xref: /openbmc/linux/fs/btrfs/block-group.h (revision babbdf5b)
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 enum btrfs_chunk_alloc_enum {
40 	CHUNK_ALLOC_NO_FORCE,
41 	CHUNK_ALLOC_LIMITED,
42 	CHUNK_ALLOC_FORCE,
43 };
44 
45 struct btrfs_caching_control {
46 	struct list_head list;
47 	struct mutex mutex;
48 	wait_queue_head_t wait;
49 	struct btrfs_work work;
50 	struct btrfs_block_group *block_group;
51 	u64 progress;
52 	refcount_t count;
53 };
54 
55 /* Once caching_thread() finds this much free space, it will wake up waiters. */
56 #define CACHING_CTL_WAKE_UP SZ_2M
57 
58 struct btrfs_block_group {
59 	struct btrfs_fs_info *fs_info;
60 	struct inode *inode;
61 	spinlock_t lock;
62 	u64 start;
63 	u64 length;
64 	u64 pinned;
65 	u64 reserved;
66 	u64 used;
67 	u64 delalloc_bytes;
68 	u64 bytes_super;
69 	u64 flags;
70 	u64 cache_generation;
71 
72 	/*
73 	 * If the free space extent count exceeds this number, convert the block
74 	 * group to bitmaps.
75 	 */
76 	u32 bitmap_high_thresh;
77 
78 	/*
79 	 * If the free space extent count drops below this number, convert the
80 	 * block group back to extents.
81 	 */
82 	u32 bitmap_low_thresh;
83 
84 	/*
85 	 * It is just used for the delayed data space allocation because
86 	 * only the data space allocation and the relative metadata update
87 	 * can be done cross the transaction.
88 	 */
89 	struct rw_semaphore data_rwsem;
90 
91 	/* For raid56, this is a full stripe, without parity */
92 	unsigned long full_stripe_len;
93 
94 	unsigned int ro;
95 	unsigned int iref:1;
96 	unsigned int has_caching_ctl:1;
97 	unsigned int removed:1;
98 	unsigned int to_copy:1;
99 	unsigned int relocating_repair:1;
100 
101 	int disk_cache_state;
102 
103 	/* Cache tracking stuff */
104 	int cached;
105 	struct btrfs_caching_control *caching_ctl;
106 	u64 last_byte_to_unpin;
107 
108 	struct btrfs_space_info *space_info;
109 
110 	/* Free space cache stuff */
111 	struct btrfs_free_space_ctl *free_space_ctl;
112 
113 	/* Block group cache stuff */
114 	struct rb_node cache_node;
115 
116 	/* For block groups in the same raid type */
117 	struct list_head list;
118 
119 	refcount_t refs;
120 
121 	/*
122 	 * List of struct btrfs_free_clusters for this block group.
123 	 * Today it will only have one thing on it, but that may change
124 	 */
125 	struct list_head cluster_list;
126 
127 	/* For delayed block group creation or deletion of empty block groups */
128 	struct list_head bg_list;
129 
130 	/* For read-only block groups */
131 	struct list_head ro_list;
132 
133 	/*
134 	 * When non-zero it means the block group's logical address and its
135 	 * device extents can not be reused for future block group allocations
136 	 * until the counter goes down to 0. This is to prevent them from being
137 	 * reused while some task is still using the block group after it was
138 	 * deleted - we want to make sure they can only be reused for new block
139 	 * groups after that task is done with the deleted block group.
140 	 */
141 	atomic_t frozen;
142 
143 	/* For discard operations */
144 	struct list_head discard_list;
145 	int discard_index;
146 	u64 discard_eligible_time;
147 	u64 discard_cursor;
148 	enum btrfs_discard_state discard_state;
149 
150 	/* For dirty block groups */
151 	struct list_head dirty_list;
152 	struct list_head io_list;
153 
154 	struct btrfs_io_ctl io_ctl;
155 
156 	/*
157 	 * Incremented when doing extent allocations and holding a read lock
158 	 * on the space_info's groups_sem semaphore.
159 	 * Decremented when an ordered extent that represents an IO against this
160 	 * block group's range is created (after it's added to its inode's
161 	 * root's list of ordered extents) or immediately after the allocation
162 	 * if it's a metadata extent or fallocate extent (for these cases we
163 	 * don't create ordered extents).
164 	 */
165 	atomic_t reservations;
166 
167 	/*
168 	 * Incremented while holding the spinlock *lock* by a task checking if
169 	 * it can perform a nocow write (incremented if the value for the *ro*
170 	 * field is 0). Decremented by such tasks once they create an ordered
171 	 * extent or before that if some error happens before reaching that step.
172 	 * This is to prevent races between block group relocation and nocow
173 	 * writes through direct IO.
174 	 */
175 	atomic_t nocow_writers;
176 
177 	/* Lock for free space tree operations. */
178 	struct mutex free_space_lock;
179 
180 	/*
181 	 * Does the block group need to be added to the free space tree?
182 	 * Protected by free_space_lock.
183 	 */
184 	int needs_free_space;
185 
186 	/* Flag indicating this block group is placed on a sequential zone */
187 	bool seq_zone;
188 
189 	/*
190 	 * Number of extents in this block group used for swap files.
191 	 * All accesses protected by the spinlock 'lock'.
192 	 */
193 	int swap_extents;
194 
195 	/* Record locked full stripes for RAID5/6 block group */
196 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
197 
198 	/*
199 	 * Allocation offset for the block group to implement sequential
200 	 * allocation. This is used only on a zoned filesystem.
201 	 */
202 	u64 alloc_offset;
203 	u64 zone_unusable;
204 	u64 meta_write_pointer;
205 };
206 
207 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
208 {
209 	return (block_group->start + block_group->length);
210 }
211 
212 static inline bool btrfs_is_block_group_data_only(
213 					struct btrfs_block_group *block_group)
214 {
215 	/*
216 	 * In mixed mode the fragmentation is expected to be high, lowering the
217 	 * efficiency, so only proper data block groups are considered.
218 	 */
219 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
220 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
221 }
222 
223 #ifdef CONFIG_BTRFS_DEBUG
224 static inline int btrfs_should_fragment_free_space(
225 		struct btrfs_block_group *block_group)
226 {
227 	struct btrfs_fs_info *fs_info = block_group->fs_info;
228 
229 	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
230 		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
231 	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
232 		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
233 }
234 #endif
235 
236 struct btrfs_block_group *btrfs_lookup_first_block_group(
237 		struct btrfs_fs_info *info, u64 bytenr);
238 struct btrfs_block_group *btrfs_lookup_block_group(
239 		struct btrfs_fs_info *info, u64 bytenr);
240 struct btrfs_block_group *btrfs_next_block_group(
241 		struct btrfs_block_group *cache);
242 void btrfs_get_block_group(struct btrfs_block_group *cache);
243 void btrfs_put_block_group(struct btrfs_block_group *cache);
244 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
245 					const u64 start);
246 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
247 bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
248 void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
249 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
250 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
251 				           u64 num_bytes);
252 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
253 int btrfs_cache_block_group(struct btrfs_block_group *cache,
254 			    int load_cache_only);
255 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
256 struct btrfs_caching_control *btrfs_get_caching_control(
257 		struct btrfs_block_group *cache);
258 u64 add_new_free_space(struct btrfs_block_group *block_group,
259 		       u64 start, u64 end);
260 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
261 				struct btrfs_fs_info *fs_info,
262 				const u64 chunk_offset);
263 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
264 			     u64 group_start, struct extent_map *em);
265 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
266 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
267 void btrfs_reclaim_bgs_work(struct work_struct *work);
268 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
269 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
270 int btrfs_read_block_groups(struct btrfs_fs_info *info);
271 int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
272 			   u64 type, u64 chunk_offset, u64 size);
273 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
274 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
275 			     bool do_chunk_alloc);
276 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
277 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
278 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
279 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
280 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
281 			     u64 bytenr, u64 num_bytes, int alloc);
282 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
283 			     u64 ram_bytes, u64 num_bytes, int delalloc);
284 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
285 			       u64 num_bytes, int delalloc);
286 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
287 		      enum btrfs_chunk_alloc_enum force);
288 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
289 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
290 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
291 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
292 int btrfs_free_block_groups(struct btrfs_fs_info *info);
293 void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,
294 				struct btrfs_caching_control *caching_ctl);
295 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
296 		       struct block_device *bdev, u64 physical, u64 **logical,
297 		       int *naddrs, int *stripe_len);
298 
299 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
300 {
301 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
302 }
303 
304 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
305 {
306 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
307 }
308 
309 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
310 {
311 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
312 }
313 
314 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
315 {
316 	smp_mb();
317 	return cache->cached == BTRFS_CACHE_FINISHED ||
318 		cache->cached == BTRFS_CACHE_ERROR;
319 }
320 
321 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
322 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
323 
324 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
325 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
326 
327 #endif /* BTRFS_BLOCK_GROUP_H */
328