]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/btrfs/block-group.h
btrfs: fix a race between scrub and block group removal/allocation
[mirror_ubuntu-hirsute-kernel.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
2bee7eb8
DZ
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 */
22enum btrfs_discard_state {
23 BTRFS_DISCARD_EXTENTS,
24 BTRFS_DISCARD_BITMAPS,
25 BTRFS_DISCARD_RESET_CURSOR,
26};
27
07730d87
JB
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 */
39enum btrfs_chunk_alloc_enum {
40 CHUNK_ALLOC_NO_FORCE,
41 CHUNK_ALLOC_LIMITED,
42 CHUNK_ALLOC_FORCE,
43};
44
aac0023c
JB
45struct btrfs_caching_control {
46 struct list_head list;
47 struct mutex mutex;
48 wait_queue_head_t wait;
49 struct btrfs_work work;
32da5386 50 struct btrfs_block_group *block_group;
aac0023c
JB
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
32da5386 58struct btrfs_block_group {
aac0023c
JB
59 struct btrfs_fs_info *fs_info;
60 struct inode *inode;
61 spinlock_t lock;
b3470b5d
DS
62 u64 start;
63 u64 length;
aac0023c
JB
64 u64 pinned;
65 u64 reserved;
bf38be65 66 u64 used;
aac0023c
JB
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
b0643e59 132 /* For discard operations */
aac0023c 133 atomic_t trimming;
b0643e59
DZ
134 struct list_head discard_list;
135 int discard_index;
136 u64 discard_eligible_time;
2bee7eb8
DZ
137 u64 discard_cursor;
138 enum btrfs_discard_state discard_state;
aac0023c
JB
139
140 /* For dirty block groups */
141 struct list_head dirty_list;
142 struct list_head io_list;
143
144 struct btrfs_io_ctl io_ctl;
145
146 /*
147 * Incremented when doing extent allocations and holding a read lock
148 * on the space_info's groups_sem semaphore.
149 * Decremented when an ordered extent that represents an IO against this
150 * block group's range is created (after it's added to its inode's
151 * root's list of ordered extents) or immediately after the allocation
152 * if it's a metadata extent or fallocate extent (for these cases we
153 * don't create ordered extents).
154 */
155 atomic_t reservations;
156
157 /*
158 * Incremented while holding the spinlock *lock* by a task checking if
159 * it can perform a nocow write (incremented if the value for the *ro*
160 * field is 0). Decremented by such tasks once they create an ordered
161 * extent or before that if some error happens before reaching that step.
162 * This is to prevent races between block group relocation and nocow
163 * writes through direct IO.
164 */
165 atomic_t nocow_writers;
166
167 /* Lock for free space tree operations. */
168 struct mutex free_space_lock;
169
170 /*
171 * Does the block group need to be added to the free space tree?
172 * Protected by free_space_lock.
173 */
174 int needs_free_space;
175
176 /* Record locked full stripes for RAID5/6 block group */
177 struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
178};
179
b0643e59
DZ
180static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
181{
182 return (block_group->start + block_group->length);
183}
184
5cb0724e
DZ
185static inline bool btrfs_is_block_group_data_only(
186 struct btrfs_block_group *block_group)
187{
188 /*
189 * In mixed mode the fragmentation is expected to be high, lowering the
190 * efficiency, so only proper data block groups are considered.
191 */
192 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
193 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
194}
195
aac0023c
JB
196#ifdef CONFIG_BTRFS_DEBUG
197static inline int btrfs_should_fragment_free_space(
32da5386 198 struct btrfs_block_group *block_group)
aac0023c
JB
199{
200 struct btrfs_fs_info *fs_info = block_group->fs_info;
201
202 return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
203 block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
204 (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
205 block_group->flags & BTRFS_BLOCK_GROUP_DATA);
206}
207#endif
208
32da5386 209struct btrfs_block_group *btrfs_lookup_first_block_group(
2e405ad8 210 struct btrfs_fs_info *info, u64 bytenr);
32da5386 211struct btrfs_block_group *btrfs_lookup_block_group(
2e405ad8 212 struct btrfs_fs_info *info, u64 bytenr);
32da5386
DS
213struct btrfs_block_group *btrfs_next_block_group(
214 struct btrfs_block_group *cache);
215void btrfs_get_block_group(struct btrfs_block_group *cache);
216void btrfs_put_block_group(struct btrfs_block_group *cache);
3eeb3226
JB
217void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
218 const u64 start);
32da5386 219void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
3eeb3226
JB
220bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
221void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
32da5386
DS
222void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
223void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
676f1f75 224 u64 num_bytes);
32da5386
DS
225int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
226int btrfs_cache_block_group(struct btrfs_block_group *cache,
676f1f75 227 int load_cache_only);
e3cb339f
JB
228void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
229struct btrfs_caching_control *btrfs_get_caching_control(
32da5386
DS
230 struct btrfs_block_group *cache);
231u64 add_new_free_space(struct btrfs_block_group *block_group,
9f21246d 232 u64 start, u64 end);
e3e0520b
JB
233struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
234 struct btrfs_fs_info *fs_info,
235 const u64 chunk_offset);
236int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
237 u64 group_start, struct extent_map *em);
238void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
32da5386 239void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
4358d963
JB
240int btrfs_read_block_groups(struct btrfs_fs_info *info);
241int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
242 u64 type, u64 chunk_offset, u64 size);
243void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
b12de528
QW
244int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
245 bool do_chunk_alloc);
32da5386 246void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
77745c05
JB
247int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
248int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
249int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
ade4b516
JB
250int btrfs_update_block_group(struct btrfs_trans_handle *trans,
251 u64 bytenr, u64 num_bytes, int alloc);
32da5386 252int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
ade4b516 253 u64 ram_bytes, u64 num_bytes, int delalloc);
32da5386 254void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
ade4b516 255 u64 num_bytes, int delalloc);
07730d87
JB
256int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
257 enum btrfs_chunk_alloc_enum force);
258int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
259void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
878d7b67 260u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
3e43c279
JB
261void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
262int btrfs_free_block_groups(struct btrfs_fs_info *info);
878d7b67
JB
263
264static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
265{
266 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
267}
268
269static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
270{
271 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
272}
273
274static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
275{
276 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
277}
676f1f75 278
32da5386 279static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
676f1f75
JB
280{
281 smp_mb();
282 return cache->cached == BTRFS_CACHE_FINISHED ||
283 cache->cached == BTRFS_CACHE_ERROR;
284}
2e405ad8 285
96a14336
NB
286#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
287int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
288 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
289#endif
290
aac0023c 291#endif /* BTRFS_BLOCK_GROUP_H */