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 /* Block group flags set at runtime */
50 enum btrfs_block_group_flags {
51 BLOCK_GROUP_FLAG_IREF,
52 BLOCK_GROUP_FLAG_REMOVED,
53 BLOCK_GROUP_FLAG_TO_COPY,
54 BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
55 BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
56 BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
57 BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
58 /* Does the block group need to be added to the free space tree? */
59 BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
60 /* Indicate that the block group is placed on a sequential zone */
61 BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
62 /*
63 * Indicate that block group is in the list of new block groups of a
64 * transaction.
65 */
66 BLOCK_GROUP_FLAG_NEW,
67 };
68
69 enum btrfs_caching_type {
70 BTRFS_CACHE_NO,
71 BTRFS_CACHE_STARTED,
72 BTRFS_CACHE_FINISHED,
73 BTRFS_CACHE_ERROR,
74 };
75
76 struct btrfs_caching_control {
77 struct list_head list;
78 struct mutex mutex;
79 wait_queue_head_t wait;
80 struct btrfs_work work;
81 struct btrfs_block_group *block_group;
82 /* Track progress of caching during allocation. */
83 atomic_t progress;
84 refcount_t count;
85 };
86
87 /* Once caching_thread() finds this much free space, it will wake up waiters. */
88 #define CACHING_CTL_WAKE_UP SZ_2M
89
90 /*
91 * Tree to record all locked full stripes of a RAID5/6 block group
92 */
93 struct btrfs_full_stripe_locks_tree {
94 struct rb_root root;
95 struct mutex lock;
96 };
97
98 struct btrfs_block_group {
99 struct btrfs_fs_info *fs_info;
100 struct inode *inode;
101 spinlock_t lock;
102 u64 start;
103 u64 length;
104 u64 pinned;
105 u64 reserved;
106 u64 used;
107 u64 delalloc_bytes;
108 u64 bytes_super;
109 u64 flags;
110 u64 cache_generation;
111 u64 global_root_id;
112
113 /*
114 * If the free space extent count exceeds this number, convert the block
115 * group to bitmaps.
116 */
117 u32 bitmap_high_thresh;
118
119 /*
120 * If the free space extent count drops below this number, convert the
121 * block group back to extents.
122 */
123 u32 bitmap_low_thresh;
124
125 /*
126 * It is just used for the delayed data space allocation because
127 * only the data space allocation and the relative metadata update
128 * can be done cross the transaction.
129 */
130 struct rw_semaphore data_rwsem;
131
132 /* For raid56, this is a full stripe, without parity */
133 unsigned long full_stripe_len;
134 unsigned long runtime_flags;
135
136 unsigned int ro;
137
138 int disk_cache_state;
139
140 /* Cache tracking stuff */
141 int cached;
142 struct btrfs_caching_control *caching_ctl;
143
144 struct btrfs_space_info *space_info;
145
146 /* Free space cache stuff */
147 struct btrfs_free_space_ctl *free_space_ctl;
148
149 /* Block group cache stuff */
150 struct rb_node cache_node;
151
152 /* For block groups in the same raid type */
153 struct list_head list;
154
155 refcount_t refs;
156
157 /*
158 * List of struct btrfs_free_clusters for this block group.
159 * Today it will only have one thing on it, but that may change
160 */
161 struct list_head cluster_list;
162
163 /* For delayed block group creation or deletion of empty block groups */
164 struct list_head bg_list;
165
166 /* For read-only block groups */
167 struct list_head ro_list;
168
169 /*
170 * When non-zero it means the block group's logical address and its
171 * device extents can not be reused for future block group allocations
172 * until the counter goes down to 0. This is to prevent them from being
173 * reused while some task is still using the block group after it was
174 * deleted - we want to make sure they can only be reused for new block
175 * groups after that task is done with the deleted block group.
176 */
177 atomic_t frozen;
178
179 /* For discard operations */
180 struct list_head discard_list;
181 int discard_index;
182 u64 discard_eligible_time;
183 u64 discard_cursor;
184 enum btrfs_discard_state discard_state;
185
186 /* For dirty block groups */
187 struct list_head dirty_list;
188 struct list_head io_list;
189
190 struct btrfs_io_ctl io_ctl;
191
192 /*
193 * Incremented when doing extent allocations and holding a read lock
194 * on the space_info's groups_sem semaphore.
195 * Decremented when an ordered extent that represents an IO against this
196 * block group's range is created (after it's added to its inode's
197 * root's list of ordered extents) or immediately after the allocation
198 * if it's a metadata extent or fallocate extent (for these cases we
199 * don't create ordered extents).
200 */
201 atomic_t reservations;
202
203 /*
204 * Incremented while holding the spinlock *lock* by a task checking if
205 * it can perform a nocow write (incremented if the value for the *ro*
206 * field is 0). Decremented by such tasks once they create an ordered
207 * extent or before that if some error happens before reaching that step.
208 * This is to prevent races between block group relocation and nocow
209 * writes through direct IO.
210 */
211 atomic_t nocow_writers;
212
213 /* Lock for free space tree operations. */
214 struct mutex free_space_lock;
215
216 /*
217 * Number of extents in this block group used for swap files.
218 * All accesses protected by the spinlock 'lock'.
219 */
220 int swap_extents;
221
222 /* Record locked full stripes for RAID5/6 block group */
223 struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
224
225 /*
226 * Allocation offset for the block group to implement sequential
227 * allocation. This is used only on a zoned filesystem.
228 */
229 u64 alloc_offset;
230 u64 zone_unusable;
231 u64 zone_capacity;
232 u64 meta_write_pointer;
233 struct map_lookup *physical_map;
234 struct list_head active_bg_list;
235 struct work_struct zone_finish_work;
236 struct extent_buffer *last_eb;
237 };
238
btrfs_block_group_end(struct btrfs_block_group * block_group)239 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
240 {
241 return (block_group->start + block_group->length);
242 }
243
btrfs_is_block_group_used(const struct btrfs_block_group * bg)244 static inline bool btrfs_is_block_group_used(const struct btrfs_block_group *bg)
245 {
246 lockdep_assert_held(&bg->lock);
247
248 return (bg->used > 0 || bg->reserved > 0 || bg->pinned > 0);
249 }
250
btrfs_is_block_group_data_only(struct btrfs_block_group * block_group)251 static inline bool btrfs_is_block_group_data_only(
252 struct btrfs_block_group *block_group)
253 {
254 /*
255 * In mixed mode the fragmentation is expected to be high, lowering the
256 * efficiency, so only proper data block groups are considered.
257 */
258 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
259 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
260 }
261
262 #ifdef CONFIG_BTRFS_DEBUG
btrfs_should_fragment_free_space(struct btrfs_block_group * block_group)263 static inline int btrfs_should_fragment_free_space(
264 struct btrfs_block_group *block_group)
265 {
266 struct btrfs_fs_info *fs_info = block_group->fs_info;
267
268 return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
269 block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
270 (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
271 block_group->flags & BTRFS_BLOCK_GROUP_DATA);
272 }
273 #endif
274
275 struct btrfs_block_group *btrfs_lookup_first_block_group(
276 struct btrfs_fs_info *info, u64 bytenr);
277 struct btrfs_block_group *btrfs_lookup_block_group(
278 struct btrfs_fs_info *info, u64 bytenr);
279 struct btrfs_block_group *btrfs_next_block_group(
280 struct btrfs_block_group *cache);
281 void btrfs_get_block_group(struct btrfs_block_group *cache);
282 void btrfs_put_block_group(struct btrfs_block_group *cache);
283 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
284 const u64 start);
285 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
286 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
287 u64 bytenr);
288 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
289 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
290 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
291 u64 num_bytes);
292 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
293 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
294 struct btrfs_caching_control *btrfs_get_caching_control(
295 struct btrfs_block_group *cache);
296 int add_new_free_space(struct btrfs_block_group *block_group,
297 u64 start, u64 end, u64 *total_added_ret);
298 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
299 struct btrfs_fs_info *fs_info,
300 const u64 chunk_offset);
301 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
302 u64 group_start, struct extent_map *em);
303 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
304 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
305 void btrfs_reclaim_bgs_work(struct work_struct *work);
306 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
307 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
308 int btrfs_read_block_groups(struct btrfs_fs_info *info);
309 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
310 u64 bytes_used, u64 type,
311 u64 chunk_offset, u64 size);
312 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
313 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
314 bool do_chunk_alloc);
315 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
316 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
317 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
318 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
319 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
320 u64 bytenr, u64 num_bytes, bool alloc);
321 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
322 u64 ram_bytes, u64 num_bytes, int delalloc);
323 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
324 u64 num_bytes, int delalloc);
325 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
326 enum btrfs_chunk_alloc_enum force);
327 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
328 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
329 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
330 bool is_item_insertion);
331 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
332 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
333 int btrfs_free_block_groups(struct btrfs_fs_info *info);
334 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
335 struct block_device *bdev, u64 physical, u64 **logical,
336 int *naddrs, int *stripe_len);
337
btrfs_data_alloc_profile(struct btrfs_fs_info * fs_info)338 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
339 {
340 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
341 }
342
btrfs_metadata_alloc_profile(struct btrfs_fs_info * fs_info)343 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
344 {
345 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
346 }
347
btrfs_system_alloc_profile(struct btrfs_fs_info * fs_info)348 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
349 {
350 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
351 }
352
btrfs_block_group_done(struct btrfs_block_group * cache)353 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
354 {
355 smp_mb();
356 return cache->cached == BTRFS_CACHE_FINISHED ||
357 cache->cached == BTRFS_CACHE_ERROR;
358 }
359
360 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
361 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
362
363 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
364 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
365
366 #endif /* BTRFS_BLOCK_GROUP_H */
367