xref: /linux/fs/btrfs/extent_io.h (revision 8e07e0e3964ca4e23ce7b68e2096fe660a888942)
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 #include "misc.h"
13 
14 struct btrfs_trans_handle;
15 
16 enum {
17 	EXTENT_BUFFER_UPTODATE,
18 	EXTENT_BUFFER_DIRTY,
19 	EXTENT_BUFFER_CORRUPT,
20 	/* this got triggered by readahead */
21 	EXTENT_BUFFER_READAHEAD,
22 	EXTENT_BUFFER_TREE_REF,
23 	EXTENT_BUFFER_STALE,
24 	EXTENT_BUFFER_WRITEBACK,
25 	/* read IO error */
26 	EXTENT_BUFFER_READ_ERR,
27 	EXTENT_BUFFER_UNMAPPED,
28 	EXTENT_BUFFER_IN_TREE,
29 	/* write IO error */
30 	EXTENT_BUFFER_WRITE_ERR,
31 	EXTENT_BUFFER_NO_CHECK,
32 	/* Indicate that extent buffer pages a being read */
33 	EXTENT_BUFFER_READING,
34 };
35 
36 /* these are flags for __process_pages_contig */
37 enum {
38 	ENUM_BIT(PAGE_UNLOCK),
39 	/* Page starts writeback, clear dirty bit and set writeback bit */
40 	ENUM_BIT(PAGE_START_WRITEBACK),
41 	ENUM_BIT(PAGE_END_WRITEBACK),
42 	ENUM_BIT(PAGE_SET_ORDERED),
43 };
44 
45 /*
46  * page->private values.  Every page that is controlled by the extent
47  * map has page->private set to one.
48  */
49 #define EXTENT_PAGE_PRIVATE 1
50 
51 /*
52  * The extent buffer bitmap operations are done with byte granularity instead of
53  * word granularity for two reasons:
54  * 1. The bitmaps must be little-endian on disk.
55  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
56  *    single word in a bitmap may straddle two pages in the extent buffer.
57  */
58 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
59 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
60 #define BITMAP_FIRST_BYTE_MASK(start) \
61 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
62 #define BITMAP_LAST_BYTE_MASK(nbits) \
63 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
64 
65 struct btrfs_root;
66 struct btrfs_inode;
67 struct btrfs_fs_info;
68 struct extent_io_tree;
69 struct btrfs_tree_parent_check;
70 
71 int __init extent_buffer_init_cachep(void);
72 void __cold extent_buffer_free_cachep(void);
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 	int read_mirror;
83 	/* >= 0 if eb belongs to a log tree, -1 otherwise */
84 	s8 log_index;
85 	struct rcu_head rcu_head;
86 
87 	struct rw_semaphore lock;
88 
89 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
90 #ifdef CONFIG_BTRFS_DEBUG
91 	struct list_head leak_list;
92 	pid_t lock_owner;
93 #endif
94 };
95 
96 struct btrfs_eb_write_context {
97 	struct writeback_control *wbc;
98 	struct extent_buffer *eb;
99 	/* Block group @eb resides in. Only used for zoned mode. */
100 	struct btrfs_block_group *zoned_bg;
101 };
102 
103 /*
104  * Get the correct offset inside the page of extent buffer.
105  *
106  * @eb:		target extent buffer
107  * @start:	offset inside the extent buffer
108  *
109  * Will handle both sectorsize == PAGE_SIZE and sectorsize < PAGE_SIZE cases.
110  */
111 static inline size_t get_eb_offset_in_page(const struct extent_buffer *eb,
112 					   unsigned long offset)
113 {
114 	/*
115 	 * For sectorsize == PAGE_SIZE case, eb->start will always be aligned
116 	 * to PAGE_SIZE, thus adding it won't cause any difference.
117 	 *
118 	 * For sectorsize < PAGE_SIZE, we must only read the data that belongs
119 	 * to the eb, thus we have to take the eb->start into consideration.
120 	 */
121 	return offset_in_page(offset + eb->start);
122 }
123 
124 static inline unsigned long get_eb_page_index(unsigned long offset)
125 {
126 	/*
127 	 * For sectorsize == PAGE_SIZE case, plain >> PAGE_SHIFT is enough.
128 	 *
129 	 * For sectorsize < PAGE_SIZE case, we only support 64K PAGE_SIZE,
130 	 * and have ensured that all tree blocks are contained in one page,
131 	 * thus we always get index == 0.
132 	 */
133 	return offset >> PAGE_SHIFT;
134 }
135 
136 /*
137  * Structure to record how many bytes and which ranges are set/cleared
138  */
139 struct extent_changeset {
140 	/* How many bytes are set/cleared in this operation */
141 	u64 bytes_changed;
142 
143 	/* Changed ranges */
144 	struct ulist range_changed;
145 };
146 
147 static inline void extent_changeset_init(struct extent_changeset *changeset)
148 {
149 	changeset->bytes_changed = 0;
150 	ulist_init(&changeset->range_changed);
151 }
152 
153 static inline struct extent_changeset *extent_changeset_alloc(void)
154 {
155 	struct extent_changeset *ret;
156 
157 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
158 	if (!ret)
159 		return NULL;
160 
161 	extent_changeset_init(ret);
162 	return ret;
163 }
164 
165 static inline void extent_changeset_release(struct extent_changeset *changeset)
166 {
167 	if (!changeset)
168 		return;
169 	changeset->bytes_changed = 0;
170 	ulist_release(&changeset->range_changed);
171 }
172 
173 static inline void extent_changeset_free(struct extent_changeset *changeset)
174 {
175 	if (!changeset)
176 		return;
177 	extent_changeset_release(changeset);
178 	kfree(changeset);
179 }
180 
181 struct extent_map_tree;
182 
183 int try_release_extent_mapping(struct page *page, gfp_t mask);
184 int try_release_extent_buffer(struct page *page);
185 
186 int btrfs_read_folio(struct file *file, struct folio *folio);
187 void extent_write_locked_range(struct inode *inode, struct page *locked_page,
188 			       u64 start, u64 end, struct writeback_control *wbc,
189 			       bool pages_dirty);
190 int extent_writepages(struct address_space *mapping,
191 		      struct writeback_control *wbc);
192 int btree_write_cache_pages(struct address_space *mapping,
193 			    struct writeback_control *wbc);
194 void extent_readahead(struct readahead_control *rac);
195 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
196 		  u64 start, u64 len);
197 int set_page_extent_mapped(struct page *page);
198 void clear_page_extent_mapped(struct page *page);
199 
200 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
201 					  u64 start, u64 owner_root, int level);
202 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
203 						  u64 start, unsigned long len);
204 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
205 						u64 start);
206 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
207 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
208 					 u64 start);
209 void free_extent_buffer(struct extent_buffer *eb);
210 void free_extent_buffer_stale(struct extent_buffer *eb);
211 #define WAIT_NONE	0
212 #define WAIT_COMPLETE	1
213 #define WAIT_PAGE_LOCK	2
214 int read_extent_buffer_pages(struct extent_buffer *eb, int wait, int mirror_num,
215 			     struct btrfs_tree_parent_check *parent_check);
216 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
217 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
218 				u64 bytenr, u64 owner_root, u64 gen, int level);
219 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
220 
221 static inline int num_extent_pages(const struct extent_buffer *eb)
222 {
223 	/*
224 	 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
225 	 * sectorsize, it's just eb->len >> PAGE_SHIFT.
226 	 *
227 	 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
228 	 * thus have to ensure we get at least one page.
229 	 */
230 	return (eb->len >> PAGE_SHIFT) ?: 1;
231 }
232 
233 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
234 {
235 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
236 }
237 
238 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
239 			 unsigned long start, unsigned long len);
240 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
241 			unsigned long start,
242 			unsigned long len);
243 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb,
244 				       void __user *dst, unsigned long start,
245 				       unsigned long len);
246 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
247 			 unsigned long start, unsigned long len);
248 
249 static inline void write_extent_buffer_chunk_tree_uuid(
250 		const struct extent_buffer *eb, const void *chunk_tree_uuid)
251 {
252 	write_extent_buffer(eb, chunk_tree_uuid,
253 			    offsetof(struct btrfs_header, chunk_tree_uuid),
254 			    BTRFS_FSID_SIZE);
255 }
256 
257 static inline void write_extent_buffer_fsid(const struct extent_buffer *eb,
258 					    const void *fsid)
259 {
260 	write_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid),
261 			    BTRFS_FSID_SIZE);
262 }
263 
264 void copy_extent_buffer_full(const struct extent_buffer *dst,
265 			     const struct extent_buffer *src);
266 void copy_extent_buffer(const struct extent_buffer *dst,
267 			const struct extent_buffer *src,
268 			unsigned long dst_offset, unsigned long src_offset,
269 			unsigned long len);
270 void memcpy_extent_buffer(const struct extent_buffer *dst,
271 			  unsigned long dst_offset, unsigned long src_offset,
272 			  unsigned long len);
273 void memmove_extent_buffer(const struct extent_buffer *dst,
274 			   unsigned long dst_offset, unsigned long src_offset,
275 			   unsigned long len);
276 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
277 			   unsigned long len);
278 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
279 			   unsigned long pos);
280 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
281 			      unsigned long pos, unsigned long len);
282 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
283 				unsigned long start, unsigned long pos,
284 				unsigned long len);
285 void set_extent_buffer_dirty(struct extent_buffer *eb);
286 void set_extent_buffer_uptodate(struct extent_buffer *eb);
287 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
288 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
289 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
290 				  struct page *locked_page,
291 				  u32 bits_to_clear, unsigned long page_ops);
292 int extent_invalidate_folio(struct extent_io_tree *tree,
293 			    struct folio *folio, size_t offset);
294 void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans,
295 			      struct extent_buffer *buf);
296 
297 int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array);
298 
299 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
300 bool find_lock_delalloc_range(struct inode *inode,
301 			     struct page *locked_page, u64 *start,
302 			     u64 *end);
303 #endif
304 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
305 					       u64 start);
306 
307 #ifdef CONFIG_BTRFS_DEBUG
308 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
309 #else
310 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
311 #endif
312 
313 #endif
314