
208 					  u64 start);
209 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
210 						  u64 start, unsigned long len);
211 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
212 						u64 start);
213 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
214 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
215 					 u64 start);
216 void free_extent_buffer(struct extent_buffer *eb);
217 void free_extent_buffer_stale(struct extent_buffer *eb);
218 #define WAIT_NONE	0
219 #define WAIT_COMPLETE	1
220 #define WAIT_PAGE_LOCK	2
221 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
222 			     int mirror_num);
223 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
224 
225 static inline int num_extent_pages(const struct extent_buffer *eb)
226 {
227 	return (round_up(eb->start + eb->len, PAGE_SIZE) >> PAGE_SHIFT) -
228 	       (eb->start >> PAGE_SHIFT);
229 }
230 
231 static inline int extent_buffer_uptodate(struct extent_buffer *eb)
232 {
233 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
234 }
235 
236 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
237 			 unsigned long start, unsigned long len);
238 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
239 			unsigned long start,
240 			unsigned long len);
241 int read_extent_buffer_to_user(const struct extent_buffer *eb,
242 			       void __user *dst, unsigned long start,
243 			       unsigned long len);
244 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *src);
245 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
246 		const void *src);
247 void write_extent_buffer(struct extent_buffer *eb, const void *src,
248 			 unsigned long start, unsigned long len);
249 void copy_extent_buffer_full(struct extent_buffer *dst,
250 			     struct extent_buffer *src);
251 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
252 			unsigned long dst_offset, unsigned long src_offset,
253 			unsigned long len);
254 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
255 			   unsigned long src_offset, unsigned long len);
256 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
257 			   unsigned long src_offset, unsigned long len);
258 void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
259 			   unsigned long len);
260 int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
261 			   unsigned long pos);
262 void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
263 			      unsigned long pos, unsigned long len);
264 void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
265 				unsigned long pos, unsigned long len);
266 void clear_extent_buffer_dirty(struct extent_buffer *eb);
267 bool set_extent_buffer_dirty(struct extent_buffer *eb);
268 void set_extent_buffer_uptodate(struct extent_buffer *eb);
269 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
270 int extent_buffer_under_io(struct extent_buffer *eb);
271 int map_private_extent_buffer(const struct extent_buffer *eb,
272 			      unsigned long offset, unsigned long min_len,
273 			      char **map, unsigned long *map_start,
274 			      unsigned long *map_len);
275 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
276 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
277 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
278 				  struct page *locked_page,
279 				  unsigned bits_to_clear,
280 				  unsigned long page_ops);
281 struct bio *btrfs_bio_alloc(u64 first_byte);
282 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
283 struct bio *btrfs_bio_clone(struct bio *bio);
284 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
285 
286 struct btrfs_fs_info;
287 struct btrfs_inode;
288 
289 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
290 		      u64 length, u64 logical, struct page *page,
291 		      unsigned int pg_offset, int mirror_num);
292 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
293 int btrfs_repair_eb_io_failure(struct extent_buffer *eb, int mirror_num);
294 
295 /*
296  * When IO fails, either with EIO or csum verification fails, we
297  * try other mirrors that might have a good copy of the data.  This
298  * io_failure_record is used to record state as we go through all the
299  * mirrors.  If another mirror has good data, the page is set up to date
300  * and things continue.  If a good mirror can't be found, the original
301  * bio end_io callback is called to indicate things have failed.
302  */
303 struct io_failure_record {
304 	struct page *page;
305 	u64 start;
306 	u64 len;
307 	u64 logical;
308 	unsigned long bio_flags;
309 	int this_mirror;
310 	int failed_mirror;
311 	int in_validation;
312 };
313 
314 
315 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
316 			    struct io_failure_record *failrec, int fail_mirror);
317 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
318 				    struct io_failure_record *failrec,
319 				    struct page *page, int pg_offset, int icsum,
320 				    bio_end_io_t *endio_func, void *data);
321 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
322 bool find_lock_delalloc_range(struct inode *inode,
323 			     struct page *locked_page, u64 *start,
324 			     u64 *end);
325 #endif
326 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
327 					       u64 start);
328 
329 #ifdef CONFIG_BTRFS_DEBUG
330 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
331 #else
332 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
333 #endif
334 
335 #endif
336