xref: /openbmc/linux/fs/btrfs/extent_io.h (revision 802b8362)
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 <linux/fiemap.h>
9 #include "ulist.h"
10 
11 /*
12  * flags for bio submission. The high bits indicate the compression
13  * type for this bio
14  */
15 #define EXTENT_BIO_COMPRESSED 1
16 #define EXTENT_BIO_FLAG_SHIFT 16
17 
18 enum {
19 	EXTENT_BUFFER_UPTODATE,
20 	EXTENT_BUFFER_DIRTY,
21 	EXTENT_BUFFER_CORRUPT,
22 	/* this got triggered by readahead */
23 	EXTENT_BUFFER_READAHEAD,
24 	EXTENT_BUFFER_TREE_REF,
25 	EXTENT_BUFFER_STALE,
26 	EXTENT_BUFFER_WRITEBACK,
27 	/* read IO error */
28 	EXTENT_BUFFER_READ_ERR,
29 	EXTENT_BUFFER_UNMAPPED,
30 	EXTENT_BUFFER_IN_TREE,
31 	/* write IO error */
32 	EXTENT_BUFFER_WRITE_ERR,
33 };
34 
35 /* these are flags for __process_pages_contig */
36 #define PAGE_UNLOCK		(1 << 0)
37 #define PAGE_CLEAR_DIRTY	(1 << 1)
38 #define PAGE_SET_WRITEBACK	(1 << 2)
39 #define PAGE_END_WRITEBACK	(1 << 3)
40 #define PAGE_SET_PRIVATE2	(1 << 4)
41 #define PAGE_SET_ERROR		(1 << 5)
42 #define PAGE_LOCK		(1 << 6)
43 
44 /*
45  * page->private values.  Every page that is controlled by the extent
46  * map has page->private set to one.
47  */
48 #define EXTENT_PAGE_PRIVATE 1
49 
50 /*
51  * The extent buffer bitmap operations are done with byte granularity instead of
52  * word granularity for two reasons:
53  * 1. The bitmaps must be little-endian on disk.
54  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
55  *    single word in a bitmap may straddle two pages in the extent buffer.
56  */
57 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
58 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
59 #define BITMAP_FIRST_BYTE_MASK(start) \
60 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
61 #define BITMAP_LAST_BYTE_MASK(nbits) \
62 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
63 
64 struct btrfs_root;
65 struct btrfs_inode;
66 struct btrfs_io_bio;
67 struct io_failure_record;
68 struct extent_io_tree;
69 
70 typedef blk_status_t (submit_bio_hook_t)(struct inode *inode, struct bio *bio,
71 					 int mirror_num,
72 					 unsigned long bio_flags);
73 
74 typedef blk_status_t (extent_submit_bio_start_t)(void *private_data,
75 		struct bio *bio, u64 bio_offset);
76 
77 struct extent_io_ops {
78 	/*
79 	 * The following callbacks must be always defined, the function
80 	 * pointer will be called unconditionally.
81 	 */
82 	submit_bio_hook_t *submit_bio_hook;
83 	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
84 				    struct page *page, u64 start, u64 end,
85 				    int mirror);
86 };
87 
88 
89 #define INLINE_EXTENT_BUFFER_PAGES 16
90 #define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_SIZE)
91 struct extent_buffer {
92 	u64 start;
93 	unsigned long len;
94 	unsigned long bflags;
95 	struct btrfs_fs_info *fs_info;
96 	spinlock_t refs_lock;
97 	atomic_t refs;
98 	atomic_t io_pages;
99 	int read_mirror;
100 	struct rcu_head rcu_head;
101 	pid_t lock_owner;
102 
103 	int blocking_writers;
104 	atomic_t blocking_readers;
105 	bool lock_nested;
106 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
107 	short log_index;
108 
109 	/* protects write locks */
110 	rwlock_t lock;
111 
112 	/* readers use lock_wq while they wait for the write
113 	 * lock holders to unlock
114 	 */
115 	wait_queue_head_t write_lock_wq;
116 
117 	/* writers use read_lock_wq while they wait for readers
118 	 * to unlock
119 	 */
120 	wait_queue_head_t read_lock_wq;
121 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
122 #ifdef CONFIG_BTRFS_DEBUG
123 	int spinning_writers;
124 	atomic_t spinning_readers;
125 	atomic_t read_locks;
126 	int write_locks;
127 	struct list_head leak_list;
128 #endif
129 };
130 
131 /*
132  * Structure to record how many bytes and which ranges are set/cleared
133  */
134 struct extent_changeset {
135 	/* How many bytes are set/cleared in this operation */
136 	unsigned int bytes_changed;
137 
138 	/* Changed ranges */
139 	struct ulist range_changed;
140 };
141 
142 static inline void extent_changeset_init(struct extent_changeset *changeset)
143 {
144 	changeset->bytes_changed = 0;
145 	ulist_init(&changeset->range_changed);
146 }
147 
148 static inline struct extent_changeset *extent_changeset_alloc(void)
149 {
150 	struct extent_changeset *ret;
151 
152 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
153 	if (!ret)
154 		return NULL;
155 
156 	extent_changeset_init(ret);
157 	return ret;
158 }
159 
160 static inline void extent_changeset_release(struct extent_changeset *changeset)
161 {
162 	if (!changeset)
163 		return;
164 	changeset->bytes_changed = 0;
165 	ulist_release(&changeset->range_changed);
166 }
167 
168 static inline void extent_changeset_free(struct extent_changeset *changeset)
169 {
170 	if (!changeset)
171 		return;
172 	extent_changeset_release(changeset);
173 	kfree(changeset);
174 }
175 
176 static inline void extent_set_compress_type(unsigned long *bio_flags,
177 					    int compress_type)
178 {
179 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
180 }
181 
182 static inline int extent_compress_type(unsigned long bio_flags)
183 {
184 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
185 }
186 
187 struct extent_map_tree;
188 
189 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
190 					  struct page *page, size_t pg_offset,
191 					  u64 start, u64 len);
192 
193 int try_release_extent_mapping(struct page *page, gfp_t mask);
194 int try_release_extent_buffer(struct page *page);
195 
196 int extent_read_full_page(struct page *page, get_extent_t *get_extent,
197 			  int mirror_num);
198 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
199 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
200 			      int mode);
201 int extent_writepages(struct address_space *mapping,
202 		      struct writeback_control *wbc);
203 int btree_write_cache_pages(struct address_space *mapping,
204 			    struct writeback_control *wbc);
205 void extent_readahead(struct readahead_control *rac);
206 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
207 		  u64 start, u64 len);
208 void set_page_extent_mapped(struct page *page);
209 
210 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
211 					  u64 start);
212 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
213 						  u64 start, unsigned long len);
214 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
215 						u64 start);
216 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
217 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
218 					 u64 start);
219 void free_extent_buffer(struct extent_buffer *eb);
220 void free_extent_buffer_stale(struct extent_buffer *eb);
221 #define WAIT_NONE	0
222 #define WAIT_COMPLETE	1
223 #define WAIT_PAGE_LOCK	2
224 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
225 			     int mirror_num);
226 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
227 
228 static inline int num_extent_pages(const struct extent_buffer *eb)
229 {
230 	return (round_up(eb->start + eb->len, PAGE_SIZE) >> PAGE_SHIFT) -
231 	       (eb->start >> PAGE_SHIFT);
232 }
233 
234 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
235 {
236 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
237 }
238 
239 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
240 			 unsigned long start, unsigned long len);
241 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
242 			unsigned long start,
243 			unsigned long len);
244 int read_extent_buffer_to_user(const struct extent_buffer *eb,
245 			       void __user *dst, unsigned long start,
246 			       unsigned long len);
247 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src);
248 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb,
249 		const void *src);
250 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
251 			 unsigned long start, unsigned long len);
252 void copy_extent_buffer_full(const struct extent_buffer *dst,
253 			     const struct extent_buffer *src);
254 void copy_extent_buffer(const struct extent_buffer *dst,
255 			const struct extent_buffer *src,
256 			unsigned long dst_offset, unsigned long src_offset,
257 			unsigned long len);
258 void memcpy_extent_buffer(const struct extent_buffer *dst,
259 			  unsigned long dst_offset, unsigned long src_offset,
260 			  unsigned long len);
261 void memmove_extent_buffer(const struct extent_buffer *dst,
262 			   unsigned long dst_offset, unsigned long src_offset,
263 			   unsigned long len);
264 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
265 			   unsigned long len);
266 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
267 			   unsigned long pos);
268 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
269 			      unsigned long pos, unsigned long len);
270 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
271 				unsigned long start, unsigned long pos,
272 				unsigned long len);
273 void clear_extent_buffer_dirty(const struct extent_buffer *eb);
274 bool set_extent_buffer_dirty(struct extent_buffer *eb);
275 void set_extent_buffer_uptodate(struct extent_buffer *eb);
276 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
277 int extent_buffer_under_io(const struct extent_buffer *eb);
278 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
279 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
280 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
281 				  struct page *locked_page,
282 				  unsigned bits_to_clear,
283 				  unsigned long page_ops);
284 struct bio *btrfs_bio_alloc(u64 first_byte);
285 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
286 struct bio *btrfs_bio_clone(struct bio *bio);
287 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
288 
289 struct btrfs_fs_info;
290 struct btrfs_inode;
291 
292 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
293 		      u64 length, u64 logical, struct page *page,
294 		      unsigned int pg_offset, int mirror_num);
295 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
296 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num);
297 
298 /*
299  * When IO fails, either with EIO or csum verification fails, we
300  * try other mirrors that might have a good copy of the data.  This
301  * io_failure_record is used to record state as we go through all the
302  * mirrors.  If another mirror has good data, the page is set up to date
303  * and things continue.  If a good mirror can't be found, the original
304  * bio end_io callback is called to indicate things have failed.
305  */
306 struct io_failure_record {
307 	struct page *page;
308 	u64 start;
309 	u64 len;
310 	u64 logical;
311 	unsigned long bio_flags;
312 	int this_mirror;
313 	int failed_mirror;
314 	int in_validation;
315 };
316 
317 
318 blk_status_t btrfs_submit_read_repair(struct inode *inode,
319 				      struct bio *failed_bio, u64 phy_offset,
320 				      struct page *page, unsigned int pgoff,
321 				      u64 start, u64 end, int failed_mirror,
322 				      submit_bio_hook_t *submit_bio_hook);
323 
324 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
325 bool find_lock_delalloc_range(struct inode *inode,
326 			     struct page *locked_page, u64 *start,
327 			     u64 *end);
328 #endif
329 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
330 					       u64 start);
331 
332 #ifdef CONFIG_BTRFS_DEBUG
333 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
334 #else
335 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
336 #endif
337 
338 #endif
339