1 /* SPDX-License-Identifier: GPL-2.0 */
2
3 #ifndef BTRFS_BLOCK_GROUP_H
4 #define BTRFS_BLOCK_GROUP_H
5
6 #include "free-space-cache.h"
7
8 enum btrfs_disk_cache_state {
9 BTRFS_DC_WRITTEN,
10 BTRFS_DC_ERROR,
11 BTRFS_DC_CLEAR,
12 BTRFS_DC_SETUP,
13 };
14
15 /*
16 * This describes the state of the block_group for async discard. This is due
17 * to the two pass nature of it where extent discarding is prioritized over
18 * bitmap discarding. BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
19 * between lists to prevent contention for discard state variables
20 * (eg. discard_cursor).
21 */
22 enum btrfs_discard_state {
23 BTRFS_DISCARD_EXTENTS,
24 BTRFS_DISCARD_BITMAPS,
25 BTRFS_DISCARD_RESET_CURSOR,
26 };
27
28 /*
29 * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
30 * only allocate a chunk if we really need one.
31 *
32 * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
33 * chunks already allocated. This is used as part of the clustering code to
34 * help make sure we have a good pool of storage to cluster in, without filling
35 * the FS with empty chunks
36 *
37 * CHUNK_ALLOC_FORCE means it must try to allocate one
38 *
39 * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
40 * find_free_extent() that also activaes the zone
41 */
42 enum btrfs_chunk_alloc_enum {
43 CHUNK_ALLOC_NO_FORCE,
44 CHUNK_ALLOC_LIMITED,
45 CHUNK_ALLOC_FORCE,
46 CHUNK_ALLOC_FORCE_FOR_EXTENT,
47 };
48
49 struct btrfs_caching_control {
50 struct list_head list;
51 struct mutex mutex;
52 wait_queue_head_t wait;
53 struct btrfs_work work;
54 struct btrfs_block_group *block_group;
55 u64 progress;
56 refcount_t count;
57 };
58
59 /* Once caching_thread() finds this much free space, it will wake up waiters. */
60 #define CACHING_CTL_WAKE_UP SZ_2M
61
62 struct btrfs_block_group {
63 struct btrfs_fs_info *fs_info;
64 struct inode *inode;
65 spinlock_t lock;
66 u64 start;
67 u64 length;
68 u64 pinned;
69 u64 reserved;
70 u64 used;
71 u64 delalloc_bytes;
72 u64 bytes_super;
73 u64 flags;
74 u64 cache_generation;
75 u64 global_root_id;
76
77 /*
78 * If the free space extent count exceeds this number, convert the block
79 * group to bitmaps.
80 */
81 u32 bitmap_high_thresh;
82
83 /*
84 * If the free space extent count drops below this number, convert the
85 * block group back to extents.
86 */
87 u32 bitmap_low_thresh;
88
89 /*
90 * It is just used for the delayed data space allocation because
91 * only the data space allocation and the relative metadata update
92 * can be done cross the transaction.
93 */
94 struct rw_semaphore data_rwsem;
95
96 /* For raid56, this is a full stripe, without parity */
97 unsigned long full_stripe_len;
98
99 unsigned int ro;
100 unsigned int iref:1;
101 unsigned int has_caching_ctl:1;
102 unsigned int removed:1;
103 unsigned int to_copy:1;
104 unsigned int relocating_repair:1;
105 unsigned int chunk_item_inserted:1;
106 unsigned int zone_is_active:1;
107 unsigned int zoned_data_reloc_ongoing:1;
108
109 int disk_cache_state;
110
111 /* Cache tracking stuff */
112 int cached;
113 struct btrfs_caching_control *caching_ctl;
114 u64 last_byte_to_unpin;
115
116 struct btrfs_space_info *space_info;
117
118 /* Free space cache stuff */
119 struct btrfs_free_space_ctl *free_space_ctl;
120
121 /* Block group cache stuff */
122 struct rb_node cache_node;
123
124 /* For block groups in the same raid type */
125 struct list_head list;
126
127 refcount_t refs;
128
129 /*
130 * List of struct btrfs_free_clusters for this block group.
131 * Today it will only have one thing on it, but that may change
132 */
133 struct list_head cluster_list;
134
135 /* For delayed block group creation or deletion of empty block groups */
136 struct list_head bg_list;
137
138 /* For read-only block groups */
139 struct list_head ro_list;
140
141 /*
142 * When non-zero it means the block group's logical address and its
143 * device extents can not be reused for future block group allocations
144 * until the counter goes down to 0. This is to prevent them from being
145 * reused while some task is still using the block group after it was
146 * deleted - we want to make sure they can only be reused for new block
147 * groups after that task is done with the deleted block group.
148 */
149 atomic_t frozen;
150
151 /* For discard operations */
152 struct list_head discard_list;
153 int discard_index;
154 u64 discard_eligible_time;
155 u64 discard_cursor;
156 enum btrfs_discard_state discard_state;
157
158 /* For dirty block groups */
159 struct list_head dirty_list;
160 struct list_head io_list;
161
162 struct btrfs_io_ctl io_ctl;
163
164 /*
165 * Incremented when doing extent allocations and holding a read lock
166 * on the space_info's groups_sem semaphore.
167 * Decremented when an ordered extent that represents an IO against this
168 * block group's range is created (after it's added to its inode's
169 * root's list of ordered extents) or immediately after the allocation
170 * if it's a metadata extent or fallocate extent (for these cases we
171 * don't create ordered extents).
172 */
173 atomic_t reservations;
174
175 /*
176 * Incremented while holding the spinlock *lock* by a task checking if
177 * it can perform a nocow write (incremented if the value for the *ro*
178 * field is 0). Decremented by such tasks once they create an ordered
179 * extent or before that if some error happens before reaching that step.
180 * This is to prevent races between block group relocation and nocow
181 * writes through direct IO.
182 */
183 atomic_t nocow_writers;
184
185 /* Lock for free space tree operations. */
186 struct mutex free_space_lock;
187
188 /*
189 * Does the block group need to be added to the free space tree?
190 * Protected by free_space_lock.
191 */
192 int needs_free_space;
193
194 /* Flag indicating this block group is placed on a sequential zone */
195 bool seq_zone;
196
197 /*
198 * Number of extents in this block group used for swap files.
199 * All accesses protected by the spinlock 'lock'.
200 */
201 int swap_extents;
202
203 /* Record locked full stripes for RAID5/6 block group */
204 struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
205
206 /*
207 * Allocation offset for the block group to implement sequential
208 * allocation. This is used only on a zoned filesystem.
209 */
210 u64 alloc_offset;
211 u64 zone_unusable;
212 u64 zone_capacity;
213 u64 meta_write_pointer;
214 struct map_lookup *physical_map;
215 struct list_head active_bg_list;
216 struct work_struct zone_finish_work;
217 struct extent_buffer *last_eb;
218 };
219
btrfs_block_group_end(struct btrfs_block_group * block_group)220 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
221 {
222 return (block_group->start + block_group->length);
223 }
224
btrfs_is_block_group_data_only(struct btrfs_block_group * block_group)225 static inline bool btrfs_is_block_group_data_only(
226 struct btrfs_block_group *block_group)
227 {
228 /*
229 * In mixed mode the fragmentation is expected to be high, lowering the
230 * efficiency, so only proper data block groups are considered.
231 */
232 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
233 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
234 }
235
236 #ifdef CONFIG_BTRFS_DEBUG
btrfs_should_fragment_free_space(struct btrfs_block_group * block_group)237 static inline int btrfs_should_fragment_free_space(
238 struct btrfs_block_group *block_group)
239 {
240 struct btrfs_fs_info *fs_info = block_group->fs_info;
241
242 return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
243 block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
244 (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
245 block_group->flags & BTRFS_BLOCK_GROUP_DATA);
246 }
247 #endif
248
249 struct btrfs_block_group *btrfs_lookup_first_block_group(
250 struct btrfs_fs_info *info, u64 bytenr);
251 struct btrfs_block_group *btrfs_lookup_block_group(
252 struct btrfs_fs_info *info, u64 bytenr);
253 struct btrfs_block_group *btrfs_next_block_group(
254 struct btrfs_block_group *cache);
255 void btrfs_get_block_group(struct btrfs_block_group *cache);
256 void btrfs_put_block_group(struct btrfs_block_group *cache);
257 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
258 const u64 start);
259 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
260 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
261 u64 bytenr);
262 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
263 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
264 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
265 u64 num_bytes);
266 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
267 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
268 struct btrfs_caching_control *btrfs_get_caching_control(
269 struct btrfs_block_group *cache);
270 u64 add_new_free_space(struct btrfs_block_group *block_group,
271 u64 start, u64 end);
272 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
273 struct btrfs_fs_info *fs_info,
274 const u64 chunk_offset);
275 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
276 u64 group_start, struct extent_map *em);
277 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
278 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
279 void btrfs_reclaim_bgs_work(struct work_struct *work);
280 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
281 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
282 int btrfs_read_block_groups(struct btrfs_fs_info *info);
283 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
284 u64 bytes_used, u64 type,
285 u64 chunk_offset, u64 size);
286 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
287 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
288 bool do_chunk_alloc);
289 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
290 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
291 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
292 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
293 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
294 u64 bytenr, u64 num_bytes, bool alloc);
295 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
296 u64 ram_bytes, u64 num_bytes, int delalloc);
297 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
298 u64 num_bytes, int delalloc);
299 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
300 enum btrfs_chunk_alloc_enum force);
301 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
302 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
303 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
304 bool is_item_insertion);
305 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
306 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
307 int btrfs_free_block_groups(struct btrfs_fs_info *info);
308 void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,
309 struct btrfs_caching_control *caching_ctl);
310 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
311 struct block_device *bdev, u64 physical, u64 **logical,
312 int *naddrs, int *stripe_len);
313
btrfs_data_alloc_profile(struct btrfs_fs_info * fs_info)314 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
315 {
316 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
317 }
318
btrfs_metadata_alloc_profile(struct btrfs_fs_info * fs_info)319 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
320 {
321 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
322 }
323
btrfs_system_alloc_profile(struct btrfs_fs_info * fs_info)324 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
325 {
326 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
327 }
328
btrfs_block_group_done(struct btrfs_block_group * cache)329 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
330 {
331 smp_mb();
332 return cache->cached == BTRFS_CACHE_FINISHED ||
333 cache->cached == BTRFS_CACHE_ERROR;
334 }
335
336 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
337 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
338
339 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
340 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
341
342 #endif /* BTRFS_BLOCK_GROUP_H */
343