Merge tag 'trace-v6.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace...
[linux-block.git] / fs / btrfs / block-group.h
CommitLineData
aac0023c
JB
1/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_BLOCK_GROUP_H
4#define BTRFS_BLOCK_GROUP_H
5
67b61aef
DS
6#include "free-space-cache.h"
7
aac0023c
JB
8enum btrfs_disk_cache_state {
9 BTRFS_DC_WRITTEN,
10 BTRFS_DC_ERROR,
11 BTRFS_DC_CLEAR,
12 BTRFS_DC_SETUP,
13};
14
52bb7a21
BB
15enum btrfs_block_group_size_class {
16 /* Unset */
17 BTRFS_BG_SZ_NONE,
18 /* 0 < size <= 128K */
19 BTRFS_BG_SZ_SMALL,
20 /* 128K < size <= 8M */
21 BTRFS_BG_SZ_MEDIUM,
22 /* 8M < size < BG_LENGTH */
23 BTRFS_BG_SZ_LARGE,
24};
25
2bee7eb8
DZ
26/*
27 * This describes the state of the block_group for async discard. This is due
28 * to the two pass nature of it where extent discarding is prioritized over
29 * bitmap discarding. BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
30 * between lists to prevent contention for discard state variables
31 * (eg. discard_cursor).
32 */
33enum btrfs_discard_state {
34 BTRFS_DISCARD_EXTENTS,
35 BTRFS_DISCARD_BITMAPS,
36 BTRFS_DISCARD_RESET_CURSOR,
37};
38
07730d87
JB
39/*
40 * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
41 * only allocate a chunk if we really need one.
42 *
43 * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
44 * chunks already allocated. This is used as part of the clustering code to
45 * help make sure we have a good pool of storage to cluster in, without filling
46 * the FS with empty chunks
47 *
48 * CHUNK_ALLOC_FORCE means it must try to allocate one
760e69c4
NA
49 *
50 * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
51 * find_free_extent() that also activaes the zone
07730d87
JB
52 */
53enum btrfs_chunk_alloc_enum {
54 CHUNK_ALLOC_NO_FORCE,
55 CHUNK_ALLOC_LIMITED,
56 CHUNK_ALLOC_FORCE,
760e69c4 57 CHUNK_ALLOC_FORCE_FOR_EXTENT,
07730d87
JB
58};
59
3349b57f
JB
60/* Block group flags set at runtime */
61enum btrfs_block_group_flags {
62 BLOCK_GROUP_FLAG_IREF,
3349b57f
JB
63 BLOCK_GROUP_FLAG_REMOVED,
64 BLOCK_GROUP_FLAG_TO_COPY,
65 BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
66 BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
67 BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
68 BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
0d7764ff
DS
69 /* Does the block group need to be added to the free space tree? */
70 BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
961f5b8b
DS
71 /* Indicate that the block group is placed on a sequential zone */
72 BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
3349b57f
JB
73};
74
16708a88
JB
75enum btrfs_caching_type {
76 BTRFS_CACHE_NO,
77 BTRFS_CACHE_STARTED,
78 BTRFS_CACHE_FINISHED,
79 BTRFS_CACHE_ERROR,
80};
81
aac0023c
JB
82struct btrfs_caching_control {
83 struct list_head list;
84 struct mutex mutex;
85 wait_queue_head_t wait;
86 struct btrfs_work work;
32da5386 87 struct btrfs_block_group *block_group;
aac0023c
JB
88 refcount_t count;
89};
90
91/* Once caching_thread() finds this much free space, it will wake up waiters. */
92#define CACHING_CTL_WAKE_UP SZ_2M
93
32da5386 94struct btrfs_block_group {
aac0023c
JB
95 struct btrfs_fs_info *fs_info;
96 struct inode *inode;
97 spinlock_t lock;
b3470b5d
DS
98 u64 start;
99 u64 length;
aac0023c
JB
100 u64 pinned;
101 u64 reserved;
bf38be65 102 u64 used;
aac0023c
JB
103 u64 delalloc_bytes;
104 u64 bytes_super;
105 u64 flags;
106 u64 cache_generation;
f7238e50 107 u64 global_root_id;
aac0023c 108
7248e0ce
QW
109 /*
110 * The last committed used bytes of this block group, if the above @used
111 * is still the same as @commit_used, we don't need to update block
112 * group item of this block group.
113 */
114 u64 commit_used;
aac0023c
JB
115 /*
116 * If the free space extent count exceeds this number, convert the block
117 * group to bitmaps.
118 */
119 u32 bitmap_high_thresh;
120
121 /*
122 * If the free space extent count drops below this number, convert the
123 * block group back to extents.
124 */
125 u32 bitmap_low_thresh;
126
127 /*
128 * It is just used for the delayed data space allocation because
129 * only the data space allocation and the relative metadata update
130 * can be done cross the transaction.
131 */
132 struct rw_semaphore data_rwsem;
133
134 /* For raid56, this is a full stripe, without parity */
135 unsigned long full_stripe_len;
3349b57f 136 unsigned long runtime_flags;
aac0023c
JB
137
138 unsigned int ro;
aac0023c
JB
139
140 int disk_cache_state;
141
142 /* Cache tracking stuff */
143 int cached;
144 struct btrfs_caching_control *caching_ctl;
aac0023c
JB
145
146 struct btrfs_space_info *space_info;
147
148 /* Free space cache stuff */
149 struct btrfs_free_space_ctl *free_space_ctl;
150
151 /* Block group cache stuff */
152 struct rb_node cache_node;
153
154 /* For block groups in the same raid type */
155 struct list_head list;
156
48aaeebe 157 refcount_t refs;
aac0023c
JB
158
159 /*
160 * List of struct btrfs_free_clusters for this block group.
161 * Today it will only have one thing on it, but that may change
162 */
163 struct list_head cluster_list;
164
165 /* For delayed block group creation or deletion of empty block groups */
166 struct list_head bg_list;
167
168 /* For read-only block groups */
169 struct list_head ro_list;
170
6b7304af
FM
171 /*
172 * When non-zero it means the block group's logical address and its
173 * device extents can not be reused for future block group allocations
174 * until the counter goes down to 0. This is to prevent them from being
175 * reused while some task is still using the block group after it was
176 * deleted - we want to make sure they can only be reused for new block
177 * groups after that task is done with the deleted block group.
178 */
179 atomic_t frozen;
180
b0643e59 181 /* For discard operations */
b0643e59
DZ
182 struct list_head discard_list;
183 int discard_index;
184 u64 discard_eligible_time;
2bee7eb8
DZ
185 u64 discard_cursor;
186 enum btrfs_discard_state discard_state;
aac0023c
JB
187
188 /* For dirty block groups */
189 struct list_head dirty_list;
190 struct list_head io_list;
191
192 struct btrfs_io_ctl io_ctl;
193
194 /*
195 * Incremented when doing extent allocations and holding a read lock
196 * on the space_info's groups_sem semaphore.
197 * Decremented when an ordered extent that represents an IO against this
198 * block group's range is created (after it's added to its inode's
199 * root's list of ordered extents) or immediately after the allocation
200 * if it's a metadata extent or fallocate extent (for these cases we
201 * don't create ordered extents).
202 */
203 atomic_t reservations;
204
205 /*
206 * Incremented while holding the spinlock *lock* by a task checking if
207 * it can perform a nocow write (incremented if the value for the *ro*
208 * field is 0). Decremented by such tasks once they create an ordered
209 * extent or before that if some error happens before reaching that step.
210 * This is to prevent races between block group relocation and nocow
211 * writes through direct IO.
212 */
213 atomic_t nocow_writers;
214
215 /* Lock for free space tree operations. */
216 struct mutex free_space_lock;
217
195a49ea
FM
218 /*
219 * Number of extents in this block group used for swap files.
220 * All accesses protected by the spinlock 'lock'.
221 */
222 int swap_extents;
223
08e11a3d
NA
224 /*
225 * Allocation offset for the block group to implement sequential
226 * allocation. This is used only on a zoned filesystem.
227 */
228 u64 alloc_offset;
169e0da9 229 u64 zone_unusable;
8eae532b 230 u64 zone_capacity;
0bc09ca1 231 u64 meta_write_pointer;
dafc340d 232 struct map_lookup *physical_map;
afba2bc0 233 struct list_head active_bg_list;
56fbb0a4
NA
234 struct work_struct zone_finish_work;
235 struct extent_buffer *last_eb;
52bb7a21 236 enum btrfs_block_group_size_class size_class;
aac0023c
JB
237};
238
b0643e59
DZ
239static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
240{
241 return (block_group->start + block_group->length);
242}
243
5cb0724e
DZ
244static inline bool btrfs_is_block_group_data_only(
245 struct btrfs_block_group *block_group)
246{
247 /*
248 * In mixed mode the fragmentation is expected to be high, lowering the
249 * efficiency, so only proper data block groups are considered.
250 */
251 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
252 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
253}
254
aac0023c 255#ifdef CONFIG_BTRFS_DEBUG
06d61cb1 256int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
aac0023c
JB
257#endif
258
32da5386 259struct btrfs_block_group *btrfs_lookup_first_block_group(
2e405ad8 260 struct btrfs_fs_info *info, u64 bytenr);
32da5386 261struct btrfs_block_group *btrfs_lookup_block_group(
2e405ad8 262 struct btrfs_fs_info *info, u64 bytenr);
32da5386
DS
263struct btrfs_block_group *btrfs_next_block_group(
264 struct btrfs_block_group *cache);
265void btrfs_get_block_group(struct btrfs_block_group *cache);
266void btrfs_put_block_group(struct btrfs_block_group *cache);
3eeb3226
JB
267void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
268 const u64 start);
32da5386 269void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
2306e83e
FM
270struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
271 u64 bytenr);
272void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
32da5386
DS
273void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
274void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
676f1f75 275 u64 num_bytes);
ced8ecf0 276int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
e3cb339f
JB
277void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
278struct btrfs_caching_control *btrfs_get_caching_control(
32da5386
DS
279 struct btrfs_block_group *cache);
280u64 add_new_free_space(struct btrfs_block_group *block_group,
9f21246d 281 u64 start, u64 end);
e3e0520b
JB
282struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
283 struct btrfs_fs_info *fs_info,
284 const u64 chunk_offset);
285int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
286 u64 group_start, struct extent_map *em);
287void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
32da5386 288void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
18bb8bbf
JT
289void btrfs_reclaim_bgs_work(struct work_struct *work);
290void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
291void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
4358d963 292int btrfs_read_block_groups(struct btrfs_fs_info *info);
79bd3712 293struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
5758d1bd 294 u64 type,
79bd3712 295 u64 chunk_offset, u64 size);
4358d963 296void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
b12de528
QW
297int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
298 bool do_chunk_alloc);
32da5386 299void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
77745c05
JB
300int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
301int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
302int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
ade4b516 303int btrfs_update_block_group(struct btrfs_trans_handle *trans,
11b66fa6 304 u64 bytenr, u64 num_bytes, bool alloc);
32da5386 305int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
52bb7a21
BB
306 u64 ram_bytes, u64 num_bytes, int delalloc,
307 bool force_wrong_size_class);
32da5386 308void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
ade4b516 309 u64 num_bytes, int delalloc);
07730d87
JB
310int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
311 enum btrfs_chunk_alloc_enum force);
312int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
313void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
2bb2e00e
FM
314void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
315 bool is_item_insertion);
878d7b67 316u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
3e43c279
JB
317void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
318int btrfs_free_block_groups(struct btrfs_fs_info *info);
138082f3 319int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
1eb82ef8 320 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
878d7b67
JB
321
322static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
323{
324 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
325}
326
327static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
328{
329 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
330}
331
332static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
333{
334 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
335}
676f1f75 336
32da5386 337static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
676f1f75
JB
338{
339 smp_mb();
340 return cache->cached == BTRFS_CACHE_FINISHED ||
341 cache->cached == BTRFS_CACHE_ERROR;
342}
2e405ad8 343
684b752b
FM
344void btrfs_freeze_block_group(struct btrfs_block_group *cache);
345void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
346
195a49ea
FM
347bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
348void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
349
52bb7a21
BB
350enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
351int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
352 enum btrfs_block_group_size_class size_class,
353 bool force_wrong_size_class);
cb0922f2 354bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
52bb7a21 355
aac0023c 356#endif /* BTRFS_BLOCK_GROUP_H */