Loading...
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6#ifndef BTRFS_TRANSACTION_H
7#define BTRFS_TRANSACTION_H
8
9#include <linux/refcount.h>
10#include "btrfs_inode.h"
11#include "delayed-ref.h"
12#include "ctree.h"
13#include "misc.h"
14
15/* Radix-tree tag for roots that are part of the trasaction. */
16#define BTRFS_ROOT_TRANS_TAG 0
17
18enum btrfs_trans_state {
19 TRANS_STATE_RUNNING,
20 TRANS_STATE_COMMIT_PREP,
21 TRANS_STATE_COMMIT_START,
22 TRANS_STATE_COMMIT_DOING,
23 TRANS_STATE_UNBLOCKED,
24 TRANS_STATE_SUPER_COMMITTED,
25 TRANS_STATE_COMPLETED,
26 TRANS_STATE_MAX,
27};
28
29#define BTRFS_TRANS_HAVE_FREE_BGS 0
30#define BTRFS_TRANS_DIRTY_BG_RUN 1
31#define BTRFS_TRANS_CACHE_ENOSPC 2
32
33struct btrfs_transaction {
34 u64 transid;
35 /*
36 * total external writers(USERSPACE/START/ATTACH) in this
37 * transaction, it must be zero before the transaction is
38 * being committed
39 */
40 atomic_t num_extwriters;
41 /*
42 * total writers in this transaction, it must be zero before the
43 * transaction can end
44 */
45 atomic_t num_writers;
46 refcount_t use_count;
47
48 unsigned long flags;
49
50 /* Be protected by fs_info->trans_lock when we want to change it. */
51 enum btrfs_trans_state state;
52 int aborted;
53 struct list_head list;
54 struct extent_io_tree dirty_pages;
55 time64_t start_time;
56 wait_queue_head_t writer_wait;
57 wait_queue_head_t commit_wait;
58 struct list_head pending_snapshots;
59 struct list_head dev_update_list;
60 struct list_head switch_commits;
61 struct list_head dirty_bgs;
62
63 /*
64 * There is no explicit lock which protects io_bgs, rather its
65 * consistency is implied by the fact that all the sites which modify
66 * it do so under some form of transaction critical section, namely:
67 *
68 * - btrfs_start_dirty_block_groups - This function can only ever be
69 * run by one of the transaction committers. Refer to
70 * BTRFS_TRANS_DIRTY_BG_RUN usage in btrfs_commit_transaction
71 *
72 * - btrfs_write_dirty_blockgroups - this is called by
73 * commit_cowonly_roots from transaction critical section
74 * (TRANS_STATE_COMMIT_DOING)
75 *
76 * - btrfs_cleanup_dirty_bgs - called on transaction abort
77 */
78 struct list_head io_bgs;
79 struct list_head dropped_roots;
80 struct extent_io_tree pinned_extents;
81
82 /*
83 * we need to make sure block group deletion doesn't race with
84 * free space cache writeout. This mutex keeps them from stomping
85 * on each other
86 */
87 struct mutex cache_write_mutex;
88 spinlock_t dirty_bgs_lock;
89 /* Protected by spin lock fs_info->unused_bgs_lock. */
90 struct list_head deleted_bgs;
91 spinlock_t dropped_roots_lock;
92 struct btrfs_delayed_ref_root delayed_refs;
93 struct btrfs_fs_info *fs_info;
94
95 /*
96 * Number of ordered extents the transaction must wait for before
97 * committing. These are ordered extents started by a fast fsync.
98 */
99 atomic_t pending_ordered;
100 wait_queue_head_t pending_wait;
101};
102
103enum {
104 ENUM_BIT(__TRANS_FREEZABLE),
105 ENUM_BIT(__TRANS_START),
106 ENUM_BIT(__TRANS_ATTACH),
107 ENUM_BIT(__TRANS_JOIN),
108 ENUM_BIT(__TRANS_JOIN_NOLOCK),
109 ENUM_BIT(__TRANS_DUMMY),
110 ENUM_BIT(__TRANS_JOIN_NOSTART),
111};
112
113#define TRANS_START (__TRANS_START | __TRANS_FREEZABLE)
114#define TRANS_ATTACH (__TRANS_ATTACH)
115#define TRANS_JOIN (__TRANS_JOIN | __TRANS_FREEZABLE)
116#define TRANS_JOIN_NOLOCK (__TRANS_JOIN_NOLOCK)
117#define TRANS_JOIN_NOSTART (__TRANS_JOIN_NOSTART)
118
119#define TRANS_EXTWRITERS (__TRANS_START | __TRANS_ATTACH)
120
121struct btrfs_trans_handle {
122 u64 transid;
123 u64 bytes_reserved;
124 u64 delayed_refs_bytes_reserved;
125 u64 chunk_bytes_reserved;
126 unsigned long delayed_ref_updates;
127 unsigned long delayed_ref_csum_deletions;
128 struct btrfs_transaction *transaction;
129 struct btrfs_block_rsv *block_rsv;
130 struct btrfs_block_rsv *orig_rsv;
131 /* Set by a task that wants to create a snapshot. */
132 struct btrfs_pending_snapshot *pending_snapshot;
133 refcount_t use_count;
134 unsigned int type;
135 /*
136 * Error code of transaction abort, set outside of locks and must use
137 * the READ_ONCE/WRITE_ONCE access
138 */
139 short aborted;
140 bool adding_csums;
141 bool allocating_chunk;
142 bool removing_chunk;
143 bool reloc_reserved;
144 bool in_fsync;
145 struct btrfs_fs_info *fs_info;
146 struct list_head new_bgs;
147 struct btrfs_block_rsv delayed_rsv;
148};
149
150/*
151 * The abort status can be changed between calls and is not protected by locks.
152 * This accepts btrfs_transaction and btrfs_trans_handle as types. Once it's
153 * set to a non-zero value it does not change, so the macro should be in checks
154 * but is not necessary for further reads of the value.
155 */
156#define TRANS_ABORTED(trans) (unlikely(READ_ONCE((trans)->aborted)))
157
158struct btrfs_pending_snapshot {
159 struct dentry *dentry;
160 struct inode *dir;
161 struct btrfs_root *root;
162 struct btrfs_root_item *root_item;
163 struct btrfs_root *snap;
164 struct btrfs_qgroup_inherit *inherit;
165 struct btrfs_path *path;
166 /* block reservation for the operation */
167 struct btrfs_block_rsv block_rsv;
168 /* extra metadata reservation for relocation */
169 int error;
170 /* Preallocated anonymous block device number */
171 dev_t anon_dev;
172 bool readonly;
173 struct list_head list;
174};
175
176static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
177 struct btrfs_inode *inode)
178{
179 spin_lock(&inode->lock);
180 inode->last_trans = trans->transaction->transid;
181 inode->last_sub_trans = btrfs_get_root_log_transid(inode->root);
182 inode->last_log_commit = inode->last_sub_trans - 1;
183 spin_unlock(&inode->lock);
184}
185
186/*
187 * Make qgroup codes to skip given qgroupid, means the old/new_roots for
188 * qgroup won't contain the qgroupid in it.
189 */
190static inline void btrfs_set_skip_qgroup(struct btrfs_trans_handle *trans,
191 u64 qgroupid)
192{
193 struct btrfs_delayed_ref_root *delayed_refs;
194
195 delayed_refs = &trans->transaction->delayed_refs;
196 WARN_ON(delayed_refs->qgroup_to_skip);
197 delayed_refs->qgroup_to_skip = qgroupid;
198}
199
200static inline void btrfs_clear_skip_qgroup(struct btrfs_trans_handle *trans)
201{
202 struct btrfs_delayed_ref_root *delayed_refs;
203
204 delayed_refs = &trans->transaction->delayed_refs;
205 WARN_ON(!delayed_refs->qgroup_to_skip);
206 delayed_refs->qgroup_to_skip = 0;
207}
208
209bool __cold abort_should_print_stack(int error);
210
211/*
212 * Call btrfs_abort_transaction as early as possible when an error condition is
213 * detected, that way the exact stack trace is reported for some errors.
214 */
215#define btrfs_abort_transaction(trans, error) \
216do { \
217 bool first = false; \
218 /* Report first abort since mount */ \
219 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED, \
220 &((trans)->fs_info->fs_state))) { \
221 first = true; \
222 if (WARN(abort_should_print_stack(error), \
223 KERN_ERR \
224 "BTRFS: Transaction aborted (error %d)\n", \
225 (error))) { \
226 /* Stack trace printed. */ \
227 } else { \
228 btrfs_err((trans)->fs_info, \
229 "Transaction aborted (error %d)", \
230 (error)); \
231 } \
232 } \
233 __btrfs_abort_transaction((trans), __func__, \
234 __LINE__, (error), first); \
235} while (0)
236
237int btrfs_end_transaction(struct btrfs_trans_handle *trans);
238struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
239 unsigned int num_items);
240struct btrfs_trans_handle *btrfs_start_transaction_fallback_global_rsv(
241 struct btrfs_root *root,
242 unsigned int num_items);
243struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root);
244struct btrfs_trans_handle *btrfs_join_transaction_spacecache(struct btrfs_root *root);
245struct btrfs_trans_handle *btrfs_join_transaction_nostart(struct btrfs_root *root);
246struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root);
247struct btrfs_trans_handle *btrfs_attach_transaction_barrier(
248 struct btrfs_root *root);
249int btrfs_wait_for_commit(struct btrfs_fs_info *fs_info, u64 transid);
250
251void btrfs_add_dead_root(struct btrfs_root *root);
252void btrfs_maybe_wake_unfinished_drop(struct btrfs_fs_info *fs_info);
253int btrfs_clean_one_deleted_snapshot(struct btrfs_fs_info *fs_info);
254int btrfs_commit_transaction(struct btrfs_trans_handle *trans);
255void btrfs_commit_transaction_async(struct btrfs_trans_handle *trans);
256int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans);
257bool btrfs_should_end_transaction(struct btrfs_trans_handle *trans);
258void btrfs_throttle(struct btrfs_fs_info *fs_info);
259int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
260 struct btrfs_root *root);
261int btrfs_write_marked_extents(struct btrfs_fs_info *fs_info,
262 struct extent_io_tree *dirty_pages, int mark);
263int btrfs_wait_tree_log_extents(struct btrfs_root *root, int mark);
264int btrfs_transaction_blocked(struct btrfs_fs_info *info);
265int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
266void btrfs_put_transaction(struct btrfs_transaction *transaction);
267void btrfs_add_dropped_root(struct btrfs_trans_handle *trans,
268 struct btrfs_root *root);
269void btrfs_trans_release_chunk_metadata(struct btrfs_trans_handle *trans);
270void __cold __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
271 const char *function,
272 unsigned int line, int error, bool first_hit);
273
274int __init btrfs_transaction_init(void);
275void __cold btrfs_transaction_exit(void);
276
277#endif
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#ifndef __BTRFS_TRANSACTION__
20#define __BTRFS_TRANSACTION__
21#include "btrfs_inode.h"
22#include "delayed-ref.h"
23#include "ctree.h"
24
25enum btrfs_trans_state {
26 TRANS_STATE_RUNNING = 0,
27 TRANS_STATE_BLOCKED = 1,
28 TRANS_STATE_COMMIT_START = 2,
29 TRANS_STATE_COMMIT_DOING = 3,
30 TRANS_STATE_UNBLOCKED = 4,
31 TRANS_STATE_COMPLETED = 5,
32 TRANS_STATE_MAX = 6,
33};
34
35#define BTRFS_TRANS_HAVE_FREE_BGS 0
36#define BTRFS_TRANS_DIRTY_BG_RUN 1
37#define BTRFS_TRANS_CACHE_ENOSPC 2
38
39struct btrfs_transaction {
40 u64 transid;
41 /*
42 * total external writers(USERSPACE/START/ATTACH) in this
43 * transaction, it must be zero before the transaction is
44 * being committed
45 */
46 atomic_t num_extwriters;
47 /*
48 * total writers in this transaction, it must be zero before the
49 * transaction can end
50 */
51 atomic_t num_writers;
52 atomic_t use_count;
53 atomic_t pending_ordered;
54
55 unsigned long flags;
56
57 /* Be protected by fs_info->trans_lock when we want to change it. */
58 enum btrfs_trans_state state;
59 struct list_head list;
60 struct extent_io_tree dirty_pages;
61 unsigned long start_time;
62 wait_queue_head_t writer_wait;
63 wait_queue_head_t commit_wait;
64 wait_queue_head_t pending_wait;
65 struct list_head pending_snapshots;
66 struct list_head pending_chunks;
67 struct list_head switch_commits;
68 struct list_head dirty_bgs;
69 struct list_head io_bgs;
70 struct list_head dropped_roots;
71 u64 num_dirty_bgs;
72
73 /*
74 * we need to make sure block group deletion doesn't race with
75 * free space cache writeout. This mutex keeps them from stomping
76 * on each other
77 */
78 struct mutex cache_write_mutex;
79 spinlock_t dirty_bgs_lock;
80 /* Protected by spin lock fs_info->unused_bgs_lock. */
81 struct list_head deleted_bgs;
82 spinlock_t dropped_roots_lock;
83 struct btrfs_delayed_ref_root delayed_refs;
84 int aborted;
85};
86
87#define __TRANS_FREEZABLE (1U << 0)
88
89#define __TRANS_USERSPACE (1U << 8)
90#define __TRANS_START (1U << 9)
91#define __TRANS_ATTACH (1U << 10)
92#define __TRANS_JOIN (1U << 11)
93#define __TRANS_JOIN_NOLOCK (1U << 12)
94#define __TRANS_DUMMY (1U << 13)
95
96#define TRANS_USERSPACE (__TRANS_USERSPACE | __TRANS_FREEZABLE)
97#define TRANS_START (__TRANS_START | __TRANS_FREEZABLE)
98#define TRANS_ATTACH (__TRANS_ATTACH)
99#define TRANS_JOIN (__TRANS_JOIN | __TRANS_FREEZABLE)
100#define TRANS_JOIN_NOLOCK (__TRANS_JOIN_NOLOCK)
101
102#define TRANS_EXTWRITERS (__TRANS_USERSPACE | __TRANS_START | \
103 __TRANS_ATTACH)
104
105#define BTRFS_SEND_TRANS_STUB ((void *)1)
106
107struct btrfs_trans_handle {
108 u64 transid;
109 u64 bytes_reserved;
110 u64 chunk_bytes_reserved;
111 unsigned long use_count;
112 unsigned long blocks_reserved;
113 unsigned long blocks_used;
114 unsigned long delayed_ref_updates;
115 struct btrfs_transaction *transaction;
116 struct btrfs_block_rsv *block_rsv;
117 struct btrfs_block_rsv *orig_rsv;
118 short aborted;
119 short adding_csums;
120 bool allocating_chunk;
121 bool can_flush_pending_bgs;
122 bool reloc_reserved;
123 bool sync;
124 unsigned int type;
125 /*
126 * this root is only needed to validate that the root passed to
127 * start_transaction is the same as the one passed to end_transaction.
128 * Subvolume quota depends on this
129 */
130 struct btrfs_root *root;
131 struct seq_list delayed_ref_elem;
132 struct list_head qgroup_ref_list;
133 struct list_head new_bgs;
134};
135
136struct btrfs_pending_snapshot {
137 struct dentry *dentry;
138 struct inode *dir;
139 struct btrfs_root *root;
140 struct btrfs_root_item *root_item;
141 struct btrfs_root *snap;
142 struct btrfs_qgroup_inherit *inherit;
143 struct btrfs_path *path;
144 /* block reservation for the operation */
145 struct btrfs_block_rsv block_rsv;
146 u64 qgroup_reserved;
147 /* extra metadata reseration for relocation */
148 int error;
149 bool readonly;
150 struct list_head list;
151};
152
153static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
154 struct inode *inode)
155{
156 spin_lock(&BTRFS_I(inode)->lock);
157 BTRFS_I(inode)->last_trans = trans->transaction->transid;
158 BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
159 BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
160 spin_unlock(&BTRFS_I(inode)->lock);
161}
162
163/*
164 * Make qgroup codes to skip given qgroupid, means the old/new_roots for
165 * qgroup won't contain the qgroupid in it.
166 */
167static inline void btrfs_set_skip_qgroup(struct btrfs_trans_handle *trans,
168 u64 qgroupid)
169{
170 struct btrfs_delayed_ref_root *delayed_refs;
171
172 delayed_refs = &trans->transaction->delayed_refs;
173 WARN_ON(delayed_refs->qgroup_to_skip);
174 delayed_refs->qgroup_to_skip = qgroupid;
175}
176
177static inline void btrfs_clear_skip_qgroup(struct btrfs_trans_handle *trans)
178{
179 struct btrfs_delayed_ref_root *delayed_refs;
180
181 delayed_refs = &trans->transaction->delayed_refs;
182 WARN_ON(!delayed_refs->qgroup_to_skip);
183 delayed_refs->qgroup_to_skip = 0;
184}
185
186int btrfs_end_transaction(struct btrfs_trans_handle *trans,
187 struct btrfs_root *root);
188struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
189 unsigned int num_items);
190struct btrfs_trans_handle *btrfs_start_transaction_fallback_global_rsv(
191 struct btrfs_root *root,
192 unsigned int num_items,
193 int min_factor);
194struct btrfs_trans_handle *btrfs_start_transaction_lflush(
195 struct btrfs_root *root,
196 unsigned int num_items);
197struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root);
198struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root);
199struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root);
200struct btrfs_trans_handle *btrfs_attach_transaction_barrier(
201 struct btrfs_root *root);
202struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root);
203int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid);
204
205void btrfs_add_dead_root(struct btrfs_root *root);
206int btrfs_defrag_root(struct btrfs_root *root);
207int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root);
208int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
209 struct btrfs_root *root);
210int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
211 struct btrfs_root *root,
212 int wait_for_unblock);
213int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
214 struct btrfs_root *root);
215int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
216 struct btrfs_root *root);
217void btrfs_throttle(struct btrfs_root *root);
218int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
219 struct btrfs_root *root);
220int btrfs_write_marked_extents(struct btrfs_root *root,
221 struct extent_io_tree *dirty_pages, int mark);
222int btrfs_wait_marked_extents(struct btrfs_root *root,
223 struct extent_io_tree *dirty_pages, int mark);
224int btrfs_transaction_blocked(struct btrfs_fs_info *info);
225int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
226void btrfs_put_transaction(struct btrfs_transaction *transaction);
227void btrfs_apply_pending_changes(struct btrfs_fs_info *fs_info);
228void btrfs_add_dropped_root(struct btrfs_trans_handle *trans,
229 struct btrfs_root *root);
230#endif