xref: /openbmc/linux/fs/btrfs/extent_io.h (revision 1e70d57e)
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 <linux/btrfs_tree.h>
10 #include "ulist.h"
11 
12 /*
13  * flags for bio submission. The high bits indicate the compression
14  * type for this bio
15  */
16 #define EXTENT_BIO_COMPRESSED 1
17 #define EXTENT_BIO_FLAG_SHIFT 16
18 
19 enum {
20 	EXTENT_BUFFER_UPTODATE,
21 	EXTENT_BUFFER_DIRTY,
22 	EXTENT_BUFFER_CORRUPT,
23 	/* this got triggered by readahead */
24 	EXTENT_BUFFER_READAHEAD,
25 	EXTENT_BUFFER_TREE_REF,
26 	EXTENT_BUFFER_STALE,
27 	EXTENT_BUFFER_WRITEBACK,
28 	/* read IO error */
29 	EXTENT_BUFFER_READ_ERR,
30 	EXTENT_BUFFER_UNMAPPED,
31 	EXTENT_BUFFER_IN_TREE,
32 	/* write IO error */
33 	EXTENT_BUFFER_WRITE_ERR,
34 	EXTENT_BUFFER_NO_CHECK,
35 	EXTENT_BUFFER_ZONE_FINISH,
36 };
37 
38 /* these are flags for __process_pages_contig */
39 #define PAGE_UNLOCK		(1 << 0)
40 /* Page starts writeback, clear dirty bit and set writeback bit */
41 #define PAGE_START_WRITEBACK	(1 << 1)
42 #define PAGE_END_WRITEBACK	(1 << 2)
43 #define PAGE_SET_ORDERED	(1 << 3)
44 #define PAGE_SET_ERROR		(1 << 4)
45 #define PAGE_LOCK		(1 << 5)
46 
47 /*
48  * page->private values.  Every page that is controlled by the extent
49  * map has page->private set to one.
50  */
51 #define EXTENT_PAGE_PRIVATE 1
52 
53 /*
54  * The extent buffer bitmap operations are done with byte granularity instead of
55  * word granularity for two reasons:
56  * 1. The bitmaps must be little-endian on disk.
57  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
58  *    single word in a bitmap may straddle two pages in the extent buffer.
59  */
60 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
61 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
62 #define BITMAP_FIRST_BYTE_MASK(start) \
63 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
64 #define BITMAP_LAST_BYTE_MASK(nbits) \
65 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
66 
67 struct btrfs_root;
68 struct btrfs_inode;
69 struct btrfs_io_bio;
70 struct btrfs_fs_info;
71 struct io_failure_record;
72 struct extent_io_tree;
73 
74 typedef blk_status_t (submit_bio_hook_t)(struct inode *inode, struct bio *bio,
75 					 int mirror_num,
76 					 unsigned long bio_flags);
77 
78 typedef blk_status_t (extent_submit_bio_start_t)(struct inode *inode,
79 		struct bio *bio, u64 dio_file_offset);
80 
81 #define INLINE_EXTENT_BUFFER_PAGES     (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE)
82 struct extent_buffer {
83 	u64 start;
84 	unsigned long len;
85 	unsigned long bflags;
86 	struct btrfs_fs_info *fs_info;
87 	spinlock_t refs_lock;
88 	atomic_t refs;
89 	atomic_t io_pages;
90 	int read_mirror;
91 	struct rcu_head rcu_head;
92 	pid_t lock_owner;
93 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
94 	s8 log_index;
95 
96 	struct rw_semaphore lock;
97 
98 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
99 	struct list_head release_list;
100 #ifdef CONFIG_BTRFS_DEBUG
101 	struct list_head leak_list;
102 #endif
103 };
104 
105 /*
106  * Structure to record info about the bio being assembled, and other info like
107  * how many bytes are there before stripe/ordered extent boundary.
108  */
109 struct btrfs_bio_ctrl {
110 	struct bio *bio;
111 	unsigned long bio_flags;
112 	u32 len_to_stripe_boundary;
113 	u32 len_to_oe_boundary;
114 };
115 
116 /*
117  * Structure to record how many bytes and which ranges are set/cleared
118  */
119 struct extent_changeset {
120 	/* How many bytes are set/cleared in this operation */
121 	u64 bytes_changed;
122 
123 	/* Changed ranges */
124 	struct ulist range_changed;
125 };
126 
127 static inline void extent_changeset_init(struct extent_changeset *changeset)
128 {
129 	changeset->bytes_changed = 0;
130 	ulist_init(&changeset->range_changed);
131 }
132 
133 static inline struct extent_changeset *extent_changeset_alloc(void)
134 {
135 	struct extent_changeset *ret;
136 
137 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
138 	if (!ret)
139 		return NULL;
140 
141 	extent_changeset_init(ret);
142 	return ret;
143 }
144 
145 static inline void extent_changeset_release(struct extent_changeset *changeset)
146 {
147 	if (!changeset)
148 		return;
149 	changeset->bytes_changed = 0;
150 	ulist_release(&changeset->range_changed);
151 }
152 
153 static inline void extent_changeset_free(struct extent_changeset *changeset)
154 {
155 	if (!changeset)
156 		return;
157 	extent_changeset_release(changeset);
158 	kfree(changeset);
159 }
160 
161 static inline void extent_set_compress_type(unsigned long *bio_flags,
162 					    int compress_type)
163 {
164 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
165 }
166 
167 static inline int extent_compress_type(unsigned long bio_flags)
168 {
169 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
170 }
171 
172 struct extent_map_tree;
173 
174 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
175 					  struct page *page, size_t pg_offset,
176 					  u64 start, u64 len);
177 
178 int try_release_extent_mapping(struct page *page, gfp_t mask);
179 int try_release_extent_buffer(struct page *page);
180 
181 int __must_check submit_one_bio(struct bio *bio, int mirror_num,
182 				unsigned long bio_flags);
183 int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
184 		      struct btrfs_bio_ctrl *bio_ctrl,
185 		      unsigned int read_flags, u64 *prev_em_start);
186 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
187 int extent_write_locked_range(struct inode *inode, u64 start, u64 end);
188 int extent_writepages(struct address_space *mapping,
189 		      struct writeback_control *wbc);
190 int btree_write_cache_pages(struct address_space *mapping,
191 			    struct writeback_control *wbc);
192 void extent_readahead(struct readahead_control *rac);
193 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
194 		  u64 start, u64 len);
195 int set_page_extent_mapped(struct page *page);
196 void clear_page_extent_mapped(struct page *page);
197 
198 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
199 					  u64 start, u64 owner_root, int level);
200 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
201 						  u64 start, unsigned long len);
202 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
203 						u64 start);
204 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
205 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
206 					 u64 start);
207 void free_extent_buffer(struct extent_buffer *eb);
208 void free_extent_buffer_stale(struct extent_buffer *eb);
209 #define WAIT_NONE	0
210 #define WAIT_COMPLETE	1
211 #define WAIT_PAGE_LOCK	2
212 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
213 			     int mirror_num);
214 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
215 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
216 				u64 bytenr, u64 owner_root, u64 gen, int level);
217 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
218 
219 static inline int num_extent_pages(const struct extent_buffer *eb)
220 {
221 	/*
222 	 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
223 	 * sectorsize, it's just eb->len >> PAGE_SHIFT.
224 	 *
225 	 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
226 	 * thus have to ensure we get at least one page.
227 	 */
228 	return (eb->len >> PAGE_SHIFT) ?: 1;
229 }
230 
231 static inline int extent_buffer_uptodate(const 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_nofault(const struct extent_buffer *eb,
242 				       void __user *dst, unsigned long start,
243 				       unsigned long len);
244 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src);
245 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb,
246 		const void *src);
247 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
248 			 unsigned long start, unsigned long len);
249 void copy_extent_buffer_full(const struct extent_buffer *dst,
250 			     const struct extent_buffer *src);
251 void copy_extent_buffer(const struct extent_buffer *dst,
252 			const struct extent_buffer *src,
253 			unsigned long dst_offset, unsigned long src_offset,
254 			unsigned long len);
255 void memcpy_extent_buffer(const struct extent_buffer *dst,
256 			  unsigned long dst_offset, unsigned long src_offset,
257 			  unsigned long len);
258 void memmove_extent_buffer(const struct extent_buffer *dst,
259 			   unsigned long dst_offset, unsigned long src_offset,
260 			   unsigned long len);
261 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
262 			   unsigned long len);
263 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
264 			   unsigned long pos);
265 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
266 			      unsigned long pos, unsigned long len);
267 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
268 				unsigned long start, unsigned long pos,
269 				unsigned long len);
270 void clear_extent_buffer_dirty(const struct extent_buffer *eb);
271 bool set_extent_buffer_dirty(struct extent_buffer *eb);
272 void set_extent_buffer_uptodate(struct extent_buffer *eb);
273 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
274 int extent_buffer_under_io(const struct extent_buffer *eb);
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 btrfs_inode *inode, u64 start, u64 end,
278 				  struct page *locked_page,
279 				  u32 bits_to_clear, unsigned long page_ops);
280 struct bio *btrfs_bio_alloc(unsigned int nr_iovecs);
281 struct bio *btrfs_bio_clone(struct bio *bio);
282 struct bio *btrfs_bio_clone_partial(struct bio *orig, u64 offset, u64 size);
283 
284 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
285 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num);
286 
287 /*
288  * When IO fails, either with EIO or csum verification fails, we
289  * try other mirrors that might have a good copy of the data.  This
290  * io_failure_record is used to record state as we go through all the
291  * mirrors.  If another mirror has good data, the sector is set up to date
292  * and things continue.  If a good mirror can't be found, the original
293  * bio end_io callback is called to indicate things have failed.
294  */
295 struct io_failure_record {
296 	struct page *page;
297 	u64 start;
298 	u64 len;
299 	u64 logical;
300 	unsigned long bio_flags;
301 	int this_mirror;
302 	int failed_mirror;
303 };
304 
305 int btrfs_repair_one_sector(struct inode *inode,
306 			    struct bio *failed_bio, u32 bio_offset,
307 			    struct page *page, unsigned int pgoff,
308 			    u64 start, int failed_mirror,
309 			    submit_bio_hook_t *submit_bio_hook);
310 
311 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
312 bool find_lock_delalloc_range(struct inode *inode,
313 			     struct page *locked_page, u64 *start,
314 			     u64 *end);
315 #endif
316 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
317 					       u64 start);
318 
319 #ifdef CONFIG_BTRFS_DEBUG
320 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
321 #else
322 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
323 #endif
324 
325 #endif
326