xref: /openbmc/linux/fs/btrfs/extent_io.h (revision 9dbbc3b9)
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 };
36 
37 /* these are flags for __process_pages_contig */
38 #define PAGE_UNLOCK		(1 << 0)
39 /* Page starts writeback, clear dirty bit and set writeback bit */
40 #define PAGE_START_WRITEBACK	(1 << 1)
41 #define PAGE_END_WRITEBACK	(1 << 2)
42 #define PAGE_SET_ORDERED	(1 << 3)
43 #define PAGE_SET_ERROR		(1 << 4)
44 #define PAGE_LOCK		(1 << 5)
45 
46 /*
47  * page->private values.  Every page that is controlled by the extent
48  * map has page->private set to one.
49  */
50 #define EXTENT_PAGE_PRIVATE 1
51 
52 /*
53  * The extent buffer bitmap operations are done with byte granularity instead of
54  * word granularity for two reasons:
55  * 1. The bitmaps must be little-endian on disk.
56  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
57  *    single word in a bitmap may straddle two pages in the extent buffer.
58  */
59 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
60 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
61 #define BITMAP_FIRST_BYTE_MASK(start) \
62 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
63 #define BITMAP_LAST_BYTE_MASK(nbits) \
64 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
65 
66 struct btrfs_root;
67 struct btrfs_inode;
68 struct btrfs_io_bio;
69 struct btrfs_fs_info;
70 struct io_failure_record;
71 struct extent_io_tree;
72 
73 typedef blk_status_t (submit_bio_hook_t)(struct inode *inode, struct bio *bio,
74 					 int mirror_num,
75 					 unsigned long bio_flags);
76 
77 typedef blk_status_t (extent_submit_bio_start_t)(struct inode *inode,
78 		struct bio *bio, u64 dio_file_offset);
79 
80 #define INLINE_EXTENT_BUFFER_PAGES     (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE)
81 struct extent_buffer {
82 	u64 start;
83 	unsigned long len;
84 	unsigned long bflags;
85 	struct btrfs_fs_info *fs_info;
86 	spinlock_t refs_lock;
87 	atomic_t refs;
88 	atomic_t io_pages;
89 	int read_mirror;
90 	struct rcu_head rcu_head;
91 	pid_t lock_owner;
92 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
93 	s8 log_index;
94 
95 	struct rw_semaphore lock;
96 
97 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
98 	struct list_head release_list;
99 #ifdef CONFIG_BTRFS_DEBUG
100 	struct list_head leak_list;
101 #endif
102 };
103 
104 /*
105  * Structure to record info about the bio being assembled, and other info like
106  * how many bytes are there before stripe/ordered extent boundary.
107  */
108 struct btrfs_bio_ctrl {
109 	struct bio *bio;
110 	unsigned long bio_flags;
111 	u32 len_to_stripe_boundary;
112 	u32 len_to_oe_boundary;
113 };
114 
115 /*
116  * Structure to record how many bytes and which ranges are set/cleared
117  */
118 struct extent_changeset {
119 	/* How many bytes are set/cleared in this operation */
120 	unsigned int bytes_changed;
121 
122 	/* Changed ranges */
123 	struct ulist range_changed;
124 };
125 
126 static inline void extent_changeset_init(struct extent_changeset *changeset)
127 {
128 	changeset->bytes_changed = 0;
129 	ulist_init(&changeset->range_changed);
130 }
131 
132 static inline struct extent_changeset *extent_changeset_alloc(void)
133 {
134 	struct extent_changeset *ret;
135 
136 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
137 	if (!ret)
138 		return NULL;
139 
140 	extent_changeset_init(ret);
141 	return ret;
142 }
143 
144 static inline void extent_changeset_release(struct extent_changeset *changeset)
145 {
146 	if (!changeset)
147 		return;
148 	changeset->bytes_changed = 0;
149 	ulist_release(&changeset->range_changed);
150 }
151 
152 static inline void extent_changeset_free(struct extent_changeset *changeset)
153 {
154 	if (!changeset)
155 		return;
156 	extent_changeset_release(changeset);
157 	kfree(changeset);
158 }
159 
160 static inline void extent_set_compress_type(unsigned long *bio_flags,
161 					    int compress_type)
162 {
163 	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
164 }
165 
166 static inline int extent_compress_type(unsigned long bio_flags)
167 {
168 	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
169 }
170 
171 struct extent_map_tree;
172 
173 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
174 					  struct page *page, size_t pg_offset,
175 					  u64 start, u64 len);
176 
177 int try_release_extent_mapping(struct page *page, gfp_t mask);
178 int try_release_extent_buffer(struct page *page);
179 
180 int __must_check submit_one_bio(struct bio *bio, int mirror_num,
181 				unsigned long bio_flags);
182 int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
183 		      struct btrfs_bio_ctrl *bio_ctrl,
184 		      unsigned int read_flags, u64 *prev_em_start);
185 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
186 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
187 			      int mode);
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(u64 first_byte);
281 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs);
282 struct bio *btrfs_bio_clone(struct bio *bio);
283 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size);
284 
285 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
286 		      u64 length, u64 logical, struct page *page,
287 		      unsigned int pg_offset, int mirror_num);
288 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
289 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num);
290 
291 /*
292  * When IO fails, either with EIO or csum verification fails, we
293  * try other mirrors that might have a good copy of the data.  This
294  * io_failure_record is used to record state as we go through all the
295  * mirrors.  If another mirror has good data, the sector is set up to date
296  * and things continue.  If a good mirror can't be found, the original
297  * bio end_io callback is called to indicate things have failed.
298  */
299 struct io_failure_record {
300 	struct page *page;
301 	u64 start;
302 	u64 len;
303 	u64 logical;
304 	unsigned long bio_flags;
305 	int this_mirror;
306 	int failed_mirror;
307 };
308 
309 int btrfs_repair_one_sector(struct inode *inode,
310 			    struct bio *failed_bio, u32 bio_offset,
311 			    struct page *page, unsigned int pgoff,
312 			    u64 start, int failed_mirror,
313 			    submit_bio_hook_t *submit_bio_hook);
314 
315 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
316 bool find_lock_delalloc_range(struct inode *inode,
317 			     struct page *locked_page, u64 *start,
318 			     u64 *end);
319 #endif
320 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
321 					       u64 start);
322 
323 #ifdef CONFIG_BTRFS_DEBUG
324 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
325 #else
326 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
327 #endif
328 
329 #endif
330