xref: /openbmc/linux/fs/btrfs/block-group.h (revision 715f23b6)
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  * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
17  * only allocate a chunk if we really need one.
18  *
19  * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
20  * chunks already allocated.  This is used as part of the clustering code to
21  * help make sure we have a good pool of storage to cluster in, without filling
22  * the FS with empty chunks
23  *
24  * CHUNK_ALLOC_FORCE means it must try to allocate one
25  */
26 enum btrfs_chunk_alloc_enum {
27 	CHUNK_ALLOC_NO_FORCE,
28 	CHUNK_ALLOC_LIMITED,
29 	CHUNK_ALLOC_FORCE,
30 };
31 
32 struct btrfs_caching_control {
33 	struct list_head list;
34 	struct mutex mutex;
35 	wait_queue_head_t wait;
36 	struct btrfs_work work;
37 	struct btrfs_block_group *block_group;
38 	u64 progress;
39 	refcount_t count;
40 };
41 
42 /* Once caching_thread() finds this much free space, it will wake up waiters. */
43 #define CACHING_CTL_WAKE_UP SZ_2M
44 
45 struct btrfs_block_group {
46 	struct btrfs_fs_info *fs_info;
47 	struct inode *inode;
48 	spinlock_t lock;
49 	u64 start;
50 	u64 length;
51 	u64 pinned;
52 	u64 reserved;
53 	u64 used;
54 	u64 delalloc_bytes;
55 	u64 bytes_super;
56 	u64 flags;
57 	u64 cache_generation;
58 
59 	/*
60 	 * If the free space extent count exceeds this number, convert the block
61 	 * group to bitmaps.
62 	 */
63 	u32 bitmap_high_thresh;
64 
65 	/*
66 	 * If the free space extent count drops below this number, convert the
67 	 * block group back to extents.
68 	 */
69 	u32 bitmap_low_thresh;
70 
71 	/*
72 	 * It is just used for the delayed data space allocation because
73 	 * only the data space allocation and the relative metadata update
74 	 * can be done cross the transaction.
75 	 */
76 	struct rw_semaphore data_rwsem;
77 
78 	/* For raid56, this is a full stripe, without parity */
79 	unsigned long full_stripe_len;
80 
81 	unsigned int ro;
82 	unsigned int iref:1;
83 	unsigned int has_caching_ctl:1;
84 	unsigned int removed:1;
85 
86 	int disk_cache_state;
87 
88 	/* Cache tracking stuff */
89 	int cached;
90 	struct btrfs_caching_control *caching_ctl;
91 	u64 last_byte_to_unpin;
92 
93 	struct btrfs_space_info *space_info;
94 
95 	/* Free space cache stuff */
96 	struct btrfs_free_space_ctl *free_space_ctl;
97 
98 	/* Block group cache stuff */
99 	struct rb_node cache_node;
100 
101 	/* For block groups in the same raid type */
102 	struct list_head list;
103 
104 	/* Usage count */
105 	atomic_t count;
106 
107 	/*
108 	 * List of struct btrfs_free_clusters for this block group.
109 	 * Today it will only have one thing on it, but that may change
110 	 */
111 	struct list_head cluster_list;
112 
113 	/* For delayed block group creation or deletion of empty block groups */
114 	struct list_head bg_list;
115 
116 	/* For read-only block groups */
117 	struct list_head ro_list;
118 
119 	atomic_t trimming;
120 
121 	/* For dirty block groups */
122 	struct list_head dirty_list;
123 	struct list_head io_list;
124 
125 	struct btrfs_io_ctl io_ctl;
126 
127 	/*
128 	 * Incremented when doing extent allocations and holding a read lock
129 	 * on the space_info's groups_sem semaphore.
130 	 * Decremented when an ordered extent that represents an IO against this
131 	 * block group's range is created (after it's added to its inode's
132 	 * root's list of ordered extents) or immediately after the allocation
133 	 * if it's a metadata extent or fallocate extent (for these cases we
134 	 * don't create ordered extents).
135 	 */
136 	atomic_t reservations;
137 
138 	/*
139 	 * Incremented while holding the spinlock *lock* by a task checking if
140 	 * it can perform a nocow write (incremented if the value for the *ro*
141 	 * field is 0). Decremented by such tasks once they create an ordered
142 	 * extent or before that if some error happens before reaching that step.
143 	 * This is to prevent races between block group relocation and nocow
144 	 * writes through direct IO.
145 	 */
146 	atomic_t nocow_writers;
147 
148 	/* Lock for free space tree operations. */
149 	struct mutex free_space_lock;
150 
151 	/*
152 	 * Does the block group need to be added to the free space tree?
153 	 * Protected by free_space_lock.
154 	 */
155 	int needs_free_space;
156 
157 	/* Record locked full stripes for RAID5/6 block group */
158 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
159 };
160 
161 #ifdef CONFIG_BTRFS_DEBUG
162 static inline int btrfs_should_fragment_free_space(
163 		struct btrfs_block_group *block_group)
164 {
165 	struct btrfs_fs_info *fs_info = block_group->fs_info;
166 
167 	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
168 		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
169 	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
170 		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
171 }
172 #endif
173 
174 struct btrfs_block_group *btrfs_lookup_first_block_group(
175 		struct btrfs_fs_info *info, u64 bytenr);
176 struct btrfs_block_group *btrfs_lookup_block_group(
177 		struct btrfs_fs_info *info, u64 bytenr);
178 struct btrfs_block_group *btrfs_next_block_group(
179 		struct btrfs_block_group *cache);
180 void btrfs_get_block_group(struct btrfs_block_group *cache);
181 void btrfs_put_block_group(struct btrfs_block_group *cache);
182 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
183 					const u64 start);
184 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
185 bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
186 void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
187 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
188 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
189 				           u64 num_bytes);
190 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
191 int btrfs_cache_block_group(struct btrfs_block_group *cache,
192 			    int load_cache_only);
193 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
194 struct btrfs_caching_control *btrfs_get_caching_control(
195 		struct btrfs_block_group *cache);
196 u64 add_new_free_space(struct btrfs_block_group *block_group,
197 		       u64 start, u64 end);
198 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
199 				struct btrfs_fs_info *fs_info,
200 				const u64 chunk_offset);
201 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
202 			     u64 group_start, struct extent_map *em);
203 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
204 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
205 int btrfs_read_block_groups(struct btrfs_fs_info *info);
206 int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
207 			   u64 type, u64 chunk_offset, u64 size);
208 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
209 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
210 			     bool do_chunk_alloc);
211 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
212 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
213 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
214 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
215 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
216 			     u64 bytenr, u64 num_bytes, int alloc);
217 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
218 			     u64 ram_bytes, u64 num_bytes, int delalloc);
219 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
220 			       u64 num_bytes, int delalloc);
221 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
222 		      enum btrfs_chunk_alloc_enum force);
223 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
224 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
225 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
226 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
227 int btrfs_free_block_groups(struct btrfs_fs_info *info);
228 
229 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
230 {
231 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
232 }
233 
234 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
235 {
236 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
237 }
238 
239 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
240 {
241 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
242 }
243 
244 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
245 {
246 	smp_mb();
247 	return cache->cached == BTRFS_CACHE_FINISHED ||
248 		cache->cached == BTRFS_CACHE_ERROR;
249 }
250 
251 #endif /* BTRFS_BLOCK_GROUP_H */
252