xref: /openbmc/linux/fs/btrfs/block-group.h (revision a16be368)
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 
99 	int disk_cache_state;
100 
101 	/* Cache tracking stuff */
102 	int cached;
103 	struct btrfs_caching_control *caching_ctl;
104 	u64 last_byte_to_unpin;
105 
106 	struct btrfs_space_info *space_info;
107 
108 	/* Free space cache stuff */
109 	struct btrfs_free_space_ctl *free_space_ctl;
110 
111 	/* Block group cache stuff */
112 	struct rb_node cache_node;
113 
114 	/* For block groups in the same raid type */
115 	struct list_head list;
116 
117 	/* Usage count */
118 	atomic_t count;
119 
120 	/*
121 	 * List of struct btrfs_free_clusters for this block group.
122 	 * Today it will only have one thing on it, but that may change
123 	 */
124 	struct list_head cluster_list;
125 
126 	/* For delayed block group creation or deletion of empty block groups */
127 	struct list_head bg_list;
128 
129 	/* For read-only block groups */
130 	struct list_head ro_list;
131 
132 	/*
133 	 * When non-zero it means the block group's logical address and its
134 	 * device extents can not be reused for future block group allocations
135 	 * until the counter goes down to 0. This is to prevent them from being
136 	 * reused while some task is still using the block group after it was
137 	 * deleted - we want to make sure they can only be reused for new block
138 	 * groups after that task is done with the deleted block group.
139 	 */
140 	atomic_t frozen;
141 
142 	/* For discard operations */
143 	struct list_head discard_list;
144 	int discard_index;
145 	u64 discard_eligible_time;
146 	u64 discard_cursor;
147 	enum btrfs_discard_state discard_state;
148 
149 	/* For dirty block groups */
150 	struct list_head dirty_list;
151 	struct list_head io_list;
152 
153 	struct btrfs_io_ctl io_ctl;
154 
155 	/*
156 	 * Incremented when doing extent allocations and holding a read lock
157 	 * on the space_info's groups_sem semaphore.
158 	 * Decremented when an ordered extent that represents an IO against this
159 	 * block group's range is created (after it's added to its inode's
160 	 * root's list of ordered extents) or immediately after the allocation
161 	 * if it's a metadata extent or fallocate extent (for these cases we
162 	 * don't create ordered extents).
163 	 */
164 	atomic_t reservations;
165 
166 	/*
167 	 * Incremented while holding the spinlock *lock* by a task checking if
168 	 * it can perform a nocow write (incremented if the value for the *ro*
169 	 * field is 0). Decremented by such tasks once they create an ordered
170 	 * extent or before that if some error happens before reaching that step.
171 	 * This is to prevent races between block group relocation and nocow
172 	 * writes through direct IO.
173 	 */
174 	atomic_t nocow_writers;
175 
176 	/* Lock for free space tree operations. */
177 	struct mutex free_space_lock;
178 
179 	/*
180 	 * Does the block group need to be added to the free space tree?
181 	 * Protected by free_space_lock.
182 	 */
183 	int needs_free_space;
184 
185 	/* Record locked full stripes for RAID5/6 block group */
186 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
187 };
188 
189 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
190 {
191 	return (block_group->start + block_group->length);
192 }
193 
194 static inline bool btrfs_is_block_group_data_only(
195 					struct btrfs_block_group *block_group)
196 {
197 	/*
198 	 * In mixed mode the fragmentation is expected to be high, lowering the
199 	 * efficiency, so only proper data block groups are considered.
200 	 */
201 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
202 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
203 }
204 
205 #ifdef CONFIG_BTRFS_DEBUG
206 static inline int btrfs_should_fragment_free_space(
207 		struct btrfs_block_group *block_group)
208 {
209 	struct btrfs_fs_info *fs_info = block_group->fs_info;
210 
211 	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
212 		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
213 	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
214 		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
215 }
216 #endif
217 
218 struct btrfs_block_group *btrfs_lookup_first_block_group(
219 		struct btrfs_fs_info *info, u64 bytenr);
220 struct btrfs_block_group *btrfs_lookup_block_group(
221 		struct btrfs_fs_info *info, u64 bytenr);
222 struct btrfs_block_group *btrfs_next_block_group(
223 		struct btrfs_block_group *cache);
224 void btrfs_get_block_group(struct btrfs_block_group *cache);
225 void btrfs_put_block_group(struct btrfs_block_group *cache);
226 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
227 					const u64 start);
228 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
229 bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
230 void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
231 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
232 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
233 				           u64 num_bytes);
234 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
235 int btrfs_cache_block_group(struct btrfs_block_group *cache,
236 			    int load_cache_only);
237 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
238 struct btrfs_caching_control *btrfs_get_caching_control(
239 		struct btrfs_block_group *cache);
240 u64 add_new_free_space(struct btrfs_block_group *block_group,
241 		       u64 start, u64 end);
242 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
243 				struct btrfs_fs_info *fs_info,
244 				const u64 chunk_offset);
245 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
246 			     u64 group_start, struct extent_map *em);
247 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
248 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
249 int btrfs_read_block_groups(struct btrfs_fs_info *info);
250 int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
251 			   u64 type, u64 chunk_offset, u64 size);
252 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
253 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
254 			     bool do_chunk_alloc);
255 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
256 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
257 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
258 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
259 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
260 			     u64 bytenr, u64 num_bytes, int alloc);
261 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
262 			     u64 ram_bytes, u64 num_bytes, int delalloc);
263 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
264 			       u64 num_bytes, int delalloc);
265 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
266 		      enum btrfs_chunk_alloc_enum force);
267 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
268 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
269 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
270 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
271 int btrfs_free_block_groups(struct btrfs_fs_info *info);
272 
273 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
274 {
275 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
276 }
277 
278 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
279 {
280 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
281 }
282 
283 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
284 {
285 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
286 }
287 
288 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
289 {
290 	smp_mb();
291 	return cache->cached == BTRFS_CACHE_FINISHED ||
292 		cache->cached == BTRFS_CACHE_ERROR;
293 }
294 
295 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
296 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
297 
298 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
299 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
300 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
301 #endif
302 
303 #endif /* BTRFS_BLOCK_GROUP_H */
304