Merge tag 'for-6.5/io_uring-2023-06-23' of git://git.kernel.dk/linux
[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
aadb164b
FM
165 /*
166 * Used for several lists:
167 *
168 * 1) struct btrfs_fs_info::unused_bgs
169 * 2) struct btrfs_fs_info::reclaim_bgs
170 * 3) struct btrfs_transaction::deleted_bgs
171 * 4) struct btrfs_trans_handle::new_bgs
172 */
aac0023c
JB
173 struct list_head bg_list;
174
175 /* For read-only block groups */
176 struct list_head ro_list;
177
6b7304af
FM
178 /*
179 * When non-zero it means the block group's logical address and its
180 * device extents can not be reused for future block group allocations
181 * until the counter goes down to 0. This is to prevent them from being
182 * reused while some task is still using the block group after it was
183 * deleted - we want to make sure they can only be reused for new block
184 * groups after that task is done with the deleted block group.
185 */
186 atomic_t frozen;
187
b0643e59 188 /* For discard operations */
b0643e59
DZ
189 struct list_head discard_list;
190 int discard_index;
191 u64 discard_eligible_time;
2bee7eb8
DZ
192 u64 discard_cursor;
193 enum btrfs_discard_state discard_state;
aac0023c
JB
194
195 /* For dirty block groups */
196 struct list_head dirty_list;
197 struct list_head io_list;
198
199 struct btrfs_io_ctl io_ctl;
200
201 /*
202 * Incremented when doing extent allocations and holding a read lock
203 * on the space_info's groups_sem semaphore.
204 * Decremented when an ordered extent that represents an IO against this
205 * block group's range is created (after it's added to its inode's
206 * root's list of ordered extents) or immediately after the allocation
207 * if it's a metadata extent or fallocate extent (for these cases we
208 * don't create ordered extents).
209 */
210 atomic_t reservations;
211
212 /*
213 * Incremented while holding the spinlock *lock* by a task checking if
214 * it can perform a nocow write (incremented if the value for the *ro*
215 * field is 0). Decremented by such tasks once they create an ordered
216 * extent or before that if some error happens before reaching that step.
217 * This is to prevent races between block group relocation and nocow
218 * writes through direct IO.
219 */
220 atomic_t nocow_writers;
221
222 /* Lock for free space tree operations. */
223 struct mutex free_space_lock;
224
195a49ea
FM
225 /*
226 * Number of extents in this block group used for swap files.
227 * All accesses protected by the spinlock 'lock'.
228 */
229 int swap_extents;
230
08e11a3d
NA
231 /*
232 * Allocation offset for the block group to implement sequential
233 * allocation. This is used only on a zoned filesystem.
234 */
235 u64 alloc_offset;
169e0da9 236 u64 zone_unusable;
8eae532b 237 u64 zone_capacity;
0bc09ca1 238 u64 meta_write_pointer;
dafc340d 239 struct map_lookup *physical_map;
afba2bc0 240 struct list_head active_bg_list;
56fbb0a4
NA
241 struct work_struct zone_finish_work;
242 struct extent_buffer *last_eb;
52bb7a21 243 enum btrfs_block_group_size_class size_class;
aac0023c
JB
244};
245
b0643e59
DZ
246static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
247{
248 return (block_group->start + block_group->length);
249}
250
5cb0724e
DZ
251static 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
aac0023c 262#ifdef CONFIG_BTRFS_DEBUG
06d61cb1 263int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
aac0023c
JB
264#endif
265
32da5386 266struct btrfs_block_group *btrfs_lookup_first_block_group(
2e405ad8 267 struct btrfs_fs_info *info, u64 bytenr);
32da5386 268struct btrfs_block_group *btrfs_lookup_block_group(
2e405ad8 269 struct btrfs_fs_info *info, u64 bytenr);
32da5386
DS
270struct btrfs_block_group *btrfs_next_block_group(
271 struct btrfs_block_group *cache);
272void btrfs_get_block_group(struct btrfs_block_group *cache);
273void btrfs_put_block_group(struct btrfs_block_group *cache);
3eeb3226
JB
274void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
275 const u64 start);
32da5386 276void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
2306e83e
FM
277struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
278 u64 bytenr);
279void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
32da5386
DS
280void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
281void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
676f1f75 282 u64 num_bytes);
ced8ecf0 283int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
e3cb339f
JB
284void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
285struct btrfs_caching_control *btrfs_get_caching_control(
32da5386
DS
286 struct btrfs_block_group *cache);
287u64 add_new_free_space(struct btrfs_block_group *block_group,
9f21246d 288 u64 start, u64 end);
e3e0520b
JB
289struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
290 struct btrfs_fs_info *fs_info,
291 const u64 chunk_offset);
292int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
293 u64 group_start, struct extent_map *em);
294void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
32da5386 295void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
18bb8bbf
JT
296void btrfs_reclaim_bgs_work(struct work_struct *work);
297void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
298void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
4358d963 299int btrfs_read_block_groups(struct btrfs_fs_info *info);
79bd3712 300struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
5758d1bd 301 u64 type,
79bd3712 302 u64 chunk_offset, u64 size);
4358d963 303void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
b12de528
QW
304int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
305 bool do_chunk_alloc);
32da5386 306void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
77745c05
JB
307int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
308int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
309int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
ade4b516 310int btrfs_update_block_group(struct btrfs_trans_handle *trans,
11b66fa6 311 u64 bytenr, u64 num_bytes, bool alloc);
32da5386 312int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
52bb7a21
BB
313 u64 ram_bytes, u64 num_bytes, int delalloc,
314 bool force_wrong_size_class);
32da5386 315void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
ade4b516 316 u64 num_bytes, int delalloc);
07730d87
JB
317int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
318 enum btrfs_chunk_alloc_enum force);
319int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
320void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
2bb2e00e
FM
321void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
322 bool is_item_insertion);
878d7b67 323u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
3e43c279
JB
324void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
325int btrfs_free_block_groups(struct btrfs_fs_info *info);
138082f3 326int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
1eb82ef8 327 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
878d7b67
JB
328
329static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
330{
331 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
332}
333
334static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
335{
336 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
337}
338
339static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
340{
341 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
342}
676f1f75 343
32da5386 344static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
676f1f75
JB
345{
346 smp_mb();
347 return cache->cached == BTRFS_CACHE_FINISHED ||
348 cache->cached == BTRFS_CACHE_ERROR;
349}
2e405ad8 350
684b752b
FM
351void btrfs_freeze_block_group(struct btrfs_block_group *cache);
352void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
353
195a49ea
FM
354bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
355void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
356
52bb7a21
BB
357enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
358int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
359 enum btrfs_block_group_size_class size_class,
360 bool force_wrong_size_class);
cb0922f2 361bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
52bb7a21 362
aac0023c 363#endif /* BTRFS_BLOCK_GROUP_H */