xref: /linux/fs/btrfs/extent_io.h (revision 24bce201d79807b668bf9d9e0aca801c5c0d5f78)
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 "compression.h"
11 #include "ulist.h"
12 
13 enum {
14 	EXTENT_BUFFER_UPTODATE,
15 	EXTENT_BUFFER_DIRTY,
16 	EXTENT_BUFFER_CORRUPT,
17 	/* this got triggered by readahead */
18 	EXTENT_BUFFER_READAHEAD,
19 	EXTENT_BUFFER_TREE_REF,
20 	EXTENT_BUFFER_STALE,
21 	EXTENT_BUFFER_WRITEBACK,
22 	/* read IO error */
23 	EXTENT_BUFFER_READ_ERR,
24 	EXTENT_BUFFER_UNMAPPED,
25 	EXTENT_BUFFER_IN_TREE,
26 	/* write IO error */
27 	EXTENT_BUFFER_WRITE_ERR,
28 	EXTENT_BUFFER_NO_CHECK,
29 };
30 
31 /* these are flags for __process_pages_contig */
32 #define PAGE_UNLOCK		(1 << 0)
33 /* Page starts writeback, clear dirty bit and set writeback bit */
34 #define PAGE_START_WRITEBACK	(1 << 1)
35 #define PAGE_END_WRITEBACK	(1 << 2)
36 #define PAGE_SET_ORDERED	(1 << 3)
37 #define PAGE_SET_ERROR		(1 << 4)
38 #define PAGE_LOCK		(1 << 5)
39 
40 /*
41  * page->private values.  Every page that is controlled by the extent
42  * map has page->private set to one.
43  */
44 #define EXTENT_PAGE_PRIVATE 1
45 
46 /*
47  * The extent buffer bitmap operations are done with byte granularity instead of
48  * word granularity for two reasons:
49  * 1. The bitmaps must be little-endian on disk.
50  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
51  *    single word in a bitmap may straddle two pages in the extent buffer.
52  */
53 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
54 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
55 #define BITMAP_FIRST_BYTE_MASK(start) \
56 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
57 #define BITMAP_LAST_BYTE_MASK(nbits) \
58 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
59 
60 struct btrfs_root;
61 struct btrfs_inode;
62 struct btrfs_io_bio;
63 struct btrfs_fs_info;
64 struct io_failure_record;
65 struct extent_io_tree;
66 
67 typedef void (submit_bio_hook_t)(struct inode *inode, struct bio *bio,
68 					 int mirror_num,
69 					 enum btrfs_compression_type compress_type);
70 
71 typedef blk_status_t (extent_submit_bio_start_t)(struct inode *inode,
72 		struct bio *bio, u64 dio_file_offset);
73 
74 #define INLINE_EXTENT_BUFFER_PAGES     (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE)
75 struct extent_buffer {
76 	u64 start;
77 	unsigned long len;
78 	unsigned long bflags;
79 	struct btrfs_fs_info *fs_info;
80 	spinlock_t refs_lock;
81 	atomic_t refs;
82 	atomic_t io_pages;
83 	int read_mirror;
84 	struct rcu_head rcu_head;
85 	pid_t lock_owner;
86 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
87 	s8 log_index;
88 
89 	struct rw_semaphore lock;
90 
91 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
92 	struct list_head release_list;
93 #ifdef CONFIG_BTRFS_DEBUG
94 	struct list_head leak_list;
95 #endif
96 };
97 
98 /*
99  * Structure to record how many bytes and which ranges are set/cleared
100  */
101 struct extent_changeset {
102 	/* How many bytes are set/cleared in this operation */
103 	u64 bytes_changed;
104 
105 	/* Changed ranges */
106 	struct ulist range_changed;
107 };
108 
109 static inline void extent_changeset_init(struct extent_changeset *changeset)
110 {
111 	changeset->bytes_changed = 0;
112 	ulist_init(&changeset->range_changed);
113 }
114 
115 static inline struct extent_changeset *extent_changeset_alloc(void)
116 {
117 	struct extent_changeset *ret;
118 
119 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
120 	if (!ret)
121 		return NULL;
122 
123 	extent_changeset_init(ret);
124 	return ret;
125 }
126 
127 static inline void extent_changeset_release(struct extent_changeset *changeset)
128 {
129 	if (!changeset)
130 		return;
131 	changeset->bytes_changed = 0;
132 	ulist_release(&changeset->range_changed);
133 }
134 
135 static inline void extent_changeset_free(struct extent_changeset *changeset)
136 {
137 	if (!changeset)
138 		return;
139 	extent_changeset_release(changeset);
140 	kfree(changeset);
141 }
142 
143 struct extent_map_tree;
144 
145 typedef struct extent_map *(get_extent_t)(struct btrfs_inode *inode,
146 					  struct page *page, size_t pg_offset,
147 					  u64 start, u64 len);
148 
149 int try_release_extent_mapping(struct page *page, gfp_t mask);
150 int try_release_extent_buffer(struct page *page);
151 
152 int btrfs_read_folio(struct file *file, struct folio *folio);
153 int extent_write_full_page(struct page *page, struct writeback_control *wbc);
154 int extent_write_locked_range(struct inode *inode, u64 start, u64 end);
155 int extent_writepages(struct address_space *mapping,
156 		      struct writeback_control *wbc);
157 int btree_write_cache_pages(struct address_space *mapping,
158 			    struct writeback_control *wbc);
159 void extent_readahead(struct readahead_control *rac);
160 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
161 		  u64 start, u64 len);
162 int set_page_extent_mapped(struct page *page);
163 void clear_page_extent_mapped(struct page *page);
164 
165 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
166 					  u64 start, u64 owner_root, int level);
167 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
168 						  u64 start, unsigned long len);
169 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
170 						u64 start);
171 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
172 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
173 					 u64 start);
174 void free_extent_buffer(struct extent_buffer *eb);
175 void free_extent_buffer_stale(struct extent_buffer *eb);
176 #define WAIT_NONE	0
177 #define WAIT_COMPLETE	1
178 #define WAIT_PAGE_LOCK	2
179 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
180 			     int mirror_num);
181 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
182 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
183 				u64 bytenr, u64 owner_root, u64 gen, int level);
184 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
185 
186 static inline int num_extent_pages(const struct extent_buffer *eb)
187 {
188 	/*
189 	 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
190 	 * sectorsize, it's just eb->len >> PAGE_SHIFT.
191 	 *
192 	 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
193 	 * thus have to ensure we get at least one page.
194 	 */
195 	return (eb->len >> PAGE_SHIFT) ?: 1;
196 }
197 
198 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
199 {
200 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
201 }
202 
203 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
204 			 unsigned long start, unsigned long len);
205 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
206 			unsigned long start,
207 			unsigned long len);
208 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb,
209 				       void __user *dst, unsigned long start,
210 				       unsigned long len);
211 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src);
212 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb,
213 		const void *src);
214 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
215 			 unsigned long start, unsigned long len);
216 void copy_extent_buffer_full(const struct extent_buffer *dst,
217 			     const struct extent_buffer *src);
218 void copy_extent_buffer(const struct extent_buffer *dst,
219 			const struct extent_buffer *src,
220 			unsigned long dst_offset, unsigned long src_offset,
221 			unsigned long len);
222 void memcpy_extent_buffer(const struct extent_buffer *dst,
223 			  unsigned long dst_offset, unsigned long src_offset,
224 			  unsigned long len);
225 void memmove_extent_buffer(const struct extent_buffer *dst,
226 			   unsigned long dst_offset, unsigned long src_offset,
227 			   unsigned long len);
228 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
229 			   unsigned long len);
230 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
231 			   unsigned long pos);
232 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
233 			      unsigned long pos, unsigned long len);
234 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
235 				unsigned long start, unsigned long pos,
236 				unsigned long len);
237 void clear_extent_buffer_dirty(const struct extent_buffer *eb);
238 bool set_extent_buffer_dirty(struct extent_buffer *eb);
239 void set_extent_buffer_uptodate(struct extent_buffer *eb);
240 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
241 int extent_buffer_under_io(const struct extent_buffer *eb);
242 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
243 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
244 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
245 				  struct page *locked_page,
246 				  u32 bits_to_clear, unsigned long page_ops);
247 
248 int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array);
249 struct bio *btrfs_bio_alloc(unsigned int nr_iovecs);
250 struct bio *btrfs_bio_clone(struct block_device *bdev, struct bio *bio);
251 struct bio *btrfs_bio_clone_partial(struct bio *orig, u64 offset, u64 size);
252 
253 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
254 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num);
255 
256 /*
257  * When IO fails, either with EIO or csum verification fails, we
258  * try other mirrors that might have a good copy of the data.  This
259  * io_failure_record is used to record state as we go through all the
260  * mirrors.  If another mirror has good data, the sector is set up to date
261  * and things continue.  If a good mirror can't be found, the original
262  * bio end_io callback is called to indicate things have failed.
263  */
264 struct io_failure_record {
265 	struct page *page;
266 	u64 start;
267 	u64 len;
268 	u64 logical;
269 	enum btrfs_compression_type compress_type;
270 	int this_mirror;
271 	int failed_mirror;
272 };
273 
274 int btrfs_repair_one_sector(struct inode *inode,
275 			    struct bio *failed_bio, u32 bio_offset,
276 			    struct page *page, unsigned int pgoff,
277 			    u64 start, int failed_mirror,
278 			    submit_bio_hook_t *submit_bio_hook);
279 
280 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
281 bool find_lock_delalloc_range(struct inode *inode,
282 			     struct page *locked_page, u64 *start,
283 			     u64 *end);
284 #endif
285 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
286 					       u64 start);
287 
288 #ifdef CONFIG_BTRFS_DEBUG
289 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
290 #else
291 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
292 #endif
293 
294 #endif
295