blob: 29678426247deb52a687232f93e574c8aaaee9b5 [file] [log] [blame]
Josef Bacikaac00232019-06-20 15:37:44 -04001/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_BLOCK_GROUP_H
4#define BTRFS_BLOCK_GROUP_H
5
David Sterba67b61ae2019-08-21 19:57:04 +02006#include "free-space-cache.h"
7
Josef Bacikaac00232019-06-20 15:37:44 -04008enum btrfs_disk_cache_state {
9 BTRFS_DC_WRITTEN,
10 BTRFS_DC_ERROR,
11 BTRFS_DC_CLEAR,
12 BTRFS_DC_SETUP,
13};
14
Josef Bacik07730d82019-06-20 15:38:04 -040015/*
Dennis Zhou2bee7eb2019-12-13 16:22:16 -080016 * 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
28/*
Josef Bacik07730d82019-06-20 15:38:04 -040029 * 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
Josef Bacikaac00232019-06-20 15:37:44 -040045struct btrfs_caching_control {
46 struct list_head list;
47 struct mutex mutex;
48 wait_queue_head_t wait;
49 struct btrfs_work work;
David Sterba32da53862019-10-29 19:20:18 +010050 struct btrfs_block_group *block_group;
Josef Bacikaac00232019-06-20 15:37:44 -040051 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
David Sterba32da53862019-10-29 19:20:18 +010058struct btrfs_block_group {
Josef Bacikaac00232019-06-20 15:37:44 -040059 struct btrfs_fs_info *fs_info;
60 struct inode *inode;
61 spinlock_t lock;
David Sterbab3470b52019-10-23 18:48:22 +020062 u64 start;
63 u64 length;
Josef Bacikaac00232019-06-20 15:37:44 -040064 u64 pinned;
65 u64 reserved;
David Sterbabf38be62019-10-23 18:48:11 +020066 u64 used;
Josef Bacikaac00232019-06-20 15:37:44 -040067 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;
Naohiro Aota78ce9fc2021-02-04 19:22:11 +090098 unsigned int to_copy:1;
Naohiro Aotaf7ef5282021-02-04 19:22:16 +090099 unsigned int relocating_repair:1;
Josef Bacikaac00232019-06-20 15:37:44 -0400100
101 int disk_cache_state;
102
103 /* Cache tracking stuff */
104 int cached;
105 struct btrfs_caching_control *caching_ctl;
106 u64 last_byte_to_unpin;
107
108 struct btrfs_space_info *space_info;
109
110 /* Free space cache stuff */
111 struct btrfs_free_space_ctl *free_space_ctl;
112
113 /* Block group cache stuff */
114 struct rb_node cache_node;
115
116 /* For block groups in the same raid type */
117 struct list_head list;
118
Josef Bacik48aaeeb2020-07-06 09:14:11 -0400119 refcount_t refs;
Josef Bacikaac00232019-06-20 15:37:44 -0400120
121 /*
122 * List of struct btrfs_free_clusters for this block group.
123 * Today it will only have one thing on it, but that may change
124 */
125 struct list_head cluster_list;
126
127 /* For delayed block group creation or deletion of empty block groups */
128 struct list_head bg_list;
129
130 /* For read-only block groups */
131 struct list_head ro_list;
132
Filipe Manana6b7304a2020-05-08 11:01:47 +0100133 /*
134 * When non-zero it means the block group's logical address and its
135 * device extents can not be reused for future block group allocations
136 * until the counter goes down to 0. This is to prevent them from being
137 * reused while some task is still using the block group after it was
138 * deleted - we want to make sure they can only be reused for new block
139 * groups after that task is done with the deleted block group.
140 */
141 atomic_t frozen;
142
Dennis Zhoub0643e52019-12-13 16:22:14 -0800143 /* For discard operations */
Dennis Zhoub0643e52019-12-13 16:22:14 -0800144 struct list_head discard_list;
145 int discard_index;
146 u64 discard_eligible_time;
Dennis Zhou2bee7eb2019-12-13 16:22:16 -0800147 u64 discard_cursor;
148 enum btrfs_discard_state discard_state;
Josef Bacikaac00232019-06-20 15:37:44 -0400149
150 /* For dirty block groups */
151 struct list_head dirty_list;
152 struct list_head io_list;
153
154 struct btrfs_io_ctl io_ctl;
155
156 /*
157 * Incremented when doing extent allocations and holding a read lock
158 * on the space_info's groups_sem semaphore.
159 * Decremented when an ordered extent that represents an IO against this
160 * block group's range is created (after it's added to its inode's
161 * root's list of ordered extents) or immediately after the allocation
162 * if it's a metadata extent or fallocate extent (for these cases we
163 * don't create ordered extents).
164 */
165 atomic_t reservations;
166
167 /*
168 * Incremented while holding the spinlock *lock* by a task checking if
169 * it can perform a nocow write (incremented if the value for the *ro*
170 * field is 0). Decremented by such tasks once they create an ordered
171 * extent or before that if some error happens before reaching that step.
172 * This is to prevent races between block group relocation and nocow
173 * writes through direct IO.
174 */
175 atomic_t nocow_writers;
176
177 /* Lock for free space tree operations. */
178 struct mutex free_space_lock;
179
180 /*
181 * Does the block group need to be added to the free space tree?
182 * Protected by free_space_lock.
183 */
184 int needs_free_space;
185
Johannes Thumshirn08f455592021-02-04 19:22:03 +0900186 /* Flag indicating this block group is placed on a sequential zone */
187 bool seq_zone;
188
Josef Bacikaac00232019-06-20 15:37:44 -0400189 /* Record locked full stripes for RAID5/6 block group */
190 struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
Naohiro Aota08e11a32021-02-04 19:21:50 +0900191
192 /*
193 * Allocation offset for the block group to implement sequential
194 * allocation. This is used only on a zoned filesystem.
195 */
196 u64 alloc_offset;
Naohiro Aota169e0da2021-02-04 19:21:52 +0900197 u64 zone_unusable;
Naohiro Aota0bc09ca2021-02-04 19:22:08 +0900198 u64 meta_write_pointer;
Josef Bacikaac00232019-06-20 15:37:44 -0400199};
200
Dennis Zhoub0643e52019-12-13 16:22:14 -0800201static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
202{
203 return (block_group->start + block_group->length);
204}
205
Dennis Zhou5cb07242020-01-02 16:26:40 -0500206static inline bool btrfs_is_block_group_data_only(
207 struct btrfs_block_group *block_group)
208{
209 /*
210 * In mixed mode the fragmentation is expected to be high, lowering the
211 * efficiency, so only proper data block groups are considered.
212 */
213 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
214 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
215}
216
Josef Bacikaac00232019-06-20 15:37:44 -0400217#ifdef CONFIG_BTRFS_DEBUG
218static inline int btrfs_should_fragment_free_space(
David Sterba32da53862019-10-29 19:20:18 +0100219 struct btrfs_block_group *block_group)
Josef Bacikaac00232019-06-20 15:37:44 -0400220{
221 struct btrfs_fs_info *fs_info = block_group->fs_info;
222
223 return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
224 block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
225 (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
226 block_group->flags & BTRFS_BLOCK_GROUP_DATA);
227}
228#endif
229
David Sterba32da53862019-10-29 19:20:18 +0100230struct btrfs_block_group *btrfs_lookup_first_block_group(
Josef Bacik2e405ad2019-06-20 15:37:45 -0400231 struct btrfs_fs_info *info, u64 bytenr);
David Sterba32da53862019-10-29 19:20:18 +0100232struct btrfs_block_group *btrfs_lookup_block_group(
Josef Bacik2e405ad2019-06-20 15:37:45 -0400233 struct btrfs_fs_info *info, u64 bytenr);
David Sterba32da53862019-10-29 19:20:18 +0100234struct btrfs_block_group *btrfs_next_block_group(
235 struct btrfs_block_group *cache);
236void btrfs_get_block_group(struct btrfs_block_group *cache);
237void btrfs_put_block_group(struct btrfs_block_group *cache);
Josef Bacik3eeb3222019-06-20 15:37:47 -0400238void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
239 const u64 start);
David Sterba32da53862019-10-29 19:20:18 +0100240void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
Josef Bacik3eeb3222019-06-20 15:37:47 -0400241bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
242void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
David Sterba32da53862019-10-29 19:20:18 +0100243void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
244void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
Josef Bacik676f1f72019-06-20 15:37:48 -0400245 u64 num_bytes);
David Sterba32da53862019-10-29 19:20:18 +0100246int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
247int btrfs_cache_block_group(struct btrfs_block_group *cache,
Josef Bacik676f1f72019-06-20 15:37:48 -0400248 int load_cache_only);
Josef Bacike3cb3392019-06-20 15:37:50 -0400249void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
250struct btrfs_caching_control *btrfs_get_caching_control(
David Sterba32da53862019-10-29 19:20:18 +0100251 struct btrfs_block_group *cache);
252u64 add_new_free_space(struct btrfs_block_group *block_group,
Josef Bacik9f212462019-08-06 16:43:19 +0200253 u64 start, u64 end);
Josef Bacike3e05202019-06-20 15:37:55 -0400254struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
255 struct btrfs_fs_info *fs_info,
256 const u64 chunk_offset);
257int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
258 u64 group_start, struct extent_map *em);
259void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
David Sterba32da53862019-10-29 19:20:18 +0100260void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
Josef Bacik4358d9632019-06-20 15:37:57 -0400261int btrfs_read_block_groups(struct btrfs_fs_info *info);
262int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used,
263 u64 type, u64 chunk_offset, u64 size);
264void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
Qu Wenruob12de522019-11-15 10:09:00 +0800265int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
266 bool do_chunk_alloc);
David Sterba32da53862019-10-29 19:20:18 +0100267void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
Josef Bacik77745c02019-06-20 15:38:00 -0400268int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
269int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
270int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
Josef Bacikade4b512019-06-20 15:38:01 -0400271int btrfs_update_block_group(struct btrfs_trans_handle *trans,
272 u64 bytenr, u64 num_bytes, int alloc);
David Sterba32da53862019-10-29 19:20:18 +0100273int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
Josef Bacikade4b512019-06-20 15:38:01 -0400274 u64 ram_bytes, u64 num_bytes, int delalloc);
David Sterba32da53862019-10-29 19:20:18 +0100275void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
Josef Bacikade4b512019-06-20 15:38:01 -0400276 u64 num_bytes, int delalloc);
Josef Bacik07730d82019-06-20 15:38:04 -0400277int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
278 enum btrfs_chunk_alloc_enum force);
279int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
280void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
Josef Bacik878d7b62019-06-20 15:38:05 -0400281u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
Josef Bacik3e43c272019-06-20 15:38:06 -0400282void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
283int btrfs_free_block_groups(struct btrfs_fs_info *info);
Josef Bacike7478532020-10-23 09:58:10 -0400284void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,
285 struct btrfs_caching_control *caching_ctl);
Naohiro Aota138082f2021-02-04 19:22:02 +0900286int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
287 struct block_device *bdev, u64 physical, u64 **logical,
288 int *naddrs, int *stripe_len);
Josef Bacik878d7b62019-06-20 15:38:05 -0400289
290static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
291{
292 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
293}
294
295static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
296{
297 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
298}
299
300static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
301{
302 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
303}
Josef Bacik676f1f72019-06-20 15:37:48 -0400304
David Sterba32da53862019-10-29 19:20:18 +0100305static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
Josef Bacik676f1f72019-06-20 15:37:48 -0400306{
307 smp_mb();
308 return cache->cached == BTRFS_CACHE_FINISHED ||
309 cache->cached == BTRFS_CACHE_ERROR;
310}
Josef Bacik2e405ad2019-06-20 15:37:45 -0400311
Filipe Manana684b7522020-05-08 11:01:59 +0100312void btrfs_freeze_block_group(struct btrfs_block_group *cache);
313void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
314
Josef Bacikaac00232019-06-20 15:37:44 -0400315#endif /* BTRFS_BLOCK_GROUP_H */