1 /* SPDX-License-Identifier: GPL-2.0 */
3 #ifndef BTRFS_EXTENT_IO_H
4 #define BTRFS_EXTENT_IO_H
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"
14 struct btrfs_trans_handle;
17 EXTENT_BUFFER_UPTODATE,
19 EXTENT_BUFFER_CORRUPT,
20 /* this got triggered by readahead */
21 EXTENT_BUFFER_READAHEAD,
22 EXTENT_BUFFER_TREE_REF,
24 EXTENT_BUFFER_WRITEBACK,
26 EXTENT_BUFFER_READ_ERR,
27 EXTENT_BUFFER_UNMAPPED,
28 EXTENT_BUFFER_IN_TREE,
30 EXTENT_BUFFER_WRITE_ERR,
31 /* Indicate the extent buffer is written zeroed out (for zoned) */
32 EXTENT_BUFFER_ZONED_ZEROOUT,
33 /* Indicate that extent buffer pages a being read */
34 EXTENT_BUFFER_READING,
37 /* these are flags for __process_pages_contig */
39 ENUM_BIT(PAGE_UNLOCK),
40 /* Page starts writeback, clear dirty bit and set writeback bit */
41 ENUM_BIT(PAGE_START_WRITEBACK),
42 ENUM_BIT(PAGE_END_WRITEBACK),
43 ENUM_BIT(PAGE_SET_ORDERED),
47 * Folio private values. Every page that is controlled by the extent map has
48 * folio private set to this value.
50 #define EXTENT_FOLIO_PRIVATE 1
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.
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)))
69 struct extent_io_tree;
70 struct btrfs_tree_parent_check;
72 int __init extent_buffer_init_cachep(void);
73 void __cold extent_buffer_free_cachep(void);
75 #define INLINE_EXTENT_BUFFER_PAGES (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE)
76 struct extent_buffer {
80 struct btrfs_fs_info *fs_info;
83 * The address where the eb can be accessed without any cross-page handling.
84 * This can be NULL if not possible.
91 /* >= 0 if eb belongs to a log tree, -1 otherwise */
93 struct rcu_head rcu_head;
95 struct rw_semaphore lock;
98 * Pointers to all the folios of the extent buffer.
100 * For now the folio is always order 0 (aka, a single page).
102 struct folio *folios[INLINE_EXTENT_BUFFER_PAGES];
103 #ifdef CONFIG_BTRFS_DEBUG
104 struct list_head leak_list;
109 struct btrfs_eb_write_context {
110 struct writeback_control *wbc;
111 struct extent_buffer *eb;
112 /* Block group @eb resides in. Only used for zoned mode. */
113 struct btrfs_block_group *zoned_bg;
117 * Get the correct offset inside the page of extent buffer.
119 * @eb: target extent buffer
120 * @start: offset inside the extent buffer
122 * Will handle both sectorsize == PAGE_SIZE and sectorsize < PAGE_SIZE cases.
124 static inline size_t get_eb_offset_in_folio(const struct extent_buffer *eb,
125 unsigned long offset)
128 * 1) sectorsize == PAGE_SIZE and nodesize >= PAGE_SIZE case
129 * 1.1) One large folio covering the whole eb
130 * The eb->start is aligned to folio size, thus adding it
131 * won't cause any difference.
132 * 1.2) Several page sized folios
133 * The eb->start is aligned to folio (page) size, thus
134 * adding it won't cause any difference.
136 * 2) sectorsize < PAGE_SIZE and nodesize < PAGE_SIZE case
137 * In this case there would only be one page sized folio, and there
138 * may be several different extent buffers in the page/folio.
139 * We need to add eb->start to properly access the offset inside
142 return offset_in_folio(eb->folios[0], offset + eb->start);
145 static inline unsigned long get_eb_folio_index(const struct extent_buffer *eb,
146 unsigned long offset)
149 * 1) sectorsize == PAGE_SIZE and nodesize >= PAGE_SIZE case
150 * 1.1) One large folio covering the whole eb.
151 * the folio_shift would be large enough to always make us
153 * 1.2) Several page sized folios
154 * The folio_shift() would be PAGE_SHIFT, giving us the correct
157 * 2) sectorsize < PAGE_SIZE and nodesize < PAGE_SIZE case
158 * The folio would only be page sized, and always give us 0 as index.
160 return offset >> folio_shift(eb->folios[0]);
164 * Structure to record how many bytes and which ranges are set/cleared
166 struct extent_changeset {
167 /* How many bytes are set/cleared in this operation */
171 struct ulist range_changed;
174 static inline void extent_changeset_init(struct extent_changeset *changeset)
176 changeset->bytes_changed = 0;
177 ulist_init(&changeset->range_changed);
180 static inline struct extent_changeset *extent_changeset_alloc(void)
182 struct extent_changeset *ret;
184 ret = kmalloc(sizeof(*ret), GFP_KERNEL);
188 extent_changeset_init(ret);
192 static inline void extent_changeset_release(struct extent_changeset *changeset)
196 changeset->bytes_changed = 0;
197 ulist_release(&changeset->range_changed);
200 static inline void extent_changeset_free(struct extent_changeset *changeset)
204 extent_changeset_release(changeset);
208 struct extent_map_tree;
210 int try_release_extent_mapping(struct page *page, gfp_t mask);
211 int try_release_extent_buffer(struct page *page);
213 int btrfs_read_folio(struct file *file, struct folio *folio);
214 void extent_write_locked_range(struct inode *inode, struct page *locked_page,
215 u64 start, u64 end, struct writeback_control *wbc,
217 int extent_writepages(struct address_space *mapping,
218 struct writeback_control *wbc);
219 int btree_write_cache_pages(struct address_space *mapping,
220 struct writeback_control *wbc);
221 void extent_readahead(struct readahead_control *rac);
222 int extent_fiemap(struct btrfs_inode *inode, struct fiemap_extent_info *fieinfo,
224 int set_folio_extent_mapped(struct folio *folio);
225 int set_page_extent_mapped(struct page *page);
226 void clear_page_extent_mapped(struct page *page);
228 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
229 u64 start, u64 owner_root, int level);
230 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
231 u64 start, unsigned long len);
232 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
234 struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src);
235 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
237 void free_extent_buffer(struct extent_buffer *eb);
238 void free_extent_buffer_stale(struct extent_buffer *eb);
240 #define WAIT_COMPLETE 1
241 #define WAIT_PAGE_LOCK 2
242 int read_extent_buffer_pages(struct extent_buffer *eb, int wait, int mirror_num,
243 struct btrfs_tree_parent_check *parent_check);
244 void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
245 void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info,
246 u64 bytenr, u64 owner_root, u64 gen, int level);
247 void btrfs_readahead_node_child(struct extent_buffer *node, int slot);
249 static inline int num_extent_pages(const struct extent_buffer *eb)
252 * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to
253 * sectorsize, it's just eb->len >> PAGE_SHIFT.
255 * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE,
256 * thus have to ensure we get at least one page.
258 return (eb->len >> PAGE_SHIFT) ?: 1;
262 * This can only be determined at runtime by checking eb::folios[0].
264 * As we can have either one large folio covering the whole eb
265 * (either nodesize <= PAGE_SIZE, or high order folio), or multiple
266 * single-paged folios.
268 static inline int num_extent_folios(const struct extent_buffer *eb)
270 if (folio_order(eb->folios[0]))
272 return num_extent_pages(eb);
275 static inline int extent_buffer_uptodate(const struct extent_buffer *eb)
277 return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
280 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
281 unsigned long start, unsigned long len);
282 void read_extent_buffer(const struct extent_buffer *eb, void *dst,
285 int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb,
286 void __user *dst, unsigned long start,
288 void write_extent_buffer(const struct extent_buffer *eb, const void *src,
289 unsigned long start, unsigned long len);
291 static inline void write_extent_buffer_chunk_tree_uuid(
292 const struct extent_buffer *eb, const void *chunk_tree_uuid)
294 write_extent_buffer(eb, chunk_tree_uuid,
295 offsetof(struct btrfs_header, chunk_tree_uuid),
299 static inline void write_extent_buffer_fsid(const struct extent_buffer *eb,
302 write_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid),
306 void copy_extent_buffer_full(const struct extent_buffer *dst,
307 const struct extent_buffer *src);
308 void copy_extent_buffer(const struct extent_buffer *dst,
309 const struct extent_buffer *src,
310 unsigned long dst_offset, unsigned long src_offset,
312 void memcpy_extent_buffer(const struct extent_buffer *dst,
313 unsigned long dst_offset, unsigned long src_offset,
315 void memmove_extent_buffer(const struct extent_buffer *dst,
316 unsigned long dst_offset, unsigned long src_offset,
318 void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start,
320 int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start,
322 void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start,
323 unsigned long pos, unsigned long len);
324 void extent_buffer_bitmap_clear(const struct extent_buffer *eb,
325 unsigned long start, unsigned long pos,
327 void set_extent_buffer_dirty(struct extent_buffer *eb);
328 void set_extent_buffer_uptodate(struct extent_buffer *eb);
329 void clear_extent_buffer_uptodate(struct extent_buffer *eb);
330 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
331 void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end,
332 struct page *locked_page,
333 u32 bits_to_clear, unsigned long page_ops);
334 int extent_invalidate_folio(struct extent_io_tree *tree,
335 struct folio *folio, size_t offset);
336 void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans,
337 struct extent_buffer *buf);
339 int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array,
342 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
343 bool find_lock_delalloc_range(struct inode *inode,
344 struct page *locked_page, u64 *start,
347 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
350 #ifdef CONFIG_BTRFS_DEBUG
351 void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info);
353 #define btrfs_extent_buffer_leak_debug_check(fs_info) do {} while (0)