xref: /openbmc/linux/fs/btrfs/extent_io.h (revision e65e175b)
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 enum {
15 	EXTENT_BUFFER_UPTODATE,
16 	EXTENT_BUFFER_DIRTY,
17 	EXTENT_BUFFER_CORRUPT,
18 	/* this got triggered by readahead */
19 	EXTENT_BUFFER_READAHEAD,
20 	EXTENT_BUFFER_TREE_REF,
21 	EXTENT_BUFFER_STALE,
22 	EXTENT_BUFFER_WRITEBACK,
23 	/* read IO error */
24 	EXTENT_BUFFER_READ_ERR,
25 	EXTENT_BUFFER_UNMAPPED,
26 	EXTENT_BUFFER_IN_TREE,
27 	/* write IO error */
28 	EXTENT_BUFFER_WRITE_ERR,
29 	EXTENT_BUFFER_NO_CHECK,
30 };
31 
32 /* these are flags for __process_pages_contig */
33 enum {
34 	ENUM_BIT(PAGE_UNLOCK),
35 	/* Page starts writeback, clear dirty bit and set writeback bit */
36 	ENUM_BIT(PAGE_START_WRITEBACK),
37 	ENUM_BIT(PAGE_END_WRITEBACK),
38 	ENUM_BIT(PAGE_SET_ORDERED),
39 	ENUM_BIT(PAGE_SET_ERROR),
40 	ENUM_BIT(PAGE_LOCK),
41 };
42 
43 /*
44  * page->private values.  Every page that is controlled by the extent
45  * map has page->private set to one.
46  */
47 #define EXTENT_PAGE_PRIVATE 1
48 
49 /*
50  * The extent buffer bitmap operations are done with byte granularity instead of
51  * word granularity for two reasons:
52  * 1. The bitmaps must be little-endian on disk.
53  * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a
54  *    single word in a bitmap may straddle two pages in the extent buffer.
55  */
56 #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
57 #define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
58 #define BITMAP_FIRST_BYTE_MASK(start) \
59 	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
60 #define BITMAP_LAST_BYTE_MASK(nbits) \
61 	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
62 
63 struct btrfs_bio;
64 struct btrfs_root;
65 struct btrfs_inode;
66 struct btrfs_fs_info;
67 struct io_failure_record;
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 	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  * Get the correct offset inside the page of extent buffer.
100  *
101  * @eb:		target extent buffer
102  * @start:	offset inside the extent buffer
103  *
104  * Will handle both sectorsize == PAGE_SIZE and sectorsize < PAGE_SIZE cases.
105  */
106 static inline size_t get_eb_offset_in_page(const struct extent_buffer *eb,
107 					   unsigned long offset)
108 {
109 	/*
110 	 * For sectorsize == PAGE_SIZE case, eb->start will always be aligned
111 	 * to PAGE_SIZE, thus adding it won't cause any difference.
112 	 *
113 	 * For sectorsize < PAGE_SIZE, we must only read the data that belongs
114 	 * to the eb, thus we have to take the eb->start into consideration.
115 	 */
116 	return offset_in_page(offset + eb->start);
117 }
118 
119 static inline unsigned long get_eb_page_index(unsigned long offset)
120 {
121 	/*
122 	 * For sectorsize == PAGE_SIZE case, plain >> PAGE_SHIFT is enough.
123 	 *
124 	 * For sectorsize < PAGE_SIZE case, we only support 64K PAGE_SIZE,
125 	 * and have ensured that all tree blocks are contained in one page,
126 	 * thus we always get index == 0.
127 	 */
128 	return offset >> PAGE_SHIFT;
129 }
130 
131 /*
132  * Structure to record how many bytes and which ranges are set/cleared
133  */
134 struct extent_changeset {
135 	/* How many bytes are set/cleared in this operation */
136 	u64 bytes_changed;
137 
138 	/* Changed ranges */
139 	struct ulist range_changed;
140 };
141 
142 static inline void extent_changeset_init(struct extent_changeset *changeset)
143 {
144 	changeset->bytes_changed = 0;
145 	ulist_init(&changeset->range_changed);
146 }
147 
148 static inline struct extent_changeset *extent_changeset_alloc(void)
149 {
150 	struct extent_changeset *ret;
151 
152 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
153 	if (!ret)
154 		return NULL;
155 
156 	extent_changeset_init(ret);
157 	return ret;
158 }
159 
160 static inline void extent_changeset_release(struct extent_changeset *changeset)
161 {
162 	if (!changeset)
163 		return;
164 	changeset->bytes_changed = 0;
165 	ulist_release(&changeset->range_changed);
166 }
167 
168 static inline void extent_changeset_free(struct extent_changeset *changeset)
169 {
170 	if (!changeset)
171 		return;
172 	extent_changeset_release(changeset);
173 	kfree(changeset);
174 }
175 
176 struct extent_map_tree;
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 btrfs_read_folio(struct file *file, struct folio *folio);
182 int extent_write_locked_range(struct inode *inode, u64 start, u64 end);
183 int extent_writepages(struct address_space *mapping,
184 		      struct writeback_control *wbc);
185 int btree_write_cache_pages(struct address_space *mapping,
186 			    struct writeback_control *wbc);
187 void extent_readahead(struct readahead_control *rac);
188 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
189 		  u64 start, u64 len);
190 int set_page_extent_mapped(struct page *page);
191 void clear_page_extent_mapped(struct page *page);
192 
193 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
194 					  u64 start, u64 owner_root, int level);
195 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
196 						  u64 start, unsigned long len);
197 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
198 						u64 start);
199 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
200 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
201 					 u64 start);
202 void free_extent_buffer(struct extent_buffer *eb);
203 void free_extent_buffer_stale(struct extent_buffer *eb);
204 #define WAIT_NONE	0
205 #define WAIT_COMPLETE	1
206 #define WAIT_PAGE_LOCK	2
207 int read_extent_buffer_pages(struct extent_buffer *eb, int wait, int mirror_num,
208 			     struct btrfs_tree_parent_check *parent_check);
209 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
210 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
211 				u64 bytenr, u64 owner_root, u64 gen, int level);
212 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
213 
214 static inline int num_extent_pages(const struct extent_buffer *eb)
215 {
216 	/*
217 	 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
218 	 * sectorsize, it's just eb->len >> PAGE_SHIFT.
219 	 *
220 	 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
221 	 * thus have to ensure we get at least one page.
222 	 */
223 	return (eb->len >> PAGE_SHIFT) ?: 1;
224 }
225 
226 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
227 {
228 	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
229 }
230 
231 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
232 			 unsigned long start, unsigned long len);
233 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
234 			unsigned long start,
235 			unsigned long len);
236 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb,
237 				       void __user *dst, unsigned long start,
238 				       unsigned long len);
239 void write_extent_buffer_fsid(const struct extent_buffer *eb, const void *src);
240 void write_extent_buffer_chunk_tree_uuid(const struct extent_buffer *eb,
241 		const void *src);
242 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
243 			 unsigned long start, unsigned long len);
244 void copy_extent_buffer_full(const struct extent_buffer *dst,
245 			     const struct extent_buffer *src);
246 void copy_extent_buffer(const struct extent_buffer *dst,
247 			const struct extent_buffer *src,
248 			unsigned long dst_offset, unsigned long src_offset,
249 			unsigned long len);
250 void memcpy_extent_buffer(const struct extent_buffer *dst,
251 			  unsigned long dst_offset, unsigned long src_offset,
252 			  unsigned long len);
253 void memmove_extent_buffer(const struct extent_buffer *dst,
254 			   unsigned long dst_offset, unsigned long src_offset,
255 			   unsigned long len);
256 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
257 			   unsigned long len);
258 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
259 			   unsigned long pos);
260 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
261 			      unsigned long pos, unsigned long len);
262 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
263 				unsigned long start, unsigned long pos,
264 				unsigned long len);
265 void clear_extent_buffer_dirty(const struct extent_buffer *eb);
266 bool set_extent_buffer_dirty(struct extent_buffer *eb);
267 void set_extent_buffer_uptodate(struct extent_buffer *eb);
268 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
269 int extent_buffer_under_io(const struct extent_buffer *eb);
270 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
271 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
272 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
273 				  struct page *locked_page,
274 				  u32 bits_to_clear, unsigned long page_ops);
275 int extent_invalidate_folio(struct extent_io_tree *tree,
276 			    struct folio *folio, size_t offset);
277 
278 int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array);
279 
280 void end_extent_writepage(struct page *page, int err, u64 start, u64 end);
281 
282 /*
283  * When IO fails, either with EIO or csum verification fails, we
284  * try other mirrors that might have a good copy of the data.  This
285  * io_failure_record is used to record state as we go through all the
286  * mirrors.  If another mirror has good data, the sector is set up to date
287  * and things continue.  If a good mirror can't be found, the original
288  * bio end_io callback is called to indicate things have failed.
289  */
290 struct io_failure_record {
291 	/* Use rb_simple_node for search/insert */
292 	struct {
293 		struct rb_node rb_node;
294 		u64 bytenr;
295 	};
296 	struct page *page;
297 	u64 len;
298 	u64 logical;
299 	int this_mirror;
300 	int failed_mirror;
301 	int num_copies;
302 };
303 
304 int btrfs_repair_one_sector(struct btrfs_inode *inode, struct btrfs_bio *failed_bbio,
305 			    u32 bio_offset, struct page *page, unsigned int pgoff,
306 			    bool submit_buffered);
307 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start, u64 end);
308 int btrfs_clean_io_failure(struct btrfs_inode *inode, u64 start,
309 			   struct page *page, unsigned int pg_offset);
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