Commit | Line | Data |
---|---|---|
9888c340 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
9888c340 DS |
6 | #ifndef BTRFS_TRANSACTION_H |
7 | #define BTRFS_TRANSACTION_H | |
9b64f57d | 8 | |
602035d7 | 9 | #include <linux/atomic.h> |
9b64f57d | 10 | #include <linux/refcount.h> |
602035d7 DS |
11 | #include <linux/list.h> |
12 | #include <linux/time64.h> | |
13 | #include <linux/mutex.h> | |
14 | #include <linux/wait.h> | |
31f3c99b | 15 | #include "btrfs_inode.h" |
56bec294 | 16 | #include "delayed-ref.h" |
602035d7 DS |
17 | #include "extent-io-tree.h" |
18 | #include "block-rsv.h" | |
19 | #include "messages.h" | |
cc37ea61 | 20 | #include "misc.h" |
e089f05c | 21 | |
602035d7 DS |
22 | struct dentry; |
23 | struct inode; | |
24 | struct btrfs_pending_snapshot; | |
25 | struct btrfs_fs_info; | |
26 | struct btrfs_root_item; | |
27 | struct btrfs_root; | |
28 | struct btrfs_path; | |
29 | ||
b321a52c BB |
30 | /* Radix-tree tag for roots that are part of the trasaction. */ |
31 | #define BTRFS_ROOT_TRANS_TAG 0 | |
32 | ||
4a9d8bde | 33 | enum btrfs_trans_state { |
bbe339cc | 34 | TRANS_STATE_RUNNING, |
77d20c68 | 35 | TRANS_STATE_COMMIT_PREP, |
bbe339cc DS |
36 | TRANS_STATE_COMMIT_START, |
37 | TRANS_STATE_COMMIT_DOING, | |
38 | TRANS_STATE_UNBLOCKED, | |
d0c2f4fa | 39 | TRANS_STATE_SUPER_COMMITTED, |
bbe339cc DS |
40 | TRANS_STATE_COMPLETED, |
41 | TRANS_STATE_MAX, | |
4a9d8bde MX |
42 | }; |
43 | ||
3204d33c JB |
44 | #define BTRFS_TRANS_HAVE_FREE_BGS 0 |
45 | #define BTRFS_TRANS_DIRTY_BG_RUN 1 | |
2968b1f4 | 46 | #define BTRFS_TRANS_CACHE_ENOSPC 2 |
3204d33c | 47 | |
79154b1b CM |
48 | struct btrfs_transaction { |
49 | u64 transid; | |
0860adfd MX |
50 | /* |
51 | * total external writers(USERSPACE/START/ATTACH) in this | |
52 | * transaction, it must be zero before the transaction is | |
53 | * being committed | |
54 | */ | |
55 | atomic_t num_extwriters; | |
b7ec40d7 CM |
56 | /* |
57 | * total writers in this transaction, it must be zero before the | |
58 | * transaction can end | |
59 | */ | |
13c5a93e | 60 | atomic_t num_writers; |
9b64f57d | 61 | refcount_t use_count; |
b7ec40d7 | 62 | |
3204d33c | 63 | unsigned long flags; |
13212b54 | 64 | |
4a9d8bde MX |
65 | /* Be protected by fs_info->trans_lock when we want to change it. */ |
66 | enum btrfs_trans_state state; | |
5302e089 | 67 | int aborted; |
8fd17795 | 68 | struct list_head list; |
d1310b2e | 69 | struct extent_io_tree dirty_pages; |
a944442c | 70 | time64_t start_time; |
79154b1b CM |
71 | wait_queue_head_t writer_wait; |
72 | wait_queue_head_t commit_wait; | |
3063d29f | 73 | struct list_head pending_snapshots; |
bbbf7243 | 74 | struct list_head dev_update_list; |
9e351cc8 | 75 | struct list_head switch_commits; |
ce93ec54 | 76 | struct list_head dirty_bgs; |
45ae2c18 NB |
77 | |
78 | /* | |
79 | * There is no explicit lock which protects io_bgs, rather its | |
80 | * consistency is implied by the fact that all the sites which modify | |
81 | * it do so under some form of transaction critical section, namely: | |
82 | * | |
83 | * - btrfs_start_dirty_block_groups - This function can only ever be | |
84 | * run by one of the transaction committers. Refer to | |
85 | * BTRFS_TRANS_DIRTY_BG_RUN usage in btrfs_commit_transaction | |
86 | * | |
87 | * - btrfs_write_dirty_blockgroups - this is called by | |
88 | * commit_cowonly_roots from transaction critical section | |
89 | * (TRANS_STATE_COMMIT_DOING) | |
90 | * | |
91 | * - btrfs_cleanup_dirty_bgs - called on transaction abort | |
92 | */ | |
1bbc621e | 93 | struct list_head io_bgs; |
2b9dbef2 | 94 | struct list_head dropped_roots; |
fe119a6e | 95 | struct extent_io_tree pinned_extents; |
1bbc621e CM |
96 | |
97 | /* | |
98 | * we need to make sure block group deletion doesn't race with | |
99 | * free space cache writeout. This mutex keeps them from stomping | |
100 | * on each other | |
101 | */ | |
102 | struct mutex cache_write_mutex; | |
ce93ec54 | 103 | spinlock_t dirty_bgs_lock; |
348a0013 | 104 | /* Protected by spin lock fs_info->unused_bgs_lock. */ |
e33e17ee | 105 | struct list_head deleted_bgs; |
2b9dbef2 | 106 | spinlock_t dropped_roots_lock; |
56bec294 | 107 | struct btrfs_delayed_ref_root delayed_refs; |
ab8d0fc4 | 108 | struct btrfs_fs_info *fs_info; |
48778179 FM |
109 | |
110 | /* | |
111 | * Number of ordered extents the transaction must wait for before | |
112 | * committing. These are ordered extents started by a fast fsync. | |
113 | */ | |
114 | atomic_t pending_ordered; | |
115 | wait_queue_head_t pending_wait; | |
79154b1b CM |
116 | }; |
117 | ||
cc37ea61 DS |
118 | enum { |
119 | ENUM_BIT(__TRANS_FREEZABLE), | |
120 | ENUM_BIT(__TRANS_START), | |
121 | ENUM_BIT(__TRANS_ATTACH), | |
122 | ENUM_BIT(__TRANS_JOIN), | |
123 | ENUM_BIT(__TRANS_JOIN_NOLOCK), | |
124 | ENUM_BIT(__TRANS_DUMMY), | |
125 | ENUM_BIT(__TRANS_JOIN_NOSTART), | |
126 | }; | |
0860adfd | 127 | |
0860adfd MX |
128 | #define TRANS_START (__TRANS_START | __TRANS_FREEZABLE) |
129 | #define TRANS_ATTACH (__TRANS_ATTACH) | |
130 | #define TRANS_JOIN (__TRANS_JOIN | __TRANS_FREEZABLE) | |
131 | #define TRANS_JOIN_NOLOCK (__TRANS_JOIN_NOLOCK) | |
a6d155d2 | 132 | #define TRANS_JOIN_NOSTART (__TRANS_JOIN_NOSTART) |
0860adfd | 133 | |
bcf3a3e7 | 134 | #define TRANS_EXTWRITERS (__TRANS_START | __TRANS_ATTACH) |
a698d075 | 135 | |
e089f05c CM |
136 | struct btrfs_trans_handle { |
137 | u64 transid; | |
f0486c68 | 138 | u64 bytes_reserved; |
28270e25 | 139 | u64 delayed_refs_bytes_reserved; |
4fbcdf66 | 140 | u64 chunk_bytes_reserved; |
56bec294 | 141 | unsigned long delayed_ref_updates; |
adb86dbe | 142 | unsigned long delayed_ref_csum_deletions; |
f0486c68 YZ |
143 | struct btrfs_transaction *transaction; |
144 | struct btrfs_block_rsv *block_rsv; | |
2a1eb461 | 145 | struct btrfs_block_rsv *orig_rsv; |
28b21c55 FM |
146 | /* Set by a task that wants to create a snapshot. */ |
147 | struct btrfs_pending_snapshot *pending_snapshot; | |
1ca4bb63 DS |
148 | refcount_t use_count; |
149 | unsigned int type; | |
bf31f87f DS |
150 | /* |
151 | * Error code of transaction abort, set outside of locks and must use | |
152 | * the READ_ONCE/WRITE_ONCE access | |
153 | */ | |
a698d075 | 154 | short aborted; |
7c2871a2 | 155 | bool adding_csums; |
c6b305a8 | 156 | bool allocating_chunk; |
79bd3712 | 157 | bool removing_chunk; |
20dd2cbf | 158 | bool reloc_reserved; |
d0c2f4fa | 159 | bool in_fsync; |
64b63580 | 160 | struct btrfs_fs_info *fs_info; |
ea658bad | 161 | struct list_head new_bgs; |
28270e25 | 162 | struct btrfs_block_rsv delayed_rsv; |
e089f05c CM |
163 | }; |
164 | ||
bf31f87f DS |
165 | /* |
166 | * The abort status can be changed between calls and is not protected by locks. | |
167 | * This accepts btrfs_transaction and btrfs_trans_handle as types. Once it's | |
168 | * set to a non-zero value it does not change, so the macro should be in checks | |
169 | * but is not necessary for further reads of the value. | |
170 | */ | |
171 | #define TRANS_ABORTED(trans) (unlikely(READ_ONCE((trans)->aborted))) | |
172 | ||
3063d29f | 173 | struct btrfs_pending_snapshot { |
3de4586c | 174 | struct dentry *dentry; |
e9662f70 | 175 | struct inode *dir; |
3063d29f | 176 | struct btrfs_root *root; |
b0c0ea63 | 177 | struct btrfs_root_item *root_item; |
a22285a6 | 178 | struct btrfs_root *snap; |
6f72c7e2 | 179 | struct btrfs_qgroup_inherit *inherit; |
8546b570 | 180 | struct btrfs_path *path; |
a22285a6 YZ |
181 | /* block reservation for the operation */ |
182 | struct btrfs_block_rsv block_rsv; | |
01327610 | 183 | /* extra metadata reservation for relocation */ |
a22285a6 | 184 | int error; |
2dfb1e43 QW |
185 | /* Preallocated anonymous block device number */ |
186 | dev_t anon_dev; | |
b83cc969 | 187 | bool readonly; |
3063d29f CM |
188 | struct list_head list; |
189 | }; | |
190 | ||
15ee9bc7 | 191 | static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans, |
d9094414 | 192 | struct btrfs_inode *inode) |
15ee9bc7 | 193 | { |
d9094414 NB |
194 | spin_lock(&inode->lock); |
195 | inode->last_trans = trans->transaction->transid; | |
6008859b | 196 | inode->last_sub_trans = btrfs_get_root_log_transid(inode->root); |
bc0939fc | 197 | inode->last_log_commit = inode->last_sub_trans - 1; |
d9094414 | 198 | spin_unlock(&inode->lock); |
15ee9bc7 JB |
199 | } |
200 | ||
9086db86 QW |
201 | /* |
202 | * Make qgroup codes to skip given qgroupid, means the old/new_roots for | |
203 | * qgroup won't contain the qgroupid in it. | |
204 | */ | |
205 | static inline void btrfs_set_skip_qgroup(struct btrfs_trans_handle *trans, | |
206 | u64 qgroupid) | |
207 | { | |
208 | struct btrfs_delayed_ref_root *delayed_refs; | |
209 | ||
210 | delayed_refs = &trans->transaction->delayed_refs; | |
211 | WARN_ON(delayed_refs->qgroup_to_skip); | |
212 | delayed_refs->qgroup_to_skip = qgroupid; | |
213 | } | |
214 | ||
215 | static inline void btrfs_clear_skip_qgroup(struct btrfs_trans_handle *trans) | |
216 | { | |
217 | struct btrfs_delayed_ref_root *delayed_refs; | |
218 | ||
219 | delayed_refs = &trans->transaction->delayed_refs; | |
220 | WARN_ON(!delayed_refs->qgroup_to_skip); | |
221 | delayed_refs->qgroup_to_skip = 0; | |
222 | } | |
223 | ||
ed164802 | 224 | bool __cold abort_should_print_stack(int error); |
fccf0c84 JB |
225 | |
226 | /* | |
227 | * Call btrfs_abort_transaction as early as possible when an error condition is | |
228 | * detected, that way the exact stack trace is reported for some errors. | |
229 | */ | |
ed164802 | 230 | #define btrfs_abort_transaction(trans, error) \ |
fccf0c84 JB |
231 | do { \ |
232 | bool first = false; \ | |
233 | /* Report first abort since mount */ \ | |
234 | if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED, \ | |
235 | &((trans)->fs_info->fs_state))) { \ | |
236 | first = true; \ | |
ed164802 | 237 | if (WARN(abort_should_print_stack(error), \ |
fccf0c84 JB |
238 | KERN_ERR \ |
239 | "BTRFS: Transaction aborted (error %d)\n", \ | |
ed164802 | 240 | (error))) { \ |
fccf0c84 JB |
241 | /* Stack trace printed. */ \ |
242 | } else { \ | |
f8d1b011 FM |
243 | btrfs_err((trans)->fs_info, \ |
244 | "Transaction aborted (error %d)", \ | |
ed164802 | 245 | (error)); \ |
fccf0c84 JB |
246 | } \ |
247 | } \ | |
248 | __btrfs_abort_transaction((trans), __func__, \ | |
ed164802 | 249 | __LINE__, (error), first); \ |
fccf0c84 JB |
250 | } while (0) |
251 | ||
3a45bb20 | 252 | int btrfs_end_transaction(struct btrfs_trans_handle *trans); |
79154b1b | 253 | struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, |
5aed1dd8 | 254 | unsigned int num_items); |
8eab77ff FM |
255 | struct btrfs_trans_handle *btrfs_start_transaction_fallback_global_rsv( |
256 | struct btrfs_root *root, | |
7f9fe614 | 257 | unsigned int num_items); |
7a7eaa40 | 258 | struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root); |
8d510121 | 259 | struct btrfs_trans_handle *btrfs_join_transaction_spacecache(struct btrfs_root *root); |
a6d155d2 | 260 | struct btrfs_trans_handle *btrfs_join_transaction_nostart(struct btrfs_root *root); |
354aa0fb | 261 | struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root); |
d4edf39b MX |
262 | struct btrfs_trans_handle *btrfs_attach_transaction_barrier( |
263 | struct btrfs_root *root); | |
2ff7e61e | 264 | int btrfs_wait_for_commit(struct btrfs_fs_info *fs_info, u64 transid); |
08607c1b | 265 | |
cfad392b | 266 | void btrfs_add_dead_root(struct btrfs_root *root); |
b4be6aef | 267 | void btrfs_maybe_wake_unfinished_drop(struct btrfs_fs_info *fs_info); |
33c44184 | 268 | int btrfs_clean_one_deleted_snapshot(struct btrfs_fs_info *fs_info); |
3a45bb20 | 269 | int btrfs_commit_transaction(struct btrfs_trans_handle *trans); |
fdfbf020 | 270 | void btrfs_commit_transaction_async(struct btrfs_trans_handle *trans); |
3a45bb20 | 271 | int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans); |
a2633b6a | 272 | bool btrfs_should_end_transaction(struct btrfs_trans_handle *trans); |
2ff7e61e | 273 | void btrfs_throttle(struct btrfs_fs_info *fs_info); |
5d4f98a2 YZ |
274 | int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans, |
275 | struct btrfs_root *root); | |
2ff7e61e | 276 | int btrfs_write_marked_extents(struct btrfs_fs_info *fs_info, |
8cef4e16 | 277 | struct extent_io_tree *dirty_pages, int mark); |
bf89d38f | 278 | int btrfs_wait_tree_log_extents(struct btrfs_root *root, int mark); |
8929ecfa | 279 | int btrfs_transaction_blocked(struct btrfs_fs_info *info); |
724e2315 | 280 | void btrfs_put_transaction(struct btrfs_transaction *transaction); |
2b9dbef2 JB |
281 | void btrfs_add_dropped_root(struct btrfs_trans_handle *trans, |
282 | struct btrfs_root *root); | |
fb6dea26 | 283 | void btrfs_trans_release_chunk_metadata(struct btrfs_trans_handle *trans); |
fccf0c84 JB |
284 | void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans, |
285 | const char *function, | |
ed164802 | 286 | unsigned int line, int error, bool first_hit); |
9888c340 | 287 | |
956504a3 JB |
288 | int __init btrfs_transaction_init(void); |
289 | void __cold btrfs_transaction_exit(void); | |
290 | ||
e089f05c | 291 | #endif |