xref: /openbmc/linux/fs/btrfs/extent_io.h (revision bfb41e46)
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 	ENUM_BIT(PAGE_LOCK),
44 };
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_fs_info;
69 struct extent_io_tree;
70 struct btrfs_tree_parent_check;
71 
72 int __init extent_buffer_init_cachep(void);
73 void __cold extent_buffer_free_cachep(void);
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 	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 #ifdef CONFIG_BTRFS_DEBUG
93 	struct list_head leak_list;
94 #endif
95 };
96 
97 /*
98  * Get the correct offset inside the page of extent buffer.
99  *
100  * @eb:		target extent buffer
101  * @start:	offset inside the extent buffer
102  *
103  * Will handle both sectorsize == PAGE_SIZE and sectorsize < PAGE_SIZE cases.
104  */
105 static inline size_t get_eb_offset_in_page(const struct extent_buffer *eb,
106 					   unsigned long offset)
107 {
108 	/*
109 	 * For sectorsize == PAGE_SIZE case, eb->start will always be aligned
110 	 * to PAGE_SIZE, thus adding it won't cause any difference.
111 	 *
112 	 * For sectorsize < PAGE_SIZE, we must only read the data that belongs
113 	 * to the eb, thus we have to take the eb->start into consideration.
114 	 */
115 	return offset_in_page(offset + eb->start);
116 }
117 
118 static inline unsigned long get_eb_page_index(unsigned long offset)
119 {
120 	/*
121 	 * For sectorsize == PAGE_SIZE case, plain >> PAGE_SHIFT is enough.
122 	 *
123 	 * For sectorsize < PAGE_SIZE case, we only support 64K PAGE_SIZE,
124 	 * and have ensured that all tree blocks are contained in one page,
125 	 * thus we always get index == 0.
126 	 */
127 	return offset >> PAGE_SHIFT;
128 }
129 
130 /*
131  * Structure to record how many bytes and which ranges are set/cleared
132  */
133 struct extent_changeset {
134 	/* How many bytes are set/cleared in this operation */
135 	u64 bytes_changed;
136 
137 	/* Changed ranges */
138 	struct ulist range_changed;
139 };
140 
141 static inline void extent_changeset_init(struct extent_changeset *changeset)
142 {
143 	changeset->bytes_changed = 0;
144 	ulist_init(&changeset->range_changed);
145 }
146 
147 static inline struct extent_changeset *extent_changeset_alloc(void)
148 {
149 	struct extent_changeset *ret;
150 
151 	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
152 	if (!ret)
153 		return NULL;
154 
155 	extent_changeset_init(ret);
156 	return ret;
157 }
158 
159 static inline void extent_changeset_release(struct extent_changeset *changeset)
160 {
161 	if (!changeset)
162 		return;
163 	changeset->bytes_changed = 0;
164 	ulist_release(&changeset->range_changed);
165 }
166 
167 static inline void extent_changeset_free(struct extent_changeset *changeset)
168 {
169 	if (!changeset)
170 		return;
171 	extent_changeset_release(changeset);
172 	kfree(changeset);
173 }
174 
175 struct extent_map_tree;
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 btrfs_read_folio(struct file *file, struct folio *folio);
181 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
182 			      struct writeback_control *wbc);
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 set_extent_buffer_dirty(struct extent_buffer *eb);
266 void set_extent_buffer_uptodate(struct extent_buffer *eb);
267 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
268 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
269 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
270 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
271 				  struct page *locked_page,
272 				  u32 bits_to_clear, unsigned long page_ops);
273 int extent_invalidate_folio(struct extent_io_tree *tree,
274 			    struct folio *folio, size_t offset);
275 void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans,
276 			      struct extent_buffer *buf);
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 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
283 bool find_lock_delalloc_range(struct inode *inode,
284 			     struct page *locked_page, u64 *start,
285 			     u64 *end);
286 #endif
287 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
288 					       u64 start);
289 
290 #ifdef CONFIG_BTRFS_DEBUG
291 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
292 #else
293 #define btrfs_extent_buffer_leak_debug_check(fs_info)	do {} while (0)
294 #endif
295 
296 #endif
297