Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
9888c340 DS |
2 | |
3 | #ifndef BTRFS_EXTENT_IO_H | |
4 | #define BTRFS_EXTENT_IO_H | |
d1310b2e CM |
5 | |
6 | #include <linux/rbtree.h> | |
b7ac31b7 | 7 | #include <linux/refcount.h> |
10c5db28 | 8 | #include <linux/fiemap.h> |
deb67895 | 9 | #include <linux/btrfs_tree.h> |
602035d7 DS |
10 | #include <linux/spinlock.h> |
11 | #include <linux/atomic.h> | |
12 | #include <linux/rwsem.h> | |
13 | #include <linux/list.h> | |
14 | #include <linux/slab.h> | |
0ff40013 | 15 | #include "compression.h" |
84cda1a6 | 16 | #include "messages.h" |
ac467772 | 17 | #include "ulist.h" |
fd8d2951 | 18 | #include "misc.h" |
d1310b2e | 19 | |
602035d7 DS |
20 | struct page; |
21 | struct file; | |
22 | struct folio; | |
23 | struct inode; | |
24 | struct fiemap_extent_info; | |
25 | struct readahead_control; | |
26 | struct address_space; | |
27 | struct writeback_control; | |
28 | struct extent_io_tree; | |
29 | struct extent_map_tree; | |
6b0a63a4 | 30 | struct extent_state; |
602035d7 DS |
31 | struct btrfs_block_group; |
32 | struct btrfs_fs_info; | |
33 | struct btrfs_inode; | |
34 | struct btrfs_root; | |
98c8d683 | 35 | struct btrfs_trans_handle; |
602035d7 | 36 | struct btrfs_tree_parent_check; |
98c8d683 | 37 | |
80cb3836 DS |
38 | enum { |
39 | EXTENT_BUFFER_UPTODATE, | |
40 | EXTENT_BUFFER_DIRTY, | |
80cb3836 DS |
41 | EXTENT_BUFFER_TREE_REF, |
42 | EXTENT_BUFFER_STALE, | |
43 | EXTENT_BUFFER_WRITEBACK, | |
80cb3836 | 44 | EXTENT_BUFFER_UNMAPPED, |
80cb3836 DS |
45 | /* write IO error */ |
46 | EXTENT_BUFFER_WRITE_ERR, | |
cbf44cd9 JT |
47 | /* Indicate the extent buffer is written zeroed out (for zoned) */ |
48 | EXTENT_BUFFER_ZONED_ZEROOUT, | |
d7172f52 CH |
49 | /* Indicate that extent buffer pages a being read */ |
50 | EXTENT_BUFFER_READING, | |
80cb3836 | 51 | }; |
b4ce94de | 52 | |
da2c7009 | 53 | /* these are flags for __process_pages_contig */ |
fd8d2951 DS |
54 | enum { |
55 | ENUM_BIT(PAGE_UNLOCK), | |
56 | /* Page starts writeback, clear dirty bit and set writeback bit */ | |
57 | ENUM_BIT(PAGE_START_WRITEBACK), | |
58 | ENUM_BIT(PAGE_END_WRITEBACK), | |
59 | ENUM_BIT(PAGE_SET_ORDERED), | |
fd8d2951 | 60 | }; |
a791e35e | 61 | |
d1310b2e | 62 | /* |
cfbf07e2 QW |
63 | * Folio private values. Every page that is controlled by the extent map has |
64 | * folio private set to this value. | |
d1310b2e | 65 | */ |
cfbf07e2 | 66 | #define EXTENT_FOLIO_PRIVATE 1 |
d1310b2e | 67 | |
2fe1d551 OS |
68 | /* |
69 | * The extent buffer bitmap operations are done with byte granularity instead of | |
70 | * word granularity for two reasons: | |
71 | * 1. The bitmaps must be little-endian on disk. | |
72 | * 2. Bitmap items are not guaranteed to be aligned to a word and therefore a | |
73 | * single word in a bitmap may straddle two pages in the extent buffer. | |
74 | */ | |
75 | #define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE) | |
05a6ec86 | 76 | #define BYTE_MASK ((1U << BITS_PER_BYTE) - 1) |
2fe1d551 OS |
77 | #define BITMAP_FIRST_BYTE_MASK(start) \ |
78 | ((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK) | |
79 | #define BITMAP_LAST_BYTE_MASK(nbits) \ | |
80 | (BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1))) | |
81 | ||
a758781d | 82 | |
a62a3bd9 JB |
83 | int __init extent_buffer_init_cachep(void); |
84 | void __cold extent_buffer_free_cachep(void); | |
85 | ||
deb67895 | 86 | #define INLINE_EXTENT_BUFFER_PAGES (BTRFS_MAX_METADATA_BLOCKSIZE / PAGE_SIZE) |
d1310b2e CM |
87 | struct extent_buffer { |
88 | u64 start; | |
84cda1a6 QW |
89 | u32 len; |
90 | u32 folio_size; | |
b4ce94de | 91 | unsigned long bflags; |
f28491e0 | 92 | struct btrfs_fs_info *fs_info; |
397239ed QW |
93 | |
94 | /* | |
95 | * The address where the eb can be accessed without any cross-page handling. | |
96 | * This can be NULL if not possible. | |
97 | */ | |
98 | void *addr; | |
99 | ||
3083ee2e | 100 | spinlock_t refs_lock; |
727011e0 | 101 | atomic_t refs; |
5cf1ab56 | 102 | int read_mirror; |
656f30db | 103 | /* >= 0 if eb belongs to a log tree, -1 otherwise */ |
dc516164 | 104 | s8 log_index; |
84cda1a6 | 105 | u8 folio_shift; |
150cce2d | 106 | struct rcu_head rcu_head; |
dc516164 DS |
107 | |
108 | struct rw_semaphore lock; | |
bd681513 | 109 | |
082d5bb9 QW |
110 | /* |
111 | * Pointers to all the folios of the extent buffer. | |
112 | * | |
113 | * For now the folio is always order 0 (aka, a single page). | |
114 | */ | |
115 | struct folio *folios[INLINE_EXTENT_BUFFER_PAGES]; | |
6d49ba1b ES |
116 | #ifdef CONFIG_BTRFS_DEBUG |
117 | struct list_head leak_list; | |
150cce2d | 118 | pid_t lock_owner; |
6d49ba1b | 119 | #endif |
d1310b2e CM |
120 | }; |
121 | ||
861093ef NA |
122 | struct btrfs_eb_write_context { |
123 | struct writeback_control *wbc; | |
124 | struct extent_buffer *eb; | |
7db94301 NA |
125 | /* Block group @eb resides in. Only used for zoned mode. */ |
126 | struct btrfs_block_group *zoned_bg; | |
861093ef NA |
127 | }; |
128 | ||
84cda1a6 QW |
129 | static inline unsigned long offset_in_eb_folio(const struct extent_buffer *eb, |
130 | u64 start) | |
131 | { | |
132 | ASSERT(eb->folio_size); | |
133 | return start & (eb->folio_size - 1); | |
134 | } | |
135 | ||
9b48adda JB |
136 | /* |
137 | * Get the correct offset inside the page of extent buffer. | |
138 | * | |
139 | * @eb: target extent buffer | |
140 | * @start: offset inside the extent buffer | |
141 | * | |
142 | * Will handle both sectorsize == PAGE_SIZE and sectorsize < PAGE_SIZE cases. | |
143 | */ | |
8d993618 QW |
144 | static inline size_t get_eb_offset_in_folio(const struct extent_buffer *eb, |
145 | unsigned long offset) | |
9b48adda JB |
146 | { |
147 | /* | |
8d993618 QW |
148 | * 1) sectorsize == PAGE_SIZE and nodesize >= PAGE_SIZE case |
149 | * 1.1) One large folio covering the whole eb | |
150 | * The eb->start is aligned to folio size, thus adding it | |
151 | * won't cause any difference. | |
152 | * 1.2) Several page sized folios | |
153 | * The eb->start is aligned to folio (page) size, thus | |
154 | * adding it won't cause any difference. | |
9b48adda | 155 | * |
8d993618 QW |
156 | * 2) sectorsize < PAGE_SIZE and nodesize < PAGE_SIZE case |
157 | * In this case there would only be one page sized folio, and there | |
158 | * may be several different extent buffers in the page/folio. | |
159 | * We need to add eb->start to properly access the offset inside | |
160 | * that eb. | |
9b48adda | 161 | */ |
8d993618 | 162 | return offset_in_folio(eb->folios[0], offset + eb->start); |
9b48adda JB |
163 | } |
164 | ||
8d993618 QW |
165 | static inline unsigned long get_eb_folio_index(const struct extent_buffer *eb, |
166 | unsigned long offset) | |
9b48adda JB |
167 | { |
168 | /* | |
8d993618 QW |
169 | * 1) sectorsize == PAGE_SIZE and nodesize >= PAGE_SIZE case |
170 | * 1.1) One large folio covering the whole eb. | |
171 | * the folio_shift would be large enough to always make us | |
172 | * return 0 as index. | |
173 | * 1.2) Several page sized folios | |
84cda1a6 | 174 | * The folio_shift would be PAGE_SHIFT, giving us the correct |
8d993618 | 175 | * index. |
9b48adda | 176 | * |
8d993618 QW |
177 | * 2) sectorsize < PAGE_SIZE and nodesize < PAGE_SIZE case |
178 | * The folio would only be page sized, and always give us 0 as index. | |
9b48adda | 179 | */ |
84cda1a6 | 180 | return offset >> eb->folio_shift; |
9b48adda JB |
181 | } |
182 | ||
ac467772 QW |
183 | /* |
184 | * Structure to record how many bytes and which ranges are set/cleared | |
185 | */ | |
186 | struct extent_changeset { | |
187 | /* How many bytes are set/cleared in this operation */ | |
b642b52d | 188 | u64 bytes_changed; |
ac467772 QW |
189 | |
190 | /* Changed ranges */ | |
53d32359 | 191 | struct ulist range_changed; |
ac467772 QW |
192 | }; |
193 | ||
364ecf36 QW |
194 | static inline void extent_changeset_init(struct extent_changeset *changeset) |
195 | { | |
196 | changeset->bytes_changed = 0; | |
197 | ulist_init(&changeset->range_changed); | |
198 | } | |
199 | ||
200 | static inline struct extent_changeset *extent_changeset_alloc(void) | |
201 | { | |
202 | struct extent_changeset *ret; | |
203 | ||
204 | ret = kmalloc(sizeof(*ret), GFP_KERNEL); | |
205 | if (!ret) | |
206 | return NULL; | |
207 | ||
208 | extent_changeset_init(ret); | |
209 | return ret; | |
210 | } | |
211 | ||
33336c18 BB |
212 | static inline void extent_changeset_prealloc(struct extent_changeset *changeset, gfp_t gfp_mask) |
213 | { | |
214 | ulist_prealloc(&changeset->range_changed, gfp_mask); | |
215 | } | |
216 | ||
364ecf36 QW |
217 | static inline void extent_changeset_release(struct extent_changeset *changeset) |
218 | { | |
219 | if (!changeset) | |
220 | return; | |
221 | changeset->bytes_changed = 0; | |
222 | ulist_release(&changeset->range_changed); | |
223 | } | |
224 | ||
225 | static inline void extent_changeset_free(struct extent_changeset *changeset) | |
226 | { | |
227 | if (!changeset) | |
228 | return; | |
229 | extent_changeset_release(changeset); | |
230 | kfree(changeset); | |
231 | } | |
232 | ||
046c0d65 | 233 | bool try_release_extent_mapping(struct folio *folio, gfp_t mask); |
b8ae2bfa | 234 | int try_release_extent_buffer(struct folio *folio); |
cd716d8f | 235 | |
fdaf9a58 | 236 | int btrfs_read_folio(struct file *file, struct folio *folio); |
01e11841 | 237 | void extent_write_locked_range(struct inode *inode, const struct folio *locked_folio, |
778b8785 CH |
238 | u64 start, u64 end, struct writeback_control *wbc, |
239 | bool pages_dirty); | |
c66f2afc | 240 | int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc); |
0b32f4bb JB |
241 | int btree_write_cache_pages(struct address_space *mapping, |
242 | struct writeback_control *wbc); | |
5e121ae6 | 243 | void btrfs_btree_wait_writeback_range(struct btrfs_fs_info *fs_info, u64 start, u64 end); |
7938d38b | 244 | void btrfs_readahead(struct readahead_control *rac); |
dfba9f47 | 245 | int set_folio_extent_mapped(struct folio *folio); |
266a9361 | 246 | void clear_folio_extent_mapped(struct folio *folio); |
d1310b2e | 247 | |
f28491e0 | 248 | struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info, |
3fbaf258 | 249 | u64 start, u64 owner_root, int level); |
3f556f78 | 250 | struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info, |
da17066c | 251 | u64 start); |
2b48966a | 252 | struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src); |
f28491e0 | 253 | struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info, |
452c75c3 | 254 | u64 start); |
d1310b2e | 255 | void free_extent_buffer(struct extent_buffer *eb); |
3083ee2e | 256 | void free_extent_buffer_stale(struct extent_buffer *eb); |
248c4ff3 | 257 | int read_extent_buffer_pages(struct extent_buffer *eb, int mirror_num, |
2917f741 | 258 | const struct btrfs_tree_parent_check *parent_check); |
248c4ff3 DS |
259 | int read_extent_buffer_pages_nowait(struct extent_buffer *eb, int mirror_num, |
260 | const struct btrfs_tree_parent_check *parent_check); | |
261 | ||
075adeeb DS |
262 | static inline void wait_on_extent_buffer_writeback(struct extent_buffer *eb) |
263 | { | |
264 | wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK, | |
265 | TASK_UNINTERRUPTIBLE); | |
266 | } | |
267 | ||
bfb484d9 | 268 | void btrfs_readahead_tree_block(struct btrfs_fs_info *fs_info, |
3fbaf258 | 269 | u64 bytenr, u64 owner_root, u64 gen, int level); |
bfb484d9 | 270 | void btrfs_readahead_node_child(struct extent_buffer *node, int slot); |
479ed9ab | 271 | |
6e7c2838 DS |
272 | /* Note: this can be used in for loops without caching the value in a variable. */ |
273 | static inline int __pure num_extent_pages(const struct extent_buffer *eb) | |
479ed9ab | 274 | { |
4a3dc938 QW |
275 | /* |
276 | * For sectorsize == PAGE_SIZE case, since nodesize is always aligned to | |
277 | * sectorsize, it's just eb->len >> PAGE_SHIFT. | |
278 | * | |
279 | * For sectorsize < PAGE_SIZE case, we could have nodesize < PAGE_SIZE, | |
280 | * thus have to ensure we get at least one page. | |
281 | */ | |
282 | return (eb->len >> PAGE_SHIFT) ?: 1; | |
479ed9ab RD |
283 | } |
284 | ||
13df3775 QW |
285 | /* |
286 | * This can only be determined at runtime by checking eb::folios[0]. | |
287 | * | |
288 | * As we can have either one large folio covering the whole eb | |
289 | * (either nodesize <= PAGE_SIZE, or high order folio), or multiple | |
290 | * single-paged folios. | |
6e7c2838 DS |
291 | * |
292 | * Note: this can be used in for loops without caching the value in a variable. | |
13df3775 | 293 | */ |
6e7c2838 | 294 | static inline int __pure num_extent_folios(const struct extent_buffer *eb) |
13df3775 | 295 | { |
d6fe0c69 BB |
296 | if (!eb->folios[0]) |
297 | return 0; | |
13df3775 QW |
298 | if (folio_order(eb->folios[0])) |
299 | return 1; | |
300 | return num_extent_pages(eb); | |
301 | } | |
302 | ||
2b48966a | 303 | static inline int extent_buffer_uptodate(const struct extent_buffer *eb) |
ba020491 AJ |
304 | { |
305 | return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); | |
306 | } | |
307 | ||
1cbb1f45 JM |
308 | int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv, |
309 | unsigned long start, unsigned long len); | |
310 | void read_extent_buffer(const struct extent_buffer *eb, void *dst, | |
d1310b2e CM |
311 | unsigned long start, |
312 | unsigned long len); | |
a48b73ec JB |
313 | int read_extent_buffer_to_user_nofault(const struct extent_buffer *eb, |
314 | void __user *dst, unsigned long start, | |
315 | unsigned long len); | |
2b48966a | 316 | void write_extent_buffer(const struct extent_buffer *eb, const void *src, |
d1310b2e | 317 | unsigned long start, unsigned long len); |
730c374e QW |
318 | |
319 | static inline void write_extent_buffer_chunk_tree_uuid( | |
320 | const struct extent_buffer *eb, const void *chunk_tree_uuid) | |
321 | { | |
322 | write_extent_buffer(eb, chunk_tree_uuid, | |
323 | offsetof(struct btrfs_header, chunk_tree_uuid), | |
324 | BTRFS_FSID_SIZE); | |
325 | } | |
326 | ||
327 | static inline void write_extent_buffer_fsid(const struct extent_buffer *eb, | |
328 | const void *fsid) | |
329 | { | |
330 | write_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), | |
331 | BTRFS_FSID_SIZE); | |
332 | } | |
333 | ||
2b48966a DS |
334 | void copy_extent_buffer_full(const struct extent_buffer *dst, |
335 | const struct extent_buffer *src); | |
336 | void copy_extent_buffer(const struct extent_buffer *dst, | |
337 | const struct extent_buffer *src, | |
d1310b2e CM |
338 | unsigned long dst_offset, unsigned long src_offset, |
339 | unsigned long len); | |
2b48966a DS |
340 | void memcpy_extent_buffer(const struct extent_buffer *dst, |
341 | unsigned long dst_offset, unsigned long src_offset, | |
342 | unsigned long len); | |
343 | void memmove_extent_buffer(const struct extent_buffer *dst, | |
344 | unsigned long dst_offset, unsigned long src_offset, | |
b159fa28 | 345 | unsigned long len); |
2b48966a | 346 | void memzero_extent_buffer(const struct extent_buffer *eb, unsigned long start, |
b159fa28 | 347 | unsigned long len); |
2b48966a | 348 | int extent_buffer_test_bit(const struct extent_buffer *eb, unsigned long start, |
3e1e8bb7 | 349 | unsigned long pos); |
2b48966a | 350 | void extent_buffer_bitmap_set(const struct extent_buffer *eb, unsigned long start, |
3e1e8bb7 | 351 | unsigned long pos, unsigned long len); |
2b48966a DS |
352 | void extent_buffer_bitmap_clear(const struct extent_buffer *eb, |
353 | unsigned long start, unsigned long pos, | |
354 | unsigned long len); | |
f18cc978 | 355 | void set_extent_buffer_dirty(struct extent_buffer *eb); |
09c25a8c | 356 | void set_extent_buffer_uptodate(struct extent_buffer *eb); |
69ba3927 | 357 | void clear_extent_buffer_uptodate(struct extent_buffer *eb); |
ad7ff17b | 358 | void extent_clear_unlock_delalloc(struct btrfs_inode *inode, u64 start, u64 end, |
a67f5405 | 359 | const struct folio *locked_folio, |
6b0a63a4 | 360 | struct extent_state **cached, |
f97e27e9 | 361 | u32 bits_to_clear, unsigned long page_ops); |
d6f65c27 JB |
362 | int extent_invalidate_folio(struct extent_io_tree *tree, |
363 | struct folio *folio, size_t offset); | |
98c8d683 JB |
364 | void btrfs_clear_buffer_dirty(struct btrfs_trans_handle *trans, |
365 | struct extent_buffer *buf); | |
dd137dd1 | 366 | |
09e6cef1 | 367 | int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array, |
0fbf6cbd | 368 | bool nofail); |
fea91134 | 369 | int btrfs_alloc_folio_array(unsigned int nr_folios, struct folio **folio_array); |
4a54c8c1 | 370 | |
294e30fe | 371 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
9978059b | 372 | bool find_lock_delalloc_range(struct inode *inode, |
c987f1e6 | 373 | struct folio *locked_folio, u64 *start, |
ce9f967f | 374 | u64 *end); |
0d4cf4e6 | 375 | #endif |
faa2dbf0 | 376 | struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info, |
da17066c | 377 | u64 start); |
3fd63727 JB |
378 | |
379 | #ifdef CONFIG_BTRFS_DEBUG | |
380 | void btrfs_extent_buffer_leak_debug_check(struct btrfs_fs_info *fs_info); | |
381 | #else | |
382 | #define btrfs_extent_buffer_leak_debug_check(fs_info) do {} while (0) | |
383 | #endif | |
384 | ||
294e30fe | 385 | #endif |