xref: /linux/fs/btrfs/extent_io.h (revision 3c4fc7bf4c9e66fe71abcbf93f62f4ddb89b7f15)
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_bio;
61 struct btrfs_root;
62 struct btrfs_inode;
63 struct btrfs_io_bio;
64 struct btrfs_fs_info;
65 struct io_failure_record;
66 struct extent_io_tree;
67 
68 typedef void (submit_bio_hook_t)(struct inode *inode, struct bio *bio,
69 					 int mirror_num,
70 					 enum btrfs_compression_type compress_type);
71 
72 typedef blk_status_t (extent_submit_bio_start_t)(struct inode *inode,
73 		struct bio *bio, u64 dio_file_offset);
74 
75 #define INLINE_EXTENT_BUFFER_PAGES     (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE)
76 struct extent_buffer {
77 	u64 start;
78 	unsigned long len;
79 	unsigned long bflags;
80 	struct btrfs_fs_info *fs_info;
81 	spinlock_t refs_lock;
82 	atomic_t refs;
83 	atomic_t io_pages;
84 	int read_mirror;
85 	struct rcu_head rcu_head;
86 	pid_t lock_owner;
87 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
88 	s8 log_index;
89 
90 	struct rw_semaphore lock;
91 
92 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
93 	struct list_head release_list;
94 #ifdef CONFIG_BTRFS_DEBUG
95 	struct list_head leak_list;
96 #endif
97 };
98 
99 /*
100  * Structure to record how many bytes and which ranges are set/cleared
101  */
102 struct extent_changeset {
103 	/* How many bytes are set/cleared in this operation */
104 	u64 bytes_changed;
105 
106 	/* Changed ranges */
107 	struct ulist range_changed;
108 };
109 
110 static inline void extent_changeset_init(struct extent_changeset *changeset)
111 {
112 	changeset->bytes_changed = 0;
113 	ulist_init(&changeset->range_changed);
114 }
115 
116 static inline struct extent_changeset *extent_changeset_alloc(void)
117 {
118 	struct extent_changeset *ret;
119 
120 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
121 	if (!ret)
122 		return NULL;
123 
124 	extent_changeset_init(ret);
125 	return ret;
126 }
127 
128 static inline void extent_changeset_release(struct extent_changeset *changeset)
129 {
130 	if (!changeset)
131 		return;
132 	changeset->bytes_changed = 0;
133 	ulist_release(&changeset->range_changed);
134 }
135 
136 static inline void extent_changeset_free(struct extent_changeset *changeset)
137 {
138 	if (!changeset)
139 		return;
140 	extent_changeset_release(changeset);
141 	kfree(changeset);
142 }
143 
144 struct extent_map_tree;
145 
146 int try_release_extent_mapping(struct page *page, gfp_t mask);
147 int try_release_extent_buffer(struct page *page);
148 
149 int btrfs_read_folio(struct file *file, struct folio *folio);
150 int extent_write_locked_range(struct inode *inode, u64 start, u64 end);
151 int extent_writepages(struct address_space *mapping,
152 		      struct writeback_control *wbc);
153 int btree_write_cache_pages(struct address_space *mapping,
154 			    struct writeback_control *wbc);
155 void extent_readahead(struct readahead_control *rac);
156 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
157 		  u64 start, u64 len);
158 int set_page_extent_mapped(struct page *page);
159 void clear_page_extent_mapped(struct page *page);
160 
161 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
162 					  u64 start, u64 owner_root, int level);
163 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
164 						  u64 start, unsigned long len);
165 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
166 						u64 start);
167 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
168 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
169 					 u64 start);
170 void free_extent_buffer(struct extent_buffer *eb);
171 void free_extent_buffer_stale(struct extent_buffer *eb);
172 #define WAIT_NONE	0
173 #define WAIT_COMPLETE	1
174 #define WAIT_PAGE_LOCK	2
175 int read_extent_buffer_pages(struct extent_buffer *eb, int wait,
176 			     int mirror_num);
177 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
178 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
179 				u64 bytenr, u64 owner_root, u64 gen, int level);
180 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
181 
182 static inline int num_extent_pages(const struct extent_buffer *eb)
183 {
184 	/*
185 	 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
186 	 * sectorsize, it's just eb->len >> PAGE_SHIFT.
187 	 *
188 	 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
189 	 * thus have to ensure we get at least one page.
190 	 */
191 	return (eb->len >> PAGE_SHIFT) ?: 1;
192 }
193 
194 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
195 {
196 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
197 }
198 
199 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
200 			 unsigned long start, unsigned long len);
201 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
202 			unsigned long start,
203 			unsigned long len);
204 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb,
205 				       void __user *dst, unsigned long start,
206 				       unsigned long len);
207 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src);
208 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb,
209 		const void *src);
210 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
211 			 unsigned long start, unsigned long len);
212 void copy_extent_buffer_full(const struct extent_buffer *dst,
213 			     const struct extent_buffer *src);
214 void copy_extent_buffer(const struct extent_buffer *dst,
215 			const struct extent_buffer *src,
216 			unsigned long dst_offset, unsigned long src_offset,
217 			unsigned long len);
218 void memcpy_extent_buffer(const struct extent_buffer *dst,
219 			  unsigned long dst_offset, unsigned long src_offset,
220 			  unsigned long len);
221 void memmove_extent_buffer(const struct extent_buffer *dst,
222 			   unsigned long dst_offset, unsigned long src_offset,
223 			   unsigned long len);
224 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
225 			   unsigned long len);
226 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
227 			   unsigned long pos);
228 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
229 			      unsigned long pos, unsigned long len);
230 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
231 				unsigned long start, unsigned long pos,
232 				unsigned long len);
233 void clear_extent_buffer_dirty(const struct extent_buffer *eb);
234 bool set_extent_buffer_dirty(struct extent_buffer *eb);
235 void set_extent_buffer_uptodate(struct extent_buffer *eb);
236 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
237 int extent_buffer_under_io(const struct extent_buffer *eb);
238 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
239 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
240 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
241 				  struct page *locked_page,
242 				  u32 bits_to_clear, unsigned long page_ops);
243 
244 int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array);
245 struct bio *btrfs_bio_alloc(unsigned int nr_iovecs);
246 struct bio *btrfs_bio_clone_partial(struct bio *orig, u64 offset, u64 size);
247 
248 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
249 int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, int mirror_num);
250 
251 /*
252  * When IO fails, either with EIO or csum verification fails, we
253  * try other mirrors that might have a good copy of the data.  This
254  * io_failure_record is used to record state as we go through all the
255  * mirrors.  If another mirror has good data, the sector is set up to date
256  * and things continue.  If a good mirror can't be found, the original
257  * bio end_io callback is called to indicate things have failed.
258  */
259 struct io_failure_record {
260 	struct page *page;
261 	u64 start;
262 	u64 len;
263 	u64 logical;
264 	int this_mirror;
265 	int failed_mirror;
266 	int num_copies;
267 };
268 
269 int btrfs_repair_one_sector(struct inode *inode, struct btrfs_bio *failed_bbio,
270 			    u32 bio_offset, struct page *page, unsigned int pgoff,
271 			    submit_bio_hook_t *submit_bio_hook);
272 
273 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
274 bool find_lock_delalloc_range(struct inode *inode,
275 			     struct page *locked_page, u64 *start,
276 			     u64 *end);
277 #endif
278 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
279 					       u64 start);
280 
281 #ifdef CONFIG_BTRFS_DEBUG
282 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
283 #else
284 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
285 #endif
286 
287 #endif
288